ABSTRACT
Bending of DNA induced by M.MspI, one of the m5C-DNA methyltransferases, has been investigated using circular permutation analysis. The M.MspI MTase induced sharp bends in DNA containing its recognition sequence 5'-CCGG-3' which was estimated to be 142 +- 4o and 132 +- 4o for circularly permuted DNA fragments of 127 and 1459 bp respectively. The bend centre was found to be asymmetric with respect to the CCGG sequence and appeared to exclude the `target cytosine'. An estimate of ~15 kcal/mol was obtained for the free energy associated with M.MspI-induced DNA bending.
Protein-induced bending of DNA is said to play an integral role in many biological processes involving protein-DNA interactions that govern high precision DNA transactions, notably replication, transcription initiation, recombination and DNA packaging (1 ,2 ). Transcription factor MerR can mediate repression as well as activation through stereospecific modulation of DNA structure. The repressor form of MerR bends DNA towards itself when bound to a promoter between the -10 and -35 regions. The activator conformation, Hg-MerR, relaxes these bends. This activator-induced unbending coupled with untwisting of the operator, remodels the promoter and makes it a better template for the poised polymerase (3 ). The significance of DNA bending in transcriptional regulation is further substantiated by observations that TATA-binding protein (TBP) bound to DNA which was pre-bent towards the major groove with 100-fold higher affinity than linear DNA of identical sequence and 300-fold higher affinity than DNA pre-bent towards the minor groove. A similar discrimination was observed with the holo-TFIID transcription complex (4 ).
Some restriction endonucleases belonging to type II R-M systems have also been found to induce bends in DNA. For example EcoRI, is reported to induce a bend angle of 50 +- 2o to 56 +- 6o in substrate DNA (1 ). The RsrI endonuclease bends DNA by ~50o and also unwinds the DNA by 25 +- 5o (5 ), a value close to that reported for EcoRI endonuclease. The bend angle induced by EcoRV endonuclease was found to be 44 +- 2o (6 ). The magnitude and orientation of DNA bending induced by various single proteins have been estimated by gel mobility shift methods. Upon recognition of the specific DNA sequence, the protein may simply stop at the cognate sequence without substantial conformational changes in DNA, as in passive recognition or a different conformation of DNA and/or protein in the recognition complex can be achieved compared to the average conformation of free DNA and/or protein, as in interactive recognition (7 ).
The m5C-DNA methyltransferases catalyze transfer of a methyl group from S-adenosyl-L-methionine to the C-5 position of cytosine within the recognition sequence of the substrate DNA (8 ). The catalytic mechanism of the enzymes is now understood. It involves formation of a covalent bond between a cysteine residue of the conserved PC motif of the MTase and carbon 6 of the target cytosine (9 ,10 ). This activates C-5 of the cytosine ring which, serving as a carbanion, accepts the methyl group from the sulfonium centre of AdoMet. Following methyl transfer, the proton at C-5 is abstracted by a basic residue of the enzyme which is [beta]-eliminated from C-6. Crystal structures for binary and ternary complexes involving the m5C-MTase HhaI, cofactor-AdoMet and substrate DNA (11 ,12 ) suggest that all m5C-MTases have an overall similar architecture (8 ) since the conserved motifs formed the core of the structure. This would therefore suggest well defined structural roles for the conserved motifs which could perhaps be extended to other m5C-MTases. Since the three dimensional structure and conformation of the enzyme is key to its mechanistic specificity, it may be proposed that the mechanism employed by HhaI for DNA modification might hold for other m5C-MTases. The HhaI methyltransferase flipped its target base by 180o out of the helix into the catalytic pocket of the enzyme. Flipping of the target base is a new phenomenon and its mechanism is yet to be understood. Another m5C-MTase has been found to flip out the target base in a crystal structure determined recently (13 ).
We have been investigating recognition and binding of DNA by MspI MTase, a m5C- DNA methyltransferase, which recognizes the sequence 5'-CCGG-3' and methylates the outer cytosine (14 ). The enzyme has been cloned, overexpressed, purified and characterized (15 ,16 ). M.MspI has been previously shown to form a sequence-specific complex with DNA where the recognition sequence CCGG was located asymmetrically within the binding site (17 ). In the present paper we have addressed the question of DNA distortion by M.MspI. The bend angle, locus and free energy change associated with the process are estimated.
Homogeneous MspI methyltransferase was purified as previously described (16 ). Restriction endonucleases, bacteriophage [lambda] DNA and Klenow fragment of DNA polymerase I were from New Englands Biolabs, USA. T4 polynucleotide kinase, DNA ligase and [Phi]X174 DNA were from Bangalore Genei, India. Reagents for SDS gel electrophoresis were from Sigma, USA and used as recommended. Deoxyadenosine 5'[[gamma]-32P]triphosphate (specific activity 2500 Ci/mmol) was procured from Bhabha Atomic Research Centre, India. Deoxynucleoside triphosphates were obtained from USB, USA. All other chemicals used were of the highest purity analytical grade.
Plasmid pBend2 (18 ), a kind gift from Dr Sankar Adhya, was used to generate 127 bp fragments containing 5'-CCGG-3' sequence for permutation analysis. The plasmid was maintained and propagated in Escherichia coli strain ER1727 under selection pressure of ampicillin at a concentration of 50 [mu]g/ml in LB medium. The plasmid was prepared from host cells using the alkaline lysis method (19 ) and was subsequently purified on agarose gel (0.8% w/v). The various steps employed to obtain the 127 bp permutated substrate are outlined in Figure 1 . In brief, these included purification of 121 bp EcoRI-SalI fragment from pBend2 on native polyacrylamide gel followed by ligation of two oligonucleotides 5'-AATTC-3' and 5'-TCGAC-3' to create EcoRI and SalI sites respectively. The non-clonable, radiolabelled 127 bp circular DNA molecule was obtained by employing the following reactions: filling in, kinasing with [[gamma]-32P]dATP and end ligation (19 ). Permutation of the 5'-CCGG-3' sequence was achieved by treating the above 127 bp circular DNA molecule with restriction endonucleases such as EcoRI, BglII, ClaI, XhoI, PvuII, KpnI, BamHI and SalI. The linear permuted substrates were subsequently purified on a 10% polyacrylamide gel under native conditions. A larger DNA fragment of 1459 bp which was permuted with respect to the recognition sequence, was obtained using BanI fragment from [Phi]X174 DNA. The BanI fragment was filled in and circularized as before and unique restriction sites for BstXI, MluI, DraI, TaqI and BanI were used to generate linear fragments (1459 bp) containing the permuted CCGG sequence. In contrast to the 127 bp fragments, the 1459 bp fragments were non-radioactive and were purified on a 3.5% polyacrylamide gel.
The DNA binding reactions were carried out as described earlier (17 ) except for a few modifications as mentioned below. A typical reaction mixture for M.MspI-DNA complex formation included 50 mM Tris-HCl (pH 8.0), 10 mM Na2EDTA, 7 mM [beta]-mercaptoethanol, 0.26 mg/ml BSA and 10% glycerol in a reaction volume of 10 [mu]l that contained an appropriate amount of M.MspI and [[gamma]-32P] labelled (127 bp) or unlabelled (1459 bp) DNA substrate with the permuted CCGG sequence. The reactions were carried out at 37oC for 5 min and were analyzed by electrophoresis on a 10% polyacrylamide gel under native conditions (127 bp fragment) and on a 0.8% agarose gel (1459 bp fragment). The bending angle was estimated from the autoradiograms (127 bp fragment) and the gel photographs (1459 bp fragment) by plotting the relative mobility of the M.MspI-DNA complex versus distance (in base pairs) of the CCGG sequence to the left end of the fragment (Fig. 1 ) following procedures described elsewhere (1 ). The bending locus was estimated from these experiments as described by Wu and Crothers (20 ). The effects of temperature (10-37oC), salt (0-250 mM) and MgCl2 (0-50 mM) on bend angle was investigated. Free energy changes ([Delta]Gbend) associated with M.MspI-induced bend in DNA were computed according to methods described earlier (21 ,22 ).
Two sets of circularly permuted DNA fragments, 127 bp radiolabelled and 1459 bp non-radiolabelled, were constructed to investigate the M.MspI-induced bending of DNA. The EcoRI-SalI fragment containing a single M.MspI site from plasmid pBend2 was used to design a set of 127 bp fragments containing the permuted 5'-CCGG-3' sequence. From the sequence of [Phi]X174, one 5'-CCGG-3' sequence out of a total of five M.MspI sites was located between bases 1101-1106. Digestion of this [Phi]X174 molecule with BanI produced a 1459 bp linear fragment (between base positions 1019-2478) which contained many unique restriction sites. Cleaving these sites followed by cyclization led to permutation of the CCGG sequence along the 1459 bp fragment.
The gel mobility shifts caused by M.MspI binding to the two sets of circularly permuted restriction fragments (127 and 1459 bp) were analyzed. For both, the complex with the highest mobility resulted from the fragment with protein bound nearest the end, such as the PvuII fragment of 127 bp and the BanI fragment of 1459 bp. The complex with lowest mobility resulted from protein bound nearest the middle. The complexes where protein is bound at other locations, produced intermediate mobilities (Fig. 2 ). The mobility of a DNA fragment is related to its end-to-end distance (23 ) and, for a rigid DNA of contour length L with a bend exactly in the middle of the fragment, this distance can be shown to be L cos([alpha]/2), with `[alpha]' defined as the angle at which the DNA is bent from linearity. The end-to-end distance of a fragment with the bend at an end will be virtually unchanged and equal to its contour length (L). An empirical relation from these considerations was developed which suggested that the ratio of relative mobility of the complex with protein bound in the middle and that with protein bound at the end equalled cos([alpha]/2). Bend angle ([alpha]) for M.MspI-induced bends in DNA was determined using the above equation. The bend angles of 142 +- 4o and 132 +- 4o were estimated for the 127 and 1459 bp fragments respectively. However, the values calculated using this equation may be different from absolute bending angles, since factors other than end-to-end distance may influence the mobility of the free- or protein-bound DNA fragments.
Quantitative measurement of the relative gel mobility of a set of circularly permutated fragments allowed us to extrapolate to the position of the cut which would yield the maximum gel mobility and thereby locate the centre of the molecular bend (20 ). The relative gel mobilities of the complexes were the ratio of distances travelled from the origin by the DNA in the complex to that of free DNA. These values were plotted as a function of the distance of permuted sequence CCGG from the left end as represented in Figure 1 . Extrapolation of the nearly linear portions of this graph produced an estimate of ~68 bp for the bend locus in the 127 bp fragment (Fig. 4 A). The data contained in Figure 4 B further proved bending of DNA upon binding by M.MspI. However, due to the effect of large size of 1459 bp fragment on the conformation of the M.MspI-DNA complex as reflected in the gel mobility pattern (Fig. 4 B), an unambiguous determination of the bend locus in the 1459 bp fragment was not possible.
The free energy associated with the M.MspI-induced DNA bending could be estimated using the following equation (21 ):[Delta]Gbend = RTP/2 F ([Delta][theta])2 1
where P is the persistence length of DNA (~127 bp), F is the number of base pairs in which the bend occurs, and [Delta][theta] is the total bend angle in radians. Inserting numbers and converting to [theta] in degrees, the above equation would become (22 ): [Delta]Gbend = 0.012/ F (bp) ([Delta][theta])2 kcal/mol2
Since M.MspI protected 16 bp when bound to DNA containing the CCGG sequence (17 ), the number of base pairs in the bend region could be estimated as 16 bp as suggested by Liu-Johnson et al. (22 ). Using 142o for the bend angle, the [Delta]Gbend was computed to be 15 kcal/mol.
The induction of a bend upon binding to DNA is a well documented feature for many proteins and enzymes that interact with DNA (5 ,6 ,20 ,22 ,25 ). This phenomenon has not been reported for any m5C-DNA methyltransferase to date. However, an N6-adenine DNA methyltransferase, M.EcoRI, has been shown to bend DNA by 52o (26 ). The available crystal structures of MTase-DNA complexes for M.HhaI (12 ) and M.HaeIII (13 ) did not demonstrate either a bend or a kink in the bound DNA. However, a comparison of the structures suggested variation in the mode of recognition by these two MTases. This indicated that although m5C-DNA MTases share a common architecture, they might differ with regard to their interaction with DNA. Particularly, larger MTases that have longer N-termini could be expected to behave differently. This has been further substantiated by the observation that M.MspI displayed topoisomerase-like activity while its isoschizomer M.HpaII and the well-characterized M.HhaI did not (27 ). This encouraged us to analyse the M.MspI-DNA complex in solution to examine distortion, if any, in the bound DNA.
The results demonstrated that M.MspI, whose DNA binding properties were described earlier, caused severe deflection of DNA in its sequence-specific complex. Circular permutation analysis has now become a well established procedure to estimate the magnitude of bends in DNA that are either protein-induced or sequence-directed. We have used this technique to investigate M.MspI-induced bends in DNA fragments of 127 and 1459 bp in length. The estimated magnitude of deviation (from linearity) of 142 +- 4o for the 127 bp fragment and 132 +- 4o for the 1459 bp fragment were comparable to those of regulatory proteins, such as 90-180o by E.coli CAP protein (22 ), but were significantly higher than the 86-90o shown by HMG domain proteins (28 ). The bend angles induced by restriction enzymes, 44 +- 4o by EcoRV and 50o by isoschizomers RsrI (5 ) and EcoRI (1 ), are much less compared to M.MspI-induced bend angles. These differences in protein-induced bend angles in DNA could arise due to functional needs of the various proteins. For instance, the restriction enzymes may not need to cause extreme distortion in DNA since they carry out hydrolysis of the phosphodiester bond on the DNA backbone which could be easily accessible. In contrast, modification methyltransferases have to perform catalysis on a base embedded in the interior of the DNA helix and so might require DNA to be distorted to a great extent as demonstrated by M.MspI. However, other m5C MTases: M.HhaI (12 ) and M.HaeIII (13 ) appear to minimize such a distortion by base flipping.
A reduction of ~10o in bend angle was observed with the larger DNA fragment. This decrease might be due to the fact that the dynamic bending becomes significant and the apparent extent of protein-induced bending gets reduced in the case of longer DNA fragments, as was observed for IHF and EcoRI-induced DNA bends (1 ). Temperature, sodium chloride and magnesium chloride were reported to influence the k-factor, an index of migration through the gel, of the sequence-directed static bend in kinetoplast DNA fragments (29 ) and in synthetic oligonucleotides carrying `A-tracts' (24 ). The effect of temperature on the k-factor might be due to its influence on DNA twist angles (30 ) and local DNA structure (31 ). In contrast the M.MspI-induced bend was unaffected by temperature; the concentration of the sodium chloride did not affect this bending either. The lack of an NaCl effect on DNA bending in the present study was consistent with previous studies on CAP and lac repressor-induced bending of DNA, but contrary to the sequence-dependent bends in kinetoplast DNA (30 ).
Magnesium chloride was, however, found to enhance the M.MspI-induced bend angle (Fig. 3 ). The effect of Mg2+ ions could be explained by its site specific binding causing a change in the curvature of the bent DNA that might result in an increase in the magnitude of DNA deflection (24 ). A similar effect of Mg2+ was reported for bends in kinetoplast DNA and synthetic oligonucleotides with `A-tracts'. The relative mobility pattern of M.MspI complexes with a circularly permuted 127 bp fragment was symmetrical (Fig. 4 A) suggesting that there was only one major bending locus in the fragment. Asymmetric variation in gel mobility pattern was, however, recorded with the larger fragment (1459 bp) (Fig. 4 B) indicating a more complicated shape whose migration through the gel was being affected by some unknown factors. It was therefore not possible to assign an unambiguous bend locus for the longer DNA. But the centre of the molecular bend in the 127 bp fragment was located at base position ~68 corresponding to the point of intersection of linear portion of the curve that would yield maximum gel mobility. When analyzed with respect to the M.MspI recognition sequence 5'-CCGG-3', the bend centre did not seem to include the target `C' (outer cytosine) but contained bases toward the 3' end including, perhaps, some non-canonical bases. Thus the molecular bend induced by M.MspI in its sequence-specific complex with DNA was located to one side of CCGG sequence. This would be consistent with our earlier demonstration of the asymmetric location of the recognition sequence within the M.MspI binding site in a DNaseI footprint (17 ).
Investigation and analysis of a thermodynamic parameter, the free energy, often becomes crucial to understand and explain molecular interactions. For M.MspI-DNA interaction, the [Delta]Gbend (~15 kcal/mol) was approximately double the value of the [Delta]Gbend (8.5 kcal/mol) for CAP-DNA interaction. This would occur due to the small size of the M.MspI binding site (16 bp) compared to that of CAP (30 bp) since more energy is required to bend a shorter fragment than a longer one as evident from eqn. 2 (22 ). A higher value of [Delta]Gbend for M.MspI suggests that local DNA structure in the protein bound region is significantly altered as a consequence of bending. Such structural alteration and accompanying energy changes might be required for subsequent steps in the multiple interactions involving protein, DNA and cofactor AdoMet. The question of why this member of the m5C-DNA MTases needs to bend DNA while others do not (12 ,13 ,26 ), is intriguing. Further investigations along these lines would, however, be necessary to address such questions.
We are grateful to Dr R. J. Roberts for his support, encouragement and criticism of the manuscript. Critical comments from Dr P.H.von Hippel and Dr Ashok S. Bhagwat proved immensely helpful and we express our sincere thanks to them for the same.
REFERENCES
