Characterization of BcgI, a new kind of restriction-modification system.

The BcgI restriction enzyme from Bacillus coagulans is unusual in that it cleaves on both sides of its recognition site, CGAN6TGC, releasing a fragment that includes the site and several bases on each side. We report the organization and nucleotide sequences of the genes for the BcgI restriction-modification system and the properties of the proteins that they encode. The system comprises two adjacent, similarly oriented genes. The proximal gene, bcgIA, codes for a 637-amino acid protein (molecular mass = 71.6 kDa) that resembles certain m6A-specific DNA-methyltransferases, particularly those that constitute the modification subunits of type I restriction-modification systems. The distal gene, bcgIB, codes for a 341-amino acid protein (molecular mass = 39.2 kDa) that resembles none of the sequences in the sequence data bases. The two genes overlap by several nucleotides. Alone, neither protein restricts or modifies DNA, but, together, they form a complex in the proportion A2B that does both. DNA binding assays showed that the DNA-protein complex can be formed only in the presence of both subunits, suggesting that the association of inactive subunits generates the active BcgI enzyme that can bind DNA and then either cleaves or methylates at target site.

The nucleotide sequence(s) reported in this paper has been submitted L17341.

to the GenBankTM/EMBL Data Bank with accession number(s)
$ T O whom correspondence should be addressed. Tel.: 508-927-5054; Fax: 508-921-1350; e-mail: kong@neb.com. tree, that R-M systems evolved only a few times, and that present day systems retained these ancestral organizations, while their target specificities diverged. The general lack of homology among restriction enzymes, in particular, argues against this explanation, however, and suggests instead that R-M systems evolved numerous times (4)(5)(6). Rather than reflecting separate evolutionary trees, then, the classes seem to represent nodes of evolutionary convergence, organizational formats of particular effectiveness toward which initially disparate systems gravitated. If this latter explanation is true, then we might expect to find rather more variety among R-M systems than the three-or four-class scheme would otherwise imply. And, indeed, as the number of well characterized systems increases, so too does our appreciation of their diversity (7-16).
Here we describe the properties of an R-M system with propertic distinct from those of other described systems, the BcgI system from Bacillus coagulans. BcgI enzyme recognizes the discontinuous, asymmetric sequence 5'-CGAN6TGC-3', and, in the presence of MgZ+ and S-adenosylmethionine (AdoMet), it cleaves bilaterally and symmetrically outside the sequence to release a 34-base pair fragment (17). The unique properties of BcgI enzyme make it an attractive system for further study. We have cloned the genes for the BcgI R-M system into Escherichia coli. In this paper, we report the organization and nucleotide sequences of the two genes, bcgIA and bcgIB, and the properties of the related A and B subunits that they encode. The results of this study suggest that the BcgI R-M system is different from all characterized R-M systems not only in its unique enzymatic properties but also in its unique gene composition and subunit functions.
BcgI Endonuclease and Methyltransferase Assays-One unit of BcgI endonuclease is the minimum amount needed to completely digest 1 pg of phage X DNA in 50 pl of BcgI cleavage buffer in 1 h at 37 "C. BcgI cleavage buffer contained 10 mM Tris-HC1 (pH 8.4), 10 mM MgClz, 100 mM NaC1, 1 m M dithiothreitol (DTT), and 20 pM AdoMet. One unit of BcgI methylase is the minimum amount needed to completely modify 1 pg of phage X DNA in 50 pl of Beg1 methylation buffer in 1 h at 37 "C. BcgI methylation buffer contained 10 mM Tris-HCl (pH 8.41, 10 mM Na'EDTA, 100 mM NaCl, 1 mM DTT, 80 p~ AdoMet. Cloning and Sequencing the BcgI R-M Genes-The genes encoding the BcgI R-M system were cloned on an 8-kilobase ClaI fragment of B. coagulans DNA by selecting for protectively modified, BcgI-resistant plasmid recombinants (21). A 3.5-kilobase HpaI-RsrII subfragment was sequenced using the Circumvent thermo-cycle sequencing kit (New England Biolabs). The first strand was sequenced from serial deletion derivatives made using exonuclease 111, and the second strand was sequenced using custom synthesized primers.
Purification of BcgI Enzyme-BcgI endonuclease activity was purified from E. coli ER1821 containing pbcgIAB-10, a recombinant plasmid containing the BcgI R-M genes downstream of the lac promoter in pUC19. All operations were performed at 4 "C unless otherwise noted. 1) Frozen cells (327 g) from cultures grown at 37 'C were thawed and suspended in 673 ml of buffer A (20 mM Tris-HCl (pH 7.5), 0.1 mM Na'EDTA, 1 mM DTT, 5% glycerol) containing 100 mM NaCl. Phenylmethylsulfonyl fluoride was added to 25 pg/ml, and the suspension was sonicated. Following cell rupture, NaCl was added to a final concentration of 300 mM, and the insoluble material was removed by centrifugation at 30,000 X g for 40 min.
2) The supernatant was applied to a DEAE cellulose column (15 X 5 cm) equilibrated with buffer A containing 300 m M NaCl. Following overnight application, the column was washed with 400 ml of the same buffer. BcgI endonuclease activity eluted in the flow-through; the majority of the nucleic acids bound to the resin. 3) Ammonium sulfate (288 g) was slowly added to the flow-through fraction (950 ml) with stirring, and the precipitate was recovered by centrifugation at 22,000 X g for 30 min. The pellet was dissolved in 250 ml of buffer B (10 mM potassium phosphate (pH 7.0), 100 mM NaCl, 10 mM 2-mercaptoethanol, 0.1 mM Na'EDTA) and dialyzed against two changes of 4 liters of the same buffer. 4) The dialyzate was applied to a heparin-Sepharose column (2.5 X 20 cm) equilibrated with buffer B. The column was washed with 250 ml of buffer B and then eluted with a 1-liter linear gradient of 0.1-1 M NaCl in buffer B. BcgI activity eluted around 0.35 M NaCl. 5) Active fractions were pooled and applied to a hydroxylapatite column (2.5 X 40 cm) equilibrated with buffer C (10 mM potassium phosphate (pH 7.0), 0.5 M NaCI, 1 mM DTT, 0.1 mM NaBDTA). The column was washed with 400 ml of the same buffer and then eluted with a 2-liter linear gradient of 0.01-0.5 M potassium phosphate in buffer C. Active fractions lacking nonspecific exonuclease were pooled and dialyzed against buffer D (20 mM Tris-HC1 (pH 7.4), 10 mM 2-mercaptoethanol). 6) The dialyzate was applied to a Q-Sepharose HR 10/10 high performance liquid chromatography column (Pharmacia LKB Biotechnology Inc.) equilibrated with buffer D. The column was washed with 10 ml of the same buffer and then eluted with a 100-ml linear gradient of 0.1-1.0 M NaCl in buffer D. BcgI activity eluted around 0.35 M NaCl. Active fractions were pooled and again dialyzed against buffer D. 7) The dialyzate was applied to a heparin Tsk 1 column (Toso Haas) equilibrated with buffer D. The column was washed with 10 ml of the same buffer and then eluted with a 65-ml linear gradient of 0.1-1 M NaCl in buffer D. BcgI endonuclease activity eluted around 0.3 M NaCl. Active fractions were pooled, dialyzed against storage buffer (10 mM Tris-HC1 (pH 7.41, 100 mM NaCl, 1 mM DTT, 0.1 mM Na'EDTA, and 50% glycerol), and kept at -20 "C. Subunit Composition Analysis-The purified BcgI enzyme (10 r g in a 250-pl injection volume) was subjected to a final chromatography . . on a Waters liquid chromatograph using a POROS R (4.5 X 150 mm) reverse-phase column, developed with a linear gradient of 5-100% acetonitrile in 0.1% trifluroacetic acid over 10 min at a flow rate of 5 ml/min with detection at 214 nm at a sensitivity of 0.5 absorbance unit. The peaks, which eluted at 5.70 and 6.18 min, were identified as the 40-and 70-kDa subunits, respectively, by SDS-polyacrylamide gel electrophoresis. The mass ratio of the absorbances of the integrated peaks from the POROS R column, 7040 kDa, was 3.653, indicating the presence of twice as many large subunits in BcgI complex as small subunits (2 X 71559/39161 = 3.655).
Electrophoretic Mobility Shift Assays-DNA binding and mobilityshift assays were performed with either labeled proteins or labeled DNA in 10 mM Tris-HC1 (pH 8.0), 50 mM NaCl, 10 mM 2-mercaptoethanol, 1 mM Na'EDTA, 20 PM Sinefungin at 30 "C for 30 min. The reaction mixtures were incubated at 30 "C for 1 h. Glycerol was added to lo%, and the DNA-protein complexes were loaded onto a 6% polyacrylamide gel cast in 10 mM NaH2P04 (pH 6.0) and 0.1 mM L-tryptophan (22) or 45 mM Tris, 45 mM boric acid, 2 mM Na'EDTA and electrophoresed at 150 volts for 3 h. The gels were fixed in 50% methanol, 10% acetic acid; 10% acetic acid; then 25% methanol, 10% acetic acid. The fixed gels were dried and exposed to x-ray films. To enhance the retardation of DNA-protein complexes, electrophoresis was performed using gels prepared at different pH values.

Cloning and
Sequencing the Beg1 Genes-The genes for the BcgI R-M system were cloned into E. coli by the methylase selection procedure as detailed in Ref. 21. ClaI fragments of B. coagulans DNA were ligated to pBR322 and transformed into E. coli RR1. The transformants were plated and grown overnight to allow plasmids carrying the methyltransferase gene to become protectively modified, and then the plasmid population was purified, digested with BcgI, and transformed back into RR1 to recover BcgI-insensitive survivors (pBR322 contains three BcgI sites, and, therefore, cleavage by BcgI destroys unmodified plasmids). Plasmids thus recovered carried an 8-kilobase ClaI fragment in common, bore BcgIspecific modification, and expressed BcgI endonuclease activity. The fragment from one of these clones was transferred to pUC19 to form pbcglAB-10 ( Fig. 1).
The ClaI fragment was subcloned to determine the boundaries of the BcgI genes and t o remove extraneous DNA. The minimum subclone that expressed both endonuclease and methyltransferase activities, pbcgIAB-HR, comprised a 3.5kilobase HpaI-RsrII fragment in pUC19. The nucleotide sequence of this fragment was determined (Fig. 2). It was found to include two large, similarly oriented open reading frames (ORFs), bcgIA (nt 173-2086) and bcg1B (nt 2079-3104), coding for proteins of 637 amino acids (71,559 Da) and 341 amino acids (39,161 Da), respectively. Both ORFS are preceded by putative ribosome binding sites, AGGTG (nt 160-164) 8 nt upstream of the A gene and AAGGAG (nt 2064-2069) 9 nt upstream of the B gene. The two ORFs overlap by several nt, and, therefore, the ribosome-binding site and the start codon of the B ORF lies within the 3' terminus of the A ORF. Such an intimate gene association is common among R-M systems.
Purifkation and Characterization of the BcgI Enzyme-The BcgI endonuclease activity was purified to near-homogeneity from E. coli ER1821 cells carrying pbcg1AB-10. It was found to comprise two proteins of 70-and 40-kDa in the molar ratio 2:l. The molecular masses of these proteins are close to those predicted for the BcgI A and B proteins on the basis of nt sequence (Fig. 3). Five-column chromatography steps were performed during the purification, and the fact that the two proteins stayed together throughout suggested that they were tightly associated. In an unsuccessful attempt to dissociate them, the endonuclease preparation was treated with 1 M urea and 1 M NaCl and applied to a Sephadex G-100 gel filtration column equilibrated in the same buffer. The two proteins again eluted in the same fractions, confirming their tight association. Previous experiments showed that the BcgI endonuclease requires AdoMet for activity and that it also catalyzes BcgI-specific methylation (17). Taken together, this suggests that BcgI is a trimer of composition A2B (or a higher multimer of 2 A1 B subunits) that manifests both endonuclease and methyltransferase activities.
N-terminal Amino Acid Sequence Analysis-The BcgI enzyme, prepared from ER1821 (pbcgZAB-lo), was subjected to electrophoresis and electroblotted according to the procedure of Matsudaira (23), with modifications as previously described (24). The membrane was stained with Coomassie Blue R-250, and the two protein bands of 70 and 40 kDa were excised and each subjected to sequential degradation (24). The first 18 residues of the 40-kDa subunit were found to be Met-Asn-

Asn-Leu-Ile-Lys-Tyr-Ser-Thr-Phe-Leu-Ile-Ser-Asp-Leu-
Phe-Asp-Val, corresponding to the first 18 codons of the bcgZB ORF. The first 14 residues of the 70-kDa subunit were likewise consistent with the sequence Met-Val-Asn-Glu-Lys-Thr-Ser-Thr-Asp-Gln-Leu-Val-Arg-Arg, corresponding to the first 14 codons of the bcgZA ORF. The 70-kDa subunit was found to be ragged, as it produced 2 PTH residues in each cycle, i.e. Met and Val were observed in cycle one, and Val and Asn were observed in cycle two, and so on. Approximately half of the 70-kDa subunit lacked the initial methionine residue, suggesting that this protein had undergone incomplete processing.
Comparative Sequence Analysis-The deduced amino acid sequences of the BcgI A and B proteins were compared with the contents of the Genbank, EMBL, SWISS-PROT, and PIR data bases using the BLAST network service from the National Center for Biotechnology Information (25, 26). The A protein was found to resemble several methyltransferases that form N-6 methyladenine in DNA, most notably those belonging to the y class (27). Members of this class include the modification enzymes of the type I1 R-M systems AccI, BanIII, HincII, PaeR71, and TaqI (28-32) and the modification subunits of the E. coli and Salmonella type I systems (33, 34), among others. These proteins are relatively large and heterogeneous, but all include the two-amino acid sequence motifs that are hallmarks of m'Ay-MTases, . . .LEP-G-G-F. . . and. . .NPPY. . ., or equivalents, 50-70 amino acids apart at their N termini. The match with the type I M subunits was the closest, registering a score of 25% amino acid sequence identity in alignment with EcoA M (Fig. 4). This, albeit scant, similarity between BcgI A and characterized modification enzymes implies that it functions as a methylase domain in the BcgI system. The B protein resembled none of the sequences in the NCBI data bases, nor did it resemble any of those in our personal data base of restriction and modification enzymes. Restriction enzymes are, as a rule, quite different from one another, and, therefore, the very absence of homology between BcgI B and other proteins might indicate, perversely as it were, that BcgI B functions as the endonucleolytic domain.
BcgI Methylation Produces N-6 Methylodenine-To identify the product of BcgI methylation, double-stranded duplexes were incubated with 3H-labeled AdoMet, purified BcgI enzyme, and analyzed for modified bases. Methylation was found to be sequence-specific in that 10 times more radioactivity was incorporated into the specific duplex that contained a BcgI site than the nonspecific duplex (data not shown). Following methylation, the specific duplex was digested to mononucleosides with phosphodiesterase I and alkaline phosphatase, unlabeled methyldeoxynucleosides were added to provide visual markers, and the mixture was separated by thin-layer chromatography (35, 36). The m6A, m4C, and mSC spots were counted for radioactivity, 95% of which was found to comigrate with m6A, indicating that N-6 methyladenine is the sole product of BcgI methylation (Table I). This result confirms the identification of BcgI as an m6A-methyltransferase, based on amino acid sequence comparisons.
The BcgI recognition sequence, CGAN6TGC, includes 1 adenine residue in each strand. We do not know whether the adenines are methylated in both strands or in only one. In type I systems, both strands are methylated (37, 38), but in type I11 systems, only one strand is methylated (39).

~~ ~~~ ~ ~ 3421 ATTCMTMGCT~~~~AGAGCCTTTAGCGGTCCG 3454
Actiuities of the Individual BcgI Proteins-Most characterized R-M systems comprise two genes, one for the endonuclease and one for methyltransferase. Usually, these genes occur side-by-side, and they code for enzymes that function independently. The BcgI system comprises two adjacent genes; to

FIG. 3. Coomassie-stained 10-20% gradient SDS-polyacrylamide gel electrophoresis of purified BcgI endonuclease (Bcglproteim lane).
The sizes of the molecular weight markers ( M W ) are given in kDa. Two proteins are present in the endonuclease preparation, BcgI A (70 kDa) and BcgI B (40 kDa). The ratio of the absorbances of the two peaks was found by scanning densitometry to be -3.3, indicating that the two proteins are present roughly in the proportion 3.3 + (70/40) = 1.9 A 1 B. This is close to the value of 2:l obtained by POROS R chromatography. test whether they code for enzymes that also function independently, polymerase chain reaction fragments containing the 6cgIA and bcgZR genes were separately cloned into the plasmid expression vector, pLM1, to form p6cgIA-2 and p6cgIB-7, respectively ( Fig.  1). Plasmid p6cgIA-2 (carrying just bcgIA, the putative modification gene) expressed n o detectable BcgI endonuclease, nor did it express RcgI methyltransferase, judging by its complete sensitivity to HcgI digestion. This was true whether the adjacent, upstream, phage T7 promoter was induced or uninduced. Plasmid p6cgIR-7 (carrying just bcgIB, the putative endonuclease gene) behaved similarly and expressed neither endonuclease nor methyltransferase, again regardless of induction.
We were not able to test the combination of p6cgIA-2 and p6cgIB-7 in the same cell, because their origins of replication were incompatible, but we were able to do so with two comparable plasmids, pbcgIA-31 and p6cgIR-8. p6cgIA-31 is an exonuclease 111 deletion derivative of pbcgIAR-10 containing the 6cgIA gene; pbcgIE-8 is a pSClOl derivative expressing the 6cgIB gene in pSYX19 (Fig. 1). When p6cgIA-31 and pbcgIB-8 were transformed into the same cell, they complemented; the plasmid DNA became resistant to RcgI digestion, signifying methyltransferase expression, and RcgI endonuclease activity was readily detected in cell extracts. This suggests that the BcgI proteins function in a mutually dependent fashion, such that both genes, in either cis or trans, are needed for the manifestation of either activity.
Complementation between the A and R proteins was also demonstrated in uitro. RcgI A and B subunits were separately purified by heparin-Sepharose chromatography, and the fractions were assayed by complementation using crude cell extracts of cells expressing the alternate subunit from E. coli cultures carrying either plasmid p6cgIA-2 or plasmid p6cgIR-  ..   Identification of the base methylated by BcgZ Purified BcgI enzyme complex was incubated with [3H]-labeled AdoMet and the specific duplex containing BcgI site. Following methylation, the duplex was digested to mononucleosides, mixed with unlabeled methyldeoxynucleosides, and separated by thin-layer chromatography. Two different solvents were used to distinguish all possible methylated bases (35). The spots, visible under ultraviolet light, were scraped from the TLC plates and counted for radioactivity. Greater than 95% of the radioactivity comigrated with m6A, confirming that N-6 methyladenine is the sole product of BcgI modification. lane number 1 2 3 4 5 6 7 8 9 A protein + + + ---+ + + B protein ---+ + + + + + specific duplex + --+ --+ -nonspecific duplex -+ --+ --+ -

. O P T K E I W F Y E H P Y P A G V K N Y S K T K P M K F E E F Q A E I D
Relative mobility' 1.3 1.1 1.0 0.9 3H incorporationd 1067 24 1119 28 Solvent G is 6633:l isobutyric acidwater:ammonium hydroxide, v/v; solvent D is 80:20 ethanol:water, v/v (45). *The methylated deoxynucleosides are as follows: m'C, N4-methyldeoxycytosine; m6C, 5-methyldeoxycytosine; m6A, N6-methyldeoxyadenosine.
Mobility shown is relative to thymidine. m6C + m' C effectively comigrate in solvent G; m6A + m' C comigrate in solvent D.
Data given in counts/min. Raw data is shown; background counts of 20 cpm have not been subtracted.

B A A+B
' H L"H L"H

L'
FIG. 5. BcgI subunit endonuclease assays. Partially purified BcgI B protein (lunes 1 and 2), BcgI A protein (lanes 3 and 4 ) , and BcgI A plus BcgI B proteins (lunes 5 and 6) were incubated with phage X DNA and then electrophoresed on an agarose gel and stained with ethidium bromide. Lanes marked H contain 3-fold more protein than lanes marked L. Only the reactions containing both proteins exhibit site-specific cleavage. 7. Individually, neither of the proteins exhibited site-specific cleavage activity, but the B subunit appears to degrade DNA in a nonspecific manner (Fig. 5), indicating it could have endonucleolytic domain. When the B was mixed with the A, specific DNA banding pattern appeared, suggesting that both subunits were required for site-specific DNA cleavage (Fig. 5).
DNA Binding Assays-The binding of the A and B proteins to DNA was examined by gel electrophoresis. 35S-labeled A, B, or A and B proteins were incubated with the unlabeled specific duplex containing a BcgI site or the nonspecific duplex that differed by 1 base pair in the absence of M e to prevent DNA cleavage and then were electrophoresed on a polyacrylamide gel, at pH 6.0, and visualized by autoradiograph. Alone, neither protein bound to the duplexes, but they did so together regardless of whether the duplex contains BcgI site or not (Fig. 6, lanes 7 and 8). Because binding can be influenced by isoelectric point (22), the samples were re-run at pH 8.2 and 10.0, but the results were unchanged; binding occurred only and A + B proteins (lanes 7-9, mixed after transcription/translation) were incubated with excess unlabeled DNA containing a BcgI site (specific duplex; lanes 1 , 4 , 7), lacking a BcgI site (nonspecific duplex; lanes 2,5,8), or with no DNA (lanes 3, 6,9) and then were electrophoresed and autoradiographed. Unbound protein did not enter the gel (because the gel buffer pH was 6.0). Only the mixture of proteins binds to DNA, and the binding is nonspecific.
when the proteins were present together. This suggests that not only does BcgI function as a complex but that it binds to DNA as a complex, too.
To confirm that BcgI binds to its recognition sequence nonspecifically, 32P-labeled specific or nonspecific duplexes were incubated with varying amounts of the purified, unlabeled BcgI enzyme. The duplexes were found to'be equally retarded, suggesting that binding was not sequence-specific (Fig. 7). Despite binding to the enzyme, the nonspecific duplex was resistant to BcgI digestion; the specific duplex was sensitive, as expected (not shown). To determine whether BcgI binds to its recognition site preferentially, the binding reactions were repeated in the presence of excess alternate unlabeled duplex (Fig. 7). Again, no differences were detected, suggesting that under conditions tested, BcgI does not bind to its recognition sequence preferentially. Similar observations have been reported previously by Taylor et al. for EcoRV endonuclease (40) and Zebala et al. for TaqI endonuclease (41).

DISCUSSION
We show here that BcgI is a bifunctional restriction and modification enzyme comprising two subunits of 637 and 341 amino acids in the proportion 2:1. The subunits are encoded by adjacent, similarly oriented genes; the larger 1914-base pair bcgIA gene lies upstream of the smaller 1026-base pair bcgIB gene. The two genes overlap by several nucleotides. Separately, neither gene expresses endonuclease or modification activity in vivo, but, together, they complement and express both activities. The large subunit (calculated PI = 5.07) and the small subunit (calculated PI = 9.66) copurify as a tight complex. Both subunits are required for specific endonuclease and methyltransferase activity. Neither subunit alone binds to DNA, but, together, they bind in a nonspecific manner. This suggests, therefore, that the formation of the A2B complex somehow changes the conformation of the proteins such that they could now bind to and cleave or methylate DNA containing its recognition sequence.
BcgI and it can be competed away equally well by excess specific or nonspecific oligos.
type I1 system; with respect to subunit composition, a type I11 system; with respect to recognition sequence, a type I system. However, it differs from each of these classes in important ways (42). Thus, type I1 enzymes act separately and recognize symmetric or continuous asymmetric (type 11s) DNA sequences; BcgI acts as a combined enzyme and recognizes an asymmetric discontinuous sequence. Type I systems recognize asymmetric, discontinuous sequences, but they cleave nonspecifically, and they comprise three subunits; BcgI cleaves specifically and comprises only two subunits. Type I11 enzymes comprise two subunits and recognize asymmetric, albeit continuous, sequences, but, like type I systems, they require ATP for cleavage, and they methylate in the absence of the R subunit; BcgI does not require ATP for cleavage, and neither of its subunits can methylate in the absence of the other. Finally, BcgI differs from another unique system, Eco57I (suggested as type IV in Refs. 10 and ll), which cleaves DNA on one side of its asymmetric recognition sequence with its bifunctional R subunit that also methylates one DNA strand, while its M subunit methylates both strands. Distinct from all type R-M systems, BcgI is the only characterized R-M system that cleaves bilaterally at a specific site with its unique A,B complex composed of two individually inactive subunits. Therefore, BcgI should be regarded as a new kind of R-M system. Because the BcgI subunits are catalytically inert on their own, assigning a function to each is problematic. The A subunit includes amino acid sequence motifs characteristic of m6A-methyltransferases, and, therefore, it doubtless contributes much, if not all, of the modification capacity of the complex. In the simplest scenario, the A subunit would catalyze modification and determine the DNA sequence specificity (A = M + S ) , and the B subunit would catalyze cleavage (B = R). This scenario is unlikely, because it implies that the A subunit should bind to DNA and methylate on its own, abilities that the A subunit does not possess. For similar reasons, it is unlikely that the A subunit catalyzes modification (A = M), and that the B subunit catalyzes cleavage and determines specificity (B = R + S ) . Considering that the complex cleaves on both sides of the recognition sequence and that the subunit stoichiometry is 2 A1 B, the A subunit might catalyze both methylation and cleavage (A = M + R), and the B subunit alone might determine specificity (B = S ) . That the B subunit does not bind to DNA separately could be explained by invoking a conformational change upon its association with the A subunit, but this explanation could equally well be applied to the other scenarios, too. Last, and, to our minds, most likely, in view of the discontinuous nature of the recognition sequence, it is possible that the A subunit catalyzes modification and contributes part of the specificity (A = M + S'), and the B subunit catalyzes restriction and contributes the rest of the specificity (B = R + S"). This model might explain why the B protein alone degrades DNA in a nonspecific manner (Fig. 5), and the A2B complex directs site-specific cleavage. It also accounts for the increased DNA binding affinity of BcgI when the A2B complex is formed (Fig. 6).
The unusual features of BcgI are (i) AdoMet is required for cleavage, (ii) the subunits are individually inactive, and (iii) cleavage is bilateral. What is the biological relevance of these features? The requirement for AdoMet might serve as a safety measure that couples restriction to modification. Thus, if the cell became undermodified for want of AdoMet, the capacity of the endonuclease to digest the cell's DNA is reduced. The interdependency of the subunits might also have safety aspects (the endonuclease is active only when the system is capable of modification) and it also promotes AdoMet conservation, since the methyltransferase is active only when the system is able to restrict. In addition, if the B subunit does determine specificity, the system is modular and has the potential to switch specificities by the acquisition of new B subunits. Finally, bilateral cleavage could prove to be a particularly effective means for restricting infectious DNA molecules, because it diminishes the opportunities for cleaved fragments to rejoin. Two more restriction endonucleases have been discovered recently that, like BcgI, require AdoMet and cleave bilaterally. They are CjeI and CjeII from Cumpylobacter jejuni.' Thus, BcgI-like systems may not be as rare as were once thought.
BcgI is a bifunctional enzyme that contains both cleavage and modification activities. The cleavage reaction occurs preferentially in vitro (17). What factor(s) and how it (they) is (are) involved in the regulation of BcgI cleavage and methylation activities in vivo is still a puzzle. The unique bilateral J. M. B. Vitor, R. D. Morgan, and I. Schildkraut (New England Biolabs), personal communication.
cleavage and other enzymatic properties may be potentially useful in manipulating DNA, for example, in the construction of overlapping clone libraries in genomic research (43) and in analyzing polymerase chain reaction products to detect mutations with a different number of identical nucleotides (44).