Divalent metal ions play a crucial role in forming the catalytic centres of DNA endonucleases. Substitution of Mg2+ ions by Fe2+ ions in two archaeal intron‐encoded homing endonucleases, I‐DmoI and I‐PorI, yielded functional enzymes and enabled the generation of reactive hydroxyl radicals within the metal ion binding sites. Specific hydroxyl radical‐induced cleavage was observed within, and immediately after, two conserved LAGLIDADG motifs in both proteins and at sites at, and near, the scissile phosphates of the corresponding DNA substrates. Titration of Fe2+‐containing protein–DNA complexes with Ca2+ ions, which are unable to support endonucleolytic activity, was performed to distinguish between the individual metal ions in the complex. Mutations of single amino acids in this region impaired catalytic activity and caused the preferential loss of a subset of hydroxyl radical cleavages in both the protein and the DNA substrate, suggesting an active role in metal ion coordination for these amino acids. The data indicate that the endonucleases cleave their DNA substrates as monomeric enzymes, and contain a minimum of four divalent metal ions located at or near the catalytic centres of each endonuclease. The metal ions involved in cleaving the coding and the non‐coding strand are positioned immediately after the N‐ and C‐terminally located LAGLIDADG motifs, respectively. The dual protein/nucleic acid footprinting approach described here is generally applicable to other protein–nucleic acid complexes when the natural metal ion can be replaced by Fe2+.
Insight into the structure of the catalytic site of deoxyribonucleases is mainly based on crystallographic studies of, for example, DNase I, 3′‐5′ exonuclease of DNA polymerase I, nuclease P1 and the restriction endonucleases EcoRV and EcoRI, which show that divalent metal ions are coordinated to the cleavage site phosphates (Suck and Oefner, 1986; Beese and Steitz, 1991; Volbeda et al., 1991; Vipond et al., 1995). In the present study, we examine for the first time the coordination and functional significance of divalent metal ions in another class of endonucleases, the intron‐encoded homing enzymes. These enzymes are encoded by group I introns, archaeal introns or are expressed as inteins (reviewed in Lambowitz and Belfort, 1993), and they specifically cleave their intron−/intein− DNA alleles near the site of intron/intein insertion, as a first step in an intron/intein homing reaction, that was first described for the omega intron of Saccharomyces cerevisiae mitochondrial 23S‐like rDNA (Jacquier and Dujon, 1985). Many homing enzymes—including all inteins and those encoded by archaeal introns—contain two copies of an, often highly degenerate, LAGLIDADG motif. Four LAGLIDADG‐type proteins encoded by hyperthermophilic archaeal introns are known (Kjems and Garrett, 1985; Dalgaard and Garrett, 1992; Burggraf et al., 1993). Of these, the proteins encoded by the intron of Desulfurococcus mobilis 23S‐like rRNA, I‐DmoI and intron 1 of Pyrobaculum organotrophum 23S‐like rRNA, I‐PorI, are known to cleave DNA from intron− alleles site‐specifically in vitro (Dalgaard et al., 1993; Lykke‐Andersen et al., 1994). Moreover, the D.mobilis intron can move intercellularly and home in an intron− population of Sulfolobus acidocaldarius (Aagaard et al., 1995).
LAGLIDADG‐type homing enzymes are highly specific for their target sequences, recognizing up to 20 base pairs, although some sequence redundancy is allowed (Colleaux et al., 1988; Sargueil et al., 1990; Marshall and Lemieux, 1992; Wernette et al., 1992; Schapira et al., 1993; Lykke‐Andersen et al., 1994; Aagaard et al., 1997). With one exception, the recognition sequences are non‐palindromic and experimental evidence suggests that the endonucleases bind as monomers to their DNA substrates, primarily in the major groove (Lykke‐Andersen et al., 1996; Aagaard et al., 1997). DNA cleavage requires divalent cations and generates four nucleotide 3′‐staggered cuts with 5′‐phosphates. Insight into the functional domain structure of the homing endonucleases is limited. A biochemical study, based on protein footprinting, identified four regions which may be involved in DNA recognition (Lykke‐Andersen et al., 1996). Furthermore, studies with enzyme mutants have indicated that the conserved LAGLIDADG motifs are important for catalysis (Gimble and Stephens, 1995; Henke et al., 1995), and it has been speculated that they may be involved in the metal ion coordination (Gimble and Stephens, 1995; Turmel et al., 1995).
In order to address this question, we employed an Fe2+‐hydroxyl radical approach, previously used to map metal binding sites on the Escherichia coli glutamine synthetase (Farber and Levine, 1986), the Tet repressor (Ettner et al., 1995), pigeon liver malic enzyme (Wei et al., 1995) and the E.coli RNA polymerase (Zaychikov et al., 1996). The divalent metal ion binding sites on the two homing endonucleases, I‐DmoI and I‐PorI, were mapped in the absence and presence of their DNA substrates, using a dual Fe2+‐hydroxyl radical approach for both 32P‐end‐labelled proteins and DNA. The results were combined with those of competition experiments, using Ca2+ ions as competitor, and analyses of endonucleases mutated at potential metal ion coordinating amino acids, to yield insight into metal ion coordination within the catalytic sites of these enzymes.
Divalent cation requirement for endonucleolytic activity
The archaeal homing enzymes, I‐DmoI and I‐PorI, were expressed in E.coli from vectors pET–HTG–DmoI and pET–HTG–PorI (Lykke‐Andersen et al., 1996). They contain a glutathione S‐transferase (GST) tag at the C–terminus, and were purified in one step on glutathione–Sepharose, before removing the GST tag by cleavage with thrombin endoproteinase. Each protein can also be 32P‐end‐labelled at a heart muscle kinase site, positioned at the N‐terminus, using [γ‐32P]ATP and heart muscle kinase. After removal of the GST tag, the endonucleases contain a 5‐ to 10‐amino acid extension at each end that does not impair cleavage activity in vitro (Lykke‐Andersen et al., 1996).
The divalent cation requirements of I‐DmoI and I‐PorI, for cleavage of their respective DNA target sites, were investigated (Figure 1). Enzyme–substrate complexes were incubated in the presence of 1 mM of different divalent cations at the optimal cleavage temperatures of 65°C and 80°C for I‐DmoI and I‐PorI, respectively (Dalgaard et al., 1993; Lykke‐Andersen et al., 1994). I‐DmoI and I‐PorI were found to exhibit similar divalent cation dependence, following the relative order of descending reactivity at 1 mM: Mn2+, Fe2+>Mg2+≥Co2+, Ni2+>Zn2+. Addition of Cu2+, Ca2+ and Ba2+ produced no activity. Initial cleavage rate experiments were also performed for both endonucleases at Mg2+ and Ca2+ concentrations from 1 to 20 mM, and while Ca2+ did not support cleavage at any concentration tested, raising the Mg2+ concentration from 1 to 20 mM increased the cleavage rate ∼15‐fold (data not shown). Attempts to determine optimal Fe2+ concentrations failed due to degradation and aggregation of DNA and protein at high Fe2+ concentrations and at long incubation times (data not shown). Importantly, however, for this study, both endonucleases were highly active in the presence of Fe2+ ions at 1 mM.
Localizing metal binding sites in the endonucleases
In the presence of hydrogen peroxide (H2O2) and ascorbate, Fe2+ ions generate hydroxyl radicals that cleave the backbones of polypeptides and polynucleotides. The capacity of Fe2+ to function as a metal ion cofactor in the cleavage reactions enabled us to induce reactive hydroxyl radicals at the divalent cation‐binding sites of I‐DmoI and I‐PorI. By mapping the resulting cleavage of the polypeptide chain and the DNA substrate, it was possible to localize the metal ions in the protein–DNA complex. Hydroxyl radical cleavage sites in the protein moieties were mapped using a recently developed protein footprinting approach that involves site‐specific 32P labelling at either the N‐ or C‐terminus of the proteins (Jensen et al., 1995a). A protein sequence ladder, run alongside the hydroxyl radical cleavage reactions, was generated by partial cleavage of the end‐labelled proteins with sequence‐specific proteinases, allowing rapid identification of the positions of the metal ion cleavage sites, with an accuracy of about two amino acids. N‐terminally labelled I‐DmoI and I‐PorI were incubated with Fe2+/H2O2/ascorbate, in the absence or presence of their respective DNA substrates, and degradation products were separated on high‐resolution SDS–polyacrylamide gels (Figure 2A and C). As expected, the presence of free radicals in the solution gave rise to a weak background of unspecific cleavages for large parts of the proteins (Figure 2A and C, lane 2). Four major hydroxyl radical cleavage sites were always observed for both I‐DmoI and I‐PorI and, in contrast to the other cleavages, these sites were all strongly enhanced in the presence of the DNA substrate (Figure 2A and C, compare lanes 2 and 3). Experiments using C‐terminally 32P‐labelled I‐DmoI (Figure 2B) and I‐PorI (data not shown) yielded the same hydroxyl radical cleavage pattern as N‐terminally labelled endonucleases, confirming that the four major hydroxyl radical cleavages are primary cleavage events, independent of cleavage elsewhere in the protein. From the migration of the cleavage products relative to the sequence marker bands, they were localized at or near L14, V39, L113 and K130 in I‐DmoI and at or near K8, V28, L86 and V109 in I‐PorI (Figure 2). Cleavage at the site closest to the C‐terminus of I–DmoI (K130) invariably produced a smeared band using N–terminally labelled I‐DmoI (Figure 2A, lane 3) and, therefore, the location of this site was determined using C‐terminally labelled I‐DmoI which produced a sharp band (Figure 2B). The cleavages induced by hydroxyl radicals generally produced doublet bands for both endonucleases, probably due to the formation of products with plus or minus one amino acid (Platis et al., 1993). The hydroxyl radical cleavages could be inhibited by adding EDTA to chelate the Fe2+ ions (Figure 2A–C, lane 4).
Locating metal sites on the DNA substrates
The observation that the presence of the DNA substrate specifically enhanced the four main hydroxyl radical cleavages for each endonuclease, suggests that the DNA actively contributes to the coordination of the metal ions in the protein. To map the metal ion sites on the DNA substrates, 25 bp DNA fragments containing the cleavage sequences for I‐DmoI and I‐PorI, were 5′ end‐labelled on the coding, or the non‐coding strand (defined as the coding and non‐coding strand in the natural substrate rRNA operon), complexed with the corresponding homing endonuclease, and incubated in the absence of divalent metal ion, or with Mg2+, Fe2+ or Fe2+/H2O2, before subjecting to denaturing polyacrylamide gel electrophoresis (Figure 3). Both I‐DmoI and I‐PorI exhibited normal endonucleolytic activity in the presence of Mg2+ and Fe2+, producing products of the expected size (marked with an asterisk in Figure 3). Upon addition of Fe2+ and H2O2, hydroxyl radical cleavages appeared at two major sites on each strand (Figure 3). Hydroxyl radical cleavage generates 3′‐phosphates, and these products are easily discriminated from the endonucleolytic cleavage which generates 5′‐phosphates (Figure 3). The I‐DmoI–DNA substrate was cleaved by hydroxyl radicals at the scissile phosphate on both the coding and non‐coding strands (denoted as cd and nc, respectively, in Figure 3). In addition, strong cleavage was observed on both strands at phosphates two nucleotides further upstream from the scissile phosphates [cd(−2) and nc(−2), respectively, Figure 3A]. Weak, but reproducible cleavage was also observed at cd(−1) and cd(+1) on the coding strand. The I‐PorI substrate was also cleaved at the scissile phosphates (cd and nc), in addition to cleavage at phosphates positioned one nucleotide downstream and upstream on the coding and non‐coding strands [cd(+1) and nc(−1), respectively, Figure 3B]. Hydroxyl radical cleavage of the DNA was strictly dependent on the presence of the corresponding endonuclease (Figure 3), implying that the enzyme positioned the metal ions at the cleavage sites. In some instances the radical cleavages were induced by Fe2+ in the absence of H2O2, possibly due to O2 dissolved in the water substituting for H2O2 (see Figure 3A).
Metal ion competition experiments
In order to investigate the correlation between hydroxyl radical cleavage of the endonucleases and their DNA substrates more closely, Fe2+ was gradually replaced by titration with Ca2+, which neither supports endonucleolytic activity nor can generate hydroxyl radicals. Ca2+ was added in concentrations ranging from 1 to 4 mM, and the effects on both endonucleolytic cleavages and hydroxyl radical cleavage at the four main sites in the protein and the DNA, were examined (Figure 4). In order to complex all labelled molecules, hydroxyl radical experiments on protein and DNA were done with a large excess of unlabelled protein or DNA ligand over the labelled DNA or protein partner (∼10‐ to 25‐fold the Km value of the cleavage reaction). This ensures the presence of a stoichiometric protein–DNA complex in the assays. Furthermore, to avoid secondary effects, all assays were stopped early in the reaction, when cleavage rates at the individual sites were still constant. The results were quantified in an Instant Imager and semi‐quantitative values averaged from at least three independent assays are summarized (see Figure 6, below).
These experiments showed that Ca2+ interfered with endonucleolytic cleavage and with Fe2+ binding to both DNA and protein, for both endonucleases (Figure 4; see also Figure 6). However, individual sites were affected to different degrees, allowing a distinction to be made between individual metal ions. Thus, for both I‐DmoI and I‐PorI, endonucleolytic cleavage of the non‐coding strand was more affected by the presence of Ca2+‐ions than cleavage of the coding strand (Figure 4A). Moreover, assaying the I‐DmoI protein, two hydroxyl radical cleavages on the protein (V39 and L113) and the DNA [nc and cd(−2)] were strongly reduced by Ca2+ ions, whereas the remaining two sites were less affected (Figure 4B left and middle panels, and C, left panels). I‐DmoI was also tested at Ca2+ concentrations up to 20 mM, resulting in the gradual loss of all radical cleavage sites. This, however, gave no additional distinction between individual metal ions (data not shown). For I‐PorI, the presence of Ca2+ strongly reduced the hydroxyl radical cleavage at three sites on both protein (K8, V28 and L86) and DNA [nc, nc(−1) and cd(+1)], whereas a single site in the protein (V109) and the DNA (cd) was unaffected (Figure 4B and C, right panels).
Mutagenizing the LAGLIDADG motifs of I‐DmoI and I‐PorI
Hydroxyl radical cleavage of I‐DmoI and I‐PorI occurred within, and 10–15 amino acids after, each of the conserved LAGLIDADG motifs (Figure 2). This motif was shown earlier, by mutagenesis experiments on other LAGLIDADG‐like endonucleases, to be important for catalysis but not for DNA binding (Gimble and Stephens, 1995; Henke et al., 1995). Furthermore, all known LAGLIDADG endonucleases contain an acidic residue at the penultimate position of their LAGLIDADG motif (D21 and E117 in I‐DmoI and D14 and E93 in I‐PorI). Therefore, we reasoned that these amino acids are the most likely candidates for participating in the metal ion binding. To test this experimentally, these residues were changed to the corresponding amides, that are poor metal ion coordinators, in the first or the second LAGLIDADG motif of both I‐DmoI (D21N and E117Q) and I‐PorI (D14N and E93Q). The mutants were assayed for endonucleolytic cleavage activity (Figure 5A) and metal ion coordination to both protein (Figure 5B) and the DNA substrate (Figure 5C), using the Fe2+‐hydroxyl radical approach described above. The results averaged from at least three independent experiments are shown in Figure 6.
All mutations severely affected the endonucleolytic cleavage activity of both strands, although some residual activity was detected with the I‐PorI D14N mutant (Figure 5A and 6). In agreement with reports for other homing enzymes (Gimble and Stephens, 1995; Henke et al., 1995), the lack of cleavage was not a result of disabled DNA binding, since all four mutant proteins were found to bind specifically to the cognate DNA substrate using a proteinase footprinting approach described earlier (unpublished observation; Lykke‐Andersen et al., 1996). The mutants were all assayed for their ability to induce Fe2+‐hydroxyl radical cleavage of both the protein and the DNA, and the results were compared with those of the unmutated proteins (Figures 5B and C and 6). The I‐DmoI D21N mutant exhibited strongly reduced hydroxyl radical cleavage at two sites in the protein (V39 and K130) and the two scissile phosphates of the DNA substrate (cd and nc), whereas the other cuts were less affected. The I‐DmoI E117Q mutant showed strong reduction of hydroxyl radical cleavage only at K130 on the protein and at the coding strand scissile phosphate (cd). In I‐PorI, the D14N mutation led to a reduced hydroxyl radical cleavage at the non‐coding strand scissile phosphate (nc) with no significant reductions at the protein, while the E93Q mutant showed highly reduced cleavage at two sites in the protein (V109 and V28) and at the two scissile phosphates of the DNA (cd and nc). A double mutant of I‐PorI (D14N/E93Q) was also tested; this was catalytically inactive and showed hydroxyl radical cleavage that was additive relative to the two single mutants (data not shown).
DNA endonucleases require divalent metal ions to cleave DNA
Structural, biochemical and theoretical studies suggest that metal ions in DNases play a direct role in catalysis, although the exact function of the metal ions remains to be determined. The ability of the metal ion to act as a Lewis base may be important. Thus, a water molecule coordinated to the metal ion is deprotonated and the resulting hydroxyl ion can then perform a nucleophilic attack. A second potential function of the metal ions may be to stabilize the oxoanion leaving group of the cleavage reaction (Steitz and Steitz, 1993 and references therein). We have investigated two members of the LAGLIDADG‐type homing endonuclease family, which resemble restriction endonucleases in several ways. They cleave both strands of the DNA within a specific nucleotide sequence yielding 5′‐phosphates. However, in contrast to type II restriction endonucleases which recognize 4–8 base pair palindromic sequences, the homing endonucleases recognize more extended sequences of up to 20 base pairs which are generally non‐palindromic, a property they share with type I restriction endonucleases. Another feature shared with restriction enzymes is the low degree of specificity for divalent metal ions. I‐DmoI and I‐PorI were active in the presence of Mn2+, Fe2+, Mg2+, Co2+, Ni2+ and Zn2+, whereas Cu2+, Ca2+ and Ba2+ ions were unable to support catalysis. A similar pattern was observed for EcoRI and EcoRV (Vipond et al., 1995). That Ca2+ is unable to function as a metal ion cofactor in the cleavage is somewhat surprising, since Mg2+ and Ca2+ are chemically very similar, although Ca2+ may fail to enter the catalytic site due to its larger ionic diameter. This is unlikely, however, since Ca2+ inhibited both endonucleolytic cleavage of DNA by I‐DmoI and I‐PorI and also Fe2+‐induced hydroxyl radical cleavages at protein and DNA substrate. Another possibility is that Ca2+ ions disturb the overall structure of the enzyme, although this is also rendered unlikely by the observation that Ca2+ only efficiently inhibited enzymatic cleavage of one strand, for both endonucleases. Furthermore, the simple change in ionic strength does not lead to structural changes, since increasing the concentration of monovalent cations (K+) above the ionic strength of the highest Ca2+ concentration inhibited neither endonucleolytic cleavage nor Fe2+‐induced hydroxyl radical cleavage. Moreover, increasing Mg2+ concentrations increased the endonucleolytic cleavage (data not shown). A third explanation is that Ca2+ binds to the active site of I‐DmoI and I‐PorI but fails to support the catalytic reaction. All the data presented here are consistent with this model; in particular, the observation that Ca2+ inhibits a Fe2+‐catalysed cleavage reaction suggests that it competes directly for Fe2+ binding to the proteins. An attempt was made to establish that Ca2+ binds directly at the catalytic centre of each endonuclease, by performing competition experiments with Ca2+ and Mg2+. However, since no Michaelis–Menten kinetics were observed with respect to Mg2+ (data not shown), a clear distinction between competitive and non‐competitive binding could not be made. Interestingly, Ca2+ seems to be an acceptable cofactor for the non‐LAGLIDADG nuclear homing endonuclease I‐PpoI, indicating a different geometry of the catalytic site in that protein (Wittmayer and Raines, 1996). However, which metal ion is used in vivo remains an unanswered question; possibly the enzyme uses whichever suitable divalent metal ion is available.
Mapping metal binding sites
To map the divalent metal ion binding sites of the two archaeal homing endonucleases, I‐DmoI and I‐PorI, we employed Fenton chemistry, involving the generation of reactive intermediates from H2O2 in the vicinity of Fe2+ (for review, see Price and Tullins, 1992). This method has previously been applied successfully to map divalent metal binding sites of E.coli glutamine synthetase (Farber and Levine, 1986), the Tet repressor (Ettner et al., 1995), pigeon liver malic enzyme (Wei et al., 1995) and E.coli RNA polymerase (Zaychikov et al., 1996). In the Tet repressor study, a close correlation between cleavage efficiency and the distance of the cleaved bonds to Mg2+ in the crystal structure of the Tet repressor was found, implying that the reactive intermediates are very short‐lived compared with the speed of diffusion, and that the cleavage reaction is a valid measurement for the proximity of the Fe2+ ion (Ettner et al., 1995). The mechanism of Fe2+/H2O2‐induced cleavage has been discussed (Rana and Meares, 1991; Price and Tullins, 1992; Platis et al., 1993; Ettner et al., 1995). It is generally assumed that cleavage mainly is a result of hydroxyl radical attack. However, some evidence also points towards a non‐radical mechanism involving a nuclear attack of an activated oxo intermediate (Rana and Meares, 1991). In our investigations, cleavage of both the DNA and the protein were strictly Fe2+‐dependent but, in some experiments, H2O2 was dispensable (e.g. Figure 2A). This could be explained by a non‐radical cleavage mechanism in which nearby hydroxyl groups, coordinated by the ferrous ion, are involved in a nucleophilic attack or, perhaps more likely, by the presence of soluble O2‐forming radicals (Price and Tullins, 1992).
Both the homing enzymes and their DNA substrates were specifically cleaved at four major positions in the presence of Fe2+/H2O2 (Figures 2 and 3). The DNA cleavages were dependent on the presence of the homing enzymes and the protein cleavages were strongly enhanced by the presence of the DNA substrates. This suggests that the metal ions are coordinated primarily to the protein, but stabilized by a phosphodiester group of the DNA substrate as has been shown for the EcoRV–DNA complex (Kostrewa and Winkler, 1995). Alternatively, the DNA may induce perturbations in the protein structure, making the adjacent amino acids more susceptible to cleavage by free Fe2+ ions. Several observations are compatible with the Fe2+ ion(s) contributing to the function of the catalytic site. Firstly, the endonucleases were highly active in the presence of Fe2+ ions. Secondly, the Fe2+/H2O2‐dependent cleavages occurred at or near the scissile phosphates, and within and immediately after the LAGLIDADG regions of the enzymes, that constitute a conserved and functionally important region of these homing endonucleases (Gimble and Stephens, 1995; Henke et al., 1995). Thirdly, I‐DmoI and I‐PorI both bind specifically to the DNA substrates in the absence of divalent metal ions (Lykke‐Andersen et al., 1996). Finally, a strong correlation exists between enzymatic activity and Fe2+ binding under a variety of experimental conditions (see below).
Locating individual metal ions in the endonuclease–DNA complex
To localize individual Fe2+ ions within the protein and DNA structures, a combination of Ca2+ competition and mutagenesis experiments was used. The former exploit the observation that Ca2+ ions cannot induce endonucleolytic cleavage of the DNA. Titration of Ca2+ ions into a reaction containing a fixed concentration of Fe2+ produced unequal competition for enzymatic cleavage of the two DNA strands. For both I‐DmoI and I‐PorI, enzymatic cleavage of the coding strands was least sensitive to addition of Ca2+, implying that more than one metal ion is responsible for catalytic cleavage and that the Fe2+ responsible for cleaving the coding strand is more tightly bound than the non‐coding strand‐cleaving Fe2+. This resembles the observation for the homing endonuclease, I‐CpaII, encoded by the chloroplast rRNA intron of Chlamydomonas pallidostigmatica, for which cleavage of the non‐coding strand was impaired more at low Mg2+ than cleavage of the coding strand (Turmel et al., 1995). In contrast, DNA cleavage by restriction endonucleases EcoRV and SalI shows no strand specificity at low Mg2+ concentration, reflecting that each subunit in the homodimeric complex has similar affinities for metal ions (Halford and Goodall, 1988).
When combining the results from the Ca2+ competition and mutagenesis experiments, the differential behaviour of individual Fe2+ ions indicates that the I‐DmoI–DNA substrate complex contains a minimum of four metal ions, with the metal ions near L14, V39, L113 and K130 of the protein positioned at the nc(−2), nc, cd(−2) and cd phosphates of the DNA substrate, respectively (Figures 6 and 7). Thus, the V39/nc and L113/cd(−2) metal sites are more sensitive to Ca2+ than the other two sites, the K130/cd site is sensitive to mutation at D21 and the V39/nc site is sensitive to mutations at both D21 and E117 (Figures 6 and 7). I‐PorI may also contain four metal ions. Here, the metal ions near V28 and V109 are probably positioned at the nc and cd phosphates, respectively, whereas the metal ions at K8 and L86 of the protein and nc(−1) and cd(+1) of the DNA cannot be discriminated by our data (Figure 6).
Taken together, the data are consistent with a model where each endonuclease cleaves its DNA substrate as a monomeric enzyme with four metal ions positioned at, or near, the catalytic centre. The metal ions coordinated to the scissile phosphates of the non‐coding and the coding strand map 10–15 amino acids C‐terminal to the first and second LAGLIDADG motifs of the proteins, respectively, and the identity of these metal ions is decisive for endonucleolytic cleavage at the corresponding phosphate, as seen from the Ca2+ titrations. Moreover, judging from the mutational data, the conserved penultimate acidic residues of the LAGLIDADG motifs are probably involved in coordinating the metal ions at the scissile phosphates. Two other metal ions, which are placed 1–2 nucleotides from the cleavage site, are not coordinated by these amino acids but are positioned near the N‐terminus of the LAGLIDADG motifs. Although we cannot conclude that all the metal ions have a specific role in the catalytic reaction, it is an attractive hypothesis that the monomeric endonuclease coordinates two metal ions at each of the two active sites, where one metal ion activates the attacking hydroxyl group, while the other ion stabilizes the oxyanion leaving group as shown for a number of other DNases, and proposed for catalytic RNAs (Steitz and Steitz, 1993).
Materials and methods
Vectors and substrates
Vectors for expression of GST–endonuclease fusion proteins in E.coli pET–HTG–DmoI, pGEX–GTH–DmoI, pET–HTG–PorI and pGEX–GTH–PorI were described earlier (Lykke‐Andersen et al., 1996). The plasmids pUC–D.muc and pUC–P.isl, used in the cleavage assays, are pUC19 derivatives containing 245 and 205 bp inserts with cleavage sites for I‐DmoI and I‐PorI, respectively (Lykke‐Andersen et al., 1996). 25 bp DNA substrates were prepared by annealing the oligodeoxynucleotides 5′‐GCCTTGCCGGGTAAGTTCCGG‐CGCG and 5′‐GCGAGCCCGTAAGGGTGTGTACGGG with their complementary sequences, for I–DmoI and I‐PorI respectively (Lykke‐Andersen et al., 1996) and a 30 bp substrate for I‐PorI was prepared by annealing the oligodeoxynucleotide 5′‐CGCGAGCCCGTAAGGGTGTGTACGGGGGCT with its complementary sequence.
Plasmids pUC–P.isl and pUC–D.muc were cleaved with PvuII, dephosphorylated with calf intestinal alkaline phosphatase (Boehringer), and 5′‐end‐labelled using T4 polynucleotide kinase (Amersham) and [γ–32P]ATP (Amersham). About 10 fmol I‐DmoI and I‐PorI were mixed with ∼50 c.p.s. 32P‐labelled substrate (∼50 fmol) in 10 μl binding buffer (25 mM HEPES–KOH, pH 8.0, 50 mM KCl, 5 mM dithiothreitol, 2% glycerol, 0.05% Triton X‐100), and the mixture was overlaid with mineral oil. A divalent cation chloride was added at 1 mM, and reaction mixtures were incubated at 65°C (I‐DmoI) or 80°C (I‐PorI) for 10 min. DNA was precipitated with ethanol, redissolved in formamide, denatured at 95°C for 2 min and run on a 5% polyacrylamide–7 M urea gel.
Determination of metal ion binding sites on proteins and DNA
In order to determine metal binding sites on I‐DmoI and I‐PorI, N‐ and C‐terminally 32P‐labelled endonucleases were generated from vectors pET–HTG–DmoI, pGEX–GTH–DmoI, pET–HTG–PorI and pGEX–GTH–PorI as described earlier (Lykke‐Andersen et al., 1996), and ∼200 c.p.s. (∼1 pmol) labelled endonuclease was incubated with 20 pmol 25 bp DNA substrate in 20 μl binding buffer at 65°C for 5 min. Samples were cooled to 50°C and 2 μl of 20 mM Fe(NH4)2SO4, 1% H2O2 and 100 mM sodium ascorbate was added. In Ca2+‐competition experiments, CaCl2 was added at different concentrations together with the Fe(NH4)2SO4. After 5 min of incubation, the reaction was stopped by adding 10 μl 100 mM Tris base, 15% glycerol, 1.5% sodium dodecylsulphate, 200 mM dithiothreitol, 20 mM EDTA and 100 mM thiourea and the samples were run on a large 7% stacking/20% separation polyacrylamide–SDS–Tricin gel, which was then subjected to autoradiography. Gels of Ca2+‐competition experiments were quantified on an Instant Imager (Packard, Meriden, USA).
To determine the metal coordination sites on the DNA substrates, 25–mer oligodeoxynucleotides encompassing the cleavage sites for I–DmoI and I‐PorI were 5′ end‐labelled using T4 polynucleotide kinase (Amersham) and [γ‐32P]ATP (Amersham), and complexed with the complementary unlabelled 25‐mer by mixing at 0.1 μM in 10 mM HEPES–KOH, pH 8.0, 100 mM KCl, heating at 95°C for 2 min, followed by incubation at 65°C for 5 min and slowly cooling to room temperature. In some experiments, 30‐mer oligonucleotides were used for I‐PorI. Approximately 100 c.p.s. substrate (∼0.1 pmol) was mixed with ∼1 pmol purified I‐DmoI and I‐PorI, expressed from pET–HTG derivatives (Lykke‐Andersen et al., 1996), in 10 μl binding buffer at 65°C for 5 min. Samples were cooled to 50°C and 1 μl of a mixture of 20 mM Fe(NH4)2SO4, 1% H2O2, 100 mM ascorbate was added. In Ca2+‐competition experiments, CaCl2 was added in different concentrations together with the Fe(NH4)2SO4. After 5 min incubation the reaction was stopped by addition of 40 μl of 0.3 M Na acetate, pH 6.0, 5 mM EDTA, 100 mM thiourea, 0.1 mg/ml tRNA and 125 μl ethanol. DNA was precipitated, washed in 80% ethanol, redissolved in formamide, denatured at 95°C for 2 min and run on 20% polyacrylamide–7 M urea gels which were then subjected to autoradiography. Gels of Ca2+‐competition and endonuclease mutant experiments were also quantified on an Instant Imager (Packard, Meriden, USA).
Mutagenesis of I‐DmoI and I‐PorI
The open reading frames of I‐DmoI and I‐PorI were excised from pET–HTG–DmoI and pET–HTG–PorI vectors (Lykke‐Andersen et al., 1996) using BamHI and EcoRI, and ligated with BamHI–EcoRI‐cleaved M13mp19. Site‐directed mutagenesis was performed according to the method of Kunkel et al. (1987). For I‐DmoI, oligonucleotides 5′‐GTAAAGTCCTCCATTACCTATTATCAATCC and 5′‐GGTTTTATCTCCTTGAGCTACAT‐ATAGCCC (position of mutation underlined) were used to mutagenize the first and second LAGLIDADG motif, respectively. For I‐PorI an internal EcoRI site in the open reading frame was first destroyed by a silent mutation, changing one glutamate codon (GAA) into another (GAG), using the oligodeoxynucleotide 5′‐CTCGAGGAACTCCTTAGA. The resulting EcoRI‐minus clone was then used to mutagenize the LAGLIDADG motifs, using oligodeoxynucleotides 5′‐AACATAGCCATTA‐AAAGCGCT for the first and 5′‐CACATTCCCTTGGGCATCTATT for the second motif. The mutated open reading frames were cut out using BamHI and EcoRI and inserted into pET‐HTG and pGEX‐GTH vectors (Jensen et al., 1995b) and the proteins were expressed and purified as described earlier for the wild‐type protein (Lykke‐Andersen et al., 1996).
We thank Torben Heick Jensen for technical advice and critical reading of the manuscript and Kaj Frank Jensen for advice on cleavage kinetics. This work was supported by the Karen Elise Jensen Foundation and the Biotechnology program of the Danish Research Councils. J.L. was supported by Copenhagen University.
- Copyright © 1997 European Molecular Biology Organization