ABSTRACT
The type I DNA methyltransferase M.EcoR124I is a multi-subunit enzyme that binds to the sequence GAAN6RTCG, transferring a methyl group from S- adenosyl methionine to a specific adenine on each DNA strand. We have investigated the protein-DNA interactions in the complex by DNase I and hydroxyl radical footprinting. The DNase I footprint is unusually large: the protein protects the DNA on both strands for at least two complete turns of the helix, indicating that the enzyme completely encloses the DNA in the complex. The higher resolution hydroxyl radical probe shows a smaller, but still extensive, 18 bp footprint encompassing the recognition site. Within this region, however, there is a remarkably hyper-reactive site on each strand. The two sites of enhanced cleavage are co-incident with the two adenines that are the target bases for methylation, showing that the DNA is both accessible and highly distorted at these sites. The hydroxyl radical footprint is unaffected by the presence of the cofactor S-adenosyl methionine, showing that the distorted DNA structure induced by M.EcoR124I is formed during the initial DNA binding reaction and not as a transient intermediate in the reaction pathway.
DNA methylation plays a vital role in controlling gene expression and in maintaining the integrity of the genome. Sequence specific methylation of DNA is carried out by a wide variety of enzymes collectively known as DNA methyltransferases (MTases). Type I DNA methyltransferases are complex multisubunit enzymes which bind to and methylate a specific target recognition sequence in order to protect the host DNA from restriction by the corresponding endonuclease (1 ,2 ). Unlike the smaller type II systems, the target recognition sequence for type I systems is asymmetric, consisting of two half sites 3-5 bp in length, separated by a non-specific `spacer' sequence 6-8 bp long. Methylation occurs at the N6 of two specific adenines on opposite strands, one within each half-site of the DNA recognition sequence. Type I restriction-modification (R-M) systems are encoded by three genes, encoding the subunits HsdS, HsdM and HsdR (responsible for specificity, methylation and restriction respectively). For methylation of the target sequence, only the HsdS and HsdM subunits are required (3 ).
Type I R-M systems have been classified into three families (IA, IB and IC) using a variety of genetic and biochemical criteria (3 -6 ). DNA sequence comparisons have shown that the HsdS subunit consists of two highly variable domains of 150-180 amino acid residues, and two or more regions that are well conserved within a given family (7 -9 ). The two variable regions of the HsdS subunit form independent target recognition domains (TRDs), each being responsible for recognition of one half of the bipartite DNA recognition sequence (10 ,11 ). Based on analysis of repeated sequences in HsdS, we have proposed a circular model for the domain organisation in HsdS which provides the required pseudo-dyad symmetry for the interaction with HsdM and with the DNA (12 ).
Two type I methyltransferases, EcoR124I and EcoKI, have been over-expressed and purified in sufficient quantities for biochemical and structural analysis. Both enzymes have been well characterised in terms of their subunit composition, DNA binding and enzymatic properties (13 -16 ). The two enzymes have a similar subunit stoichiometry, and share common features in their domain organisation, although they differ in a number of important respects, for example in their response to co-factor binding.
The EcoR124I methyltransferase (M.EcoR124I) consists of two copies of the HsdM subunit (each 58 kDa) and one HsdS subunit (46 kDa), to form a trimeric enzyme (162 kDa) with a subunit stoichiometry of M2S1. The binding affinity of M.EcoR124I for its cognate DNA recognition sequence (GAAN6RTCG) is 108 M-1; methylation of either strand of the DNA recognition sequence increases the catalytic activity of the enzyme at least 100-fold, but reduces the DNA binding affinity ~30-fold (16 ). We have shown for M.EcoR124I that DNA binding confers considerable protection from proteolysis (17 ). Likewise, chemical modification experiments show that a large fraction of the lysine residues on the surface of the protein, principally those in the HsdS subunit, are inaccessible in the DNA-protein complex (18 ). These results all indicate that the methyltransferase undergoes a significant conformational change when it binds to DNA.
X-ray solution scattering and circular dichroism have been used to determine the structural parameters of M.EcoR124I and its complex with DNA (19 ). A dramatic reduction (~70 Å) is observed in the overall dimensions of the enzyme following DNA binding. This structural transition appears to involve a large rotation of the HsdM subunits to clamp the DNA, and involves additional non-sequence specific interactions outside of the DNA recognition sequence. The circular dichroism spectrum shows that this structural transition in the MTase is accompanied by considerable distortion of the DNA structure as it is bound by the enzyme.
In order to elucidate the precise region of the DNA that is protected when the methyltransferase is bound, we have carried out DNA footprinting experiments using DNase I and hydroxyl radical cleavage. Unexpectedly, we also found evidence for two localised regions of distortion in the DNA recognition sequence in the vicinity of the two adenines that are the target bases for methylation by the enzyme.
M.EcoR124I was purified from an over-expressing plasmid pJS4M (20 ) in Escherichia coli JM109(DE3) in which the HsdM and HsdS genes are expressed from a T7 promoter. The enzyme was purified to homogeneity as described elsewhere (13 ). The protein was judged to be at least 99% pure by gel electrophoresis.
Oligonucleotides were purchased from Oswel DNA Services and Oligo Express. The M.EcoR124I recognition sequence is shown in bold type. The 85 bp oligonucleotide used for the DNase I footprinting was generated by an extension reaction of two 50mer oligonucleotides complementary over the 15 bases at the 3' end of each oligonucleotide. This complementary region contained the M.EcoR124I binding site, thereby eliminating any errors introduced by the `filling-out' procedure. The 50mer oligonucleotides were end-labelled on either the A or the B strand using T4 polynucleotide kinase (New England Biolabs) with [[gamma]-32P]ATP (ICN Flow). Any free [[gamma]-32P]ATP was subsequently removed using a Nuctraptm column (Stratagene) prior to precipitation with ethanol. The oligonucleotide was resuspended in 1* Taq polymerase reaction buffer (Promega) containing an equal ratio of all 4 nt. The 85mers were formed by extension using PCR and subsequently run on a 12% non-denaturing acrylamide gel in 1* TAE buffer (40 mM Tris-acetate pH 7.4, 1 mM EDTA). The radiolabelled bands were visualised and the correct products eluted from the gel following the crush-soak method (21 ).
The 57mer oligonucleotides used for the hydroxyl radical footprinting were labelled as described previously and a slight excess of the complementary unlabelled strand added to the labelled oligonucleotide in a buffer containing 10 mM Tris-HCl, pH 8.2, 5 mM MgCl2, 1 mM EDTA. The mixture was heated to 90oC for 10 min and left to cool slowly to room temperature before purification on a non-denaturing gel, as described previously.
DNase I protection footprinting was performed as described elsewhere (22 ). A titration of M.EcoR124I with the 85 bp DNA fragment labelled on the either the A or the B strand was prepared. These complexes were formed at 4oC in a buffer containing 50 mM Tris-HCl, pH 8.2, 5 mM MgCl2, 2 mM DTT prior to the addition of 0.05 U DNase I (Pharmacia). The DNA was digested for 2 min at room temperature in the same buffer, after which time the reaction was quenched by the addition of DNase I stop solution [1% (w/v) SDS, 50 mM EDTA, 20 [mu]g/ml glycogen and 150 mM NaCl]. The samples were precipitated with ethanol and resuspended in formamide gel loading buffer, normalised for counts, and run on a 12% sequencing gel. The gel was visualised using autoradiography.
The procedures adopted were as described elsewhere (23 ). A titration of M.EcoR124I with either the A or the B strand labelled 57mer DNA was made. Complexes were formed by incubation at 0oC for 15 min in a buffer containing 50 mM Tris-HCl, pH 8.2, 5 mM MgCl2, 2 mM DTT. Where used, the co-factor AdoMet was added to a concentration of 1 mM. The protein-DNA complexes were digested for 2 min at 37oC. The reactions were quenched by the addition of 10% glycerol, and were run on a 6% native polyacrylamide gel to separate free and bound fractions (24 ). The samples were subsequently precipitated with ethanol, resuspended in an appropriate volume of formamide gel loading buffer to normalise for counts, and run on a 12% sequencing gel. The gels were visualised by autoradiography and the bands quantitated using a Molecular Dynamics Phosphorimager and ImageQuanttm software on a Macintosh PowerPC.
The complex of M.EcoR124I with a 30 bp DNA duplex has been extensively characterised (13 ,16 ,17 ,19 ). With the longer DNA sequences required for DNA footprinting studies, there is the possibility of multiple binding of the protein to the DNA due to additional non-specific interactions, especially when high concentrations of protein are added. Thus we first checked binding of the MTase to the 85 and 57 bp fragments to be used in subsequent footprinting experiments, by means of gel retardation assays. Conditions were optimised to ensure that non-specific binding was minimised in each case.
DNase I footprinting is a relatively low resolution technique, but has the advantage of being one of the mildest footprinting procedures and one least likely to perturb the binding (22 ). We initially used DNase I footprinting to investigate the DNA binding site on an 85 bp fragment. An 85 bp oligonucleotide duplex including the recognition site for M.EcoR124I was constructed by the extension of two overlapping 50 bp oligonucleotides (as described in Materials and Methods) and labelled with 32P on either the A or the B strand. The MTase was added in increasing concentrations to each of the labelled DNA duplexes, up to a 5:1 molar ratio. Following limited digestion of the complexes with DNase I, the fragments were run on a DNA sequencing gel (Fig. 1 ). The resulting DNase I footprint reveals a clear protected region at protein:DNA ratios >1. The addition of further protein has no appreciable affect on the size of the protected region.
Hydroxyl radical footprinting was performed on a complex of the MTase with a 57 bp oligonucleotide duplex containing the M.EcoR124I recognition sequence. Figure 2 shows the cleavage patterns of the bound DNA in the complex, compared with that from the free DNA. It is immediately obvious that the cleavage patterns obtained differ markedly from those generated by DNase I. In addition to clear regions of protection, within these regions 2-3 bases on each strand are cleaved much more readily in the complex compared with the free DNA. To allow accurate comparison of the free and bound cleavage patterns, the intensities of the bands were measured using a Phosphorimager (Molecular Dynamics) and analysed using the ImageQuanttm software. Line scans were obtained using data taken from the Phosphorimager for each track on the gel (Fig. 3 ). This clearly shows the sites which are preferentially cleaved in the protein-DNA complex, but the precise extent of the protected region is less well defined.
It has been suggested that type I methyltransferases undergo a structural change when the methyl group donor AdoMet is present in the ternary complex with DNA. For the related enzyme, M.EcoKI, AdoMet is required for efficient DNA binding and appears to affect sequence discrimination (15 ). Moreover, for M.EcoKI, AdoMet has been reported to affect certain DNA contacts, as judged by methylation interference (25 ). For M.EcoR124I, however, AdoMet does not affect DNA binding, and the co-factor has no effect on the structural parameters of the enzyme or its complex with DNA (19 ).
Nevertheless, in view of the very large perturbation to the DNA structure when M.EcoR124I binds to its recognition sequence, it was important to establish whether or not this phenomenon would be affected by the presence of AdoMet. Thus, we performed an identical set of experiments to observe the effects of AdoMet on the hydroxyl radical footprints. Figure 5 shows quite clearly that the co-factor has no significant effect, either on the extent of the footprint or on the anomalously high rates of cleavage of nucleotides in the vicinity of the target adenines.
We have investigated the protein-DNA interactions in the complex formed between M.EcoR124I and its cognate DNA, using DNase I and hydroxyl radical footprinting techniques. DNase I footprinting reveals a large protected region of the DNA with an apparent size of 27-28 bases on each strand, staggered by 3-4 bases in the 3' direction, with no accessibility to the enzymatic probe within this region. Clearly, the extent of the true footprint is overestimated due to the relatively large size of the DNase I and attempts have been made to allow for this in other systems (26 ). Allowing for the two phosphates either side of the cleavage site, and additional phosphates required for binding on the opposite strand, one can estimate that the region bound by the methyltransferase covers ~23 bases on either strand.
Hydroxyl radical footprinting also reveals a pronounced footprint on both strands of the DNA helix, although this is complicated by the internal regions of enhanced cleavage. The hydroxyl radical footprint is 17-18 bases on either strand. Compared with the DNase I footprint, this is ~4 bases shorter on the 5' side and 6 bases shorter on the 3' side. This difference is rather more than can be accounted for by the size of the enzymatic probe, and could arise from weaker DNA-protein interactions at the edge of the footprint which might allow the smaller hydroxyl radical probe to gain access.
It is instructive to compare these results with those obtained for the type II MTase M.HhaI (and the related enzyme M.SssI) (27 ). The observed DNase I footprint for M.HhaI extends to 14-15 bases on each strand, some 13 bases shorter than the DNase I footprint we observe for M.EcoR124I. For M.HhaI, the recognition sequence is only 4 bp, whereas that for M.EcoR124I is embedded in a 13 bp sequence. Moreover, the former enzyme is considerably smaller than the multisubunit type I MTases such as M.EcoR124I. The much larger footprint observed for the latter is therefore not unexpected, but does indicate quite clearly the more extensive interface between the protein and DNA in the larger type I enzymes. Likewise, if the hydroxyl radical footprints are compared, M.HhaI protects 9-10 bases on each strand, compared with 18 bases for M.EcoR124I.
For M.HhaI (and M.SssI), the cutting pattern observed shows that these enzymes bind to one face of the helix. For M.EcoR124I, this is clearly not the case. Indeed, the footprint extends over both strands of the DNA for a considerable distance-approximately two turns of the DNA double helix-with the exception of the two hyper-accessible sites within this region (Fig. 6 ). Unlike M.HhaI(and indeed most other DNA binding proteins), M.EcoR124I must completely enclose this region of the DNA in the complex, covering a length of some 60-80 Å. So how does the DNA helix enter into the active site of the MTase? To accomplish this, the enzyme must undergo a dramatic structural change as it binds the DNA. X-ray scattering experiments have previously shown that the MTase undergoes a very large conformational change as it binds to its recognition sequence, folding around the DNA to form a considerably more compact structure with a reduction of almost 70 in the overall dimensions of the complex, compared with the free enzyme (19 ). Our DNA footprinting experiments show the extent of the DNA-protein contacts that result in this more compact structure.
We thank our colleagues Drs Ian Taylor, Iain Manfield and Michelle Webb for their generous advice on aspects of this work, and to Damian Watts for assistance with protein preparation. This work has been funded by the Wellcome Trust through the award of a studentship, a project grant and an equipment grant.
*To whom correspondence should be addressed. Tel: +44 1705 842678; Fax: +44 1705 842053; Email: knealeg@biol.port.ac.uk
REFERENCES
Return