Alteration of Sequence Specificity of the Type II Restriction Endonuclease HincII through an Indirect Readout Mechanism*

The functional and structural consequences of a mutation of the DNA intercalating residue of HincII, Q138F, are presented. Modeling has suggested that the DNA intercalation by Gln-138 results in DNA distortions potentially used by HincII in indirect readout of its cognate DNA, GTYRAC (Y = C or T, R = AorG) (Horton, N. C., Dorner, L. F., and Perona, J. J. (2002) Nat. Struct. Biol. 9, 42-47). Kinetic data presented here indicate that the mutation of glutamine 138 to phenylalanine (Q138F) results in a change in sequence specificity at the center two base pairs of the cognate recognition site. We show that the preference of HincII for cutting, but not binding, the three cognate sites differing in the center two base pairs has been altered by the mutation Q138F. Five new crystal structures are presented including Q138F HincII bound to GTTAAC and GTCGAC both with and without Ca2+ as well as the structure of wild type HincII bound to GTTAAC. The Q138F HincII/DNA structures show conformational changes in the protein, bound DNA, and at the protein-DNA interface, consistent with the formation of adaptive complexes. Analysis of these structures and the effect of Ca2+ binding on the protein-DNA interface illuminates the origin of the altered specificity by the mutation Q138F in the HincII enzyme.

High resolution x-ray crystal structures of protein-DNA complexes have revealed a rational basis for sequence specificity in the form of direct contacts between amino acid side chains and DNA bases in the major groove (1,2). However, cases have been reported where specificity of a protein for a particular base pair (or base pairs) within its recognition sequence exists even when no direct contacts are seen between the protein and DNA (3)(4)(5)(6)(7)(8)(9)(10)(11). In these cases the specificity exhibited by the DNA-binding protein appears to be mani-fested by indirect means. This "indirect readout" of the DNA sequence may involve water-mediated contacts, contacts to the sugar-phosphate backbone, or the utilization of sequence-dependent DNA structure and/or deformability (3,12).
The crystal structure of the type II restriction endonuclease HincII reveals distortion from canonical B-form of its bound cognate DNA (13). HincII recognizes and cleaves the sequence GTYRAC, having degeneracy in its specificity for the center two nucleotides by allowing the presence of any pyrimidinepurine, i.e. TA, TG, CA, or CG. Sequence-specific contacts between the protein and DNA can explain this specificity. The contact to the center YR step (CG in the original crystal structure (13)) consists of a hydrogen bond from Asn-141 to the N7 of the purine base. Such a contact can occur when either G or A, but not C or T, occur in the fourth position of the recognition sequence. No contacts are made to the pyrimidine base. Typically, the high degree of specificity achieved by type II restriction endonucleases is due in part to the fact that contacts occur to both bases of each base pair. The question remains of whether or not one hydrogen bond to each purine of the center base pairs can explain the HincII specificity against sequences containing GTYYAC, GTRYAC, or GTRRAC.
The crystal structure of wild type HincII bound to one of the three cognate DNA sequences, GTCGAC, shows that HincII distorts the bound DNA at three loci, two as a result of intercalation of the Gln-138 side chain into the DNA duplex just outside of its recognition sequence (Fig. 1A, light brown spacefilling atoms ϭ the recognition sequence, dark brown spacefilling atoms ϭ flanking DNA), and the third site at the center YR base pair (Fig. 1, A, red ϭ Y (C or T), and blue ϭ R (A or G), B, right). A hypothesis has been put forth that links the distortion of DNA at the center CG to indirect readout of this sequence by HincII. Furthermore, modeling of the DNA distortion revealed a connection between the Gln-138 intercalation and the center base pair distortion three base pairs away (13).
The studies herein describe the characterization of HincII structure and function when glutamine 138 has been substituted with phenylalanine, Q138F. Preliminary velocity measurements showed that Q138F HincII favors TA over CG at the center YR step to a greater extent than wild type. Subsequent DNA binding assays and single turnover cleavage rate measurements revealed that the alteration in specificity toward cognate sequences containing either CG or TA at the center step occurs at the cleavage and not the binding step. The structural studies of Q138F HincII bound to cognate DNA containing either CG or TA at the center step (GTYRAC) with and without Ca 2ϩ showed numerous conformational changes in the protein and DNA as a result of the mutation that differ with the various DNA sequences as well as the presence or absence of Ca 2ϩ . The divalent cation Ca 2ϩ inhibits DNA cleavage while increasing the affinity and specificity of type II restriction endonucleases for their cognate DNA (14 -16) and is found (pink ball, Fig. 1A) to bind very close to the identified Mg 2ϩ binding sites in the active sites of enzymes studied (17). Thus, Ca 2ϩ serves as a mimic of the in vivo cofactor Mg 2ϩ , allowing the observation of a pretransition state structure. The structure of wild type HincII bound to cognate DNA containing TA at the center step showed no significant conformational differences from the previously determined structure with CG at the center step. These five new structures have been analyzed with respect to the observed functional effects of the mutation to address the following questions. 1) Why does the Q138F mutation reduce the single turnover cleavage rate of cognate DNA by HincII? 2) Why is the DNA binding affinity of HincII for cognate DNA altered very little by the Q138F mutation? 3) Why does the Q138F mutation of HincII reduce the single turnover rate of cleavage more for GTCGAC than for GTTAAC?
The overlapping primers, P 1 and P 2 , were selected in the vicinity of targeted position 414 nucleotides of the hincII gene to allow the introduction of a desired codon change (CAA to TTT) in the primer P 2 sequence (marked as bold). Each overlapping primer contained one deoxyuridine residue that flanked the overlap region on the 3Ј side. 420-and 380-bp fragments of hincIIR gene were amplified with AmpliTaqDNA polymerase (Applied Biosystems) following the manufacturer's recommendations as follows: initial denaturation at 94°C for 5 min, denaturation at 94°C for 10 s, annealing at 55°C for 1 min, extension at 72°C for 20 s, final extension at 72°C for 5 min, for 25 number. PCR products were purified by phenol-chloroform extraction and alcohol precipitation and dissolved in 50 l of Tris-EDTA buffer (10 mM Tris, pH 7.5, 1 mM EDTA). For a 50-l assembly reaction, 20 pmol (20 l) of each PCR fragment were mixed with 5 l of 10ϫ T4 DNA ligase buffer, 1 l of 10 mM dNTPs, and 1 l (5 units) of DNA polymerase I large (Klenow) fragment. The reaction was incubated for 10 min at 37°C to remove unpaired nucleotides from the 3Ј-termini of PCR fragments added by TaqDNA polymerase. The Klenow Fragment was then inactivated by incubating for 20 min at 80°C. 1 l (1 unit) of USER enzyme (New England Biolabs) was added to the reaction and incubated for 15 min at 37°C. The USER enzyme is a mixture of two DNA glycosylases, uracil DNA glycosylase and Endo-VIII DNA glycosylase/AP lyase. Uracil DNA glycosylase catalyzes the excision of a uracil base, forming an abasic (apyrimidinic) site (18,19), whereas lyase activity of Endo-VIII breaks the phosphodiester backbone at the 3Ј and 5Ј sides of the abasic site (20,21). After phosphodiester bond breakage, the terminal 8-mer oligonucleotides dissociate, leaving PCR products flanked with complementary 3Ј single-stranded extentions (Fig. 1). Next, 1 l (400 units) of T4 DNA ligase was added, and fragments were ligated for 15 min at room temperature. T4 DNA ligase was inactivated by incubating for 20 min at 80°C. Ligated 780-bp PCR product carrying the complete hincIIR/Q138F gene was then digested with NdeI and SapI restriction endonucleases and subcloned into NdeI-SapI-cleaved dephosphorylated pTXB1 vector (New England Biolabs). The new construct, pLC/HincIIQ138F, was introduced into Escherichia coli ER2566 strain carrying the pLysS and pSYX20/HincIIM plasmids.
Protein Purification-Purification of wild type and mutant HincII was performed as described (22). Purified enzyme was dialyzed exhaustively against buffer A (50 mM bistris propane, pH 7.5, at room temperature, 140 mM NaCl, 1 mM DTT 3 ) made 50% in glycerol, flash-frozen in liquid nitrogen, and stored at Ϫ80°C until just before use. Purity has been estimated from overloaded SDS-PAGE (stained with Coomassie Blue) as greater than 99.6 and 98.5% for Q138F and wild type HincII, respectively.
Preparation of DNA Substrates-Plasmids containing a single HincII cleavage site, GTCGAC, GTTGAC/GTCAAC, or GTTAAC were prepared using Qiagen miniprep kits (Qiagen) using pUC19, pACYC177, or a modified pUC19 with GTTAAC in place of the SalI/HincII site, respectively. Each plasmid was cleaved with either BsrFI (both pUC19 plasmids) or XhoI (pACYC177) before the multiple turnover assays such that the HincII site was located approximately in the center of the linear DNA.
Both 6-hexachlorofluorescein (HEX)-labeled and non-HEXlabeled oligonucleotides were obtained from commercial synthetic sources and purified using C18 reverse phase high performance liquid chromatography (23). The concentration of each single strand was then measured spectrophotometrically, with extinction coefficients calculated from standard values for the nucleotides (24) and 32.2 mM Ϫ1 cm Ϫ1 (at 260 nm) for the HEX molecule. Equimolar amounts of each strand (typically 0.5-2 mM each) were mixed and annealed by heating to 90°C for 10 min followed by slow-cooling to 4°C over 4 -5 h in an Eppindorf thermocycler. Radiolabeled oligonucleotides were prepared using [␥-32 P]ATP (MP Biomedicals, Irvine, CA) and T4 polynucleotide kinase (Fermentas, Hercules, CA). Excess [ 32 P]ATP was removed using Micro Bio-spin 6 columns P-polyacrylamide P-6 gel (Bio-Rad).
DNA substrates are shown below, with the recognition sequence in bold, center step underlined, and cleavage site marked by the following sequences.
Multiple Turnover Velocity Measurements-Linearized plasmid DNA (2 nM) was incubated at 37°C with 0.6 nM concentrations of wild type HincII or Q138F HincII in 80 l of reaction buffer (50 mM Tris-HCl, pH 7.9, at 25°C, 10 mM MgCl 2 , 100 mM NaCl, 1 mM DTT). 10-l samples were withdrawn at the indicated times from the reaction, quenched by the addition of 5 l of quench solution (60 mM EDTA, 0.2% SDS, 50% glycerol, 0.02% bromphenol blue), and subjected to electrophoresis on 1% agarose-gel containing ethidium bromide (3 ng/ml) ( Fig.  2A). The density of each band was quantitated from the fluorescence of ethidium bromide using NIH ImageJ Version 1. 31 program. The product (cleaved DNA) concentration was then calculated using a calibration curve constructed from the known concentrations of pUC19 DNA cleaved with BsrFI and HincII. The rates of cleavage, or multiple turnover velocities, were measured as the slope of product concentration versus time of incubation (Fig. 2, B and C).  Binding Assays-The dissociation constant K D was measured against the 5Ј-32 P-end-labeled 16-bp DNA containing the relevant sequence by a gel shift assay (10,15,(25)(26)(27). The binding buffer was composed of mixing buffer B with buffer C until the conductivity (19.00 millisiemens/cm) of buffer A was achieved. Buffer compositions were as follows: buffer B, 50 mM bistris propane (pH 7.5 at 25°C), 500 mM NaCl, 5 mM CaCl 2 , 1 mM DTT; buffer C, 50 mM bis-tris propane (pH 7.5 at 25°C), 5 mM CaCl 2 , 1 mM DTT. Protein concentration (in terms of monomers) was measured spectrophotometrically using the extinction coefficient ⑀ ϭ 36,840 cm Ϫ1 M Ϫ1 at 280 nm. The calculated concentration was then adjusted for molarity of dimers. Protein and DNA were incubated at 37°C for 30 min, rapidly mixed with prewarmed loading buffer composed of binding buffer and 20% glycerol, and immediately loaded onto a 10% native PAGE gel undergoing electrophoresis. The running buffer was composed of 89 mM Tris, 89 mM boric acid, 5 mM CaCl 2 , pH 8.3. The gels were run at 190 V and 4°C with recirculation of the running buffer. Gels were autoradiographed using a phosphorimage plate (GE Healthcare) (Fig. 3A). The data were fit by nonlinear regression (Fig. 3B) to Equation 1 (derived from the quadratic solution to the equilibrium association binding equation).
where A represents the intensity of the protein bound DNA band. Both A max and A min are fit in addition to K D . The DNA concentration, O T , was held constant, and P T , the total protein under "Experimental Procedures," giving a K D of 1.19 nM. C, scan of a gel used to measure the quantity of cleaved TA HEX-DNA after incubation with wild type HincII. D, a plot of the integrated intensity of product bands in C versus time of incubation for the top (black squares) and bottom (gray circles) strands. E, chart showing the relative ratios of k cleave /K D values for wild type HincII (black bars) and Q138F HincII (hatched bars) for the TA and CG substrates. The ratio is calculated using the corresponding value with CG for the wild type or mutant protein as the reference (values from the second to last column in Table 2). concentration, is the independent variable, varying with each measurement of A. Each reported K D is the average of at least three independent measurements.
Single Turnover Reactions-HEX or 32 P-labeled DNA containing the cognate sequence GTCGAC or GTTAAC at concentrations ranging from 0.62 to 77 nM were incubated with wild type or Q138F HincII at a concentration of 8 -5000-fold molar excess of protein to DNA. The reaction conditions included a buffer prepared by mixing buffer D with buffer E until the conductivity of buffer AЈ was achieved. Buffer concentrations were as follows: buffer D, 50 mM bis-tris propane (pH 7.5 at 25°C), 10 mM MgCl 2 , 0.1 mg/ml bovine serum albumin, 1 mM DTT; buffer E, 50 mM bis-tris propane (pH 7.5 at 25°C), 10 mM MgCl 2 , 0.1 mg/ml bovine serum albumin, 1 mM DTT, 500 mM NaCl; buffer AЈ, 50 mM bis-tris propane (pH 7.5 at 25°C), 0.1 mg/ml bovine serum albumin, 1 mM DTT, 140 mM NaCl. This procedure for preparing the reaction buffer was used to ensure constant ionic strength between various binding and reaction buffers having different concentrations of Ca 2ϩ or Mg 2ϩ . Initiation of the reaction was performed by mixing preheated (at 37°C) enzyme and DNA followed by incubation of the reaction at 37°C. At various indicated times after initiation, 10 l of the reaction were removed, and the reaction was quenched by the addition of 10 l of 4 M urea, 50 mM EDTA, 80% formamide. The quenched reaction aliquots were then separated on a 20% acrylamide denaturing PAGE gel (containing 8 M urea) in 1ϫ TBE (89 mM Tris, 89 mM boric acid, 10 mM EDTA, pH 8.3). A Typhoon scanner (GE Healthcare) was used to detect the HEX-DNA within the gels or the bands on a phosphorimage plate after exposure to the gel containing 32 P-labeled DNA. The product bands were integrated using Image-Quant (GE Healthcare), then plotted against the length of reaction. The intensity of the product band was fit to a single exponential: p ϭ C(1 Ϫ e Ϫkt ). Each reported rate constant is the average of at least three independent measurements. The cleavage site is off-center in the 16-base pair substrate, and the products of cleavage of both strands are distinguishable on the denaturing PAGE gel (Fig. 3C). The single turnover cleavage rate constants were determined for each strand (Fig. 3D).
Crystallization, Diffraction Data Collection, and Structure Refinement-Crystals of wild type HincII bound to cognate DNA containing TA at the center step were prepared in a manner identical to those containing CG at the center step (22). Crystals of mutant HincII were prepared by cocrystallization with two different recognition sequences, GTTAAC or GTC-GAC. Both DNA substrates were 14 base pairs in length and contained the sequence 5Ј-GCCGGTTAACCGGC-3Ј (top and bottom) or 5Ј-GCCGGTCGACCGGC-3Ј (top and bottom), respectively (the HincII recognition sequence is underlined). DNA containing these recognition sequences was mixed at a molar ratio of 1:2 with 16 -23 mg/ml starting concentration of protein and 9.42 mg/ml TA or 7 mg/ml CG DNA concentration. The crystallization conditions included 1.5-3.0 l of the protein:DNA mixture with 1.0 -1.5 l of the precipitating solution (25-21% polyethylene glycol 4000, 0.1 M imidazole buffer, pH 6.5, or 0.1 M HEPES buffer, pH 7.5, 0.15 M NaCl) per drop and placed over 1 ml of the precipitating solution. Crystals appear in drops between 17 h to 1 week at 17°C. The Ca 2ϩ -soaked crystals were prepared by soaking crystals in 25% polyethylene glycol 4000, 0.1 M buffer, pH 6.5 or 7.5, 0.3 M NaCl, and 50 mM CaCl 2 for 5 h. The crystals were then exchanged into a cryoprotection buffer (25% polyethylene glycol 4000, 0.1 M buffer, pH 6.5 or 7.5, 0.3 M NaCl, 30% glycerol) and flash-frozen in liquid nitrogen. X-ray diffraction amplitudes were measured using synchrotron radiation at the Stanford Synchrotron Light Source (SSRL) BL9 -2 (HQ9, HQ51, see Table 3), APS BL14-BMC (H56 , Table 3), or SSRL BL 1-5 (HQ61, HQ65, Table 3). Data collection was performed while maintaining the crystal at 100 K. Image processing and data reduction were performed with MOSFLM (28) and SCALA (29) (HQ9, Table 3) or HKL2000 (HKL Research, Inc.) (H56, HQ51, HQ61, HQ65, Table 3).
In the case of the structure of wild type HincII bound to cognate DNA-containing TA, the previously solved isomorphous structure of HincII bound to cognate DNA-containing CG was used beginning with rigid body refinement as implemented in CNS (30) followed by iterative model rebuilding using XtalView (31), simulated annealing, temperature factor, and positional refinement using CNS (30). All structures of Q138F HincII bound to DNA with and without Ca 2ϩ are nonisomorphous with the previously solved wild type HincII/cognate DNA structure. Three (HQ9, HQ65, HQ61; Table 3) have space group P2 1 symmetry and are isomorphous. The binding of Ca 2ϩ by the Q138F/CG crystals changes the space group symmetry from P2 1 to I4 in data set HQ51. The HQ9 (Q138F/CG) structure was solved by molecular replacement using a monomer from the wild type DNA-bound form (1KC6, wtHincII/CG) as a search model and the program CNS (30). The same solution was used as starting models for HQ61 and HQ65. The HQ51 structure was solved independently also with the wild type HincII bound to CG DNA (1KC6) monomer and MOLREP (32). Models were then refined using CNS (30) and rigid body refinement, simulated annealing, temperature factor refinement, and manual model building using XtalView (31).
Analysis of Protein Conformation-r.m.s.d. were calculated using LSQKAB (33). Parts of the protein molecules which move as rigid bodies were determined using the difference distance matrix method implemented in DDMP. 4 Rotations and translations of parts of a structure relative to another were calculated using LSQKAB (33). Rotation angles describing the opening up or closing down of one dimer relative to another are calculated by first superimposing one subunit using all C␣ atoms, then calculating the rotation angle necessary to align the second subunit.

RESULTS
Multiple Turnover Velocity Measurements-An example of a gel used to measure the amount of cleaved DNA as a function of the length of incubation with HincII is shown in Fig. 2A. The plots of product versus time for wild type HincII or Q138F HincII with each of the three cognate sequences are shown in Fig. 2, B and C, respectively, and the resulting rates are provided in Table 1. Multiple rounds of cleavage take place as evidenced by the linear production of cleaved DNA with time. From the appearance of a single cleavage in each plasmid bearing one of the three cognate HincII recognition sites, it is clear that the endonuclease activity of Q138F HincII is specific for the HincII recognition sites. However, Q138F HincII is reduced in overall cleavage activity and presents an altered preference for the sequence at the center step of the six-base pair recognition sequence. Although wild type HincII prefers TA:(TG/CA):CG at the center step on the order of 6:2:1, the mutant, Q138F HincII, prefers the same sequences on the order 36:8:1. Thus, the mutant prefers TA relative to CG 6-fold more than does the wild type HincII protein.
Binding Measurements-The measurement of binding affinities of the two proteins, wild type and Q138F HincII, for the cognate sequences containing either TA or CG at their center step was performed using the gel shift assay. The divalent cation Ca 2ϩ was included in the incubation and running buffer because numerous studies have shown that this ion inhibits cleavage while increasing the affinity and specificity of the enzyme for its cognate DNA (14,15,35). Fig. 3A shows an example of a scan of one gel shift assay, and Fig. 3B shows the fit of the data from Fig. 3A used to determine the K D . The measured K D values for wild type and Q138F HincII with the cognate sequences containing CG or TA are shown in Table 2, as are the standard deviations determined from three independent binding measurements.
Wild type HincII binds to its cognate DNA containing TA or CG at the center step with a K D of 0.58 -1.14 nM, respectively, indicating a 2-fold preference for the TA substrate. Q138F HincII bound the cognate sequences with roughly the same affinity as the wild type HincII. Therefore, the substitution of phenylalanine for glutamine at the DNA intercalating residue did not appreciably affect either the affinity or sequence preference (TA versus CG) of HincII. In contrast, no binding of CG DNA (1 nM) was detected using the identical gel shift assay without Ca 2ϩ in the binding or running buffers with either wild type or Q138F HincII up to a concentration of 1 ϫ 10 Ϫ6 M (supplemental data).
Single Turnover Cleavage Rate Constant Measurements-The single turnover DNA cleavage rate measurements were performed with HEX-or 32 P-labeled DNA. To assess any possible effect of the HEX label on HincII activity, a single turnover cleavage rate measurement was performed on 32 P-labeled DNA identical to the HEX-labeled DNA-containing CG at the center step and the Q138F enzyme. Identical rates were observed (data not shown).
The single turnover DNA cleavage rate constants, k cleave , were determined in the presence of an 11-5000-fold molar excess of protein over DNA and either 8.1 ϫ 10 Ϫ7 M (wild type HincII reactions) or 3.0 ϫ 10 Ϫ6 M (Q138F HincII reactions) protein. Given the dissociation constants determined above (in the presence of Ca 2ϩ ), these conditions ensure Ն99% binding of the DNA (if the binding constants with Mg 2ϩ are similar to those with Ca 2ϩ ), with only a single round of cleavage. In all cases the quantified product at specific time points followed a single exponential, indicating the detection of a single step. For all single turnover measurements both with wild type or Q138F HincII, the product release step was not of interest, and the quenching of the reaction with the denaturant urea before electrophoreses ensured that all product was measured. In this way the product release step did not contribute to the k cleave measurement.
The single turnover cleavage rate constants are given in Table 2 for both wild type and Q138F HincII, with cognate DNA containing either CG or TA at the center step. The reactions were initiated by the mixing of HincII enzyme with DNA at time 0, both in buffer containing Mg 2ϩ . The mutation impairs the single turnover cleavage rate constant by 37-39fold in the case of the TA substrate and 84 -128-fold in the case of the CG substrate. Although there is a 1.6 -2.5-fold difference found between k cleave of wild type HincII and these two cognate substrates, 5.6-fold differences occur with the mutant.
The data of Table 2 show that HincII exhibits sequence preferences at both the binding and cleavage steps. A specificity constant composed of k cleave /K D (albeit different from a standard k cleave /K D , because the K D was measured with Ca 2ϩ , and the cleavage rate constant k cleave was measured with Mg 2ϩ ) can be calculated for each enzyme/substrate pair, also shown in Table  2. Using this parameter, wild type HincII prefers TA over CG by 3:1. The mutation, Q138F, designed to test the connection between DNA intercalation and sequence specificity, alters this preference to 14:1, differing from the wild type preference by nearly 5-fold (Fig. 3E).
Overall Structure of Wild Type HincII with TA (wtHincII/TA)-The structure of wild type HincII bound to cognate DNA containing TA at the center step (GTTAAC) has been solved to 2.6 Å in the I2 1 2 1 2 1 space group (Table 3) and is isomorphous with wild type HincII bound to cognate DNA containing CG (GTC-GAC) (13). Like the isomorphous structure, the asymmetric unit contains two protein dimers and two duplexes of DNA. Only two protein residues could not be located in the electron density maps, both occurring at the carboxyl terminus and only in chains B and D. However, all 13 nucleotides of all four DNA strands were visible in the maps. The model also includes 292 ͗I hkl ͘ is the average intensity over symmetry related and equivalent reflections, and I hkl is the observed intensity for reflection hkl. b R cyst Ϫ ͚ hkl (ʈF obs ͉ Ϫ ͉F calc ʈ)/͚ hkl ͉F obs ͉, where ͉F obs ͉ and ͉F calc ͉ are the observed and calculated structure factor amplitude for reflection hkl. The sum is carried out over the 95% of the observed reflections that are used in refinement. c R free refers to the R factor for the test reflection set (5% of the total observed) that was excluded from refinement.  (16)). The coordinate error (using the Luzzati method (37) as implemented in CNS (30)) for these structures is 0.4 and 0.2 Å; therefore, these large r.m.s.d. indicate the presence of significant conformational changes within the structures. High r.m.s.d. are evident at the monomeric as well as at the dimeric level, indicating a change both within as well as between the two monomers of the dimers. The r.m.s.d. of both subunits A and B as well as in the dimer are greater in the structure with TA (Q138F/TA) than that with CG (Q138F/CG). Significant conformational differences are also evident in the bound DNA. Judging by the lower r.m.s.d., it appears that the core 6 base pairs of the CG DNA bound to Q138F HincII is more like the CG DNA bound to wild type HincII than the TA DNA bound to Q138F HincII, as is apparent in viewing the superposition (see Fig. 5A, blue is Q138F/CG, gray is wtHincII/CG/Ca 2ϩ , red is Q138F/TA). Because very little difference is found in the wild type HincIIbound DNA whether CG or TA or with or without Ca 2ϩ , the larger r.m.s.d. in the DNA of the Q138F/TA structure is not an inherent property of the DNA sequence alone or due to the bound Ca 2ϩ . We can also look at the r.m.s.d. of the DNA when the protein dimer is used for the superposition onto wtHincII/CG/Ca 2ϩ , which gives additional information as to how differently (relative to wtHincII/ CG/Ca 2ϩ ) the DNA sits in the Q138F HincII dimer, and find that it is significant in both mutant structures. However, it is much greater for the core six base pairs of TA bound to Q138F HincII than for the CG DNA. Again, the larger numbers indicate a mispositioning of the DNA within the dimer that is greater in Q138F/TA than Q138F/CG. Thus, the high r.m.s.d. in these two structures, Q138F/TA and Q138F/CG, indicates numerous differences (from wild type) in conformation, and in addition, in all measures of r.m.s.d., the Q138F/TA structure contains the greatest deviations from wild type HincII bound to cognate DNA.
Overall Structure; Q138F HincII Bound to GTTAAC with Ca 2ϩ (Q138F/TA/Ca 2ϩ ), and Q138F HincII Bound to GTC-GAC with Ca 2ϩ (Q138F/CG/Ca 2ϩ )-Structures of Q138F Hin-cII bound to DNA and Ca 2ϩ were prepared by soaking the Q138F/DNA crystals in 50 mM CaCl 2 for 5 h. The Q138F/TA/   Fig. 4. The r.m.s.d. of different parts of these structures when compared with the wtHincII/CG/Ca 2ϩ structure is given in Table 4. The r.m.s.d. of residues in each monomer have either changed very little or decreased in response to binding Ca 2ϩ in both mutant HincII structures, Q138F/TA/Ca 2ϩ and Q138F/CG/ Ca 2ϩ . Significantly, the r.m.s.d. of the whole dimer has decreased significantly in response to binding Ca 2ϩ (Table 4, compare the r.m.s.d. of whole dimer of Q138F/TA/Ca 2ϩ to Q138F/TA and Q138F/TA/Ca 2ϩ to Q138F/TA). Thus, on first inspection, the effect of Ca 2ϩ binding appears to have brought the protein structure, especially that of the protein dimer (last column, Table 4), closer to that of wild type. In addition, the structure containing CG at the center step, Q138F/CG/Ca 2ϩ , is closer to the wild type structure at both the monomer and dimer level. Looking at the DNA, the r.m.s.d. of all atoms of the six base pairs of the recognition sequence has decreased in the mutant structure bound to TA in response to binding Ca 2ϩ but has increased in the structure with CG on binding Ca 2ϩ . Thus, the effect of Ca 2ϩ binding to the mutant protein/DNA complexes has had opposite effects on the two different DNA sequences, making the TA DNA appear more like that bound to wild type HincII but making the CG DNA look less like it (Fig. 5,  B-D). Looking at the way the DNA sits in the protein dimer, namely the r.m.s.d. of the DNA after superposition of the protein dimer, these values have decreased relative to the corresponding structures without Ca 2ϩ ; therefore, the Ca 2ϩ has had the effect of positioning both DNA sequences within the Q138F HincII dimer much more similarly to that within wild type HincII (see Fig. 9). In summary, we find that Ca 2ϩ binding has had the effect of bringing both the dimer conformation and the positioning of the DNA within the dimer much closer to that found in wtHincII/CG/Ca 2ϩ in the mutant structures with either DNA sequence. In addition, it appears to bring the core six base pair recognition sequence of DNA of the TA DNA closer to that of the DNA bound to the wild type enzyme. The dimer conformation still deviates more from that of wild type with TA DNA than with CG, even with bound Ca 2ϩ . The r.m.s.d. of the DNA itself and of its positioning also shows that it is still greater in the mutant structures bound to TA DNA than CG DNA (i.e. in Q138F/TA/Ca 2ϩ compared with Q138F/ CG/Ca 2ϩ ). These observations will be important below because it appears that the distortion of the protein dimer as a result of loss of the center step base stacking in the TA DNA allows for better recovery of the alignment of the active sites in Q138F/ TA/Ca 2ϩ as compared with those in Q138F/CG/Ca 2ϩ .
DNA Structure-Comparison of the structures of the sixbase pair recognition site of DNA from the Q138F/DNA structures to that bound to wtHincII/CG/Ca 2ϩ is shown in Fig. 5. The superposition of these base pairs from the mutant structures without Ca 2ϩ onto that from wtHincII/CG/Ca 2ϩ is shown in Fig. 5A. It is quite apparent here that the DNA from Q138F/TA (red) possesses much greater deviation from that bound to wtHincII/CG/Ca 2ϩ (gray) than that from Q138F/CG (dark blue). The deviations are especially apparent in the center two base pairs. It appears that stress on the protein-DNA complex as a result of the mutation produced distortion of the TA DNA much more than the CG DNA. Fig. 5B shows the superposition of the DNA from the Ca 2ϩ -bound mutant structures, Q138F/CG/ Ca 2ϩ (light blue) and Q138F/TA/Ca 2ϩ (pink) onto that of wtHin-cII/CG/Ca 2ϩ (gray). The TA DNA now resembles that bound to wild type much more closely than without Ca 2ϩ (Fig. 5A, red). Thus, Ca 2ϩ "corrects" the DNA structure as well. Fig. 5, C-D, shows the r.m.s.d. for atoms of each nucleotide when all atoms are used for the superposition. Fig. 5C shows the r.m.s.d. of only backbone atoms, whereas Fig. 5D shows the r.m.s.d. for only the base atoms. The DNA backbone r.m.s.d. (Fig. 5C) shows a systematic trend of increasing differences from nucleotides 8 to 10 in structures Q138F/CG (dark blue), Q138F/CG/Ca 2ϩ (light blue), Q138F/TA/Ca 2ϩ (pink) and increasing differences from nucleotides 5 to 9, then falling slightly at nucleotide 10, in Q138F/TA (red). The r.m.s.d. are much larger at nearly all nucleotides in the Q138F/TA structure (red, Fig. 5, C-D). The intercalation site and the location of the mutation Q138F is between nucleotides 10 and 11. Thus, the larger deviations in the DNA backbone structure near nucleotide 10 appear to be a result of the mutation and changes in contacts at the protein-DNA interface (described in more detail below). The r.m.s.d. of the base atoms (Fig. 5D) show a similar pattern with respect to each structure; however, the greatest deviations in Q138F/TA (red) occur at the center two base pairs (nucleotides 7-8), as seen in the superposition (Fig. 5A).
The unusual distortion of cognate DNA bound to wild type HincII at the center two base pairs (GTYRAC) is shown in Fig.  1B, right. For comparison, the same DNA, but with B-form conformation, is shown in Fig. 1B, left. When the DNA is bound to wild type HincII, the pyrimidine base of the center step, Cyt 7, is unstacked from its 3Ј neighbor, Gua 8. In addition the center step purines from opposite strands, Gua 8 and Gua 8Ј, are shifted toward one another, increasing their stacking surface area. This type of conformation is known as a cross-strand purine stack (CSPS). The CSPS was identified in the original structure and may play a role in indirect readout by HincII. Furthermore, modeling studies suggest a connection between DNA intercalation by glutamine 138 (green, Fig. 1A) and the CSPS (13). The conformations of the CG and TA DNA when bound to Q138F HincII are shown in Fig. 6, A-D. The CG DNA bound to wild type HincII is shown in black. and the superpositions were done with the base atoms of one of the two purine bases (Pur 8, Fig. 6, A-D). The view is down the helical axis to illustrate the amount of base stacking by the two center step purines (Pur 8 and Pur 8Ј, Fig. 6, A-D). The DNA closest to that bound to wild type is the CG DNA of Q138F/CG (Fig. 6C), and Ca 2ϩ binding to this complex appears to actually decrease the amount of CSPS (Fig. 6D). The center step purines do not stack at all in Q138F/TA (Fig. 6A) but come closer when Ca 2ϩ is bound (Fig. 6B), although the stacking is still worse than in Q138F/CG/Ca 2ϩ (Fig. 6B).
Local Conformational Effects of the Mutation-Both tertiary and quaternary conformational changes relative to wild type HincII bound to cognate DNA are found to varying degrees in the four structures of Q138F HincII with cognate DNA with and without bound Ca 2ϩ . This is surprising, because residue 138 extends away from the protein in both the unliganded (38), and DNA-bound structures of wild type HincII (13). These changes differ depending on the sequence of the bound DNA as well as on the presence or absence of Ca 2ϩ . At the r.m.s.d. level, the changes are greatest when comparing whole dimers to that of wild type and especially in the absence of bound Ca 2ϩ . The subunits are opened, widening the DNA binding domain. The increased rotation relative to the wtHincII/CG/Ca 2ϩ of one subunit relative to the other is 10.9°and 9.2°in Q138F/TA and Q138F/CG, respectively, without bound Ca 2ϩ . The binding of Ca 2ϩ causes the Q138F HincII dimer to close up around the bound DNA, reducing the subunit rotations (relative to wild type) to 7.0°and 4.0°, respectively. The locations and magnitudes of differences in conformation between two structures can be viewed using difference distance plots (Fig. 7). These figures highlight changes in the protein conformation relative to the wtHin-cII/CG/Ca 2ϩ structure. They are calculated by first meas-uring the distances between all C␣ atoms within the wtHin-cII/CG/Ca 2ϩ structure and measuring the same distances within a Q138F/DNA structure. These distances are then compared. Large differences in C␣-C␣ distances indicate a change in the protein conformation within the mutant structure relative to wild type. The differences in C␣-C␣ distances are plotted using colored boxes inside a matrix where the residue number of the first C␣ of the pair is specified along the horizontal, and the residue number of the second C␣ is specified along the vertical. Blue indicates an increase in the distance between selected C␣ atoms relative to the wild type structure, whereas red indicates a decrease. The upper left and lower right quadrants refer to changes in C␣-C␣ distances within the same monomeric subunits, subunits A or B, respectively, whereas the upper right and lower left quad- rants show changes in the C␣-C␣ distances of the two different subunits in the dimer, A and B and in B and A, respectively.
The difference distance plots of Fig. 7 indicate that the Q138F mutation has caused disruptions in the protein structure that depend on the sequence of the bound DNA, which is further changed by the binding of Ca 2ϩ . To better understand the origin of these changes, we looked closely at the protein-DNA interface at the site of mutation (Fig. 8). It is apparent from these superpositions that relative to wild type (gray, Fig. 8,  A-D), the backbone around residue 138 is pulled back away from the DNA in the Q138F/DNA structures. This appears to allow better stacking between the phenylalanine side chain and the two neighboring cytosines, Cyt 10 and Cyt 11.
Probably as a response to the shifting of the protein main chain at 138, the protein main chain at the neighboring residue Ala-137 is flipped, such that the side chain of Ala-137 no longer points toward the DNA where it had made a van der Waals contact to the deoxyribose of Cyt 10 in the wtHincII/CG/Ca 2ϩ structure (Fig. 8). The flipping of Ala-137 occurs in both subunits of all DNA-bound Q138F HincII structures. The similarities within the different subunits of the different DNA-bound structures ends here, with some structures showing a recovery of the protein trace to follow that of wild type, other subunits shift with a structural portion active as a rigid domain, whereas still others shift smaller segments of each subunit. Rotation of a greater portion of the subunit (i.e. rigid body rotation versus local distortion) is more likely to occur in complexes without bound Ca 2ϩ .
To understand why distortions occur in the structures of Q138F HincII bound to DNA, with and without Ca 2ϩ , and also why they involve different segments of the protein in each, we took the DNA structure from each Q138F/DNA structure, superimposed it onto the bound DNA of wtHincII/CG/Ca 2ϩ , and asked where steric conflicts occur. The most consistent conflicts occurred between the side chain of Thr-130 and the phosphate of Ade 9. Interestingly, this did not occur in the DNA bound to the subunits of Q138F HincII that show less subunit rotation relative to wild type ( Table 5). The reason for this appears to be a conformation taken on by the phosphate of Ade 9 (Table  5). When the gamma (C5Ј-C4Ј) angle of the Ade 9 phosphate is close to 170°, the phosphate appears twisted, or rotated, relative to the wild type conformation (gamma angle of 30 -40°or Ϫ90°, Figs. 8 and  9). This rotation moves the nonesterified oxygens of the Ade 9 phosphate away from the side chain of Thr-130. The twisting of the phosphate is present only in subunits with 3.2°or less of rotation (Table  5), which is always the subunit arbitrarily assigned as A in these structures.
To understand why the conformation of the phosphate of Ade 9 should be affected, we looked closely at the DNA around the site of intercalation. The change in contacts to the DNA backbone, namely the loss of Ala-137 to Cyt 10 and gain of S136 to Cyt 11, appear to be responsible for twisting the sugar-phosphate DNA backbone (Fig.  8, A-D). The base planes of the outer two base pairs of the recognition sequence (i.e. GTYRAC) are unmoved relative to the intercalating segment (Figs. 5B and 8), and the positions of the phosphorus atoms of Cyt 11 and Cyt 10 are not significantly shifted; however, the sugar rings of Cyt 11 and Cyt 10 as well as Ade 9 are rotated, causing a shift and rotation of the Ade 9 phosphate (Figs. 8 and 9). This would result in a steric conflict between this phosphate and Thr-130 unless one of two things happens; either 1) segments of the protein subunit shift to increase the distance between Thr-130 and Ade 9 phosphate, or 2) the phosphate of Ade 9 rotates to move the offending phosphate oxygen out of the way of the Thr-130 side chain.
Active Site Conformations-To understand how the binding of Ca 2ϩ has appeared to alter the conformations of the protein, DNA, and the protein-DNA interface and to investigate the origins of the reduced cleavage activity by the mutant enzyme, the area in the vicinity of the Ca 2ϩ was examined. DNA-cleaving enzymes utilize divalent cations in their catalytic mechanism, and the natural cofactor of HincII is Mg 2ϩ . Within type II restriction endonucleases, Ca 2ϩ acts as a competitive inhibitor of Mg 2ϩ and is usually seen to occupy a position within the active site at the protein-DNA interface (16). Sometimes one Ca 2ϩ is seen bound per active site, sometimes two, depending on the actual enzyme (39). In wild type HincII, a single Ca 2ϩ ion was seen to bind each active site when bound to cognate CG DNA (wtHincII/ CG/Ca 2ϩ ) (16), which caused very little change in the protein or DNA structure.
In the case of the DNA-bound Q138F HincII structures, each active site binds one Ca 2ϩ ion, similar to that found in wtHincII/CG/Ca 2ϩ (Figs. 4 and 9, C-D).
Although two divalent cations are expected to bind in the case of Mg 2ϩ and as observed with Mn 2ϩ (17), Ca 2ϩ binds in only one of the two positions seen to be occupied by Mn 2ϩ in both wild type and Q138F HincII. The Ca 2ϩ bound in the active site bridges the protein and DNA. It is ligated to both active site aspartate residues, Asp-114 and Asp-127, as well as to the scissile phosphate (the phosphate to be cleaved) found between the central pyrimidine-purine step. For this reason the Ca 2ϩ -bound structures appear to resist displacement of the protein near these residues to maintain these contacts. In addition, the efficiency of catalysis of DNA cleavage will be dependent on the precise positioning of any groups within the active site, and therefore, mispositioning of important active site side chains, the divalent cation, and/or the DNA can result in a decrease in the rate of DNA cleavage.
We wanted to characterize exactly how much, if any, mispositioning occurred within the active sites of the subunits of the various structures. Fig. 9 shows each active site superimposed onto wild type HincII bound to CG DNA with Ca 2ϩ (wtHincII/CG/Ca 2ϩ ). The superpositions were performed with the C␣ atoms of the active site residues Asp-114, Asp-127, and Lys-129 to show the mispositioning of the DNA relative to these residues. The misalignment is much greater with-  AUGUST 18, 2006 • VOLUME 281 • NUMBER 33 out bound Ca 2ϩ (Fig. 9, A and B) than with bound Ca 2ϩ (Fig.  9, C and D). On the whole mispositioning is greater in Q138F/CG/Ca 2ϩ than in Q138F/TA/Ca 2ϩ . This is significant as the single turnover rate of cleavage is depressed to a greater degree with CG than with TA at the center step of the recognition sequence, GTYRAC.

Indirect Readout of DNA by HincII Endonuclease
Mispositioning in the active site is a consequence of distortions of the protein and alterations in the protein-DNA interface. The protein distortion appears to be linked to the clash between the shifted phosphate of Ade 9 and the side chain of Thr-130 (see above). Because it is the B subunits of the two Ca 2ϩ -bound structures that exhibit differences in active site alignments, we wondered how differences in the DNA conformation of the Q138F/TA/Ca 2ϩ structure could lead to differences in protein segment displacements. Fig. 6E utilizes a superposition based on the atoms of the base of one of the two central purine rings and shows the position of the Ade 9 phosphate on the opposite strand. It is clear that the stacking between the superimposed Pur 8 and the opposing Pur 8Ј is not the same in all structures and especially the stacking of Pur 8Ј and its neighboring Ade 9. In addition to the different DNA stacking, the response by the protein also differs.
The ability of the Q138F/TA/Ca 2ϩ complex to better accommodate the effects of the mutation yet still maintain a fairly well aligned active site compared with Q138F/CG/Ca 2ϩ must be attributable to the center step TA sequence. More distortions occur at the local level within the protein and DNA subunits in the TA structures in general. Because the protein sequence does not change in comparing Q138F/TA/Ca 2ϩ and Q138F/ CG/Ca 2ϩ , the changes in DNA sequence and consequently structure must be critical to allowing the adjustments of the protein and protein-DNA interface that in the end result in better alignment of the active site in Q138F/TA/Ca 2ϩ . Fig. 6B shows that the CSPS is much less in Q138F/TA/Ca 2ϩ ( pink), and the stacking of Ade 8 3 Ade 9 also differs from wild type, resulting in a differing positioning of the phosphate of Ade 9 (Fig. 6E ). The position of the Ade 9 phosphate is less at odds with the side chain of Thr-130, resulting in less protein displacement and leading to a better aligned active site. Thus flexibility of the DNA stacking (or rather, greater deviation from the wild type-induced DNA structure) results in less steric con-flict with the protein in the case of the TA DNA, a smaller protein segment displacement, and a better aligned active site. Thus, the greater stability of the wild type-induced conformation of the cognate DNA with CG at the center step has resulted in DNA less able to adapt to the distortions induced by the Q138F mutation.

DISCUSSION
The residue at position 138 of HincII was targeted for mutagenesis to better understand the role of DNA intercalation by Gln-138 in DNA sequence discrimination. The mutation of glutamine 138 to phenylalanine resulted in several important functional consequences. First, the mutant HincII is reduced in its cleavage rate by 16 -214-fold. Second, the cleavage rate is reduced to a greater extent for DNA containing CG at the center step (GTCGAC) than TA (GTTAAC). However, the binding affinity of the enzyme for these DNA sequences was not greatly affected. The crystal structures of Q138F HincII bound to the two cognate DNA sequences as well as wild type HincII with TA DNA were solved to better understand how the mutation Q138F resulted in these functional results. Three questions to be addressed by the structural analyses were posed in the introduction. These will be addressed in order.
Why Does the Q138F Mutation Reduce the Single Turnover Cleavage Rate of Cognate DNA by HincII?-The positioning of important groups in the active sites of these complexes is misaligned relative to the wild type HincII structure bound to cognate DNA (and Ca 2ϩ ). The misalignment is worse without bound Ca 2ϩ , indicating that the binding of Ca 2ϩ serves to better align these groups. This is in contrast to the wild type HincII structures, where little difference is seen with and without bound Ca 2ϩ (16). The requirement of realignment to bind Ca 2ϩ in the mutant structures may decrease the overall affinity for Ca 2ϩ . In addition, the misalignment is expected to decrease the rate of DNA cleavage through either a higher energy transition state for cleavage or a requirement for conformational changes to better align active site groups. Alternatively, the mutation may also alter the rates of other steps in the reaction pathway preceding cleavage, which include binding, translocating, DNA distortion including intercalation, and protein conformational changes (see below).
Why Is the DNA Binding Affinity of HincII for Cognate DNA Altered Very Little by the Q138F Mutation?-When a modification is introduced into an endonuclease at the protein-DNA interface, particularly if distant from the active site, it is expected, due to the similarity in structure, that the mutation will destabilize both the ground and transition states to the same extent (40). In such cases the effect of the destabilizing modification will be exhibited at the binding, rather than cleavage step of the reaction pathway. However, in many cases it is known that such modifications can lead to conformational readjustments in the ground state protein-DNA complex to minimize the energetic cost of that modification (10,34,40). Such structural adjustments result in a less than expected effect on the binding affinity, and the phenomenon has been termed adaptation (40). Because of the requirement for a precise arrangement of active site groups, the transition state is more structurally stringent The rotation angle was calculated by first superimposing the 6 base pairs of the DNA recognition sequence of the listed structure onto that of wild type HincII bound to CG and Ca 2ϩ , then calculating the rotation required to bring residues 126 -132 and 167-175 of the listed structure onto the same residues of wild type HincII/CG/Ca 2ϩ .
than the ground state complex, and thus, the destabilization caused by the modification cannot be minimized by structural adaptation. With the destabilized transition state and stabilized ground state, effects will be seen on the cleavage, rather than binding step of the reaction pathway. Thus, structural adaptation is consistent with the results obtained here for the Q138F mutant of HincII. The mutation had nearly no effect on the binding affinity of HincII for cognate DNA but had a 16 -214-fold effect on the single turnover rate constant for cleavage of DNA. We find that there are many changes in the protein, DNA, and protein-DNA interface structure, consistent with structural adaptation. These changes must work to minimize the detrimental effects of the mutation on the DNA binding affinity. Why Does the Q138F Mutation of HincII Reduce the Single Turnover Rate of Cleavage More for GTCGAC than for GTTAAC?-The mutation Q138F of HincII affected the overall and relative multiple turnover rates of the enzyme for its cognate sequences. However, the mutation did not affect the mag-nitude or specificity of binding of the enzyme for these substrates but did affect the single turnover rate of cleavage. Before any structural interpretations can be made of this result, we must first consider the molecular step or steps measured by the single turnover rate of cleavage.
The single turnover rate of cleavage as implemented here is a measurement meant to isolate a single round of DNA cleavage up to and including the actual chemical cleavage of DNA. The product release step is not measured because all product, whether bound to the enzyme or released into solution, is measured in the assay. When the reaction is initiated by mixing enzyme and DNA, both in buffer containing Mg 2ϩ , steps including binding, DNA bending, DNA intercalation, and protein conformational changes as well as the actual cleavage step contribute to the observed rate of product formation. In many cases, and especially at higher concentrations of enzyme, the actual binding step can be very fast relative to later steps. Alternatively, the reaction can be initiated by the addition of Mg 2ϩ to premixed enzyme and DNA. If the enzyme binds to the recog- FIGURE 9. Stereo diagrams of active site superpositions, wild type HincII/CG/Ca 2؉ (gray) and Q138F/DNA structures (colored). A, subunit A of Q138F/ TA (red) and Q138F/CG (blue). B, subunit B of Q138F/ TA (red) and Q138F/CG (blue). C, subunit A of Q138F/TA/Ca 2ϩ (pink) and Q138F/CG/Ca 2ϩ (cyan). D, subunit B of Q138F/TA/Ca 2ϩ (pink) and Q138F/CG/Ca 2ϩ (cyan). nition sequence strongly in the absence of the divalent cation, then the pathway to DNA cleavage will involve binding of the Mg 2ϩ as well as the DNA hydrolysis. However, no binding to cognate DNA containing CG at the center step was detected in gel shift assays without Ca 2ϩ with either Q138F or wild type HincII. The DNA concentration was 1 nM in these experiments, and enzyme levels up to 1 M were tested, indicating that the equilibrium dissociation constant would be higher than 1 M. Thus, binding of the DNA by the HincII enzymes before the addition of Mg 2ϩ would not occur to any great extent in the single turnover assays using 0.8 -3.0 M HincII. The fact that both wild type and Q138F HincII have been crystallized with cognate DNA in the absence of divalent cation indicates that binding does take place under the crystallization conditions; however, the enzyme and DNA concentrations are very high (90 and 180 M), and solution conditions are radically different (25% polyethylene glycol) from the reaction buffer.
Both for the mutant and for wild type HincII, the single turnover cleavage rate constants were found to be independent of enzyme concentration (in the range of 1 ϫ 10 Ϫ7 to 3 ϫ 10 Ϫ6 M). In addition, we found very little dependence on Mg 2ϩ concentration (in the range of 10 -50 mM, with buffer adjusted to constant ionic strength); therefore, protein binding to DNA and Mg 2ϩ binding to the protein-DNA complex is probably not rate-limiting. The data fit well to a single exponential, implying a single rate-limiting step. We do not have any information to indicate whether this step is Mg 2ϩ -dependent target site location, Mg 2ϩ -dependent DNA distortion, Mg 2ϩ -dependent protein conformational changes, or the actual (Mg 2ϩ -dependent) rate of DNA cleavage. Further studies will be needed to isolate these steps.
From the structural results, it is apparent that although the protein and DNA conformations overall are more distorted in the Q138F/TA structures, the active sites in the presence of Ca 2ϩ are better aligned in the Q138F/TA/Ca 2ϩ structure than in the Q138F/CG/Ca 2ϩ structure. The effect of Mg 2ϩ binding on the structures is likely to be similar to that of Ca 2ϩ binding. If the increased depression of the single turnover rate of cleavage with the CG substrate relative to the TA as a result of the mutation is due to effects on the rate of chemical cleavage, then the greater misalignment in the Q138F/CG/Ca 2ϩ structure is consistent with this result. In addition, if it is due to a slower conformational change on the part of the Q138F/CG/Ca 2ϩ structure, then the structural observations are also consistent with this result, as we find much more flexibility in terms of structural deviations from the wild type in the Q138F/TA and Q138F/TA/Ca 2ϩ structure than in Q138F/CG and Q138F/CG/ Ca 2ϩ . Although the structures do not directly report on conformational dynamics, much greater conformational freedom is suggested in the Q138F/TA/Ca 2ϩ and Q138F/TA structures by the greater differences in structure from wild type. Both of these effects seem to originate in the ability of the DNA bases to take on different stacking around the center TA step (Fig. 6, A,  B, and E), by either allowing the multiple conformations themselves, or by allowing multiple structural modifications to compensate for the protein distortions that otherwise lead to poor active site alignment.
There is little distinction between these two arguments if one believes that the cleavage reaction transition state is purely stringent. This is because the actual reaction rate of the chemical cleavage event would be the same in both cases, since the active site arrangement by definition must be identical and the difference in single turnover cleavage rate would be due to the differences in the rate of attaining that stringent conformation. That is, the rates must differ due to a difference in the energy of a transition state for attaining a near attack conformation (the ground state reaction just before reaction), such as protein or DNA structural rearrangement. The energy of this transition state could be dependent on the degree of misalignment and/or the flexibility of the protein-DNA complex. Of course, it is entirely possible that the actual cleavage reaction (with bond making and breaking) proceeds through transition states of differing energies due to differences in the ability of the enzyme to stabilize the reaction transition state. Further studies into the kinetics of the cleavage reactions by the Q138F HincII and its various substrates will distinguish between these possibilities.
Summary and Conclusion-A DNA intercalating residue of HincII was mutated from glutamine to phenylalanine, Q138F, to test the connection between intercalation of DNA by this residue and indirect readout of the center YR step of the cognate recognition sequence GTYRAC. Multiple turnover velocity measurements showed a 6-fold decrease in preference for cognate DNA containing CG at the center step than for that containing TA, as a result of the mutation. In contrast, the binding affinity measurements of the wild type and mutant enzyme to the two cognate DNA sequences showed little difference between wild type and the mutant enzyme. Single turnover cleavage measurements showed a change of 3.5-fold in ratio of the rate constant of cleavage of cognate DNA containing TA relative to CG. The single turnover cleavage rate constant is also reduced 16 -214-fold by the mutation and reduced more for cognate DNA containing CG than that containing TA at the center step. A high resolution crystal structure of wild type HincII bound to TA DNA showed no significant differences from that containing CG. High resolution crystal structures of the mutant HincII, Q138F, bound to DNA showed marked changes in conformations at the tertiary and quaternary (i.e. dimeric) level as well as changes in DNA conformation and the protein-DNA interface, which were dependent on the center DNA step sequence and the presence or absence of bound Ca 2ϩ . The perturbations in structure were traced from the site of mutation showing a change in position of residue 138 in the intercalated DNA. This shift results in a marked shift in the position of the side chain of the adjacent residue, Ala-137, which may explain the observed change in a deoxyribose conformation normally contacted by Ala-137 in the wild type structures. The altered sugar conformation propagates changes in the sugar-phosphate backbone to adjacent nucleotides, leading to a steric conflict between the O5Ј oxygen of the phosphate of Ade 9, just adjacent to the scissile phosphate, and the side chain of Thr-130. The conflict is avoided in different ways in different subunits of the different structures. In three of the four structures, one subunit shows a rotation about the ␥ angle (at C5Ј-C4Ј) of the sugar ring, moving the phosphate away from Thr-130. The other subunits of these dimeric structures have residue 130 shifted away from the bound DNA by varying degrees. The binding of Ca 2ϩ at the active site very near to this point of steric collision acts to favor a conformation closer to that seen in the wild type HincII DNA-bound structures. The realignment of the protein-DNA interface at the active site as a result of Ca 2ϩ binding is more successful when TA is found at the center step of the recognition sequence than with CG. This difference in the ability of Ca 2ϩ to correct misalignment appears to be a result of an alteration in the stacking of the two center step adenine bases with each other as well as with the adjacent adenines (Ade 9)