PROTEINS: Structure, Function, and Genetics 7:185-197 (1990)

Substrate Recognition by the EcoRI Endonuclease Joseph Heitman and Peter Model The Rockefeller University, New York, New York 10021

ABSTRACT The EcoRI restriction endonuclease is one of the most widely used tools for recombinant DNA manipulations. Because the EcoRI enzyme has been extremely well characterized biochemically and its structure is known at 3 A resolution as an enzyme-DNA complex, EcoRI also serves as a paradigm for other restriction enzymes and as an important model of DNA-protein interactions. To facilitate a genetic analysis of the EcoRI enzyme, we devised an in vivo DNA scission assay based on our finding that DNA double-strand breaks induce the Escherichia coli SOS response and thereby increase P-galactosidase expression from S0S::ZacZ gene fusions. By site-directed mutagenesis, 50 of 60 possible point mutations were generated at three amino acids (E144, R145, and R200) implicated in substrate recognition by the crystal structure. Although several of these mutant enzymes retain partial endonuclease activity, none are altered in substrate specificity in vivo or in vitro. These findings argue that, in addition to the hydrogen bond interactions revealed by the crystal structure, the EcoRI enzyme must make additional contacts to recognize its substrate. Key words: restriction enzyme, DNA-protein interactions, DNA recognition, enzyme-substrate interaction, active site, site-directed mutagenesis, protein structure-function, DNA double-strand breaks INTRODUCTION Our ability to manipulate and study DNA molecules depends upon a special class of enzymes, the restriction enzymes, which recognize and cleave within specific DNA sequences with exceedingly high fidelity.' We have studied the EcoRI restriction endonuclease as a model of DNA-protein interactions, to understand how these enzymes recognize their substrates with such precision, and with the idea that we might be able to design restriction enzymes with new specificities. The EcoRI endonuclease cleaves double-stranded DNA molecules at the sequence GAATTC and enables the host bacterial cell to destroy foreign DNA that enters the Its compatriot, the EcoRI methylase, protects the host chromosome from destruc0 1990 WILEY-LISS, INC.

tion by adding methyl groups to the recognition site.3 The EcoRI endonuclease is a symmetric homodimer of known sequence and requires Mg2+ as a cofactor for enzymatic activity.*x5 Reaction kinetic^,^,^ DNA binding proper tie^,',^ activity with oligonucleotides containing modified nucleoand stereochemistry of cleavage14 have been well described for EcoRI. They reveal that the enzyme initially binds DNA nonspecifically, diffuses in a one-dimensional search to locate its recognition site, and then binds DNA sequence specifically by contacting six base pairs within the major groove as well as the phosphodiester backbone that encompasses and flanks these bases. Within this complex, the structure of the enzyme-bound DNA is significantly distorted from B-form DNA.15,16 kine ti^,^ and gene ti^,'^^'^ eviCry~tallographic,'~ dence suggests that DNA scission by EcoRI is allosterically activated by conformational changes induced upon substrate binding. The two DNA strands are cleaved sequentially, and whereas release of the doubly cleaved product is often the rate-limiting step, for some substrates and reaction conditions the enzyme releases the singly cleaved intermediate and cuts the other strand after a second binding event." DNA scission proceeds with inversion of configuration a t the scissile phosphodiester bond; this is indicative of an odd number of reaction steps and is consistent with cleavage by a hydroxyl ion rather than through a covalent enzymeDNA intermediate.14 The crystal structure of an EcoRI endonucleaseDNA ~ o m p l e x ' ~(solved . ~ ~ to 3 A resolution in the absence of Mg2+ to prevent cleavage) demonstrated that the EcoRI enzyme is not structurally homologous to either the h e l i x - t ~ r n - h e l i xo~r ~the zinc finger21 class of DNA-binding proteins. Instead, two a-helices from each subunit of the symmetric dimer project steeply into the major groove of the DNA and carry amino acids that hydrogen bond to the edges of the nucleotides of the substrate (shown in Fig. 1). Glutamic acid 144 and arginine 145 lie at the end of

Received June 19,1989; revision accepted October 19, 1989. Joseph Heitman's present address is Department of Biochemistry, The Biozentrum, 70 Klingelbergstrasse, CH-4056 Basel, Switzerland. Address reprint requests to Peter Model, The Rockefeller University, 1230 York Ave., New York, NY 10021.

186

J. HEITMAN AND P. MODEL

C

9p

N7

I

'

i rI

-"+\ 4,"

Arg 200a

Fig. 1. Crystal structure and substrate interaction model for the EcoRl endonuclease. The 3 A resolution X-ray crystal structure of an EcoRl endonuclease-DNA complex is shown on the left. One monomer of the symmetric homodimer is omitted for clarity. Each monomer bears two a-helices (the inner and outer helices) that project steeply into the major groove of the DNA

and deliver amino acids that bind the DNA. The three amino acids (glu 144, arg 145, and arg 200) predicted to interact with the nucleotides of the recognition site are indicated by arrows. On the right is a detailed model of the hydrogen bonds between the protein and the DNA (see text). The crystal structure and model shown here are adapted from McClarin et aI.l5

the inner a-helix, and each, from opposite subunits, spans and makes two hydrogen bonds with the central adenines (gAAttc). Arginine 200 lies at the end of the outer a-helix and makes two hydrogen bonds to the outer guanine nucleotide (Gaattc). Contacts to the right and left halves of the palindromic substrate are symmetric. Thus, for the dimer, a total of 12 hydrogen bonds are predicted to form the substrate binding pocket. Because any nucleotide substitution within the recognition site disrupts one or more of these interactions, this hydrogen bond network could enable the enzyme to recognize uniquely this DNA sequence. One test of this model is to isolate mutants altered a t the amino acids implicated in substrate recognition. Previously, conservative amino acids substitutions (E144D, R145K, R200K) were shown to reduce enzymatic activity without altering substrate s p e c i f i ~ i t y . To ~ ~ test , ~ ~ whether less conservative amino acid substitutions would alter or disrupt the substrate specificity of the enzyme, we developed an in vivo endonuclease assay based on our finding that EcoRI DNA breaks induce the SOS DNA repair re~ponse.'~-'~ SOS induction is conveniently monitored with strains in which the lactose operon has been fused to a DNA damage-inducible promoter.27

Here we employ this assay to assess the phenotypes of EcoRI mutants altered in the amino acids predicted to bind the substrate. In addition, several mutant enzymes that are active in vivo were purified to homogeneity and studied in vitro. In vivo and in vitro, these mutant proteins are reduced in specific activity and retain their specificity for the wild-type substrate. Therefore, the hydrogen bond network revealed by the crystal structure is insufficient to account fully for substrate recognition, and additional amino acids must contact the DNA to help discern the substrate. Similar findings have been reported for a collection of site-directed mutants altered at R200.28

MATERIALS AND METHODS Bacterial Strains and Methods Bacterial strains used include K38 [ =Hf& phoA(am)(X)], K902 ( = K38 recA56 srZ300::TnlO supE), JH137 [ = K38 AEacZ dinDl ::Mu dI(Ap'Zac)], JH138 ( = JH137 rep71 iZuY864::TnlO), JH69 [ =HB101 recA dinDl ::Mu dI(Ap' lac)], JH153 (FendAl thil hsdRl7 supE44 cZ' Agal nadA::TnlO PL-lac TI1 AcZI-uurB =MM294 with the defective h lysogen of strain AR120; see below), and the dut ung strain K1053 (=BW313 30).24-26 The dinD1::Mu dI +

SUBSTRATE RECOGNITION BY EcoRI ENDONUCLEASE

(Ap' lac) fusion allows one to monitor SOS induction by measuring p-galactosidase activity. In some cases, p-galactosidase activity was assayed by growing colonies on X-gal indicator plates (35 kg/ml) and scoring for blue color intensity. For quantitative pgalactosidase assays, bacteria were grown in K120 minimal media2* supplemented with 0.2% glucose and 0.4% casamino acids and assayed as described by Miller."

Site-Directed Mutagenesis Plasmids expressing the EcoRI endonuclease (pJH15a/b) and methylase (pJC1) have been d e s ~ r i b e d . ~Plasmid ~ , ~ ~ , pJH15a/b ~~ can be separated from plasmid pJCl by cleavage with BamHI endonuclease (cleaves p J C l once and does not cleave pJH15aib) and transformation. Plasmids pJH15a and pJH15b carry the fl intergenic region in opposite orientations and can be packaged into defective fl transducing particles following fl helper phage infection. In some cases, these transducing particles were isolated free of contaminating helper phage by transduction into a host strain (JH138, rep-), which does not support fl replication. Site-directed mutagenesis was by the Kunkel method,30 in which uracil is partially incorporated into the single-stranded template DNA (by growth in a dut- ung- host), second strand synthesis is primed with a mutagenic oligonucleotide, and the template strand is then selectively destroyed (by growth in a dut' ung+ host) to increase the yield of mutants (typically 70-80%). For this purpose, the endonuclease bearing plasmic pJH15a was rendered single stranded by infecting the dut ung host strain K1053/pJH15a and pJCl with fl helper phage R176.31 Reaction conditions were as described previously, except that in some cases separate annealing reactions were carried out at several different temperatures (room temperature, 45"C, and 65°C). As mutagens, we employed degenerate oligonucleotides in which the codon of the mutagenized amino acid is replaced by the mixture 5'-NNX-3', (where N is a equal mixture of all four bases and X is a mixture of G and C), which is 32-fold degenerate and encompasses a t least one codon for each amino acid. Specific mutagens were: 5'-CTTATGAGATCTXNNGATAGCATTACC (E144X), 5'-CTTATGAGAXNNTTCGATAGC (R145X), and 5'-GTCGATCTAAXNNATTTAATATACC (R200X). For the R200 position, several additional oligonucleotides were used that, instead of XNN, bore CGC (R200A), CTX (RZOOEQ),ATX (RBOOHD),or AA/TA/T (RZOOFYIN) as the mutagenic codon. Strains JH137 (without the EcoRI methylase) and JH137/pJCl (with the methylase) were transformed with plasmid DNA mutagenized as above. Reducedactivity endonuclease mutants were obtained either 1)in strain JH137 by selecting for colony formation in the absence of the EcoRI methylase or 2) by trans-

187

forming strain JH137/pJCl with mutagenized DNA, sequencing random isolates to identify mutants (70-80%) and subsequently assessing the mutant phenotype in the absence of the methylase. The second screen ensured that mutants lethal to the host in the absence of the methylase would not be missed. Notably, we found that mutants with substantial endonuclease activity (R200K, E144D) could be stably isolated only in the presence of the EcoRI methylase. Single-stranded DNA templates isolated after fl helper phage infection were sequenced by the dideoxynucleotide method using as primer either 5'GATTACTATTATAGGC (for R200X) or 5'-GTGGCTCTCAGGAGAGC (for E144XIR145X). In general, 150-200 base pairs spanning the mutated site were sequenced (roughly one-quarter of the endonuclease gene). When mutagenic oligonucleotides had been annealed at 65"C, mutations most commonly resulted from single base pair mismatches. With lower annealing temperatures, a greater proportion of mutants arose from double and triple mismatches. Among approximately 250 mutants sequenced, no extraneous silent mutations were observed. We established that these site-directed lesions confer the observed phenotypes by the following criteria. The entire gene was sequenced for two mutants (R200K and R145K). For the majority of the remaining mutants, independent isolates, obtained either in the absence or the presence of the EcoRI methylase, did not differ in phenotype. In the only case where two mutants with the same site-directed mutation (R145K) differed in phenotype, sequencing the entire gene of both mutants revealed that the more active isolate carried a spontaneous suppressor that increased the activity of the R145K mutant. Finally, uracil containing templates of each mutant were subjected to site-directed reversion with oligonucleotides bearing the wild-type (WT) codon. Following site-directed reversion, plasmid DNA was introduced into a host strain expressing the EcoRI methylase (either strain JH137lpJC1 or JH138/ pJC1). Because none of the site-directed mutants restrict phage A, revertants could be readily detected by cross streaking colonies harboring putative revertants across a line of A-vir phage spread on an agar plate; growth beyond the line of phage indicates that restriction occured. With high frequency (25-75%), restriction of phage was restored. For at least six R' isolates from each reversion reaction, restriction of phage A-vir was determined quantitatively by spotting portions of serially diluted phage on a lawn of the strain to be tested. Reversion restored the WT level of phage restriction plaques/phage) for every mutant tested.

Trans-Dominant In Vivo Inhibition Assay To determine if the mutant endonucleases inhibit the action of the WT enzyme, both the WT EcoRI

188

J. HEITMAN AND P. MODEL

restriction system (endonuclease and methylase) and a site-directed EcoRI mutant were coexpressed from the compatible plasmids pJH97 and pJH15a in the recA- host K902, and phage restriction was measured. Plasmid pJH97 (Cm', EcoRI R + M + )is a pACYC184 derivative that bears the WT EcoRI endonuclease and methylase genes (the PuuII to CZaI fragment of pAN4 was inserted into the pACYC184 BarnHI site with BarnHI linkers). Plasmid pJH15a (Kan' and Amp', EcoRI R+M-, pBR322 derivat i ~ ecarries ) ~ ~ the EcoRI endonuclease site-directed alleles.Cultures were grown to late logarithmic phase (OD,,, = 1-21, mixed with A-vir phage, and the phage titer was determined after overnight incubation. The decrease in A restriction was calculated by assaying the titer on strain K902/pJH97 expressing a site-directed mutant and dividing by the titer on the control strain K902ipJH97 bearing a pJH15a EcoRI- deletion (pJH72) that removes the N-terminal three-quarters of the endonuclease gene. The values in Table I11 are the average of four to eight independent determinations.

Protein Purification and Enzyme Assays The EcoRI site-directed mutants were overexpressed from the strong X promoter P, and purified. The mutant endonuclease genes were fused to the P, promoter in two cloning steps. First, the mutant genes were rejoined to the EcoRI methylase gene by subcloning from plasmid pJH15a into plasmid PAN^^ (the large PstI fragment of pJH15a was ligated to the small PstI fragment of pAN4). Second, by ligating appropriate BgZII restriction fragments, the EcoRI site-directed mutants were transferred from plasmid pAN4 to plasmid pJH89, a derivative of plasmid pAN4 that bears the EcoRI restriction system fused to the A P, promoter. Plasmid pJH89 itself was constructed by partially digesting plasmid pAN4 with NdeI (there are two NdeI sites in pAN4), purifying full-length linear molecules, filling in with Klenow polymerase, and blunt-end ligating to a filled-in BgZII-HpaII fragment carrying A PL(from plasmid P K C ~ O ~From ~ ) . this procedure, one isolate was chosen (pJH89) in which A P, was inserted at the NdeI site immediately upstream of the EcoRI genes in the proper orientation and in which the BgZIl site flanking the PL fragment was not recreated. The overproduction host strain JH153 was constructed by introducing the defective A lysogen of strain AR12033 into the e n d - strain MM294 by transduction with phage P1 and selection for tetracycline resistance conferred by the linked nadA:: TnlO marker. One liter cultures of strain JH153/ pJH89 were grown in FB at 30°C to OD,oo=l, induced with 160 pg/ml of nalidixic acid,34 and grown for 12 hr. The WT and mutant EcoRI enzymes were purified to homogeneity (single band on coomassie blue-stained SDS-polyacrylamide gel) as described previously for the WT enzyme.35

The specific activities of the wild-type, the RZOOK, and the R200C enzymes were determined from partial restriction digests at several concentrations of plasmid pUC18 DNA. Negatives of EtBr-stained agarose gels (Polaroid 665 film) were scanned with a Joyce-Loebl microdensitometer to determine the initial rate of product formation.

RESULTS In Vivo DNA Scission Assay The EcoRI restriction-modification system consists of the EcoRI endonuclease and methylase. Normally the genes for the two proteins lie adjacent to one another and may consititute a n o p e r ~ n .To ~.~ mutagenize and express the endonuclease and methylase independently, the two genes were previously cloned on separate compatible plasm id^.'^^^^,^^ Plasmids pJH15a and pJCl express the EcoRI endonuclease and methylase, respectively. Methylation protects the cellular DNA from EcoRI cleavage and therefore expression of the wild-type (WT) endonuclease from plasmid pJH15a is lethal to cells that lack the EcoRI methylase. Endonuclease mutants with reduced or null activity can be isolated by selecting colonies that survive transformation by the endonuclease plasmid pJH15a.25,26 To facilitate a genetic analysis of the EcoRI restriction enzyme, we developed a n assay for in vivo endonuclease activity, which is shown in Figure 2. Earlier we found that DNA scission induces the E. coZi SOS DNA repair re~ponse.'~-'~ SOS induction is conveniently monitored with strains that bear the lactose operon fused to DNA damage inducible (din) After DNA damage, these strains make pgalactosidase, which can be assayed either qualitatively (on X-gal indicator medium) or quantitatively (by ONPG cleavage with permeabilized cells).

EcoRI Mutants Altered at Amino Acids 144, 145, or 200 Retain Activity As is shown in Figure 1,the crystal structure of a n EcoRI-DNA complex revealed that glutamic acid 144, arginine 145, and arginine 200 form hydrogen bonds with the DNA, which may mediate substrate recognition. By site-directed mutagenesis using multiply degenerate oligonucleotides and uracil containing templates (the Kunkel method3'), we isolated many substitutions at amino acids 144, 145, and 200 (see Materials and Methods). The phenotypes of these mutants were determined as follows. First, cells that express the WT EcoRI endonuclease in the absence of the methylase suffer DNA degradation and die. Each of the site-directed mutants was tested for a similar effect on cell viability. As a second assay of endonuclease activity, we measured the ability of each mutant enzyme to induce the SOS response, a sensitive indicator of in vivo DNA s c i ~ s i o n . For ~ ~ -both ~ ~ assays, we employed two host strains, JH69 and JH137, which 1)carry the lactose

189

SUBSTRATE RECOGNITION BY EcoRI ENDONUCLEASE

0 E. coli

Chromosome

@ACHl

1

PdinD

all

+

DNA Scission Induces SOS

lac2 I

Fig. 2. In vivo DNA scission assay. When the E. colf chromosome suffers DNA scission, the SOS DNA repair response is SOS induction is a cascade whereby damaged DNA activates the RecA protein to stimulate the autodigestion of LexA, the transcriptional repressor of the SOS genes. When the lac2 gene is fused to an SOS promoter, DNA damage activates

(3-galactosidase expression to yield blue colonies on X-gal indicator media. EcoRl mutants that retain endonuclease activity induce the SOS response. If expression of the EcoRl methylase blocksthis SOS induction,the mutant endonuclease retains specificity for the wild-type recognition site. TS, temperature sensitive; SD, site-directed; alt, altered substrate specificity.

operon fused to the DNA damage-inducible locus dinD (for measuring SOS induction, see references 24-27) and 2) differ in sensitivity to EcoRI DNA scission (strain JH69, an HBlOl derivative, is less sensitive to EcoRI scission than strain JH137). Finally, cell viability and SOS induction were measured in the presence and absence of the EcoRI methylase (expressed from plasmid pJC 1) to determine the specificity of DNA scission by the mutant endonucleases. The phenotypes of 50 of the 60 possible point mutants a t residues 144, 145, and 200 are listed in Tables I and I1 and summarized in Figure 3. Like the WT EcoRI enzyme, the three most active mutant proteins (E144D, E144C, and R200K) are lethal to strain JH137 in the absence of the protective methylase (Table I). In these cases, the mutants were initially isolated in strain JH137 expressing the EcoRI methylase. When plasmid DNA encoding these mutants was introduced into strain JH137 lacking the methylase, transformants could be obtained at low efficiency but only when grown at 42°C. Moreover, the resulting colonies were induced for the SOS response, exhibited cold-sensitive growth, and grew extremely poorly, giving rise to faster growing variants. We conclude that these alleles are lethal to strain JH137, most likely as a result of endonuclease action. These EcoRI mutants could be expressed under some conditions in a strain that is more resistant to DNA scission (JH69). In strain JH69, these mutants again exhibited a temperature-sensitive phenotype, whereby the mutant

enzyme is less active at higher temperature and can be expressed whereas a t lower temperatures the host strain grows more poorly or dies. At intermediate temperature conditions, where strain JH69 expresses a mutant endonuclease but remains viable, the SOS response is induced (Table 11).Because the WT EcoRI endonuclease is fully lethal to either strain JH69 or JH137 at all temperatures, we conclude that the endonuclease activity of the E144D, E144C, and R200K mutant enzymes is reduced compared to the WT enzyme. None of the remaining mutant alleles was lethal to strain JH137 or JH69. However, several additional alleles at each position do retain endonuclease activity as determined by the more sensitive in vivo SOS induction assay (Tables I, 11, Fig. 3). Colonies of the S0S::lacZ fusion strain JH137 expressing several of these EcoRI mutants are shown in Figure 4. The medium contains X-gal, a chromogenic substrate that yields a blue dye when cleaved by P-galactosidase. In that the LacZ gene is now controlled by a DNA damage-inducible promoter, a blue colony color indicates that the SOS response is induced, in this case by in vivo DNA scission. Cells expressing EcoRI mutants bearing cysteine, serine, or valine at position 200 produce blue colonies. Because these mutant enzymes induce the SOS response, we conclude that they cleave DNA in vivo. By the same criteria, the serine and glycine mutants a t position 144 and the lysine and cysteine substitutions a t position 145 display some

190

J . HEITMAN AND P. MODEL

TABLE I. SOS Induction by E144X, R145X, and R200X EcoRI Mutants in Strain JH1371:

MAllele WT R200K R200C R200V R200S R200Xfi E 144D E144C E144S E144G E144Xfi R145K R145C R145XO

Phenotype Lethal wlo M Lethal wio M* TS TS TS Null Lethal wlo M* TS, lethal TS TS Null TS TS Null

42 Dead Dead LB (25) W-LB (20) W-LB (12) W Dead DB (240) W-LB (24)

w (9)

W W-LB (28) W (13) W

37 Dead Dead MB LB W-LB W Dead DB, sick (350) DB, sick W-LB W LB W-LB W

34 Dead Dead MB L-MB LB W Dead Dead DB, sick W-LB W LB W-LB W

30 Dead Dead DB, sick (100) MB (83) LB (41) W Dead Dead DB, sick (130) W-LB (25) W LB (61) W-LB (30)

W

M' 30-42" w-LBI W W W W W W W W W W

W W W

$Indicator plates contained 35 pgiml X-gal. Under these conditions, color intensity corresponded to the following units of P-galactosidase activity: W, -10 units; W-LB, -10-30 units; LB, 25-60 units; MB, 60-100 units; and DB, 100+ units. In some cases, units of P-galactosidase activity determined by the ONPG assay are listed here in parentheses. *The R200K and E144D alleles are lethal without methylase in strain JH137 but viable at high temperature in the HB101-derived strain JH69. In strain JH69, both alleles are temperature sensitive (see Table 11). t W = white or faint blue; LB = light blue; MB = medium blue; DB = dark blue. $The following null alleles (R200X, E144X, R145X) showed no induction of the SOS response: R200P, T, N, W, L, G, M, A, E, Q , D, H, I, F, and Y; E144K, T, A, R, V, L, W, M, F, Y, N, and P (E144Q, I, and H not tested); and R145E, F, M, D, S, T, I, G , N, A, and UAG (amber) (R145P, W, L, V, Q, H, and Y not tested).

TABLE 11. SOS Induction by EcoRI Mutants in Strain JH69* M-

Allele WT R200K R200C E144D E144C

Phenotype Lethal wlo M TS, lethal TS TS TS, lethal

42 Dead LB-MB LB DB LB-MB

37 Dead DB, sick LB DB, sick DB, sick

34 Dead Dead LB DB, sick DB, sick

30 Dead Dead LB-MB DB, sick Dead

M+ 30-42°C W-LBt W-LB W-LB W-LB W-LB

*The background level of lac expression is higher in strain JH69 (lacZ+ LUCY-)than in JH137 (AlacZ).Indicator plates contained 35 pgiml X-Gal. $W, white or faint blue; LB, light blue; MB, medium blue; DB, dark blue.

endonuclease activity. Every other amino acid substitution at position 200 and many other changes at positions 144 and 145 induced no more P-galactosidase expression than an EcoRI nonsense mutant and hence result in a null phenotype. Furthermore, double mutants constructed from the most active single mutants (E144D RZOOK, E144D + RZOOC, E144C + R200K, and E144C + R200C) were completely inactive. In summary, at each position every mutation reduces activity compared to the WT amino acid. The most conservative amino acid substitutions (E144D, R145K, R200K) exhibit greater activity than other changes, but several less conservative changes are also partially active (RZOOC, V, and S; E144C, S, and G; and R145C). Furthermore, all the active mutants display a temperature-sensitive phenotype (see Tables I, 11). Since altering these three amino acids cripples enzyme activity, these residues play an important but not essential role for enzyme action.

+

Site-Directed Mutants Retain WT Substrate Specificity As an in vivo measure of cleavage specificity, the endonuclease activity of the mutant enzymes was assayed in the presence of the EcoRI methylase. When the WT EcoRI endonuclease and methylase were expressed together (from plasmids pJH15a and pJCl), both strains JH137 and JH69 were viable, and the SOS response was not induced (white colonies on X-gal indicator medium). Thus expression of the EcoRI methylase effectively counteracts the lethal effect of the WT endonuclease. Similarly, the lethal and SOS-inducing phenotypes of all the active site-directed EcoRI mutants were completely blocked when the EcoRI methylase plasmid was present (Tables I, I1 and Fig. 4). We conclude that these mutant enzymes recognize and cleave the WT EcoRI recognition site. As a control, we checked that the EcoRI methylase

191

SUBSTRATE RECOGNITION BY EcoRI ENDONUCLEASE

144

145

200

Wild-type

lethal

strong

increasing activity

ser

CYS

SOS Induction

val ser

decreasing

I

inactive

~~

lYS thr ala arg val leu

trp met phe tyr asn pro

glu phe met asp ser

thr ile gly asn ala

trp pro leu gly met thr ala glu

gln asp asn his ile phe tyr

Fig. 3. Summary of phenotypes of site-directed EcoRl mutants. To assay endonuclease activity in vivo, the site-directed EcoRl mutants were expressed in strain JH137 in the absence of

cleavage assays as de~cribed.’~ Mutants with a high level of activity are, like the wild-type enzyme, lethal to host strain JH137. SOS induction was scored as strong (loo+ units p-glactosidase), moderate (60-100 units), or weak (10-60 units). Mutants that induce no more p-galactosidase expression (10 units) than an EcoRl nonsense mutant (R145 UAG) were scored as null mutants.

did not inhibit SOS induction by other DNA damaging agents such as nalidixic acid or mitomycin C (data not shown). One further consideration was that a n altered specificity of the mutants might be blocked if the EcoRI methylase can modify sites other than GAATTC in vivo, as has been described for certain in vitro condition^.^^,^^ However, we find no evidence for this methylase star activity in vivo. Neither a plasmid expressing the EcoRI methylase (pJC1) nor fl phage (RFI form) grown on a host producing the EcoRI methylase was protected from EcoRI* cleavage activity in vitro (data not shown). In addition, several EcoRI endonuclease mutants that clearly induce the SOS response and cleave DNA in the presence of the EcoRI methylase have been isolated (Heitman and Model, submitted). Thus, if the site-directed mutants were cleaving sites other than the WT substrate, this assay should have detected it. We conclude that these sitedirected EcoRI mutants are not dramatically altered in substrate recognition. Because the activity of these mutants is blocked by the EcoRI methylase, we can also rule out the possibility that their phe-

notypes (lethality, SOS induction) arise from nonspecific effects, such a s precipitation or denaturation of the mutant enzymes.

the EcoRl methylase. Strains were grown at 30°C on X-gal indicator medium (35 pglml) to monitor increased p-galactosidase expression from the dinD7::lacZ fusion as a measure of SOS induction. p-Galactosidase activity was quantified by ONPG

Site-Directed Mutants Do Not Restrict A As a further in vivo measure of the mutant enzymes activity, their competence to restrict phage growth was assessed. The WT EcoRI restriction system (R’M’) decreased the efficiency of A-vir plaque formation to lOp4/phage. In contrast, A-vir phage plated with unit efficiency on a strain (JH137) expressing both a mutant EcoRI endonuclease and the methylase. Thus, although some site-directed EcoRI mutants have sufficient activity to kill the host cell in the absence of the methylase, none restrict h in a host expressing the methylase. By this measure, we again conclude that these mutants are decreased in activity compared to the WT enzyme. Site-Directed Mutants Inhibit the WT Enzyme in Trans Using the same A restriction assay, we asked if expression of a mutant endonuclease would behave

192

J. HEITMAN AND P. MODEL

Fig. 4. Site-directed EcoRl mutants induce p-galactosidase expression from an S0S::lacZ fusion. These colonies express site-directed EcoRl mutants: clockwise from the top are the E144S, R145K, R200C, R200V, R200S, and R200W mutants. The media contains t h e chromogenic substrate X-gal and the host strain (JH137) bears an S0S::lacZfusion. The strains in the left half lack the EcoRl methylase, and, because the wild-type endonuclease would belethal in this case, these mutant endonu-

cleases are reduced in activity. Several mutants (R200S, R200V, R200C, R145K, and E144S) induce the SOS response and therefore retain some endonuclease activity. The R200W mutant is a null mutant that induces no more p-galactosidase expression than an EcoRl nonsense mutant. The right half shows that coexpression of t h e EcoRl methylase protects the cell from DNA scission by the mutant endonucleases (no SOS induction).

as a codominant negative allele3' and inhibit the WT enzyme. For this purpose, the WT EcoRI restriction-modification system was expressed from plasmid pJH97, a pACYC184 derivative compatible with the plasmid that bears the site-directed EcoRI endonuclease mutants (pJH15a). Cells harboring both plasmids were tested for their ability to restrict the growth of phage A-vir. As controls, we showed that plasmids bearing an amber mutant (R145UAG) or a deletion derivative of EcoRI did not significantly interfere with restriction by the WT enzyme (see Table 111).In contrast, as is shown in Table 111, all the site-directed mutants decrease restriction of A by the WT enzyme from 20- to 1,500-fold. There are two possible explanations for this effect: the mutants act either on the target DNA or on the WT enzyme. The purified R200K and R200C mutant proteins do not inhibit DNA cleavage by the WT enzyme in vitro (data not shown), suggesting that this inhibition is not by competition for substrate. Alternatively, the WT and mutant proteins could form defective heterodimers. The E144K mutant is known to be defective in dimerization in ~ i t r o . ~ One ' might expect this mutant to be a monomer in vivo and therefore not to inhibit the WT enzyme in trans. Unexpectedly, the E144K mutant (as well as the similar E144R mutant) inhibits restriction to a greater degree than the other mutants. We suggest

that, because the E144K mutant self-dimerizes poorly, i t presents a larger pool of monomer subunits to compete with WT monomers for dimerization. Either this larger pool compensates for the dimerization defect or the mutant is less impaired to dimerize with a WT partner than with itself. Since the E144, R145, and R200 residues form salt bridges that stabilize the dimer interface,l5z3' all these mutants may be partially defective in dimerization. This could contribute to the magnitude of the observed inhibition. From this in vivo assay, we infer that all the mutant proteins are stably synthesized and folded properly and that they inhibit the WT enzyme by forming defective heterodimers. This assertion is further supported by pulse-labeling and immunoprecipitation assays in which none of the mutants tested (R200K, R200M, R200P) was reduced i n stability compared to the WT protein (data not shown). Codominant negative inhibitory alleles have been described for the A, trp, and lac repressors and attributed to defective heterodimer f o r m a t i ~ n . ~ ' . ~ ~ This has been demonstrated directly for mutants of the trp r e p r e ~ s o r . ~ '

Purification of Mutant Proteins The in vivo properties of the EcoRI site-directed mutants suggest that several retain partial endonuclease activity of WT substrate specificity. To test

SUBSTRATE RECOGNITION BY EcoRI ENDONUCLEASE

TABLE 111. EcoRI Mutants Inhibit Restriction b y the Wild-Type Enzyme Fold decreased 1-4 20-50 50-150 150-300

500-1500

144 D,Y restf T,P K,R

Amino acid 145 (UAG)stop all*

Efficiency 200

of restriction

10-4 all? 10-2 10-1

*All other 145 = KCEFMDSTIGNA. iAll other 200 = all amino acids except R (wild-type). $Rest of 144 = AVLWMFNCSG.

this, the WT and the R200K and R200C mutant enzymes were purified to homogeneity and studied further in vitro. The proteins were first overproduced from the strong A promoter P,. Expression from P, is usually induced by thermal denaturation of a temperature-sensitive c l repressor (~1857). Because these EcoRI mutants are temperature-sensitive in vivo, cultures were grown at 30°C and expression from P, was instead induced with nalidixic acid.34 For this purpose, the host strain JH153 was constructed with two features: first, an endA- mutation to reduce the level of contaminating nonspecific endonuclease and, second, a defective A lysogen bearing a WT cl repressor gene. This is important because most cI857 alleles also carry the ind- mutation, which prevents induction by nalidixic acid. The WT and the R200K and R200C mutant endonucleases were purified as described previously for the WT enzyme.35

DNA Cleavage by WT and Mutant EcoRI Endonucleases From restriction digests, the R200K and R200C mutant proteins cleave substrates (A, pBR322, pUC18) at the canonical recognition site (GAATTC) with specific activity reduced by three to four orders of magnitude compared with the WT enzyme (see Table IV). For example, Figure 5 shows that plasmid pUC18 DNA (one EcoRI site) is cleaved by the WT enzyme and the R200K and R200C mutant proteins to produce a single linear DNA species. To determine the site at which cleavage occurs, plasmid pUC18 was treated with EcoRI (WT, or R200K or R200C mutant enzyme) and a second enzyme (either SmaI or XrnnI) that cleaves one additional site to yield two fragments of diagnostic lengths. Under these conditions, both the R200K and the R200C mutant enzymes yielded two fragments that comigrated with those observed with the WT enzyme (shown in Fig. 5 for EcoRI star buffer conditions). Furthermore, substrates lacking EcoRI sites (fl, pBR322ARI) were not cleaved a t detectable levels by the mutant enzymes, even under conditions in

193

which significant nicking by the WT enzyme occurs (data not shown). We conclude that the EcoRI mutant enzymes retain WT recognition specificity in vitro as was observed in vivo. In standard EcoRI buffer, the WT EcoRI enzyme cleaves its substrate (GAATTC) -lo7-fold faster than any other DNA sequence.43 However, in star buffer (higher pH, low salt, glycerol) EcoRI cleaves additional sites usually related to the canonical recognition site by one or more nucleotide substitutions, a phenomenon called EcoRI" a ~ t i v i t y . As ~~-~~ is shown in Figure 5, EcoRI star buffer strongly enhances DNA scission by the R200K mutant protein. By titering the extent to which A phage DNA was digested by serial dilutions of the enzyme, we find that the R200K enzyme is activated 100-fold by EcoRI* buffer compared to the standard EcoRI buffer. DNA scission by the R200C mutant was weakly activated by star buffer (five-fold). For both the R200K and R200C mutant enzymes, DNA cleavage under star buffer conditions was still exclusively at the WT EcoRI recognition site (Fig. 5). When incubated in standard EcoRI or EcoRI* buffer, both the R200K and the R200C mutant proteins produce a higher proportion of the nicked product than does the WT enzyme. Compared to the WT enzyme, these mutant proteins may dissociate more readily from the nicked intermediate. From a Michaelis-Menton-type analysis of DNA scission, the cleavage defect of the R200K mutant enzyme was shown to be largely attributable to a decreased V,, (see Table IV). That the K, of the mutant enzyme is unaltered does not necessarily indicate that DNA binding has not been affected, because K, for the EcoRI enzyme is not strictly related to the DNA binding constant. In a separate study, several EcoRI mutants with enhanced star activity were isolated (Heitman and Model, in preparation). To test if these enhanced star activity mutations would activate cleavage by the R200K and R200C mutants as do star buffer conditions, the mutants were recombined together in vitro. By monitoring SOS induction, we find that in vivo DNA scission by the R200C mutant is increased by a n additional mutation that enhances the star activity of the WT enzyme. The EcoRI methylase blocks this increased DNA scission. Therefore, DNA scission still occurs at only the WT substrate. In contrast to the R200C mutant, the R200K mutant is not activated by the star mutations, even though it is strongly activated by EcoRI" buffer in vitro. Because the lysine and cysteine containing mutants are affected differently by EcoRI" buffer and mutations that enhance star activity, they may not recognize the substrate in the same way. For example, the R200K mutant may make one or both of the hydrogen bonds like the arginine of the WT enzyme, whereas the R200C mutant may make only one or neither.

194

J. HEITMAN AND P. MODEL

TABLE IV. Specific Activity and Kinetic Parameters of Enzyme Activity Allele Wild-type R200K R200C

EcoRI buffer

Star buffer ND 7 x 104-1.1 x 2-4 x lo3

1-2 x IO'Uimg 9 x 102-1.8 x lo3

2-4 x 10'

lo5

K, (nm) 3.8 2.7 ND

V,,,

(min-l) 7.2 0.4* ND ~~

RFI1:RFIII -1:4 -2:l

-2:l ~~~

*For the RFII product under conditions of maximal cleavage activity (25 mM Tris C1, pH 8.5, 2 mM MgCl,, 5% glycerol). ND, not determined.

Enzyme: - R R R Second Enzyme: - Buffer:

sx-

R200K -

- R R R R *

R200C

sx* *

-

R *

sx * *

RFIHRFI-

Lane:

1 2 3 4 5 6 7 8 9 1011 12

Fig. 5. The R200K and R200C mutant EcoRl endonucleases cleave DNA at EcoRl sites. Restriction digests with plasmid pUC18 demonstrate the in vitro cleavage specificity of these mutant enzymes; 500 ng of CsCI-purified RFI form plasmid DNA was incubated with the purified wild-type (15 ng = 30 units) or the mutant endonucleases (R200K 15 ng; R200C, 5 ng) for 1 hr at 30°C in standard EcoRl buffer conditions (lanes 2, 5, and 9; 100 mM Tris HCI, pH 7.5,5 mM MgCI,, 50 mM NaCI, 100 Fgiml BSA) or EcoRI' buffer conditions (lanes 6-8, 10-12; 25 mM Tris HCI, pH 8.5, 2 mM MgCI,, 5% glycerol). In some reactions, a second enzyme was included to cleave the plasmid at one additional site and provide a point of reference (5 units of Smal in lanes 3, 7, 11; 5 units of Xmnl in lanes 4, 8, 12). The reaction products were displayed on a 0.6% agarose gel containing 0.5 Fg/ml ethidium bromide. R, EcoRI; S, Smal; X, Xmnl.

DISCUSSION To determine how the EcoRI restriction enzyme recognizes its substrate, we mutated those amino acids that the EcoRI crystal structure implicated in substrate binding. The resulting mutant enzymes are crippled in enzyme activity but are not altered in substrate specificity in vivo or in vitro. Since the activity of these mutants is dramatically reduced compared to the WT enzyme, these amino acids clearly play a n important role in enzyme action as predicted by the crystal structure. It is perhaps not surprising that the most active mutant at each residue bears the most conservative amino acid change (E144D, R145K, and R200K). These mutant enzymes may make similar but weaker contacts than the WT protein (see also references 22, 23, 28). We also find that several mutants with nonconservative amino acid replacements retain partial enzyme activity and still recognize the WT substrate (R200C, V, or S; E144C, S, or G; and R145C). Strikingly, at each of the three mutated residues the cysteine substitution is the next best amino acid following the conservative mutations. For all the active mutants

that bear nonconservative substitutions, the WT amino acid has been replaced by a smaller amino acid which may simply be compatible with the existing protein structure and make no direct contact to the substrate. These site-directed mutants provide insight into how EcoRI recognizes its substrate. Consider for example that at the R200 position mutants bearing lysine, cysteine, serine, and valine cleave DNA and are not altered in substrate specificity. In the WT enzyme, R200 is proposed to recognize the outer guanine nucleotides of the substrate by making four of the twelve specific hydrogen bonds in the enzymesubstrate complex. In contrast, the hydrophobic amino acid valine can make no hydrogen bonds. Therefore the R200V mutant enzyme must recognize the WT EcoRI substrate without the interactions normally provided by R200. Similarly, in the WT enzyme residues E l 4 4 and R145 form hydrogen bonds proposed to recognize the central AT basepairs of the substrate. However, we again find that nonconservative substitutions of these residues (E144C, S, or G and R145C) reduce enzyme activity without altering substrate specificity. Thus it is a general property that changing the substrate binding amino acids of the EcoRI enzyme does not alter substrate recognition. We are forced to the conclusion that, if EcoRI does not require these specific hydrogen bond interactions to recognize its substrate, then the enzyme must make additional contacts to the DNA. These findings are in agreement with studies that show EcoRI cleaves (at reduced rate) oligonucleotides that contain unusual bases lacking functional groups that contact the enzyme, such as the N-7 nitrogen of guanine and the N6 amino groups of the central adenines.l0-l3 Furthermore, by a somewhat different method, Needels et a1.28 reached similar conclusions based on a collection of R200X mutants. The overall congruence of our findings for the R200 mutants suggests that we have both correctly identified those mutants that retain partial enzyme activity. We conclude that in addition to the hydrogen bond interactions predicted from the crystal structure model, the EcoRI enzyme must make additional contacts to recognize its substrate. What is the nature of these additional interactions? As was originally suggested by D i ~ k e r s o n , ~ ~ protein-DNA interactions can be influenced by sequence dependent variations in the structure of the

SUBSTRATE RECOGNITION BY EcoRI ENDONUCLEASE

195

The EcoRI enzyme-DNA crystal ~ t r u c t u r e ' re~ sugar-phosphate backbone. This could play a critical veals several hydrophobic amino acids (1197, L198, role in sequence recognition by the EcoRI enzyme, L10) as well as the methylene side chains of several because naked DNA bearing a n EcoRI site is lysines and arginines (R9, K15, R183, R187) that lie slightly bent4' and the structure is further disnear the substrate pyrimidines. Some of these resitorted upon protein binding.l5,l6 The crystal strucdues may make additional contacts that participate ture reveals that the DNA major groove widens to in substrate recognition. These contacts could form allow the protein's recognition machinery access to concomitantly with those described from the crystal the bases, the two AT base pairs of each half-site roll structure and may become apparent with further towards each other to facilitate the glu144larg145 structural refinements. Alternatively, because the bridging interactions, the base pairs flanking the EcoRI enzyme undergoes conformational changes recognition site are driven to high propeller twist, during binding and catalysis, the different conforand the sugar-phosphate backbone is kinked a t the mations of the enzyme could make distinct sets of edges (type I1 neokinks) and the center (type I contacts that act together to enhance the fidelity of neokink) of the recognition site. By the sequencesubstrate recognition. We call this the sequential dependent conformational model, the EcoRI enzyme contact model. In the kinetic pathway from EcoRI could recognize the sequence GAATTC because it binding to DNA scission, the proposed contacts to adopts this unusual conformation more readily than the pyrimidines could either precede or follow the other DNA sequences. For both the 434 and the trp purine contacts observed in the crystal structure. repressors, DNA secondary structure effects the These models could be tested by mutagenizing those complementary fit between protein and DNA, and amino acids that may bind the pyrimidines or by protein-backbone contacts play a role in sequence solving the structure of EcoRI-DNA complexes that specific re~ognition.~'-~' It would not be surprising might be expected to be in different conformational if EcoRI took advantage of similar interactions as states [EcoRI bound to its cleaved substrate, DNA well. complexed with EcoRI mutants defective or altered In addition, we propose that the EcoRI enzyme in allosteric activation17.1a(Heitman and Model, in may also contact the pyrimidines of its recognition preparation), or EcoRI bound to a n oligonucleotide site. For example, when the cytosines at the outer that is a competitive inhibitor but not a substrate"]. base pairs of the EcoRI site are methylated, scission The enzyme may make a by EcoRI is It has been argued that two contacts per base pair steric or hydrophobic contact to these cytosine resiare necessary and sufficient for a DNA binding protein to discriminate a basepair uniquely.55 Most dues that is blocked by the 5-methyl group. Such a n interaction would allow the enzyme to discriminate models of repressor-DNA interactions invoke two or against thymidine with its 5-methyl group and guaonly one contact per base In some cases, nine and adenine, each with a n imidazole nitrogen single hydrophobic or hydrogen bond interactions a t this position. One observation that must be recappear sufficient for unique discrimination. For exonciled with this model is a report that DNA conample, a 434 repressor mutant that recognizes an taining the unusual base 5-hydroxymethyl cytosine AT instead of a GC base pair has a hydrophobic (HMC DNA; present in unglucosylated T-even amino acid substitution (Q43A) proposed to bind the . ~model ~ of subphages) is cleaved by the EcoRI enzyme in ~ i t r o . ~ ~5-methyl group of the t h ~ m i d i n eA However, in this case, there was no determination of strate recognition by the EcoRI restriction enzyme, the rate at which HMC DNA is cleaved compared based on both the rate and specificity of star cleavwith cytosine-containing DNA. We find that HMC age and the DNA-enzyme cocrystal DNA (unglucosylated T4 or T6) is not restricted by structure,15 suggests that the enzyme binds its subEcoRI in vivo (Heitman and Model, unpublished strate by making two hydrogen bonds per base pair. results). Together these findings suggest that this Our analysis of the substrate specificity of EcoRI modified base inhibits but does not prevent scission mutants lacking functional groups that interact by EcoRI. This is again consistant with a contact with the substrate argues that the WT EcoRI endobetween the enzyme and the 5-position of the cynuclease must make additional contacts. In contrast tosine residues. Because DNA cleavage by EcoRI is to repressors where association with nonspecific inhibited when uracils replace thymidines within sites can be tolerated and binding to closely related the recognition site,'".'l the enzyme may also conoperators is required, restriction enzymes may make tact the 5-methyl groups of the substrate thymore contacts than seems necessary as a failsafe midines. By this model, the EcoRI enzyme interacts mechanism to reduce cleavage of noncanonical subwith both members of each base pair. In the E144X, strates. R145X, and R200X mutants, recognition of the puThese findings suggest that DNA-protein interrine nucleotides is disrupted, but because baseactions are more complex than simply counting hypairing is complementary, contacts to the pyridrogen bonds. Proteins bind DNA by taking advantage of whatever contacts are a ~ a i l a b l e ~ ' and ~~' midines would still permit recognition of the in this respect resemble antigen-antibody interaccanonical base pair.

196

J. HEITMAN AND P. MODEL

tions. Even with a detailed cocrystal structure, it is exceedingly difficult to determine which interactions contribute to sequence specific DNA recognition. To understand how a particular protein recognizes DNA will require not only a detailed crystal structure but also biochemical and genetic analyses that stringently test structural predictions.

ACKNOWLEDGMENTS We thank Richard Roberts, Norton Zinder, and Benedicte Michel for stimulating - discussions and critical review; Tracy Ripmaster for technical tan% John Anderson for discussions; Alan Schatzman and Francis Barany for strains; and Paul Modrich for plasmids and communication of results prior to publication. This work was supported by grants from NIH and NSF and by an MSTP fellowship awarded to J.H. REFERENCES 1. Roberts, R.J. Restriction endonucleases. CRC Crit. Rev. Biochem. 4:123-164, 1976. 2. Hedgpeth, J., Goodman, H.M., Boyer, H.W. DNA nucleotide sequence restricted by the RI endonuclease. Proc. Natl. Acad. Sci. USA 69:3448-3452, 1972. 3. Rubin, R.A., Modrich, P. EcoRI Methylase. J. Biol. Chem. 252:7265-7272, 1977. 4. Newman, A.K., Rubin, R.A., Kim, S.-H., Modrich, P. DNA sequences of structural genes for EcoRI DNA restriction and modification enzymes. J . Biol. Chem. 256:2131-2139, 1981. 5. Greene, P.J., Gupta, M., Boyer, H.W., Brown, W.E., Rosenberg, J.M. Sequence analysis of the DNA encoding the EcoRI endonuclease and methylase. J . Biol. Chem. 256: 2143-2153, 1981. 6. Jack, W.E., Terry, B.J., Modrich, P. Involvement ofoutside DNA sequences in the major kinetic path by which EcoRI endonuclease locates and leaves its recognition sequence. Proc. Natl. Acad. Sci. USA 79:4010-4014, 1982. 7. Terry, B.J., Jack, W.E., Modrich, P. Mechanism of specific site location and DNA cleavage by EcoRI endonuclease. In: “Gene Amplification and Analysis,” Vol. 5. J.G. Chirikjian, Ed. New York: Elsevier, 1987: 103-118. 8. Terry, B.J., Jack, W.E., Rubin, R.A., Modrich, P. Thermodynamic parameters governing interaction of EcoRI endonuclease with specific and nonspecific DNA sequences. J . Biol. Chem. 2589820-9825,1983. 9. Terry, B.J., Jack, W.E., Modrich, P. Facilitated diffusion during catalysis by EcoRI endonuclease. J . Biol. Chem. 260:13130 -131 37, 1985. 10. Brennan, C.A., Van Cleve, M.D., Gumport, R.I. The effects of base analogue substitutions on the cleavage by the EcoRI restriction endonuclease of octadeoxyribonucleotides containing modified EcoRI recognition sequences. J. Biol. Chem. 261:7270-7278, 1986. 11. McLaughlin, L.W., Benseler, F., Graeser, E., Piel, N., Scholtissek, S. Effects of functional group changes in the EcoRI recognition site on the cleavage reaction catalyzed by the endonuclease. Biochemistry 267238-7245, 1987. 12. Seela, F.,Driller, H. Palindromic oligonucleotides containing 7-deaza-2‘-deoxyguanosine:Solid phase synthesis of d(pG*GAATTCC)octamers and recognition by the endodeoxyribonuclease EcoRI. Nucleic Acids Res. 14:2319-2332, 1986. 13. Seela, F., Kehne, A. Palindromic octa- and dodecanucleotides containing 2‘-deoxytubercidin: Synthesis, hairpin formation, and recognition by the endodeoxyribonuclease EcoRI. Biochemistry 26:2232-2238, 1987. 14. Connolly, B.A., Eckstein, F., Pingoud, A. The stereochemical course of the restriction endonuclease EcoRI catalyzed reaction. J. Biol. Chem. 259:10760-10763, 1984. 15. McClarin, J.A., Frederick, C.A., Wang, B.-C., Greene, P., Boyer, H.W., Grable, J., Rosenberg, J.M. Structure of the

DNA-EcoRI endonuclease recognition complex a t 3 A Resolution. Science 234:1526-1541, 1986. 16. Frederick, C.A., Grable, J., Melia, M., Samudzi, C., JenJacobsen, L., Wang, B.-C., Greene, P., Boyer, H.W., Rosenberg, J.M. Kinked DNA in crystalline complex with EcoRI endonuclease. Nature 309:327-331, 1984. 17. Wright, D.J., King, K., Modrich, P. The negative charge of Glu-111 is required to activate the cleavage center of EcoRIendonuclease. J. Biol. Chem. 26411816-11821,1989. 18. King, K., Benkovic, S.J., Modrich, P. Glu-111 is required for activation of the DNA cleavage center of EcoRI endonuclease. J. Biol. Chem. 26411807-11815, 1989. 19. Rubin, R.A., Modrich, P. Substrate dependence of the mechanism of EcoRI endonuclease. Nucleic Acids Res. 5: 2991-2997. 1978. 20. Pabo, C.O., Sauer, R.T. Protein-DNA recognition. Annu. Rev. Biochem. 53:293-321, 1984. 21. Evans, R.M., Hollenberg, S.M. Zinc fingers: Gilt by association. Cell 52:l-3, 1988. 22. Wolfes, H., Alves, J., Fleiss, A., Geiger, R., Pingoud, A. Site-directed mutagenesis experiments suggest that Glu 111, Glu 144, and Arg 145 are essential for endonucleolytic activity of EcoRI. Nucleic Acids Res. 14:9063-9080, 1986. 23. Greene, P.J., Yanofsky, s., Reich, N., Day, J., Hager, P., Boyer,H.W., McClarin, J.A., Frederick, C.A., Wang, B-C., Grable, J., Love, R.L., Rosenberg, J.: Structure and function of the EcoRI endonuclease. UCLA SvmDosia on Molecular and Cellular Biology, New Series,“Vdume69. Oxender, D., ed. New York, NY: Alan R. Liss, Inc. 1987:3-7. 24. Heitman, J.,Model, P. Site-specific methylases induce the SOS DNA repair response in Escherichia coli. J . Bacteriol. 169:3243-3250, 1987. 25. Heitman, J., Zinder, N.D., Model, P. Repair of the Escherichia coli chromosome after in vivo scission by the EcoRI endonuclease. Proc. Natl. Acad. Sci. USA 86:2281-2285, 1989. 26. Heitman, J . “On the Repair of DNA Breaks and the Specificity of the EcoRI Restriction Endonuclease.” New York: The Rockefeller University, PhD. Thesis, 1989. 27. Kenyon, C.J., Walker, G.C. DNA-damaging agents stimulate gene expression at specific loci in Escherichia coli. Proc. Natl. Acad. Sci. USA 77:2819-2823, 1980. 28. Needels, M.C., Fried, S.R., Love, R., Rosenberg, J.M., Boyer, H.W., Greene, P.J. Determinants of EcoRI endonuclease sequence discrimination. Proc. Natl. Acad. Sci. USA 86:3579-3583, 1989. 29. Miller, J.H. Assay of 6-galactosidase. In: “Experiments in Molecular Genetics.” Cold Spring Harbor, NY: Cold Spring Harbor Laboratory, 1972: 352-359. 30. Kunkel, T.A. Rapid and efficient site-specific mutagenesis without phenotypic selection. Proc. Natl. Acad. Sci. USA 82:488-492, 1985. 31. Russel, M., Kidd, S., Kelley, M.R. An improved filamentous helper phage for generating single-stranded plasmid DNA. Gene 45333-338, 1986. 32. Russel, M., Model, P. The role of thioredoxin in filamentous phage assembly. J. Biol. Chem. 261:14997-15005, 1986. 33. Rosenberg, M., Ho, Y . 3 , Shatzman, A. The use of pKC3O and its derivatives for controlled expression of genes. Methods Enzymol. 101:123-138, 1983. 34. Mott, J.E., Grant, R.A., Ho, Y.-S., Platt, T. Maximizing gene expression from plasmid vectors containing the A P , promoter: Strategies for over-producing transcription termination factor p . Proc. Natl. Acad. Sci. USA 82:88-92, 1985. 35. Cheng, S.-C., Kim, R., King, K., Kim, K., Modrich, P. Isolation of gram quantities of EcoRI restriction and modification enzymes from a n over-producing strain. J . Biol. Chem 259:11571-11575, 1984. 36. Berkner, K.L., Folk, W.R. Overmethylation of DNAs by the EcoRI methylase. Nucleic Acids Res. 5435-450,1978, 37. Woodbury, C.P. Jr., Downey, R.L., von Hippel, P.H. DNA site recognition and overmethylation by the EcoRI methylase. J . Biol. Chem. 255:11526-11533, 1980. 38. Herskowitz, I. Functional inactivation of genes by dominant negative mutations. Nature 329:219-222, 1987. 39. Yanofsky, S.D., Love, R., McClarin, J.A., Rosenberg, J.M., Boyer, H.W., Greene, P.J. Clustering of null mutations in the EcoRI endonuclease. Proteins 2:273-282, 1987.

SUBSTRATE RECOGNITION BY EcoRI ENDONUCLEASE 40. Kelley, R.L., Yanofsky, C. Mutational studies with the trp repressor of Escherzchia coli support the helix turn helix model of repressor recognition of operator DNA. Proc. Natl. Acad. Sci. USA 82:483-487, 1985. 41. Miller, J.H. The lacl gene: Its role in lac operon control and its use as a genetic system. In: “The Operon.” Miller, J.H., Reznikoff, W.S., eds. Cold Spring Harbor, NY: Cold Spring Harbor Laboratory, 1978, 31-88. 42. Graddis, T.J., Klig, L.S., Yanofsky, C., Oxender, D.L. Formation of heterodimers between wild-type and mutant trp aporepressor polypeptides of Escherichia coli. Proteins 4: 173-i81, 1988. 43. Halford. S.E., Johnson, N.P. The EcoRI restriction endonuclease with bacteriophage k DNA. Biochem. J . 191593604, 1980. 44. Polisky, B., Greene, P., Garfin, D.E., McCarthy, B.J., Goodman, H.M., Boyer, H.W. Specificity of substrate recognition by the EcoRI restriction endonuclease. Proc. Natl. Acad. Sci. USA 72:3310-3314, 1975. 45. Hsu, M.-T., Berg, P. Altering the specificity of restriction endonucleases: Effect of replacing Mg”+ with Mn”. Biochemistry 17:131-138, 1978. 46. Woodbury, C.P. Jr., Hagenbuchle, O., von Hippel, P.H. DNA site recognition and reduced specificity of the EcoRI endonuclease. J . Biol. Chem. 255:11534-11546, 1980. 47. Rosenberg, J.M., Greene, P. EcoRI* specificity and hydrogen bonding. DNA 1:117-124, 1982. 48. Dickerson, R.E. Base sequence and helix structure variation in B and A DNA. J. Mol. Biol. 166:419-441, 1983. 49. Wing, R., Drew, H., Takano, T., Broka, C., Tanaka, S., ~

50.

51. 52.

~

53.

54. 55. 56.

57.

197

Itakura, K., Dickerson, R.E. Crystal structure analysis of a complete turn of B-DNA. Nature 287:755-758, 1980. Aggarwal, A.K., Rodgers, D.W., Drottar, M., Ptashne, M., Harrison, S.C. Recognition of a DNA operator by the repressor of phage 434: A view at high resolution. Science 242:899-907, 1988. Anderson, J.E., Ptashne, M., Harrison, S.C. Structure of the repressor-operator complex of bacteriophage 434. Nature 326:846-852, 1987. Otwinowski, Z., Schevitz, R.W., Zhang, R.-G., Lawson, C.L., Joachimiak, A., Marmorstein, R.Q., Luisi, B.F., Sigler, P.B. Crystal structure of trp repressorioperator complex at atomic resolution. Nature 335:321-329, 1988. Tasseron-de Jong, J., Aker, J., Giphart-Gassler, M. The ability of the restriction endonuclease EcoRI to digest hemi-methylated versus fully cytosine methylated DNA of the herpes TK promoter region. Gene 74:147-149, 1988. Kaplan, D.A., Nierlich, D.P. Cleavage of nonglucosylated bacteriophage T4 deoxyribonucleic acid by restriction endonuclease EcoRI. J. Biol. Chem. 250:2395-2397, 1975. Seeman, N.C., Rosenberg, J.M., Rich, A. Sequence-specific recognition of double helical nucleic acids by proteins. Proc. Natl. Acad. Sci. USA 73:804-808, 1976. Wharton, R.P., Ptashne, M. A new-specificity mutant of 434 repressor that defines a n amino acid-base pair contact. Nature 320:888-891, 1987. Cheng, S-C., Modrich, P. Positive-selection cloning vehicle useful for over production of hybrid proteins. J . Bacteriol. 154:1005-1008, 1983.