Cleavage of Phosphorothioate-substituted DNA by Restriction Endonucleases*

M13 RF DNA was synthesized in uitro in the pres- ence of various single deoxynucleoside 5’-0-( l-thiotri-phosphate) phosphorothioate analogues, and the three other appropriate deoxynucleoside triphosphates using a M13 (+)-single-stranded template, Escherichia coli DNA polymerase I and T4 DNA ligase. The resulting DNAs contained various restriction endonuclease rec- ognition sequences which had been modified at their cleavage points in the (-)-strand by phosphorothioate substitution. The behavior of the restriction enzymes Aval, BamHI, EcoRI, HindIII, and SalI towards these substituted DNAs was investigated. EcoRI, BamHI, and HindIII were found to cleave appropriate phos-phorothioate-substituted DNA at a reduced rate compared to normal M13 RF DNA, and by a two-step process in which all of the DNA is converted to an isolable intermediate nicked molecule containing a specific discontinuity at the respective recognition site presumably in the (+)-strand. By contrast, S d cleaved substituted DNA effectively without the intermediacy of a nicked form. AuaI, however, is only capable of cIeaving thd unsubstituted (+)-strand in appropriately modified DNA.

internucleotidic linkage of the R, configuration is produced in the new DNA chain. The polymerization rates of these analogues are comparable with those observed using normal substrates (3, 4). In this way double-stranded circular fd RF DNA was prepared in which only the (-)-strand had been modified by the incorporation of dAMPS residues instead of dAMP (2), thus demonstrating the feasibility of modifying DNA using these analogues. Drastic changes in the physicochemical properties of this modified DNA were not observed, and the DNA remained infective. A later report (5) showed in a semiquantitative fashion that such DNA is often cIeaved considerably more slowly by restriction endonucleases than unsubstituted DNA. However, from the data presented it was not possible to ascertain whether the eventual cleavage of this hybrid DNA proceeded, as might be expected, first by attack on the unsubstituted template strand to give a nicked intermediate. If this were the case, as might be predicted from the general properties of phosphorothioate analogues, then phosphorothioate substitution might form a general basis for the generation of a specific nick at any desired restriction site in one strand of a duplex DNA molecule, and permit mechanistic information on the mode of cleavage of DNA by restriction endonucleases to be obtained.
We wish to report here that phosphorothioate substitution, by altering relative cleavage rates, can be an effective probe of restriction enzyme cleavage mechanisms under optimum conditions for cleavage. Forschung, Heidelberg and were obtained from denatured calf thymus DNA by limited digestion with pancreatic DNAase, and subsequent fractionation on DEAE-Sephadex G-100. The primer molecules had an average length of 15-20 nucleotides as judged by polyacrylamide gel electrophoresis. On the average, three primer molecules were present per DNA template during in uitro synthesis. dCTP, dTTP, dGTP, dATP, ATP, and ethidium bromide were obtained from Boehringer Mannheim, GmbH. E. coli acetate kinase (200 units/mg, 5 mg/ ml), E. coli DNA polymerase I (endonuclease free, 5 unitslpl), BamHI (11 units/pl), HaeIII (5 units/pl), HindIII (10 units/pl), and SalI (9 units/pl) were purchased from Boehringer Manheim GmbH. Aual (6 units/rl) was from Amersham Buchler GmbH, Braunschweig. T4 DNA ligase (3-5 units/l), isolated from the overproducing strain E.

M13mp9
coli NM 989 was a gift of Dr. F. Winkler, European Molecular Biology Laboratory, Heidelberg. EcoRI was a product of Renner GmbH, Heidelberg. Microfiltration apparatus and appropriate nitrocellulose filters (pore size 0.45 pm, diameter 9 mm) used for DNA purification were obtained from Schleicher and Schiill GmbH, Dassel, W. Germany.
dNMI'Ss were prepared by a slight modification of the method described for AMPS (6): deoxyadenosine, deoxycytidine, and deoxyguanosine were dried a t 80 "C in uacuo overnight and thymidine by repeated evaporation of dry pyridine. Thiophosphorylation on a 10mmol scale was carried out in triethylphosphate (40 ml), except for deoxyguanosine for which trimethyl phosphate was used. Before addition of PSCI:, the deoxynucleosides were dissolved by heating the reaction mixture gently over an open flame. Usually 2.2 eq of PSCln were employed except for deoxyadenosine where 1.1 eq was sufficient. Workup was as described (6). Final purification was by chromatography on DEAE-Sephadex using a linear gradient of 1.5 liters each of H20 and 0.45 M triethylammonium bicarbonate as eluent. Yields ranged from 30 to 50%. "P NMR (D,O/EDTA, pH 7.0) showed typically A, 43.8 ppm.
The mixtures of diastereomers of dNDPnS were prepared by activation of dNMPS with diphenylphosphorochloridate and reaction with phosphoric acid as described for the synthesis of ADPnS (7). Yields were between 20 and 40%.
All dNDPtvS and dNTPnS were also checked for purity by reversephase high-performance liquid chromatography using 50 mM KH2P04 as mobile phase (8).
Synthesis of M13rnp9 RF and M13rnp2 RF-Typical was the following experiment: M13mp9 (+)-single-stranded DNA (100 p1, 2 pg) was mixed with calf thymus primer (1 pl, 4.6 OD,,/ml), the solution heated to 45 "C for 15-20 min, and allowed to cool to room temperature over 0.5 h. This mixture was diluted to 230 p l and made 1.2 mM in ATP, 0.8 mM each in MTP, dCTP, dGTP, and dTTP, 50 mM in Tris.HCI, pH 8.0, 50 mM in NaCl 10 mM in MS12, and DNA polymerase I (4 p l , 20 units) and T4 DNA ligase (2 p l , 6-10 units) were added. The final volume was 237 p l , and this mixture was incubated at 16 "C. After 18-20 h, gel electrophoresis generally showed >90% conversion to RF. The solution was then made approximately 10 mM in EDTA and heated a t 70 "C for 10 min. The resulting solution containing approximately 0.14 pg of M13 RF DNA/pl was used directly for restriction experiments. If the DNA contained appreciable amounts of contaminating RF I1 and RF 111 forms it was purified by denaturation of these forms and passage through a nitrocellulose filter according to Gray et al. (9).
Synthesis of M13rnp9 (S)N RF and M13rnp2 (S)N RF-The above protocol was used, with the difference that one dNTP was replaced by the corresponding dNTPnS. No differences in the reaction or quality of the product were noted as a result of this substitution.

RESULTS
Covalently closed circular relaxed M13 RF IV DNAs containing various single phosphorothioate substitutions in the (-)-strand were synthesized in vitro by polymerization of a dNTPaS phosphorothioate analogue and the three appropriate normal deoxynucleoside triphosphates on a M13 (+)single-stranded template using E. coli DNA polymerase I.
Closure of the newly synthesized circles was effected by T4 DNA ligase. Internucleotidic phosphorothioate linkages were thus introduced base specifically in the (-)-strand and substituted cleavage sites for several DNA restriction endonucleases. The effects of these substitutions on DNA cleavage by these enzymes were studied.
EcoRZ-The enzyme EcoRI recognizes the palindromic duplex sequence GAATTC (10) and cleaves between the dG and dA residues. After (-)-strand synthesis in the presence of dATPaS, 2 dAMPS residues are incorporated into this sequence in the (-)-strand, and there is consequently an internucleotidic phosphorothioate bridge of the R, configuration   1 min under these conditions. A d -T h e enzyme AuaI recognizes the sequence CPy-CGPuG (10) which is not necessarily palindromic and cleaves between the dC and the pyrimidine residues. There is a unique site for AuaI in M13mp2 RF DNA possessing the non-palindromic sequence CTCGGG. Thus in order to substitute at the cutting position in the (-)-strand, second-strand synthesis must be performed in the presence of dCTPaS.
M13mp2 (S)C RF IV DNA was tested as a substrate for AuaI. The results are shown in Fig. 4. It is observed that nicking of the DNA to an open circle form readily occurs, presumably at the unsubstituted AuaI site in the (+)-strand, but that this is the end product of the reaction (Fig. 4. ~o D ) .
bation. Unmodified M13mp2 RF IV DNA was cleaved by No nicks were introduced into phage 6x174 DNA, which has no EcoRI site, when this was treated under the same conditions, thus ruling out the possibility to random nicking by possible endonuclease impurities in the EcoRI.
HindZZZ-The enzyme HindIII cleaves between the dAMP residues in the sequence AAGCTT. When the cleavage of M13mp9 (S)A RF IV DNA at the unique site by HindIII was investigated a similar reaction profile was obtained as for EcoRI (results not shown).
BarnHZ-The enzyme BarnHI recognizes the sequence GGATCC (IO) and cleaves between the two dG residues. M13mp9 (S)G RF IV DNA was synthesized using dGTPaS and tested as a substrate for BamHI. The results are shown in Fig. 3. A similar pattern to that observed for EcoRI was observed. Cleavage of the phosphorothioate DNA clearly pro-  AuaI under these conditions in approximately 2.5 min. AuaI introduced no nicks into SV40 DNA, which has no AuaI site, under the same conditions. That the enzyme is not rapidly deactivated under the conditions used was tested by incubation of the enzyme alone in reaction buffer for 1 h. DNA was then added and incubation continued. The characteristic cleavage pattern of AuaI was observed, confirming the presence of active enzyme (results not shown). That this DNA itself possessed no other irregularity other than the desired phosphorothioate substitution was tested as shown in Fig. 4, bottom. The DNA was first converted fully to the nicked form by AuaI, and then HaeIII was added. As can be seen from Fig.  4, bottom, the DNA was rapidly cleaved at the phosphorothioate to give HaeIII fragments.
Sall-The enzyme SalI recognizes the DNA sequence GTCGAC and cleaves between the dG and dT residues (10). M13mp9 (S)T RF IV DNA was prepared using dTTPaS and tested as a substrate for SalI. This DNA contains a unique restriction site for SalI in the multiple cloning region. The results are shown in Fig. 5. It can be seen that in contrast to EcoRI, RamHI, and AuaI essentially no open circle intermediate is observed, and cleavage proceeds directly from the RF IV to the RF 111 form with only minimal formation of a nicked RF I1 intermediate. The cleavage reaction was slower than for unmodified DNA which under the condition used was linearized in less than 1 min. SalI introduced no nicks into 4x174 DNA, which has no SalI site, under the same conditions.

DISCUSSION
Phosphorothioate-containing polynucleotides are generally more slowly hydrolyzed by enzymes than those containing phosphate groups (1). Thus, snake venom phosphodiesterase cleaves polynucleotides with phosphorothioate linkages of the R, configuration approximately 10 times more slowly than normal polynucleotides (3, l l ) , and exonuclease I11 (12, 13) as well as the 5' + 3' exonuclease activity of E. coli DNA polymerase I (3, 14, 15) cleave them so slowly that hydrolysis can normally not be detected. Also, the 3' + 5' exonuclease -substituted DNA activity of the latter hydrolyses these groups extremely slowly (4, 16). It is this aspect which justifies a more detailed investigation of the cleavage of phosphorothioate groups in DNA by restriction endonucleases as it might open a way of achieving cleavage of a particular restriction enzyme site only in one strand by protecting it by phosphorothioate substitution in the other. The earlier report on this subject described the inhibition of cleavage of hybrid phosphorothioate-substituted DNA by six restriction enzymes (5). It established that inhibition was most pronounced when an internucleotidic phosphorothioate linkage was introduced exactly at the cleavage point in the appropriate recognition sequence. In this work, electrophoresis was performed on agarose gels in the absence of ethidium bromide and it was not possible to distinguish between RF I1 DNA containing one or more nicks and covalently closed circular relaxed RF IV DNA. Consequently, it was not possible to examine how cleavage of the hybrid DNA molecules thus prepared had been effected.
The reaction catalyzed by Type I1 restriction enzymes requires subsequent introduction of two single-stranded breaks within a recognition sequence contained in a DNA. The two cleavage events can occur simultaneously or be separated in time. In the latter case, the enzyme will either stay bound or will dissociate between the cuts, and an open circle form I1 intermediate will be formed. The observation of this intermediate will be indicative of a mechanism where the two cleavages are distinct. When no intermediate is observed the two cleavages can be said to occur simultaneously.
It is thought that sequential cleavage of DNA strands may be a general property of restriction endonucleases. Evidence in the case of the most studied endonuclease, EcoRI, is rather well-founded (17-20). In several other cases the intermediate possessing a single-strand scission at the recognition sequence is isolable under suboptimal reaction conditions, i.e. with limiting amounts of enzyme (18), a t low temperatures (21, 22), or in the presence of ethidium bromide (17).
There are several reports concerning the isolation of RF I1 DNA with single-strand nicks generated by restriction endonucleases in the presence of ethidium bromide (23-26). One of the main uses of such DNA has been its application to sitedirected mutagenesis using nick translation in the presence of base-modified nucleotides (27), by the bisulfite method (25) and by gap misrepair mutagenesis including the use of dNTPaS (29). A common limitation of these methods, however, is the great variability in the behavior of restriction enzymes towards DNA in the presence of ethidium bromide (26). Thus, it is sometimes only possible to nick between 50 and 90% of the input DNA, and some linearization and unchanged starting material is almost always observed. The exact conditions must be determined by careful titration for each enzyme. More serious, however, is the fact that many enzymes do not exhibit this effect a t all, and some only very poorly. Thus it is often difficult to generate reasonable quantities of nicked DNA and many potential mutagenesis sites on a given genome are not available for this technique. In addition, of course, the partial cleavage by restriction enzymes is not strand specific so that the gaps are presumably equally distributed between the (+)-and the (-)-strand. Other methods for generating specific nicks would clearly be most useful.
We suspected that cleavage of a hybrid DNA molecule possessing phosphorothioate substitutions a t a restriction cleavage site in the (-)-strand, should proceed by an enhanced two-step process for restriction enzymes which are supposed to cleave only by a two-step mechanism, since cleavage of the unsubstituted viral (+)-template strand should be favored, followed by a considerably slower cleavage of the phosphoro-thioate internucleotidic linkage in the (-)-strand. This should permit isolation of a DNA RF I1 form with a nick only in the (+)-strand.
To this end we decided to reinvestigate the previously reported inhibitory effects of phosphorothioate-substituted DNA on restriction endonucleases (5) using first the enzyme EcoR1, since more mechanistic data is available for this enzyme than any other (17) and also because these data indicate a two-step mechanism for this enzyme.
In this reinvestigation it was important to monitor cleavage products on agarose gels containing ethidium bromide (231, which facilitate the separation of open circle and covalently closed circular forms of DNA. We chose the single-stranded DNA of bacteriophage M13. Since this is currently one of the most popular cloning vehicles (30), it has frequently been employed for site-specific mutagenesis experiments (31) and M13mp9 possesses many unique restriction enzyme cleavage sites in the multiple cloning region.
The mechanism of cleavage by EcoRI has been extensively studied (17)(18)(19)(20)(21)(22). Evidence is strongly in favor of the enzyme acting in a two-step fashion with the participation of an opencircle intermediate. Second-strand cleavage is slower than first-strand cleavage by a factor of about two, although this may be accounted for by a statistical factor (19,20), and the enzyme may dissociate from the DNA between first-and second-strand cleavages, although at 37 "C the cleavage process appears to be coupled. However, slightly different behavior has been observed with different DNA.
When M13mp9 RF IV DNA and M13mp9 (S)A RF IV DNA were compared as substrates for Ec6RI under identical conditions, the former DNA was completely cleaved to the linear form, whereas the latter phosphorothioate-substituted DNA was at first only converted to the RF I1 nicked form (Fig. 1). This result can be explained on the basis of the generally slower rate of enzymatic hydrolysis of phosphorothioates in comparison to phosphates as mentioned above. Indeed, for the synthetic octamer d(GGsAATTCC) containing the EcoRI recognition sequence and a phosphorothioate group at the cleavage site in both strands a decrease in rate of approximately 20 in comparison to the unmodified octamer could be observed (32). The conclusion that EcoRI has placed a nick in its recognition sequence in the unsubstituted viral (+)-strand is therefore most reasonable. The observed rate difference is sufficient to obtain the nicked material as the sole product in the first phase of the reaction.
Incubation with larger amounts of enzyme, when monitored over a period of time, demonstrated that after nicking to form I1 DNA has taken place, subsequent second-strand cleavage of the phosphorothioate-containing strand occurs resulting in a linear RF 111 product (Fig. 2). This result is thus in agreement with what has been found with the phosphorothioate octamer d(GGsAATTCC), namely that a phosphorothioate linkage of the R, configuration can be hydrolyzed by EcoRI, albeit slowly. Moreover, the fact that in this small substrate analogue cleave by EcoRI occurred at the correct position in the recognition sequence provides reassurance that the second strand of M13mp9 (S)A RF IV DNA is indeed also cleaved at the correct site. The overall result demonstrates that even a t 37 "C double-strand cleavage by EcoRI is not an obligatory coupled process.
The cleavage of this DNA by Hind111 also leads to formation of a nicked intermediate. As this enzyme can nick DNA, like EcoRI, in the presence of ethidium bromide (28) this is probably not surprising, and indeed a two-step cleavage mechanism has been proposed (19).
Evidence in favor of a two-step mechanism for the endo-nuclease BamHI has been provided by Smith and Chirikjian (33) who showed that form 11 DNA accumulates during the digestion of plasmic pmB9 with the enzyme at 1 "C with limiting amounts of nuclease. Also Shortle and Botstein (28) have demonstrated that BamHI displays an efficient nicking reaction in the presence of ethidium bromide. On the other hand, Halford et al. (19) have shown that under optimal conditions BamHI displays a higher reactivity towards second-strand cleavage once the first strand has been cut than towards first-strand cleavage. Recent support for this has been provided by examination of cleavage of a supercoiled DNA substrate ( (Fig. 3). Thus, cleavage at the second, the phosphorothioate-containing strand is now slowed down and an isolable form I1 intermediate is observed suggesting a two-step mechanism a t least for this artificial DNA. A simultaneous strand cleavage mechanism similar to that for BamHI has also been proposed for the enzyme AuaI (35). If this were correct, one would not expect to observe a nicked intermediate, even in the presence of a phosphorothioate group in one strand. The cleavage of M13mp2 (SIC RF IV DNA by this enzyme, however, shows not only this unexpected intermediate but also that it is in this case the endproduct of the reaction. This preparation of DNA could be cleaved by HaeIII, an enzyme which attacks M13mp2 DNA a t multiple sites cleaving between dG and dC in the sequence GGCC, and is known to cleave M13mp2 (S)C RF IV DNA.' As can be seen from Fig. 4, bottom, the DNA was rapidly cleaved to give HaeIII fragments indicating that it could be cleaved by another restriction enzyme sensitive to dCMPS substitution.
Several explanations for this behavior of AuaI are possible. First, it is important to note that of the two possible configurations of a phosphorothioate diastereomer in general, normally only one will be cleaved by an enzyme at a reasonable rate. It is normally not possible to predict in advance which one this may be. Thus, among the nonspecific DNA endonucleases, there are enzymes which cleave R, linkages and those which cleave S, linkages (1). It is thus quite possible that AuaI might be one of the enzymes cleaving S, rather than R, phosphorothioate linkages. As the phosphorothioate groups introduced into DNA by polymerization with DNA polymerases have only the R, configuration, this would explain the inability of the enzyme to linearize this DNA. At present, only data on the stereospecificity of retriction enzymes are available for EcoRI and this enzyme has been shown to recognize a R, internucleotidic linkage (32). However, this does not exclude the possibility that other restriction enzymes may show a different stereospecificity. A second explanation could be that the row of three dCMPS residues, in the middle of which the cleavage site resides, in the recognition sequence of the (-)-strand of this particular DNA causes an inhibition in an as yet unexplained way. That in a recognition sequence other residues substituted with phosphorothioate but not directly at the cleavage point can have an inhibitory effect on cleavage was noted by Vosberg and Eckstein (5). Third, the possibility exists that DNA nicked in one strand is not a substrate for AuaI. However, it is not yet known whether AuaI exhibits a good nicking reaction in the presence of ethidium bromide so this cannot at present straightforwardly be tested. That AuaI does, however, nick one strand of a DNA although not natural does seem to rule out a mechanism whereby it is * B. V. L. Potter and F. Eckstein, unpublished results. obligatory for the enzyme to cut both strands simultaneously, and makes a two-step mechanism of action more likely for this enzyme. Clarification of this aspect, however, must await further work.
A particularly challenging enzyme to investigate with hybrid phosphorothioate-phosphate DNA with the view of isolating nicked intermediates is the restriction endonuclease SalI. On cleavage of normal DNA by this enzyme under optimal conditions no nicked intermediate can be observed (19, 20, 35). It is only under suboptimal conditions at low pH or low MgC1, concentration that some nicked DNA can be seen but this DNA is cleaved at a slower rate that the intact duplex and is therefore not an obligatory intermediate. It is therefore suggested that the preferred pathway for this enzyme is a concerted reaction to cleave both strands of the DNA within one enzyme-DNA complex but some nicked DNA can dissociate before cutting of the second strand. T o be able to isolate nicked DNA in the cleavage of the phosphorothioate-phosphate DNA would then require that this leakage would have to be increased by virtue of the slow rate of cleavage of the second strand. However, cleavage of M13mp9 (S)T RF IV DNA by SalI gave no indication of such a process.
Except for a time point taken at 3 min incubation, either only starting material or linearized DNA as product can be detected. Thus, the observation that even in a DNA where cleavage of one strand is most likely more difficult than the other, the reaction still proceeds without dissociation of the enzyme-substrate complex and provides further evidence for the assumption that the SulI-catalyzed reaction proceeds in a concerted fashion.
These studies demonstrate that for restriction enzymes which cleave both strands of DNA in a stepwise fashion, nicked DNA can in many cases be isolated by employing phosphorothioate-phosphate DNA as substrate when the proper choice of dNTPolS has been made for the synthesis of the second strand. As the phosphorothioate groups can be cleaved eventually by most enzymes, AuaI being at present the exception, the amount of enzyme and the time of incubation have to be determined so as to obtain this nicked DNA as the sole product. Exceptions for the preparation of such nicked DNA by this method seem to be enzymes which cleave the DNA in a concerted fashion as exemplified in this respect by SalI.
This method of protecting restriction sites against cleavage in one strand of double-stranded DNA might prove to be an interesting method for the creation of gaps for the performance of single-site mutagenesis.