that specifically cleaves Holliday junctions - NCBI

2 downloads 0 Views 2MB Size Report
Oct 7, 1994 - targeted to junctions (Sharples et al., 1994; Whitby et al.,. 1994). As with ruvA and ruvB, null mutations in recG confer only a slight deficiency in ...
The EMBO Journal vol.13 no.24 pp.6133-6142, 1994

Processing of intermediates in recombination and DNA repair: identification of a new endonuclease that specifically cleaves Holliday junctions Gary J.Sharples, Sau N.Chan, Akeel A.Mahdi, Matthew C.Whitby and Robert G.Lloyd1 Department of Genetics, University of Nottingham, Queens Medical Centre, Nottingham NG7 2UH, UK 'Coffesponding author Communicated by S.West

The formation and subsequent resolution of Holliday junctions are critical stages in recombination. We describe a new Escherichia coli endonuclease that resolves Holliday intermediates by junction cleavage. The 14 kDa Rus protein binds DNA containing a synthetic four-way junction (X-DNA) and introduces symmetrical cuts in two strands to give nicked duplex products. Rus also processes Holliday intermediates made by RecA into products that are characteristic of junction resolution. The cleavage activity on X-DNA is remarkably similar to that of RuvC. Both proteins preferentially cut the same two strands at the same location. Increased expression of Rus suppresses the DNA repair and recombination defects of ruvA, ruvB and ruvC mutants. We conclude that all ruv strains are defective in junction cleavage, and discuss pathways for Holliday junction resolution by RuvAB, RuvC, RecG and Rus. Key words: E. coli/Holliday junctions/resolvase/Rus protein

Introduction Recombination is a fundamental process in biology that serves continually to shape and reshape the genomes of all organisms and to promote repair of damaged DNA. The genetic control of this process has received particular attention in Escherichia coli where a large number of genes have been identified; the activities associated with their products can be related to specific stages in repair and recombination (Kowalczykowski et al., 1994). Recombination in wild-type E.coli shows an absolute requirement for the product of the recA gene, an inducible protein that also functions as an activator of the SOS response to DNA damage (Clark and Margulies, 1965; Walker, 1984). RecA polymerizes on DNA to form a helical nucleoprotein filament which promotes homologous pairing and strand exchange with a homologous duplex (West, 1992). In vitro, and presumably in vivo, the assembly of the filament requires a region of singlestranded DNA (ssDNA) to initiate RecA binding; however, once formed, the filament can extend to adjacent duplex regions. Pairing occurs within the filament. When homologous contacts have been established between two duplexes and free ends are available, RecA catalyses a

reciprocal exchange of single strands to form a heteroduplex joint, or Holliday junction, in which the paired molecules are held together by the exchanged strands, an intermediate first proposed by Holliday (1964). Recent studies have identified three E.coli proteins, specified by ruvA, ruvB and ruvC, which process these intermediates into mature products. The RuvA and RuvB proteins together (RuvAB) catalyse branch migration of the junction, while RuvC resolves the joint molecule into duplex products by cleavage across the point of strand exchange (Connolly et al., 1991; Dunderdale et al., 1991; Iwasaki et al., 1991, 1992; Shiba et al., 1991; Parsons et al., 1992; Tsaneva et al., 1992; Parsons and West, 1993). The molecular mechanisms involved are beginning to emerge. RuvA is a highly specific DNA binding protein whose function is to target RuvB to the junction, while RuvB itself assembles around DNA as hexamer rings that are thought to translocate along DNA using the energy derived from hydrolysis of ATP (Stasiak et al., 1994; West, 1994). The mechanism of branch migration by RuvAB is still not clear, but is thought to involve a specialized form of DNA helicase activity that partially unwinds duplex DNA at the junction (Tsaneva et al., 1993; Stasiak et al., 1994). RuvC is an endonuclease and, like RuvA, has a high affinity for Holliday junctions. Cleavage of the junction involves a dual incision mechanism in which ligatable nicks are introduced into two strands of the same polarity (Bennett et al., 1993). RuvC shows a distinct sequence specificity for junction cleavage and it is possible that the primary function of the branch migration activity of RuvAB is to drive the junction, with or without RuvC, to the preferred sequence (Bennett et al., 1993; West, 1994). The properties of RecA, RuvAB and RuvC suggest a model of recombination in which recombinant chromosomes arise through the formation and subsequent resolution of Holliday junctions. However, null mutations in any of the ruv genes reduce the frequency of recombination by no more that 2- to 3-fold (Otsuji et al., 1974; Lloyd et al., 1984). Holliday junctions are either unessential intermediates in E.coli recombination, or they can be processed by alternative proteins. We described recently an activity encoded by the recG gene that may be able to replace RuvAB. The 76 kDa RecG protein is a DNAdependent ATPase that binds specifically synthetic Xjunctions, and drives branch migration of these junctions and Holliday intermediates made by RecA (Lloyd and Sharples, 1993a,b; Whitby et al., 1993). The mechanism of branch migration appears similar to that of RuvAB in that it seems to involve a specialized helicase activity targeted to junctions (Sharples et al., 1994; Whitby et al., 1994). As with ruvA and ruvB, null mutations in recG confer only a slight deficiency in recombination, but ruvA recG and ruvB recG double mutants are extremely

6133

G.J.Sharples et al.

deficient, supporting the notion that RuvAB and RecG provide alternative activities for branch migration of junctions (Lloyd, 1991; Lloyd and Buckman, 1991). Surprisingly, the inactivation of recG also blocks recombination in ruvC strains, although RecG itself does not cleave junctions (Lloyd, 1991; Lloyd and Sharples, 1993a). A possible solution to this apparent paradox was provided by the observation that when RecG is added to an in vitro strand exchange reaction being catalysed by RecA, it blocks the reaction by driving branch migration in the reverse direction to that initiated by RecA (Whitby et al., 1993). Presumably, RecA is in some way providing an asymmetry at the junction which is recognized by RecG. RuvC and RecG may therefore provide alternative ways of removing junctions. Reverse branch migration by RecG would avoid unnecessary crossovers during recovery from UV damage when strand exchanges between daughter chromosomes provide the means to resume DNA synthesis and to close any gaps left in the template DNA (West et al., 1981; Echols and Goodman, 1991). This possible role for RecG would be consistent with the extreme UV sensitivity of ruv recG strains compared with ruv or recG single mutants (Lloyd, 1991). In certain circumstances, reversing branch migration may also help eliminate unproductive exchanges in genetic crosses (Ryder et al., 1994). A different interpretation of the synergism observed between ruv and recG mutations has emerged from studies of a suppressor mutation called rus-l that restores UV resistance to ruv strains, and which also eliminates any deficiency in recombination, particularly in the recBC sbcBC genetic background (Mandal et al., 1993). DNA cloning and sequencing revealed that rus-l is an insertion of a 1.4 kb DNA element upstream of an open reading frame encoding a polypeptide of 14 kDa. Multicopy plasmids that overexpress this protein alone, which we call Rus, suppress all the ruv mutations we have tested, from which we deduce that the rus-] insertion activates transcription of the normally poorly expressed rus reading frame (A.A.Mahdi, T.N.Mandal, S.N.Chan, G.J.Sharples and R.G.Lloyd, manuscript in preparation). To account for the ability of rus-l to alleviate defects in either RuvAB or RuvC, or both, Mandal et al. (1993) hypothesized that the rus region encodes a Holliday junction resolvase, and that ruvA, ruvB and ruvC mutants are all deficient in junction cleavage. RecG could possibly provide branch migration for this new resolvase in ruv mutant strains. The recombination deficiency and extreme UV sensitivity of ruv recG strains could then be explained by the inability to process Holliday junctions by either branch migration or cleavage. In this paper, we show that Rus is indeed a resolvase that cleaves Holliday junctions in a manner remarkably similar to RuvC.

Results Overexpression and purification of Rus protein Rus is encoded by an open reading frame that forms part of a poorly transcribed complex operon (details of the sequence and structural organization of the rus region of the chromosome will be published elsewhere; A.A.Mahdi, T.N.Mandal, S.N.Chan, G.J.Sharples and R.G.Lloyd, manuscript in preparation). To facilitate the purification 6134

of Rus, we made a plasmid construct designed to overexpress the rus reading frame. Two oligonucleotide primers incorporating suitable restriction sites were used to facilitate PCR cloning of rus into the expression vector pT7-7. The resulting construct, pAM 151, fuses the start codon of rus to a consensus ribosome binding sequence and places the gene under the control of the bacteriophage T7 o 10 promoter (Figure 1 A). Nucleotide sequencing of the insert from pAM151 confirmed that the sequence was identical to that of the wild-type gene. When this plasmid was introduced into a AruvAC strain, or into ruvA, ruvB or ruvC strains, it conferred resistance to UV light (Figure 1 B and data not shown). The same plasmid makes a ruv+ strain slightly sensitive to UV light, from which we conclude that too much Rus is detrimental to repair. Since the strains used do not express T7 RNA polymerase, we assume that rus is being expressed from an adventitious promoter in pAM15 1. Overexpression of Rus was achieved by introducing pAM151 into N3757, a derivative of BL21 (DE3) plysS. This strain carries a AruvAC mutation that prevents synthesis of both RuvA and RuvC proteins. A 11 culture of this strain was grown and phage T7 RNA polymerase was induced with isopropyl P-D-thiogalactoside (IPTG) to express rus. A strongly induced protein migrating with an apparent Mr of -15 kDa accumulated to -20-30% total cell protein after 3 h incubation (Figure 1C, lane c). We assumed this protein to be Rus. The cells were lysed and the - 15 kDa protein was purified by chromatography through DEAE Bio-Gel A, phosphocellulose, doublestranded DNA (dsDNA) cellulose and Mono Q matrices, as described in Materials and methods (Figure 1C, lanes d-g). The very strong interactions with phosphocellulose and dsDNA matrices (eluting at 0.8 and 0.5 M KCI, respectively) greatly facilitated purification. The peak fractions eluting from the Mono Q column contained a total of 0.8 mg of the - 15 kDa protein at >99% purity. The N-terminus of the purified protein revealed the sequence MNTYSITLP, which confirmed its identity as Rus (Figure IA).

Specific binding of Rus to synthetic X-junctions The ability of pAM 151 to restore UV resistance to ruv mutants demonstrates very clearly the ability of Rus protein to provide a DNA repair function that can replace activities normally provided by RuvAB and RuvC. Since the Ruv proteins process Holliday intermediates, it was natural to consider that Rus might provide a similar function. To investigate this possibility, we first used a simple bandshift assay to see if Rus could bind a DNA substrate containing a synthetic four-way junction (Xjunction, X-DNA). Previous studies have shown that RuvA, RuvC and RecG form well-defined complexes with a substrate made by annealing four partially complementary oligonucleotides of 49-51 nucleotides to form an X structure, with a central 12 bp homologous core free to branch migrate (Connolly et al., 1991; Parsons et al., 1992; Lloyd and Sharples, 1993a). When Rus protein was mixed on ice with 32P-labelled X-DNA in the absence of divalent cations, several discrete complexes were formed which could be detected by subsequent electrophoresis through a non-denaturing gel (Figure 2, lanes c-g). Rus was also able to bind to a linear duplex control, as

Holliday junction resolution by Rus protein

C

A

=-

-If

. I. Z:

z

3

EU.

--. . .

B AM5S47 .ruvACE ABfl^ ½

,

.n- 1

_

_99-0

K., .

w

.u

. ~ 4. 110 1

_

.LL.

0 .:

m4

R

'k

-.

C 00. ,f

IJI.

C. DCCCO1;I .1

-1

UV Dose J

-

--

Fig. 1. Cloning, overproduction and purification of Rus protein. (A) Diagram showing the DNA sequences at the 5' and 3' ends of rus used for construction of pAM151 by PCR cloning. The predicted N-terminal amino acid sequence of the Rus protein is shown above the sequence. The Ndel and HindIll sites do not correspond to the rus sequence and were introduced into the PCR primers to facilitate cloning. The italicized nucleotides are pT7-7 vector sequences linking the rus gene to the T7 l10 promoter and ribosome binding site (RBS). (B) Effect of pAM151 on the UV sensitivity of strains AM547 (AruvAC65) and ABI 157 (ruv+). The same strains transformed with pT7-7 vector were used to provide controls. (C) Overproduction and purification of Rus protein. Lane a, molecular weight markers (Pharmacia); lane b, uninduced N3757 pAM151 cells; lane c, N3757 pAM151 cells induced with IPTG; lane d, pooled fractions from a DEAE column (13 gg protein); lane e, pooled fractions from a phosphocellulose column (6 ,tg protein); lane f, pooled fractions from a dsDNA cellulose column (6 pg protein); lane g, pooled fractions from a Mono Q column (6 tg protein). Proteins were resolved by 13.5% SDS-PAGE and stained with Coomassie brilliant blue.

expected from the procedure used for its purification. However, binding in this case requires a higher concentration of protein. The single complex formed is also much less stable than those formed with the X-junction, and tends to dissociate during the electrophoresis to produce a smear of faster migrating material (Figure 2, lanes 1-n). Densitometric analysis of the exposed X-ray film revealed that at a protein concentration of 16 nM, Rus retarded 50% of the X-junction but only 4% of linear duplex control. A DNA substrate containing a three-way junction (Y-junction, Y-DNA), which has been shown to bind RuvA and RecG (Lloyd and Sharples, 1993b), was also retarded by Rus. The same number of complexes were formed at similar protein concentrations to those used to show binding to the X-DNA (data not shown). To examine further the specificity of binding, 100 nM Rus was incubated with 0.15 ng X-DNA in the presence of increasing amounts of poly(dI-dC)-poly(dI-dC) competitor DNA. The slowest migrating complexes observed in the absence of competitor (Figure 2, lanes e-g) were not observed in the presence of 31 ng of the competitor DNA; the faster migrating complex (Figure 2, lanes c and d) was retained, suggesting that it is much more stable (data not shown). About 10% of the X-junction was retained in this stable complex with Rus when the concentration of competitor DNA was increased to 250 ng (data not shown). These results reveal that Rus has a 10to 100-fold higher affinity for junction DNA over linear

.

cupiex -

^ .: 25--

R7s

-M

-

ex es~

-

X

D

Fig. 2. Bandshift assay showing binding of Rus protein to X-junctions. Reactions (20 id) contained -0.15 ng 32P-labelled X-DNA (lanes a-g) or linear duplex DNA (lanes h-n), and Rus protein as indicated.

duplex DNA. We conclude that Rus is a DNA binding protein that has a specific affinity for junctions.

Rus resolves X-junctions to nicked duplex products The DNA binding reactions were performed on ice. To determine what happens under more physiological conditions, Rus protein was incubated with the X-junction for 60 min at 37°C in the presence of divalent cations. Reactions were deproteinized and the products were separated on a non-denaturing polyacrylamide gel. Under 6135

G.J.Sharples et al.

A

A X Junction

linear

4 ?0 1 0 500 0 4

+

+

+

X Junction

Y

20'.., 50

2:. 52

iM

-

Mn

-

+

+

-

-

a

D

+

d

e

9g

k

a

,.E.7

.! 3a'

c

g

e

r

v

7".

.L 'b

0

X L",

jt,

U. 1)

-r 1-

3J 21 22 23 24 25 26 27 28 29 303 3v

3

o °numer trac br)

Fig. 3. Resolution of a synthetic X-junction. (A) Autoradiograph showing cleavage of X-junction DNA by Rus. Reactions contained -0.15 ng 32P-labelled DNA, Rus protein and either 10 mM MnCI2, MgCI2 or EDTA as indicated. (B) Co-elution of Rus protein with junction-cleavage activity. Fractions from a dsDNA cellulose column were diluted 300-fold in reaction buffer. 2 ,l samples were assayed for cleavage activity in buffer containing 10 mM MnCl2 as described in (A). The amount of linear product was determined using laser densitometry (0). Total protein (0) in each fraction was determined using the Bradford method, with BSA as a standard.

these conditions, the X-junction was converted to a DNA product that migrated to the same position as a linear duplex control (Figure 3A, lanes b-e and h-j). Examination of the reaction requirements revealed that Mn2+ was -7-fold more effective than Mg2+ (Figure 3A, compare lanes b-e with lanes h-j). No product was detected in the absence of divalent cations (lane k), when the X-DNA was replaced with linear duplex DNA, or with a substrate containing a three-way junction (lanes g and m). A DNA product with identical mobility to that produced by Rus was detected when the X-junction was incubated with RuvC. The reaction conditions, including the stimulatory effect of Mn>, and substrate specificity were very similar for the two proteins (data not shown). We suspected therefore that Rus was an endonuclease that cleaves junction DNA to produce a nicked duplex product. Since Rus was purified from a strain carrying a deletion of ruvC, we could be certain that the activity we detected was not due to contamination with RuvC. However, as with RuvC, the specific activity of Rus is very low and we considered the possibility that the cleavage product detected was produced by a contaminating endonuclease with much higher activity. We therefore examined cleavage activity in protein fractions eluting from the columns used to

6136

c

B

40 -

0

slrand 4 GA R T C

m

B

(-

s'ard 3 G ARTC

nETA

-

i

s47and 2 GA R TC

2M

,

-

c

s ran,d 1 GA R TC

Fig. 4. Mapping Rus cleavage sites in a synthetic X-junction. (A) Autoradiograph showing reaction products (R) from Rus cleavage reactions in Mg2+ buffer with four separate X-junctions, each labelled on a different strand (strands 1-4). In each case, the reaction product is flanked by G+A and T+C Maxam-Gilbert sequence ladders for the labelled strand. Only the relevant part of the gel is shown. (B) Sequence of the X-junction showing the position of Rus cleavage sites. The boxed regions are heterologous sequences that flank the 12 bp central region of homology. The cleavage sites detected in buffer containing Mg2+ are marked with a solid arrowhead. Open arrowheads mark additional minor cleavage sites detected in reactions containing Mn2+.

purify Rus. The peak of activity coincided exactly with the peak of Rus protein in the fractions taken from both the phosphocellulose (data not shown) and dsDNA cellulose columns (Figure 3B). We conclude that the cleavage activity is due to Rus. Rus cleaves X-junctions by introducing symmetrical nicks in two strands of the same polarity RuvC protein resolves X-junctions by introducing symmetrical nicks in opposing DNA strands (Dunderdale et al., 1991; Bennett et al., 1993). To see if the same was true of Rus, we made four X-junctions, each labelled in a different strand. These were incubated with Rus in reactions containing Mg2+; DNA products were separated on denaturing polyacrylamide gels. Initial studies revealed that Rus cleaves the junction by introducing single nicks into strands 2 and 4; no cleavage was detected in strand 1 or 3 (data not shown). The positions of these nicks were located using A+G and T+C Maxam-Gilbert sequencing reactions of each labelled oligonucleotide as markers. The results shown in Figure 4A reveal that the cleavage sites are located symmetrically in strands 2 and 4, and occur at the same location (5'-GT4-CC-3') as those generated by RuvC (Figure 4B). The identity of the cleavage sites

Holliday junction resolution by Rus protein

was confirmed by running the Rus and RuvC reaction products in parallel (data not shown). These were the only cleavage sites detected with Rus in buffer containing Mg2 . In the presence of Mn2+ we detected the same major cleavage sites. However, in this case we also detected minor symmetrical cleavages at 5'-TGI1CT-3' and 5'-GGIkAC-3' in strands 1 and 3 (Figure 4B; data not shown), and a trace of symmetrical cleavage at the adjacent 5'-GA4-CA-3' and 5'-AC&AT-3' sequences in these same strands. These minor cleavage sites are different from any reported for RuvC, and show no requirement for a thymine residue 5' of the cleavage point; a feature that is characteristic of RuvC (Bennett et al., 1993). All the Rus cleavage sites we have detected reside in the region of the homologous core. From these data, we conclude that Rus, like RuvC, is a junction-specific endonuclease.

Nicks introduced into X-junctions can be resealed by Ecoli DNA ligase If Rus is to act as a resolvase of Holliday junctions in vivo, the nicks introduced into the DNA by symmetrical junction cleavage should be re-ligatable (Bennett et al., 1993). To determine if this is the case, X-DNA labelled on strand 2 was incubated at 37°C with 100 nM Rus in ligation buffer containing 9 mM MgCl2. After 45 min, various amounts of E.coli DNA ligase were added and incubation continued for a further 30 min. Samples were then deproteinized and the reaction products were analysed by electrophoresis on a denaturing gel. In the absence of DNA ligase, a single Rus-generated cleavage product of 23 nucleotides is detected (Figure 5A, lane b). When increasing amounts of DNA ligase are added after cleavage, there is a corresponding increase in the amount of the full-length strand of 50 nucleotides (Figure 5A, lanes c-g). In all, 86% of the cleavage product was converted to the fulllength oligonucleotide using 1 U of DNA ligase (Figure SB). The results demonstrate that the Rus cleavage product is a nicked strand with 5' phosphate and 3' hydroxyl residues that can be repaired by DNA ligase. Resolution of Holliday junctions made by RecA We next examined the ability of Rus to resolve Holliday junctions made by RecA. Junctions were made using a standard strand exchange (four-strand) reaction between gapped circular duplex and 32P-labelled linear duplex substrates based on OX 174 DNA. In these reactions, partial strand exchange produces a-structures containing a single Holliday junction (Figure 6A). Cleavage of the junction in the a + c orientation is expected to resolve the aintermediate into a linear dimer DNA species, while cleavage in the b + d orientation is expected to give nicked circular and linear products, of which the latter migrates in the same position as the linear substrate DNA (Dunderdale et al., 1991; Muller et al., 1990, 1992). We first examined the ability of Rus to cleave RecA-free intermediates. The substrates were incubated for 20 min with RecA and the intermediates were then deproteinized and purified as described in Materials and methods (Figure 6B, lane a). These intermediates were then incubated for 30 min at 37°C with increasing amounts of Rus, or with RuvC as a control, in reaction buffer containing 10 mM MgCl2. With Rus, both linear dimer and nicked circular

A

ceo

-Z fz. GO r--

IGOli,O

CD

G C

GO DNA ligase +

Rus

-

--

fL.I1 lengst 5C oases

cea,age

iica' c-

-

_

_

-

cieavage product I

a

b

z

d

e f

o ases

g

B 8C

.

4A

--

-

2. 9

C 2_. 92

08 A4 6c gase -units)

19!

Fig. 5. Re-ligation of X-junctions cleaved by Rus. (A) Autoradiograph showing re-ligation of the nicked duplex product of the cleavage reaction by Ecoli DNA ligase. The junction labelled in strand 2 (lane a) was cleaved with Rus (lane b) and then incubated with increasing concentrations of ligase, as indicated (lanes c-g) before denaturing the DNA and separating the products on a denaturing gel. (B) Quantification of the autoradiograph shown above by laser densitometry.

products were detected, showing that it can cleave Holliday junctions in either of the two possible orientations (Figure 6B, lanes b-g). From the relative amounts of these two products, it appears that Rus has no particular bias. The control reactions with RuvC (Figure 6B, lanes h-m) revealed a bias towards the linear dimer product, as reported previously (Dunderdale et al., 1991). RuvC also appears to be more active than Rus in these reactions. Similar results were obtained when Rus or RuvC were added directly to the four-strand reaction 30 min after RecA. In both cases, the formation of the linear dimer band provides clear evidence of junction resolution in the presence of RecA (Figure 6C). Again, RuvC seems more active than Rus (compare lanes c-e with f-h). The lower activity for Rus detected in these experiments is in contrast to the results obtained with synthetic X-junctions. Since most of the OX174 substrate is in the form of linear duplex DNA, to which Rus binds more effectively than RuvC, this difference may simply reflect the amount of Rus free to cleave junctions. It is clear, nevertheless, that Rus is able to resolve Holliday junctions made by RecA. Rus was also tested on deproteinized intermediates from a RecA strand exchange between linear duplex DNA and circular ssDNA (three-strand reaction; West, 1992). Neither Rus nor RuvC had any detectable activity on these intermediates (data not shown), although previous studies have shown that RuvC will cleave deproteinized

6137

G.J.Sharples et aL

A

B

RuvC(

Rus

___

gNA

L n) C)

U;) . (J

C

.

LiU

LU

.-

.-

-J) '

-

C_)

C

LU

Y-

L- .n C\

Y

!:'!near DNA

tlUf)] rIm - ongiri

RecA

inintermediates

_rcked circUl IC Jer

F ..

r

I I%

R.IvC or RLJS

RecA Rus RuvC

C(

0 .ia. cI

....

n.ea If'

,

_

,r=

)

.;

-ILi

-'

-

RuvCl riN oricrirl

prod;.;cD,s

v

V -^

H- inte,rmediates

,...- linear diriler nicked cIrcLJ !ar _inear -.

.1., -.k-ec 1,

S rr -

w

r5:. ...

-_ a

linrear dimr .-r

linear

b

-."-__ c

d

e

f

g

DNA

Fig. 6. Resolution of Holliday junction intermediates made by RecA in vitro. (A) Diagram showing stages of the strand exchange and resolution reactions. The substrates in the reaction are gapped circular DNA (gDNA) and 32P-labelled (*) linear DNA. Complete strand exchange (not shown) generates labelled nicked circular and gapped linear DNA. Cleavage in the a-c orientation gives labelled linear dimer DNA, which is diagnostic of 'splice' resolution. Cleavage in the b-d orientation gives labelled nicked circular and gapped linear DNAs, which are diagnostic of 'patch' resolution, but are indistinguishable from the products of complete strand exchange. (B) Autoradiograph showing resolution of RecA-free intermediates. Deproteinized recombination intermediates (5 ng total DNA in a 20 gl reaction) were incubated at 37°C for 1 h with Rus (lanes b-g) or RuvC (lanes h-m), as indicated. (C) Autoradiograph showing resolution of recombination intermediates during RecA-mediated strand exchange. Strand exchange was initiated by the addition of 8 gM RecA protein to a reaction (20 ,l final volume) containing 200 ng gDNA and 50 ng linear duplex DNA, followed by incubation at 37°C for 30 min before adding various amounts of Rus (lanes d-f) or RuvC (lanes f-h) and incubating for a further 30 min before analysing the products.

three-strand intermediates where the junction point has been driven to a heterologous block (Benson and West, 1994). From these studies, we conclude that Rus activity is directed specifically at a Holliday junction.

Discussion We have described a new E.coli activity capable of resolving Holliday intermediates by junction cleavage. The identification of Rus as a resolvase is supported by: (i) the fact that it forms specific protein-DNA complexes with substrates containing a four-way junction; (ii) its ability to convert these X-junctions to duplex products by cleaving two strands of similar polarity in a symmetrical fashion, leaving nicks that can be repaired by DNA ligase; and (iii) the fact that it can resolve Holliday intermediates made by RecA in either of the two possible orientations to generate 'patch' or 'splice' products. It explains neatly why activation of rus expression is able to suppress the defects in DNA repair and recombination associated with strains lacking the activity of RuvC (Mandal et al., 1993). The resolvase activity of Rus is in fact very similar to that of RuvC. This is particularly true in terms of its sequence specificity. In the vast majority of cases, both 6138

proteins resolve the X-junction used in these studies by cleaving the same two strands at precisely the same location. We do not know whether this sequence specificity is a property of the junction or a common feature of the two proteins. DNA molecules containing a four-way junction fold in a particular stacked configuration in the presence of divalent cations (Duckett et al., 1990). This folding may dictate how and where the junction is cleaved. However, recent studies with RuvC indicate that protein binding alters the folding of the junction. These studies have also revealed a distinct preference for cleaving strands 3' of a T residue (Bennett et al., 1993). It remains to be established whether cleavage occurs within a particular sequence motif. Sequence specificity of strand cutting would perhaps explain why both Rus and RuvC can resolve junctions in Holliday intermediates made in oX174 DNA in either of the two possible orientations with reasonable frequency. In this case, the junctions are located at random along the -5 kb length of the DNA and any preferred sequence could be present in either of the two strands. With the X-DNA used, junction cleavage is restricted to sequences within the 12 bp homologous core. Sequence specificity is also consistent with the fact that the activity of both Rus and RuvC on X-junctions is

Holliday junction resolution by Rus protein strongly enhanced by replacing Mg2+ in the reactions with Mn2+ (this work; Shah et al., 1994). Under these conditions cleavage occurs at other sequences much more readily. With Rus, these additional cleavage sites show no obvious sequence preference; cuts were observed 3' to all four nucleotides. This relaxed sequence specificity in the presence of Mn2+ is also seen with other enzymes that cut DNA. The restriction endonuclease EcoRV normally has an extremely high specificity for 5'-GATATC-3', but will cut at related sequences in reaction buffer containing Mn2+ (Vermote and Halford, 1992). The low specific activity of X-junction cleavage observed with both Rus and RuvC may therefore reflect the absence of the most favoured recognition sequence, or the failure to present this sequence in the correct position relative to the crossover point. It would be remarkable if Rus and RuvC had exactly the same sequence preference for cutting at junctions. Both proteins are small and highly basic, but otherwise show no obvious sequence similarity to suggest a common origin (A.A.Mahdi, T.N.Mandal, S.N.Chan, G.J.Sharples and R.G.Lloyd, manuscript in preparation). There are also significant differences between the two activities. In the presence of Mg2+, Rus cuts the X-junction used at only one site, while RuvC cuts at additional sites, albeit at low efficiency. Rus binds linear duplex DNA much more readily then RuvC, and shows a higher affinity for YDNA. It also forms multiple complexes with X-DNA, whereas RuvC consistently produces a single complex (Dunderdale et al., 199 1; Iwasaki et al., 1991). Nevertheless, both proteins show a clear specificity for cleaving four-way junctions. Neither cuts DNA containing a threeway junction. This specificity distinguishes Rus and RuvC from other enzymes such as T4 endonuclease VII and T7 endonuclease I that also resolve Holliday junctions. These phage enzymes act on a much broader range of substrates and probably function to remove branched structures from DNA prior to packaging into phage particles (de Massy et al., 1987; Dickie et al., 1987; Kemper et al., 1990). The identification of Rus as a Holliday junction resolvase able to suppress the DNA repair defect in AruvAC strains confirms the hypothesis of Mandal et al. (1993) that ruvA, ruvB and ruvC mutants are deficient in the resolution of junctions by endonuclease cleavage. Activation of Rus synthesis by the rus-J mutation, or by cloning the rus sequence in a multicopy plasmid, restores UV resistance to all ruv mutants and eliminates any deficiency in recombination (Mandal et al., 1993; A.A.Mahdi, T.N.Mandal, S.N.Chan, G.J.Sharples and R.G.Lloyd, manuscript in preparation; and this work). This observation leads to the inescapable conclusion that junction cleavage by RuvC depends normally on the activity of RuvAB. Why this should be is not clear. RuvAB may be needed to translocate RuvC to the junction, or to move the junction to a specific DNA sequence that is then recognized and cleaved by RuvC. Alternatively, RuvAB may be needed to remove RecA from the junction before RuvC can act (Stasiak et al., 1994). Although RuvC has been shown to resolve junction intermediates in the presence of RecA in vitro (Dunderdale et al., 1991), it is not known whether the junctions cleaved are free of RecA or embedded in a RecA-DNA filament. The studies reported here and by Mandal et al. (1993)

provide strong support for recombinational models of DNA repair that involve the formation and subsequent processing of Holliday junctions. Howard-Flanders and colleagues were the first to propose the idea that strand exchange between daughter chromosomes could help to close gaps left in the newly synthesized strands following replication of a DNA duplex containing UV-induced lesions (Rupp and Howard-Flanders, 1968; Rupp et al., 1971). When the replisome encounters a pyrimidine dimer, DNA synthesis comes to an abrupt halt, but resumes again after a short delay by processes that involve RecA (Khidhir et al., 1985). Several mechanisms have been proposed (Echols and Goodman, 1991; Asai et al., 1993), but the

molecular details remain uncertain. One possibility is that resumes downstream, leaving a gap opposite the lesion in the template strand (Rupp and HowardFlanders, 1968). RecA could polymerize at the gap and promote homologous pairing and strand exchange with the intact sister duplex. Strand transfer past the lesion into the adjacent duplex DNA could then close the gap and allow excision enzymes to remove the lesion (HowardFlanders et al., 1984). This would leave the DNA with a single Holliday junction to be resolved before repair could be completed. RuvC could cleave the junction. Alternatively, RecG could remove the junction by driving branch migration in the reverse direction to restore the exchanged strands to their original partners (Whitby et al., 1993). This model accounts for the extreme UV sensitivity of ruv recG double mutants (Lloyd, 1991). In the absence of the Ruv and RecG proteins, Holliday junctions would remain unresolved. We assume the level of Rus is normally too low to make a significant impact. However, it is clear that activation of rus expression increases the UV resistance of ruv recG strains (Mandal et al., 1993). This is a critical observation which demonstrates unambiguously that junction cleavage by Rus can promote repair in the absence of RuvAB and RecG proteins. Presumably, the strand exchange and branch migration activities of RecA alone can generate the appropriate intermediates for repair. It is possible that RecA is assisted in this role by one of the numerous other helicases in E.coli (Matson and Kaiser-Rogers, 1990). The Dda helicase of bacteriophage T4 has been shown to enhance branch migration by UvsX, an analogue of RecA (Kodadek and Alberts, 1987). Dda assists in the reaction by aiding UvsX filament assembly on ssDNA rather than by direct interaction with the Holliday junction. We noticed, however, that a multicopy rus+ plasmid makes a wild-type strain somewhat sensitive to UV light. Overexpression of Rus may tend to resolve junctions before strand exchange has had a chance to properly close gaps in DNA. The discovery of Rus also has important implications for recombination that question a simple pathway based on strand exchange by RecA, branch migration by RuvAB and junction resolution by RuvC. In strains expressing Rus, we can envisage four possible pathways for processing recombination intermediates made by RecA (Figure 7). Three of these can serve to promote recombination by resolving junctions into 'patch' or 'splice' products. For the sake of simplicity, only 'patch' products are shown. The reverse branch migration pathway catalysed by RecG also removes the junction, but is essentially antirecombinogenic, though under certain circumstances it

replication

6139

G.J.Sharples et al. reverse branch migration

.~ ~.

brarnch rmiorat 3n and

,.

,, F77

/...........-..____

RecG \ :.....:

RecG/Rus

::

.. tn.l-..-..-. .:...........................:....:..I .M =

_ 1 ''.:'.'.: ': .. .-' F'.'::-...........

branch mig at

-dd pleavaqa

,,

cleavage

2,,,,.72.n

........Ru .ABC

'RuvABC ..:

Rus 4k

_;

=......... C!@ Vnne

Fig. 7. Possible pathways for processing Holliday junction in Ecoli. A Holliday junction formed by RecA-mediated homologous pairing and strand exchange, from left to right, is shown in the centre. With RecA present, RuvAB promotes branch migration in the same direction as RecA strand exchange and resolution occurs by RuvC cleavage (top right). RecG drives the Holliday junction backwards in the presence of RecA (top left), but could drive the reaction forward when RecA is released from the junction. This forward branch migration could result in resolution by Rus protein (bottom left). Rus cleavage alone depends on strand exchange and branch migration by RecA (bottom right). Endonuclease resolution is shown as a patch but could equally be a splice.

may help to promote recombination by directing DNA substrates towards other mechanisms of recombination that do not involve junction intermediates (Ryder et al., 1994). Also, gene conversion could occur in this situation if mismatch repair of heteroduplex DNA precedes the reversal of branch migration. Mismatch repair may be sufficiently rapid for this to be a realistic possibility, at least in yeast (Haber et al., 1993). However, reverse branch migration would be able to remove junctions only if nicks persist in the exchanged strands. In E.coli, these nicks might also be expected to direct the mismatch repair specified by MutHLS to the invading strand, thus preventing conversion (Modrich, 1987; Rayssiguier et al., 1989). Given that RuvAB is needed for RuvC cleavage activity in vivo, there is some ambiguity about the role of branch migration and its catalysis. Extensive branch migration may be required to repair gaps in damaged DNA, and a specific branch migration protein may help drive strand exchange across lesions (Bianchi and Radding, 1983; Tsaneva et al., 1992), but it is not obvious why in genetic crosses heteroduplex joints would have to extend much beyond the estimated 30-50 bp minimum needed to establish homology (Watt et al., 1985; Shen and Huang, 1986), RecA alone should have no difficulty driving strand exchange this distance. However, there is no doubt that RecG is essential for recombination in the absence of RuvABC (Lloyd, 1991; Mandal et al., 1993), which implies that activities specific to branch migration do have a role to play. RecG may in this instance function with Rus to drive forward branch migration and resolution, but there is no evidence of direct interplay between these proteins, as is possibly the case with RuvAB and RuvC. As with repair, activation of Rus by rus-l stimulates recombination in ruv recG strains by at least 20-fold

6140

(Mandal et al., 1993). Rus must therefore be able to act with RecA to promote recombination in the absence of any other recognized branch migration activity, and a question mark must remain against the idea of forward branch migration by RecG followed by Rus cleavage. The question also remains as to how recombinants are produced in ruv mutants. Since these strains are sensitive to UV light, there is clearly not enough Rus to promote efficient repair. However, a low level of this protein may be sufficient to catalyse the few exchanges needed to produce recombinant chromosomes in genetic crosses. We are currently testing this possibility.

Materials and methods Strains and plasmids Ecoli strains AB1 157 (ruv+), AM547 (AruvAC65), AM561 (AruvAC65 eda-51::TnlO) and BL21 (DE3) plysS have been described (Studier and Moffat, 1986; Bachmann, 1987; Mandal et al., 1993). N3757 is a AruvAC eda-51 derivative of BL21 made by transduction with phage P1 from AM561, with selection for resistance to tetracycline. Plasmid pAM149 is a derivative of pUCI9 carrying a 2.4 kb HindIII-XmnI fragment from the Ecoli chromosomal rus+ region located near coordinate 580 on the Kohara et al. (1987) physical map (A.A.Mahdi, T.N.Mandal, S.N.Chan, G.J.Sharples and R.G.Lloyd, manuscript in preparation). The Rus overexpression plasmid pAM151 was made by PCR cloning of the rus reading frame into the pT7-7 expression vector (Tabor and Richardson, 1985). The rus gene was amplified by PCR from pAM 149 using the oligonucleotide primers 5'-GGGTTATTAACATATGAATACCTAC-3' and 5'-TGGCGAAGAAGCTTTGCCATATTAA-3' based on the 5' and 3' sequences, respectively, of the rus gene. These primers incorporated NdeI and HindIII sites (bold) at the 5' and 3' ends, respectively. PCRs followed standard protocols. The product was purified from an agarose gel, cleaved with NdeI and HindIII, and ligated into pT7-7 to give pAM 151, which places rus under the control of the T7 phage p10 promoter and a consensus ribosome binding site (Tabor and Richardson, 1985). The insert from pAM151 was excised using XbaI and HindlIl and inserted into pGEM-7Zf (+ and -) (Promega). The resulting constructs were used to sequence both strands of the insert using a T7 sequencing kit (Pharmacia). DNA manipulations, genetic procedures and methods for measuring sensitivity to UV light followed published protocols (Silhavy et al., 1984; Sambrook et al., 1989; Mandal et al., 1993).

Proteins Ecoli RecA and RuvC proteins were purified as described (Cox et al., 1981; Dunderdale et al., 1991). Rus protein was purified from a 11 culture of strain N3757 carrying the overexpression construct pAM151. Cells were grown at 370C to an A650 of 0.5 in Luria-Bertani broth containing 125 gg/ml ampicillin and 50 sg/ml chloramphenicol. IPTG was then added to 1 mM, and incubation continued for a further 3 h at 37°C. The induced cells were harvested by centrifugation, resuspended in 100 mM Tris-HCI, pH 8.0, 2 mM EDTA, 5% glycerol and stored at -80°C. The cells (3.4 g, wet weight) were then thawed on ice-water, mixed with 0.25 vol. 5 M NaCI, 0.01 vol. 0.1 M dithiothreitol (DTT) and 0.01 vol. 10% Triton X-100, and lysed by three freeze-thaw cycles in liquid nitrogen. The lysate was centrifuged at 40 000 r.p.m. for I h using a Kontron TST 41 rotor; the upper half of the supernatant was retained (Fraction I, 24 ml, 4.64 mg/ml protein), dialysed against buffer A (20 mM Tris-HCI, pH 8.0, 1 mM EDTA, 0.5 mM DTT, 10% glycerol) and loaded onto a 30 ml DEAE BioGel A (Bio-Rad) column. Bound proteins were eluted with a 300 ml, 0-500 mM gradient of KCI in the same buffer. Fractions containing the induced Rus protein, which migrated as an -15 kDa species on SDS gels, eluted with a peak at -300 mM KCI. These were pooled (Fraction II, 50 ml, 0.31 mg/ml), dialysed against buffer A and loaded onto an 8 ml phosphocellulose (Whatman PI11) column. A 150 ml linear gradient of 0- 1000 mM KCI in buffer A was used for elution; peak fractions (eluting at 800 mM KCI) were pooled (Fraction III, 16 ml, 0.24 mg/ml protein), dialysed against buffer A and loaded onto a 4 ml dsDNA cellulose (Sigma) column. Bound proteins were eluted with an 80 ml linear gradient of 0-1000 mM KCI. Rus eluted at 500 mM KCI; fractions were pooled (Fraction IV, 6 ml, 0.17 mg/ml), dialysed against buffer A and loaded

Holliday junction resolution by Rus protein onto a I ml Mono Q column (Pharmacia). Approximately one third of the Rus protein appeared in the wash (Fraction Va). The remainder of the protein bound to the column and eluted in a broad peak at -150 mM KCI (Fraction Vb). Fractions Va and Vb were dialysed against storage buffer (20 mM Tris-HCI, pH 8.0, 1 mM EDTA, 0.5 mM DTT, 50% glycerol) and stored in aliquots at -80°C. Fraction Va consisted of 0.85 ml at a concentration of 0.3 mg/ml; Fraction Vb was 1I.1 ml at 0.5 mg/ ml. Both protein fractions had similar activities. Fraction Va was used in the experiments described here. Protein concentrations were determined by the Bradford method, using a kit from Bio-Rad and bovine serum albumin (BSA; Sigma) as a standard. The sequence of the N-terminus of purified Rus protein was determined using a pulsed liquid amino acid sequencer (Model 473, Applied Biosystems).

DNA substrates Synthetic X-junction, Y-junction and linear duplex DNA substrates were made by annealing synthetic 49-51mer oligonucleotides purified by denaturing 12% PAGE (Parsons et al., 1990). The numbering and sequences of the oligonucleotides used have been described (Connolly et al., 1991; Lloyd and Sharples, 1993a,b). Unless stated otherwise, oligonucleotide 1 was 32P end-labelled at the 5' end using T4 polynucleotide kinase (BRL) and [y-32PIATP (Amersham) before annealing. Concentrations of junction and linear duplex DNAs used as controls were measured using DNA DipSticks (Invitrogen, San Diego, CA). Circular OX]174 viral (+) strand and RF DNA were obtained from PharmaciaLKB. Gapped circular DNA (gDNA) and PstI-linearized duplex DNA were made from the OX1 74 RF DNA as described (Muller et al., 1990). The linear duplex was labelled at the 3' end using terminal transferase and [a-32PJddATP. Poly(dI-dC)-poly(dI-dC) competitor dsDNA (average length 951 bp) was purchased from Pharmacia.

Gel retardation assays DNA-protein interactions were measured using a standard gelshift assay. Reactions (20 ,ul) contained 32P-labelled X-DNA, Y-DNA or linear duplex DNA in 50 mM Tris-HCI, pH 8.0, 5 mM EDTA, 1 mM DTT and 100 tg/ml BSA. After 15 min on ice, 5 Id loading buffer (40 mM Tris-HCI, pH 7.5, 4 mM EDTA, 25% glycerol, 400 ,ug/ml BSA) was added and the samples loaded onto 4% non-denaturing polyacrylamide gels in low ionic strength buffer (6.7 mM Tris-HCI, pH 8.0, 3.3 mM sodium acetate, 2 mM EDTA). Electrophoresis was at 160 V for - 105 min. Gels were dried and subjected to autoradiography. In competition studies, synthetic X-junction and poly(dI-dC).poly(dI-dC) DNAs were mixed before the addition of Rus protein. Assay for cleavage of synthetic Holliday junctions Standard reactions (20 1t) contained 0.15 ng of 5' 32P-labelled junction DNA in 50 mM Tris-HCI, pH 8.0, 1 mM DTT, 10 mM MgCl2 (or MnCl2) and 100 jg/ml BSA. EDTA, when used, was present at 10 mM. Reactions were incubated at 37°C (usually for I h) before adding 5 ,ul stop buffer [2.5% (w/v) SDS, 200 mM EDTA, 10 mg/ml proteinase KI and incubating for a further 10 min at 37°C. The DNA products were then electrophoresed at 160 V for 150 min through a 10% native polyacrylamide gel using a Tris-borate buffer system (Parsons et al., 1990), followed by autoradiography.

Mapping X-junction cleavage sites The four oligonucleotides used to make the X-junction were labelled at the 5' end using [y-32P]ATP and T4 polynucleotide kinase (BRL). Unincorporated label was removed using JETnick suspension (Genomed) before annealing each one to the appropriate combination of three unlabelled oligonucleotides. This gave four X-junctions, each labelled in a different strand. Cleavage reactions (both Rus and RuvC) were performed as described using buffer containing 10 mM MgCl2 or MnCI2. Reactions were stopped by the addition of 20 mM EDTA and the addition of 1.5 vol. of gel loading buffer (0.3% bromophenol blue, 0.3% xylene cyanol FF, 10 mM EDTA, pH 7.5, 97.5% de-ionized formamide). Samples were boiled for 6 min and the products were separated by electrophoresis through a 12% denaturing polyacrylamide gel. Gels were dried and the labelled products visualized by autoradiography. Labelled oligonucleotides were sequenced (A+G and T+C tracks) using a Maxam-Gilbert sequencing kit (Sigma) to provide a ladder of marker bands.

Ligation of cleaved Holliday junctions

Rus protein (100 nM) and synthetic X-junction DNA (0.15 ng) 32Plabelled on strand 2 were incubated at 37°C for 45 min in ligase buffer (30 mM Tris-HCI, pH 8.0, 9 mM MgClI, 0.1 mM EDTA, I mM DTT,

26 jM DPN, 50 ,ug/ml BSA). Escherichia coli DNA ligase (NEB) was added and incubation continued for a further 30 min. Reactions were stopped by the addition of 20 mM EDTA and gel loading buffer, as described in the previous section. Samples were heated at 95°C for 2 min before electrophoresis on a 9% denaturing polyacrylamide gel followed by autoradiography.

Cleavage of Holliday intermediates made by RecA RecA was used to catalyse strand exchange between OX174 gapped circular duplex DNA and 32P-labelled linear duplex DNA (four-strand reaction), as described (Muller et al., 1992). Reaction conditions for strand exchange and procedures for the purification of strand exchange intermediates (ax-structures) followed the protocols described (Muller et al., 1992; Whitby et al., 1993). Cleavage reactions in which Rus or RuvC were added to the RecA strand exchange reaction were performed in buffer containing 20 mM Tris-HCI, pH 7.5, 15 mM MgCI2, 2 mM DTT, 100 gg/ml BSA, 2 mM ATP, 20 mM phosphocreatine and 12.5 U/ ml creatine phosphokinase. Cleavage of the purified intermediates was conducted in 20 mM Tris-HCl, pH 7.5, 10 mM MgCI2, 2 mM DTT, 100 jig/ml BSA and protein, as indicated. In both cases, reactions were deproteinized and the products analysed on 0.8% agarose gels using a Tris -acetate buffer system (40 mM Tris -acetate, pH 8.0, 1 mM EDTA) at 60 V for 2.5 h. Gels were dried and the labelled products visualized

by autoradiography.

Acknowledgements We thank Lynda Harris, Carol Buckman and Lisa Corbett for excellent technical assistance and Steve West for communicating results prior to publication, particularly the effect of Mn2+ on junction cleavage by RuvC. This work was supported by grants to G.J.S. from the Royal Society, and to R.G.L. from the Medical Research Council and the Science and Engineering Research Council.

References Asai,T., Sommer,S., Bailone,A. and Kogoma,T. (1993) EMBO J., 12, 3287-3295. Bachmann,B.J. (1987) In Neidhardt,F.C., Ingraham,J.L., Low,K.B., Magasanik,B., Schaechter,M. and Umbarger,H.E. (eds), Escherichia coli and Salmonella typhimurium Cellular and Molecular Biology. American Society for Microbiology, Washington, DC, pp. 1190-1219. Bennett,R.J., Dunderdale,H.J. and West,S.C. (1993) Cell, 74, 1021-103 1. Benson,F.E. and West,S.C. (1994) J. Biol. Chem., 269, 5195-5201. Bianchi,M.E. and Radding,C.M. (1983) Cell, 35, 511-520. Clark,A.J. and Margulies,A.D. (1965) Proc. Natl Acad. Sci. USA, 53, 451-459.

Connolly,B., Parsons,C., Benson,F.E., Dunderdale,H.J., Sharples,G.J., Lloyd,R.G. and West,S.C. (1991) Proc. Natl Acad. Sci. USA, 88, 6063-6067.

Cox,M.M., McEntee,K. and Lehman,I.R. (1981) J. Biol. Chem., 256, 4676-4678. de Massy,B., Weisberg,R.A. and Studier,F.W. (1987) J. Mol. Biol., 193, 359-376. Dickie,P., McFadden,G. and Morgan,A.R. (1987) J. Biol. Chem., 262,

14826-14836. Duckett,D.R., Murchie,A.I.H. and Lilley,D.M.J. (1990) EMBO J., 9, 583-590.

Dunderdale,H.J., Benson,F.E., Parsons,C.A., Sharples,G.J., Lloyd,R.G. and West,S.C. (1991) Nature, 354, 506-510. Echols,H. and Goodman,M.F. (1991) Annu. Rev. Biochem., 60, 477-511. Haber,J.E., Ray,B.L., Kolb,J.M. and White,C.I. (1993) Proc. Natl Acad. Sci. USA, 90, 3363-3367. Holliday,R. (1964) Genet. Res. Camb., 5, 282-304. Howard-Flanders,P., West,S.C. and Stasiak,A.J. (1984) Nature, 309, 215-220. Iwasaki,H., Takahagi,M., Shiba,T., Nakata,A. and Shinagawa,H. (1991) EMBO J., 10, 4381-4389. Iwasaki,H., Takahagi,M., Nakata,A. and Shinagawa,H. (1992) Genes Den, 6, 2214-2220. Kemper,B., Pottmeyer,S., Solaro,P. and Kosak,H.G. (1990) In Sarma,R.H. and Sarma,M.H. (eds), Structure and Methods: Human Genome Initiative and DNA Recombination. Adenine Press, New York, pp. 215-229.

Khidhir,M.A., Casaregola,S. and Holland,I.B. (1985) Mol. Gen. Genet., 199, 133-140.

6141

G.J.Sharples et al. Kodadek,T. and Alberts,B.M. (1987) Nature, 326, 312-314. Kohara,Y., Akiyama,K. and Isono,K. (1987) Cell, 50, 495-508. Kowalczykowski,S.C., Dixon,D.A., Eggleston,A.K., Lauder,S.D. and Rehrauer,W.M. (1994) Micro. Rev., 58, 401-465. Lloyd,R.G. (1991) J. Bacteriol., 173, 5414-5418. Lloyd,R.G. and Buckman,C. (1 99 1) J. Bacteriol., 173, 1004-1 01 1. Lloyd,R.G. and Sharples,G.J. (1993a) EMBO J., 12, 17-22. Lloyd,R.G. and Sharples,G.J. (1 993b) Nucleic Acids Res., 21, 1719-1725. Lloyd,R.G., Benson,F.E. and Shurvinton,C.E. (1984) Mol. Gen. Genet., 194, 303-309. Mandal,T.N., Mahdi,A.A., Sharples,G.J. and Lloyd,R.G. (1993) J. Bacteriol., 175, 4325-4334. Matson,S.W. and Kaiser-Rogers,K.A. (1990) Annu. Rev. Biochem., 59, 289-329. Modrich,P. (1987) Annu. Rev. Biochem., 56, 435-466. Muller,B., Jones,C., Kemper,B. and West,S.C. (1990) Cell, 60, 329-336. Muller,B., Burdett,I. and West,S.C. (1992) EMBO J., 11, 2685-2693. Otsuji,N., Iyehara,H. and Hideshima,Y. (1974) J. Bacteriol., 117, 337344. Parsons,C.A. and West,S.C. (1993) J. Mol. Biol., 232, 397-405. Parsons,C.A., Kemper,B. and West,S.C. (1990) J. Biol. Chem., 265, 9285-9289. Parsons,C.A., Tsaneva,I., Lloyd,R.G. and West,S.C. (1992) Proc. Natl Acad. Sci. USA, 89, 5452-5456. Rayssiguier,C., Thaler,D.S. and Radman,M. (1989) Nature, 342, 396401. Rupp,W.D. and Howard-Flanders,P. (1968) J. Mol. Biol., 31, 291-304. Rupp,W.D., Wilde,C.E., Reno,D.L. and Howard-Flanders,P. (1971) J. Mol. Biol., 61, 25-44. Ryder,L., Whitby,M.C. and Lloyd,R.G. (1994) J. Bacteriol., 176, 1570-1577. Sambrook,J., Fritsch,E.F. and Maniatis,T. (1989) Molecular Cloning: A Laboratorv Manual. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY. Shah,R., Bennett,R.J. and West,S.C. (1994) Nucleic Acids Res., 22, 2490-2497. Sharples,G.J., Whitby,M.C., Ryder,L. and Lloyd,R.G. (1994) Nucleic Acids Res., 22, 308-313. Shen,P. and Huang,H.V. (1986) Genetics, 112, 441-457. Shiba,T., Iwasaki,H., Nakata,A. and Shinagawa,H. (1991) Proc. Natl Acad. Sci. USA, 88, 8445-8449. Silhavy,T.J., Berman,M.L. and Enquist,L.W. (1984) Experiments with Gene Fusions. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY. Stasiak,A., Tsaneva,I.R., West,S.C., Benson,C.J.B., Yu,X. and Egelman,E.H. (1994) Proc. Natl Acad. Sci. USA, 91, 7618-7622. Studier,F.W. and Moffat,B. (1986) J. Mol. Biol., 189, 113-130. Tabor,S. and Richardson,C.C. (1985) Proc. Natl Acad. Sci. USA, 82, 1074-1078. Tsaneva,I.R., Muller,B. and West,S.C. (1992) Cell, 69, 1171-1180. Tsaneva,I.R., Muller,B. and West,S.C. (1993) Proc. Natl Acad. Sci. USA, 90, 1315-1319. Vermote,C.L.M. and Halford,S.E. (1992) Biochemistr, 31, 6082-6089. Walker,G.C. (1984) Microbiol. Rev., 48, 60-93. Watt,V.M., Ingles,C.J., Urdea,M.S. and Rutter,W.J. (1985) Proc. Natl Acad. Sci. USA, 82, 4768-4772. West,S.C. (1992) Annu. Rei. Biochem., 61, 603-640. West,S.C. (1994) Cell, 76, 9-15. West,S.C., Cassuto,E. and Howard-Flanders,P. (1981) Nature, 294, 659-662. Whitby,M.C., Ryder,L. and Lloyd,R.G. (1993) Cell, 75, 341-350. Whitby,M.C., Vincent,S. and Lloyd,R.G. (1994) EMBO J., 13, 52205228. Received on September 9, 1994; revised on October 7, 1994

6142