Gene, 104 (1991) 75-80 0 1991 Elsevier Science

GENE

Publishers

B.V. All rights reserved

75

0378-l 119/91/$03.50

05064

DNA synthesis on discontinuous (Nontemplated

nucleotide

addition;

templates by DNA polymerase I of Escherichia

non-homologous

coli

recombination)

James M. Clark Laboratory of Molecular Genetics, National Institute of Environmental Health Sciences, Research Triangle Park. NC 27709 (U.S.A.) Received by R.E. Yasbin: 7 December 1990 Revised/Accepted: 1 April/8 April 1991 Received at publishers: 21 May 1991

SUMMARY

DNA polymerases normally catalyze DNA synthesis in a template-directed manner. Generally, the continuity of the phosphodiester backbone of the template strand was thought to be an absolute requirement for DNA synthesis. Here, I demonstrate that a 3’-exonuclease-deficient derivative of the Klenow (large) fragment of Escherichia colt’ DNA polymerase I (PolIk) can carry out DNA synthesis on discontinuous templates in vitro. Addition of multiple nucleotides (nt) to the 3’ end of a blunt-end duplex, templated by unlinked single-stranded oligodeoxyribonucleotides (oligos), was monitored electrophoretically. The reaction was demonstrable with either homopolymers or mixed-sequence oligos, but showed a requirement for complementarity between the Iirst nt added to the duplex and the 3’ nt of the unlinked oligo. These results demonstrate that continuity of the phosphodiester backbone of the template strand is not absolutely required for in vitro DNA synthesis by a 3’-exonuclease-deficient form of PolIk.

INTRODUCTION

DNA polymerase I of E. colt’ (PolI) has served as a paradigm for enzymes that synthesize DNA in a templatedirected manner (Kornberg, 1980). It is the only DNA polymerase for which detailed structural information is presently available (Ollis et al., 1985). Moreover, mutant forms of the protein that retain normal levels of PolI activity but lack the 3’-5’ (proofreading) and 5’-3’ exonuclease activities found in the wt PolI have recently been constructed by site-directed mutagenesis. The availability of Correspondenceto: Dr. J.M. Clark,

Laboratory

N.I.E.H.S.,

Triangle

P.O. Box 12233, Research

Tel. (919)541-0118; Abbreviations: phate;

dNTP,

high-performance Klenow

Molecular

Genetics,

Park, NC 27709 (U.S.A.)

Fax (919)541-7593.

bp, base pair(s); deoxynucleoside liquid

(large) fragment

oligodeoxyribonucleotide;

dNMP,

deoxynucleoside

triphosphate;

chromatography;

monophos-

Exo, exonuclease;

HPLC,

nucleotide(s);

PolIk,

of PolI; PolI, E. co/i DNA polymerase

I; oligo,

ss, single strand(ed);

nt,

wt, wild type.

these mutant forms of PolI has facilitated the analysis of DNA polymerase structure-function relationships at high resolution (Derbyshire et al., 1988). DNA synthesis catalyzed by PolIk is generally a highly accurate process in which the specificity of nt addition is dictated by the sequence of the template strand. Recently, we described a novel blunt-end addition reaction catalyzed by PolIk and other DNA polymerases (Clark et al., 1987; Clark, 1988). In this reaction, a single nt was added to the 3’ terminus of a blunt-ended DNA duplex. Any one of the four dNTPs could be added when provided individually; however, when all four dNTPs were present, the nt added was almost exclusively dAMP. Addition of dAMP, in particular, appeared to be a nontemplated event since standard base pairing information from thymidine residues was not required. We also observed that PolIk used dNMPs to provide coding information for templated, single-nt addition to a duplex DNA substrate (Clark et al., 1987). Here, I report that a 3’-exonuclease-deficient form of PolIk uses ss oligos, physically unlinked to a DNA duplex, as

76 templates duplex.

for multiple

EXPERIMENTAL

nt additions

to the 3’ end(s) of the

Since ternplating had been previously observed when dNMPs were included in the reaction, it was of interest to determine whether or not unlinked oligos could also serve as templates for synthesis. Fig. 2 shows the results of such

AND DISCUSSION

(a) Discontinuous homopolymers

templating occurs with small, unlinked

The substrate used to study templated addition consisted of two complementary pentadecanucleotides annealed to form the following blunt-end duplex: 51-32 P-GTCCGTCTCTGCCTC 3’-

CAGGCAGAGACGGAG

(see Fig. 1, top). Addition of nt to this substrate was monitored by electrophoresis of the reaction products through high-resolution, denaturing polyacrylamide gels. This duplex was used in earlier studies to characterize nontemplated and dNMPternplated nt addition reactions catalyzed by PolIk. Both reactions were demonstrated with wt PolIk having intrinsic 3’ + 5’ exonuclease activity, and with a mutant form of the enzyme that is deficient in this activity (Clark et al., 1987).

5' 32P-GTCCGTCTCTGCCTC 3' CAGGCAGAGACGGAG

(3'-Exo-)

J

5'32P-GTCCGTCTCTGCCTCG CAGGCAGAGACGGAG 3'

5' 32P-GTCCGTCTCTGCCTC CAGGCAGAGACGGAG3'

5' 32P-GTCCGTCTCTGCCTCG CAGGCAGAGACGGAGU2X 3'

5' 32P-GTCCGTCTCTGCCTCGGGG CAGGCAGAGACGGAG&XiX 3'

5' 32P-GTCCGTCTCTGCCTCGGGG CAGGf2AGAGACGGAGEX.C 3'

Fig. 1. A schematic view ofthe representative products of the nt addition reactions demonstrated in Fig. 2, exemplified by pd(C),. Alternative reaction

pathways

templating: (left branch)

are

the pathways

illustrated

for

one

differ by the presence

of an obligatory

+ 1 intermediate.

cycle

of discontinuous

(right branch)

or absence

In either case the starting

substrates and final products are the same. The unlinked homopolymer sequence is underlined; it remains unlinked even after the reaction and, in the position

shown,

cannot

be ligated to the duplex since the 5’ end

of the adjoining oligo is not phosphorylated. the 3’ + 5’-exonuclease-deficient PolIk.

The experiments described below were carried out with the 3’ + 5’-exonuclease-deficient mutant PolIk.

PolIk (3’-Exe-)

represents

an experiment in which unlinked, poly(dC) homopolymers of varying lengths [pd(C),] were used as templates. In the absence of the ss oligos, the 3’-exonuclease-deficient PolIk generally added only one nt to the 3’ end of the labeled strand (Fig. 2, lane 2). However, in the presence of oligos of increasing length, the enzyme added correspondingly greater numbers of nt. When pd(C), was included in the reaction, a major band corresponding to the addition of two nt was observed (Fig. 2, lane 3). With pd(C),, numerous products were synthesized corresponding to the addition of one to more than 11 nt (Fig. 2, lane 4) and with pd(C),,, the principal products formed represent the addition of more than 20 nt (Fig. 2, lane 5). The ternplated synthesis reaction requires the presence of the blunt-end duplex since control experiments in which the labeled strand of the duplex (as an ss molecule) was incubated with the polymerase in the presence of dGTP and pd(C),, did not result in any extension of the primer (data not shown). The addition of more nt than would be expected for a single cycle of ternplated addition [i.e., >4 nt for p(dC), and > 10 nt for pd(C),,] suggests that multiple cycles of synthesis occurred with these two homopolymers. That is, synthesis proceeded to the end of one template homopolymer, after which the process was repeated with another homopolymer template. Alternatively, a single homopolymer molecule could be used as the template for continued elongation via a repetitive slippage/synthesis mechanism. These two pathways are not mutually exclusive. Discontinuous ternplating promoted by pd(C), was observed at a homopolymer concentration of 100 PM (Fig. 2) but was almost undetectable at a concentration of 1 PM (data not shown). In contrast, discontinuous templating was readily observed at 1 PM pd(C),,, (Fig. 2). These data suggest that the efficiency with which the enzyme uses unlinked templates increases with increasing length of the oligos. Moreover, the use of homopolymers as unlinked templates is not restricted to pd(C),, since pd(G),,, pd(T),,, and pd(A),, are all utilized by the enzyme to carry out the reaction (data not shown). Fig. 1 provides a schematic overview of the various nt addition reactions demonstrated in Fig. 2 and illustrates two possible pathways for the discontinuous ternplating reaction. In one pathway, discontinuous ternplating proceeds through direct juxtaposition of the duplex and ss DNA substrates (Fig. 1, left branch); in the other, a (nonternplated) + 1 product is formed as an obligatory intermediate (Fig. 1, right branch).

77 I

2

(b) Complementarity

3

4

100

10

requirement

for multiple

templated

additions 1

Since the 3’Exop

PM

mutant

PolIk shows a strong bias for

the nontemplated addition of dAMP (Clark et al., 1987), it seemed likely that the ternplated and nontemplated addition reactions would be competitive if the first nt added was not complementary to the homopolymer template. To test this hypothesis directly, two stage reactions were carried out in which the blunt-end duplex was first extended (by one nt) with either dATP or dGTP alone. Subsequently, a single homopolymer (either pdT,,, complementary to the 3’ dAMP residue or pdC,,, complementary to the 3’ dGMP residue) was added, along with all four dNTPs, to assess the ability of the added homopolymers to act as templates. As shown in Fig. 3, the presence of a noncomplementary nt

nt 37

23

2

I

+I

nt

3

A

5

4

A

G

G

15

Fig. 2. Autoradiogram directed

by unlinked,

end duplex, and

synthesized,

Beardsley,

1987)

HPLC-purified mutant

showing the addition homopolymer s2P-labeled,

and annealed

were obtained

PolIk (designated

D355A,

from

E357A;

by Catherine

Joyce

available

commercially

from U.S. Biochemical

DNA

DNA/400

synthesis

tions of ss homopolymers were carried

Pharmacia.

(Yale University).

reactions

nM dGTP/excess

Corporation,

contained

approx.

(0.4 PM) enzyme,

mamide

solution,

denaturing

Reactions

were terminated

heat-denatured,

1986; 1987). Autoradiography

35 nM

and variable

duplex

concentra-

HCI pH 8.0/10 mM

X-Omatic

and analyzed

was carried

Regular

of the indicated

by addition

gels as described

are given in of a dye/for-

by electrophoresis

on

(Clark and Beardsley,

out at -20°C

intensifying

unextended 15-mer marker (duplex Lanes Z-5 show the reaction products or presence

is now

Cleveland,

as noted at the top of the figure. The reactions

20% polyacrylamide

with a Kodak

and

et al., 1988) was

(This enzyme

out at 37°C for 60 min in 25 mM Tris units.

(Clark

The 3’-Exe-

MgCl, in 5-10 g1 sample volumes. All DNA concentrations molecular

duplex

ofthe blunt-

as described

Derbyshire

provided OH).

The sequence

in section a. The homopolymers

is given

dNTPs

ofnt to the blunt-end

templates.

screen.

for approx.

2h

Lane 1 shows the

DNA incubated without PolIk). synthesized in the absence (lane 2)

concentrations

of the homopolymers

pd(C),

Fig. 3. Autoradiogram reactions.

These

showing

reactions

DNA/O.4 PM enzyme, tration

homopolymer

products

A schematic

representation

is given in Fig. 1 for pd (C,).

of the reaction

substrates

and

70 nM duplex (G) at a concen-

at 37”C, the appropriate

was added (10 PM final concentration)

for an additional45 were carried

min at 37°C. Sample preparation

out as described

(G) (lanes 4 and 5) followed

radiogram.

in two-stage

approx.

along with all four

dNTPs at 400 PM each, resulting in a dilution of the duplex DNA to a final concentration of approx. 40 nM. The incubation was then continued

extended

the auto-

synthesized

and either dATP (A) or dGTP

visible on the original

and are given alongside

contained

of 330 PM. After a 15-min incubation

(lane 3). pd(C), (lane 4), and pd(C),, (lane 5). Product lengths, in nt, were determined by counting bands (including bands at intermediate positions autoradiogram)

the products

initially

overnight

at -20°C

without

in Fig. 2. Autoradiography a screen.

The blunt-end

and electrophoresis was carried duplex

out

was first

by one nt ( + 1 nt), either dAMP (A) (lanes 2 and 3) or dGMP

(lanes 2 and 5) or pd(C),, 15-mer marker

by addition

of the homopolymers

pd(T),,

(lanes 3 and 4). Lane I shows the unextended

(Fig. 1, top).

78 on the 3’ end of the duplex

effectively

blocked

further

12

synthesis on the homopolymer templates (lanes 3 and 5). These results are consistent with, but do not prove, the hypothesis that discontinuous ternplating proceeds through an obligatory + 1 intermediate as illustrated in Fig. 1. Somewhat longer products were synthesized on the homopurine template than on the homopyrimidine template (Fig. 3). The reason for this difference in product size distribution is not known but may reflect differential stability of DNA structures formed as intermediates during the reaction. It should also be noted that, except for the + 1

345

dNTPs

-

4

G

4

G,C,T

Qmer

-

-

-

+

+

residue, the sequence of the added nt has not been specified; direct sequence analysis of the reaction products will be required to determine the accuracy with which discontinuous ternplating occurs. (c) Discontinuous templating occurs with oligos of mixed sequence Since the unlinked oligos used in the preceding experiments were homopolymers, it is possible that some of the products were generated by translocation of a single, homopolymer template, e.g., pd(C),,, relative to the extended pd(G), portion of the primer strand. Therefore, I assayed the ability of a synthetic oligo of mixed sequence to act as a template. As shown in Fig. 4, a synthetic 9-mer having the sequence 5’-CGGCCCATC also served as a substrate for multiple cycles of ternplated synthesis when all four dNTPs were included in the reaction (lane 4). A substantial amount of + 1 product was also formed under these conditions. Since the nontemplated pathway shows a strong bias for the addition of dAMP when all four dNTPs are provided (Clark et al., 1987), most of the + 1 product seen in lane 4 terminates with a 3’-dAMP residue. Subsequent discontinuous ternplating is presumably blocked by the lack of complementarity between the + 1 nt (dAMP) and the dCMP-terminated oligo of mixed sequence. If dATP is omitted from the reaction, the ability of the 9-mer to serve as a template is virtually abolished and synthesis terminates predominantly after the addition of a single nt (Fig. 4, lane 5). It is likely that discontinuous templating is blocked, in part, because of the addition of a noncomplementary 3’ nt as described above. However, this mechanism cannot be the only explanation for the effects observed when dATP was omitted. One would not expect to see significant discontinuous ternplating when all four dNTPs were present because the strong bias for dAMP addition noted above would favor the formation of a + 1 product having a noncomplementary 3’ terminus. Yet discontinuous ternplating was clearly observed under these conditions (Fig. 4, lane 4). In addition, it should be noted that the + 1 product seen in lane 5 is heterogeneous, consisting of two distinct species that differ slightly in electrophoretic mobility. These bands correspond to + 1 products

Fig. 4. Autoradiogram directed marker absence

showing the addition

by a mixed-sequence

of the 9-mer when either all four (4) dNTPs

(G) alone (lane 3) were present. lane 5 shows the corresponding carried

from the reaction. out as described

synthesized

products

synthesized

9-mer ohgo were present;

synthesized

when dATP was

Sample preparation

and electrophoresis

in Fig. 2; autoradiography

was carried

night as in Fig. 3. The dNTPs

were present

in the

(lane 2) or dGTP

Lane 4 shows the products

and the mixed sequence

duplex

Lane 1 shows the 15-mer

(Fig. 1, top). Lanes 2 and 3 show the products

when all four dNTPs omitted

ofnt to the blunt-end

oligo template.

were

out over-

at 400 nM in all reactions

except that shown in lane 5 in which 200 pM dCTP, dTTP, and dGTP were used. The 9-mer (5’-CGGCCCATC) concentration was 47 PM.

that terminate with different 3’ nt. In earlier work it was shown that the 3’ dGMP-terminated + 1 product migrated more slowly than + 1 products terminated with other nt (Clark et al., 1987). Moreover, dGTP is used more efIi-

79

ciently than either of the pyrimidine nt by PolIk to carry out the + 1 addition reaction (unpublished data). Taken together, these observations suggest that a substantial fraction of the + 1 products synthesized in the absence of dATP terminates with a 3’ dGMP residue (Fig. 4, lane 5), thereby providing complementarity with the oligo of mixed sequence. Therefore, factors other than noncomplementarity must contribute to the reduction in discontinuous ternplating observed in the absence of dATP. Since the unlinked oligo contains a single thymine residue at the penultimate position from the 3’ end, it is likely that templated synthesis occurs most efficiently when the unlinked oligo is oriented in such a way as to pair its 3’ end to the + 1 nucleotide of the (extended) duplex, i.e., 5’-1ZP-GTCCGTCTCTGCCTCG CAGGCAGAGACGGAG 3’-

l

+ 3’-CTACCCGGC.

“P-GTCCGTCTCTGCCTCG CAGGCAGAGACGGAGCTACCCGGC-5’

-3’

(d) Conclusions (1) A 3’-exonuclease-deficient form of E. coli PolIk is capable of using small, unlinked oligos as templates for the addition of multiple nt to the 3’ end of a blunt-ended DNA duplex. The most likely mechanism for this novel synthesis is a polymerase-mediated juxtaposition of the unlinked homopolymers with the blunt-end duplex in such a way as to permit the unlinked oligo to serve as a template, in effect, DNA synthesis on a discontinuous template. Alternative possibilities such as synthesis by anomalous self-priming or on substrates formed by annealing of the homopolymer to the labeled strand of the duplex were eliminated by the control experiment with an ss primer. Therefore, continuity of the template strand is not absolutely required for template-directed synthesis by this 3’-exonuclease-deficient form of PolIk. (2) Discontinuous ternplating probably proceeds through a + 1 intermediate (Fig. 1, right branch) (presumably generated by the nontemplated pathway) in which the first nt added must be complementary to the 3’ nt of the unlinked oligo. (3) The efliciency of discontinuous ternplating appears to increase with increasing length of the unlinked homopolymer (Fig. 2) perhaps because of a greater number of stabilizing contacts between the polymerase and the longer oligos. Consistent with this hypothesis, Allen et al. (1989) have shown that contacts between PolIk and the ss template DNA extend at least eleven nt distal to the primer terminus. (4) The band intensities of the products synthesized on the mixed sequence template varied in a quasi-periodic manner, with clusters of more intense bands followed by regions of less intense bands (Fig. 4). A similar periodicity

was observed

when homopolymers,

notably

pd(G),,

and

were used as templates (data not shown). The pd(T),,, reason for this periodicity is unknown but may reflect a shift by the enzyme from distributive to semi-processive synthesis as nt are added during each cycle. Termination of synthesis at some sites on a mixed-sequence oligo could also result from insertion of an incorrect nt to form a mismatched primer terminus that cannot 3’-exonuclease-deficient PolIk. (5) The biological significance of continuous templates is unclear. The machinery is unlikely to encounter a

be extended

by the

synthesis on discellular replication blunt-ended DNA

substrate during the process of normal semiconservative DNA synthesis. It is possible that such ends could be formed during recombination or as a consequence of DNA damage introduced, for example, by ionizing radiation. In this context, the reaction described here could be considered an extreme example of ‘recombinational synthesis’ in which two DNA molecules (one duplex, the other ss), capable of forming only a single bp overlap, are joined together. (6) Extracts from Xenopus laevis eggs efficiently rejoin the ends of linear DNA molecules containing mismatched termini in vitro (Pfeiffer and Vielmetter, 1988). More recently, these workers have shown that molecules containing one blunt end and one protruding 3’ ss end can be rejoined in a reaction in which sequence information at the junction is preserved by ‘fill-in’ DNA synthesis prior to ligation (Thode et al., 1990). Since the end having a protruding 3’ extension does not provide the appropriate recessed 3’ OH terminus required by all known DNA polymerases (Kornberg, 1980) it was suggested that the exposed 3’ ss is juxtaposed to the blunt end by an ‘alignment protein’ and that synthesis then proceeds from the 3’ OH group on the blunt end even though the two ends are not covalently linked (Thode et al., 1990). This reaction appears to be formally equivalent to the synthesis on a discontinuous template reaction carried out by the 3’-exonucleasedeficient PolIk. Mammalian cells also appear capable of rejoining transfected DNA molecules with mismatched termini in a manner analogous to that observed with X. laevis extracts (Roth and Wilson, 1988). In any event, the mutant PolIk clearly has an unexpected ability to join unlinked molecules together, in a reaction that could be exploited as an alternative to conventional methods for G-C tailing or linker addition in cloning experiments.

ACKNOWLEDGMENTS

I thank Dr. Cathy Joyce for generously providing the exonuclease-deficient PolIk. I am also grateful to Drs. Ken Tindall and Tom Kunkel for critical reading of the manuscript.

80 Derbyshire,

REFERENCES

V., Freemont,

J.M., Joyce, Allen, D.J., Darke,

P.L. and Benkovic,

and deoxynucleotide with the large merase Clark,

triphosphates:

(Klenow)

I. Biochemistry

J.M.:

Novel

catalyzed

S.J.: Fluorescent preparation

by procaryotic

nucleotide

and eucaryotic

poly-

elongation

by DNA

polymerase

addition

reactions

DNA polymerases.

Nucleic

glycol lesions terminate

I in vitro.

Nucleic

Acids

chain Res.

14

Clark, J.M. and Beardsley, on

DNA

G.P.: Functional synthesis

in vitro.

effects of cis-thymine Biochemistry

26

catalyzed

G.P.: Novel blunt-end

by DNA polymerase

Biol. 198 (1987) 123-127.

Freeman,

San Francisco,

of Escherichia coli DNA polymerase

of large fragment with dTMP. strand

Nature

breaks

I. Science CA, 1980.

glycol

American pp. 621-653.

addition

I of Escherichia coli. J. Mol.

Structure

I complexed

313 (1985) 762-766. W.: Joining of nonhomologous

in vitro. Nucleic

Roth, D. and Wilson, J.: Illegitimate

(1987)

5398-5403. Clark, J.M., Joyce, C.M. and Beardsley, reactions

A.: DNA Replication.

In: Kucherlapati,

(1986) 737-749. lesions

Kornberg,

Pfeiffer, P. and Vielmetter,

G.P.: Thymine

and crystallographic

site of DNA polymerase

Ollis, D.L., Brick, P., Hamlin, R., Xuong, N.G. and Steitz,T.A.:

Acids Res. 16 (1988) 9677-9686. Clark, J.M. and Beardsley,

M.R., Beese, L., Friedman,

T.A.: Genetic

240 (1988) 199-201.

and their interaction

28 (1989) 4601-4607.

non-ternplated

Steitz,

studies of the 3’ ,5’-exonucleolytic

oligonucleotides

of Escherikhiu co/i DNA

fragment

P.S., Sanderson,

C.M. and

DNA double

Acids Res. 16 (1988) 907-924. recombination

in mammalian

ceils.

R. and Smith, G.R. (Eds.), Genetic Recombination.

Society

for

Microbiology,

Washington,

DC,

1988,

Thode, S., Schafer, A., Pfeiffer, P. and Vielmetter, W.: A novel pathway of DNA end-to-end joining. Cell 60 (1990) 921-928.