Specificity Changes in the Evolution of Type II Restriction Endonucleases

(1)

Specificity Changes in the Evolution of Type II Restriction Endonucleases

A BIOCHEMICAL AND BIOINFORMATIC ANALYSIS OF RESTRICTION ENZYMES THAT RECOGNIZE UNRELATED SEQUENCES*

Received for publication, August 6, 2004, and in revised form, November 22, 2004 Published, JBC Papers in Press, November 24, 2004, DOI 10.1074/jbc.M409020200

Vera Pingoud‡§, Anna Sudina^¶, Hildegard Geyer储, Janusz M. Bujnicki**‡‡, Rudi Lurz§§, Gerhild Lu¨ der§§, Richard Morgan^¶¶, Elena Kubareva^¶, and Alfred Pingoud‡

From the ‡Institut fu¨ r Biochemie, Justus-Liebig-Universita¨t, Heinrich-Buff-Ring 58, D-35392 Giessen, Germany, the

¶A. N. Belozersky Institute of Physico-chemical Biology and Chemistry Department, Moscow State University, Moscow 119899, Russia, the储Biochemisches Institut, Justus-Liebig-Universita¨t, Friedrichstrasse 24, D-35392 Giessen, Germany, the **Bioinformatics Laboratory, International Institute of Molecular and Cell Biology, Trojdena 4, 02-108 Warsaw, Poland, the §§Max-Planck-Institut fu¨ r Molekulare Genetik, Ihnestrasse 63-73, D-14195 Berlin, Germany, and

¶¶New England Biolabs, Beverly, Massachusetts 01915

How restriction enzymes with their different specific- ities and mode of cleavage evolved has been a long standing question in evolutionary biology. We have re- cently shown that several Type II restriction endonucle- ases, namely SsoII (2CCNGG), PspGI (2CCWGG), Eco- RII (2CCWGG), NgoMIV (G2CCGGC), and Cfr10I (R2CCGGY), which recognize similar DNA sequences (as indicated, where the downward arrows denote cleav- age position), share limited sequence similarity over an interrupted stretch of ⬃70 amino acid residues with MboI, a Type II restriction endonuclease from Moraxella bovis (Pingoud, V., Conzelmann, C., Kinzebach, S., Su- dina, A., Metelev, V., Kubareva, E., Bujnicki, J. M., Lurz, R., Luder, G., Xu, S. Y., and Pingoud, A. (2003) J. Mol.

Biol. 329, 913–929). Nevertheless, MboI has a dissimilar DNA specificity (2GATC) compared with these en- zymes. In this study, we characterize MboI in detail to determine whether it utilizes a mechanism of DNA rec- ognition similar to SsoII, PspGI, EcoRII, NgoMIV, and Cfr10I. Mutational analyses and photocross-linking ex- periments demonstrate that MboI exploits the stretch of

⬃70 amino acids for DNA recognition and cleavage. It is therefore likely that MboI shares a common evolution- ary origin with SsoII, PspGI, EcoRII, NgoMIV, and Cfr10I. This is the first example of a relatively close evolutionary link between Type II restriction enzymes of widely different specificities.

Restriction enzymes are components of restriction-modification systems that protect bacteria and archaea from invading

foreign DNA (1–3). Four different types of restriction enzymes are known (4) that differ in quaternary structure, cofactor dependence, and mode of DNA recognition and cleavage. Type II enzymes recognize short, typically palindromic sequences and cleave the DNA within or in close proximity to the recognition site (5–7). More than 3570 different Type II restriction enzymes with ⬃240 unique specificities have been biochemi- cally characterized (8). Remarkably, little sequence similarity has been observed between the more than 200 Type II restriction enzymes that have been sequenced to date. The few exceptions consist of isoschizomers that cleave the same sequence at the same position, e.g. EcoRI and RsrI (G2AATTC) (9), MthTI and NgoPII (GG2CC) (10), XmaI and CfrI (C2CCGGG) (11), and Cfr10I and Bse634I (R2CCGGY) (12). Still, most isoschizomers do not share significant sequence similarity. Limited sequence similarity has also been observed in some cases among restriction enzymes that recognize partially related se- quences, e.g. EcoRI (G2AATTC) and MunI (C2AATTG) (13) and SsoII (2CCWGG) and PspGI (2CCNGG) (14). Thus, prior to 1995 the prevailing view was that restriction enzymes are not evolutionarily related (15, 16). This notion changed as additional crystal structures of restriction enzymes became available, clearly demonstrating these proteins to have a highly similar structural core that consists of a four-stranded

␤-sheet flanked by ␣-helices and that harbors the characteristic PD . . . (D/E)XK motif active site (17–19). Furthermore, a sta- tistical analysis revealed a significant correlation between the amino acid sequences (“genotype”) of restriction enzymes and their recognition sequences and mode of cleavage (“phenotype”);

these findings were interpreted as evidence for an evolutionary relationship among Type II restriction endonucleases (20). There is little doubt now that Type II restriction enzymes of the PD . . . (D/E)XK superfamily evolved via divergent evolution (21), a process that was accelerated by frequent exchange of restriction- modification systems through horizontal gene transfer among bacteria and archaea (22). Nevertheless, only slowly do we begin to understand the evolution of type II restriction enzymes.

As mentioned above, amino acid sequence similarities among Type II restriction enzymes appear to be restricted to enzymes that recognize the same or similar DNA sequences and that cleave the DNA at the same position. The paradigm is the correlation first detected for SsoII (2CCNGG) and PspGI (2CCWGG) (14) that was subsequently verified and extended to include EcoRII (2CCWGG), NgoMIV (G2CCGGC), and

* This work was supported by the Deutsche Forschungsgemeinschaft (Grants Pi 122/13-4 and SFB535,Z1), the Fonds der Chemischen Indus- trie, the Deutscher Akademischer Austauschdienst (International Quality Network Biochemistry of Nucleic Acids), Grant 07.03.066 from the State Program Universities of Russia, and Grant 3P04A001124 from the State Committee for Scientific Research in Poland. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertise- ment” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

This report is dedicated to Prof. Dr. G. Maass on the occasion of his 70th birthday.

§ To whom correspondence should be addressed: Tel.: 49-641-99-35402;

Fax.: 49-641-99-35409; E-mail: [email protected].

‡‡ An EMBO/Howard Hughes Medical Institute Young Investigator and a Recipient of the fellowship for young scientists from the Founda- tion for Polish Science.

This paper is available on line at http://www.jbc.org

4289

by on April 14, 2008 www.jbc.orgDownloaded from

(2)

Cfr10I (R2CCGGY) (23–25). It is noteworthy that this group of related enzymes and their close relatives (SsoII: Kpn2kI/

Ecl18kI/StyD4I/SenPI; Cfr10I: Bse634I/BsrFI) comprises: (i) Type IIP enzymes such as SsoII and PspGI, homodimeric enzymes that recognize palindromic sequences; (ii) Type IIE enzymes such as EcoRII, homodimeric enzymes that need two recognition sequences for DNA cleavage, one sequence being cleaved and the other serving as an effector (26 –29); and (iii) Type IIF enzymes such as NgoMIV and Cfr10I, homotetrameric enzymes that cleave, in a concerted manner, DNA with two recognition sites (26, 27, 29, 30). These enzymes share sequence similarities over a stretch of⬃70 amino acid residues, interrupted by two gaps of variable length (Fig. 1). This region is part of the conserved core of the PD . . . (D/E)XK superfamily of Type II restriction endonucleases. In NgoMIV, for which a co-crystal structure is available (31), this region encompasses helices 3 and 7 and␤-strands 2 and 3 that are involved in DNA recognition and cleavage. The involvement of this region in DNA binding and cleavage by Cfr10I and EcoRII is also apparent from crystal structures (32, 33). For SsoII and PspGI, mutational analysis and protein-DNA cross-linking determined this region to be likewise important (23, 25).

SsoII, PspGI, EcoRII, NgoMIV, and Cfr10I belong to the superfamily of PD . . . (D/E)XK enzymes but have a variant configuration of the active site in which the second carboxylate (D/E) is replaced by serine (SsoII, NgoMIV, or Cfr10I), threo- nine (PspGI), or alanine (EcoRII). The catalytic role of the second carboxylate in these enzymes is assumed by a gluta- mate that is 15 or 16 residues downstream, as shown for Cfr10I by mutational replacement (34). That these five enzymes (and their close relatives, see above) belong to different families (Type IIP, Type IIE, and Type IIF) is interpreted as their having a common ancestor, presumably a homodimeric enzyme that already had the variant active site and that acquisition of an effector domain (Type IIE) or generation of a dimer-dimer interface (Type IIF) is a more recent event in evolution (23).

Although this evolutionary relationship is well supported by a previous structure-function relationship study of SsoII and PspGI, the suggestion that other Type II restriction enzymes

that recognize unrelated sequences, such as 2GATC, G2GYRCC, and GG2CC (25), are evolutionarily related to SsoII, PspGI, EcoRII, NgoMIV, and Cfr10I has not yet been explored experimentally.

Here, we present a detailed biochemical analysis of MboI, a Type II restriction enzyme from Moraxella bovis that recog- nizes and cleaves the sequence 2GATC (35, 36). MboI shares sequence similarity with EcoRII, NgoMIV, Cfr10I, SsoII, and PspGI over precisely the region in these enzymes involved in DNA binding and cleavage. However, as evident in Fig. 1, unlike these enzymes, MboI does not have the variant PD . . . (D/E)XK motif but rather another variation, namely FD . . . ETN, related to the catalytic motif of BglII, i.e. ID . . . EVQ (37).

Our results demonstrate that MboI, like SsoII and PspGI, is a Type IIP enzyme: a homodimer, both in the absence and presence of DNA, that cleaves DNA with one site as efficiently as DNA with two sites. Extensive mutational analysis reveals amino acid residues that appear conserved in the alignment with SsoII, PspGI, EcoRII, NgoMIV, and Cfr10I are critical for DNA binding and/or catalysis by MboI. Lys-209 in MboI, suggested by the alignment with Arg-194 in NgoMIV to make a base-specific contact, forms a zero-length cross-link with 5-iododeoxyuridine (5-IdU)¹in an oligodeoxyribonucleotide carrying a GATC recognition site in which the A is substituted by 5-IdU.

We conclude from these results that MboI shares a common structure and evolutionary origin with SsoII, PspGI, EcoRII, NgoMIV, and Cfr10I. This is the first example of a close evolutionary link between Type II restriction enzymes of com- pletely different specificities.

EXPERIMENTAL PROCEDURES

Site-directed Mutagenesis of MboI Variants—Site-directed mutagen- esis of the MboI gene was performed using a PCR-based technique (38). The mutant genes were sequenced and found to contain only the desired mutation.

1The abbreviations used are: 5-IdU, 5-iododeoxyuridine; MES, 4-morpholineethanesulfonic acid; IMAC, immobilized metal-affinity chromatography; HF, hydrogen fluoride; MALDI, matrix-assisted laser desorption ionization; MS, mass spectrometry; TOF, time-of-flight.

FIG. 1. Partial alignment of amino acid sequences of MboI with 20 other restriction enzymes. The alignment is subdivided into three groups, based on the similarity of the recognition sequence. Putative Type II restriction enzymes are indicated with a “P ” postfix. Amino acids are shaded according to the similarity of their chemical properties. Conserved residues are highlighted. Residues subjected to a mutational analysis in MboI are shown above the alignment. Helices and strands are shown as tubes and arrows. The location of the R-box and of amino acid residues of the catalytic motif (PD . . . ExK) is indicated.

(3)

Protein Expression and Purification—His6-tagged MboI and its vari- ants were expressed in Escherichia coli BL21 and purified to near homogeneity using nickel-nitrilotriacetic acid-agarose. Fractions containing pure MboI were collected and dialyzed against buffer D (10 mM

Tris-HCl, 0.1 mMEDTA, 100 mMNaCl, 50% (v/v) glycerol, pH 8.0).

DNA Binding Assay—MboI protein dimer (0.5 nMto 10 ␮^M) was mixed with 5 nM³²P-labeled 188-bp PCR product (containing one MboI recognition site) in binding buffer (10 mMTris-HCl, 100 mMNaCl, 5 mM

CaCl2, 100␮g/ml bovine serum albumin, 10% (v/v) glycerol, pH 8.5) supplemented with 0.1␮g of poly(dI-dC) (Amersham Biosciences). After incubation for 15 min at ambient temperature, the samples were applied to a 6% polyacrylamide gel, electrophoresed for 3.5 h at 80 V (10 V/cm) in 20 mMTris acetate, 5 mMCaCl2, pH 8.5, and the gels subjected to autoradiography using an instant imager (Packard Instrument Co.).

Binding constants were determined by a linear least-square fit of the data to a simple A⫹ B ⫽ C binding model. Gel-shift experiments to compare DNA binding by MboI to a substrate containing two recognition sites were done with a 5 nM³²P-labeled 352-bp PCR product.

DNA Cleavage Assay—DNA cleavage activity of wild-type MboI and its variants was determined by incubating protein with ³²P-labeled 188-bp PCR product in cleavage buffer (50 mMTris-HCl, 100 mMNaCl, 10 mMMgCl2, 100␮g/ml bovine serum albumin, pH 8.0) at 37 °C for different times depending on the activity of the variant. Protein and DNA concentrations are indicated in Table I. Experiments designed to compare the rates of cleavage by wild-type MboI of one- versus two-site substrates were performed with two different PCR products containing one (188 bp) or two recognition sites (352 bp), respectively, in the absence or presence of a specific double-stranded oligodeoxyribonucleotide d(CATACGAGGATCCATTCGCT) (US20/LS20). Reaction conditions are given in the legend to Fig. 4. Reaction products were analyzed by electrophoresis on 10% polyacrylamide gels, which were subsequently subjected to autoradiography using an instant imager (Packard Instrument Co.).

Determination of the Mg²^⫹- and Mn²^⫹Concentration Dependence—

Mg^2⫹(0.05–50 mM) and Mn^2⫹(0.005–2.5 mM) concentration dependence of DNA cleavage by MboI was measured with cleavage buffer (10 mMTris-HCl, 1 mMdithiothreitol, 100␮g/ml bovine serum albumin, pH 7.5) supplemented with 0 –150 mMNaCl to maintain a constant ionic strength. Cleavage was measured with an enzyme concentration of 0.5 nM(Mg²^⫹dependence) or 1 nM(Mn²^⫹dependence) MboI dimer and a substrate concentration of 37 nM 188-bp PCR product containing a single cleavage site. Cleavage reactions were performed for 15 (Mg²^⫹ dependence) or 30 (Mn^2⫹dependence) min at 37 °C.

Determination of the Salt Concentration Dependence—Salt concen- tration dependence of DNA cleavage by MboI was measured in buffers (10 mMTris-HCl, 10 mMMgCl2, 1 mMdithiothreitol, 100␮g/ml bovine serum albumin, pH 8.0) containing varying concentrations of NaCl (0 –200 mM). Cleavage was measured at an enzyme concentration of 1.25 nMMboI dimer and a substrate concentration of 25 nM188-bp PCR product containing a single cleavage site. Cleavage reactions were performed for 4 min at 37 °C.

Determination of the pH Dependence—A triple buffer containing 25 mMMES, 25 mMsodium acetate, and 50 mMTris-HCl (39) was prepared at various pH values. This buffer was chosen to maintain constant ionic strength across the pH range examined. The cleavage rate was measured within the range pH 4.5–10 with an enzyme concentration of 2.5 nMMboI dimer and a substrate concentration of 25 nM188-bp PCR product containing a single cleavage site. Cleavage reactions were performed for 4 min at 37 °C.

Gel Filtration—Gel filtration experiments were performed at room temperature on a Merck-Hitachi high-performance liquid chromatography system using a Superdex 75 column (Amersham Biosciences) equil- ibrated with 10 mMTris-HCl, pH 8.5, 500 mMNaCl, 5 mMCaCl2. Wild type MboI was diluted in the same buffer, which was loaded onto the column, and chromatography was carried out at a flow rate of 0.5 ml/min. Elution was monitored by absorbance at 280 nm. The molecular mass was determined by interpolation, using a calibration curve of proteins of known molecular mass (molecular weight marker kit, Sigma).

Transmission Electron Microscopy—For electron microscopy, DNA fragments of 188 and 352 bp in length, containing one or two recognition sites for MboI, respectively, were produced by PCR and purified. In a reaction volume of 20␮l, 30 ng of DNA was incubated with 30 or 60 ng of MboI (the molar ratio of recognition sites to MboI monomers was

⬃3.5 or 7, respectively) in 10 m^MTris-HCl, pH 8.5, 100 mMNaCl, 5 mM

CaCl2for 15 min at 23 °C, fixed for 10 min with 0.2% (w/v) glutaralde- hyde, and prepared by adsorption to mica as described previously (40).

Photocross-linking—For analytical scale photocross-linking, 10␮^M MboI dimer was preincubated with 10 ␮^M double-stranded eico-

sadeoxyribonucleotide monosubstituted with 5-IdU at various positions in the recognition site (underlined) or the two flanking positions, namely 5⬘-CATACGAGGATCCATTCGCT-3⬘ (Thermo Hybaid) in buffer P (10 mMTris-HCl, 100 mMNaCl, 5 mMCaCl2, pH 8.0) for 10 min at ambient temperature in a volume of 25 ␮l. Photocross-linking was carried out with a 40-milliwatt helium/cadmium laser emitting at 325 nm (Laser 2000). The total irradiation time was typically 45 min, or 0 –90 min in kinetic experiments. Samples of 5␮l were taken before and after cross-linking and analyzed by electrophoresis on a 15% SDS- polyacrylamide gel. Gels were silver-stained, and radioactive bands were visualized by autoradiography with intensifying screens or using an imager. For preparative isolation of a cross-linked MboI-oligodeoxyribonucleotide complex (see below), 20␮^MMboI dimer was preincubated with 20␮^Mdouble-stranded dodecadeoxyribonucleotide monosubstituted with 5-IdU at the A of the recognition sequence, namely 5⬘- CGAGGATCCATT-3⬘ (US12-A-1/LS12, Biomers) in buffer P in a volume of 100␮l. A small amount of the oligodeoxyribonucleotide was radioactively 5⬘-end-labeled with³²P-phosphate.

Digestion of the Cross-linked MboI-dodecadeoxyribonucleotide Com- plex—Cross-linked complex (125␮l and 2.5 nmol) of MboI and dodecadeoxyribonucleotide US12-A-1/LS12 was incubated at 37 °C overnight with 15.5␮g of trypsin (sequencing grade, Roche Applied Science) or a combination of 9.5␮g of trypsin and 9.5 ␮g of chymotrypsin (sequencing grade, Roche Applied Science) or with 12.5␮g of proteinase K (Fermen- tas). Aliquots were analyzed by electrophoresis on a 15% polyacrylamide gel containing 0.5⫻ TBE (89 m^MTris base, 89 mMboric acid, 2 mMEDTA, pH 8.3) and 2Murea (41). After complete digestion, samples were precipitated in the presence of 1/10 volume of 4MLiCl and 3 volumes of ethanol, dissolved in 0.1 Macetic acid, and loaded onto Fe³^⫹-IMAC columns prepared from nickel-nitrilotriacetic acid spin columns (Qiagen) by Ni^2⫹/Fe^3⫹exchange according to a previous study (42). Only unreacted oligodeoxyribonucleotides and peptide-DNA heteroconjugates were retained on the IMAC column when washed as described before (43). The retained DNA-containing components were eluted with 2⫻ 600␮l of H2O, pH 10.5 (adjusted with NH3), and evaporated in a SpeedVac. The dried material was dissolved in 40␮l of H2O.

Mass Spectrometry—Hydrogen fluoride (HF) treatment: selective cleavage of phosphodiester linkages was performed according to a previous study (44) using a modified protocol (45). Briefly, dried samples were incubated with 50␮l of 48% HF (Merck) at 4 °C overnight. The reagent was removed under a stream of nitrogen, and dried samples were dissolved in 50␮l of methanol and dried again; this procedure was repeated once.

MALDI-MS and MS/MS spectra were recorded using an Ultraflex time-of-flight mass spectrometer (Bruker) equipped with a LIFT- MS/MS facility. Samples were prepared on an AnchorChip^TM600/384 target plate. For oligodeoxyribonucleotides and oligodeoxyribonucleotide-peptide heteroconjugates, a saturated solution of 3-hydroxypico- linic acid (Sigma-Aldrich) in acetonitrile/water (45:65, v/v), containing 10% diammoniumhydrogen citrate was used as matrix. Matrix solution (0.5␮l) was applied to the anchor chip and allowed to dry. Then 0.5–1

␮l of analyte was added, and the chip was dried again. HF-treated cross-linked complexes were analyzed as described in detail previously (46). The MALDI mass spectra obtained were externally calibrated using either an oligodeoxyribonucleotide calibration standard or peptide calibration standard II (both from Bruker). Fragment ion analysis in the tandem time-of-flight (TOF-TOF) mode was performed as described (47).

Protein Structure Prediction—Secondary structure prediction and protein-fold recognition was carried out via the GeneSilico metaserver gateway (genesilico.pl/meta/ (48)). The reported alignments between the MboI sequence and structures of PD . . . (D/E)XK nucleases were used as starting points for homology modeling using the “Franken- stein’s monster” approach, comprising cycles of model building, evalu- ation/scoring, realignment in poorly scored regions, and merging of best scoring fragments (49).

RESULTS

The present study was carried out with His₆-tagged MboI (wild type and mutants) isolated from an overproducing E. coli strain and purified to near homogeneity by nickel-nitrilotriacetic acid-agarose.

Analysis of DNA Cleavage by MboI—Optimum DNA cleav- age conditions, a prerequisite for the analysis of structure- function relationships, were determined for MboI. Experiments carried out to assess the Mg^2⫹/Mn^2⫹, ionic strength, and pH

(4)

dependence of DNA cleavage by MboI show that this enzyme, like many Type II restriction endonucleases, has an Mg^2⫹

(Mn^2⫹) optimum between 1 and 10 mM(0.1 and 1 mM, corrected for ionic strength), requires⬃100 m^Msalt for optimal activity, and shows an increase in cleavage activity with pH in the range between 5 and 10 (data not shown).

Classification of MboI as a Type IIP Enzyme—The presump- tive distant relatives of MboI are Type IIP (SsoII and PspGI), Type IIE (EcoRII), and Type IIF (NgoMIV and Cfr10I) restriction enzymes. To ascertain the subtype of MboI, four types of experiments were carried out: (i) gel filtration, to determine the quaternary structure of MboI; (ii) electron microscopy, to es- tablish whether MboI binds two recognition sites simultaneously; (iii) gel electrophoretic mobility shift experiments, to observe if two recognition sites on one DNA molecule are bound cooperatively by MboI and; and (iv) DNA cleavage experiments, to decide whether the rate of cleavage of a substrate having one site can be increased by addition of an oligodeoxyribonucleotide substrate (activation in trans).

Gel filtration runs with MboI and marker proteins show that MboI is eluted between bovine serum albumin and ovalbumin (data not shown). The apparent molecular mass of ⬃60 kDa indicates that MboI is a homodimer of the 32-kDa subunit, excluding it being a homotetrameric Type IIF enzyme. Electron microscopy experiments show that MboI does not bind to two recognition sites simultaneously but rather that on DNA with two MboI sites, one site is occupied preferentially (Fig. 2). Gel electrophoretic mobility shift experiments confirm that MboI does not bind cooperatively to two sites on DNA (Fig. 3). These results already suggest that MboI most likely is neither a Type IIF nor a Type IIE enzyme, which normally bind two sites simultaneously. This was confirmed by DNA cleavage experiments in the presence of increasing amounts of an oligodeoxyribonucleotide substrate (Fig. 4) that clearly demonstrate there is no activation in trans, as would have been expected for a Type IIE enzyme (50, 51). Furthermore, MboI does not cleave a substrate with two sites more readily than a substrate with one site (data not shown). Taken together, these data allow classi- fying MboI as an orthodox Type IIP enzyme, i.e. a homodimer that binds to and cleaves one site at a time and does not require allosteric activation by an effector site.

Mutational Analysis of Residues in MboI Considered Impor- tant for DNA Binding and Cleavage—Based on the alignment shown in Fig. 1, we selected 15 amino acids for a mutational analysis, which included DNA binding and cleavage experiments. The amino acid residues selected fall into three groups:

(i) residues 138 –149 (R-box), whose counterparts in NgoMIV are involved in DNA base and backbone contacts on the 3⬘-side of the recognition sequence; (ii) residues 186 –201, comprising the PD . . . (D/E)XK motif; and (iii) residues 203–215, compris- ing a region that in NgoMIV is responsible for some of the nucleobase contacts. The results of the mutational analysis are summarized in Table I.

Asn-138, conserved among the close MboI relatives (which all recognize 2GATC) and also present in EcoRII, can be replaced by Ala without affecting DNA binding and cleavage significantly. In contrast, Arg-140 and Arg-143, which define the R-box and are conserved in the MboI relatives as well as in EcoRII, SsoII, and PspGI, are very critical for DNA binding and cleavage. Their replacement by Ala but also by Lys leads to catalytically inactive (Ala) or almost inactive (Lys) variants that have a 1000-fold lower affinity for their DNA substrate.

Asn-142, a residue within the R-box and conserved among the close MboI relatives but corresponding to Ser in EcoRII, SsoII, and PspGI, appears to be as important as its neighboring Arg residues; intriguingly, it cannot be substituted by Ser. In con-

FIG. 2. Transmission electron microscopy of MboI-DNA com- plexes. A, the frequency distribution (left) shows specific binding of MboI to its recognition site (TGATCA) in the 188-bp fragment. The gallery on the right shows typical micrographs of the DNA fragment with bound enzyme. B, DNA binding by MboI was studied with a 352-bp fragment with two recognition sites for MboI (115 bp from one end, TGATCA, and 65 bp from the other end, GGATCA). Preferential bind- ing to the TGATCA site is observed. The gallery below shows examples of binding to the left site (first row), to the right site (second row), and fragments with MboI protein bound at both sites (third row). Loop formation by interaction of the two bound proteins was not observed.

The last micrograph in row three of the gallery shows a rare example of two fragments crossed at the bound protein.

(5)

trast, Asp-146, a few amino acids C-terminal of the R-box, and a residue that is not conserved among the MboI relatives, can be replaced by Lys, the corresponding residue in EcoRII, SsoII, and PspGI; the D146K variant has wild type binding affinity and cleavage activity. Ser-149, which aligns with Glu in the close MboI relatives as well as EcoRII, SsoII, and PspGI, can be substituted by Glu without impairment of the DNA binding affinity, but with a 100-fold reduction in DNA cleavage activity.

MboI has an unusual PD . . . (D/E)XK motif, in which Pro is replaced by Phe and Lys by Asn. The FD . . . EXN sequence is, however, fully conserved among the MboI relatives. To demonstrate that this sequence is indeed representing the catalytic center, Asp-186, Glu-199, and Asn-201 were substituted by Ala:

the resulting variants are catalytically inactive, but show un- altered DNA binding affinity, which is not unusual for Ala substitutions of the PD . . . (D/E)XK residues of Type II restric- tion enzymes (compare with the “Discussion”). Interestingly, the N201K variant is catalytically inactive and impaired in DNA binding. Replacing a nearby Tyr (Tyr-203) by Ala leads to a variant with similar properties to the N201K variant.

Based on the alignment with NgoMIV, which uses Asp-193 and Arg-194 for recognition of the central 4 bp of its 6-bp recognition sequence, we have replaced in MboI the corresponding residues Ser-208 and Lys-209, which are conserved among the close relatives of MboI, as well as three adjacent residues, which are partially conserved among the close MboI relatives and have functional groups that could be used for base recognition (e.g.

Asn-211, Glu-212, and Arg-215). The S208A and K209A variants are catalytically inactive and show a strong reduction in DNA binding affinity, as expected for residues involved in base recognition. Intriguingly, the K209R variant is not compromised in DNA binding, but is almost inactive in DNA cleavage, suggesting that Lys-209 is required for DNA binding (a function that can be taken over by Arg) and coupling of recognition to catalysis (a

function that requires a precise interaction, which cannot be taken over by Arg). The N211A variant has near wild-type properties, whereas the E212Q as well as the R215A and R215K variants bind to DNA with wild-type affinity but are almost inactive in DNA cleavage.

Photocross-linking of MboI with 5-IdU-substituted Oligode- oxyribonucleotides—A photocross-linking approach utilizing 5-iododeoxyuridine (5-IdU)-substituted oligodeoxyribonucleotides was used to determine whether the DNA target site GATC is in close contact to amino acid residues of MboI that are suggested by the alignment with EcoRII, SsoII, PspGI, and NgoMIV (Fig. 1) to be involved in DNA binding and cleavage.

For this purpose, six double-stranded eicosadeoxyribonucleo- tides with a central GATC site and monosubstituted in one strand with 5-IdU were synthesized. The substitutions were localized in the recognition site and the two adjacent positions.

These oligodeoxyribonucleotides were tested for cleavage by MboI. It was found that oligodeoxyribonucleotides with 5-IdU substitutions at the positions G, A, and C of the GATC recognition sequence cannot be cleaved by MboI but were all demonstrated to bind to MboI in the presence of Ca²^⫹(41). Results of the photocross-linking reactions are shown in Fig. 5A. Only the oligodeoxyribonucleotide that in one strand had the 5-IdU substitution at the adenine in the GATC recognition site showed a detectable cross-link on a silver-stained gel. A time course of the photocross-linking reaction with a dodecadeoxyribonucleotide with a GATC site, in which in one strand the alanine was substituted by 5-IdU, indicated that⬃30–60 min of irradiation under the given conditions was sufficient to ob- tain the maximum yield of 15% (data not shown).

To identify the amino acid in MboI involved in the cross-link with the oligodeoxyribonucleotide, preparative scale photo- FIG. 4. DNA cleavage by MboI in the absence or presence of a specific double-stranded oligodeoxyribonucleotide. A, 25 nM³²P- labeled 188-bp PCR product containing one recognition site was incubated with 1.25 nMMboI-wt dimer at 37 °C. Aliquots were taken after 0, 0.5, 1, 2, 3, 5, 10, 20, 40, and 60 min, respectively, and analyzed by electrophoresis, followed by autoradiography using an instant imager.

B, cleavage of the one-site substrate in the absence (Œ) or presence of 25 nM(●) or 250 nM(f) of a 20-bp double-stranded oligodeoxyribonucleotide containing one recognition site was evaluated quantitatively. Ad- dition of oligodeoxyribonucleotide substrates leads to a concentration- dependent inhibition of cleavage of the 188-bp PCR product.

FIG. 3. Gel electrophoretic mobility shift experiment to com- pare the binding of MboI to a one-site and a two-site substrate.

A 188-bp DNA containing one TGATCA site (A) or a 325-bp DNA containing two sites (TGATCA and GGATCA) (B) was incubated with increasing concentrations of MboI. Free and bound DNA were sepa- rated by electrophoresis under non-denaturing conditions and subse- quently analyzed using an instant imager. UB denotes unbound DNA;

C1 and C2 denote 1:1 and 2:1 complexes, respectively, of protein and DNA. It is apparent from B that MboI does not bind cooperatively to DNA with two sites.

(6)

cross-linking was carried out with a double-stranded dodecadeoxyribonucleotide having a central GATC site in which the adenine was substituted in one strand by 5-IdU. After photocross-linking, trypsin, chymotrypsin plus trypsin, or proteinase K was added to aliquots of the irradiation mixture. The progress of proteolysis was analyzed by electrophoresis on denaturing gels containing 2^Murea (41). Peptide-oligodeoxyribonucleotide heteroconjugates were isolated from the proteolytic digestion mixture by Fe^3⫹-IMAC and analyzed by MALDI-TOF mass spectrometry (Fig. 6). The mass differences between the heteroconjugates and the unreacted DNA, corrected for the loss of hydrogen iodide during the reaction, allowed us to calculate the masses of the peptide portions. The trypsin digestion gave a fragment with a molecular weight of 2379.6, which was assigned to the peptide TYVIETNYYNSGGSKLNEVAR, having a theoretical average molecular weight of 2379.59. The corresponding results with the chymotrypsin plus trypsin and proteinase K digestions were 1231.9 (observed molecular weight)/NSGGSKLNEVAR/1232.34 (theoretical molecular weight) and 574.8 (observed molecular weight)/GGSKLN/

575.64 (theoretical mass), respectively (Fig. 6A). These results allowed narrowing down the cross-link site to Ser-208, Lys-209, and Asn-211, which are reasonable candidates for an attack by a uridyl radical produced through homolytic cleavage of the I–C bond in iodouracil. To determine the exact site of cross-linking, the peptide-oligodeoxyribonucleotide heteroconjugates were treated with HF to cleave all phosphodiester bonds in the peptide-oligodeoxyribonucleotide heteroconjugates. The MALDI- TOF analyses of the resulting dU-modified peptides are shown in Fig. 6B. Observed masses closely match the theoretical masses.

Fig. 6C shows the result of the MS/MS analysis of the tryptic peptide-dU heteroconjugate, in which Lys-209 was identified as the amino acid residue cross-linked to the 5-IdU-substituted oligodeoxyribonucleotide.

We produced the K209A variant to confirm that Lys-209 is

the cross-linking site in MboI and tried to cross-link this variant with the 5-IdU-substituted oligodeoxyribonucleotide. As shown in Fig. 5B, K209A cannot be cross-linked to this oligode- oxyribonucleotide. The variant K209R, however, which has wild type binding affinity (despite the fact that it is almost inactive in DNA cleavage, see above), is able to form a cross- link with the oligodeoxyribonucleotide in which the deoxyade- nosine had been replaced by 5-IdU, at a reduced yield compared with wild-type MboI.

To facilitate experimental data interpretation in the sequence-structure-function context, we constructed a theoretical model of the MboI catalytic domain (residues 137–280) in complex with target DNA. The homology model of the monomer is based on the fold recognition analysis (see “Experimental Pro- cedures”). The mutual orientation of the two MboI monomers as well as the coordinates of the DNA molecule and the Mg^2⫹

ions was derived from superposition onto subunits A and B in the NgoMIV co-crystal structure. The GATC site was generated by “mutating” bases in the original NgoMIV target (CCGG) and optimizing their geometry using HyperChem 7.1 (Hypercube, Inc.). The final model of the MboI dimer-GATC sequence complex (available from genesilico.pl/iamb/models/R.MboI/R.Mbo- I.model.pdb) was obtained after removing steric clashes between residues from both monomers and the DNA by selecting different rotamers of the respective side chains (Figs. 7 and 8).

DISCUSSION

The PD . . . (D/E)XK superfamily of Type II restriction endo- nucleases is divided into an EcoRI and EcoRV branch (17). This division is based on structural (i.e. dimerization scheme and topology of secondary structure elements) and functional dif- FIG. 5. Photocross-linking of MboI to 5-IdU-substituted oligode- oxyribonucleotides. A, results of cross-linking reactions, in which 10

␮^MMboI dimer was incubated with 10␮^Mdouble-stranded eicosadeoxyri- bonucleotides for 10 min at room temperature, and then irradiated with a laser for 45 min. Double-stranded oligodeoxyribonucleotides comprising the specific MboI recognition sequence were monosubstituted with 5-IdU at six individual positions (underlined): CATACGAGGATCCATTCGCT.

Reaction products generated with the six different substituted oligodeoxyribonucleotides (the position of the 5-IdU-substitution is indicated above the lanes) were analyzed by SDS-PAGE. Bands representing cross- linked MboI (CL-MboI) and uncross-linked MboI are indicated. S, the molecular mass standard. Second lane, the 5-IdU-substituted oligode- oxyribonucleotide (with the 5⬘-flanking G being substituted) was incubated with MboI but not irradiated (⫺h␯, G). B, MboI-wt and the variants K209A and K209R were cross-linked to the 5-IdU-substituted oligodeoxyribonucleotide 5⬘-CATACGAGGXTCCATTCGCT-3⬘, where X denotes the position of the 5-IdU-substitution in the upper strand, the lower strand contains a T in this position). Lane S shows size markers. In contrast to MboI-wt, the K209A variant cannot be photocross-linked, whereas the K209R variant can be cross-linked with a reduced yield compared with wild-type MboI

TABLE I

Relative DNA cleavage and binding activity of MboI variants Cleavage activity was determined using cleavage assays with 25 nM 32P-labeled 188-bp PCR product containing one recognition site (TGATCA) and 1.25–50 nMwild type (wt) or mutant MboI dimer. kapp

values were calculated by measuring the initial velocity of cleavage. A value of 2.4⫾ 0.4 min^⫺1was obtained for the wild-type enzyme. Binding activity was determined using electrophoretic mobility shift assays by incubating 5 nM³²P-labeled 188-bp PCR product with increasing concentrations of MboI or its variants. The binding isotherms were evalu- ated in terms of Kdvalues.

Construct Cleavage activity Binding activity

% Kd

MboI-wt 100 7 nM

MboI-N138A 56 10 nM

MboI-R140A ⬍0.01 ⬎10␮^M

MboI-R140K 0.1 ⬎10␮^M

MboI-N142A ⬍0.01 ⬎1␮^M

MboI-N142S ⬍0.01 ⬎1␮^M

MboI-R143A ⬍0.01 ⬎10␮^M

MboI-R143K 0.05 ⬎10␮^M

MboI-D146K 88 7 nM

MboI-S149E 0.75 6 nM

MboI-D186A ⬍0.01 6.5 nM

MboI-E199A ⬍0.01 7 nM

MboI-N201A 3 10 nM

MboI-N201K ⬍0.01 ⬎440 n^M

MboI-Y203A 0.02 ⬎1␮^M

MboI-S208A ⬍0.01 ⬎1␮^M

MboI-K209A 0.2 330 nM

MboI-K209R 0.4 5.4 nM

MboI-N211A 64 7.4 nM

MboI-E212Q 0.3 6.2 nM

MboI-R215A 0.04 6 nM

MboI-R215K 0.4 5.6 nM

(7)

ferences: enzymes that belong to the EcoRI branch (BamHI, BglII, Bse634I, BsoBI, Cfr10I, EcoRI, EcoRII, FokI, MunI, and NgoMIV) usually approach the DNA from the major groove side (17), recognize the DNA mainly via an ␣-helix and a loop (␣-class (52)) and in general produce 5⬘-staggered ends, whereas enzymes of the EcoRV branch (BglI, EcoRV, HincII,

NaeI, and PvuII) usually approach the DNA from the minor groove side (17), use a ␤-strand and a ␤-like turn for DNA recognition (␤-class (52)), and in general produce blunt or 3⬘- staggered ends. Therefore, the two branches were termed “␣”

and “␤,” respectively (53). Structural variations within the EcoRI (␣) and EcoRV (␤) branch were used for a cladistic FIG. 6. MALDI-TOF mass analysis of peptide-oligodeoxyribonucleotide heteroconjugates. A, mass spectra of the photocross-linked protein-DNA complexes after digestion with trypsin, trypsin plus chymotrypsin or proteinase K, and purification by Fe^3⫹-IMAC. Average masses of the pseudomolecular ions [M⫹H]^⫹of unreacted DNA and heteroconjugates are given. B, mass spectra of the respective dU-peptides after hydrolysis of the peptide-oligodeoxyribonucleotide heteroconjugates with HF. Signals correspond to protonated pseudomolecular ions [M⫹H]^⫹; monoisotopic masses are given. C, identification of the cross-linked amino acid residue in the tryptic dU-peptide (m/z 2604.42). After loss of the deoxyribose unit, sequence-specific peptide ions are generated from the U-containing peptide (m/z 2488.2), which are labeled according to previous studies (60, 61). In addition, the series of y- and b-fragment ions are indicated, and the deduced amino acid sequences are shown.

(8)

analysis of the members of the two branches using methods based on protein structure rather than amino acid sequence (54). This analysis allows inferences regarding evolutionary distances and construction of an evolutionary tree of Type II restriction enzymes of known three-dimensional structure (21).

In the absence of detailed structural information, predictions regarding phylogenetic relationships among restriction endonucleases can be made on the basis of amino acid sequence similarities. However, with few exceptions concerning isoschizomers, restriction enzymes do not show sufficient sequence similarities to infer phylogenetic relationships without additional structural and/or functional information. Based on se-

quence alignments, we previously proposed that SsoII, PspGI, and EcoRII belong to the EcoRI branch and are structurally related to Bse634I, Cfr10I, and NgoMIV (23, 25). This hypoth- esis was supported by a detailed experimental analysis and homology modeling, which demonstrated amino acid residues of SsoII and PspGI that align with functionally important residues in Bse634I, Cfr10I, and NgoMIV to be indeed essential for DNA binding and catalysis by SsoII and PspGI. Our suggestion to include EcoRII in this comparison has been validated by the recently published crystal structure analysis of EcoRII (33). These enzymes recognize similar sequences (e.g. SsoII 2CCNGG, EcoRII 2CCWGG, and NgoMIV G2CCGGC) but belong to different subtypes (Type IIP, IIE, and IIF, respectively). This has been interpreted as the enzymes having di- verged early in evolution, presumably from a Type IIP enzyme that recognized 2CCXGG or X2CCGGX. It is intriguing to note that the sequence similarity over an interrupted stretch of

⬃70 amino acid residues can be extended to enzymes such as MboI (2GATC) that recognize unrelated sequences (25). Show- ing that MboI uses secondary structure elements composed of these⬃70 amino acid residues for DNA binding and cleavage would be the first example of an evolutionary relationship between two subfamilies of restriction endonucleases with widely differing specificities within the EcoRI branch.

In the present study, we followed the approach we developed to demonstrate an evolutionary relationship between SsoII, PspGI, and EcoRII on one side and Bse634I, Cfr10I, and NgoMIV on the other (23, 25). This approach consists of detailed biochemical analyses of the restriction enzyme, mutational analyses of amino acids assumed to be of functional importance, identification of amino acids involved in DNA binding by cross-linking, and homology modeling of the protein-DNA complex. Our results show that MboI is a typical Type II restriction enzyme with a divalent cation requirement, pH, and ionic strength dependence similar to that of PspGI (25). Like PspGI and SsoII, but unlike EcoRII (Type IIE) and Bse634I, and Cfr10I and NgoMIV (Type IIF), it is a Type IIP enzyme, because it interacts with DNA as a homodimer and does not require binding to a second recognition site for efficient DNA cleavage.

The mutational analysis shows that amino acid residues from the R-box (Table I), well conserved among the close MboI relatives, among them DpnII (Fig. 1), are essential for DNA binding and cleavage. The corresponding alanine variants of MboI are catalytically inactive and bind to DNA by approxi- mately three orders of magnitude less than the wild type enzyme. A similar result was obtained with SsoII and PspGI, suggesting that the arginine residues of the R-box play an important role for these enzymes (23, 25). This is emphasized by the finding that also the lysine for arginine substitution of Arg-140 and Arg-143 in MboI is deleterious for DNA binding and cleavage. The amino acids of the R-box are not conserved among Bse634I, Cfr10I, and NgoMIV, but rather are replaced by hydrophilic residues. The co-crystal structure of an NgoMIV-DNA product complex shows that amino acids corresponding to R-box residues are involved in contacts to the region at the 3⬘-end of the recognition sequence, e.g. Gln-63 (which aligns with Asn-142 in MboI) is involved in a hydrogen bond to the N4 of the outer cytosine (underlined) of the G2CCGGC recognition sequence. This suggests that the R-box residues of MboI could be involved in contacts to the 3⬘-end and/or the flanking region of the 2GATC recognition sequence (which is shorter than the G2CCGGC recognition sequence of NgoMIV).

Six residues C-terminal of the R-box is a glutamic acid that is conserved among the restriction enzymes under consider- FIG. 7. Structure prediction for MboI. A, predicted topology of

MboI showing the relative positions of amino acid residues known by mutational analysis to be essential for DNA cleavage. For comparison, the topologies of EcoRII and NgoMIV are shown, as derived from the crystallographic analyses of the free protein and protein-DNA complex, respectively. Amino acids of the variant PD . . . EXK motif in EcoRII and NgoMIV, as well as residues known to be involved in DNA base and backbone contacts in NgoMIV, are indicated. Black arrows and gray cylinders represent the␤-strands and ␣-helices, respectively, conserved among the three enzymes. Catalytic residues are boxed in gray; resi- dues involved in DNA recognition are boxed in white.

(9)

ation in this study; in MboI, however, it is replaced by a serine (Fig. 1). In NgoMIV, this glutamic acid (Glu-70) is within⬍0.4 nm of Asp-140, the first carboxylate of the variant PD . . . (D/E)XK motif. The equivalent residue in MboI is Ser-149. It cannot be substituted by glutamic acid without interfering with catalysis, presumably because the active site of MboI has been fine-tuned in the framework of the entire structure. The cata- lytic motif in MboI is a variation of the classic PD . . . (D/E)XK motif (PDX_{46 –55}KX₁₃E (55)), inasmuch as the conserved lysine is substituted by asparagine. Notably, it is another variation than that found in EcoRII, SsoII, PspGI, NgoMIV, Cfr10I, and Bse634I. The mutational analysis clearly shows that Asp-186, Glu-199, and Asn-201 are essential for catalysis by MboI; furthermore, Asn-201 cannot be replaced by lysine, although this replacement would have reconstituted the classic PD . . . (D/

E)XK motif. Replacements of the active site carboxylates by alanine do not affect the affinity of the MboI variants for their DNA substrate. This is in line with results obtained for other restriction enzymes (56, 57).

In NgoMIV, base-specific contacts to the two inner base pairs are made by amino acids 2– 6 residues downstream from the catalytic lysine (Lys-187), among them Thr-189, Asp-193, and Arg-194; equivalent to Tyr-203, Ser-208, and Lys-209, respectively, in MboI. Replacing these residues by alanine led to inactive mutants, which in addition show a greatly decreased affinity for their DNA substrate, as observed for the R-box variants and as expected for residues involved in recognition.

The replacement of Lys-209 by arginine did not compromise DNA binding but is deleterious for DNA cleavage activity. This indicates that the arginine for lysine substitution interferes with the coupling of recognition and catalysis.

Asn-197 in NgoMIV makes a backbone contact to the 5⬘- phosphate after cleavage of the scissile bond. The corresponding residue in MboI is Glu-212; its replacement by glutamine (as in EcoRII and Cfr10I) renders the variant almost inactive, without affecting the affinity for its substrate. A similar effect is observed for the R215A variant. Based on homology with NgoMIV, it is assumed that this region of MboI is involved in backbone interactions, presumably at the 5⬘-end of the 2GATC recognition site or in buttressing interactions for the catalytic amino acid residues. Although the results of the mutational analysis strongly support our conclusions regarding the involvement of the stretch of⬃70 amino acids in DNA binding and cleavage by MboI, one would like to have direct evidence for this region being in contact with DNA. Our photocross-link experiments show that Lys-209 makes a zero-length cross-link to the second base of the 2GATC recognition sequence.

Taken together, these results allowed us to model the structure of MboI. Fig. 7 shows a secondary structure predic-

tion for the catalytic domain of MboI in comparison with the secondary structure of EcoRII and NgoMIV, deduced from the crystal structures (31, 33). The active site residues are indicated for all three enzymes, whereas the residues involved in DNA binding are only indicated for MboI (based on the mutational analysis) and NgoMIV (based on the co-crystal structure). Fig. 8 shows the stretch of ⬃70 amino acids of the alignment (Fig. 1) projected onto the co-crystal structure of NgoMIV and the structural model of MboI (with the DNA adapted from the NgoMIV-product complex: TGCGGATC- CGC). The two␣-helices and two ␤-strands that make up the active site of these enzymes are indicated. As shown in the MboI model, Lys-209 (corresponding to Arg-194 in NgoMIV) is suitably positioned to cross-link with the second base of the MboI 2GATC recognition sequence (reminiscent of NgoMIV Arg-194, which forms hydrogen bonds to the central CG base pairs of its G2CCGGC recognition sequence).

CONCLUSIONS

In two previous reports (23, 25) we showed that the Type II restriction endonucleases SsoII and PspGI share significant sequence similarities with EcoRII, NgoMIV, Cfr10I, and Bse634I.

It was argued that these enzymes, which recognize similar sequences and have a similar catalytic center (a variant PD . . . (D/E)XK motif), are evolutionarily closely related, despite belong- FIG. 9. Phylogenetic relationship of MboI with other restric- tion enzymes. MboI, which recognizes the sequence 2GATC, is evo- lutionarily related to SsoII, PspGI, EcoRII, NgoMIV, Cfr10I, and Bse634I, all of which recognize variations of the sequence X2CCXGGX (a detailed evolutionary tree is given in Ref. 25). MboI belongs to the EcoRI (␣) branch and is not closely related to its isoschizomers MutH and Sau3AI, which belong to the EcoRV (␤) branch. White boxes, Type IIP enzymes; light gray boxes, Type IIF enzymes; dark gray boxes, Type IIE enzymes.

FIG. 8. Comparison of the NgoMIV co-crystal structure with an MboI model. Part of the NgoMIV-product com- plex structure (left) and of the MboI model (right) illustrating the secondary struc- ture elements involved in DNA recognition and catalysis. The positions of Arg- 194 (NgoMIV), involved in base recognition (G2CCGGC), and Lys-209 (MboI), identified in a cross-link with the second base of the MboI recognition sequence (2GATC), are indicated. The

␣-helices and ␤-strands correspond to the

⬃70 amino acids of the alignment shown in Fig. 1. It is suggested that this ␣ . . .

␤␤␣ unit has been used in a subset of enzymes of the PD . . . (D/E)XK superfam- ily of Type II restriction endonucleases to evolve widely different specificities.

(10)

ing to different sub-types. We have now extended this study to MboI, which recognizes a very different sequence, but makes use of the same structural elements to recognize and cleave DNA using a catalytic center that constitutes another variant of the PD . . . (D/E)XK motif. Having confirmed the functional signifi- cance of the sequence similarities between MboI, SsoII, PspGI, EcoRII, NgoMIV, Cfr10I, and Bse634I, we conclude that these enzymes are evolutionarily related (Fig. 9). They belong to the EcoRI branch of Type II restriction enzymes that use an␣-helix and its preceding loop for DNA recognition. Because MboI has only a slight variation of the PD . . . (D/E)XK (K being substituted by N) motif, we infer that MboI, a Type IIP enzyme, resembles the common ancestor more than other enzymes that have a larger variation of the catalytic motif. In these enzymes, the second carboxylate of the catalytic motif, usually found on the

␤-strand preceding the recognition helix, is recruited from this helix. Two isoschizomers of MboI, namely Sau3AI and MutH (a nuclease involved in mismatch repair, but related to Type II restriction enzymes), in contrast to MboI belong to the EcoRV branch; they are related to each other but not to MboI. Further- more, whereas MutH is a monomer that nicks DNA, Sau3AI is a monomer that dimerizes on the DNA and makes a double strand cut, activated by binding to a second recognition site (like a Type IIE enzyme) (40, 58). MboI and Sau3AI/MutH are therefore examples of independent invention of the same functionality (spec- ificity toward GATC), i.e. a bona fide case of functional conver- gence of two proteins that shared a common ancestor, but underwent extreme divergence. Finally, it appears from our study that the␣ . . . ␤␤␣ unit that characterizes the stretch of

⬃70 amino acids of sequence similarity has been used in a subset of enzymes of the EcoRI branch to evolve widely different specificities. It will be interesting to observe whether restriction endonucleases recognizing G2GYRCC, such as BanI, and GG2CC, such as FnuDI, which share sequence similarity with MboI (25, 59), indeed belong to the EcoRI branch of restriction endonucleases and to determine which elements they use for specific sequence recognition.

Acknowledgments—We thank Drs. Liqing Chen and Hwai-Chen Guo for communicating results prior to publication, Dr. Rudolf Geyer for valuable suggestions, Dr. Dagmar Niemeyer for helpful discussions, Dr. George Silva for critical reading of the manuscript, and Nadine Thome for expert technical assistance.

REFERENCES

1. Wilson, G. G., and Murray, N. E. (1991) Annu. Rev. Genet. 25, 585– 627 2. Bickle, T. A., and Kruger, D. H. (1993) Microbiol. Rev. 57, 434 – 450 3. Szybalski, W., Blumenthal, R. M., Brooks, J. E., Hattman, S., and Raleigh,

E. A. (1988) Gene (Amst.) 74, 279 –280

4. Roberts, R. J., Belfort, M., Bestor, T., Bhagwat, A. S., Bickle, T. A., Bitinaite, J., Blumenthal, R. M., Degtyarev, S., Dryden, D. T., Dybvig, K., Firman, K., Gromova, E. S., Gumport, R. I., Halford, S. E., Hattman, S., Heitman, J., Hornby, D. P., Janulaitis, A., Jeltsch, A., Josephsen, J., Kiss, A., Klaen- hammer, T. R., Kobayashi, I., Kong, H., Kruger, D. H., Lacks, S., Marinus, M. G., Miyahara, M., Morgan, R. D., Murray, N. E., Nagaraja, V., Piek- arowicz, A., Pingoud, A., Raleigh, E., Rao, D. N., Reich, N., Repin, V. E., Selker, E. U., Shaw, P. C., Stein, D. C., Stoddard, B. L., Szybalski, W., Trautner, T. A., Van Etten, J. L., Vitor, J. M., Wilson, G. G., and Xu, S. Y.

(2003) Nucleic Acids Res. 31, 1805–1812

5. Roberts, R. J., and Halford, S. E. (1993) in Nucleases (Linn, S. M., Lloyd, R. S., and Roberts, R. J., eds) pp. 35– 88, Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY

6. Pingoud, A., and Jeltsch, A. (1997) Eur. J. Biochem. 246, 1–22 7. Pingoud, A., and Jeltsch, A. (2001) Nucleic Acids Res. 29, 3705–3727 8. Roberts, R. J., Vincze, T., Posfai, J., and Macelis, D. (2003) Nucleic Acids Res.

31, 418 – 420

9. Stephenson, F. H., Ballard, B. T., Boyer, H. W., Rosenberg, J. M., and Greene, P. J. (1989) Gene (Amst.) 85, 1–13

10. Nolling, J., and de Vos, W. M. (1992) J. Bacteriol. 174, 5719 –5726 11. Withers, B. E., Ambroso, L. A., and Dunbar, J. C. (1992) Nucleic Acids Res. 20,

6267– 6273

12. Grazulis, S., Deibert, M., Rimseliene, R., Skirgaila, R., Sasnauskas, G., Lagu- navicius, A., Repin, V., Urbanke, C., Huber, R., and Siksnys, V. (2002) Nucleic Acids Res. 30, 876 – 885

13. Siksnys, V., Zareckaja, N., Vaisvila, R., Timinskas, A., Stakenas, P., Butkus, V., and Janulaitis, A. (1994) Gene (Amst.) 142, 1– 8

14. Morgan, R., Xiao, J., and Xu, S. (1998) Appl. Environ. Microbiol. 64, 3669 –3673

15. Wilson, G. G. (1991) Nucleic Acids Res. 19, 2539 –2566 16. Heitman, J. (1993) Genet. Eng. (N. Y.) 15, 57–108 17. Anderson, J. E. (1993) Curr. Opin. Struct. Biol. 3, 24 –30

18. Venclovas, C., Timinskas, A., and Siksnys, V. (1994) Proteins 20, 279 –282 19. Aggarwal, A. K. (1995) Curr. Opin. Struct. Biol. 5, 11–19

20. Jeltsch, A., Kroger, M., and Pingoud, A. (1995) Gene (Amst.) 160, 7–16 21. Bujnicki, J. M. (2004) in Restriction Endonucleases (Pingoud, A., ed) pp. 63–93,

Springer-Verlag, Berlin

22. Jeltsch, A., and Pingoud, A. (1996) J. Mol. Evol. 42, 91–96

23. Pingoud, V., Kubareva, E., Stengel, G., Friedhoff, P., Bujnicki, J. M., Urbanke, C., Sudina, A., and Pingoud, A. (2002) J. Biol. Chem. 277, 14306 –14314 24. Tamulaitis, G., Solonin, A. S., and Siksnys, V. (2002) FEBS Lett. 518, 17–22 25. Pingoud, V., Conzelmann, C., Kinzebach, S., Sudina, A., Metelev, V., Kubareva, E., Bujnicki, J. M., Lurz, R., Luder, G., Xu, S. Y., and Pingoud, A. (2003) J. Mol. Biol. 329, 913–929

26. Mucke, M., Kruger, D. H., and Reuter, M. (2003) Nucleic Acids Res. 31, 6079 – 6084

27. Halford, S. E., Bilcock, D. T., Stanford, N. P., Williams, S. A., Milsom, S. E., Gormley, N. A., Watson, M. A., Bath, A. J., Embleton, M. L., Gowers, D. M., Daniels, L. E., Parry, S. H., and Szczelkun, M. D. (1999) Biochem. Soc.

Trans. 27, 696 – 699

28. Reuter, M., Mu¨ cke, M., and Kru¨ ger, D. H. (2004) in Restriction Endonucleases (Pingoud, A., ed) pp. 261–295, Springer-Verlag, Berlin

29. Welsh, A. J., Halford, S. E., and Scott, D. J. (2004) in Restriction Endonucle- ases (Pingoud, A., ed) pp. 297–317, Springer-Verlag, Berlin

30. Siksnys, V., Grazulis, S., and Huber, R. (2004) in Restriction Endonucleases (Pingoud, A., ed) pp. 237–259, Springer-Verlag, Berlin

31. Deibert, M., Grazulis, S., Sasnauskas, G., Siksnys, V., and Huber, R. (2000) Nat. Struct. Biol. 7, 792–799

32. Bozic, D., Grazulis, S., Siksnys, V., and Huber, R. (1996) J. Mol. Biol. 255, 176 –186

33. Zhou, X. E., Wang, Y., Reuter, M., Mucke, M., Kruger, D. H., Meehan, E. J., and Chen, L. (2004) J. Mol. Biol. 335, 307–319

34. Skirgaila, R., Grazulis, S., Bozic, D., Huber, R., and Siksnys, V. (1998) J. Mol.

Biol. 279, 473– 481

35. Gelinas, R. E., Myers, P. A., and Roberts, R. J. (1977) J. Mol. Biol. 114, 169 –179

36. Ueno, T., Ito, H., Kimizuka, F., Kotani, H., and Nakajima, K. (1993) Nucleic Acids Res. 21, 2309 –2313

37. Lukacs, C. M., Kucera, R., Schildkraut, I., and Aggarwal, A. K. (2000) Nat.

Struct. Biol. 7, 134 –140

38. Kirsch, R. D., and Joly, E. (1998) Nucleic Acids Res. 26, 1848 –1850 39. Ellis, K. J., and Morrison, J. F. (1982) Methods Enzymol. 87, 405– 426 40. Friedhoff, P., Lurz, R., Luder, G., and Pingoud, A. (2001) J. Biol. Chem. 276,

23581–23588

41. Geyer, H., Geyer, R., and Pingoud, V. (2004) Nucleic Acids Res. 32, e132 42. Stensballe, A., Jensen, O. N., Olsen, J. V., Haselmann, K. F., and Zubarev,

R. A. (2000) Rapid Commun. Mass. Spectrom. 14, 1793–1800

43. Steen, H., Petersen, J., Mann, M., and Jensen, O. N. (2001) Protein Sci. 10, 1989 –2001

44. Ferguson, M. A., Homans, S. W., Dwek, R. A., and Rademacher, T. W. (1988) Science 239, 753–759

45. Lochnit, G., Dennis, R. D., Ulmer, A. J., and Geyer, R. (1998) J. Biol. Chem.

273, 466 – 474

46. Lochnit, G., and Geyer, R. (2004) Biomed. Chromatogr. 18, 841– 848 47. Wuhrer, M., Robijn, M. L., Koeleman, C. A., Balog, C. I., Geyer, R., Deelder,

A. M., and Hokke, C. H. (2004) Biochem. J. 378, 625– 632

48. Kurowski, M. A., and Bujnicki, J. M. (2003) Nucleic Acids Res. 31, 3305–3307 49. Kosinski, J., Cymerman, I. A., Feder, M., Kurowski, M. A., Sasin, J. M., and

Bujnicki, J. M. (2003) Proteins 53, Suppl. 6, 369 –379

50. Pein, C. D., Reuter, M., Cech, D., and Kruger, D. H. (1989) FEBS Lett. 245, 141–144

51. Reuter, M., Kupper, D., Pein, C. D., Petrusyte, M., Siksnys, V., Frey, B., and Kruger, D. H. (1993) Anal. Biochem. 209, 232–237

52. Huai, Q., Colandene, J. D., Chen, Y., Luo, F., Zhao, Y., Topal, M. D., and Ke, H. (2000) EMBO J. 19, 3110 –3118

53. Bujnicki, J. M. (2001) Acta Biochim. Pol. 48, 935–967 54. Bujnicki, J. M. (2000) J. Mol. Evol. 50, 39 – 44

55. Horton, J. R., Blumenthal, R. M., and Cheng, X. (2004) in Restriction Endo- nucleases (Pingoud, A., ed) pp. 361–392, Springer-Verlag, Berlin 56. Selent, U., Ruter, T., Kohler, E., Liedtke, M., Thielking, V., Alves, J., Oelge-

schlager, T., Wolfes, H., Peters, F., and Pingoud, A. (1992) Biochemistry 31, 4808 – 4815

57. Lagunavicius, A., and Siksnys, V. (1997) Biochemistry 36, 11086 –11092 58. Friedhoff, P., Thomas, E., and Pingoud, A. (2003) J. Mol. Biol. 325, 285–297 59. Bujnicki, J. M., and Rychlewski, L. (2001) J. Mol. Microbiol. Biotechnol. 3,

69 –72

60. Roepstorff, P., and Fohlman, J. (1984) Biomed. Mass Spectrom. 11, 601 61. Biemann, K. (1988) Biomed. Environ. Mass Spectrom. 16, 99 –111