Identification of the DNA-binding Domain of the FLP Recombinase”

We have subjected the FLP protein of the 2-pm plas- mid to partial proteolysis by proteinase K and have found that FLP can be digested into two major protein- ase K-resistant peptides of 21 and 13 kDa, respectively. The 21-kDa peptide contains a site-specific DNA-binding domain that binds to the FLP recognition target (FRT) site with an affinity similar to that ob- served for the native FLP protein. This peptide can induce DNA bending upon binding to a DNA fragment containing the FRT site, but the angle of the bend (approximately 24’) is smaller in magnitude than that induced by the native FLP protein (60’). The additional DNA bending induced by the interaction be- tween two native FLP molecules bound to the FRT site is not observed with the 2 1-kDa DNA-binding peptide. Amino-terminal sequencing has been used to map this peptide to an internal region of FLP that begins at residue Leu-148. It is likely that the DNA-binding peptide includes the catalytic site of the FLP protein. The FLP recombinase catalyzes site-specific recombination between two FLP recognition target sites present in inverted orientation on the 2-qm plasmid of Saccharomyces cerevisiae (Broach and Hicks, 1980). These sites contain three 13-base pair sequences to which FLP binds in a site-specific manner (Andrews et al., 1987). To understand the DNA-protein interactions in the FLP-mediated recombination, it is important to identify the elements of the protein which are required for site-specific DNA binding.

We have subjected the FLP protein of the 2-pm plasmid to partial proteolysis by proteinase K and have found that FLP can be digested into two major proteinase K-resistant peptides of 21 and 13 kDa, respectively. The 21-kDa peptide contains a site-specific DNA-binding domain that binds to the FLP recognition target (FRT) site with an affinity similar to that observed for the native FLP protein. This peptide can induce DNA bending upon binding to a DNA fragment containing the FRT site, but the angle of the bend (approximately 24') is smaller in magnitude than that induced by the native FLP protein (60'). The additional DNA bending induced by the interaction between two native FLP molecules bound to the FRT site is not observed with the 2 1-kDa DNA-binding peptide. Amino-terminal sequencing has been used to map this peptide to an internal region of FLP that begins at residue Leu-148. It is likely that the DNA-binding peptide includes the catalytic site of the FLP protein.
The FLP recombinase catalyzes site-specific recombination between two FLP recognition target sites present in inverted orientation on the 2-qm plasmid of Saccharomyces cerevisiae (Broach and Hicks, 1980). These sites contain three 13-base pair sequences to which FLP binds in a site-specific manner (Andrews et al., 1987). T o understand the DNA-protein interactions in the FLP-mediated recombination, it is important to identify the elements of the protein which are required for site-specific DNA binding.
One approach to identifying the DNA-binding domain of a protein has been to analyze truncated proteins whose synthesis is directed from deletion mutations of their genes in vitro (Hope and Struhl, 1986;Rusconi and Yamamoto, 1987;Henry et al., 1990) or in vivo (Kadonaga et al., 1987;Moskaluk and Bastia, 1988). We have used in vitro transcription and translation of mutant FLP genes in an attempt to localize the DNA-binding domain in this protein (Amin and Sadowski, 1989). However, it was found that the DNA-binding activity of FLP is dramatically sensitive to changes in the structure of the protein and this approach was not successful in the identification of a discrete DNA-binding domain.
Partial proteolysis and subsequent purification of the proteolytic peptides have also been used to identify the DNAbinding domains in some sequence-specific DNA-binding proteins (Abdel-Meguid et al., 1984;Smith et al., 1984;Marzouki * This work was supported by a grant from the Medical Research Council of Canada. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisenent" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$Held a Special Entrance Fellowship and an Open Fellowship from the University of Toronto.
f To whom correspondence should be addressed. Huet and Sentenac, 1987;de Vargas et al., 1988;Boulanger et al., 1989). In order to probe the location of the DNA-binding domain in FLP, we have subjected this protein to limited proteolysis. Digestion of FLP with the nonspecific protease proteinase K produced an internal polypeptide of molecular mass of 21 kDa. Gel mobility shift assays showed that this fragment bound to the FRT site in a site-specific manner with an affinity similar to that of the native FLP and formed DNA-peptide complexes with greater mobilities than those produced by the intact FLP protein. The proteolytic fragment induces bends in the substrate DNA to which it binds, but the bend angles are less than those induced by the intact FLP protein.

RESULTS
Proteolytic Digestion of FLP Generates a DNA-binding Peptide-Previous attempts to use deletion analysis to identify the DNA-binding domain of FLP were unsuccessful, possibly because the deletions perturbed the proper folding of the protein (Amin and Sadowski, 1989). We reasoned that digestion of the folded native protein might yield a proteaseresistant domain that would preserve DNA-binding activity. To identify such a DNA-binding domain in FLP, we incubated the purified FLP protein with various amounts of proteinase K for a fixed time period (Fig. 2). After the digestions had been terminated by the addition of PMSF, a labeled DNA fragment containing the wild-type FRT site ( Fig. 1) was added to the reactions and the incubations were continued. The DNA-protein complexes formed by the undigested, and proteolyzed FLP were resolved by electrophoresis on a nondenaturing acrylamide gel. The native FLP protein generates three specific complexes believed to be caused by the binding of one, two, or three molecules of FLP to the FRT site (Fig. 2, lane 12;Andrews et al., 1987;Beatty and Sadowski, 1988;Qian, et al., 1990). The proteolyzed FLP sample also generated three DNA-protein complexes with the FRT site, but the mobilities of the complexes were increased relative to those formed with native FLP (Fig. 2, lanes 3 and 4 ) . These complexes were abolished by further increasing the concentration of proteinase K in the reactions (lanes 5 and 6).
Similar results were obtained when FLP was incubated with the end-labeled DNA fragment before digestion with the proteinase K (Fig. 2, lanes 7-11). The FLP protein in the complexes was somewhat more resistant to proteinase K than the free FLP. This may be due to a conformational change of the protein upon binding or to the protection of FLP by the ' Portions of this paper (including "Materials and Methods," Table   1, and Figs. 8 and 9) are presented in miniprint at the end of this paper. Miniprint is easily read with the aid of a standard magnifying glass. Full size photocopies are included in the microfilm edition of the Journal that is available from Waverly Press.

' G A A G l T C C T A T T C C G A A G T l C C l A T T C T C T A G A A A G l A l A G G A A C T l C 3'
3'CTTCAAGGATAAGG-AGATCTTT?ATATCCTTGAAG - Gel mobility shift assay of proteinase K-treated FLP protein. Purified FLP (0.3 pg) was digested with no proteinase K (lanes 2, 7, and 12), 5 ng (lanes 3, 8, and 13), 10 ng (lanes 4, 9, and 1 4 ) , 20 ng (lanes 5, 10, and 15). or 40 ng (lanes 6, 11, and 16) of proteinase K for 15 min a t 25 "C. The proteinase K was then inactivated by the addition of 5 pg of PMSF (lanes [2][3][4][5][6][7][8][9][10][11]. For the reactions of lanes 2-6, digestion of FLP preceded the addition of 0.02 pmol of '"P-labeled FRT site-containing fragment. After an additional 15 min at 25 "C, the reaction mixtures were analyzed by electrophoresis. For the reactions of lanes 7-11, FLP protein was incubated with the labeled DNA fragment prior to treatment with proteinase K. For the reactions of lanes 12-16, proteinase K was inactivated with PMSF prior to the addition of FLP and the labeled DNA. The final reactions (50 p l ) contained 4 pg of sonicated denatured calf thymus DNA and binding assay buffer in addition to the above reagents. S, unreacted substrate; I , II, and III, complexes formed by native FLP. FI, FII, and FIII, complexes formed by proteolyzed FLP. DNA substrate (compare lanes 3-6 and lanes 8-1 1 ). A control experiment (Fig. 2, lanes 12-16) indicated that proteinase K was completely inhibited by PMSF.' The electric charge of a DNA protein complex in a neutral nondenaturing gel is carried mainly by the DNA component in the complex, whereas the charge on the protein does not have a great effect on the mobility (Boulanger et al., 1989). Therefore, the much faster migration of the DNA-proteolyzed FLP complexes was most likely caused by a substantial reduction of the protein mass of the FLP protein.
Products of Proteinase K Digestion of FLP-To identify the peptide(s) of FLP that retained the ability to bind to DNA, we analyzed the products of proteinase K digestion of FLP by SDS-PAGE (Laemmli, 1970). Ten minutes of digestion at 0 "C led to the disappearance of the FLP protein (-45 kDa) and to the appearance of three major peptide bands of approximately 32 (P32), 27 (P27), and 13 (P13) kDa, respectively (Fig. 3A). Determination of the amino-terminal sequences of these peptides indicated that P27 was derived from the amino-terminal end of FLP and that the amino-terminal end of P32 mapped to amino acid 124 of FLP. The 13-kDa band was found to contain two peptides: the amino-terminal The abbreviations used are: PMSF, phenylmethanesulfonyl fluoride; AMV, avian myeloblastosis virus; bp, base pair(s); DTT, dithiothreitol; FPLC, fast protein liquid chromatography; FRT, FLP recombination target; IPTG, isopropylthiogalactoside; SDS, sodium dodecyl sulfate; PAGE, polyacrylamide gel electrophoresis. A, electrophoretic profile of peptides generated by partial proteinase K digestion of FLP. Twenty-four pg of FLP was digested with 400 ng of proteinase K in 100 pl of binding assay buffer for various periods of time a t 0 "C. After digestion, 10 pg of PMSF was added to each tube to inhibit proteinase K. Forty pl of each digest (-10 pg protein) was analyzed on a 15% SDS-acrylamide gel, and the the gel was stained with 0. sequences of peptides were obtained as described under "Materials and Methods." The approximate sizes of the major peptides P32, P27, P21, P13, and P11 are 32, 27, 21, 13, and 11 kDa, respectively. The numbers at the left indicate molecular masses for protein markers that were run on the same gel. B, amino acid sequences of the partial proteolytic peptides of FLP produced by proteinase K. About 10 pg of the same proteolyzed FLP sample as in Fig. 3 was used for peptide sequencing. The first six amino terminal amino acids of the peptides from partial proteolysis of the FLP are aligned with the corresponding sequences of the native FLP protein from which they were obtained. A nonidentical residue is marked with a black dot. Brackets indicate that no amino acid could be identified during this cycle. Two 13-kDa peptides were identified by microsequencing but the P13a was 5-10fold more abundant than P13h. The first residue of P27 and P13a was shown to be proline instead of methionine, probably because the methionine of FLP is removed in Escherichia coli (Babineau et al., 1985). The numbers indicated above the FLP are the amino acid numbers in the FLP protein.
C, locations of peptides in the FLP protein. Black solid lines, peptide segments from partial proteolysis. The beginning residue is determined from amino-terminal sequencing, but the carboxyl terminus of each peptide segment is estimated from the size of each peptide on SDS-PAGE. Two regions ( I and I I ) indicated at the top are highly conserved (more than 60% match a t the nucleotide sequence level) among six FLP proteins from six 2pm-like plasmids (Utatsu et al., 1987). The residue numbers indicated are amino acids of FLP from the 2-pm plasmid of yeast. sequence of one (P13a) corresponded to that of mature FLP whereas the amino-terminal sequence of the other corresponded to amino acid 148 of FLP (Fig. 3B). However, P13a was found to be 5-10-fold more abundant than P13b in molar concentration (data not shown).
With increasing digestion time, new peptides of approximately 21 (P21) and 11 kDa (P11) appeared, and the amounts of bands P32 and P27 declined (Fig. 3A). The amino-terminal sequence of P21 was shown to be identical to the internal FLP sequence beginning at Leu-148, suggesting that P21 might be derived from P32 (Fig. 3B). After a 150-min digestion at 0 "C, three major bands were observed on an SDSpolyacrylamide gel, P21, P13, and P11 (Fig. 3A, lane 8 ) . Two different peptides were present in the 11-kDa band with their amino-terminal sequences mapping to proline 2 and leucine 148, respectively (not shown). These results indicate that the FLP protein can be digested into at least two major nonoverlapping peptides (P21 and 13a) by proteinase K. These results are summarized in Fig. 3C.
Proteolyzed Products of FLP Bind to the FRT Site-We then tested these proteolyzed products of FLP for their ability to bind to an FRT-containing DNA fragment (Fig. 4). Even the most extensively digested FLP sample retained DNA binding activity, forming complexes FI, FII, and FIII (Fig. 4, lane 6). The amount of complex FIII formed was less than that obtained with the less proteolyzed sample of FLP (Fig.  4, lanes 2-4). These observations indicated that a FLP-derived peptide of 21 kDa or less retained the ability to bind to the FRT site. At the beginning time points of the digestion, some intermediate complexes were observed (lanes 2 and 3); however, only three complexes (FI, FII, and FIII) with the increased rate of migration were found after 30 min of digestion (lanes 4-6).
Purification of the Peptides Resulting from Partial Proteolysis of FLP-To determine which peptide(s) was actually the DNA-binding species, we subjected extensively digested FLP to ion exchange chromatography (see "Materials and Methods") ( Fig. 5A). Three UV absorbing peaks ( I , 11, and IZI) were observed. Peak I consisted of a 13 kDa peptide and a smear of lower molecular weight (Fig. 5B, lane 1 ), and peak ZZ contained a protein with a molecular mass greater than FLP that was likely a contaminant (Fig. 5B, lane ZZ). Neither Ion exchange chromatography of a partial digest of FLP by proteinase K. FLP was digested with proteinase K at a mass ratio of 5 0 1 (FLP:proteinase K, w/w) for 20 min a t 25 "C, followed by addition of PMSF to a final concentration of 100 pg/ml. The proteolyzed FLP sample was filtered through a 0.45-pm filter (Millipore) and kept on ice prior to chromatography on a Mono

11350
FLP DNA-binding Domain peaks I nor II contained DNA binding activity (Fig. 5C, lanes   3 and 4 ) . Peak ZZZ contained a major 21-kDa band and a minor 19-kDa band and had the same DNA-binding activity as the unfractionated proteolysed FLP (Fig. 2, lane 3 Determination of the amino-terminal sequence of the peptides present in peaks I and III indicated that the 13-kDa peptide represented P13a and that the amino termini of both the 21-and 19-kDa peptides mapped to Leu-148 of FLP. These results indicate that FLP contains an internal 21-kDa proteinase K-resistant domain that retains DNA binding activity. The DNA Binding Specificity and Affinity of the DNAbinding Peptide-To compare the specificity and affinity of the DNA binding of the 21-kDa peptide with that of native FLP, the purified P21 or intact FLP was incubated with a FRT-containing DNA fragment or a non-FRT-containing DNA fragment. Both the native FLP and P21 formed three complexes with the FRT-containing fragment (Fig. 6A, lanes  3 and 5), although the peptide formed less complex I11 (lane 5 ) than did native FLP. We also found that both of the proteins formed a small amount of a DNA-protein complex with a non-FRT-containing fragment (lanes 4 and 6). The amount of complex produced by the native FLP was severalfold less than that produced by P21, suggesting that the DNAbinding domain contained in P21 might have less stringent specificity than the native FLP.
A competition experiment was carried out to confirm the site specificity of the P21 DNA binding peptide. A linearized FRT site-containing plasmid or nonspecific plasmid was used as competitor. The specific complexes formed by the native FLP and P21 were competed efficiently with an excess of the specific plasmid (pBL112) (Fig. 6B, lanes 2-5 and lanes 10-13). In contrast, a nonspecific plasmid (pUC19) was much less efficient in competing for the formation of the specific complexes (lanes 6- 9 and 14-1 7). The band labeled "RP" in Fig. 6B is a recombination product between the labeled FRT fragment and the specific competitor plasmid.
The Purified DNA-binding Peptide Bends DNA-We have found previously that the FLP protein bends the DNA of the FRT site upon binding (Schwartz and Sadowski, 1989). We have measured the angles of the bends induced by FLP in the FRT site and found that the binding of the native FLP to the target site with two inverted symmetry elements (a and b) separated by an 8-bp core region causes two types of DNA bending: the type I bend occurs when one FLP molecule binds to a single symmetry element; the type I1 bend occurs when two FLP molecules bind to the two symmetry elements on opposite sides of the core region (Schwartz and Sadowski, 1990). The type I bend angle is approximately 60", and the type I1 bend is greater than 140". The fact that the type I1 bend angle is much greater than the sum of two type I bends is believed to be the result of protein-protein interactions (Schwartz and Sadowski, 1990).
To investigate whether the isolated DNA-binding peptide induces bending on binding to FRT-containing DNA, the bending assay was carried out using fragments that contained a partial FRT site in the middle or at the end of the fragment. DNA bending is indicated when the mobility of the fragment with the FRT site in the middle is slower than the mobility of the DNA with the FRT site at the end. The bending angles were measured as described by Schwartz and Sadowski (1990) by comparing the mobility of the bent DNA fragments with that of a series of DNA standard fragments that contain sequence-directed bends of known magnitude (Thompson and Landy, 1988). were incubated with increasing amounts of linearized FRT site-containing plasmid (pLB112) or a nonspecific plasmid (pUC19) for 10 min a t 25 "C. Then 0.001 pmol of labeled FRT site-containing fragment (EcoRI-BarnHI fragment from pLB112) was added to each reaction, and the incubation was continued for 20 min. The reactions were analyzed on a native acrylamide gel. Lane I, substrate alone; lanes 2-5, native FLP and FRT sitecontaining plasmid (pLB112); lanes 6-9, native FLP and non-FRTcontaining plasmid (pUC19); lanes 10-13, P21 and pLB112 (specific competitor); lanes 14-17, P21 and pUC19 (nonspecific Competitor). The molar excesses of competitor plasmids: lanes 2,6,10, and 14, no competitor plasmid; lanes 3, 7, 11, and 15, 100-fold excess; lanes 4,8,   12, and 16, 500-fold excess; lanes 5, 9, 13, and 17, 1000-fold excess.
The free substrate and the complexes are indicated in the same way as in Fig. 2. RP, recombination product formed between substrate fragment and cold FRT-containing competitor plasmid.
We found that the DNA of complexes FI and FII induced by P21 contained a bend (compare the positions of complexes FI and FII in Fig. 7, lane 3 versus 4 ) . However, the bend angles for the DNA complexes FI and FII were much smaller DNA-bending assay of the purified DNA-binding peptide (P21). Two 300-bp fragments containing two FLP recognition elements (a and b) from pCS38 were used as the bending assay substrate (Schwartz and Sadowski, 1990). The complexes formed between native FLP or P21 and the fragments containing the partial  Thompson and Landy (1988 and 4 versus lanes 5 and 6 ) . We estimated the bending angles using a series of fragments containing sequence-directed bends and found that the bending angle of the DNA in complex FI was approximately 24", whereas the bend angle of the DNA in complex FII was about 37". The corresponding values for the wild-type FLP are 60 and >144" (Schwartz and Sadowski, 1990).

DISCUSSION
The initial step in the FLP recombination reaction is the site-specific binding of the FLP protein to the FRT site. To further understand the interaction between FLP and the FRT site, it is important to identify the DNA-binding domain in FLP. So far, there is no evidence that FLP contains any of the previously defined DNA binding motifs, e.g. helix-turnhelix, helix-loop-helix (Murre et al., 1989), leucine zipper (Busch and Sassone-Corsi, 1990), or zinc finger (Evans and Hollenberg, 1988).
Our previous attempts to identify the DNA-binding domain of FLP using an in vitro transcription and translation system were unsuccessful, possibly because the deletion or insertion mutations used may have disturbed the proper folding or thermostability of the protein (Amin and Sadowski, 1989).
In this paper, we report the use of partial proteolysis by proteinase K to search for a structural domain that preserves the DNA binding specificity of the FLP protein. We found that FLP could be digested into two major peptides which may represent distinct functional domains. A 21-kDa peptide was purified and was shown to contain a DNA-binding domain. A function for P13 has not yet been identified. The P21 peptide is able to interact with the full FRT site to form three specific complexes. The affinity of this peptide for the FRT site is comparable with the intact FLP protein (Fig. 6 B ) , although this peptide does not form complex I11 as well as intact FLP (Fig. 6B). However, we have also found that the specificity of this isolated DNA-binding domain may be less stringent than that of native FLP (Fig. 6 A ) . Although we have not observed a DNase I footprint produced by the 21-kDa peptide,:' it should be pointed out that even the binding of intact FLP to a single 13-bp symmetry element of the FRT site (formation of complex I) does not give a prominent DNase I footprint (Andrews et al., 1987). Apparently, the production of an identifiable DNase I footprint by intact FLP requires protein-protein interactions in complexes I1 and 111. These interactions may not be possible for peptide P21. The DNA-bendingassay has shown that the 21-kDa peptide induces a DNA bend upon binding to the FRT site, although the bending angle induced by P21 is less than that caused by native FLP. The extreme bend which is thought to be the result of the interactions between two FLP molecules bound to the same substrate (type I1 bend, Schwartz and Sadowski, 1990) is not induced by P21. We hypothesize that some regions of the FLP protein that are necessary for contact between the protein and DNA are absent from P21. These contacts are apparently necessary for bending although not specificity or even high affinity. It is possible that the regions required for protein-protein interactions needed for the type I1 bend are also absent. Alternatively, these regions may be present but are not close enough together to interact properly in the FII complex.
FLP has two regions that are highly conserved among six FLP proteins from six 2-pm-like plasmids of various yeast strains (Utatsu et aZ., 1987). They cover amino acid residues 185-203 and 295-313 of 2-pm plasmid FLP. It is likely that P21 includes both conserved regions (Fig. 3C) and the residues in these conserved regions play an important role in forming the DNA binding domain. We have found that several mutations in these two conserved regions give a binding-deficient and bending-deficient phenotype: Three residues (His-305, Arg-308, and Tyr-343) are absolutely conserved among the integrase family of recominases (e.g. Int, Cre, and FLP) (Argos et al., 1986). It has been shown that Tyr-343 is directly involved in DNA cleavage, Arg-308 may help this step and His-305 is essential for the strand exchange and religation reactions (Parsons et al., 1988). Conserved region 11 contains His-305 and Arg-308 (Fig. 3C). We do not know whether P21 contains Tyr-343 as well, but, we have not observed any DNA cleavage activity from P21. Inat shown). This band was present one hour after induction and its intensity increased markedly after three hour6 of induction.
It therefore seemed that a large amount of inactive FLP protein was being expressed from the Tl promoter. We therefore sought methods which might facilitate the synthehls of enzymatically competent protein and found that inductlan of FLP expression at low tenrperature led to the synthesis of substantla1 amounts of active protein.

When we induced FLP expression at 2S0c
we obtalned about forty-fold more FLP activity in crude cell extracts than we previously detected in cell extracts where FLP w a s expcessed from the % promoter at 3 1 ' .

Protein-induced DNA bending assay
The plasmid pcs38 was used far preparation of the hubstrate in the bending assay of the purifled 21 kUa peptlde. The preparation of the substrate, the bending assay, and the estimat~on of the magnitudes of the protein-induced bends have been detailed elsewhere ISchwactz 6 Sadowski, 1990, Thompson and Landy. 198s