Proc. Natl. Acad. Sci. USAVol. 81, pp. 1026-1029, February 1984Biochemistry

Interaction of the bacteriophage P1 recombinase Cre with therecombining site loxP

(site-specific recombination/nuclease protection pattern/DNA-protein interaction/inverted repeats)

RONALD H. HOESS AND KEN ABREMSKILaboratory of Molecular Biology, LBI-Basic Research Program, NCI-Frederick Cancer Research Facility, P.O. Box B, Frederick, MD 21701

Communicated by Hamilton 0. Smith, October 28, 1983

ABSTRACT The interaction between the P1 recombinaseprotein Cre and the DNA site at which it acts, loxP, has beenstudied by using nuclease protection techniques. The region ofDNA protected by Cre against nuclease attack by DNase I orneocarzinostatin is a 34-base-pair (bp) region containing two13-bp inverted repeats separated by an 8-bp spacer region.These protected sequences have previously been shown to berequired for efficient Cre-mediated recombination at loxP.The results of the above protection experiments suggest (i) thatno more than 34 bp may be required for loxP recombinationand (is) that the asymmetry of loxP recombination is due to the8-bp spacer sequence. With neocarzinostatin, a specific nucle-otide within the 8-bp spacer region is not protected. This nu-cleotide is located in a 2-bp sequence shown to be involved in aloxP crossover event, suggesting that this region remains ex-posed after Cre binding. Protection experiments have alsobeen done with IoxP sites that have either the left or right in-verted repeat deleted. The nuclease protection pattern of thesesites reveals that each loxP site consists of two binding domainsfor Cre, each being composed of one 13-bp inverted repeat andthe contiguous 4 bp of the 8-bp spacer region.

Bacteriophage P1 encodes its own site-specific recombina-tion system. This system is composed of a site, loxP, atwhich crossing-over takes place, and Cre, a P1-encoded pro-tein that mediates the recombination reaction between twoIoxP sites (1). No other phage or host components are neces-sary to carry out this recombination between lox sites in vivo(1) or in vitro (2).

In rare instances the loxP site of P1 will undergo Cre-medi-ated recombination with a specific site on the bacterial chro-mosome, loxB (3, 4). As a result of this recombination, twohybrid lox sites, loxL and loxR, are generated (4). Sequenceanalysis of these sites has established the point of crossing-over between loxP and loxB (5). The crossover point in loxPoccurs within an 8-base-pair (bp) region separating two per-fect 13-bp inverted repeats. Deletion analysis of surroundingP1 sequences necessary for the more efficient loxP x loxPreaction has shown that a sequence of 50 bp containing the13-bp inverted repeat structure is fully functional for recom-bination (2). Deletions that remove either inverted repeatand the 8-bp spacer region are nonfunctional.A more precise definition of the bases involved in the re-

combination process can be ascertained by knowing whatsequences are in contact with the recombinase. Recently,Cre has been purified to homogeneity (6), allowing the analy-sis of the molecular interaction between Cre and loxP usingnuclease protection techniques. These experiments have re-vealed that sequences outside the inverted repeat structure

are not in contact with Cre and that each loxP site is com-posed of two binding domains for Cre.

MATERIALS AND METHODSPlasmids. The source ofDNA containing loxP was a 50-bp

fragment of P1 DNA bounded by EcoRI and Xho I restric-tion sites that was obtained from the plasmid pRH44 (2). AloxP site in which the right inverted repeat had been deletedwas obtained by restriction of pRH44 with Xho I, which cutsjust to the right of loxP, followed by limited digestion withT4 DNA polymerase in the absence of deoxynucleotide tri-phosphates. Ends were made flush by digestion with S1 nu-clease (Sigma), and HindIII linkers (Collaborative Research,Waltham, MA) were ligated to these ends prior to reclosureof the plasmid. The resulting plasmid is pRH442. A loxP sitein which the left inverted repeat is deleted was constructedin analogous fashion, starting with a plasmid (pRH43) thatcontains a unique BamHI site just to the left of loxP (2). Theresulting plasmid is pRH438.

Chemicals and Enzymes. Restriction endonucleases, T4DNA polymerase, and T4 ligase were purchased from Be-thesda Research Laboratories. Calf intestinal alkaline phos-phatase and polynucleotide kinase were obtained fromBoehringer Mannheim. [y-32PIATP ("3000 Ci/mol; 1 Ci =37 GBq) was from Amersham. Neocarzinostatin (NCS) wasa gift from A. Landy and R. Pandey, and DNase I was pur-chased from Worthington.

Purification of Cre. Cre was purified as described (6).Preparations of purified Cre are >99% pure, with only oneband of Mr 35,000 appearing on NaDodSO4/polyacrylamidegels.

Nuclease Protection Experiments. Approximately 0.2 pmolof 32P-labeled DNA containing the lox site was incubated in a50-,ud reaction volume containing 50 mM Tris-HCl (pH 7.5),30 mM NaCl, 50 ,ug of bovine serum albumin per ml, 2 mMMgCI2, 1 mM dithiothreitol, and 5% glycerol at 30°C for 15min in the presence or absence of Cre. After the binding ofCre, 10 ng of DNase I was added and the reaction was incu-bated at 30°C for 5 min. Protection experiments using NCSwere done in the same manner except that MgCl2 and dithio-threitol were replaced with 3 mM EDTA and 30 mM 2-mer-captoethanol, respectively. Incubation with 10 ,ug of NCSwas at 20'C for 10 min. Nuclease digestion was halted by theaddition of 100 Al of a solution of 0.6 M sodium acetate/75mM EDTA/50 ,g oftRNA per ml. Reactions were extractedone time with an equal volume of phenol and then precipitat-ed with ethanol. The pelleted DNA was then resuspended inloading solution containing 0.3% xylene cyanol, 7 M urea, 45mM Tris-HCI, 45 mM boric acid, and 1.25 mM EDTA andelectrophoresed through a 10% polyacrylamide gel contain-ing 7 M urea. For size markers, the 32P-labeled lox-contain-

Abbreviations: bp, base pair(s); NCS, neocarzinostatin.

1026

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement"in accordance with 18 U.S.C. §1734 solely to indicate this fact.

Proc. Natl. Acad Sci USA 81 (1984) 1027

ing fragment was subjected to sequence analysis accordingto the procedure of Maxam and Gilbert (7).

RESULTSProtection of loxP by Cre. We have previously shown that a

functional IoxP site could be localized to a 50-bp region of P1DNA (2), and deletion analysis suggested that a functionalloxP site may be smaller than 50 bp (unpublished data). TheDNA sequences of the loxP site that interact with Cre were

determined by the nuclease protection technique (8). End-labeled restriction fragments containing loxP were incubated

A T NCS DNaseI

G G CC 1 2 3 4 1 2o *

-

lb

".

46 s f -

FIG. 1. Protection by Cre of the IoxP site. A 32P-end-labeled re-

striction fragment containing loxP was used in all lanes. From theleft are four lanes, G, A + G, C, T + C, which are markers generat-ed by Maxam and Gilbert sequence analysis (7) of the fragment.Under NCS, lanes 1-3, the DNA fragment was incubated with 63 ng,105 ng, and 210 ng of Cre, respectively, prior to treatment withNCS. Lane 4 is a control in which no Cre was added prior to treat-ment with NCS. Under DNase I, lane 1, the fragment was incubatedwith 63 ng of Cre prior to DNase I digestion; lane 2 is a control withno Cre added. The loxP site is shown schematically at the right withthe right inverted repeat (RIR) and the left inverted repeat (LIR)represented by arrows and the 8-bp spacer region indicated by a

hatched box. The solid lines connecting the gel and the schematicindicate the extent of protection by Cre against DNase I digestionand the broken lines indicate the extent of protection by Cre againstNCS digestion. The T residue within the spacer cleaved by NCS inthe presence of Cre is shown. The products of DNase I cleavagehave a 3-OH, rather than a 3'-PO4. A 3'-PO4 is characteristic of theproducts of NCS cleavage and of the chemical cleavage products ofthe sequence analysis reactions. As a consequence, bands in theDNase I lanes migrate 1.5 base equivalents more slowly than theequivalent bands in either NCS or sequenced lanes.

with purified Cre and then subjected to limited digestion witheither DNase I or NCS. The resulting cleavage productswere then analyzed by separating them on polyacryl-amide/urea gels.

Fig. 1 illustrates a nuclease protection experiment usingDNase I or NCS to probe the sequences protected by Cre atloxP. Cre protects a region of 36 ± 2 bp from DNase I diges-tion. (We are uncertain of one base at either end of the pro-tected region because, in control experiments, these baseswere not cleaved by DNase in the absence of Cre.) The re-gion of DNA protected against nuclease attack (Fig. 2) iscentered over the 13-bp inverted repeats and 8-bp spacer re-gion, sequences which had been previously implicated as be-ing critical for loxP recombination (5). The complementarystrand (Fig. 2, lower strand) exhibits a similar pattern of pro-tection against DNase I. The boundary for protection of thisstrand is at the outer edge of the right inverted repeat. Pro-tection extends through the inverted repeats and 4 or 5 bpbeyond the outer boundary of the left inverted repeat.When the experiments were repeated using NCS in place

of DNase I, the inverted repeats and the spacer region wereagain protected (Figs. 1 and 2). (Note that a number of bases5' to the left inverted repeat do not appear to be cleaved, butcomparison with the control, lane 4, shows that these basesare weakly cleaved in the absence of Cre.) However, thereare two distinct differences between the patterns generatedby DNase I and by NCS. The first difference is that withDNase I, 3-5 bp outside of the 34-bp inverted repeat andspacer region sequence appear to be protected. Note thatthis occurs on each strand beyond the outer edge of the in-verted repeat distal to the 5' labeled end of the DNA strand.With NCS, only 1 bp outside of the 34-bp sequence is pro-tected. The difference can most easily be explained by theearlier observation of Lomonossoff et al. (9) that for efficientcutting by DNase I, the enzyme must be in contact with the 3or 4 bp to the 5' side of the cleavage site; in this case thepresence of bound Cre sterically prevents contact by DNaseI. NCS, being a much smaller adduct, does not appear torequire this additional contact. The second difference is thatwithin the 8-bp spacer region and 2 bp immediately to theright of the axis of dyad symmetry (Fig. 2, top strand), a T iscleaved by NCS and not by DNase I. A corresponding cleav-age in the spacer region of the complementary strand (at C, 2bp to the left of the axis of dyad symmetry in the lowerstrand of Fig. 2) is not seen. However, an unprotected Ccould not be detected because NCS does not cleave at Cresidues (10). It is not clear whether the differences reflectthe inherent ability of the smaller NCS molecule (Mr 10,000)to cleave at bases where the larger DNase I molecule (Mr30,000) is sterically hindered by the presence of Cre orwhether Cre changes the configuration of the DNA so as toallow preferential cutting by NCS.Cre is capable of binding nonspecifically to any DNA, as

measured by filter binding (6). This nonspecific binding, as

*U...... .,--............ ............... -

SCCTTAAT ATAACTTCGTATA ATGTATGC TATACGAAGTTAT TAGGTC3 /oxPGGAATTA TATTGAAGCATAT TACATACG ATATGCTTCAATA ATCCAG

*U

FIG. 2. Sequences of loxP protected by Cre. The sequence ofloxP is shown with the patterns of protection by Cre against DNase(i) or NCS (0) cleavage indicated in each strand. The top strandhere is that shown in Fig. 1. The small squares (m) are at positionswhere DNase does not cut even in the absence of Cre, so that pro-tection of those bases is uncertain. The arrowhead (-a,) indicates thecleavage by NCS at the T position in the upper strand. The axis ofdyad symmetry is considered to be at the midpoint of the spacerregion between 5' T-A 3' of the top strand.

Biochemistry: Hoess and Abremski

1028 Biochemistry: Hoess and Abremski

loxP438

DNase I T A NCS

1 2 3 C G 1 2 3

RIR

.W b.

LIF

BDNaseI T A

+ +1 2 C G

a

_.._0 O_ _l

_- __0 & 1

4.

-4W

FIG. 3. Protection of the lox half-sites loxP438 and loxP442 by Cre. 32P-labeled restriction fragments containing the lox half-sites wereincubated in the presence or absence of Cre and then treated with either NCS or DNase I. (A) Protection of loxP438. Under DNase I, lane 1contained 210 ng of Cre, lane 2 contained 63 ng of Cre, and lane 3 is a control with no Cre added prior to DNase I digestion. The T + C and A +G lanes are markers generated by sequence analysis of the loxP438 restriction fragment. Under NCS, lane 1 represents a control with no Cre,and lanes 2 and 3 have 63 ng and 210 ng of Cre, respectively. (B) Under DNase I, lane 1 is a control with no Cre and lane 2 contained 43 ng ofCre. The T + C and A + G lanes are markers from sequenced fragments containing loxP442. Under NCS, lane 1 is a control with no Cre, lane 2contained 21 ng of Cre, and the T + C lane shows the appropriate sequenced markers. Adjacent to each gel is a schematic representation of thelox half-sites with the inverted repeat and spacer region indicated as in Fig. 1. Horizontal lines indicate boundaries of the protected region, fromcleavage by DNase I (-) and by NCS (---) as in Fig. 1.

opposed to binding to loxP, is weak and can be abolished bythe addition of the polyanion heparin. The selective protec-tion of the loxP site found in the nuclease protection experi-ments shows that binding is specific to loxP under conditionsused and is insensitive to heparin (data not shown). Onlywhen high concentrations of Cre (200 ng/50 ,ud) were useddid all the DNA become protected.

Protection of lox half-sites by Cre. Because of the dyadsymmetry exhibited by the loxP site, the possibility existedthat each of the inverted repeat sequences represented a sep-arate binding domain for Cre. This premise was tested byconstruction of lox sites that contained the spacer region andone of the inverted repeats, either left or right. The loxP half-sites were no longer proficient in carrying out Cre-mediatedrecombination with a complete loxP site (unpublished data),but they were still able to bind Cre. Restriction fragmentscontaining the lox half-sites were used to perform protectionexperiments with DNase I or NCS. Fig. 3 shows an exampleof the experiments with one strand of the lox half-site inwhich the left inverted repeat is deleted (loxP438) and withone strand of the lox half-site in which the right inverted re-peat is deleted (loxP442).A summnary of the protected sequences of the lox half-sites

is given in Fig. 4. The boundaries at the outer edges of theinverted repeats are no different than those already notedwith the complete loxP site. The boundaries within the spac-er region are quite interesting. When NCS is used, protec-tion does not extend beyond the 4 bp contiguous to the re-maining inverted repeat. Furthermore, the T residue cleavedby NCS in the complete loxP site is also cleaved in the loxsite containing the right inverted repeat in the presence ofCre. When DNase I is used, protection of either of the lox

half-sites extends almost the entire length of the spacer re-gion on both strands (Fig. 4). This presents somewhat of adilemma because it suggests that two Cre molecules bindingto the loxP site both protect the entire spacereregion. Thereare several possible explanations for this observation. First,the DNase I is sterically hindered from cutting by the pres-ence of Cre, even though the uncut sequences may not be inactual contact with Cre. Second, Cre assumes a differentconformation when binding to a lox half-site than when bind-ing to a loxP site and, as a consequence, protects additionalsequences. Third, the binding of Cre to a lox half-site some-how affects DNA conformation, making surrounding bases

U

5CCTTAAT ATAACTTCGTATA ATGTATGC CA AGC T TGCTGCCT3GGAATTA TATTGAAGCATAT TACATACG GTTCGAACGACGGAMI_1-.77

loxP442

* U

5CGTCCCAAGCTTGA ATGTATGC TATAGAAGTTATTAGGTC3 1oxP438GCAGGGTTCGAACT TACATACG TATGCTTCAAT ATCCAG

FIG. 4. Sequences protected by Cre in the lox half-sites loxP438and loxP442. The sequence of each lox half-site is shown, with-indicating Cre protection against DNase I digestion and = indicatingprotection against cleavage by NCS. Where protection is uncertainbecause either DNase I or NCS does not cleave at that position inthe absence of Cre is indicated by * (the small boxes) for DNase Iand o for NCS. The top strand of loxP442 and loxP438 are thoseillustrated in Fig. 3.

A

RIR

loxP442

NCS

1 2

T

C

0*..._do _

1,

a

_W

LIR

Proc. Nad Acad Sd USA 81 (1984)

R41.*.

Proc. NatL. Acad. Sci USA 81 (1984) 1029

refractory to DNase I cleavage, even though they are notdirectly involved in the binding. The result with NCS is like-ly to be a truer reflection of the actual binding contacts;therefore, a loxP site is probably composed of two Cre bind-ing domains, each consisting of a 13-bp inverted repeat andthe contiguous 4 bp of the spacer region.Because each loxP site consists of two bindings domains,

we performed a protection experiment to determine whetherCre binds preferentially to one domain before it binds theother. A fragment containing loxP was incubated with in-creasing amounts of Cre (from suboptimal amounts toamounts needed to observe nuclease protection), and nopreferential protection of one inverted repeat over the otherwas observed (data not shown). If we assume that the bind-ing of Cre to one domain does not significantly acceleratebinding to the adjacent domain, this result would suggestthat there is no preferential binding to either domain of loxP.Techniques other than nuclease protection will be requiredto accurately measure the binding affinity of Cre for these loxsites.

DISCUSSIONThe results presented here are a direct demonstration of theinteraction of the recombinase Cre with the recombinationsite at which it acts, loxP. The nuclease protection patternseen after Cre binding extends over two 13-bp inverted re-peats and an 8-bp spacer region separating them. The se-quences protected by Cre show an excellent correlation withthose sequences known to be required for recombination(ref. 5 and unpublished data). Sequences outside of the in-verted repeats do not appear to be in contact with Cre. Ourpresent experiments cannot exclude the possibility of otherbinding sites for Cre on the P1 genome; however, if suchsites exist, they are clearly not an essential part of the loxPrecombination system. These results are in striking contrastwith two other well-characterized site-specific recombina-tion systems, those of X and Tn3/y8. In these two systems,the respective recombination proteins have multiple bindingsites, some of which are located a large distance away fromthe actual crossover point (X attP binding sites encompass a240-bp region; Tn3/yS binding sites encompass 120 bp) (11,12), and all of these binding sites are required for efficientrecombination. Clearly, the Cre-lox recombination system issimpler than those recombination systems of X and Tn3/y8because both protein-binding domains are located within 34bp.

It has already been established that the loxP site exhibitsdirectionality. When two sites on the same DNA moleculeare in a directly repeated orientation, the DNA between thesites is excised after recombination (2). However, if the sitesare inverted with respect to each other, the DNA betweenthem is not excised after recombination but is simply invert-ed. How then does Cre recognize the directionality of theloxP site? The inverted repeats, by definition, are symmetri-cal and cannot impose a directionality to the site. BecauseCre does not protect sequences outside of the region con-taining the inverted repeats and spacer, the directionality ofthe loxP site must be a consequence of the asymmetry of thespacer region.The data obtained from nuclease protection experiments

with the lox half-sites indicate that each loxP site is com-posed of two binding domains for Cre. Each domain includesan inverted repeat and the contiguous 4 bp of the adjoiningspacer region. When NCS is used to probe the Cre-lox com-plexes, certain portions of the spacer region are apparentlynot protected by Cre. Of particular note is the T residue (2 bpimmediately to the right of the axis of dyad symmetry in thetop strand of Fig. 2), which is cleaved by NCS in the pres-ence of Cre both in the complete loxP site and the loxP438site. The T here occurs within a two-base sequence, 5' T-G

N OM 5GccA AC AGTATAAAAAAGC TGAACGAG 3

PIloxP FATAACTT CGT AT AI.ATGTATGC.nTAT A cGAA G T TA

Yeast 2, FLP IGA A G T T CC T A T AC)TTTCTAGAGAATAGGAACTTCI

FIG. 5. A comparison of the loxP sequence with two other se-quences involved in site-specific recombination. Line one showsone strand of XattP (13) with the common core sequence wherecrossing-over takes place indicated by the boxed region (c). Linetwo shows the inverted repeat sequence of loxP. To better demon-strate the homology with the yeast 2-,um FLP sequence (14) shownin line three, a one-base gap (m) was made at each end of the loxPspacer. Bases that are homologous in loxP and the other sites areindicated by large type.

3', that is homologous in the spacer regions of loxP and loxB(5), and the crossover between loxP and loxB takes placewithin this region. The results presented here suggest thatone of the DNA strands in the crossover region is exposed.Perhaps this short region of exposed bases provides a meansby which two recombining loxP sites can exchange strands.We have pointed out the simplicity of the Cre-lox system

in comparison with other site-specific recombination sys-tems; however, this system does have a feature in commonwith the X Int-attP system and the yeast 2-Am circle recom-bination system. The 13-bp inverted repeat sequence of loxPis homologous with the X att site (13) and the yeast 2-,AmFLP site (14) (Fig. 5). Furthermore, in the case of X, thehomologous sequence comprises one of the core arm junc-tions that is also recognized by the Int recombinase (15). Thesequence to which the yeast FLP recombinase binds is notknown, but the site at which it acts is similar in organizationto loxP-i.e., it contains two inverted repeats that are ho-mologous to the ones of loxP and are separated by an 8-bpspacer region. Although the sites and recombinases areprobably not interchangeable, the homologies suggest anevolutionary relatedness among these site-specific recombi-nation systems. More detailed analysis of the respective re-combinases may help to establish such relatedness.The authors thank Dr. N. Sternberg for his encouragement and

suggestions during the course of this work. We also thank Drs. Aus-tin, Pearson, and Sodergren for critically reading the manuscript andJ. Ratliff for preparation of the manuscript. This research was spon-sored by the National Cancer Institute, under contract N01-CO-23909 with Litton Bionetics, Inc.1. Sternberg, N. & Hamilton, D. (1981) J. Mol. Biol. 150, 467-

486.2. Abremski, K., Hoess, R. & Sternberg, N. (1983) Cell 32, 1301-

1311.3. Chesney, R. H., Scott, J. & Vapnek, D. (1979) J. Mol. Biol.

130, 161-173.4. Sternberg, N., Hamilton, D. & Hoess, R. (1981) J. Mol. Biol.

150, 487-507.5. Hoess, R. H., Ziese, M. & Sternberg, N. (1982) Proc. Nadl.

Acad. Sci. USA 79, 3398-3402.6. Abremski, K. & Hoess, R. H. (1984) J. Biol. Chem. 259, 1509-

1514.7. Maxam, A. M. & Gilbert, W. (1980) Methods Enzymol. 65,

499-560.8. Galas, D. J. & Schmitz, A. (1978) Nucleic Acids Res. 5, 3157-

3170.9. Lomonossoff, G. P., Butler, P. J. G. & Klug, A. (1981) J. Mol.

Biol. 149, 745-760.10. Poon, R., Beerman, T. A. & Goldberg, I. H. (1977) Biochemis-

try 16, 486-493.11. Hsu, P.-L., Ross, W. & Landy, A. (1980) Nature (London)

285, 85-91.12. Grindley, N. D. F., Lauth, M. R., Wells, R. G., Wityk, R. J.,

Salvo, J. J. & Reed, R. R. (1982) Cell 30, 19-27.13. Landy, A. & Ross, W. (1977) Science 197, 1147-1160.14. Broach, J. R., Guarascio, V. R. & Jayaram, M. (1982) Cell 29,

227-234.15. Ross, W. & Landy, A. (1982) Proc. Natl. Acad. Sci. USA 79,

7724-7728.

Biochemistry: Hoess and Abremski