Profile of the DNA recognition site of the archaeal homing endonuclease I-DmoI
Profile of the DNA recognition site of the archaeal homing endonuclease I- Dmo IClaus Aagaard+, Marianne J. Awayez and Roger A. Garrett*
Institute of Molecular Biology, Copenhagen University, Sølvgade 83 H, DK-1307 Copenhagen K, Denmark
Received January 22, 1997;Revised and Accepted March 5, 1997
ABSTRACT
I-DmoI is a homing enzyme of the LAGLI-DADG type that recognizes up to 20 bp of DNA and is encoded by an archaeal intron of the hyperthermophilic archaeon Desulfurococcus mobilis. A combined mutational and DNA footprinting approach was employed to investigate the specificity of the I-DmoI-substrate interaction. The results indicate that the enzyme binds primarily to short base paired regions that border the sites of DNA cleavage and intron insertion. The minimal substrate spans no more than 15 bp and while sequence degeneracy is tolerated in the DNA binding regions, the sequence and size of the cleavage region is highly conserved. The enzyme has a slow turnover rate and cuts the coding strand with a slight preference over the non-coding strand. Complex formation produces some distortion of the DNA double helix within the cleavage region. The data are compatible with the two DNA-binding domains of I-DmoI bridging the minor groove, where cleavage occurs, and interacting within the major groove on either side, thereby stabilizing a distorted DNA double helix. This may provide a general mode of DNA interaction at least for the LAGLIDADG-type homing enzymes.
INTRODUCTION
Homing-type enzymes have a wide phylogenetic distribution. They have been detected in group I introns of the nuclei and organelles of lower eukaryotes, in introns and inteins of the archaea, in proteobacterial phages (reviewed in 1 ) and, possibly, in cyanobacteria (2 ). These DNA endonucleases have been classified into groups on the basis of both the amino acid sequence motifs which they carry and their mechanism of DNA cleavage. The largest group, common to organelles of lower eukaryotes and archaea, contain two partially conserved LAGLIDADG boxes separated by ~100 amino acids. These homing enzymes bind to and cut sequence regions of double helical DNA, which are generally non-palindromic and extend over 15-25 bp, producing a 4 bp 3'-OH overhang (reviewed in 3 ). Their main function is to facilitate site-specific insertion, or homing, of the intron that encodes the endonuclease, into an intron-minus chromosome (4 ,5 ). Although this process inevitably requires a high level of DNA sequence specificity, a picture is emerging of some redundancy in the sequence being tolerated, at least for the homing enzymes I-CeuI from Chlamydomonas moewusii, I-SceII from Saccharomyces cerevisiae and I-PorI from Pyrobaculum organotrophum (6 -8 ).
In the present study, we have employed kinetic, mutational and footprinting approaches to investigate the DNA binding site of the homing enzyme I-DmoI (9 ) that is encoded by an intron within the single 23S rRNA gene of Desulfurococcus mobilis (10 ). This complements a parallel protein footprinting study of the DNA binding domains and metal chelating sites of I-DmoI (11 ,12 ).
MATERIALS AND METHODS
DNA cleavage assay
Aliquots of purified I-DmoI (a gift from J. Z. Dalgaard) (13 ) at 35 pmol/[mu]l were stored at -20oC in 50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 50% ethylene glycol, 1 mM DTT. The protein preparation was an equimolar mixture of the two forms of I-DmoI, expressed from linear and circular mRNAs, with sizes 22 and 22.6 kDa respectively, differing by five amino acids at the C-terminus (13 ). Their cleavage properties were indistinguishable (13 ) and it is unknown which form is expressed in vivo. The specific activity of the preparation was 1 U I-DmoI that corresponds to 1.75 pmol enzyme where 1 unit (U) is defined as the amount of enzyme required to digest 48.3 pmol of substrate to completion in 1 h at 65oC in 25 [mu]l of an assay buffer (13 ). The DNA substrate was generated from two complementary deoxyoligonucleotides with sequences 5'-GCCTTGCCGGGTAAGTTCCGGCGCG-3' and 5'-CGCGCCGGAACTTACCCGGCAAGGC-3' or a 400 bp PCR product amplified from the 23S rRNA gene of D.mucosus using the primers 5'-GGCACACCCCTGGGACCGC-3' and 5'-CCCTCCCACCTACTCTACGC-3'. Oligonucleo- tides were 5'-end-labelled using [[gamma]-32P]ATP and T4 polynucleotide kinase (both Amersham) and purified in 20% denaturing polyacrylamide gels. The two 25mer oligonucleotides were annealed at a ratio of 1:10 32P-end-labelled:unlabelled oligonucleotide in 10 [mu]l of 50 mM HEPES-KOH, pH 8.0, 50 mM KCl, 10 mM MgCl2 by heating at 85oC for 2 min and slowly cooling to room temperature. Substrates were cleaved at 65oC in 20 [mu]l cleavage buffer (50 mM HEPES-KOH, pH 8.0, 50 mM KCl, 10 mM MgCl2, 10% glycerol, 0.1% Triton X-100) and 1.5 [mu]g carrier DNA (pUC9). Reactions were stopped by adding 10 [mu]l formamide loading dye and cooling to 0oC. The substrate and cleavage products were separated in 10% denaturing polyacrylamide gels and quantified using an Instant Imager (Packard, USA).
Mutation, enrichment and cleavage of the DNA substrate
Single nucleotide mutants of the homing site were isolated by a single cleavage and PCR amplification step. A partially doped deoxyoligonucleotide 5'-CCTCTTAAGGTACCCAAATGCCTTGCCGGGTAAGTTCCGGCGCGCATGAATGGATCCACGCG-3' was synthesized where the underlined region was doped by 1.5%. The doping was performed by using a mixture of wild-type NTP (95.5%) and 1.5% of each of the other NTPs in each of the 25 rounds of synthesis. This produced substrate molecules where 32% carried wild-type sequences, 37% had a single mutation and 31% carried two or more mutations. The doped sequence was confirmed by DNA sequencing. It was hybridized to 500 pmol of 32P-5'-end-labelled oligonucleotide 5'-CGCGTGGATCCATTCATG-3' which was extended with 10 U T4 DNA polymerase in standard polymerase buffer and 25 mM of each dNTP at 30oC. The resulting double-stranded DNA was purified on a 10% polyacrylamide gel and 50 pmol were digested with 35 pmol I-DmoI for 20 min at 65oC. Substrate and cleavage products were separated in 10% polyacrylamide gels and undigested DNA was purified and amplified using the two oligonucleotide primers 5'-CCTCTTAAGGTACCCAAAT-3' and 5'-CGCGTGGATCCATTCATG-3'. The amplified DNA was digested with BamHI and KpnI (Amersham) and cloned into the M13mp19 vector digested with the same enzymes. Mutations were identified by sequencing the different clones using the universal M13 primer. Mutant substrates for I-DmoI cleavage assays were prepared as PCR fragments that were amplified from M13 clones using 20 pmol of both M13 universal and reverse primer. For each clone, two separate amplifications were performed with either the universal primer or the reverse primer being 5'-end-labelled.
A minimal 11 bp substrate was prepared by annealing 20 pmol of each of the deoxyoligonucleotides 5'-CGGGTAAGTTC-3' and 5'-GAACTTACCCG-3', as described above, and ligating the product into SmaI-digested pUC18. The ligation mixture was then amplified by PCR using 50 pmol of 32P-labelled universal sequencing primer 5'-GTAAAACGACGGCCAGT-3' and unlabelled reverse sequencing primer 5'-AACAGCTATGACCATG-3'. The amplified DNA (109 bp) resulting from the insertion of the oligonucleotides was purified on a 10% polyacrylamide gel and digested with 35 or 3.5 pmol I-DmoI at 65oC for 20 min, or with 1 U HincII or SmaI (Amersham) for 1 h at 37oC.
DNA footprinting
Probing with DNase I was performed by incubating the I-DmoI-substrate complex in 50 [mu]l with 0.001 U DNase I (Boehringer) for 10 min at room temperature. The reaction was started by adding 1 [mu]l 0.5 M MgCl2 and stopped by adding 150 [mu]l of 100 mM Tris-HCl, pH 8.0, 10 mM EDTA, 1% SDS and 1 [mu]g yeast tRNA. After precipitation, samples were lyophilized, suspended in formamide loading dye, and electrophoresed in 10% polyacrylamide sequencing gels containing 7 M urea and subjected to autoradiography.
End-labelled PCR fragments, with and without single site mutations, were complexed with I-DmoI and subjected to probing with dimethylsulphate (DMS). I-DmoI (0.1-3.5 pmol for the wild-type substrate and 3.5 pmol for the mutant substrates) was bound to 0.2 pmol substrate by heating at 65oC for 5 min in 50 [mu]l of 50 mM HEPES-KOH, pH 8.0, 100 mM KCl, 1 mM DTT, 1% glycerol and 0.1% Triton X-100. The DNA was methylated with 0.05% (vol/vol) DMS for 5 min at 65oC and the reaction was stopped by adding 150 [mu]l 100 mM Tris-HCl, pH 8.0, 250 mM 2-mercaptoethanol. After precipitation with ethanol, strand scission was induced at the methylated bases by heating at 90oC for 30 min in 1 M alkaline piperidine (Fisher, NJ). Samples were then suspended in formamide loading dye and electrophoresed as described above.
Competition binding experiments
An aliquot of 1 pmol end-labelled 25 bp substrate was mixed with 0, 1, 3 or 10 pmol non-labelled wild-type or mutant substrate generated by PCR amplification from the original M13 clones (see above) or with 0, 7, 21 or 70 ng pUC18 DNA and incubated with 1 pmol I-DmoI at 65oC for 5 min in 100 [mu]l standard cleavage buffer. The substrate and products were separated in 10% denaturing polyacrylamide gels and quantified using an Instant Imager (Packard, USA).
RESULTS
Activation of cleavage by divalent metal ions
An earlier study (13 ) indicated that although I-DmoI can bind to its DNA substrate in the absence of divalent metal ions, the presence of Mg2+ (or Mn2+) is required in order to activate the cleavage activity. We tested the influence of divalent metal ions more systematically employing a variety of divalent ions at a concentration of 1 mM. The results, which are summarized in Table 1 , demonstrate a broad metal ion specificity; strong cleavage was induced by Mn2+, Co2+ and Mg2+, weaker cleavage by Zn2+ and very weak cleavage by Ni2+. Since we had no insight into the preferred metal ion in vivo, Mg2+ was used in the subsequent studies.
. Comparison of ionic radii and the relative degree of DNA cleavage for different divalent metal ions
Metal ion
Relative cleavagea
Effective radius r (pm)b
Mg2+
82 (+-8)
72
Cu2+
n.d.
96
Ni2+
10 (+-5)
70
Zn2+
54 (+-10)
74
Ca2+
n.d.
100
Mn2+
100
67 (82)c
Co2+
93 (+-2)
73 (65)c
Ba2+
n.d.
136
None
n.d.
-
aRelative cleavage was calculated as follows: 2.5 pmol wild-type substrate (25mer) in 100 [mu]l was incubated with 1 pmol I-DmoI for 5 min at 65oC with the individual metal ions present at a concentration of 1 mM. The ion inducing the highest level of I-DmoI cleavage was arbitrarily set to 100. Non-detectable cleavage (n.d.) corresponding to I-DmoI added with no divalent metal ion was set to 0. The amount of product formed with other ions was normalized within this range. Experiments were performed in triplicate and the results averaged. bFor metal ions in octahedral coordination (14).
cTwo different radii are given for these metal ions (14).
Minimal substrate and cleavage of the coding strand
I-DmoI steady-state kinetics
The cleavage rate of the 25 bp substrate was measured at one enzyme concentration (1 pmol) and at four different substrate concentrations (Fig. 1 C). Cleavage reached completion except at the highest substrate concentration where there was a 6-fold molar excess of substrate. A half reciprocal plot of initial cleavage rates (v0) against initial substrate concentrations (S) (Fig. 1 D) yielded a kcat value of 0.5 min-1 anda Km value of 4 nM (+- 0.6), similar to a previous estimate of 4.8-5.6 obtained using a longer substrate and a different plotting procedure (13 ).
I-DmoI binds with two step kinetics
Some homing-type enzymes cleave their DNA substrates incompletely in vitro. This may reflect enzyme instability in the absence of substrate or, as was found for I-SceI, slow release of the DNA cleavage products (16 ). Therefore, the stability of I-DmoI was tested by incubating it alone in the cleavage buffer at 65oC. A strong decrease in cleavage activity was observed after incubating for 90 min with no residual activity after 4 h (Fig. 2 A). Given this relative instability, the cleavage activity of I-DmoI was investigated as a function of time at different enzyme concentrations.
The first experiment was performed using 50 pmol substrate (500 nM), ~125-fold higher than the Km for I-DmoI. Whereas 35 pmol I-DmoI produced 82-84% cleavage after 30 min, 3.5 pmol produced only 2% cleavage in the same time (Fig. 2 B). In the second experiment, 5 pmol substrate was cleaved ~90% by both 35 and 3.5 pmol I-DmoI after 30 min, while 0.35 pmol enzyme produced only 0.2% cleavage (Fig. 2 C).
These results are compatible with the following. (i) I-DmoI binds strongly to one of the products, as is found for I-SceI (16 ), and is unavailable for second cleavage round which would also explain why almost a 1:1 molar ratio of I-DmoI:substrate is required. (ii) Two copies of I-DmoI bind non-cooperatively per substrate molecule, since dimer binding at a 10-fold higher concentration of I-DmoI could enhance the binding probability by 100-fold. We favour the former explanation because there is no evidence for a symmetrical 2-fold symmetry within the DNA site (see Discussion).
Mutation-selection study of the DNA substrate
The nucleotide sequence specificity of the I-DmoI interaction was investigated by randomly mutating the 25 bp substrate and enriching for sequences that were deficient in cleavage by I-DmoI. An aliquot of 250 pmol of each substrate was derived from an oligonucleotide where the 25 nt (Fig. 1 A) were `doped' at a level of ~1.5% (see Material and Methods). I-DmoI (100 pmol) was incubated with this pool of substrates and the products were resolved on native gels. The band containing the uncleaved substrate was excised and the substrate was purified and cloned into M13mp19. Ninety-one clones were isolated from single colonies and sequenced. Of these, 22 exhibited the wild-type sequence while the remainder were mutated: 49 with single site mutations, 11 with double mutations and nine others contained deletions or insertions. The identities and frequency of occurrence of the single mutants are listed in Table 2 .
Given the relatively low level of selection in the first round (non-selection would theoretically give ~29 wild-type sequences), uncleaved substrates were cloned and sequenced after three rounds of selection (see Materials and Methods). After each selection and amplification round, 50-100 pmol substrate was digested with 70 pmol I-DmoI for 1 h at 65oC. Of 54 clones examined, most carried multiple mutations (three to eight) within the substrate sequence with at least one mutation within the central sequence adjacent to the intron homing site (see below). This indicated that many of the mutants selected for low, or non-cleavage, after one cycle were still cleavable in vitro, albeit at a much reduced rate.
This inference was tested directly by comparing the cleavage efficiencies of the 30 single-site mutants with that of the wild-type substrate. The results (Table 2 ) show that 19 were cleaved at >80% efficiency, four were cut in the 40-79% range, another four were cut in the 10-39% range; only three substrates A(-1)C, A(+1)T and G(+2)C showed very low, or no, cleavage (<10%). These data are superimposed on the DNA sequence in Figure 3 . They demonstrate that a limited sequence variation is tolerated within the DNA substrate, for both I-DmoI binding and cleavage, except around the site of cleavage and intron homing. Most mutations in this central region (-3 to +3) resulted in a strong reduction in cleavage, with transversions producing the larger effects; in contrast, relatively weak reductions in cleavage were observed for single mutations between positions -7 to -4 and -4 to +7. This also correlates with the observation (Table 2 ) that mutations producing the greatest reductions in cleavage were the most frequently isolated after selection.
. Summary of the positions of the single site mutations, their frequency and effect on I-DmoI cleavage and binding
Position
Mutation
Number of
Cleavage
Competitive
Binding/
clones
efficiency
binding
cleavage
-13
G -> T
2
100 (+-12)
nd
-12
C -> T
2
100 (+-15)
nd
-10
T -> A
1
95 (+- 4)
nd
-9
T -> A
1
100 (+- 8)
nd
-9
T -> G
1
100 (+-14)
nd
-7
C -> A
1
72 (+- 8)
63
b/c
-7
C -> T
2
45 (+-17)
58
b/c
-6
C -> G
2
90 (+-10)
90
b/c
-4
G -> C
1
85 (+-15)
91
b/c
-3
G -> T
1
44 (+-15)
59
b/c
-2
T -> A
4
12 (+- 8)
86
c
-2
T -> C
1
100 (+-20)
100
-2
T -> G
1
30 (+-15)
91
c
-1
A -> C
5
2 (+- 2)
86
c
-1
A -> T
3
34 (+- 6)
95
c
1
A -> G
1
87 (+-14)
nd
1
A -> T
3
9 (+- 6)
94
c
2
G -> C
1
6 (+- 5)
97
c
3
T -> A
1
87 (+-20)
nd
3
T -> G
4
13 (+- 4)
86
4
T -> A
1
100 (+-15)
nd
4
T -> G
1
87 (+-15)
nd
5
C -> A
1
93 (+- 4)
97
b/c
5
C -> G
2
70 (+-10)
82
b/c
5
C -> T
1
100 (+-15)
nd
6
C -> G
1
100 (+-10)
nd
6
C -> T
1
93 (+-14)
nd
7
G -> T
1
100 (+-12)
nd
9
C -> A
1
100 (+- 7)
nd
12
G -> T
1
100 (+- 8)
nd
Relative efficiencies were calculated as follows: the amount of cleavage product formed on incubating 2.5 pmol wild-type substrate in 100 [mu]l with 1 pmol I-DmoI for 5 min at 65oC corresponds to 100. No detectable cleavage corresponds to 0. The amount of product formed on incubating mutant substrates, under similar conditions, was normalized within this range. Each experiment was performed at least three times and the range of values are indicated. The relative binding capacity of mutant substrates was estimated from competition experiments in which inhibition of cleavage of the wild-type substrate was measured (Fig. 6) and was calculated similarly. Error limits were estimated within +-10%. The inhibition of cleavage seen with 10-fold molar excess of non-labelled over labelled wild-type substrate (~16 ng DNA) was set to 100 and the inhibition produced by 16 ng pUC18 DNA corresponds to 0 (Fig. 6). b: indicates that competitive binding of the DNA substrate for I-DmoI was decreased by the mutation. c: indicates that cleavage was reduced by the mutation (see also Fig. 5).
Footprinting of I-DmoI on wild-type and mutant substrates
To investigate potential contacts between I-DmoI and the DNA substrate, the complex formed in the absence of Mg2+ was probed with DNase I and DMS. DNase I has no strong nucleoside specificity and is useful for determining the physical limits of the binding site. I-DmoI complexed with the wild-type substrate was treated with DNase I and the products were analysed in denaturing polyacrylamide gels. Strong protection was observed between positions -12 to +12 on the non-coding strand and from -10 to +12 on the coding strand (Fig. 4 D, lane 1 +- I-DmoI) suggesting that most of the 25 bp fragment was required for binding.
Differentiating DNA binding and cleavage domains
The data are compatible with a model in which the central region of the DNA substrate is primarily involved in cleavage given (i) the limited I-DmoI protection effects observed in this region and (ii) the more dramatic effects that mutations in this region have on cleavage (Fig. 4 E). The data correlate further with the flanking sequences being primarily involved in I-DmoI binding (Table 2 ).
This model was tested further by examining I-DmoI binding and cleavage of substrates carrying single site deletions or insertions (Figs 4 C and D and 5 ). The footprinting data obtained by DMS probing indicated that little or no binding occurred to the substrates with deletions at positions -3, -1 or +1 and +3 or +4. This is consistent with the observation that the former was cleaved weakly (Fig. 4 C and D; lanes 4, 5 and 6) while the latter two were not cleaved (Fig. 5 ); similar results were found for a single base pair insertion between positions -6 and -3 (Fig. 5 ). Therefore, the results strongly suggest that changes in the spacing of this region prevent I-DmoI binding which could reflect either that changes in the distance between the two DNA attachment sites that flank this region are not tolerated or that the flexibility of the central region is diminished.
A minimal recognition-cleavage site
The above results imply that the minimal substrate for I-DmoI is 11-12 bp (Fig. 4 E). To test this, the two strands of an 11mer, extending from positions -6 to +5 on the coding strand, were synthesized and annealed. No cleavage was observed by I-DmoI, at 15 mM Mg2+, over a temperature range from 25 to 50oC. However, when the 11mer was cloned into the SmaI site of pUC18 and amplified using PCR, generating a substrate corresponding to a 15mer with 1 bp change (position +6) that does not impair cleavage (Fig. 3 ) and with ~50 bp flanking the 15 bp core on both sides, complete cleavage of 10 pmol substrate was observed with both 35 and 3.5 pmol I-DmoI after 20 min incubation at 65oC (Fig. 7 ).
DISCUSSION
A combined chemical and mutational analysis was performed in vitro to investigate I-DmoI-substrate interactions. The data obtained are compatible with the central region of the DNA substrate, that requires a conserved sequence and length, being primarily involved in cleavage and intron insertion, while the flanking regions contain the main DNA attachment sites. Although the I-DmoI substrate contains a palindromic sequence, that is a prerequisite for the formation of an essential helix in 23S rRNA (7 ), it is non-symmetrical with respect to the I-DmoI cleavage sites. Thus, it is unlikely that I-DmoI binds to its substrate as a dimer. This inference is supported both by the failure to detect I-DmoI dimers in solution (13 ) and by the results of a parallel protein footprinting study on I-DmoI where two separate DNA binding domains were identified, within the monomer protein, separated by ~100 amino acids (11 ).
The low turnover rate of I-DmoI (0.5 min-1) correlates with that observed earlier for another archaeal homing enzyme I-PorI (8 ) and may reflect that hyperthermophile genomes carry only one cleavage site (5 ) such that there is no need for rapid turnover. Moreover, I-DmoI may remain bound to one of the cleavage products (Fig. 1 A) as found for I-SceI and PI-SceI (16 ,17 ); this could explain the observed competitive binding of the upstream exon-intron fragment (Fig. 6 B).
All DNA endonucleases require divalent metal ions to cleave their DNA substrates although the binding specificity for the ions is generally low (12 ,18 ). I-DmoI was activated in the order Mn2+ > Co2+ > Mg2+ > Zn2+ > Ni2+ at 1 mM concentration whereas Cu2+, Ca2+ and Ba2+ did not support cleavage. From Table 1 it seems that an ionic radius of ~70 pm is needed for catalysis. A possible explanation is that the metal ion has to fit into a relatively small pocket that excludes larger metal ions. This is unlikely, however, since a parallel protein footprinting study with I-PorI and I-DmoI (12 ) showed that Ca2+ inhibited endonucleolytic cleavage of the DNA substrates. Another explanation may be that the larger metal ions can enter the catalytic pocket but are coordinated differently from the smaller metal ions and therefore do not support cleavage.
A comparison of the results from the mutation-cleavage study of the I-DmoI-substrate complex with those of other homing-type enzyme-substrate complexes did not reveal any general correlation between the location of the mutated site and the degree of cleavage inhibition, although both the PI-SceI enzyme and HO endonuclease appear to share a central region where mutations strongly inhibit cleavage (17 ,19 ). Apparently, different homing enzymes exhibit differing levels of sequence specificity. Whereas I-DmoI is less discriminating during binding but more so in the cleavage region, others, including I-PorI, discriminate strongly during binding where a single mutation in the flanking region prevents binding, and therefore cleavage, whereas the sequence within the central cleavage region is less critical (8 ). A third group that includes I-CeuI and I-SceII (6 ,7 ) appears to discriminate at relatively low levels for binding and cleavage while a fourth group, that includes PI-SceI and I-SceI, discriminates fairly strongly for both functions (17 ,20 ). Thus, the occurrence of at least two levels of discrimination, and possibly a third (see below), could explain why these homing enzymes cut so rarely.
The footprinting data (Fig. 4 ) are comparable with those reported earlier for the eukaryotic endonucleases I-PpoI and PI-SceI that cleave DNA by a similar mechanism to produce a 3'-OH 4 bp overhang, although only the latter carries the LAGLIDADG motifs (21 ,22 ). The former, I-PpoI, is also exceptional in that it binds as a dimer to an imperfect palindrome that is symmetrical with respect to the cleavage sites (23 ). I-DmoI and I-PpoI produce similar protection patterns when probed with DMS and they both produce the same hypersensitive site at A(+4) on the non-coding strand (23 ). PI-SceI-DNA complexes have not been probed with DMS but hydroxyl radical probing produced similar protection patterns on either side of, but not within, the cleavage region (24 ); these data were also reinforced by damage-selection experiments using methylation and ethylation reactions (17 ). The similarities correlate with the mutation cleavage results, described above, and suggest that, despite their different levels of sequence specificity, homing enzymes may have a common mechanism for interacting with the DNA double helix.
For I-DmoI, each of the two domains within the monomer protein (11 ) appear to interact with regions of the DNA substrate that border the cleavage region (Fig. 4 E). This suggests a working model for cleavage by homing enzymes where one domain binds first in the major groove where it recognizes a sequence flanking the cleavage site. PI-SceI has been shown to bind initially to the downstream recognition region (17 ) while I-PorI may have a preference for the upstream region (8 ); in contrast, I-DmoI and I-PpoI (23 ) seem to have no strong preference. Subsequently, the enzyme binds in the major groove containing the other flanking sequence, after inducing a distortion of the intervening DNA helix. The protein crosses the intervening minor groove of the DNA, where cleavage of the backbone takes place, to produce 3'-OH 4 bp overhangs (Fig. 4 E). The binding energy deriving from base- and backbone-specific interactions probably stabilizes an energetically unfavourable DNA conformer which places the phosphodiester bonds to be hydrolysed in the active site of the endonuclease. Bending of the cleavage domain of the DNA (Fig. 4 E) would probably involve some degree of unwinding of the double helix that increases the accessibility of the N-1 position of A(+4) on the non-coding strand of the I-DmoI- and I-PpoI-substrate complexes. In support of such a model, both I-PpoI and PI-SceI have been shown to bend their substrates (17 ,25 ). The bending and distortion would also provide a rationale for how the relatively small endonucleases (~20-25 kDa) recognize a DNA site corresponding to 1.5-2 double helical turns. Furthermore, the ease with which the DNA sequence adopts the appropriate conformation for cleavage may provide an additional level of discrimination for the homing enzyme-substrate interaction.
ACKNOWLEDGEMENTS
The research was supported by grants from the Danish Natural Science Research Council and the Danish Biotechnology II Program. J. Lykke-Andersen is thanked for critically reading the manuscript.
REFERENCES
1 Belfort,M., Reaban,M.E., Coetzee,T. and Dalgaard,J.Z. (1995) J. Bacteriol., 177, 3897-3903.MEDLINE Abstract
