Conformational isomerization of the holliday junction associated with a cruciform during branch migration in supercoiled plasmid DNA

J. Mol. Biol. (1988) 201, 1%30

Conformational Isomerization of the Holliday Junction Associated with a Cruciform During Branch Migration in Supercoiled Plasmid DNA Peter Dickie, A. Richard Morgan and Grant McFadden Department of Biochemistry University of Alberta Edmonton, Alberta Canada T6G 2H7 (Received 24 July 1987, and in revised form 7 December 1987) The variable positions of a branch-migrating cruciform junction in supercoiled plasmid DNA were mapped following cleavage of the DNA with bacteriophage T7 endonuclease I. T7 endonucleaseI specifically cleaved, and thereby resolved, the Holliday junction existing at the base of the cruciform in the circular bacterial plasmid pSAlB.56A. Cruciform extrusion of cloned sequences in pSAlB.56A (containing a 322 base-pair inverted repeat insert composed of poxvirus telomeric sequences) topologically relaxed the plasmid substrate in vitro. Thus, numerous crossover positions were identified within the region of cloned sequences, reflecting the range of superhelical densities in the native plasmid preparation. Endonuclease I-sensitive crossover positions, mapped to both strands of the viral insert following the T7 endonuclease I digestion of either plasmid preparations or individual topoisomers, were regularly separated by approximately ten nucleotides. The appearance of sensitive crossovers every ten nucleotides corresponds to a change in linking difference (ALk) of +2 in the circular core domain of the plasmid during branch point migration. In contrast, individual topoisomers of a plasmid preparation differ in linking number in increments of + 1. Thus, the observed linearization of each individual topoisomer following enzyme treatment, as a result of resolution of the crossovers associated with each topoisomer, showed that branch point migration to sensitive crossover positions must have occurred facilely. T7 endonucleaseI randomly resolved across either axis of the cruciform, though some discrimination (related to the sequencespecificity of the enzyme) was observed. The ten-nucleotide spacing between sensitive crossover positions is accounted for by an isomerization of the cruciform junction on branch point migration. An hypothesis is that this isomerization was imposed upon the cruciform junction by the change in helix twist (ATw) in the two branches that compose the topologically closed, circular domain of the plasmid. T7 endonuclease I may discriminate between the various isomeric forms and cleave a sensitive conformation that appears with every turn of branch migration which leads to the extrusion, or absorption, of two turns of helix from the circular core.

Several refinements have since been incorporated into the model describing more rigorously the possible orientations of the branches about the junction and their functional isomerization (Sigal & Alberts, 1972; Sobell, 1974; Meselson & Radding, 1975; Robinson & Seeman, 1987). Basically, three conformations for Holliday junctions have been proposed. In one version, postulated by Sigal & Alberts (1972), the branches are base-stacked and co-planar such that the two recombining molecules have unperturbed helix axes (referred to as the UHA structure by Robinson & Seeman (1987)). Another representation views it as a planar

1. Introduction analyses of Interpretive the molecular mechanisms of genetic recombination rely on the premises originally stated in the Holliday model (Holliday, 1964, 1968, 1974; Brooker & Lehman, 1971; Sigal & Alberts, 1972; Meselson & Radding, 1975; Dressler & Potter, 1982; Szostak et al., 1983; Hsu & Landy, 1984; Robinson & Seeman, 1987). Central to this model is the definition of a Holliday structure as the branched DNA complex formed by the exchange of single strands between two homologous duplexes (Holliday, 1964, 1968). 0022-2836/88/090019-12

$03.00/O

19

0

1988

Academic

Press

Limited

20

P. Dickie

structure, but not base-stacked, with the branches at right angles to each other such that the structure posseses pseudo4-fold symmetry (Sobell, 1974). Lastly, an intermediate conformation, wherein the branches of the junction are arranged in a tetrahedral fashion, has been postulated (Gough & Lilley, 1985). The only physical evidence discriminating between the various forms has come from electron microscopy (Dressler & Potter, 1982) and the electrophoretic behaviour of Holliday-like structures (Gough & Lilley, 1985). For given conformations of the junction, descriptions of the mechanics and dynamics of branch point migration have also been provided (Meselson, 1972; Warner et al., 1978; Thompson et al., 1986; Robinson & Seeman, 1987). With respect to the structural dynamics of Holliday junctions, actual experimental evidence is scarce, though there have been studies on the kinetics of branch point migration in topologically unconstrained branched structures (Warner et al., 1978; Thompson et al., 1986). The process of branch point migration is biologically significant in that it determines the extent of heterologous DNA formed during recombination, whereas an isomerization of Holliday junctions with stacked bases is postulated to determine the incidence of genetic “crossingover” during recombination (Meselson & Radding, 1975; Szostak et al., 1983). Mechanistically, both these processes involve the twisting or rotation of the helical DNA branches at the junction. Thus, both should be influenced by the topological nature of the DNA substrate. Circular bacterial plasmids, bearing large palin-

et al dromic sequences that can be extruded into stable cruciforms, are substrates in which branch migration can occur in response to torsional strain in the circular plasmid (Mizuuchi et al., 1982b). The plasmid pSAlB56A (Fig. 1) bears the telomeric sequences of Shope fibroma virus as a 322 base-pair imperfect inverted repeat (DeLange et al., 1986). Cruciform extrusion from the axis of the inverted repeat (Fig. I(c) and (d)) has been shown to relax the plasmid topologically (Dickie et al., 19873) and, in fact, the plasmid is isolated from bacteria topologically relaxed due to the extruded cruciform. Consequently, branch point migration of the cruciform crossover should occur in these molecules within the limits imposed by the superhelical energy in the topologically closed, circular core of the plasmid. Bacteriophage T7 endonuclease I (Center & Richardson,

1970; Sadowski,

1971) has been used to

probe for the cruciform crossover in pSAlB.56A (Dickie et al., 19876). It belongs to a class of recombination-specific a endonucleases with specificity for branched DNA forms. Members of this class also include Int protein (Hsu & Landy, 1984), endonucleases purified from yeast (West & Korner, 1985; Symington & Kolodner, 1985) and bacteriophage T4 endonuclease VII (Mizuuchi et al., 1982a; Lilley & Kemper, 1984; Kemper et al., 1984; Jensch 6 Kemper, 1986). The bacteriophage endonucleases have demonstrated an in-vitro specificity that makes them potentially useful as structural probes for branched DNA junctions. In fact, several

reports

have included

the conformation of Holliday

inferences

about

structures based on

(b)

(a)

A (c 1 b ‘. 3’ 5’

, , I.: .:.::‘:I

a ‘.

*.

;

.’

4’ I I ,:p&+? I E

‘.’ \

;

,=’

‘\

,’

‘\ \

.

(d) 6

J

Figure 1. The effect of cruciformation on the topology of piasmid pSAlB.56A. (a) and (b) The lineform representation of the inverted repeat insert of Shope fibroma virus telomere sequences is shown (by arrows). Vector sequences are shaded and all distances are given in base-pairs. (c) The cruciform configuration of the same sequences is illustrated, where a and b denote the 2 axes of the crossover junction. (b) and (d) Schematic representations illustrating the topology of the plasmid when bearing either configuration of the closed sequences. H, HindIII; E, EcoRI.

Isomerization

of Cruciform

21

Junctions

Unstackrd

5’

3’

Terminal

3’

5’

hairpins

Bare-stacked

Cleavage

Figure 2. Cruciform isomerization and T7 endonuclease I cleavage of cruciform crossovers. Potential cleavage sites for T7 endonuclease I are positioned at a cruciform crossover as it may exist in 2 hypothetical conformations (left side of reaction). Cleavage sites are paired (circles and arrowheads)to denote the 2 cleavageaxes and are specifically 5’ of the actual crossover.The top illustration is unstackedwith respectto the base-pairsat the crossover,and planar, whereas, in the lower representation the branches are stacked one on top of the other (the unperturbed-helix-axis structure (UHA) first described by Sigal & Alberta, 1972).The cleavage of cruciform DNA results in a linear molecule with nicked, terminal hairpins (right side of reaction, cleavedat the circled sites). Due to the imperfect nature of the viral inverted repeat in pSAlB.56A, thesehairpins will carry complementary “flip-flop” sequences.

the activity of the T4 and T7 enzymes (Kemper et al., 1984; de Messy et al., 1987; Dickie et al., 1987a). Bacteriophage T7 endonuclease I has a broad specificity for branched DNA structures (Panayotatos & Wells, 1981; de Massy et al., 1985, 1987; Dickie et al., 1987a). It is site-specific, cleaving predominantly one nucleotide to the 5’ side of the crossover, and shows a modest degree of preference for certain DNA sequences. The site preference has been defined for freely branched Holliday structures, but it is similar for fixed cruciform junctions that are structurally equivalent to Holliday crossovers. When endonuclease I was used to digest the plasmid, pSAlB56A, a linear plasmid molecule with hairpin termini was produced (Fig. 2). The termini produced are eq,uivalent to the telomere hairpins in the native virus (DeLange et al., 1986). Numerous cruciform crossover positions in the plasmid pSAlB56A were identified using T7 endonuclease I as a probe (Dickie et aZ., 1987b). The sites probed corresponded to cleavage across one chosen axis of the junction (axis a as represented in Fig. 1(c)), as mapped from restriction sites flanking either side of the inverted repeat sequences. Unexpectedly the cleavage positions were discovered to be periodically spaced, an average of ten nucleotides apart (the equivalent of roughly one turn of primary helix) along the DNA sequence. The periodicity was surprising since every

phosphodiester within the limited region of dyad symmetry of a synthetic Holliday structure capable of branch point migration was shown to be sensitive to cleavage by T7 endonuclease I (Dickie et al., 1987a).

The conformational isomerization of Holliday structures has not been reported to be linked, or coupled, to branch point migration. The isomerization of Holliday structures described by Meselson & Radding (1975), through which opposite pairs of strands are brought into the “crossed” configuration, has no apparent topological implications for a cruciform junction since the process would only involve the inversion of the cruciform structure relative to the circular domain of the molecule. Therefore, experiments were carried out to determine: (1) whether either axis could be cleaved at the “sensitive” crossover positions; (2) to verify that the cleavage pattern was exclusively related to the previously characterized cruciform structure in pSAlB.56A (Dickie et al., 1987b); and (3) if branch point migration in isolated plasmid topoisomers would lead to the generation of the same cleavage pattern. The results were found to be consistent with the interpretation that T7 endonuclease I discriminates between conformational isomers of the cruciform junction that arise as a result of a change in the helical twist of the plasmid circular core during branch migration. This presumptive

22 (a

P. Dickie et al. The viral sequences were excised from pSAlB.56A by digestion with Hind111 and EcoRI (Fig. 1). This “bridge” fragment was purified from low-melting point agarose and heat-denatured by boiling for 3 min in a small volume of TE buffer (10 mna-Tris.HCl (pH 8), 0.1 mM-EDTA). The DNA was immediately chilled on ice and the individual strands self-annealed to form “snapback” structures which are equivalent to the cruciform hairpins and the original viral telomere hairpins (DeLange et al., 1986).

)

(b)

(b)

/ * Migration

Figure 3. Plasmid topoisomer distribution before and after digestion with T7 endonuclease I. (a) Undigested pSAlB.56A, electrophoresed in 1% agarose in the presence of 2.1 pg chloroquine/ml, stained with ethidium bromide, photographed and scanned with a densitometer. Gel migration ww left to right. Open circles refers to nicked plasmid molecules. (b) Plasmid pSAlB.56A after a partial digestion with T7 endonucleaseI at 37°C. The linearized product of the reaction co-migrated with one of the topoisomers in the mixture under these electrophoresis conditions.

Enzymes

and

reaction

Restriction enzymes and phage T4 DNA polymerase were purchased from Bethesda Research Laboratories (BRL). Mung bean nuclease and phage T4 polynucleotide kinase were obtained from Pharmacia and calf intestinal phosphatase was obtained from Boehringer-Mannheim. Bacteriophage T7 endonuclease I was geneously donated by Dr P. Sadowski. It was greater than 95% pure with a reported concentration of 57 units/ml (P. Sadowski, personal communication). Digestions of plasmid DNA with T7 endonuclease I were performed as described (Dickie et al., 1987b) except that the reaction buffer contained 4 mnn-spermidine. In lo+1 reactions, 50 to 200 ng of DNA are digested with 1 ~1 of diluted enzyme (1 : 10 in 50 y. (v/v) glycerol and 1 mM-EDTA) at 37°C. The incubation times were adjusted to achieve 50 to 80% linearization of the plasmid (usually 5 to 15 min) as analyzed following electrophoresis of reaction products in agarose gels. Mung bean nuclease digestions were performed in T7 endonuclease I buffer at 37°C for 5 min. Reactions contained 200 ng of DNA and 12.5 units of enzyme in 20 ~1. The partial linearization of pSAlB.56A by cleavage with AvaII was performed in T7 endonuclease I buffer at 37°C for 10 min. All other enzyme reactions were performed according to the manufacturers specifications or standard procedures (Maniatis et al., 1982). Endlabeling of all DNA fragments was accomplished using [y32P ATP (5200 Ci/mmol) and T4 polynucleotide kinase or [@-3?P]dATP (3700 Ci/mmol), 2 mM each of dGTP, dTTP and dCTP and T4 DNA polymerase. Radiolabeled ATP and dATP was purchased from ICN Radiochemicals. (c) Gel electrophoresis

isomerization appears to be unrelated to the more familiar inversion isomerization process of Meselson & Radding

(1975).

2. Materials

and Methods

(a) Preparaticm

of DNA

The cruciform-containing plasmid substrate was pSAlB56A, which has been extensively characterized (DeLange et al., 1986; Dickie et al., 1987b). It is a pUC13 derivative bearing the telomere sequences of Shope fibroma virus in the form a 322 base-pair inverted repeat. The plasmid, with a stable cruciform extruded from the viral insert sequences was purified from low-melting point agarose (Langridge et al., 1980). The hairpins of the cruciform contain a total of 8 extrahelical bases as a consequence of the imperfect nature of the inverted repeat. The creation of 2 pSAlB.56A derivatives, pSAlB.LA and pSAlB.RA, has been described (Dickie et al., 1987b). These plasmids carry the left and right arms, respectively, of the viral inverted repeat (see Fig. l(a)).

conditions

conditions

Topoisomers of pSAIB.56A were purified from 20 cm x 25 cm, 1 y. (w/v) low-melting point agarose gels electrophoresed at room temperature in TBE buffer (90 mM-Tris-borate (pH 8-3), 2 mM-EDTA) containing 2.1 pg chloroquine diphosphate/ml (Sigma) for 72 h at 2.5 V/cm. The DNA bands were located by staining with ethidium bromide and were extracted according to published procedures (Langridge et aE., 1980). The analysis of the purified topoisomer DNA was carried out under similar conditions using standard agarose and electrophoresing for 24 h. End-labeled cleavage products were analyzed in standard 10% (w/v) polyacrylamide/‘lM-urea denaturing gels (Maxam & Gilbert, 1980) electrophoresed in TBE buffer. Polaroid negatives of agarose gels stained with ethidium bromide were scanned with a Joyce-Loebl Chromoscan 3 densitometer.

3. Results Purified pSAlB.56A co-migrates with nicked, open-circular plasmid DNA under standard electrophoresis

conditions

by virtue

of the extrusion

of a

Ismerimtion

of Cruciform

Junctions

23

cruciform that relaxes the plasmid topologically (Dickie et al., 19875). When the plasmid is electrophoresed in the presence of a titrated amount of the intercalating drug chloroquine, which partially unwinds the DNA causing the resorption of the cruciform structure, individual topoisomers can be separated on the basis of linking number. As shown in Figure 3(a), preparations of purified pSAlB.56A plasmid DNA were demonstrated to contain a normal distribution of topoisomer molecules using this procedure. Partial digestion of pSAlB.56A with T7 endonuclease I at 37”C, analyzed by electrophoresis in chloroquine-containing gels, demonstrated a reduction of every topoisomer band, and by inference, showed that each was equally sensitive to cleavage by T7 endonuclease I. The only visible product derived from the 11 discernable topoisomers was linearized plasmid (Fig. 3(b)). Cruciform crossover positions in the viral sequences of pSAlB.56A were determined by mapping T7 endonuclease I cleavage sites with respect to the EcoRI restriction site flanking the insert sequences (see Fig. 1). Plasmid DNA was consecutively digested with T7 endonuclease I and EcoRI, then end-labeled under conditions that minimized the labeling of internal nick sites. A similar experiment, mapping cleavage sites to one axis (axis a, Fig. l(c), mapped according to cleavage fragments 5’ end-labeled at the EcoRI site) has been reported (Dickie et al., 19876). Both 3’ and 5’ endlabeled cleavage fragments were analyzed so that resolution across either axis of the cruciform crossover could be identified. A comparison of the cleavage pattern along axis a and axis b (mapped according to the cleavage fragments that are 3’ endlabeled) is shown in Figure 4, lanes 2 and 1, respectively. Generally, three broad domains of large

-

311

-

249

-

200

*--

(i)

Symmrtry axis

[

(ii 1 [

(iii) I

(iv) [

cleavage (VI [

proximal

Figure 4. Mapping T7 endonucle- I cleavagesitesto the viral insert sequences.Purified pSAlB.56A was partially linearized with T7 endonuclease I, then digested to completion with EcoRI. Fragments were either 3’ or 5’-end-labeled at the EcoRI termini with 32P and electrophoresed in standard 10 o/o denaturing polyacrylamide gels. Lane 1, 3’-end-labeled fragments; lane 2, 5’.end-labeled fragments; lanes 3 and 4, 3’ and 5’-endlabeled fragments of Hi&-digested #Xl74 replicative form DNA, respectively. Fragment sizes are given in nucleotides. Three general domains of pSAlB56A cleavage fragment sizes are identified: I, II and III (for lanes 1 and 2). Within domain III are 5 sub-regions of

were

observed

within

the viral

sequences,

labeled I, II and III in Figure 4. Domains I and III are symmetrically related with respect to the inverted repeat axis along the same strand of the viral insert. Associated with domain I are cleavage positions distal to the EcoRI site, relative to the cruciform structure, whereas cleavage positions to the restriction

site are identified

by the

fragments of domain III. The two domains, within a given lane (Fig. 4, lanes 1 and 2), represent cleavage across both axes of the crossover. On the other hand, the cleavage sites identified in domain II are very near to the inverted repeat axis. Although domain II cleavages are symmetrically placed about the inverted repeat axis, band intensities

are far greater

on the proximal

side of

the axis. Cleavage in this region may be indirectly related to the presence of extrahelical bases in the cleavage fragments as marked (i) through (v). These regions arise from T7 endonuclease I cleavage at the base of the cruciform on the side proximal to the EcoRI site. The inverted repeat axis of the cloned viral sequences is 188 bases from the EcoRI site. Fragment sizes are given m .base-pairs.

24

P. Dickie et al. v 11 3’

CCATTATACCTCCTCAAAATYTCTGGAAAGAAIGGTGC

5’

GGTAATATGGAGGAGTTTTAGAGACCTTTCTT,C,CAfC ‘A4

75

+

65

Lll (iii)

-Axis

o

-Axis

b

55

y-y ( iv )

”

Figure 5. The correlation of cleavage positions to DNA sequence. Three cleavage regions ((iii), (iv) and (v)) were mapped, by the direct comparison of pSAlB.MA fragment lengths to known size markers (HinfI digests of 4x174 and pBR322). Cleavage along axis a is read off the top strand labeled at the 5’ end, whereas cleavage along axis b is read off the 3’-end-labeled lower strand. The numerals between the 2 strands give the distance, in base-pairs, from the labeled EcoRI terminus. The cleaved nhosnhodiesters are indicated with arrowheads, the size of which approximates the I L relative sensitivity of each site.

cruciform hairpins and will be discussed later. For the accurate identification of cruciform crossover positions, mapped by endonuclease I cleavage across one or the other axis, those sites contained in domain III were studied more closely. It was demonstrated that the second nick required for effective resolution along one axis could be located to the same sequence on the opposite strand by mapping relative to the Hind111 site on the other side of the insert (Dickie et al., 19875). Within domain III, five limited regions of T7 endonuclease I cleavage are defined ((i) through (v), Fig. 4), the centers of which are spaced, on average, ten nucleotides apart along the primary sequence. The relative intensity of cleavage associated with each region reflects a Gaussian distribution of fragment sizes, as might be expected for a native population of topoisomers, each with a unique cruciform crossover position when in its lowest free energy form (i.e. the ALk value in the circular core is near zero). As previously explained, a plasmid molecule with a native superhelical density (around -0.067) would have the base of its extruded cruciform within this general region of the viral insert sequence (Dickie et al., 1987b). Cleavage along either junctional axis has been mapped to the same crossover positions. T7 endonuclease I cleaves predominantly one nucieotide to the 5’ side of the crossover in synthetic Holliday junction analogs (Dickie et al., 1987a). Consequently, if cleavage across either axes occurred at the same crossover position, 3’ endlabeled fragments should be two nucleotides longer than the 5’ end-labeled fragments (corresponding to cleavage along axis b and axis a, respectively). This was generally observed for regions (iii), (iv) and (v), considering that some variability in cleavage efficiency might be expected because of the sequence preference displayed by the enzyme (Dickie et al., 1987a). In Figure 5, the cleavage positions have been aligned with the corresponding nucleotide sequence. The enzyme is known to prefer to cleave phosphodiesters 5’ to pyrimidine residues (Sadowski, 1971; Dickie et al., 1987a), particularly cytosine residues. This could explain the greater variability observed in the cutting at crossover region (v) , which resides in a polypyrimidinepolypurine stretch (Fig. 5). Most significantly,

however, within the interregion sequencesthere are numerous sites with preferred cleavage sequences that are not cleaved by T7 endonuclease I. For example, the sequence 5’-AC has been identified as a highly preferred cleavage position for the enzyme (Dickie et al., 1987a), and this sequence exists between regions (ii) and (iii), (iii) and (iv), and (iv) and (v). On the other hand, the sequence 5’-CA was poorly cleaved in synthetic Holliday structures capable of branch migration, and thus, the absence of cutting at this phosphodiester in region (iii) (top strand, Fig. 5) is consistent with this demonstrated enzyme specificity and substantiates the sequence alignment. A series of controls were performed to confirm that the observed endonuclease I cleavage sites did, in fact, correspond to variable crossover positions of a cruciform extruded from a common inverted repeat axis. Isolated hairpins, prepared by the heat denaturation and “snap-back” of excised inverted repeat insert fragments (see Materials and Methods), are not cleaved at all in domain III (Fig. 6, lane 5). Mung bean nuclease, used to probe for other non-B-DNA structures (Kowalski, 1986), only cleaved at the central axis of the cloned viral inverted repeat (Fig. 6, lane 1). These central axis sequences would have formed the single-stranded hairpins of the large cruciform branches at pSAlB.56A. Also, pSAlB.56A was partially linearized with AvaII (2 AvaII sites are positioned 222 base-pairs apart, far removed from the cloning site in pUCl3) and then immediately digested with T7 endonuclease I (Fig. 6 lane 4). Linearization destabilizes cruciform structures and would have allowed the molecules to branch-migrate into the more stable lineform configuration. Consistent with the resorption of the cruciform structure, less cleavage was observed at regions (i) through (iv) following a partial linearization of the plasmid. And finally, when clones bearing the individual arms of the viral inverted repeat (pSAlB.LA) and pSAlB.RA) were treated with T7 endonuclease I under similar conditions as digestions of pSAlB.56A, no cleavage was observed whatsoever (data not shown). Inspection of Figure 3(a) reveals that at least 11 topoisomers could be distinguished in the purified pSAlB.56A preparations. In their lowest free

Isomerization

of Cruciform

oc-DN

Junctions

25

IA *

-

-

-

46

-

Figure 6. Cleavage of pSAlB56A by T7 endonuclease I is related to the existence of the cruciform crossover. Plasmid pSAlB56A was partially digested with various endonucleases and then cleaved with EcoRI, end-labeled and electrophoresed as described in the legend to Fig. 4. Lane 1, pSAlB.56A digested with Mung bean nuclease; lane 2, pSAlB56A digested with A&I; lane 3, plasmid digested with T7 endonuclease I; lane 4, plasmid partially digested with A&I, followed by digestion with T7 endonuclease I. In lane 5, isolated cruciform hairpin structures (see Materials and Methods) were digested with T7 endonuclease I and then 5’-end-labeled. The size markers (S) were $-end-labeled Hinff fragments of 4x174 replicative form DNA.

Figure 7. Purification of pSAlB.56A topoisomers. Plasmid DNA was electrophoresed and individual topoisomers were purified from low melting point agarose gels containing 2.1 pg chloroquine/ml (see Materials and Methods). The purified topoisomers were analyzed following electrophoresis for 24 h in standard 1 o/o agarose gels containing the same concentration of intercalating drug, and staining with ethidium bromide. In lanes 1 through 8 are the individual purified topoisomers, ranked according to linking difference (Lk) with the most supercoiled species present in lane 8. oc-DNA is the nicked, circular form of the plasmid.

energy form, each should have a cruciform of unique size and crossover location. Nonetheless, as evident from the data in Figure 4, T7 endonuclease I only identifies five, perhaps six, possible crossover locations. In conjunction with the observed tennucleotide spacing between the T7 endonuclease I cleavage regions, this suggested that every other topoisomer was not cleaved when in its lowest free energy state. Instead, since every topoisomer has been shown to be cleavable (Fig. 3), the cruciforms must have branch-migrated to sensitive crossover positions and there been cleaved. To assessthe degree of branch migration in these molecules, individual topoisomers were purified (Fig. 7) and digested with T7 endonuclease I (Fig. 8). Each topoisomer, after digestion, produced a subset of the cleavage fragments previously identified in the plasmid population as a whole. As expected, the greater the ALk value, the further the average crossover position was from the inverted repeat axis. Most importantly, the crossover in each topoisomer branch-migrated a distance spanning a number of T7 endonuclease I-sensitive and insensitive regions.

The less efficient

cleavage

in region

(ii),

relative to the cleavage at regions (i) and (iii), as demonstrated with the purified topoisomers, was probably not due to a lesserfrequency of crossovers in this

region,

but

rather

was

attributable

to the

sequence preference of T7 endonuclease I. Though T7 endonuclease I prefers certain sequences, nonpreferred cleavage sequences can nevertheless still be cleaved when positioned 5’ to a fixed (nonmigrating) Holliday structure (Dickie et al., 1987~~).

26

P. Dickie et al. Cleavage very near the inverted repeat axis (in domain II, Fig. 4), was also observed in each of the purified topoisomers (Fig. 8). The efficiency and position of cleavage in domain II appeared to be independent of linking difference. Moreover, it was not significantly reduced when the plasmid was linearized (Fig. 6, lane 4). The cleavage pattern is provocative because the cut sites are phased approximately ten nucleotides apart beginning from the inverted repeat axis. Their appearance was dependent upon the presence of the inverted repeat and they are absent in digestions of the cruciform hairpins (Fig. 6). Their presence is consistent with the extrusion of a small cruciform from the same inverted repeat axis as the major cruciform under study. Consequently, it is possible that small cruciforms have become trapped in this region because the presence of the extrahelical basesin the cruciform hairpins would retard branch migration during the cruciform-to-lineform transition (Robinson & Seeman, 1987).

I[ II I

4. Discussion

(i) [ (ii)

[

(iii) C

(iv)

[

Due to the large size of the inverted repeat (322 base-pairs) in pSAlB.56A, cruciformation leads to the complete relaxation of any topoisomer within the physiological range of around 20 supercoils. Certain cruciform crossover positions might be expected to be favored, for instance those associated with completely relaxed topoisomers. Extrusion of one more turn of helix is anticipated every time the linking number of a topoisomer is

decreased by one. The extrusion of one turn of helix results from branch point migration of the crossover along half a turn of helix, since two half turns in the circular domain feed the cruciform branches from either side of the junction. Therefore, a simple model might have predicted that maximum cleavage be spaced five base-pairs apart, in association with the 11 topoisomers seen in Figure 3. However, mapping crossover positions susceptible to T7 endonuclease I resulted in an observed periodicity of around ten base-pairs. equivalent to the extrusion (or adsorption) of two turns

Figure 8. T7 endonuclease I digestion of purified pSAlB.56A topoisomers. The digestion of plasmid DNA and the analysis of 5’-end-labeled cleavage fragments was performed as described in the legend to Fig. 4. The lane headings relate directly to the numbering of the purified topoisomers in Fig. 7. For example, in lane 7 are the cleavage fragments generated by the digestion of the next to most supercoiled topoisomer purified. Domains and regions of cleavage are the same as described in the legend to Fig. 4.

of helix,

i.e. branch

point

migration

along

one

full turn of helix (Figs 5 and 6 of Dickie et al. (1987b), which shows six cleavage sites separated by around 10 base-pairs). The hypothesis that only alternate topoisomers are cleaved can be dismissed since, as shown in Figure 3, all 11 topoisomers appear equally sensitive to T7 endonuclease I. However, the energy barrier for branch point migration in relaxed DNA is quite low (many topoisomers are obtained on ligating a nicked DNA), and therefore thermally induced branch point migration in any one topoisomer will traverse positions at the energy minima for crossover topoisomers differing by a few linking numbers. Indeed this is exactly what is observed (see Fig. 8) show cleavage sites where individual topoisomers over a distance of > 30 base-pairs. Other topoisomers show exactly the same cleavage sites.

Isomerization

of Cruciform

but the intensity of cleavage increases towards the ends of the palindromic sequence as their linking numbers decrease. This is to be expected as, on average, the hairpin arms of the cruciform would be further extruded. The regular ten-base-pair spacing between cleavage sites can be concluded to be independent of topoisomer linking number. Particularly important is the finding that sequences that are very susceptible to cleavage, such as 5’ AC (Dickie et al., 1987a), occur often in either strand between the cleavage sites, and are completely refractory to cleavage. 5’ AC sequences were preferentially cleaved in synthetic cruciforms having limited branch migration but also having no topological constraints. (a) Branch migration in closedcircular plasmid DNA and isomerization processes Models for double-stranded branch migration in Holliday structures (Meselson, 1972; Meselson & Radding, 1975) propose that rotary diffusion drives

lsomerizolion

Junctions

27

the migration of a crossover, in which the basesare all assumed to be stacked (Fig. 2) (Sigal & Alberts, 1972). More recently Robinson & Seeman (1987) have called this particular Holliday structure the unperturbed helix axis (UHA) form. The UHA form can have either of two pairs of phosphodiesters crossing over one another. As discussedby Meselson & Radding (1975) the two conformational isomers can be readily interconverted by unstacking of the basesat the crossover, rotation of the arms of DNA, with subsequent restacking of different bases, such that the pair of phosphodiesters that were formerly “outside” now become the crossing-over phosphodiesters. This is the most commonly understood isomerization process when discussing recombination. However, there is another type of conformational isomerization which is discussedby Robinson & Seeman (1987). It is best illustrated in Figure 9 where the crossover is shown in a pseudotetrahedral conformation with unstacked bases at the junction, i.e. it approximates the MeselsonRadding intermediate in isomerization. (For

through

branch

ATW

w

Figure 9. Isomerization in terms of the disposition of the helical grooves of the DNA branchesadjacent to the cruciform crossover.Two unique isomerizationevents are schematically represented using the stick figure of a Holliday junction (Holliday, 1964). The helical turns have been removed for clarity and the branches of the junction are shown iu the tetrahedrally arranged conformation. Proceeding from top to bottom, junctions undergo an inversion isomerization by passing through a planar form. Left to right in the Figure are pairs of isomers that are related by a change in twist (half twist/branch) in the junctional branches. Note that in this case there is a change in the disposition of the grooves relative to the polarity of the DNA strands. In this diagram, open branches marked with the 5’/3’ polarities denote the circular core branches. Structures (a) and (b) are “enantiomorphic” forms of the junction. M, major grooves; m, minor grooves. In the unwound representations shown, the major and minor grooves are on opposite faces of the duplex DNA tracts.

28

P. Dickie

heuristic studies the Blackwell Molecular Models are very useful.) The conformation shown has the two minor grooves (m) facing one another in the upper half, and the two major grooves (M) facing one another in the lower half. As branch migration proceeds, the helices must rotate and an exchange of strands takes place. After one half turn of helix, structure (b) will be obtained, regardless of the direction of branch migration. It is evidently a conformational isomer different from the initial one (neglecting the asymmetry of the arms). If looked at from below, so that the minor grooves again face one another, the polarities of the strands have inverted. The important point is that branch point migration must proceed through another half turn of helix to restore the initial conformation. This immediately suggests why T7 endonuclease I cuts only approximately every tenth nucleotide. We cannot say which conformation the enzyme recognizes but it will only appear once in every turn of the helices during branch point migration with strand exchange. To clarify the relationship of the structures shown in Figure 9 to other proposed structures for cruciforms, the “junctional” structure is shown flattened into a plane representing a cruciform structure with tetrad symmetry of the backbone as proposed by Sobell (1974). It is immediately obvious that the minor grooves are on one side and the major grooves are on the other side if helicity is removed from the DNA, i.e. the two sides are not equivalent. Lastly, structure (a) is essentially an inversion of the original structure passing through the planar Sobell structure. Branch migration has not occurred. The original structure and (a), if base-stacked to give a UHA cross-over, would correspond to the conformational isomers of Meselson & Radding (1975). In Figure 10 a more realistic picture of the relationship of cruciformation to branch migration is shown, taking into account the DNA helicity (seethe legend to Fig. 10 for a full description). Significantly, structures (a) and (b) are pseudo-enantiomorphic. The change in twist (AT;) in the circular DNA with cruciformation must be accompanied by a change of writhe (supercoiling). Thus, for any relaxed topoisomer with a cruciform, which branch migrates thermally around the minimum energy conformation, compensating positive and negative writhes will appear such that Wr = -AT,,,. It has been shown for low writhing (supe’rhelical) densities that there is essentially no change in T, (Pulleyblack et al., 1975). Therefore the analysis should not be complicated by the effect of writhing

on twist. It does therefore suggest that were the T7 endonuclease I very specific for the conformation it recognizes, the spacing of cleavage sites will give some indication of polymorphism in the local DNA structure, i.e. the number of base-pairs/turn of helix. In the present case this seemsunlikely since the enzyme has some specificity for sequence. There is usually more than one cut for every cleavage site. This could be because the enzyme can induce small

et al.

1

1

Unwind

Strand

exchange

FQ J ) a5’

3’

I

2

Rewind

L

s

Inversion

s

5’ 3’

2

3’

5l

x

Figure 10. Branch migration and isomerization of a cruciform crossover. A plasmid molecule with 1 helical turn extruded into a cruciform structure (top figure) branch migrates so that a 2nd turn is extruded from the circular core. The left and right branches of the junction are enclosed to form the circular core of the plasmid. The 2 bottom figures are inversion isomers of the more extruded molecule. Individual strands are distinguished as open and filled ribbons. Three partial helices are numbered, -1., -2- and -3- to highlight the net movement of the partial turns during extrusion. As shown, turn -2- is extruded into a hairpin by becoming unwound, undergoing strand exchange and finally being rewound to complete the process. Note that the disposition of the helical grooves has been reversed at the cruciform junction. This is the critical difference between the twist-related isomerization and inversion isomerization (bottom reaction).

changes in T, within the thermal range e.g. w l/10 of a twist due to tighter binding to a preferred sequence. In contrast, when synthetic Holliday structures were constructed with a central core that could branch migrate (Dickie et al., 1987a), all positions were cleaved, as expected for the extra degree of rotational freedom. In order to test these ideas out more fully, and in particular to test our working hypothesis for cleavages with approximately ten-base-pair spacings, plasmids with inverted repeats, related to pSAlB.56A by small

Isomerization

of Cruciform

deletions near the central axis, will be constructed. It should be possible accurately to predict the cleavage sites. Also it would be of interest to test other enzymes that resolve Holliday structures; two others having been characterized from phage T4 (Kemper et al., 1984) and from yeast (Evans & Kolodner, personal communication). For example the precise conformational isomer recognized by these latter enzymes may be different as well as their sequence preferences. These properties could be tested out with our model system. One further point of interest that deserves study is that the calculations of Robinson & Seeman (1987) indicate that branch migration past a mismatched base-pair will be blocked. In Figures 4 and 8 three well-defined and regularly spaced cleavage sites are observed in domain II, close to the symmetry axis and about ten base-pairs apart. Several extrahelical bases are expected in this region (Dickie et aE., 1987b) and some molecules may be trapped from further branch migration. This is perhaps surprising since pSAIB56A spontaneously forms cruciforms that are relaxed on isolation and there is no evidence from previous studies (Dickie et al., 1987b) of partly relaxed molecules. Finally can we deduce anything about the structure of a Holliday junction or how T7

(a)

Junctions

29

endonuclease I recognizes the junction? Robinson & Seeman (1987) suggest the planar conformation of Sobell as being the energetically most plausible using electrostatic calculations, but the theory neglects base-stacking. Gough & Lilley (1985) have provided evidence that a small cruciform at the center of an otherwise linear DNA gives it a bend. This conclusion is obtained from anomalous gel electrophoretic mobility. This is perhaps most consistent with a “tetrahedral” crossover, but the gel conditions are different from those used in the T7 endonuclease I assay, where both Mgzf and spermidine are present. Robinson & Seeman (1987) suggest that perhaps for branch point migration the Holliday junction assumes the UHA form, although it is a high energy form, since otherwise branch migration would be opposed by the viscous drag of rotating DNA arms. This suggestion is made to rationalize the much slower branch point migration than would be expected by Meselson’s (1972) calculations for the UHA form. Enzymatic studies are more consistent with a structure closer to the Sobell (1974) structure. There are two lines of reasoning supporting this notion, though it must be emphasized that this may be a conformation imposed by the enzyme. Firstly the enzyme cleaves across the two axes a and b of Figure 1, equally within experimental accuracy. Secondly it has been

(D)

Figure 11. A hypothetical cruciform junction and the proposed binding of T7 endonuclease I. (a) The branches of the cruciform structure are shown to be arranged in a near-planar form. The circular core region of the molecule is represented by the single dotted line. The cruciform hairpins are only partially drawn and extend to the right and left of the crossover. One hairpin is identified by a “heavy” strand (to the right) while the other is “light’‘-stranded (to the left). The pairs of cleavage sites (representing the 2 cleavage axes) are denoted by the open circles and the filled triangles. In this representation, the cleavage sites can be observed to be on the crossing-over strands of the component branches of the junction. (b) and (c) The binding of 1 enzyme molecule (b) and sequentially a 2nd enzyme molecule (c) at the crossover is illustrated. The sequential binding of catalytic units along 1 cleavage axis was proposed primarily on the basis of earlier results (Dickie et al., 1987a), which defined the binding domain of the enzyme to include regions of 2 duplex branches and, most importantly, the ‘1 strand which tethers the branches at the junction.

P. Dickie

30

shown (Dickie et al., 1987a) that a trioligonucleotide complex containing two duplex branches and two single-stranded branches is cleaved on the bridging strand one nucleotide 5’ to the crossover in the duplex region. The enzyme therefore appears to recognize just two duplex arms. Figure 11 is included to give a rough indication of how the enzyme might work through two catalytic units (see the legend to Fig. 11). In summary these results are relevant to recombination in viwo, since almost all DNA in viva is topologically restricted even if not circular. The site of resolution of the Holliday junction will be restricted both by conformational isomers and some sequence specificity. This may lead to hot spots of recombination, especially since the binding of histones may further restrict the number of possible sites. The authors are indebted to Dr Paul Sadowski for providing them with bacteriophage T? endonuclease I and to Dr Wayne Anderson for his criticisms of the manuscript. They are also grateful for the technical assistance of Adrian Wills. This work was supported by operating funds from the Medical Research Council of Canada and by salary support (to P.D. and G.M.) from the Alberta Heritage Foundation for Medical Research.

References Brooker, T. R. & Lehman, I. R. (1971). J. Mol. Biol. 60, 131-149. Center, M. & Richardson, C. (1970). J. Biol. Chem. 245, 62856291. DeLange, A. M., Reddy, M., Scraba, D., Upton, C. & McFadden, G. (1986). J. Viral. 59, 249259. de Massy, B., Studier, F. W., Dorgai, L., Appelbaum, E. & Weisberg, R. A. (1985). Cold Spring Harbor Symp. Quant. Biol. 49, 715726. de Massy, B., Weisberg, R. A. & Studier, F. W. (1987). J. Mol. Biol. 193, 359376. Dickie, P., McFadden, G. t Morgan, A. R. (1987a). J. Biol. Chem. 262, 1482614836. Dickie, P., Morgan, A. R. & McFadden, G. (19876). J. Mol. Biol. l%, 541-558. Dressler, E. & Potter, H. A. (1982). Annu. Rev. Biochem. 51, 727-761. Edited

et al. Gough, G. W. & Lilley, D. M. J. (1985). Nature (London), 313, 14-156. Holliday, R. (1964). Genet. Res. 5, 282-304. Holliday, R. (1968). In Replication and Recombination of Genetic Material (Peacock, W. J. & Brook, K. D., eds), pp. 157-174, Australian Acad. Sci., Canberra. Holliday, R. (1974). Genetics, 78, 273-287. Hsu, P. L. & Landy, A. (1984). Nature (London), 311, 721-726. Jensch, F. & Kemper, B. (1986). EMBO J. 5, 181-189. Kemper, B., Jensch, F., v. Depka-Prondzynski, M., Fritz, H.-J., Borgmeyer, U. & Mizuuchi, K. (1984). Cold Spring Harbor Symp. Quant. Biol. 49, 815-825. Kowalski, D. (1986). Nucl. Acids Res. 14, 8949-8962. Langridge, J., Langridge, P. & Bergquist, P. L. (1980). Anal. Biochem. 103, 264271. Lilley, D. M. J. & Kemper, B. (1984). Cell, 36, 413422. Maniatis, T., Fritsch, E. F. & Sambrook, J. (1982). Molecular Cloning: A Laboratory Manual, Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY. Maxam, A. M. & Gilbert, W. (1980). Methods Enzymol. 65, 499-560. Meselson, M. (1972). J. Mol. Biol. 71, 795-798. Meselson, M. S. & Radding, C. M. (1975). Proc. ,Nat. Acad. Ski., [J.S.A. 72, 358-361. Mizuuchi, K., Kemper, B., Hays, J. & Weisberg, R. (1982a). Cell, 29, 357-365. Mizuuchi, K., Mizuuchi, M. 8: Gellert, M. (19826). ,J. Mol. Biol. 156, 229-243. Panayotatos, N. & Wells, R. D. (1981). Nature (London), 289, 466470. Pulleyblank, D. E., Shure, M., Tang, I)., Vinograd, J. & Vosberg, H.-P. (1975). Proc. Nat. Amd. Sci., U.S.A. 72, 42894284. Robinson, B. H. & Seeman, N. (1987). Biophys. J. 51, 611626. Sadowski, P. (1971). J. Biol. Chem. 244, 209-216. Sigal, N. & Alberts, B. (1972). J. MOE. BioE. 71, 789-793. Sobell. H. M. (1974). In Molecular Mechanisms in Genetic Recombination (Grell, R. F., ed), pp. 433438, Plenum Publishing Co., New York. Symington, L. S. & Kolodner, R. (1985). Proc. Nat. Acad. Sci., I7.S.A. 82, 7247-7251. Szostak, J. W., &r-Weaver, T. L. & Rothstein. R. J. (1983). Cell, 33, 25-35. Thompson, B. J., Camien, M. N. & Warner, R. C. (1986). Proc. Nat. Acod. Sci., U.S.A. 73, 22992303. Warner, R. C., Fishel, R. & Wheeler, F. (1978). Cold Spring Harbor Symp. Quant. Biol. 43, 957-968. West, S. C. & Korner, A. (1985). PTOC. Nat. Acad. Ski., U.S.A. 82, 644.56449.

by R. Laskey