Action of ATP-dependent DNase from Hemophilus influenzae on cross-linked DNA molecules.

The ATP-dependent DNase from Hemophilus influenzae digests double-stranded linear DNA molecules exonucleolytically while hydrolyzing large amounts of ATP to ADP. Various cross-linked linear duplex DNA molecules are partially resistant to the exonuclease action. Vaccinia DNA, containing natural terminal cross-links (probably in the form of terminal single-stranded loops), is much more slowly degraded than comparable "open-ended" DNA molecules, and ATP is consumed at a proportionately lower rate. It is postulated that the vaccinia DNA molecules undergo slow terminal cleavage by the single strand specific endonuclease activity of the enzyme, and are then rapidly degraded by the double strand exonuclease activity. Phage T7 DNA, containing an average of 100 4',5'8-trimethylpsoralen cross-links/molecule at random internal sites, is digested only to the extent of 2 to 3%. However, ATP hydrolysis continues at a linear rate long after DNA digestion has ceased. A stable enzyme-DNA complex is formed as demonstrated by co-sedimentation of DNA and ATPase activity in sucrose gradients. The hypothesis is advanced that the enzyme digests exonucleolytically to the first cross-link at each end of the DNA molecules where further movement is prevented. The enzyme then remains bound at the cross-links and functions continuously as an ATPase.

The general properties of the ATP-dependent DNase (exonuclease V) from Hemophilus influenzae have been described in previous papers from this laboratory (l-4) and its action on a variety of DNA substrates has been studied (3). Duplex linear DNA is rapidly degraded under standard reaction conditions, but duplex circular DNA is not attacked. The enzyme's classification as an exonuclease is based on this specificity. However, paradoxically it acts as an endonuclease activity on single-stranded DNA. Single-stranded circular DNA molecules are slowly cleaved even by highly purified preparations of enzyme (3,5). This endonucleolytic activity is inhibited by the presence of linear duplex or single-stranded DNA both of which are preferred as substrates to the circular single-stranded form (3). Linear single-stranded DNA is digested at one-tenth the rate of duplex DNA molecules. DNA substrates in the order of enzyme preference are therefore duplex linear DNA > singlestranded linear DNA > single-stranded circular DNA > > duplex circular DNA.
Hemophilus exonuclease V' has two other enzymatic activities on duplex DNA in addition to nuclease activity: (a) it is a very active DNA-dependent ATPase and under standard * This work was supported by United States Public Health Service Grant A107875.
1 The abbreviations used are: Hemophilus exonuclease V (11) or exo V, Hemophilus psoralen,4,5', reaction conditions as many as 30 to 40 ATP molecules are hydrolyzed to ADP per phosphodiester bond cleaved (2); (6) it possesses a DNA unwinding activity that converts doublestranded DNA into single-stranded DNA early in a reaction, prior to the accumulation of significant acid-soluble oligonucleotide products (4, 6). Putting these observations together, we have formulated a simple working model for the enzyme mechanism. We propose that enzyme initiates attack at the ends of duplex DNA, and then utilizes energy from ATP hydrolysis to move along the DNA and unwind regions of the molecule, releasing large partially or totally single-stranded fragments as cleavage occurs.
In this paper and two subsequent articles we describe evidence supporting the general features of this model. Here we analyze the action of exonuclease V on two kinds of DNA containing covalent cross-links that prevent effective strand separation: vaccinia virus DNA containing naturally occurring terminal cross-links (7), and DNA artificially cross-linked at random internal sites by 4,5,%trimethylpsoralen (8). According to our hypothesis we would expect DNA cross-links to impede processive enzyme movement, thus decreasing DNA digestion and strand separation.
We shall show: (a) that vaccinia DNA is resistant to digestion, and (6) that psoralencross-linked DNA is also resistant but binds to the enzyme and leads to prolonged hydrolysis of ATP. This uncoupling of the ATPase and DNase reactions with psoralen-cross-linked DNA (1) except as otherwise noted. Acid-soluble and acid-precipitable DNA radioactivities were determined by chromatography in 1 N HCl on polyethyleneimine-cellulose thin layer sheets as previously described (1). Conversion of ATP to ADP was measured by polyethyleneimine-cellulose chromatography in 0.5 M potassium phosphate buffer, pH 3.5, as previously described (2). Agarose Gel Electrophoresis-DNA samples were analyzed on 0.6% agarose gels prepared in glass tubes (0.6 cm inside diameter by 12 cm length). The electrode buffer was 0. was for 2 h at 48,000 rpm at 4" in a Spinco SW 50 rotor. Fractions were collected and assayed as detailed under "Results."

Hemophilus
Exonuclease V Activity on Vaccinia DNA- incubated separately with enzyme. Thus vaccinia DNA is resistant to digestion by exonuclease V as are the two circular forms of SV40 DNA. From this we conclude that enzyme initiates attack at DNA ends and that the natural cross-links at the termini of vaccinia DNA protect against enzyme attack. Enzyme Activity on Psoralen-Cross-linked DNA--Because naturally cross-linked vaccinia DNA molecules are resistant to digestion by Hemophilus exo V, it is of interest to examine enzyme activity on other cross-linked DNAs, for example, those that have been chemically cross-linked at internal sites. DNA can be cross-linked with psoralen in a photochemical reaction using 360 nm wavelength ultraviolet light (8). Covalent interstrand cross-links of psoralen are introduced at various sites along the DNA without significant strand breakage (8,9). Phage T7 [3*P]DNA was heavily cross-linked with psoralen as described under "Experimental Procedures" and then used as a substrate for exo V in a reaction containing [SH]ATP (Fig.  2). Only about 2% of the cross-linked DNA was converted to acid-soluble material during 45 min of incubation (Fig. 2, Curve A) as compared to 60% conversion of non-cross-linked DNA to acid-soluble products (representing almost a limit product) in an otherwise identical control reaction (Fig. 2, Curve B). ATP hydrolysis, on the other hand, proceeded at a linear rate throughout the observed time period in the crosslinked DNA reaction (Fig. 2, Curve C), while in the control reaction, ATP hydrolysis reached a plateau within 10 min (Fig.  2, Curve D), roughly paralleling the kinetics of DNA digestion. Thus psoralen-cross-linked DNA is resistant to degradation but supports prolonged ATPase activity. This apparent uncoupling of ATPase from DNase activity on psoralen-cross-linked DNA has also been observed by Karu and Linn (9) using the exonuclease V from Escherichia cob. A reasonable interpretation of our results is that exo V degrades DNA up to the outermost psoralen cross-link, where it remains tightly bound, forming a DNA*enzyme complex having ATPase activity. We give evidence in support of this below.
Enzyme Binding to Cross-linked DNA--To test for formation of a stable complex between enzyme and psoralen-cross-linked DNA, a less than saturating amount (10) of Hemophilus exo V was preincubated for 3 min at 37" with unlabeled, native T7 DNA or cross-linked T7 DNA in two separate complete reaction mixtures. Native T7 ['ZP]DNA was then added to each tube, and its degradation measured as an assay for any unbound, active exo V. Preincubation with native T7 DNA did not affect the activity of the enzyme, as shown by the rapid degradation of the added ['*P]DNA (Fig. 3). However, preincubation with cross-linked DNA greatly reduced the available exonucleolytic activity, as shown by the slow degradation of the added [32P]DNA. This result is compatible with our proposal that the enzyme binds strongly to cross-linked DNA. An alternate explanation is that after contact with cross-linked DNA, enzyme dissociates into a form that cannot digest DNA but can act as an ATPase.
Actual binding was demonstrated by showing that enzyme activity cosediments with cross-linked T7 DNA in sucrose gradients. Heavily cross-linked T7 [82P]DNA (6 nmol) was incubated with an excess of enzyme (8.8 units) in a reaction mixture containing both ATP and Mg*+. The mixture was then layered onto a sucrose gradient and centrifuged to separate DNA from free enzyme. Fractions were collected and assayed for DNA, for ATPase activity and for DNase activity. The results are shown in Fig. 4. Most (85%) of the DNA was detected as a peak of 32P radioactivity centering in Fractions 15 to 17. This corresponds approximately to the position of intact T7 DNA, as determined by sedimentation on parallel gradients of similar reaction mixtures lacking either ATP or Mg'+, and indicates that there was little degradation of the cross-linked DNA. The small percentage (15%) of radioactivity near the top of the gradient may represent material produced by enzymatic degradation of the region between a DNA terminus and the outermost cross-link.
ATPase activity sedimented almost exactly with the nearly intact [82P]DNA at Fractions 14 through 18, except for a small shoulder extending upward in the gradient. DNase activity was not detected in the region of the intact DNA, but was found higher in the gradient at Reaction Mixture 2 was identical except non-crosslinked DNA was used. Incubation was at 37". Five-microliter samples were taken at intervals and analyzed chromatographically for ATP hydrolysis and acid-soluble "P radioactivity as described under "Experimental Procedure." Curue A, acid-soluble 3T radioactivity from cross-linked DNA; Curve B, acid-soluble 92P radioactivity from non-cross-linked DNA; Curoe C, ATP hydrolysis on cross-linked DNA; Curve D, ATP hydrolysis on non-cross-linked DNA. DNA. ATP hydrolysis was followed for 30 min (Fig. 5). Hydrolysis was rapid in the cross-linked and non-cross-linked T7 DNA reactions (Fig. 5, Curves B and C), but only one-sixth as rapid initially with vaccinia DNA (Fig. 5, Curve A). We anticipated that if vaccinia DNA were completely resistant to digestion, no ATP hydrolysis would take place. But in fact there was significant hydrolysis, suggesting that vaccinia DNA is attacked 'at a slow rate that escaped detection in the experiment of Fig. 1. To assay for vaccinia DNA digestion samples of the vaccinia reaction mixture were analyzed by agarose gel electrophoresis and the amount of DNA in the position of the vaccinia band was determined by densitometer tracings at each time point. The per cent disappearance of vaccinia DNA is plotted in Curve D of Fig. 5 40% of the vaccinia molecules were degraded so that no intermediate products were visible on the gels. However, the remaining 60% moved in the normal position of intact molecules. Slowing of the reaction probably reflects increasing competition of the oligonucleotide products for the available enzyme rather than a resistant fraction of vaccinia DNA. The structure of the terminal vaccinia cross-links is unknown; but we conclude that they are somewhat susceptible to attack by exo V, accounting for the low ATPase activity observed. It should be noted that under the conditions of our experiment, native T7 DNA is completely digested and the psoralen-crosslinked DNA is almost completely resistant (data not shown).

DISCUSSION
Cross-linked DNA molecules prove to be interesting substrates for exonuclease V. Terminally cross-linked vaccinia DNA is only slowly attacked and supports a correspondingly low rate of ATP hydrolysis (Figs. 1 and 5). Interpretation of these results is complicated by incomplete knowledge of the vaccinia cross-link structure. However, closure of each DNA end by a short single-stranded loop is compatible with the known facts. The cross-links have been located within six nucleotides of the ends3; furthermore, they can be cleaved by a single strand specific endonuclease contained in the vaccinia virion (7).
Assuming that the ends of vaccinia DNA are sealed by short loops, a simple mechanism based on established properties of Hemophilus exo V can be put forward to explain our observations (Fig. 6A no ATPase activation.') However, exo V possesses weak single strand specific endonuclease activity (1, 5), and apparently it can slowly cleave the single-stranded terminal loops. As soon as one end of a molecule is opened the whole molecule is quickly digested to oligonucleotides with accompanying hydrolysis of ATP. Endonucleolytic terminal cleavage would appear to be the rate-limiting step. Psoralen-cross-linked DNA behaves differently from vaccinia DNA as an exo V substrate. Psoralen-cross-links are presumably introduced fairly randomly into T7 DNA so that, in general, they are situated internally leaving some portion of the ends free. It is reasonable to expect that exonuclease V can initiate attack at these ends without inhibiton by cross-links several hundred bases toward the interior. Our experiments confirm this. Our heavily cross-linked DNA contains approximately 100 randomly intercalated psoralen molecules, leaving an estimated average of 2% psoralen-free DNA at the ends. This agrees well with the observed 2 to 3% available for digestion (Fig. 2, Curve A). With less extensively cross-linked DNA a greater average length at the ends of the DNA molecules is free of cross-links. We have observed with such Unpublished results.
DNA a proportionate increase in the amount of exo V digestible DNA (data not shown). Our various observations are most easily explained by the mechanism illustrated in Fig. 6B. An enzyme molecule binds to each end of a psoralen-cross-linked DNA molecule (10) and moves along cleaving phosphodiester bonds and hydrolyzing ATP. Progress is halted at the first cross-links encountered. At these sites enzyme remains tightly bound but continues to hydrolyze ATP. Regions of the molecule enclosed by cross-links form a resistant core because exo V can neither attack duplex DNA directly as an endonuclease (Fig. 1) nor bypass the cross-links.
The preincubation experiment of Fig. 3 and the co-sedimentation experiment of Fig. 4 provide evidence of a tight complex between psoralen-cross-linked DNA and exonuclease V. Since we have argued that the psoralen-free terminal region (2%) is removed by exo V digestion, and that the enzyme cannot bypass a cross-link, we must conclude that the enzyme remains bound at the cross-link as illustrated in Fig. 6B. We have as yet been unable to directly demonstrate this by electron microscopy, nor have we ruled out dissociation of one or more enzyme subunits (1, 2). We have demonstrated that bound enzyme retains ATPase activity (Figs. 2 and 4), and that no subunit containing DNase activity is released by preincubation on psoralen-cross-linked DNA (Fig. 3). On the assumption that complete enzyme is bound at a cross-link, one must conclude that a firm stoichiometry between ATP hydrolysis and phosphodiester bond cleavage does not exist. Previously observed ratios of 30 to 40 ATP molecules hydrolyzed/DNA bond cleaved presumably do not reflect direct coupling between these two activities. In a subsequent paper (6), experiments are described that lend support to the hypothesis that ATP is utilized for movement and strand separation rather than for strand cleavage.