A site-specific recombination function in Staphylococcus aureus plasmids

All known small staphylococcal plasmids possess one or two recombination sites at which site-specific cointegrate formation occurs. One of these sites, RSA, is present on two small multicopy plasmids, pT181 and pE194; it consists of 24 base pairs of identity in the two plasmids, the "core," flanked by some 50 base pairs of decreasing homology. Here we show that recombination at RSA is recA independent and is mediated by a plasmid-encoded, trans-acting protein, Pre (plasmid recombination). Pre-mediated recombination is site specific in that it occurs within the core sequence of RSA in a recA1 host. Recombination also occurs between two intramolecular RSA sites. Unlike site-specific recombination systems encoded by other plasmids, Pre-RSA is not involved in plasmid maintenance.

All known small staphylococcal plasmnids possess one or two recombination sites at which site-specific cointegrate formation occurs. One of these sites, RSA, is present on two small multicopy plasmids, pT181 and pE194; it consists of 24 base pairs of identity in the two plasmids, the "core," flanked by some 50 base pairs of decreasing homology. Here we show that recombination at RSA is recA independent and is mediated by a plasmid-encoded, trans-acting protein, Pre (plasmid recombination). Pre-mediated recombination is site specific in that it occurs within the core sequence of RSA in a recAl host. Recombination also occurs between two intramolecular RSA sites. Unlike site-specific recombination systems encoded by other plasmids, Pre-RSA is not involved in plasmid maintenance.
Site-specific recombination is generally used to effect stable or semistable structural rearrangements of nonhomologous elements (integration of phage genomes and transposons; formation and resolution of cointegrates) or of segments within the same element (expression-related inversions) (1,13,15,32,40,43). Site-specific recombination also mediates structural rearrangements involving homologous elements such as circularization of the terminally redundant linear P1 phage genome (36) and resolution of plasmid multimers. It has been suggested that the latter type of activity is involved in stable plasmid maintenance because certain plasmid mutants defective in site-specific recombination activity accumulate multimers and are hereditarily unstable (9,38). A unit-copy replicon, such as P1, utilizes its site-specific recombination system to mediate dimer resolution into monomeric substrates for proper partitioning (2).
Site-specific recombination in Staphylococcus aureus plasmids is responsible for the formation of stable cointegrates (14). We have observed previously that most of the known small S. aureus plasmids contain one or two specific recombination sites, RSA and RSB (26). RSB, about 30 base pairs (bp) in length, is present on all of the six plasmids then analyzed, namely, pT181, pE194, pC194, pC221, pS194, and pSN2. RSB cointegrates were obtained only after cotransduction, and crossovers were later shown to map within a perfectly conserved 18-bp sequence (5'-AAGTTl-TCTC GGCATAAA-3') (28). It was therefore suggested that RSB recombination was mediated by a site-specific phage recombination function. The RSA sequence was found in only two of the six plasmids, pT181, and pE194. In contrast to RSB, plasmid transduction was not required for the formation of RSA cointegrates. These were readily obtained with established heteroplasmid strains by selecting for the rescue of a temperature-sensitive mutant plasmid. In a Rec+ host strain, recombination involving RSA occurred at different locations within a 70-bp region consisting of a 24-bp fully conserved "core" sequence and flanking regions marked by a considerable number of mismatched nucleotides. No recombination occurred utilizing adjacent regions of extensive DNA homology in the two plasmids. These results suggested that RSA recombination involved an unusual mechanism that * Corresponding author. recognized a specific short region but catalyzed strand exchange at different sites within that region.
In this communication, we identify a new plasmidencoded recombination protein, Pre (plasmid recombination), that mediates RSA, but not RSB, recombination. Both of the plasmids that contain RSA encode homologous Pre proteins, whereas plasmids lacking RSA do not. We show that Pre-mediated recombination is site specific as it always occuts within the RSA core sequence in a recAl background; this suggests that the previously observed site variability may reflect synergism between Pre and the host rec system.

MATERIALS AND METHODS
Bacterial strains and plasmids. The plasmids used in this study are listed in Table 1. The Rec+ host strain was S. aureus RN450, a derivative of NTCC 8325 cured of all known prophages. S. aureus RN1030 (recAl) (42) is lysogenic for phage +11. Escheeichia coli BL21 (F-hsdS gal) (37) was the host strain for plasmid pGEM1 (Promega Biotec) and its derivatives.
Nucleotide sequence coordinates are according to the published sequences of pT181 (16) and pE194 (11). Cloned fragments retain their original nucleotide sequence coordinates, which are suffixed (T) or (E) to indicate the plasmid of origin.
Media and culture conditions. Liquid and solid culture media for S. aureus were used as described (4). Tetracycline, chloramphenicol, and erythromycin were used at 5 ,ug/ml. Plasmid transfers were by transduction with phage +11 (23) or by protoplast transformation (5).
For the growth of E. coli we used 2x YT broth, B agar, and M9 salts minimal medium (22); where appropriate, ampicillin (100 ,ug/ml) was added to the medium. Plasmid transformation of calcium chloride-treated cells was as described by Maniatis et al. (19).
Assay of plasmid stability. S, aureus cells from a selective plate were suspended in antibiotic-free liquid medium to a turbidity of approximately 20 klett units (-2 x 108 CFU/ml) and grown to exponential phase. Dilutions (100 fold) of these cultures were grown nonselectively in liquid medium overnight. The cycle of dilution and growth in antibiotic-free broth was performed four times. Each cycle constituted approximately 20 generations. From each cycle, samples were diluted and spread onto antibiotic-free plates. The resulting colonies were replica plated on the appropriate antibiotic-containing plates and scored for resistance. The same procedure was used to detect cointegrate resolution. Colonies grown on tetracycline plus erythromycin plates at 32°C after 60 generations of growth in antibiotic-free liquid medium were replica plated on erythromycin plates at 43°C. Isolation and analysis of plasmid DNA. Plasmid DNA was isolated by ethidium bromide-cesium chloride centrifugation of cleared lysates prepared as described (27). Sheared whole-cell minilysates were prepared as described by Projan et al. (31) and analyzed by 1.0o agarose gel electrophoresis.
Restriction mapping and cloning. Restriction enzymes were purchased from New England Biolabs, Bethesda Research Laboratories, and International Biotechnologies, Inc. and used as recommended. Restriction mapping was performed with ethidium bromide-cesium chloride-purified plasmid DNA samples.
Cointegrates between pT181 and pE194 and their deletion derivatives were analyzed by single and double digestions with TaqI, MboI, Hinfl, and DdeI restriction endonucleases. The digestion products were separated by agarose or acrylamide gel electrophoresis. Junctions were identified as fragments that did not comigrate with any fragment from either parental plasmid, and, conversely, fragments of the parental plasmids containing the crossover point were identified by their absence from the cointegrate digests.
For molecular cloning, standard procedures were used as described (4). DNA sequencing. Nucleotide sequences were determined by the dideoxy chain termination method (34) using linear plasmid DNA as a template and a synthetic oligonucleotide (position 1860-1880 in the pT181 map) as a primer. The PNA sequence of the pre coding region was determined from isolated restriction fragments cloned into either mplO or mpll M13 vectors (22), using the M13 universal primer.
Analysis of cointegrate formation. Heteroplasmid strains of S. aureus RN1030 (recAl) were obtained by 411 transduction. For each heteroplasmid analyzed, 6 to 12 single colonies grown on doubly selective plates at the permissive temperature for the thermosensitive replication (Tsr) mutant of pE194 used (32°C) were picked and suspended in CY broth (4). Appropriate dilutions were plated on doubly selective plates and incubated at the permissive (32°C) and nonpermissive (43°C) temperatures for 48 h. Cointegrate formation frequencies were calculated as the ratio of the number of colonies at 43°C divided by the viable counts at 32°C. Approximately 1/4 to 1/3 of the colonies obtained at 43°C were lysed and screened by 1% agarose gel electrophoresis for the presence of cointegrate molecules, and the colony counts at 43°C were corrected for noncointegrates.
RNA isolation. Lysostaphin protoplasts were lysed with 5M guanidine thiocyanate (Fluka Chemicals). The crude lysate was layered on a 3-ml CsCl cushion (p = 1.76) and spun at 35,000 rpm for 16 h. The RNA pellet was precipitated from 8 M guanidine hydrochloride (6,8), followed by two presipitations from distilled water and then four rinses with 80% ethanol. The final RNA pellet was suspended in TE (10 mM Tris, pH 7.6, 1 mM EDTA) and stored in aliquots at -70°C. When necessary, RNA was treated with DNase I (Pharmacia) (8 U/100 ,ug of RNA) in the presence of RNase inhibitor (Pharmacia) at 15 U/U of DNase I for 30 min at 370C.
Runoff transcription. In vitro transcripts were obtained using a modification of the procedure of Fisher et al. (7). DNA templates consisted of specific restriction fragments of pT181 isolated from 5% acrylamide gels (20) and labeled with [a-32P]dATP by nick translation (33). From 500 to 750 ng of template was used in each reaction. Transcriptions were carried out in a volume of 50 ,lI. DNA, E. coli RNA polymerase (1 U; Boehringer Mannheim), and binding buffer (40 mM Tris, pH 7.6, 80 mM KCl, 8 mM MgCl2, 1 mM dithiothreitol, and 40 ,ug of bovine serum albumin per ml) were combined and incubated for 5 min at 37°C. Heparin was then added (600 ,ug/ml, final), followed by [a-32P]UTP. The reaction was started by adding cold nucleotide triphosphates (200 ,uM ATP, CTP, and GTP; 20 ,uM UTP) and incubated for a further 40 min at 370C. Reactions were terminated by adding sodium acetate to 0.3 M, 20 ,ug of carrier tRNA, and 3 volumes of 95% ethanol. After alcohol precipitation, the dried pellets were suspended in 8 M urea, denatured at 65°C for 10 min, and loaded on a thin 6% acrylamide-8 M urea sequencing gel. Electrophoresis was carried out at 1,200 V until the xylene cyanol was approximately 10 cm from the bottom. The gel was fixed in 5% acetic acid-methanol, dried, and autoradiographed at -70°C, using Fuji AR film and a DuPont intensifying screen. S1 nuclease protection. The procedure outlined by Maniatis et al. (19) was followed for nuclease protection. Samples of 20 pg of RNA were used in these experiments. The DNA probe fragments were the same as in the runoff experiments. These fragments were treated with 1 U of calf intestinal alkaline phosphatase (Boehringer Mannheim) for 60 min at 37°C and then end labeled with [.y-32P]ATP and 20 U of kinase (Pharmacia) for 60 min at 37°C. The labeled fragments were gel purified from 5% acrylamide and used for hybridization. Approximately 5 x 104 to 7 x 104 cpm was used for each hybridization. Hybridizations were carried out at 21, 30, and 39°C for 16 h. Then 200 U of S1 was added, and the samples were incubated for 30 min at 37°C, precipitated, and vacuum dried. Pellets were suspended and analyzed as described for runoff transcripts.
Lac induction. A single colony of the test strain was inoculated into 10 ml of ampicillin-containing M9 medium and grown at 37°C with shaking to 5 x 10O to 10 x 108 J. BACTEPIOL. CFU/ml. Induction was with 0.4 mM isopropyl-p-Dthiogalactopyranoside. Cells were harvested 3 h after induction, and pellets were frozen (37). Proteins were extracted by acetone-sodium dodecyl sulfate treatment of the pellets (3). Sodium dodecyl sulfate-polyacrylamide gel electrophoresis on 12.5% gels was by the method of Laemmli (17).

RESULTS
A plasmid product, Pre, is required for recombination at RSA. To determine whether a plasmid product is involved in recombination at RSA, we directed attention to the region between RSA and RSB, which contains a homologous open reading frame (ORF) on pT181 and pE194 to which no function had been assigned. Delections affecting this region were constructed by removal of restriction fragments using pT181 wild type and pRN5101, a temperature-sensitive derivative of pE194. A schematic map of the two plasmids indicating the deleted regions is presented in Fig. 1.
It has been previously shown in this and other systems (18,26) that cointegrates can be selected in heteroplasmid strains, where one plasmid of the pair is temperature sensitive, by plating at the nonpermissive temperature on medium selective for the temperature-sensitive replicon. We constructed heteroplasmid strains by transduction in both Rec+ and recAl host strains containing either the two parental plasmids or one or both deletion derivatives. Therefore, four pairs were examined: pT181 and pRN5101; pRN8157 (pT181 ADdeI-B+D) and pRN5101; pT181 and pRN6321 (pRN5101 ATaqI-C); and pRN8157 (pT181 ADdeI-B+D) and pRN6321 (pRN5101 ATaqI-C). The heteroplasmid strains were plated on erythromycin (5 ,ug/ml) plates (see Materials and Methods) at 43°C, and colonies arising were subcultured. Of 454 colonies, 284 were streened for plasmid content; 95% of them contained a single plasmid species whose size corresponded to the sum of the two original plasmids. Frequency values ( Table 2) are corrected accordingly. Separate frequencies were measured for several single-colony isolates (see Materials and Methods); frequencies obtained for each heteroplasmid strain are the mean of these independent determinations. With the pT181 pre gene intact, cointegrates were formed at frequencies of the order of 10-6 per CFU. When both plasmids had ORF deletions, the frequency of colony formation decreased by two orders of magnitude. If the pE194 ORF was intact but that of pT181 was deleted, intermediate frequencies of cointegrate formation were obtained. Although cointegrate formation in Rec+ cells occurred at a slightly higher frequency than in the recAl host, the frequency was equally depressed by the double ORF deletion.
These results led to a twofold conclusion. First, the corresponding ORFs of pT181 and pE194 each encode a recombination-promoting function; the pE194 protein seems somewhat less effective than that of pT181 for the formation of heterologous cointegrates. Second, cointegrate formation between pT181 and pE194 is not promoted to a comparable extent by any host rec function.
The recombination function responsible for the formation of rare cointegrates in the recAl mutant host and in the absence of either plasmid protein is unknown; it is noted that the residual activity of the recAl host strain for homologous interplasmid recombination is about 1i-0 of that of the Rec+ host (41). The cointegrates were found to be stable during prolonged growth on nonselective medium. Single colonies from each type were grown for approximately 40 generations in broth and then tested by replica plating at the nonpermis- Dark line inside the pT181 map represents the MboI-A fragment containing the pre gene. ori, Origin of replication; cop, copy control; repC, initiator gene; tetA, tetB, tetracycline resistance determinant. Numbers are nucleotide coordinates according to the published sequence (11,16). Only relevant restriction sites are indicated.
sive temperature. Loss of the temperature-sensitive marker was not detected. If it occurs, it does so at a frequency less than 10-4. The MboI-A fragment of pT181 encodes a trans-acting protein that nediates cointegrate formation. To test whether the ORF implicated in cointegrate formation acts in trans, we cloned the MboI-A fragment of pT181 (376-2492) into the unique HindIII site of the chloramphenicol resistance plasmid pC194 (12,14). The cloned fragment contains the entire ORF and the two flanking recombination sites, RSA and RSB. The recombinant plasmid pRN6378 was introduced by transduction into the strain containing the deleted pT181 and pE194 replicons, and the frequency of cointegrate formation was determined by selecting for rescue of the temperaturesensitive pE194 plasmid. Since the three starting plasmids have different mobilities in agarose, the composition of the  Additionally, we introduced a frameshift mutation into the ORF by inserting a synthetic octamer (commercial EcoRI linkers) at the unique FnuDII site (position 1687 in pTl8l sequence) and cloned the modified Mbol-A fragment to pC194. The resulting plasmid, pRN6404, was introduced into the heteroplasmid strain carrying the pT18i1 and pE194 deletion derivatives used to demonstrate trans activity. In this case, no increase in cointegrate formation over background was observed ( Table 2). We therefore suggest that the MboI-A fragment from pTl8l encodes a trans-acting, recombination-promoting protein.
Nucleotide and amino acid sequence analysis. The published sequence of pTl8l (16) shows two shorter, overlapping OREs in comparison to the single 403-amino acid ORF of pIE194 (11) encompassing the RSA-RSB region.
We resequenced the pTl8l region between nucleotides 700 and 1800 and fou'nd six differ'ences with the previously published sequence: an extra T at position 1540, a dinucleotide CA instead of TC at positions 1494-1495, and an extra A at position 1429 (Fig. 2). A was read instead of G at three different positions (795, 820, and 829). All changes refer to the bottom strand of pTl8l as published.
The revised sequence contains a single ORF of 413 amino acids between positions 681 and 1917 of the plasmid map. The presence of only a single ORE is co'nsistent with the protein analysis presented below.
A comparative analysis of the OREs in pTl8l and in pE194 shows a high degree of DNA homology (47%), which is reflected in the overall amino acid sequ'ence (39%1). The homology is much stronger in the N-terminal portion of the two polypeptides (approximately 60%) than in the Cterminal portion (approximately 20%') (Fig. 3). We refer to this ORF as Pre (plasmid rec'ombination) henceforth, Pre(T) and Pre(E) being the pTl8l and pE194 homologs.
The pre promoter. To map the pre promoter, we analyzed in vitro runoff transcripts by using two restriction fragments from pTl8l as templates for E. coli RNA polymerase: MboI (2492)-Taq(1726) (766 bp) and MboI(2492)-HindTII(1444) (1,048 bp). These two fragments have the same MboI end but differ by 282 bp at the opposite Taql and HindII ends. Transcription products were run on a 6% acrylamide-urea sequencing gel next to a M13 dideoxy nucleotide sequencing ladder. The results (Fig. 4A) (16). Capital letters show the differences between the published sequence and that determined in this work. An arrow indicates the putative ORF for pre. obtained from the longer MboI-HindIII fragment. This indicated that the direction of transcription is from the MboI end toward the TaqI end. By comparison with the sequencing ladder, the MboI-TaqI runoff transcript measured approximately 240 nucleotides, which maps the transcription start site at around position 1968 in pT181 sequence. The MboI-HindIII runoff, which would be about twice as long, was too large to measure accurately.
To confirm the in vitro runoff results, we performed Si nuclease protection studies. Whole-cell RNA isolated from an S. aureus strain carrying a pT181 copy mutant was hybridized to the MboI-TaqI and MboI-HindIII fragments end-labeled with kinase and [y-32P]ATP. Sl-protected hybrids were run on a 6% sequencing gel next to an M13 dideoxy sequencing ladder. The results (Fig. 4B) showed that a protected band was obtained from both fragments. Consistent with the runoffs, the band protected by the MboI-TaqI fragment was shorter than the band protected by the MboI-HindIII fragment. Measurement of the Siresistant band from the MboI-TaqI fragment gave a length of 232 nucleotides. This is in good agreement with the in vitro transcription results and maps the site of transcription initiation to position 1960 in pT181 sequence. The presence of a second, higher-molecular-weight protected band is noted here. This band, present with both probes, was seen preferentially at lower hybridization temperatures, suggesting a lower degree of homology. The origin and nature of this species are unknown. Figure 5 shows the sequences containing the promoter and initiation site of the pre gene.
A protein is made from the 413-amino acid ORF. To test whether the MboI-A transcript was translated, we cloned the HgiAI(2010)-RsaI(570) fragment, internal to the MboI-A region but containing the entire pre coding sequence, into E. coli expression vector pGEM1, a 2.9-kilobase ampicillin resistance plasmid containing a polylinker region between T7 and SP6 promoters; depending on the orientation of the insert, transcription will be driven by either promoter.
We chose the orientation in which the insert was cloned under control of the T7 promoter and transformed the recombinant plasmid to E. coli BL21 (lambda DE3 lysogen) (37), which contains the T7 RNA polymerase gene in the chromosome under control of the inducible lacUVS promoter. Addition of isopropyl-,3-D-thiogalactopyranoside induces T7 RNA polymerase, which in turn results in highlevel expression of the target gene cloned in plasmid pGEM1 (see Materials and Methods). Figure 4C shows the result of a sodium dodecyl sulfatepolyacrylamide gel electrophoresis analysis of the protein content of crude cell lysates prepared from isopropyl-,3-Dthiogalactopyranoside-induced cultures containing either the recombinant plasmid or the vector plasmid. The gel reveals a single prominent protein band in the recombinant plasmidcontaining cells that is absent in the controls. The apparent molecular weight of this protein is approximately 50,000, VOL. 169, 1987 .'rrMT . Runoff transcription experiments were carried out as described in Materials and Methods, and the reaction products were visualized by electrophoresis on a 6% polyacrylamide gel containing 8 M urea. Two fragments were used as templates for E. coli RNA polymerase-directed in vitro transcription: pT181 MboI(2492)-TaqI(1726) (lane 1) and MboI(2492)-HindIII(1444) (lane 2). pT181 MboI-D fragment  was used a template for the control reaction (lane 3), since it gives rise to transcripts of known length. GACT is an M13 ladder used as a size marker. Arrows indicate bands corresponding to a transcript of approximately 240 bases (lane 1) and a larger transcript (lane 2) whose size is not measurable under these experimental conditions. The results indicate the direction of transcription of the pre gene (see text). (B) The same pT181 fragments as in runoff transcription were used for Si protection experiments, and the products were electrophoresed on a 6% polyacrylamide-8 M urea gel: pT181 MboI(2492)-HindIII(1444) (lanes 1 through 5) and pT181 MboI(2492)-TaqI(1726) (lanes 6 through 10). GATC is an M13 ladder used as a size marker. Lanes: 1 and 10, purified fragment, no RNA and no Si; 2 and 9, purified fragment + S1, no RNA; 3 to 5 and 6 to 8, purified fragment + RNA + Si (DNA-RNA hybridization was carried out at different temperatures: 3 and 6, 21°C; 4 and 7, 30°C; 5 and 8, 39°C). Arrows indicate protected bands obtained with the two templates, corresponding to the 5' end of the pre mRNA. (C) Expression of pre gene product in E. coli. Sodium dodecyl sulfate-polyacrylamide gel electrophoresis of whole-cell lysates upon lac induction, as described in Materials and Methods, visualized by Coomassie staining. Lanes: 1, molecular weight standards (m.w.s.) (BioRad Laboratories); 2, pGEM1::pTl8l HgiAI-RsaI (T7 orientation); 3, pGEM1; 4, pGEM1::pT181 HgiAI-RsaI (SN orientation).
The arrow indicates a protein band of approximately 50,000 molecular weight produced by the recombinant plasmid pGEM1::pT181 HgiAI-RsaI (T7 orientation) (Lane 2), which is absent in the Qther lanes.
which is in agreement with the predicted molecular weight (55,700). Therefore, these results confirm the sequence data indicating a single ORF in the pre region.
Pre-mediated cointegrate formation occurs at RSA. Restriction analysis of 41 cointegrates derived from the first three classes of heteroplasmid strains listed in Table 2 (in which at least one of the parental plasmids carried pre) showed that (i) as previously observed, the two plasmids are always in the same orientation relative to each other in the cointegrate molecule, and (ii) the crossover point was always located in the RSA region and never in the flanking pre homologous regions (data not shown). Additionally, 8 cointegrates of the 48 obtained in the presence of the cloned pT181 pre gene were analyzed by restriction enzyme digestion; they were all RSA cointegrates. These observations confirmed previous findings (28) that cointegrates form primarily at RSA, despite the presence of adjacent homologous sequences, and indicated that the pre gene product promotes these events. We determined the DNA sequence of one crossover junction of seven of the RSA cointegrates formed in the presence of Pre in a recombination-deficient host. In each case, the crossover site was contained within a 24-bp sequence of identity, the RSA core (Fig. 6A), suggesting that Pre-mediated cointegrate formation is site specific. In contrast, of five cointegrates from a heteroplasmid strain containing the two Pre-plasmids, three were identified as RSA and two as RSB cointegrates, suggesting the existence of additional lowactivity rec functions capable of recognizing RSA and RSB, possibly on the basis of homology.
Intramolecular recombination occurs within the RSA core sequence. Plasmid pRN6010 is the prototype of a series of cointegrates constructed in vitro by ligating pT181 (wild type or copy mutants) and pE194 at their unique XbaI sites (25). A schematic map of pRN6010 is shown in Fig. 7. The plasmid contains the intact pT181 pre gene, whereas the corresponding pE194 pre coding region is interrupted by the cloning procedure. The tetracycline resistance (Tc9 gene of pT181 is flanked in this construct by homologous sequences in the direct orientation, including the RSA sites of each parent. These cointegrates (but not pRN6050, in which pT181 and pE194 are in the opposite orientation) give rise to spontaneous deletions of the Tcr marker at very low frequency (<10-4). Tcs spontaneous deletion derivatives could be obtained by exploiting the presence of a unique KpnI site within the tetracycline resistance gene (G. K. Adler and R. P. Novick, unpublished data). Plasmid DNA from cultures carrying pRN6010 was digested with KpnI before transformation; only KpnI-resistant molecules (i.e., missing tetracycline resistance gene sequences) could give rise to Emr transformants. As expected, these plasmids carried a spontaneous deletion in the tetracycline gene (25). We sequenced the deletion site in Tcs derivatives of seven cointegrates of the pRN6010 family. The deletion occurred, in each case, by a site-specific crossover within the 24nucleotide RSA core; its sequence is identical to that shown in Fig. 6B. It is therefore likely that these events are Pre mediated.
Pre-plasmids are not detectably defective in maintenance and do not accumulate multimers. Since other site-specific recombination systems act to resolve plasmid multimers, we tested Preor RSAplasmids (or both) for instability in S. aureus and for the accumulation of multimers. No loss of Pre-plasmids from either Rec+ or recAl host strains was detected after growth for approximately 40 generations in nonselective medium. The presence of multimers was analyzed by 0.8% agarose gel electrophoresis of whole-cell lysates containing Preor Pre' plasmids. Accumulation of multimers was greater with Pre' as opposed to Preplasmids and in a Rec+ as opposed to a recAl background (Fig. 8). Similar results were obtained with RSAplasmids (not shown) which, due to the overlapping of RSA sequences with the pre promoter, are also Pre-.

DISCUSSION
We have shown that two S. aureus plasmids, pT181 and pE194, possess a site-specific recombination system. This system comprises a cis-acting element, the recombination site RSA, and a trans-acting product which we have designated Pre. The RSA sites in the two plasmids have identical 24-bp core sequences and are 80% homologous in the flanking sequences. A perfect 7-bp inverted repeat with 6-bp spacer region (5'-GTGTGT-3') is located within the core sequence (Fig. 6A). Some dyad symmetry is present in other sites involved in site-specific recombination, such as res (Tn3), parB (CloDF13), and loxP (P1), and is common to many other protein-binding sites.
Although the two Pre proteins are similar in size [413 amino acids for Pre(T) and 403 amino acids for Pre(E), respectively] and amino acid sequence (39% overall homology; 60% in the N-terminal region), Pre(T) seems considerably more active than Pre(E) in the formation of pT181-pE194 RSA cointegrates. Indeed, Pre(E) may even interfere with Pre(T) in this situation; when both were present in the heteroplasmid the cointegration frequency was considerably lower than when Pre(T) alone was present. The Pre proteins Map of pRN6010, a pT181::pE194 XbaI cointegrate. pT181 sequences are represented by a thick line; those from pE194 are indicated by a thin line. Coding sequences in the two plasmids interrupted by the cloning procedure are shown as wavy lines [pre(E) and pT181 repC]. The spontaneous deletion of a 2.2-kilobase fragment including the tetracycline resistance determinant is indicated by the dashed lines. The sequence of the crossover point is the same as in Fig. 6B. are encoded by the largest ORF present in each plasmid. We have shown that Pre(T) is synthesized in vivo; Dubnau and co-workers (35) had previously shown that a 54-kilodalton protein, indicated as El, was synthesized from the pre coding region in Bacillus subtilis minicells containing pE194.
We have focused our attention on the pTl8l-encoded system and shown that Pre-mediated crossover events always take place within the 24-bp core sequence of RSA. This site seems to be used more efficiently for interthan for intraplasmid recombination, since multimers accumulated with Pre'-RSA' but not with Pre--RSA plasmids.
We were able to detect Pre-mediated intraplasmid recombination only in a situation permitting molecular selection (for KpnI resistance). Unfortunately, no estimate of frequency could be obtained in such experiments. We were also unable to measure resolution frequency directly, as it was below the detection limit (<10-4 per cell per generation).
Since all the experiments involved heterologous RSA sites, it was considered important to examine the behavior of homologous sites. Cointegrates were obtained between pSA0301 (a Tsr pT181 mutant) and pRN6450 (a pC194 derivative carrying a 240-bp RSA-containing fragment from pTl8l) at a frequency of 10-' per cell.
We do not know the function responsible for formation of these cointegrates, because the recAl host has residual homologous interplasmid recombination activity. However, the resolution frequency of these cointegrates was again below the limit of detectability (<10-4). While these experiments do not permit any definite conclusion about the relative frequency of cointegrate formation versus resolution, they argue strongly against the possibility that the biological function of the RSA-Pre system is multimer resolution. Any resolution function would have to operate at an efficiency many orders of magnitude greater than that observed for this system to play any meaningful role in the plasmid replication-cell division cycle. In contrast, the other plasmid-coded site-specific recombination systems, which do function in multimer resolution, have activity frequencies of the order of 100 to 10-1 per cell per generation (2,9,38).
The RSA-Pre system may be widespread among plasmids in gram-positive bacteria. pUB110, a small multicopy plasmid originally isolated from S. aureus (30)