Gene, 119 (1992) 37-48 0 1992 Elsevier Science Publishers

GENE

B.V. All rights reserved.

37

0378-l 119/92/$05.00

06603

Bending-incompetent variants of Flp recombinase mediate strand transfer in half-site recombinations: role of DNA bending in recombination (Site-specific

recombination;

recombinant

DNA;

FR T site; FBE site)

Jing-Wen Chen, Barbara Evans, Hans Rosenfeldt

and Makkuni

Jayaram

Department of Microbiology,Universityof Texas at Austin, Austin, TX 78712, USA Received

by S. Adhya:

24 January

1992; Revised/Accepted:

4 April/20

April 1992; Received

at publishers:

18 May 1992

SUMMARY

One key feature of the interaction of Flp recombinase with its target site (FR T) is the large bend introduced in the substrate as a result of protein binding. The extent of bending was found to depend on the phasing and spacing of the Flp monomers occupying the two Flp-binding elements (FBE) bordering the strand-exchange region (spacer) of the substrate. The relative mobilities of the Flp complexes formed by the two permuted substrate fragments, containing the FRT site near the end or in the middle, corresponded to a DNA bend of approx. 140” when each of the two FBEs flanking the spacer was occupied by a protein monomer. The estimated bend angle was the same when the reference DNA fragment with the FRT site at the end was substituted by one with the site in the middle, but containing a 4-bp insertion within the spacer. We used a combination of wild-type Flp and Flp variants that were competent or incompetent in DNA bending, together with full, or half FRT sites, to ask whether bending is a conformational requirement for catalysis, namely cleavage and exchange of strands. We obtained the following results: in full-site (FRT) vs. full-site recombinations or in full-site vs. half-site (half FRT) recombinations, there was a large difference in the reactivity between Flp and a bending-incompetent Flp variant. This difference virtually disappeared when reactions were done with half-FRT sites. We conclude that bending is not a prerequisite for catalysis, but represents the manner in which the substrate accommodates the Flp protomer-protomer interactions that are pertinent to catalysis.

INTRODUCTION

The Flp recombinase from S. cerevisiue catalyzes bination between a pair of 34-bp long substrates

recom(FRT),

Correspondence to: Dr. M. Jayaram, Department of Microbiology, University of Texas at Austin, Austin, TX 78712, USA. Tel. (512)471-0966; Fax (512)471-5546. Abbreviations: bp, base pair(s); deoxyribonucleoside triphosphate; I); Flp, site-specific

recombinase

BSA, bovine serum albumin; dNTP, FBE, Flp-binding element (see Table enzyme of the 2 p circle plasmid;

Flp target site (see Table I and INTRODUCTION); ture; kb, kilobase nucleotide; fragment

PEG,

or 1000 bp; nt, nucleotide(s); polyethylene

HP, hairpin

oligo, oligodeoxyribo-

glycol (M, 6000); PolIk,

of E. coli DNA polymerase

FR T, struc-

Klenow

(large)

I; S. Saccharomyces; wt, wild type.

each consisting of an almost perfect dyad of two 13-bp protein-binding elements (FBEs) bordering a strandexchange region of 8 bp (FRT spacer or core) (Jayaram, 1985; Senecoff et al., 1985; Gronostajski and Sadowski, 1985). The Flp protein binds to each of the FBEs, presumably as a monomer (Andrews et al., 1987; Prasad et al., 1987). The recombination reaction involves cooperative protein-protein interactions between Flp monomers bound to two FBEs and to two FRT sites (Prasad et al. 1986; Parsons et al., 1990; Qian et al., 1990). One striking aspect of the Flp-FR T complex in which the two FBEs have been occupied by Flp is the resultant large DNA bend of over 144” (Schwartz and Sadowski, 1990). The center of the bend has been localized to the FRT spacer region. Point mutations within Flp that reduce or

38 TABLE

I

Symbols

and sequences

0A

of FRT sites and half FRT sites a

Site

Symbol

0B

whereased:

FRT

l’b

FRT

-

sequtncesof hair

FRT andSpar

Fig. 1,2B, 3,7

la

Fig.4A,B,C&D; Fig. 6A, B; Fig. 8B, C, D; Fig. PB, C, D

l’a S’E3’ 3’ ~CAAGGAT~G

5’ GAAGTWCTATAC 3’ CTTCAAGGATATG

3’ 5’

5’ 1,

l’b 5’ GAAGTTCCTA’ITC 3’ CTKAAGGATAAG

Fig. 4A, B, C & D; Fig. SA, B; Fig. 6B,C l’* HOP

Half FRT (fight)

a The Flp-binding

elements (I%.&) are named

FBEs and an 8-bp spacer. (sandwiched truncated strand

between spacer

In the FRT

la and

la, l’a, and l’b and are symbolized

site containing

of the top strand

(short hatched

by a dashed

arrows. between

line in Fig.

bar) in the left half FRT is the trinucleotide,

Usually, 1 ‘a and

3’ 5’

FRT is composed

of two facing 13-bp

1 ‘b. The sequence

of the FRT spacer

I and by a hatched bar in Figs. 4C, D and 5. The

5’-TTT-3’.

The corresponding

sequence

in the bottom

(see Fig. 2A).

eliminate substrate bending have been identified. These map to G1y328 (G328R, G328E), Tyr6’ (Y60D, Y6OS) and Tyr343 (Y343S) (Schwartz and Sadowski, 1989; Chen et al., 1991a). mutations are represented by the wt residue, followed by the position number, followed by the mutant residue. For example, G328R means that the glytine at position 328 has been altered to arginine. Tyr343 is the active site residue of Flp that gets covalently attached to FRT DNA during the first trans-esterification (strand cleavage) step of recombination (Evans et al., 1990). It is not surprising, therefore, that alteration of Tyr343 results in a loss of catalytic activity (Prasad et al., 1987; Evans et al., 1990). However, even the other nonbending Flp variants are either inactive or extremely poor in strand cleavage (Schwartz and Sadowski, 1989; Chen et al., 1991a). The striking correspondence between bending incompetence and catalytic inefficiency raises two interesting questions. Is the bent configuration of the substrate within the synaptic complex a structural requirement for catalysis? This is possible if bending strains the phosphodiester bond at the strand-exchange point and prepares it for breakage. Or, is bending a passive m~ifestation of the structural distortion in DNA caused by the interactions of Flp monomers bound on either side of the FRT spacer region? The aim of present study was to distinguish between these alternatives. Two results presented in this paper prove the latter proposition to be correct. First, a substrate, freed virtually

by horizontal

all three FREs, there is a 1-bp spacer

I ‘a) is given in italics. The spacer is represented

of the right half FRT is 5’-TCT-3’

Spacer 5’ TTXTAGA 3’ AAAGATCT

Fig. 5A, B; Fig. 6C; Fig. SA, B, C; Fig. PA, B, C

imv

3’ 5’

from the constraints of bending by design, can still efficiently undergo recombination. Second, two Flp variants that fail to bend DNA can effectively utilize such a substrate to give the expected re~ombin~t products. RESULTS

AND DISCUSSION

(a) The FRT substrate bending by Flp: effects of spacer length and protein phasing The substrate used for the bending studies included the 34-bp minimal FRT site that consisted of two 13-bp Flpbinding arms (FBE) flanking the 8-bp core (or spacer) in a head-to-head orientation (Fig. 1, top; see Table IA, B). This fragment was placed within the EcoRV site in pBR322 and then excised as an approx. 400-bp ~~~dIII-~u~HI fragment for bending assays. The FRT site was located almost exactly at the center of this fragment. The electrophoretic profile of the complexes formed by this fragment with wt Flp is shown in Fig. 1 (lane 2). As expected, two complexes, CI and CII, were formed, corresponding to the occupancy, presumably, of one or two monomers of Flp, respectively (Andrews et al., 1987; Prasad et al., 1987). The retardation of complex CII relative to CI was disproportionately large and was consistent with a large DNA bend within this complex (Schwartz and Sadowski, 1990; see Table II. Since this large bend required the filling of both FBEs of the substrate by Flp, it seemed likely that bend-

39 la

: ____---_ l’a

matically I 5

by a 4-bp insertion

within the spacer region of the

FRT site (lane 6, Fig. 1). The normal reveal protein-induced

6

bends

mobility

by the relative

shift assays difference

in

the electrophoretic retardation of complexes formed by a pair of permuted substrates - one containing the binding site in the middle and the other containing the binding site at the end (Wu and Crothers, 1984). The mobility difference is converted to a corresponding bend angle from an empirical calibration curve obtained by a set of standard DNA fragments containing intrinsic sequence-induced

CII

bends (Thompson and Landy, 1988). For Flp, the bend angle for CII measured against a permuted fragment, in which the FRT site was near one end, was approx. 140” CII

CI

CI

S

S

Fig. 1. DNA bending consists

by the Flp recombinase.

of the Flp-binding

The Flp target (FRT) site

elements (FEE) labeled

la and l’a (top). The

strand-exchange region (spacer) is shown by the dashed line. The sequences of la, I’a and the spacer are given in Table I. The substrates for the binding reactions

consisted

of a 400-bp fragment

site close to its middle (panel M), its permuted site at the end (panel E), or the fragment dle, but containing

a 4-bp insertion

I). Binding reactions action volume,

approx.

The reactions

(see Fig. 2) within the spacer contained,

0.1 pmol of the substrate/50 pg calf thymus

were incubated

and were fractionated

the FRT

containing

the

with the FRT site in the mid-

with Flp or a Flp variant

7.5/60 mM NaC1/200

containing

equivalent

(panel

in a 50-~1 re-

mM Tris.HCl

pH

DNA per ml/100 pg BSA per ml.

at room temperature

on a 5 y0 nondenaturing

or on ice for 30 min,

polyacrylamide

gel (30: 1

cross-linked). Lanes 1, 3 and 5 are free substrates; lanes 2, 4 and 6 are binding reactions with wt Flp. CI and CII are complexes resulting from the binding, presumably, input ratio was approx.

of one or two monomers three-four

(FBE).Methods. The fragment tained by cloning

monomers

of Flp, respectively.

The

of Flp per binding element

with the FRT site in the middle was ob-

the site in the EcoRV

site of pBR322,

thus placing

it

virtually equidistant from the Hind111 and BumHI sites of the plasmid. The 400-bp substrate fragment was obtained by Hind111 + BamHI digestion.

The permuted

substrate

with the site at the end was constructed

as follows. The FRT site bordered

by EcoRI and BumHI

300-bp

from pBR322,

HindIII-BumHI

Hind111 backbone

fragment

fragment

from pUC19

overhangs,

the

and the large EcoRI-

were joined together

in a three-

fragment ligation reaction. The resulting plasmid was cut with Hind111 + EcoRI to retrieve the substrate fragment with the FRT site at the end. The 4-bp insertion the X&I

site present

reaction

in the presence

substrates

within the spacer was obtained

in the spacer,

filling-in the overhangs

of all four dNTPs,

and ligating the flush ends. The

for the bend assay were end-labeled

[ r-32P]dNTPs,

plus the other unlabeled

by cutting at by the PolIk

by PolIk using one or two

dNTPs.

ing was a direct consequence of across-the-spacer interactions between the Flp monomers. If so, these interactions could potentially be dependent on the phasing of the protein monomers on the presumed B DNA helix. Consistent with this idea, the retardation of CII was decreased dra-

(Fig. 1; Table II). The value did not change when the reference was the 4-bp insertion substrate (Fig. 1; Table II). This estimate compares well with a bend of 144” or larger derived by Schwartz and Sadowski (1990). To test further the idea of phased Flp-Flp interactions on DNA, we constructed two more insertion substrates by introducing lo- and 12-bp sequences within the FR T spacer (Fig. 2A). The gel retardation of complexes formed by the normal substrate and those containing 4, 10 and 12-bp FRT spacer insertions were compared (Fig. 2B). Notice that the 4, 10 and 12-bp additions increase the spacer length from the normal 8-bp to 12, 18 and 20-bp respectively (Fig. 2A). Relative to the 4-bp insertion substrate, the substrate with the lo-bp insertion showed a bend angle of 66” (compare lanes 6 and 4; Fig. 2B; Table II). Insertion of 12 bp resulted in a smaller DNA bend of 44” in CII (compare lanes 8 and 6 to lane 4, Fig. 2B; Table II). These results are most easily explained by the hypothesis that the Flp monomer contacts with the substrate occur on the same face of the DNA. Alternatively, the protein-protein interactions responsible for DNA bending may be mediated by two different Flp domains. For example, the DNAbinding domain of one Flp monomer may interact with a domain of the second Flp monomer that is not directly involved in substrate binding. Point mutations of Flp that eliminate or reduce DNA bending map to two distinct domains of Flp. Tyr6’, whose alterations cause bending deficiency (Chen et al., 1991a; results of this study) falls within domain IA (Chen et al., 1991b), that is not essential for DNA binding. Two other residues, G1y3** and Tyr343, that are also required for substrate bending fall within or close to domain IB, that is directly responsible for DNA binding (Schwartz and Sadowski, 1989; Chen et al., 1991a, 1991b; Pan et al., 1991). Thus, a phase shift between the interacting domains of the Flp monomers may compensate for a possible out-of-phase location of the DNA-binding domains of the protein monomers on the substrate. The oligo cassettes used ‘to obtain the lo- and 12-bp insertions within the FRT site spacer were G+C abundant (Fig. 2A). Use of A+T-rich sequences was deliberately

40 TABLE

II

Extent of substrate

bend induced

Protein il

by Flp and its variants

Substrateb

FlP

Mobility

and bend angle values’

nr

Bend angle

ur

Bend angle

N (M)

0.95

40” (36”)

0.5

140” (120”)

N (E) +lO (M)

0.93

43” (43”)

0.5

140” (120”)

0.95

40” (36”)

0.84

66”

(66”)

+12 (M)

0.95

40” (36”)

0.94

44”

(40”)

Flp(Y60F)

N (M)

0.95

40” (36”)

0.5

Flp(Y6OD)

N (W

0.96

40” (33”)

0.95

40”

(36’)

Flp(Y60S)

N

0.93

43” (43”)

0.90

52”

(52’)

a Flp is the wt recombinase.

(El

The variants,

’ N refers to the normal FRT site containing and

~12, respectively

spacer.

substrate

was a permuted

quite similar. Hence bend angles computed of a complex

(p) of a complex formed

plex formed by the reference with standard (Thompson

substrate.

bends (Thompson and Landy,

substrate fragment

containing

by changing

TyrbO to Phe, Asp and Ser, respectively.

spacer insertions

of 10 and 12-bp are indicated

with the FRT site in the middle, but containing

was the fragment

the FRT site at the end. The mobilities of the reference

containing

as the ratio of the distance

containing

it migrated

to that migrated

the FRT site in the middle was calculated

The ur values were converted

and Landy,

were obtained

a 4-bp insertion fragments

by + 10 in the

M and E were

using M and E were nearly identical.

was computed

by the substrate

and Flp(Y6OS),

the 8-nt spacer (see Fig. 2A). Substrates

(Fig. 2A). (M) The reference

(E) The reference

’ The mobility

Flp(Y6OF),Flp(Y6OD~

140” (120”)

by dividing

to bend angles by using a calibration

198). The values in parentheses

were calculated

by the free substrate. its mobility

curve obtained

by using the relation

The relative mobiIity (ur)

by that of the corresponding

com-

with a set of DNA fragments

ur = cos(cc/2), where r is the bend angle

1988).

d CI and CII are complexes

formed

by association

of one or two Flp monomers

avoided to minimize the chances of introducing sequencedependent bends. It is possible that the G+C richness may in some ways oppose the Flp-induced bend. Recall that the lo-bp insertion caused a larger CII bend with reference to the 4-bp insertion, but the bend angle of 66” was significantly smaller than the normal 140”. It is also possible that the increase in distance by a helical turn between the Flp monomers may be responsible for the reduction in bending from 140” to 66”. The 12-bp insertion is expected to shift the protein phasing and spacing in CII by approx. 68 o and 6.8 A, respectively, relative to the lo-bp insertion. The bend angle in CII for the 12-bp insertion substrate was only 44” (compare lanes 8 and 4; Fig. 2B; Table II). Overall, the estimated bend angles for the substrates with altered spacer lengths conform to phase- and, perhaps, dist~ce-dependent cooperative interactions between Flp monomers. Each Flp monomer bound to the FRT site introduces an individual bend of approx. 40”. If, within complex CII, the two individual bends are perfectly phased, the total predicted bend would be 40 + 40 = 80 * . The measured bend of 140’ (Table II) includes the additional bend (60”) contributed by protein-protein cooperativity. Insertion of 4 bp between the FBEs places Flp monomers virtually on opposite faces of the DNA. Even though the bends would not be exactly coplanar, they would nearly cancel out. Insertion of 10 bp, on the other hand, should reestablish normal protein phasing. The measured DNA bend in CII of 66” with this substrate approximates the

with the FRT site, respectively

(see Fig. 1, for example).

sum of the individual bends (80”). The extra 60” arising from Flp-Flp interaction is now lost.

(b) DNA bend angles in complexes of Tyr@ variants of Flp The Tyr6’ of Flp corresponds to a conserved residue in the yeast family site-specific recombinases (Utatsu et al., 1987). Alterations of this residue in Flp resulted in Flp variants that produced conformationally altered DNAprotein complexes (Chen et al., 1991a). Gel retardation of complexes formed with a permuted set of substrates, approx. 150 bp long, indicated that Flp(Y 60s) and Flp(Y6OD) were defective in DNA bending while Flp(Y60F) was not. The assay, however, was not sensitive enough to assign quantitative differences in bending. The extent of DNA bending by two of the Tyr6’ variants, Flp(Y60F) and Flp(Y6OD~, was measured using the 400-bp DNA fragment with the FRT site in the middle (as in Fig. 1 and Fig. 2). The bending angles were derived by comparison to the fragment containing the 4-bp spacer insertion that effectively eliminated bending (Fig. 3, lane 3). As pointed out earlier, the mobilities of Flp complexes formed by the 4-bp insertion fragment were equivalent to those formed by a permuted fragment with the FRT site at the end. The CI and CII mobilities with the insertion substrate were the same for Flp, Flp(Y60F) and Flp(Y6OD) (data not shown). The DNA bend in CII (occupancy of two Flp monomers) was estimated to be approx. 140” for Flp~6OF~ (Fig. 3, lanes 6 and 6’; Table II). Therefore, Flp(Y60F) was as

41

A

1

FOP 2 3

Flp(Y60F) 5’ 5 6

4

Flp(Y60D) 8 I’ 8’

7

6’

Spacer

Insertion

T T TCTAGA AAAGATCT +4bp

T T TCTAGCTAGA AAAGATCGATCT

+lObp

T T TCTAGCGATCGCTAGA AAAGATCGCTAGCGATCT

+12bp

T T TCTAGCGACGTCGCTAGA AAAGATCGCTGCAGCGATCT Fig. 3. Relative DNA bending by wt Flp, Flp(Y60F)

+lObp 5 6

+4bp 3 4

+12bp 7 8

assays used wt Flp and Flp variants or Asp. The binding reactions gel (30 to 1 cross-linked).

were fractionated

Lanes

and Flp(Y60D).

in which Tyr@’ was replaced

The

by Phe

on a 5% polyacrylamide

1, 4 and 7 correspond

to the fragment

with the FRT site in the middle (see Fig. 1, M) to which no protein was added.

Lanes 2, 5 and 6, and 7 and 8 represent

fragment

with Flp, Flp(Y60F)

through

8’ are overexposures

Flp(Y60D)

concentration

and Flp(Y60D),

of the corresponding

in lane 8 was one-third

in lane 2. Higher amounts

of Flp(Y60D)

plexes that were not well resolved. to the substrate

in the spacer

substrate

showed

For methods,

Fig. 2. Flp-induced (A) The spacer

bending

sequences

of substrates

of the normal

containing FRT

spacer

insertions.

site and those containing

insertions are indicated. The inserted sequences are indicated by bold letters in italics. The 4-bp insertion was obtained by filling-in the XbaI site within the spacer by the PolIk reaction lo- and 12 bp insertions 5’-CTAGCGATCG

and

or to those containing

resultant

complexes

substrate

with the 4-bp insertion

in the binding

the indicated

were fractionated

insertions

that contained

per FBE. For details,

FR T

(see A), and the

in a 50/, polyacrylamide

migrated

in a slightly aberrant

gel. Lanes 1, 3, 5 and 7 are controls.

reactions

four monomers

The

GCTGCAGCGATC,

at the XbaI site. (Panel B) Flp was bound to a normal

substrate

are binding

of all four dNTPs.

by ligating the oligo cassettes,

5’-CTAGCGACGTCG

GCTAGCGATC respectively,

in presence

were obtained

gel. The manner

Lanes 2, 4, 6 and 8

wt Flp at an input ratio of three to see Fig. 1 legend.

competent as wt Flp in effecting the large substrate bend in the CII complex (compare lanes 6 and 6’ to lane 2; Fig. 3). In CII derived from Flp(YBOD), the calculated bend was only 40” (Fig. 3, lanes 8 and 8’; Table II). Thus, substitution of Tyr6’ of Flp by Phe did not alter the degree of DNA bend in CII, whereas substitution by Asp reduced it markedly. (c) Strand-transfer reactions by half-site substrates The design of half FRT sites followed that of the half attachment sites used to study I Int-mediated recombina-

with the centrally

insertion

binding patterns respectively.

lanes 5 through

resulted in broad smeared

com-

of Flp binding

Flp site, but containing

(see Fig. 1, Table I). The uncomplexed

a slightly aberrant

8.

the Flp concentration

Lane 3 is the pattern located

of this

The lanes 5’

a 4-bp insertion

mobility in this gel system (lane 3).

see Fig. 1 legend.

tion and is described in detail elsewhere (Nunes-Duby et al., 1989; Serre et al., 1992). Each half FRT site consisted of one FBE sequence, the Flp cleavage site on one of the strands followed by 3 nt of the spacer, and the entire spacer sequence on the other strand ending in a free 5’-hydroxyl (see Table IA, B). The 5’-hydroxyl mimicked the terminus of a spacer that had been cleaved by Flp and rendered competent to function as a strand acceptor. The presence of one FBE, one strand-transfer (cleavage) site and one strand-acceptor site makes the half FRT site potentially capable of exactly half the reactions of a normal full FRT site. The FRT site could thus be split into two half FRT sites - a left half-site and a right half-site. The significant aspect of the half-site configuration pertinent to our studies is that its spacer is not covalently connected to a second Flp-binding element. The results of a cross between the left half-site and a full-site carried by a circular plasmid (see Table IA, B) are shown in Fig. 4A. The half-site(s), end-labeled with 32P, reacted efficiently in the assay giving rise to a series of recombinant bands (R). The first cross-over event, the result of one strand exchange each at the left and right ends of the spacer region, would linearize the circle (see Fig. 4C), and repeated rounds of integration of the circle into the linear product would result in a ladder of recombinants. Digestion of the recombinants with ScaI gave two labeled fragments, Rl and R2 (Fig. 4B). Fragment Rl (the pre-

42

-Flp

1

2

3

-Flp

+ScaI 1 2

3

R2 Rl

dominant one) was the digestion product expected from a normal recombination event in which the half-site and fullsite spacers were aligned in the homologous mode (Fig. 4C; Serre et al., 1992). Fragment R2 was derived from the product of a ‘contrary’ recombination in which the half-site was aligned in the ‘reverse’ mode with the full-site (Fig. 4D). It is difficult to imagine how the half-site spacer could be held in a bent form within the recombination complex. These results demonstrate that at least one of the two substrates of recombination can be unbent and still be reactive, thus challenging a direct role of spacer bending in catalysis. The reaction between a labeled left half-site, and an unlabeled right half-site (see Table IA, B) is shown in Fig. 5A. The products were fractionated on a denaturing (urea) 20% polyacrylamide gel. One of the major labeled products of the reaction was a full-site (F) expected for strand cleavage and exchange at the left and right ends of the spacer (Fig. 5B). In addition, a small amount of pseudo full-sites, corresponding to strand transfer between two left half-sites, and significantly higher amounts of hairpins (HP; see Fig. S), corresponding to intramolecular strand transfer within a half-site, were also detectable on a nondenaturing gel (data not shown). The bands labeled C migrating below the substrate band L in Fig. 5A were the result of Flp cleavages that did not result in completion of strand exchange. The heterogeneity in C can be accounted for by the

Fig. 4. Half-site

vs. full-site recombination.

The Flp-mediated

nation reactions

contained

pH 7.8/10 mM MgCl,/lOO