Three-stranded paranemic joints: Architecture, topological constraints and movement

J. Mol. Biol. (1990) 215,623-634

Three-stranded Paranemic Joints: Architecture, Topological Constraints and Movement Carl Bortner and Jack GriffithT Lineberger Cancer Research Center and Department of Microbiology and Immunology University of North Carolina Chapel Hill, NC 27514, U.S.A. (Received 27 February 1990; accepted 18 June 1990) The RecA and SSB proteins will catalyze the joining of two DNA molecules containing homologous sequences but lacking homologous ends in a reaction termed paranemic joining. The absence of homologous ends can be achieved by (1) pairing two circular DNAs or (2) using linear DNA(s) with ends lacking homology to the pairing partner. Here we have used electron microscopy (EM) to examine such pairings. Circular M13 single-stranded (ss) I)NA enveloped by RecA protein into a presynaptic filament was paired with linear Ml3mp7 double-stranded (ds) DNA containing non-M13 sequences at its ends. Joint complexes were frequently seen in which the dsDNA was joined with the presynaptic filament over several kilobase (10 a bases) lengths of the dsDNA. In this region, the presynaptic filament appeared disorganized as contrasted to the customary helical structure of the filament containing only a single strand of DNA. The same ultrastructure, but with greater detail, was observed when the samples were prepared for EM without fixation using a new method of fast-fl.eezing and freeze-drying. EM immunogold staining demonstrated the presence of SSB protein in the disorganized region containing all three strands, but not in the regular helically arranged region. Psoralen photo-crosslinking of the DNA in the joint complexes revealed that the three DNA strands were in close proximity only over a single short (200 to 300 base-pairs) region. The joining of nicked circular Ml3 dsDNA and presynaptic filaments containing circular M13 ssDNA resulted in the intertwining of the dsDNA about the circular presynaptic filament. The joints produced in this case were short, as was the single region of psoralen photo-crosslinking of the three DNA strands. A model of how these long three-stranded joints form is presented involving the movement of a short "true" paranemic joint along the presynaptic filament.

strands and dissociation of the products. What is known about the second stage is that metastable three-stranded joints, termed paranemic joints, appear to be precursors of the more stable and productive plectonemic joints (for a review, see Griffith & Harris, 1988).

1. I n t r o d u c t i o n

One of the least understood aspects of general recombination involves the step in which two DNA molecules pair at a region of homology. Our knowledge of molecular mechanisms of general recombination derives largely from studies of the RecA protein of Escherichia coli, and more recently, the UvsX protein of T4 phage. In vitro, these proteins will catalyze strand exchange reactions that can be separated into three distinct stages: first, formation of presynaptic filaments in which the protein assembles onto single-stranded (ss:~) DNA; second, the search ibr homology, followed by synapsis; and third, with appropriate templates, a net exchange of

(a) Architecture of paranemic joints A typical plectonemic joint is the D-loop in which an end of a ssDNA invades and displaces one strand of an homologous dsDNA, forming stable WatsonCrick base-pairs with its complementary strand. The net interwinding of strands from the two molecules creates a joint that is stable to deproteinization (McEntee et al., 1979; Shibata et al., 1979). Paranemic joints were first identified as ReeA protein-catalyzed joint complexes that formed between ss and dsDNAs sharing homologies but lacking homologous DNA ends. These associations were

Author to whom all correspondence should be addressed. Abbreviations used: ~I)NA. single-stranded DNA; dsDNA, double-stranded DNA; EM, electron microscopy; bp, base-pair(s); kb, l0 s bases. 0022-2836/90/200623-12 $03.00/0

623

O 1990 Academic Press Limited

C. Bortner and J. Gri~th

624

" (a)

r

(b)

Figure 1. Schematic illustration of the intertwining of dsDNA about a presynaptic filament resulting from the presence or absence of free dsDNA ends. (a) A linear dsDNA (here lacking homologous ends) pairs with a circular ssDNA contained within the RecA protein coating. If the formation or movement of the joint (bar) induces a rotation of the dsDNA and presynaptic filament about each other, then the free ends of the dsDNA will allow it to wrap freely about the circular presynaptic filament. However, in (b) if the dsDNA is circular (and in this study, nicked), then the rotation of the dsDNA and presynaptic filament about each other will produce a counterwrapping of the 2 DNAs.

found to be unstable to deproteinization (DasGupta et al., 1980; Bianchi et al., 1983). This was apparently due to the absence of homologous ends whose presence are required for the formation of a D-loop (Cunningham et al., 1981; Bianchi et al., 1983). The first direct visualization of paranemic joints utilized supertwisted M13 dsDNA and linear M13mp7 ssDNA containing non-M13 sequences at its ends assembled into presynaptic filaments (Christiansen & Griffith, 1986). Joints were observed by EM in which the protein-free dsDNA entered and exited the linear presynaptic filament over a short distance without visibly altering the structure of the filament within the region of the joint (Christiansen & Griffith, 1986). The length of the joints equalled the length of the dsDNA which, when fully melted, just relieved the supertwisting of the dsDNA; with supertwisted M13mp7 dsDNA, the joints measured 350 to 400 bp. When these same DNAs were joined paranemically by the T4 UvsX and gene 32 proteins, joints were found by EM which were even shorter in length while the dsDNA remained partially supertwisted (Harris & Griffith, 1987). When RecA protein catalyzed the paranemic joining of circular MI3 ssDNA and circular super-

twisted M13 dsDNA, joints having the same length and morphology as those formed with the linear M13mp7 ssDNA were observed. However, here, there was some indication of an intertwining of the circular dsDNA and the circular presynaptic filament (Christiansen & Griffith, 1986). (b) Topological requirements for paranemic joining The elimination of free homologous ends can be satisfied by (1) pairing two circular DNAs, or (2) by pairing linear DNA(s) but blocking free DNA ends with non-homologous sequences. Although both template pairs satisfy the biochemical criteria for paranemic joining described by Riddles & Lehman (1985), they differ topologically, since the presence of a free DNA end could allow the linear DNA to freely wrap about its partner as illustrated in Figure l(a). Were paranemic joining to involve a coiling of the dsDNA and presynaptic filament about each other, then in the pairing of two circular DNAs, a highly intertwined complex might result (Fig. l(b)). Honigberg & Radding (1988) presented an analysis of the rotational requirements for the plectonemic exchange of strands between a linear dsDNA

Three-stranded Paranemic Joints and a circular or linear presynaptic filament. T h e y concluded t h a t the only required rotation was t h a t of the d s D N A a b o u t its own axis, driven presumably b y the unwinding of the d s D N A required for the separation of its two strands. No parallel s t u d y of the rotation of the pairing D N A s during paranemic joining has been presented.

625

Ml3mp7 dsDNA was linearized with either EcoRI (a gift from Dr Paul Modrich, Duke University) or BamHI (Bethesda Research Laboratories). Nicked M13 dsDNA was obtained as a preparation of initially supertwisted M13 dsDNA that had become >75~o nicked upon storage. (b) Joint formation

(c) Growth and movement of paranemic joints The observation t h a t paranemic joints form rapidly a n d can be converted to plectonemic joints implicates t h e m as precursors to plectonemic joints (Bianchi et al., 1983; Riddles & L e h m a n , 1985; Harris & Griffith, 1988). P a r a n e m i c joints might either form and dissociate, or, once formed a paranemic joint might move along the presynaptic filam e n t until a free homologous end was reached and a plectonemic joint initiated. Evidence has y e t to be presented supporting either possibility. Our earlier studies (Christiansen & Griffith, 1986; Harris & Griffith, 1987) employed circular supertwisted dsDNA. The superhelical strain imposed a strict limit on the length of the joints and m a y have imposed other limitations, such as the ability of the joints to dissociate once formed or to move along the p r e s y n a p t i c filament. In this study, we examine the RecA proteincatalyzed p a r a n e m i c joining of two D N A pairs t h a t lack the constraint of supertwisting and which differ only in the presence or absence of a free d s D N A end as illustrated in Figure 1. We describe the architecture of these joints, and conclude t h a t ' w h e n linear d s D N A w a s used, v e r y long three-stranded joints formed while the pairing t h a t employed two circular DNAs produced only short joints. In both pairings, the s e g m e n t in which the three strands were in close p r o x i m i t y was localized to a single short region. In this work an EM p r e p a r a t i v e method (Heuser, 1983; Heuser & Griffith, 1989) involving fast-freezing and freeze-drying samples t h a t avoids the steps of chemical fixation and air-drying was e m p l o y e d together with the conventional methods. There was no difference in the conclusions t h a t could be drawn from the t w o m e t h o d s of preparation, b u t much greater detail and preservation of the helical structure of the p r e s y n a p t i c filament was clearly a p p a r e n t when fixation and air-drying were avoided. We present a model in which the d s l ) N A rotates a b o u t the p r e s y n a p t i c filament during paranemic joining; when this rotation is allowed by the presence of a free d s D N A end, the joints are free to move along the p r e s y n a p t i c filament.

2. Materials and Methods (a) DNA and protei~' MI3 ss- (3H-labeled) and dsDNAs (Register & Griffith, 1986), RecA (Griffith & Shores, 1985) and SSB (Chase el al., 1980) proteins were purified as described. Rabbit antiSSB antibody was a gift from l)r Jack Chase (U.S. Biochemical Corp.). Gold-conjugated goat anti-rabbit antibody was purchased from Jansen Life Sciences.

Ml3 ssDNA was diluted to 2 #g/ml (4/~M) in HE buffer (20 mM-Hepes, 0-1 mM-EDTA (pH 7"5)), heated to 65°C for 5 rain to disperse the DNA, cooled to 37°C, and then magnesium acetate and ATP added to 12 mM and 3 raM, respectively. A protein mixture was prepared containing 60 mM-sodium acetate, 12 ram-magnesium acetate, 3 mMATP, 8#g creatine phosphokinase/ml, 40 mM-creatine phosphate, and 160 to 240/lg RecA protein]ml (4 to 6/~M) in HE buffer. The protein mixture was incubated at 37°C for 5 rain then combined with an equal volume of ssDNA yielding a final mixture containing l #g ssDNA/ml (2/~M), and 80 to 120#g RecA protein/ml (2 to 3/gM). After l0 min at 37°C, SSB protein was added to a final concentration of 3/~g/ml (0"20/ZM) to complete the formation of the presynaptic filaments. Linear or nicked circular dsDNA had, in parallel, been diluted to between 7 and 14/~g/ml (10 to 20/~M) in HE buffer and incubated at 55°C for 5 min, then mixed with magnesium acetate and ATP to final concentrations of 12 mM and 3 mM, respectively. The dsDNA was then added to the presynaptic filaments and the full mixture incubated at 37°C for the times indicated in the text. (c) EM immunogold labeling Presynaptic filaments were formed as described above in a 200/d volume but at a 10-fold higher concentration of ssDNA, RecA and SSB proteins. Mi3mp7 dsDNA was added (at a 10-fold higher concentration) and the reaction incubated at 37°C for 20 to 40 rain. All subsequent steps were carried out at room temperature. The sample was fixed with 0"3~/o (v/v) glutaraldehyde for 5 rain and chromatographed over 2 ml of Sepharose 4B equilibrated with HE buffer. Fractions containing the complexes were adjusted to a DNA concentration of 4/~g/ml (based on the all-labeled ssDNA) in a 100]~l volume and NaCI was added to 300 mM final concentration to inhibit nonspecific antibody binding. Rabbit anti-SSB protein antibody (6 #l of 50:1 dilution of the stock antibody in HE buffer) was added for 10rain, followed by chromatography as described above. To 70#1 of the fraction containing the highest concentration of DNA, 5/zl of 5nM-gold-conjugated goat anti-rabbit antibody was added for 30 min. The sample was fixed and chromatographed as described above. (d) Psoralen photo-crosslinking Paranemic complexes employing either M13mp7 linear or nicked M13 dsDNA, were prepared as described above (at a l x concentration}, with a 20 min incubation at 37 °C. Psoralen (hydroxy-methyl trioxalen) (a gift from Dr Aziz Sancar at this University) was added to a final concentration of 3/~g/ml and the sample irradiated with ultraviolet light from a standard germicidal lamp for 20 min at a distance of 20 cm. Sodium dodecyl sulfate was added to 1 ~/o (w/v) and the sample incubated for 5 min at 37 °C. The DNA was ehromatographed over Sepharose 4B to remove the proteins and detergent and prepared for EM by surface spreading on a denatured protein film in

C. Bortner and J. Gri~ith

626

40% (v/v) formamide as described by Chow & Broker (1981). Linear M13mp7 dsDNA was treated with psoraten as described above then exposed to 9% (v/v) glyoxal (from a 40% aqueous stock) for 60 min at 65°C to disrupt hydrogen bonding and thus determine the extent of crosslinking. The DNA was prepared for EM using 40 % formamide as described above.

(e) Electron microscopy Methods for preparing paranemic complexes for EM following fixation and chromatography have been described (Griffith & Christiansen, 1978). Briefly, samples were absorbed to.a thin carbon film in the presence of 2 mM-spermidine, dehydrated through a series of graded ethanol washes, air-dried and rotary shadowed with tungsten. The preparation of samples by rapid-freezing and freeze-drying was carried out by forming paranemic complexes as described above but without fixation. The sample was then quickly filtered through 1 ml of Sephaeryl S-500 equilibrated with 12mM-magnesium acetate, 30 mM-sodium acetate, 3 mM-ATP, and 20 mMHepes (pH 7-5) to separate the complexes from the excess of free RecA protein and ATP-regenerating enzymes. The complexes were immediately adsorbed for 5 s onto a thin carbon film supported by a copper mesh grid. The excess liquid was blotted away and the sample plunged into liquid ethane chilled with liquid nitrogen. The frozen sample was transferred to a Wiltek-modified Balzers 300 freeze-etch machine and freeze-drying was carried out for 2 h at -85°C and 1 h at -50°C. Samples were then rotary shadowcast with tantalum at -170°C and at l0 -7 Torr (l Torr= 133"322 Pa). Micrographs were taken on a Philips EM400 TLG, and length measurements were made using a Summagraphics digitizer coupled to an IBM PC-AT computer.

3. Results (a) The paranemic joining of circular M13 ssDNA

and linear M13mp7 dsDNA produces very long three-stranded structures Cleavage of M13mp7 dsDNA at the center of the 840 bp lac insert produces a linear M13 dsDNA with ~400 bp of non-M13 sequences at its ends. When this DNA is paired with circular M13 ssDNA, plectonemic joining is excluded. Here, RecA protein was first assembled onto circular M13 ssDNA in the presence of SSB protein to produce presynaptic filaments. Linear M13mp7 dsDNA was then added and incubation was continued to produce joint complexes (see Materials and Methods). In the initial studies the samples were fixed, chromatographed over Sephacryl S-500, and directly mounted onto thin carbon films and rotary shadowcast (see Materials and Methods). Examination of the samples after incubations of l0 to 40 minutes revealed many joint complexes on each grid square in which a circular presynaptic filament was joined to a linear M13mp7 dsDNA (Fig. 2(a) to (c)). The protein-free dsDNA entered the circular presynaptic filament, disappeared, and then exited the filament a variable distance along the filament. In the joint complexes where the dsDNA was well laid out (30% to 5 0 ~ ) , as

contrasted to flopping over the presynaptic filament several times, the points at which the dsDNA entered or exited the filament marked the junction between two distinctly different morphologies of the presynaptie filament, One portion was indistinguishable in appearance from the circular presynaptie filaments not engaged in joining; showing a smooth-contoured and regular uitrastructure {Fig. 2{c)). The other portion of the presynaptic filament separated by the entry and exit points of the dsDNA had a much more "disorganized" and irregular appearance. The disorganized segment varied between 0-04 and 1.00/~m, equivalent to 200 to 3000 bases of ssDNA. At times, dsDNA could be seen looping out of this region {Fig. 2{b) and (c)), suggesting that the disorganized segment contained all three DNA strands. To confirm that the path of the dsDNA was through the disorganized segment as contrasted to the more regular segment, the lengths of the two dsDNA arms, the regular and disorganized segments of 24 joint complexes were measured. It was assumed that the pitch of the dsDNA was not changed within the presynaptic filament nor was the dsDNA coiled or compacted greatly. On the basis of this assumption the lengths of the two dsDNA arms plus the disorganized segment always summed to a value within 85% of the length of the protein-free M13mp7 dsDNA and did not exceed this value by more than 7 ~/o. In contrast, summing the length of the two dsDNA arms and the regular smooth-contoured segment produced values that often greatly exceeded the length of the protein-free dsDNA. Thus, we conclude that the dsDNA is present in the "disorganized" portion of the joint complex. The minor fraction of linear M13 ssDNA present ( < 1 ~/o) provided a few linear presynaptic filaments which could join with the linear M13mp7 dsDNA to form plectonemic or paranemic joints. As shown in Figure 2(d), structures could be found in which two linear molecules were joined and the region of the joint showed a disorganized structure. Arguments that these joint complexes were not the result of accidental contact between independent presynaptic filaments and free dsDNA were: first, that the concentration of presynaptic filaments and dsDNA on the EM support was adjusted to be below the level where accidental contact was common. Second, if the salt concentration in the incubation was increased to 150mM, the presynaptic filaments remained intact but no joint complexes were seen nor did the presynaptic filaments show any disorganized segments. Third, not one instance was found in which a presynaptic filament was observed containing a disorganized and organized region where there was no dsDNA associated with the filament, nor were situations found where the dsDNA clearly entered and exited the presynaptic filament at a site other than at or very close to the junction between the organized and disorganized segments. These ultrastructural conclusions are the result of over 75 separate EM experiments in which nearly 1000 complexes were

Three-stranded Paranemic Joints

627

I

Figure 2. Visualization of the paranemic joining of linear Ml3mp7 dsDNA with non-Ml3 sequences at its ends and circular M13 ssDNA. Presynaptic filaments containing M13 ssDNA were paired with linear dsDNA (see the text) and prepared for EM by a method involving fixation, dehydration and rotary shadoweasting with tungsten (see Materials and Methods). In these complexes, the dsDNA entered and exited the presynaptic filament at or near the junction between regions on the filament where the appearance changed from a smooth-contoured appearance to a much less organized structure. This disorganized region was shown to contain all 3 DNA strands (see the text) ((a) to (c)). In (d) a linear presynaptic filament is shown engaged in joint formation with a linear dsDNA. The bar represents 0"l/tm.

observed. Only when t h e assembly of the presynaptic filaments was incomplete, usually due to poor A T P regeneration, were a b e r a n t results obtained and structures other t h a n those described here observed.

EM was used to follow the kinetics of joining over a 5 to 60 minute time-course. W i t h a twofold m o l a r excess of p r e s y n a p t i c filaments over dsDNA, the fraction of d s D N A in the t h r e e - s t r a n d e d joints, as described above, increased f r o m 7 ~/o at 5 minutes to

C. Bortner and d. GriJfith

628

Table 1

Kinetics of joining over 60 minutes Lengths of joints (bp) Incubation time (min)

dsDNAin joints (%)

5

7

10 15 20 30 40 60

12 17 22 13 12 2

Minimum --

140 -610 -630 270

Maximum --

1350 -1640 -3300 3200

Average --

670 (n = 27) ±380 -1270 (n=6) ±400 -1570 (n = 15) ±740 1420 (n = 14) ±740

The lengthof the disorganizedsegmentof the 3-stranded paranemiejoints as shownin Figs 2 and 3 was measured directlyfrom electron micrographsand the length expressed in base-pairs of dsDNA. The percentage of joint complexeswas an average from several experiments.

22 % at 20 minutes, and then decreased to < 2 % at 60 minutes. Table 1 summarizes these data along with the lengths of the three-stranded regions at each time point. The wide range Of join t lengths reflects the asynchronous nature of the reaction; thus the average at each time point provides the best description of the progress of the reaction. Although the fraction of dsDNA molecules present in three-stranded joints decreased after 20 minutes, the average length of the joints increased to a maximum at 40 minutes, and did not significantly decrease until after 60 minutes. Qualitatively, the .appearance of the joints changed over the 5 to 60 minutes time-course. Incubations of five to ten minutes produced three-stranded joints that were relatively short and had a less disorganized appearance than when incubations of 20 to 40 minutes were used. Incubations of 60 minutes produced many three-stranded joints that appeared collapsed upon themselves, possibly due to a depletion of ATP (not shown). Glutaraldehyde fixation is known to obscure the helical substructure of presynaptic filaments (Shaner et al., 1987; Heuser & Griffith, 1989). To prepare joint complexes for EM without fixation or air-drying, joining reactions were carried out for 20 minutes, portions taken, and without fixation passed over 1 ml Sephacryl S-500 columns equilibrated in a buffer containing 12 raM-magnesium and 3 mM-ATP. The fraction containing the complexes was immediately applied to the EM support, rapidly frozen and slowly freeze-dried. Without exposing the samples to air, the samples were then rotary shadowcast (see Materials and Methods). As shown in Figure 3, the three-stranded joint complexes visualized by this method had the same overall morphology as those visualized after fixation. Here, however, the helical substructure of the presynaptic filament was much clearer, making the contrast between the helical segment and the disorganized segment even more apparent. The "disorganized" segment often showed a partially beaded appearance typical of segments of ssDNA complexes with SSB protein. This suggested that the disorganized segments might contain both SSB and RecA pro-

teins, since the degree of "beadedness" was not typical of ssDNA complexed with SSB protein alone. (b) S S B protein is present in the disorganized segment of the three-stranded joints Immuno-electron microscopy was used to determine if SSB protein is present in the disorganized region of the joints. A three-step procedure (see Materials and Methods) was employed in which joint complexes were first lightly fixed and chromatographed over Sepharose 4B to remove unbound proteins and fixatives; second, the sample was incubated with a polyclonal rabbit anti-SSB antibody, fixed and chromatographed to remove unbound antibody; and finally the sample was incubated with a gold-conjugated goat anti-rabbit antibody, fixed and chromatographed again. Samples were then examined by EM. Control samples used were M13 ssDNA complexed with only SSB protein or only RecA protein. Using these controls it was found that by including 300 mM-NaC1 in the incubation with the primary antibody, and keeping the incubation time to a maximum of ten minutes, that the background of non-specific staining was diminished to where the complexes containing only RecA protein showed no gold particles, but most of the SSB protein-ssDNA complexes did have clusters of gold particles bound. Thus, very stringent staining conditions were selected. The few linear M13 ssDNA molecules present provided a control for the antibody staining. We have shown that no SSB protein remains in the fully assembled circular presynaptic filaments (Thresher et al., 1988). However, SSB protein does remain at the 5' ends of linear presynaptic filaments due to the 5' to 3' polarity of RecA protein assembly on ssDNA (Register & Griffith, 1985; see Fig. 4(a), insert). Thus, the antibody staining should detect SSB protein at one end of the linear filaments but not along the circular ones. Indeed, inspection of the linear filaments (20 scored) showed no gold particles along their lengths, but 50~o of the linear filaments had gold particles at the end containing the SSB protein (Fig. 4(a), insert). Gold particles were very seidomly

Three-stranded Paranemic Joints

I

629

I

Figure 3. Visualization of 3-stranded joints preserved by rapid-freezing and freeze-drying. Joint complexes were prepared as described for Fig. 2, but frozen rapidly in the reaction buffer without fixation. This was followed by freezedrying and rotary shadowcasting with tantalum at -170°C (see Materials and Methods). This method provides a much better preservation of the 3-dimensional architecture of the complexes clearly revealing the partially beaded appearance of the disorganized joint and the smooth helical structure of the segment containing only 1 strand of DNA. The bar represents 0"l #m.

observed along the length of fully formed circular presynaptic filaments, and in the absence of the primary antibody no binding of the gold-labeled goat anti-rabbit antibody was detected. Inspection of joint complexes formed as described in section (a), above, after 40 minutes of incubation revealed that over half of the three-stranded joint complexes had clusters of gold particles bound (Fig. 4(a) and (b)) and the gold particles were always attached to the disorganized segment of the joint. Given the stringency of the staining conditions employed, the amount of gold staining over the three-stranded joints was consistent with the

conclusion that the disorganized segments contain a mixture of both SSB and RecA proteins. (c) Psoralen photo-crosslinking reveals a single short region over which the three D N A strands are closely associated Psoralen photo-crosslinking was employed to probe the spatial association of the DNA strands in the three-stranded joints. Joint complexes were formed (20 rain incubation), psoralen added, and the sample irradiated with ultraviolet light (see Materials and Methods). One half of the reaction

630

C. Bortner and J. Gri~ith

I

Figure 4. Immunological detection of SSB protein in the joint complexes. Rabbit anti-SSB antibodies were incubated with joint complexes prepared as described for Fig. 2, fixed and chromatographed before the addition of gold-conjugated goat anti-rabbit antibodies and preparation for EM as for Fig. 2. A clustering of gold particles was observed in the disorganized region containing all 3 strands ((a) and (b)), but not along the region of the presynaptic filament containing only l strand of DNA. In (a) (inset) is a linear presynaptic filament containing SSB protein at the 5' end and a cluster of gold label at that site. The bar represents 0"25 ~m.

was fixed and prepared for EM as described in section (a), above. The other half of the reaction was deproteinized and the DNA spread on a monolayer of cytochrome c protein in the presence of 4 0 % formamide (see Materials and Methods). Examination of the fixed sample with the proteins bound confirmed t h a t three-stranded joints (some of which were very long) were present and t h a t the

psoralen and ultraviolet light t r e a t m e n t had not changed their morphology. Counting fields of deproteinized DNA molecules revealed t h a t 2 5 % of the ssDNA circles were attached to a single linear M13mp7 dsDNA (Fig. 5(a) and (b)). Thus, the fraction of linear M13mp7 dsDNA molecules scored as being in joints after psoralen photo-crosslinking and deproteinization was close to the fraction scored

Figure 5. Photo-crosslinking of the DNAs engaged in paranemic joining. Presynaptie filaments containing circular M13 ssDNA were paired with linear M13mp7 dsDNA ((a) and (b)) or nicked Ml3 dsDNA ((c) and (d)), photo-crosslinked with psoralen and ultraviolet light, and the DNA purified and prepared for EM by spreading on a layer of denatured cytochrome c protein with 40o/o formamide (see Materials and Methods). The region in which the DNAs were in close enough association to be crosslinked averaged only 250 bp for both sets of templates, The bar represents 0"25 pm.

Three-stranded P aranemic Joints

631

Figure 6. Visualization of the paranemic joining of circular ssDNA and nicked circular dsDNA. (a) Nicked circular dsDNA was incubated with circular presyn~ptic filaments containing M13 ssDNA and prepared for EM by fast-freezing as described for Fig. 3 without fixation. (b) Complexes as in (a) were fixed, and the dsDNA cleaved with HpaI at a single site to allow the dsDNA to disentangle from the presynaptic filament; here the sample was prepared for EM as for Fig. 2. Only a very short region was observed where the dsDNA entered and exited the presynaptic filament. The bar represents 0-I lam. when the samples were first fixed and examined directly after the 20 minute incubation. The region over which the ss- and dsDNA were attached was short, measuring only 250 bp on average, with joints ranging from 75 to 725 bp ( n = 21). In over 100 molecules examined, none showed more than one site of association between the ss- and dsDNA. Longer irradiation times and use of higher psoralen concentrations did not increase the fraction of ssDNA molecules crosslinked to the dsDNA, the length of the crosslinked regions nor the number of crosslinked sites per complex. To assure ourselves that the short region of psoralen crosslinking and lack of multiple crosslinked sites was not due to inefficient crosslinking, linear M13mp7 dsDNA was treated with psoralen and ultraviolet light in the absence of any protein. The DNA was then treated with glyoxal (see Materials and Methods) to disrupt the hydrogen bonding of the dsDNA and the DNA spread for EM in 40~/o formamide, which spreads out the singlestranded regions. EM examination revealed that the DNA appeared mostly duplex, with 70~o of the length being indistinguishable from perfectly basepaired dsDNA, and the remaining 30~/o being in

small single-stranded bubbles. If the smallest singlestranded bubble that can be detected by this method is 100 bp, then we estimate that on average there was one crosslink every 100 to 150 bp. This is a high enough degree of crosslinking such that had the three DNA strands been in close contact over the entire lengths that ranged up to 3"3 kb, then we would have expected to have seen much longer regions of crosslinking or multiple sites of crosslinking in some DNA molecules. (d) Paranemic joints formed between circular M13 ssDNA and nicked circular M13 dsDNA are short, and the D N A s appear intertwined The pairing of circular M13 ssDNA with nicked Ml3 dsDNA can only result in paranemic joints. Joint complexes using this template pair were prepared exactly as described above and prepared for EM by fixation and air drying, and by fastfreezing and freeze-drying without fixation (Fig. 6(a)). In multiple experiments utilizing both methods of preparation, EM examination revealed paranemic joint complexes in which a single circular presynaptic filament was associated with a single M13 dsDNA (close to 60~o of the M13 dsDNA was

632

C. Bortner and J. Gri~th

present in such joint complexes). Here, the proteinfree dsDNA appeared highly twisted about the presynaptic filament and this feature remained throughout the l0 to 40 minute incubations. This twisting made it difficult to measure the length of the jomts and to inspect their structure. To disentangle the dsDNA from the presynaptic filaments: following a 20 minute incubation the complexes were fixed, chromatographed over Sephacryl S-500 and then treated with HpaI to cleave the dsDNA at a single site. Examination of the joints revealed a short compact region of joining with no disorganized segment (Fig. 6(b)) and their lengths varied from 30 to 265 bp with an average of 125 bp (n = 17). Paranemic complexes formed between the two circular DNAs (20min incubation) were photocrosslinked with psoralen. Examination of the deproteinized crosslinked DNA revealed that 25~/o of the M13 ssDNA circles were attached to relaxed Ml3 dsDNA circles (Fig. 5(c) and (d)). As described above, only single sites of attachments between the two DNAs were found and the length of the region of association averaged 250 bp, with a variation from 60 bp to 470 bp ( n = 14). This was slightly larger than the length observed above after restriction digestion of the dsDNA, but due to the tangled nature of the non-deproteinized joint a clear region of invasion was not always apparent using this set of templates, none the less, both values are close.

4. Discussion

In this study we have utilized electron microscopy to investigate the structure of three-stranded joints that form between two sets of DNA templates that differ only by the presence or absence of free dsDNA ends. The pairing of circular ssDNA with linear dsDNA produced three-stranded joints often several kilobases ( x 10a bases) in length. These joints had a relatively disorganized ultrastructure and contained both RecA and SSB protein. Within these long joints the three DNA strands were in close association only at a single site and over a relatively short (-~250 bp) distance as revealed by psoralen photo-crosslinking. When free dsDNA ends were absent, the two circular DNAs were highly twisted upon themselves, and the joints formed were Short, corresponding to the region over which the three DNA strands could be photo-crosslinked. We have avoided using the term "paranemic" to describe the long three-stranded joints formed by the pairing of the linear dsDNA and circular ssDNA since, as described below, we believe that these long joints contain a region of true paranemic association only over a single short distance followed by a long region of a much looser association. (a) Architecture of the joints Most of the samples examined in this study were chemically fixed with glutaraldehyde during their

preparation for EM. As shown elsewhere (Shaner et al., 1987; Heuser & Griffith, 1989), fixation obscures the fine ultrastructure of the RecA protein filaments. However, when samples were prepared by the rapid-freezing and freeze-drying method in which the samples were prepared directly in the joining reaction buffer without fixation, the results augmented those derived from the chemically fixed samples. Furthermore, the striking preservation of three-dimensional structure afforded by freezedrying provided a much clearer definition of the architecture of these complexes. It would not have been practical to have prepared all of the samples by this newer method and for the EM immunogold staining as fixation was required. SSB protein is necessary for the assembly of active presynaptic filaments under ionic and cofactor conditions optimal for joint formation (McEntee et al., 1980). Here, we observed that SSB (and RecA) protein was present in the disorganized regions of the long three-stranded joints, and at one end of linear presynaptic filaments, but was not present over the length of fully assembled presynaptic filaments. The latter two observations provide a useful confirmation of the earlier studies (Register & Griffith, 1985; Thresher et al., 1988). SSB protein was not required for the creation of the disorganized structure of the long three-stranded joints: when presynaptic filaments were formed in the absence of SSB protein using the magnesium shift protocol of Flory et al. (1984), disorganized joints of the same general appearance (but lacking a beadedness) were observed (data not shown). Although the immunological staining for SSB protein was not carried out with the pairing of the two circular DNAs for technical reasons, the short compact appearance of the joints formed argued strongly for the absence of SSB protein. Had we probed the structure of the joints formed by the two template pairs by psoralen photocrosslinking alone, we would have concluded that there was little difference in the joints formed. In both, there was only a single short region of close contact between the ssDNA and dsDNA, averaging 250 bp in length. A reasonable interpretation of this finding is that the region of "true" paranemic joining for both sets of templates is restricted to only ~ 250 bp, the region that could be crosslinked. The very long disorganized region produced when linear dsDNA was used may result from the movement of the joint along the presynaptic filament, as discussed below. The disorganized ultrastructure cannot be attributed solely to the presence of three DNA strands, since in the plectonemic exchange of strands between a circular ssDNA and a linear dsDNA (Stasiak et al., 1984; Register et al., 1987), the dsDNA melds into the circular presynaptic filament over many kb without altering the helical appearance of the presynaptic filament, which must contain all three DNA strands. The specific architecture of the disorganized region can be explained, we believe, from the observations of Register & Griffith (1988) on the

Three-stranded Paranemic Joints pairing of supertwisted dsDNA and fully homologous linear ssDNA. Here, the product is a D-loop, which moves around the circular dsDNA in a "rolling D-loop" reaction. When the products were examined by EM, it was found that as the D-loop moved along the dsDNA, a loosely organized coating of RecA protein was left behind on the dsDNA. Here, we would suggest that if the paranemic joint moves along the presynaptic filament (see below) the dsDNA reanneals in its wake but retains a loose coating of RecA protein. The ssDNA, which may be wrapped loosely about the dsDNA, is then fi'ee to bind SSB protein.

(b) Topological constraints on joining The intertwining of the nicked circular dsDNA about the circular presynaptic filament argues, we believe, that the dsDNA and the presynaptic filament must rotate about each other during joining. Whether the dsDNA coils about the presynaptic filament or vice versa we cannot say, although it would seem that the dsDNA could more easily rotate about the presynaptic filament. This intertwining cannot be due to the rotation of the dsDNA about its own axis, since this motion would have been relieved by the presence of the nick in the dsDNA. Also, as in previous studies, when circular M13 ssDNA was paranemically paired with circular supertwisted Ml3 dsDNA, the dsDNA appeared to be twisted about the presynaptic filament (Christiansen & Griffith, 1986). These conclusions are not in conflict with the work of Honigberg & Radding (1988), since in that study the templates used were different and plectonemic pairing was examined. The differences between the observations presented in this paper and our earlier work (Christiansen & Griffith, 1986), in which paranemic joints were formed between supertwisted dsDNA and linear presynaptic filaments, are revealing. In the former study, the paranemic joints were short and showed no significant disruption of the presynaptic filament. In the present study, the joints (with the linear dsDNA) were long and the presynaptic filament appeared disrupted. The use of circular supertwisted dsDNA imposed major constraints. The negative superhelical coiling very likely facilitated joint formation, but once the supertwisting was relieved, further extension of the joint would have produced positive supertwists. These positive supertwists may have created a counterforce acting to retain a compact structure in the joints. If, as argued below, the presence of a free dsDNA end allows the paranemic joint to move along the presynaptic filament, and this movement disrupts the helical ultrastructure of the filament, then the absence of any disruption seen in the former study would argue that the superhelical strain in the dsDNA inhibited the movement of those paranemic joints.

633

(c) Growth (movement) of paranemic joints The disruption of the presynaptic filament observed with linear dsDNA may reflect the movement of the paranemic joints along the presynaptic filaments. Does the quantitative data obtained support this suggestion? The average length of the joints formed with the linear dsDNA increased over time and the maximum joint length increased greatly. After a ten minute incubation the average joint length was 670(+380)bp, with the longest joint observed measuring 1350bp, while a 40 minute incubation produced joints with an average length of 1420(_+740)bp and joints as long as 3300 bp were observed. Given that the disorganized joints increase in length during the incubation, do they grow unidirectionally, bidirectionally or do they grow by forming multiple contacts which then fuse into a long joint? The results of the psoralen photo-crosslinking argue strongly, we believe, for unidirectional growth. First, it provides direct evidence against there being multiple regions of true paranemic joining which fuse with time; only single sites of crosslinking were seen. Secondly, it seems reasonable to assume that the region of close contact between the three DNA strands would be at the borders of the joint and the uninvolved segments of the presynaptic filament rather than in the center of the joint. The presence of a single short region over which the three DNA strands could be crosslinked thus argues for growth at only one of the two ends of the joint.

(d) A model of the long three-stranded joints These observations lead to a model of how the long three-stranded joints form and grow. In the first step, a true paranemic synapsis forms at a site of homology internal to the ends of the linear dsDNA. This may involve a fusion of all three DNA strands over a 50 to 250 bp length of the dsDNA. Whether the dsDNA must rotate about the presynaptic filament at this step we cannot say. In the second step, the paranemic joint moves along the length of the presynaptic filament. Based on the polarity of RecA protein assembly onto ssDNA (Register & Griffith, 1985), we suggest that the joints move 5' to 3' relative to the ssDNA but we have no experimental data to support this suggestion. As the joint moves the linear dsDNA is free to rotate about the presynaptic filament, resulting in a loose intertwining of the ds- and ssDNAs. Thus, in this model (Fig. 7), true paranemic association of all three DNA strands occurs within a region that remains short and lies at one end of the joint. The region of the presynaptic filament left in the wake of this movement contains the original dsDNA whose strands have separated from their close contact with the ssDNA, but the two DNAs none the less remain twisted about each other. In these studies the role of SSB protein was clearly incidental; being present in the reactions, it decorated the RecA protein-free

634

C. Bortner and J. Gri~th

Figure 7. Schematic illustration of the architecture of the 3-stranded joints. As the true paranemic joint moves along the presynaptic filament, the presence of a free dsDNA end allows the rotation of the dsDNA about the presynaptic filament, thus leaving a disorganized region in its wake. All 3 strands are contained within this disorganized segment, but are in close contact only in the region of the true paranemic joint.

ssDNA. In the cell, SSB protein could play a more i m p o r t a n t role of protecting ssDNA from e n z y m a t i c attack. This work was supported by a grant from the NIH (GM-31819) and the ACS (NP 583). References

Bianchi, M., DasGupta, C. & Radding, C. M. (1983). Cell, 34, 931-939. Chase, J. W., Whittier, R. F., Auerbach, J., Sanear, A. & Rupp, W. D. (1980). Nucl. Acids Res. 8, 3215-3227. Chow, L. T. & Broker, T. R. (1981). In Electron Microscopy in Bioloyy (Griffith, J. D., ed.), vol. 2, pp. 139-188, John Wiley, New York. Christiansen, G. & Griffith, J. (1986). Proc. Nat. Acad. Sci., U.S.A. 83, 2066-2070.

Cunningham, R. P., Wu, A. M., Shibata, T., DasGupta, C. & Radding, C. M. (1981). Cell, 24, 213-223. DasGupta, C., Shibata, T., Cunningham, R. P. & Radding, C. M. (1980). Cell, 22, 437-446, Flory, J., Tsang, S. S. & Radding, C. M. (1984). Proc, Nat. Acad. Sci., U.S.A. 81, 7026-7030. Griffith J. D. & Christiansen, G. (1978). Annu. Rev. Biophys. Bioeng. 7, 19-35. Griffith J. D. & Harris, L. D. (1988). CRC Crit. Rev. Biochem. 23, $43-$86. (Suppl. 1). Griffith J. D. & Shores, C. G. (1985). Biochemistry, 24, 158-162. Harris L. D. & Griffith, J. (1987). J. Biol. Chem. 262, 9285-9292. Harris L. D. & Griffith, J. D. (1988). Biochemistry, 27, 6954-6959. Heuser, J. (1983). J. Mol. Biol. 169, 155-195. Heuser. J. & Griffith, J. (1989). J. Mol. Biol. 210. 473-484. Honigberg, S. M. & Radding, C. M. (1988). Cell, 54, 525-532. McEntee, K., Weinstock, G. M. & Lehman, I. R. (1979). PTvc. Nat. Acad. Sci., U.S.A. 76, 2615-2619. McEntee, K., Weinstock, G. M. & Lehman, I. R. (1980). Proc. Nat. Acad. Sci., U.S.A. 77, 857-861. Register, J. C. & Griffith, J. (1985). J. Biol. Chem. 260, 12308-12312. Register, J. C. & Griffith, J. (1986). Proc. Nat. Acad. 8ci.. U.S.A. 83,624-628. Register, J. C. & Griffith, J. (1988). J. Biol. Chem. 263, ! 1029-11032. Register, J. C., Christiansen, G. & Griffith, J. D. (1987). J. Biol. Chem. 262, 12812-12820. Riddles, P. W. & Lehman, I. R. (1985). J. Biol. Chem. 260, 165-169. Shaner, S. L., Flory, J. & Radding, C. M. (1987). J. Biol. Chem, 262, 9211-9219. Shibata, T., DasGupta, C., Cunningham, R. P. & Radding, C. M. (1979). Proc. Nat. Acad. Sci., U.S.A. 76, 1638-1642. Stasiak, A., Stasiak, A. Z. & Koller, T. (1984). Cold Spring Harbor Syrup. Quant. Biol. 49, 561-570. Thresher, R. J., Christiansen, G. & Griffith, J. D. (1988). J. Mol. Biol. 201, 101-113.

Edited by P. yon Hippel