Photocross-linking of the Homing Endonuclease PI-SceI to Its Recognition Sequence*

PI-SceI is an intein-encoded protein that belongs to the LAGLIDADG family of homing endonucleases. According to the crystal structure and mutational studies, this endonuclease consists of two domains, one responsible for protein splicing, the other for DNA cleavage, and both presumably for DNA binding. To define the DNA binding site of PI-SceI, photocross-linking was used to identify amino acid residues in contact with DNA. Sixty-three double-stranded oligodeoxynucleotides comprising the minimal recognition sequence and containing single 5-iodopyrimidine substitutions in almost all positions of the recognition sequence were synthesized and irradiated in the presence of PI-SceI with a helium/cadmium laser (325 nm). The best cross-linking yield (approximately 30%) was obtained with an oligodeoxynucleotide with a 5-iododeoxyuridine at position +9 in the bottom strand. The subsequent analysis showed that cross-linking had occurred with amino acid His-333, 6 amino acids after the second LAGLIDADG motif. With the H333A variant of PI-SceI or in the presence of excess unmodified oligodeoxynucleotide, no cross-linking was observed, indicating the specificity of the cross-linking reaction. Chemical modification of His residues in PI-SceI by diethylpyrocarbonate leads to a substantial reduction in the binding and cleavage activity of PI-SceI. This inactivation can be suppressed by substrate binding. This result further supports the finding that at least one His residue is in close contact to the DNA. Based on these and published results, conclusions are drawn regarding the DNA binding site of PI-SceI.

in prokaryotes and eukaryotes as well as in archaebacteria and are encoded by introns or inteins (for review, see Ref. 3). The largest group is characterized by the presence of one or two copies of a conserved dodecapeptide sequence, the LAGLI-DADG motif (4). PI-SceI, a homing endonuclease from yeast, belongs to this group and occurs as an intein within the vacuolar H ϩ -ATPase, from which it is spliced in an autocatalytic reaction (5).
The mature protein recognizes an extraordinarily long DNA sequence of 35-45 bp, 1 bends the DNA, and cleaves the substrate to produce a 4-bp 3Ј overhang (5,6,7). The molecular details of DNA recognition and cleavage by PI-SceI are unclear, in spite of the fact that the crystal structure of PI-SceI is known (8). According to the structure analysis and mutational studies (9), PI-SceI is composed of two domains, one responsible for protein splicing (domain I) and one for DNA cleavage (domain II), which are connected by two peptide segments. The structure of the elongated domain I consists almost entirely of ␤-sheets, whereas the compact domain II is an almost equal mixture of ␣-helices and ␤-strands. Domain II is built up from two substructures that are related by local 2-fold symmetry about an axis between the two LAGLIDADG sequences. The domain II of PI-SceI is structurally very similar to the homodimeric homing endonuclease I-CreI, which contains one LAGLIDADG motif per subunit and lacks the protein splicing domain. As shown by the crystal structure analysis (10), the LAGLIDADG motifs in I-CreI form part of the dimer interface while simultaneously positioning one of the conserved Asp residues adjacent to the scissile phosphates. These residues may function to coordinate Mg 2ϩ and thereby help to attack the DNA; substitution of these residues abolishes the endonuclease activity of I-CreI (11). Mutation of the analogous residues Asp-218 and Asp-326 in PI-SceI also destroys the nucleolytic activity of this enzyme (12), whereas substrate binding of the mutated PI-SceI is not affected, suggesting that these Asp residues are involved in catalysis. Similar results were obtained with I-SceII (13), I-DmoI (14), I-CeuI (15), I-PorI (14), and PI-TliI (16), all members of the LAGLIDADG family of homing endonucleases.
All homing endonucleases have long recognition sequences (15-45 bp) that can tolerate variation of the sequence, as shown for PI-SceI (6) and other homing endonucleases, e.g. I-CreI (17), I-DmoI (18), I-PorI (19), I-PpoI (17), and I-TevI (20). Footprinting studies with PI-SceI (6), I-DmoI (18), and I-TevII (21) and their substrates show that they are involved in both major and minor groove interactions. After cleavage, PI-SceI (6,7), I-SceI (22), F-SceII (23), I-TevI (24), and I-TevII (21) remain bound to one of the two cleavage products that may be required for the subsequent recombination event which completes the homing reaction. The genetically engineered domain DI of PI-SceI binds specifically and with similar affinity as full-length PI-SceI to DNA containing the PI-SceI recognition site as well as to one of the two cleavage products (9), demonstrating that domain I is not only involved in protein splicing but also in DNA binding. In contrast, the genetically engineered domain II of PI-SceI is not able to bind DNA with strong affinity, suggesting that domain I is responsible for a decisive part of the contacts between PI-SceI and its substrate (9). A similar twodomain structure with a catalytic domain and a DNA binding domain has been proposed for I-TevI (25).
The long recognition site of PI-SceI makes it an attractive model for studying the mechanism of DNA sequence recognition by proteins but makes it difficult to model the DNA into the structure of PI-SceI, in particular as it is known that both domains, which may be linked in a flexible manner, are involved in DNA binding (9,26,27). In the study presented here, we tried to identify residues of PI-SceI in close contact to the recognition site by a photocross-linking technique using 5-iodine-substituted pyrimidines (5-IdU and 5-IdC). We show that PI-SceI can be cross-linked via His-333 to a 5-Iododeoxyuridine (5-IdU) residue located in position ϩ9 of the bottom strand, i.e. the right half of the PI-SceI recognition sequence. Substitution of the cross-linked amino acid His-333 by Ala results in a PI-SceI mutant that is not significantly impaired in its ability to bind and cleave DNA. However, no photocross-linking could be observed with the H333A mutant, demonstrating that the region around His-333 is in close contact with the DNA. With this information and knowing where the active site is located, it is possible to present a model that describes the approximate location of the DNA binding site in the structure of PI-SceI.

MATERIALS AND METHODS
Purification of Recombinant His 6 -tagged PI-SceI-His 6 -tagged PI-SceI was expressed in Escherichia coli and purified as described by Wende et al. (7).
Synthesis of Iododeoxyuridine-and Iododeoxycytidine-substituted Oligodeoxynucleotides-Oligodeoxynucleotides were chemically synthesized by automated ␤-cyanoethylphosphoramidite DNA synthesis using 5-IdU-␤-cyanoethylphosphoramidites (Glen Research) on a Cyclone plus DNA synthesizer (Millipore) or obtained by INTERACTIVA. To reduce possible deiodination of the 5-IdU and 5-IdC, the final deprotection step was carried out at ambient temperature for 24 h, as suggested by the manufacturer. We used either the minimal full-length recognition sequence (cf. Fig. 1) or the right-half cleavage product comprising the upper strand 5Ј-GGAGAAAGAGGTAATGAAATGGCAGAAGTC-T-3Ј (31-mer) and the lower strand 5Ј-GATCAGACTTCTGCCATTTCA-TTACCTCTTTCTCCGCAC-3Ј (39-mer). Some oligodeoxynucleotides were labeled at their 5Ј-terminus using T4 polynucleotide kinase and [␥-32 P]ATP.
Photocross-linking of ds Oligodeoxynucleotides Carrying a Single Iodopyrimidine Substitution and PI-SceI-For analytical scale photocross-linking, approximately 10 M PI-SceI was preincubated with 10 M radioactively labeled ds oligodeoxynucleotide monosubstituted with 5-IdU or 5-IdC at various positions of the top and bottom strand of either the full-length recognition site or the right-half cleavage product (cf. Fig. 1) in buffer P (10 mM Tris/HCl, pH 8.5, 100 mM KCl, 2.5 mM EDTA) for 30 min at ambient temperature in a volume of 50 l. Photocross-linking was carried out with a 40 milliwatt helium/cadmium laser emitting at 325 nm (Laser 2000). The total irradiation time was usually 2 h; in kinetic experiments, 0 -3 h. Samples of 2.5 l were withdrawn before and after cross-linking and analyzed on a 15% (w/v) SDS-polyacrylamide gel. Gels were silver-stained and dried, and radioactive bands were visualized by autoradiography with intensifying screens or by using an imager. For preparative isolation of the crosslinked PI-SceI/oligodeoxynucleotide complex (see below), the analytical scale was increased 10-fold.
Photocross-linking of ds Oligodeoxynucleotide T ϩ 9 and Nicked PI-SceI-PI-SceI (20 M) was digested under limiting conditions in 50 mM Tris/HCl, pH 8.0 with trypsin at a substrate:protease ratio of 500:1 (w/w) at ambient temperature, similar to that described recently (27). After 2 h of incubation, the reaction was terminated by the addition of 5 mM phenylmethylsulfonyl fluoride. The cross-linking reaction with nicked PI-SceI (10 M) was performed in the presence of radioactively labeled ds oligodeoxynucleotide T ϩ 9 (10 M, right-half cleavage product) for 2 h in buffer P. 5-l aliquots withdrawn before and after irradiation were analyzed by electrophoresis on a 15% (w/v) SDS-polyacrylamide gel with subsequent silver-staining and autoradiography. Sequencing of the cross-linked C-terminal tryptic fragment was performed as described recently (27).
Purification of the Cross-linked PI-SceI⅐Oligodeoxynucleotide T ϩ 9 Complex-The cross-linked PI-SceI⅐oligodeoxynucleotide T ϩ 9 (righthalf cleavage product) complex was purified from unreacted PI-SceI by anion exchange chromatography on a Mono Q column (HR5/5, Amersham Pharmacia Biotech). After irradiation, the reaction mixture was incubated in the presence of 2 M urea for 5 min at 60°C and directly applied to the column. The elution buffers used were A, 50 mM Tris/HCl, pH 8.0 and B, 50 mM Tris/HCl, pH 8.0 with 1 M NaCl. The gradient applied was 0 -80% B in 40 min. The flow-rate was 1.0 ml/min. The elution was monitored by measuring the absorbance at 260 nm. Fractions of 1 ml were collected, and aliquots were analyzed by electrophoresis on a 15% (w/v) SDS-polyacrylamide gel.
Protease Digestions of the Cross-linked PI-SceI⅐Oligodeoxynucleotide T ϩ 9 Complex-The purified cross-linked complex of PI-SceI with oligodeoxynucleotide T ϩ 9 obtained by anion-exchange chromatography was concentrated and washed in a Centricon 50 tube with 50 mM Tris/HCl, pH 8.0. An aliquot (5 pmol) was radioactively labeled with [␥-32 P]ATP in the presence of T4 polynucleotide kinase. Free [␥-32 P]ATP was removed using a NAP5 column (Amersham Pharmacia Biotech). The radioactively labeled PI-SceI⅐oligodeoxynucleotide T ϩ 9 complex was digested in 50 mM Tris/HCl, pH 8.0, 1 mM CaCl 2 by various proteases in a reaction volume of 50 l. Trypsin and chymotrypsin were added to give a final concentration of 2, 5, 10, 20, 40, 80, and 200 g/ml. Proteinase K and subtilisin were added to give a final concentration of 2 g/ml and 20 g/ml. The digestions were performed for 16 h at 37°C. The reactions were terminated by precipitation with 0.1 volume of 1 M sodium acetate, pH 6.8, and 2 volumes of ethanol. The samples were dissolved in 20 l of 6 M urea, 0.025% (w/v) bromphenol blue, and 0.025% xylene cyanole and subjected to electrophoresis on a 12% (w/v) polyacrylamide gel containing 0.5 ϫ TBE and 2 M urea after a pre-run for 30 min with a cathode buffer containing 1 mM thioglycolic acid. The gel was dried, and radioactive bands were visualized by autoradiography.
Identification of the Cross-link Site in the Cross-linked PI-SceI⅐Oligodeoxynucleotide T ϩ 9 Complex-5 nmol of PI-SceI were cross-linked with an equimolar amount of ds oligodeoxynucleotide T ϩ 9 (right-half cleavage product) by irradiation for 2 h at 325 nm. The DNA in the reaction mixture was subsequently radioactively labeled with [␥-32 P]ATP and T4 polynucleotide kinase. The buffer was adjusted to 50 mM Tris/HCl, pH 8.0, 1 mM CaCl 2 , and 40 g/ml chymotrypsin was added. The digestion was performed for 2 h at 37°C. To test the progress of chymotryptic proteolysis, an aliquot of the reaction mixture was precipitated with ethanol and analyzed by electrophoresis on a 12% (w/v) polyacrylamide gel containing TBE (0.5ϫ concentration) and 2 M urea. After the digestion was complete, the whole sample was ethanolprecipitated, redissolved in 50 l of 50 mM Tris/HCl, pH 8.0 and 2 M urea, and applied onto a Mono Q column. For elution, the following buffers were used: Buffer A, 50 mM Tris/HCl, pH 8.0; buffer B, 50 mM Tris/HCl, pH 8.0, and 1 M NaCl. The gradient applied was 0 -80% B in 40 min. The flow rate was 1 ml/min. Fractions of 1 ml were collected, ethanol-precipitated, and redissolved in 6 M urea, 0.025% (w/v) bromphenol blue, and 0.025% (w/v) xylene cyanole. 50-l aliquots of the fractions were analyzed by electrophoresis on a 12% (w/v) polyacrylamide, 0.5 ϫ TBE, 2 M urea gel. For further purification of the crosslinked peptide/oligodeoxynucleotide adduct, the peak fractions from the Mono Q column were combined and subjected to preparative electrophoresis under the same conditions as used for the electrophoretic analysis. Radioactive bands were visualized by an imager. The crosslinked peptide/oligodeoxynucleotide adduct was extracted from the gel and eluted into 0.5 ml of 10 mM ammonium bicarbonate, pH 8.8, by shaking for 2.5 h at 37°C. This solution was lyophilized, and the cross-linked peptide/oligodeoxynucleotide adduct was redissolved in 50 l of 10 mM ammonium bicarbonate, pH 8.8, was applied to a 4-ml column of Sephadex G25 (Amersham Pharmacia Biotech), eluted with the same buffer, and lyophilized. The recovery was 400 pmols. For sequencing, the cross-linked peptide/oligodeoxynucleotide complex was solubilized with 200 l of H 2 O and centrifuged using a ProSorb cartridge (Applied Biosystems). The membrane was cut out, washed with 5% (v/v) methanol for 5 min, and dried. The peptide was sequenced on a pulsed liquid phase sequenator Model 477A (Applied Biosystems) with a 120A on-line high performance liquid chromatography system according to Thole et al. (28). 50 pmols of the material were amenable to sequencing.
Site-directed Mutagenesis, Purification, and Characterization of the H333A Variant of PI-SceI-Site-directed mutagenesis of the PI-SceI gene was performed by a polymerase chain reaction-based technique (29) using the primer 5Ј-GCTATGTTACTGATGAGGCCGGCAT-CAAAGCAACAATAAAG-3Ј to introduce the desired mutation. The sequence of the mutated gene was confirmed by sequencing. The purification of the PI-SceI mutant H333A was carried out as described for wild type PI-SceI (7). Binding and cleavage experiments were performed as described by Pingoud et al. (27), and bending assays were performed as described by Wende et al. (7). Photocross-linking experiments with the H333A variant and ds oligodeoxynucleotide T ϩ 9 were carried out as described above for wild type PI-SceI.
Generation of a 311-bp Recognition Site Containing a 5-IdU Modification at Position ϩ9 of the Bottom Strand-A 311-bp substrate carrying the PI-SceI cleavage site in the center and the 5-IdU modification in position ϩ9 of the lower strand was generated by ligating two polymerase chain reaction products. The left half, a 187-bp-long DNA fragment with the cleavage site and the modification was produced with the primers 5Ј-GCGTCGGATCCAGGTCAAAGAGTTTTGG-3Ј and 5Ј-AGACTTCTGCCATTTCATTACCCTCXTTCTCCGCAC-3Ј (X, 5-IdU) and a 311-bp DNA fragment as template in the presence of [␣-32 P]dATP (7). For the generation of the right half, a 124-bp DNA fragment, we used the primers 5Ј-GATGGAATTCCCAGAGTTATATC-3Ј and 5Ј-GC-GTCGGATCCAAGCTTCTCTGGCTGC-3Ј and the unmodified 311-bp substrate as template, again in the presence of [␣-32 P]dATP. Both DNA fragments were annealed with the 311-bp DNA by incubation at 95°C for 10 min and, after cooling to 37°C, ligated with T4 DNA ligase (AGS). The 32 P-labeled ligation product was purified by electrophoresis on a 10% (w/v) polyacrylamide gel. The 311-bp substrate modified in position T ϩ 9 of the recognition site was incubated with 50 nM PI-SceI and irradiated for 1.5 h at 325 nm as described above. Analytical gel shift experiments were performed in the absence and presence of the 56-bp competitor oligodeoxynucleotide F (7). Preparative gel shift experiments of the PI-SceI/311-bp T ϩ 9 complex were carried out before and after photocross-linking. Bands corresponding to the upper and lower complex were excised, eluted in 50 mM Tris/HCl, pH 8.0, precipitated, and analyzed on a 6% (w/v) polyacrylamide gel containing 1% (w/v) SDS.
Modification of Histidine Residues in PI-SceI by Diethylpyrocarbonate (DEPC)-PI-SceI was dialyzed against a buffer consisting of 30 mM inorganic sodium phosphate, pH 6.4, 150 mM NaCl, and 20 mM dithiothreitol. PI-SceI (6 M) in the absence and presence of equimolar amounts of a 62-bp substrate (oligodeoxynucleotide G (7)) was treated with DEPC at final concentrations of 0.25, 0.5, 1, 2.5, 5, and 10 mM for 30 min at ambient temperature. To remove the ethoxyformyl residue from the histidine residues, an aliquot of the DEPC-treated mixture was acidified by NaH 2 PO 4 to reach a pH value of 6.25. Hydroxylamine was added to a final concentration of 250 mM, and the reaction mixture was incubated for 16 h at 4°C. Circular dichroism spectra of 17 M PI-SceI before and after incubation with DEPC (3 mM) were measured in a buffer consisting of 30 mM inorganic sodium phosphate, pH 6.4, and 150 mM NaCl on a JASCO J-710 spectrophotometer at ambient temperature.
Electrophoretic Mobility Shift Assay-For electrophoretic mobility shift assays, PI-SceI was mixed with 10,000 cpm of 32 P-labeled 311-mer polymerase chain reaction product containing the PI-SceI recognition sequence in a total volume of 10 l of binding buffer (10 mM Tris/HCl, pH 7.5, 50 mM NaCl, 1 mM EDTA, 0.05% (w/v) nonfat dry milk, 5% (v/v) glycerol, 10 mM dithiothreitol, 0.1 g of poly(dI-dC)) (27). After electrophoresis, the gels were dried, and bands were visualized and quantified in an imager.

RESULTS
To obtain detailed topological information about specific contacts between PI-SceI and its DNA substrate, we performed photocross-linking experiments using either 5-IdU or 5-IdC monosubstituted ds oligodeoxynucleotides comprising the recognition sequence for PI-SceI (Fig. 1). Photocross-linking of halogenated pyrimidines has been used successfully to identify contacts in DNA-and RNA-protein complexes (30 -34). 5-IdU is an almost perfect analogue of thymidine (35) and particularly useful for the study of DNA-binding proteins using oligode-oxynucleotides with a T 3 5-IdU substitution in defined positions. It is a photoactivated zero-length cross-linker that has the advantage that cross-linking to regions of a protein not involved in DNA binding is minimized and that irradiation with long wavelength UV light (325 nm) does not lead to the excitation of other nucleic acid and protein chromophores. Mechanistic studies of the 5-IdU chromophore relevant to its use in nucleoprotein photocross-linking have been performed by Norris et al. (36). The site-specific 5-IdU-or 5-IdC-mediated photocross-linking method does not depend on information regarding the structure of the protein or the structure of the protein⅐DNA complex. Efficient cross-linking requires the close proximity of the modified base and a reactive amino acid. Preferential targets are Phe, Tyr, Trp, His, and Met residues (35,37). In addition, the cross-link yield depends on a suitable orientation of the reacting groups. With aromatic amino acid side chains as acceptor, the yield of the cross-linking reaction is significantly enhanced when astacking interaction between the base analog and an aromatic amino acid residue is possible (36).
Analytical Photocross-linking of PI-SceI⅐DNA Complexes-We used synthetic oligodeoxynucleotides comprising either the minimal full-length recognition sequence of 36 bp (Fig. 1) or containing the right-half cleavage product of the PI-SceI recognition sequence required for specific binding by PI-SceI (7). These oligonucleotides were substituted with a single 5-IdU moiety that substituted for T, A, or G, or with 5-IdC, which substituted for C, as shown in Fig. 1. It was important to ensure that the presence of such a substitution in the PI-SceI recognition sequence does not interfere with the binding to PI-SceI. We therefore compared the binding of PI-SceI to modified and unmodified DNA in gel shift experiments and found that they are bound by PI-SceI with the same apparent K D of 5-10 nM as the wild type sequence (data not shown). For analytical cross-linking experiments designed to find out which position produces the best cross-link yield, PI-SceI was incubated with the different mono-substituted ds oligodeoxynucleotides for 30 min at ambient temperature at a PI-SceI to DNA ratio of 1:1. The photocross-linking reactions were carrried out by irradiation at 325 nm with a helium/cadmium laser for 2 h at ambient temperature. Among the various positions modified by 5-IdU or 5-IdC in the recognition site, only three positions gave rise to a substantial amount of cross-linked PI-SceI (  (6)) that were substituted are indicated by circles or arrows. Circles indicate positions that did not give rise to a photocross-link, and arrows denote photoreactive substitutions that lead to a photocross-link product with varying yields. A cross-link in very good yields is observed with an oligodeoxynucleotide substituted in position T ϩ 9 of the bottom strand, and two cross-links in good yield are detected with oligodeoxynucleotides modified in position G ϩ 4 and A ϩ 5 of the upper strand, whereas low cross-link yields are found for several additional positions. product with DNA substituted at the other positions as shown in Fig. 1 indicates that either a suitable acceptor amino acid is not available at these positions in sufficient proximity or the stereochemical requirements for cross-linking are not fulfilled. Only the cross-link to position ϩ9 was investigated further, because the yield was high enough to ensure that after purification of the cross-linked complex and its proteolytic degradation, sufficient amounts of a peptide/oligonucleotide adduct would be available for peptide sequencing.
Substitution of 5-IdU in position ϩ9 of the bottom strand results in an apparently homogenous cross-linked species as judged by SDS-PAGE (Fig. 2). The time course of irradiation of the PI-SceI⅐oligodeoxynucleotide T ϩ 9 complex is shown on a silver-stained gel ( Fig. 2A) and on its autoradiogram (Fig. 2B). The yield of this photocross-linked complex was routinely about 20 -30% after 2 h of irradiation and could not be further increased by longer irradiation. Omission of PI-SceI or incubation without irradiation failed to yield any cross-linked material, nor could unmodified DNA be cross-linked to PI-SceI to a significant extent.
Analytical Photocross-linking of Nicked PI-SceI⅐DNA Complexes-To find out whether the cross-link position is localized in the C-or N-terminal half of the endonuclease, we performed photocross-linking with nicked PI-SceI that was generated by limited tryptic digestion under native conditions (27), with the ds oligodeoxynucleotide substituted by 5-IdU in position ϩ9 of the bottom strand. Trypsin cleaves PI-SceI preferentially after Arg-277 between ␤-strand 16 and ␣-helix 6, separating the two conserved LAGLIDADG motifs. The two resulting halfs comprise residues 1-277 and 278 -454, which remain associated under native conditions but dissociate upon treatment with SDS. DNA binding and cleavage experiments had shown that nicked PI-SceI binds to substrate DNA with essentially the same affinity as intact PI-SceI but is devoid of DNA cleavage activity, possibly because it is not capable of inducing the same strong distortion in its substrate DNA as uncleaved PI-SceI, i.e. a bend of 75° (27). Photocross-linking of native and nicked PI-SceI is compared in Fig. 3. As shown in lane 2, cross-linked PI-SceI migrates on an SDS-polyacrylamide gel corresponding to a molecular species that is by 20-kDa larger than uncrosslinked PI-SceI. If the complex between nicked PI-SceI and ds oligodeoxynucleotide T ϩ 9 is irradiated and the reaction mixture analyzed by SDS-PAGE, a cross-linked protein fragment that migrates as a 40-kDa species (lane 4) is observed. Based on these results, this species can only be interpreted as a covalent adduct between the oligodeoxynucleotide modified in position ϩ9 of the bottom strand and the 20-kDa C-terminal half of PI-SceI. This was confirmed by sequencing; the analyzed sequence was NNLNTENPLWDAIVG, which corresponds to the N terminus of the cross-linked C-terminal peptide comprising residues 278 -454.
Preparative Cross-linking of PI-SceI⅐DNA Complexes, Isolation, and Proteolytic Digestion of Cross-linked PI-SceI-After preparative cross-linking of PI-SceI and oligodeoxynucleotide T ϩ 9 (right-half cleavage product), the resulting mixture was resolved using anion-exchange chromatography. The eluted fractions were analyzed by SDS-PAGE, and the covalentlylinked protein⅐DNA complex was identified as the peak eluting at 500 mM NaCl immediately before the free DNA (Fig. 4). Peak fractions contained the PI-SceI⅐oligodeoxynucleotide T ϩ 9 complex with Ͼ95% purity.
To find out which protease would be most suitable to obtain a small and defined fragment of CL PI-SceI, extensive proteolytic digestions were performed with aliquots of the radioactively labeled cross-linked PI-SceI⅐oligodeoxynucleotide T ϩ 9 complex in the presence of increasing amounts of trypsin, chymotrypsin, proteinase K, and subtilisin at 37°C overnight. The degree of digestion was analyzed by polyacrylamide gel electrophoresis in the presence of urea and visualized by autoradiography (Fig. 5). The treatment of the photocross-linked PI-SceI⅐oligodeoxynucleotide T ϩ 9 complex with specific and unspecific proteases converted most of the cross-linked PI-SceI⅐DNA complex to small cross-linked peptides. The chymotryptic degradation was chosen for further analysis, because it produced an apparently defined end product that, because of the specificity of chymotrypsin, is likely to be homogenous. (We found out later that apparently homogenous end products obtained with proteinase K upon sequencing turned out not to have a defined N-terminal sequence).
Identification of the Cross-link Position in the Cross-linked Chymotryptic Peptide-To identify the amino acid residue of the homing endonuclease PI-SceI attached to the 5-IdU residue in position ϩ9 of the bottom strand of ds oligodeoxynucleotide T ϩ 9, a preparative cross-linking experiment with equimolar amounts of PI-SceI and ds oligodeoxynucleotide T ϩ 9 was carried out. After irradiation and radioactive labeling, the re-  Fig. 1) for 30 min at ambient temperature. The mixture was irradiated with a 40-milliwatt helium/cadmium laser emitting at 325 nm for the indicated time periods. In A, the silver-stained gel of the irradiated samples analyzed by 15% (w/v) SDS-PAGE is shown. The positions of cross-linked PI-SceI (CL PI-SceI) and uncross-linked PI-SceI are indicated. A protein contaminant of approximately 25 kDa is present in the PI-SceI preparation. S represents the molecular mass standard. In B, the autoradiogram of the gel is shown .   FIG. 3. Photocross-linking of nicked PI-SceI and ds oligodeoxynucleotide T ؉ 9. PI-SceI was digested with trypsin under limiting conditions, incubated with the 32 P-labeled ds oligodeoxynucleotide T ϩ 9, and irradiated for 2 h as described for PI-SceI. Aliquots before and after irradiation were analyzed by SDS-PAGE on a 15% (w/v) gel. A shows the silver-stained gel, and B shows the corresponding autoradiogram. S is the molecular mass standard, lanes 1 and 2 represent the product mixture obtained with PI-SceI, and lanes 3 and 4 represent that with nicked PI-SceI before and after irradiation, respectively. CL PI-SceI denotes the cross-linked PI-SceI⅐DNA complex, and CL C-term represents the C-terminal fragment comprising amino acids 278 -454 cross-linked to oligodeoxynucleotide T ϩ 9. The positions on the gel representing PI-SceI and the N-and C-terminal fragments are indicated.
action mixture was incubated with chymotrypsin without prior purification of cross-linked PI-SceI. An aliquot of the digested material was precipitated and analyzed by polyacrylamide gel electrophoresis in the presence of urea to confirm complete proteolysis of the cross-linked PI-SceI⅐oligodeoxynucleotide T ϩ 9 complex. The cross-linked peptide was separated from other peptides by anion-exchange chromatography on a Mono Q column (Fig. 6). Aliquots of the fractions 28 -35 were precipitated and analyzed on a urea polyacrylamide gel (Fig. 6). Fractions containing the cross-linked peptide (fractions 32-34) were combined, concentrated, precipitated and loaded onto a preparative urea polyacrylamide gel. The cross-linked peptide/oligodeoxynucleotide T ϩ 9 adduct appeared as a homogenous radiolabeled species with higher apparent molecular weight than the free oligodeoxynucleotide strands (not shown). It was extracted from the gel, desalted and lyophilized. The recovery was 400 pmols. The identity of the photocross-linked peptide/ oligodeoxynucleotide adduct was determined by peptide sequencing. Only 50 pmols of the cross-linked peptide were accessible to sequencing. This low recovery of the chymotryptic fragment in the sequencing analysis may be explained by adsorptive losses, losses during the solubilization and washing procedure and to a large extent by peptide modification by urea that is present in the purification procedure. The peptide sequence covalently linked to position ϩ9 in the lower strand of the PI-SceI recognition sequence, VTDEXGIKA, corresponds to amino acid residues 329 to 337 in the PI-SceI protein sequence. At the fifth position (denoted by X) of the cross-linked peptide, where His-333 was expected, no standard amino acid was identified by sequencing, indicating that the major site of pho-tocross-linking is at residue His-333. This cross-link site resides in domain II, following the second LAGLIDADG motif, which is essential for catalysis.
Loss of Photocross-linking Activity in the H333A Mutant Protein-Based on the identification of His-333 as the amino acid covalently attached to the recognition site, the H333A variant of PI-SceI was produced. Both the binding and cleavage properties of the PI-SceI mutant were characterized and compared with the wild type enzyme. The H333A protein was shown in gel shift experiments to bind to a polymerase chain reaction-generated 311 bp substrate with the same affinity as wild type PI-SceI. In these experiments, two complexes can be observed, an upper complex and a lower complex with low and high mobility, respectively. The two complexes differ by the degree of DNA bending as shown for wild type PI-SceI before (6,7). Compared with wild type PI-SceI, the lower complex is slightly less populated with the H333A mutant. Cleavage properties of the H333A, however, are the same as for the wild type PI-SceI, as shown independently by He et al. (26). As expected, H333A was inactive in producing a cross-link with the T ϩ 9-substituted oligodeoxynucleotide (Fig. 7), indicating that His-333 is indeed responsible for the cross-link to ds oligodeoxynucleotide T ϩ 9.
Absence of Effects of Cross-linking on Formation of Upper and Lower Complex-To find out whether cross-linking of DNA and PI-SceI interferes with the formation of the upper or lower complex observed in gel electrophoretic shift experiments with PI-SceI (6, 7), we generated a 311-bp substrate containing a 5-IdU substitution in position ϩ9 of the bottom strand. This substrate was subjected to a photocross-linking reaction with PI-SceI. In an analytical gel shift experiment, it could be demonstrated that cross-linking had occurred, because a band shift is observed even in the presence of an excess of a specific oligodeoxynucleotide competitor added to the reaction mixture before loading the gel (data not shown). Under these conditions, the radioactively labeled 311-bp substrate T ϩ 9 in the either without or with increasing amounts of the indicated protease (trypsin, chymotrypsin, proteinase K (PK) or subtilisin (Sub)) for 16 h at 37°C. After ethanol precipitation, the digests were analyzed by 2 M urea, 12% polyacrylamide gel electrophoresis in 0.5 ϫ TBE followed by autoradiography. The lane designated ds oligo shows the positions of the two strands comprising ds oligodeoxynucleotide T ϩ 9 (right-half cleavage product). The radioactivity in the two strands is different because of the lower efficiency of labeling of the recessed 5Ј-end of the upper strand. The lower strand is involved in the cross-link, whereas the upper strand is released upon addition of urea. Small amounts of the lower strand seen on the gel may be a contaminant carried over from the purification of CL PI-SceI. Fig.  4 shows that there is no base-line separation of CL PI-SceI and free oligodeoxynucleotide.
PI-SceI⅐DNA complex that had not been irradiated is replaced by unlabeled oligodeoxynucleotide F (7) and therefore does not show a band shift. In addition, a preparative gel shift experiment was performed with PI-SceI before and after cross-linking. Both, upper and lower complexes were extracted from the gel and analyzed by SDS-PAGE. Cross-linked DNA was shown to be present in both complexes after irradiation (data not shown). This result suggests that His-333 is in close proximity to T ϩ 9 in the bottom strand in both the upper and the lower complex or that the cross-linked PI-SceI/311-bp T ϩ 9 adduct is in equilibrium between the two conformations, characteristic for the upper and the lower complex.
Modification of His Residues by Diethylpyrocarbonate-To test whether His residues of PI-SceI are involved in DNA binding and cleavage, a group-specific chemical modification of PI-SceI was performed using DEPC. The ethoxyformylation was carried out with an increasing amount of DEPC in a 2-80-fold excess over His residues. As shown in Fig. 8, the modification of His residues in the PI-SceI molecule has a dramatic effect on DNA binding. A 20-fold molar excess of DEPC over His residues completely abolishes DNA binding by the endonuclease. In the presence of equimolar amounts of DNA containing the recognition site, the modification of His residues is slowed down, indicating that these amino acids are protected from modification by the DNA and, therefore, might be involved in a specific protein-DNA interaction. The ethoxyformylation of His residues is in part reversible by treatment with hydroxylamine, as shown by restoration of the binding activity. This means that the DEPC treatment does not lead to an unspecific denaturation of the enzyme but is a specific effect of the ethoxyformylation of His residues. Experiments in which the cleavage activity of DEPC-modified and -demodified PI-SceI was measured produced nearly the same results, indicating that an impaired DNA binding correlates with a reduced DNA cleavage activity. The cleavage activity is restored by hydroxylamine as well. In fact, circular dichroism analysis demonstrates that in the presence of 3 mM DEPC (10-fold molar excess of DEPC over His residues), the secondary structure composition of PI-SceI is not significantly different from that in the absence of DEPC (Fig. 8B). This result suggests that at least one His residue must be involved in DNA binding and cleavage or located at the protein-DNA interface such that its chemical modification interferes with DNA binding and cleavage. DISCUSSION PI-SceI recognizes an extremely long asymmetrical sequence of more than 30 bp in length, as shown by a primer extension analysis (5), footprinting studies (12), and cleavage assays with substrates of different length (7). Specific binding is associated with strong bending into the major groove (6,7). The center of bending is located at position ϩ7, i.e. 5 and 9 nucleotides downstream of the sites of cleavage in the upper and lower strands, respectively (6,7). Despite the fact that the crystal structure of PI-SceI is known (8), the size of the recognition site makes it difficult to imagine how the DNA is bound. The authors of the crystal structure analysis proposed a docking model based on several criteria.
(i) The scissile phosphodiester bonds must be close to the two Asp residues of the putative active site in domain II, Asp-218 and Asp-326.
(ii) The center of the cleavage site has to be positioned such that the left and right parts are in contact with the two symmetry-related ␤-sheets 7 and 9, respectively, of domain II.
(iii) The DNA backbone must be close to the positively charged residues following the LAGLIDADG helices, at the interface between domains I and II and in the extended region of domain I.
(iv) The bend of the DNA induced by specific complex formation has to be implemented into the docking model.
The published model fulfils these criteria. Its characteristic feature is that the DNA follows the concave contour of the  1 and 2) or the H333A mutant, 5 M (lanes 3 and 4) were incubated with 5 M ds oligodeoxynucleotide (right-half cleavage product) containing 5-IdU in position ϩ9 of the bottom strand for 30 min at ambient temperature. The mixtures were irradiated for 2 h. Aliquots before (lanes 1 and 3) and after irradiation (lanes 2 and 4) were analyzed by electrophoresis on a 12% (w/v) SDS-polyacrylamide gel that was silver-stained. Lane S shows size markers. PI-SceI structure. According to this model, domain II interacts with about 14 bp, covering about 8 bp upstream and 6 bp downstream of the cleavage site. The additional 16 or more bp on the right side extend to domain I. It is conceivable that alternative docking models might satisfy the criteria given by Duan et al. (8), and it is clear that it would be very useful to know points of contacts between the protein and the DNA to test alternative models. Presumably, the most straightforward approach to determine such points of contact is to produce cross-links between PI-SceI and its substrate. We have chosen to use 5-iodopyrimidine-substituted DNA for our cross-linking study because 5-iodouracil and 5-iodocytosine are excellent chromophores to achieve photocross-linking of nucleoprotein complexes in high yield (35,38,31). In these studies up to 95% of the modified nucleic acid could be cross-linked. A recent example in which photocross-linking of 5-IdU-substituted double-stranded DNA to protein has been reported is for Thermus aquaticus MutS, which is cross-linked with a Phe residue close to the N terminus to a 5-IdU-substituted heteroduplex DNA (33).
To find out which positions when substituted by 5-iodopyrimidines in the recognition sequence of PI-SceI give rise to cross-links in good yield, we have synthesized 63 oligodeoxynucleotides monosubstituted with 5-IdU (for T, G, and A) and 5-IdC (for C) in almost all positions of the recognition sequence. These modified oligodeoxynucleotides were hybridized with their complementary unmodified sequence, incubated with PI-SceI in the absence of Mg 2ϩ (which is not required for specific binding), and irradiated. 12 of 32 positions tested in the upper strand and 3 of 31 positions tested in the lower strand proved to be reactive in photocross-linking, albeit with different yields (Fig. 1). The reactive positions thus identified must be in close contact to the protein in the PI-SceI⅐DNA complex and in the vicinity of a reactive amino acid residue; depending on the extent to which these conditions are fulfilled, different cross-linking yields are obtained. Our cross-linking results indicate that reactive positions are found preferentially in the upper strand with a cluster around the cleavage site that is in good agreement with the results of methylation interference experiments (6). The most reactive positions in the upper strand are G ϩ 4 and A ϩ 5, 2 and 3 bp downstream of the site of cleavage. The guanine residue was shown to be protected by PI-SceI in a methylation interference experiment (6). The lower strand contains only few reactive positions. The most reactive position of all positions tested in the lower and upper strand, however, was found in the lower strand at T ϩ 9, 11 bp downstream of the cleavage site. This region contains a run of pyrimidine residues and, thus, was not informative in methylation experiments; it is, however, a region shown to be in close contact with the protein by hydroxyl radical footprint (12) and ethylation interference experiments (6). In addition, photocross-linking experiments in which the phosphate group linking C ϩ 6/T ϩ 7 and T ϩ 7/T ϩ 8, respectively, in the lower strand was substituted by a phosphorothioate group and coupled with p-azidophenacylbromide produced cross-links in good yield, 2 confirming that this region is in close contact with the protein.
We have selected position T ϩ 9 in the lower strand for the identification of a point of contact between PI-SceI and its DNA substrate, as this position proved to be the most reactive of all those tested. Furthermore, it is located approximately one helix turn away from the site of cleavage and, therefore, promised to give useful information regarding the location and orientation of the DNA substrate in that part of the DNA binding site, which is far away from the putative catalytic centers around Asp-218 and Asp-326.
Cross-linking experiments with nicked PI-SceI (27) and an oligodeoxynucleotide modified at position T ϩ 9 demonstrated that cross-linking had occurred to the C-terminal half of PI-SceI (residues 278 -454). Peptide sequencing of an isolated peptide oligodeoxynucleotide adduct from the chymotryptic digest of the cross-linked PI-SceI⅐DNA complex showed that the cross-link involves His-333. In the PI-SceI structure, this residue, located between ␤-strands 19 and 20, protrudes from the protein surface into the cleft, which is generated by the junc-2 F. Christ, unpublished information.
FIG. 8. Binding activity of PI-SceI after modification and demodification of His residues with DEPC; correlation with secondary structure. A, PI-SceI (6 M) was treated with increasing concentrations of DEPC (0 -10 mM) in the absence (f) or presence (q) of equimolar amounts of a 62-mer ds oligodeoxynucleotide containing the PI-SceI recognition sequence. An aliquot of the modified PI-SceI was demodified by treatment with hydroxylamine (Ⅺ, sample modified in the absence of DNA; E, sample modified in the presence of DNA). For a gel shift analysis of the binding of the DEPC-modified and -demodified PI-SceI to a 311-bp DNA substrate, PI-SceI aliquots (50 nM) were incubated with 5 nM 32 P-labeled 311-bp substrate in the presence of 10 g/ml poly(dI-dC). Bound and unbound species were separated by electrophoresis on a native 6% (w/v) polyacrylamide gel and quantified by an instant imager. B shows the circular dichroism spectra of 17 M PI-SceI before (black curve) and after modification by a 10-fold excess of DEPC over His-residues (gray curve). tion between the domains I and II (Fig. 9). The fact that the H333A mutant cannot be cross-linked with an oligodeoxynucleotide substituted in position T ϩ 9 by 5-IdU demonstrates the specificity of the cross-link. Although His-333 is in contact with the DNA, its contribution to DNA binding and cleavage is not significant, as the H333A variant binds and cleaves DNA with similar efficiency as the wild type enzyme (this paper and Ref. 26). On the other hand, given the length of the recognition sequence, large effects are not expected for the H333A substitution, which only removes one of many interactions and is not expected to create a steric problem. This may be the reason why modification of PI-SceI with DEPC, which leads to the ethoxyformylation of His residues and introduces a bulky group at this position, produces an enzyme defective in binding and cleavage. Whether this effect is solely because of modification of His-333 is not clear, in particular as two other His residues in domain II, His-343 and His-377, are important for DNA cleavage (26). That His-333 is likely to be located at the protein-DNA interface can be inferred from a comparison of other LAGLIDADG-type homing endonucleases. Although this residue is not conserved (39), amino acid residues in I-DmoI and I-PorI, located in a similar position as His-333 in PI-SceI, i.e. 6 and 4 -10 amino acid residues, respectively, after the LAGLI-DADG motif, are protected against proteolytic attack by DNA binding (40).
Our results demonstrate that this His-333 is in touch with the DNA in the upper and lower complex. If the scenario is correct that PI-SceI binds DNA first to form a complex (lower complex) in which the DNA is bent by 45°and then undergoes a conformational change to give a catalytically competent complex (upper complex) in which the DNA is bent by 75° (7), then the implication is that the region around His-333 remains in contact with DNA during the conformational change or does not interfere with it.
Conclusions-The main purpose of the cross-linking study presented here has been to find out how the DNA might be located with respect to the enzyme in the specific PI-SceI⅐DNA complex. Given the structural similarity between PI-SceI (8) and I-CreI (10), there is no reasonable doubt that the DNA with its scissile phosphodiester bonds is close to the Asp residues at the ends of the LAGLIDADG motifs. This is one region that interacts with the DNA. Another region has now been defined by a zero-length cross-link to be around His-333. If conformational changes are not excluded, involving for example loops connecting ␤-strands 21 and 22 as well as 15 and 16 and possibly also a subdomain movement to better position ␤-sheet 7, then this means that the DNA will be located in a groove defined by ␤-sheets 7 and 9 in domain II, will leave domain II at His-333, cross the interdomain cleft, and make contact with domain I via ␤-sheet 6, ␣-helix 1, and interconnecting loops. In contrast to the docking model proposed by Duan et al. (8) in which the bent DNA follows the concave contour of PI-SceI, in our model the DNA takes up a convex curvature (Fig. 9). The comprehensive mutational analysis carried out by He et al. (26) to support the docking model of Duan et al. (8) is also in good agreement with our model as are the hydroxyl radical footprint and ethylation interference experiments (6,12). It must be pointed out that our model resembles with respect to domain II the docking model proposed for I-CreI (10), a homodimeric homing endonuclease with one LAGLIDADG motif per subunit. This similarity suggests that PI-SceI, a monomeric protein with a pseudosymmetrical catalytic domain containing two LAGLDADG motifs like I-CreI, has two catalytic sites that cooperate in cleaving the two strands of the duplex substrate in one binding event (7). In this respect, PI-SceI presumably also resembles I-PpoI, the only homing endonuclease for which a co-crystal structure has been solved and for which it is understood how the DNA is bound in the enzyme-substrate and enzyme-product complex (41). FIG. 9. Hypothetical model for the interaction between PI-SceI and its bent DNA substrate. Based on the structure of PI-SceI (Brookhaven Protein Data Bank: 1VDE), a schematic model is presented for the PI-SceI⅐DNA complex in which the DNA containing the recognition site is bent by approximately 70°and positioned such that the cleavage site is located next to the presumptive catalytic Asp residues 218 and 326 and the photocross-link position, His-333, next to thymidine in position ϩ9 of the lower strand (both indicated in red). The protein splicing domain (domain I) and the endonucleolytic domain (domain II) of PI-SceI are located at the bottom and top of the structure, respectively (8).