Crystal structures of λ exonuclease in complex with DNA suggest an electrostatic ratchet mechanism for processivity

Abstract

The λ exonuclease is an ATP-independent enzyme that binds to dsDNA ends and processively digests the 5′-ended strand to form 5′ mononucleotides and a long 3′ overhang. The crystal structure of λ exonuclease revealed a toroidal homotrimer with a central funnel-shaped channel for tracking along the DNA, and a mechanism for processivity based on topological linkage of the trimer to the DNA was proposed. Here, we have determined the crystal structure of λ exonuclease in complex with DNA at 1.88-Å resolution. The structure reveals that the enzyme unwinds the DNA prior to cleavage, such that two nucleotides of the 5′-ended strand insert into the active site of one subunit of the trimer, while the 3′-ended strand passes through the central channel to emerge out the back of the trimer. Unwinding of the DNA is facilitated by several apolar residues, including Leu78, that wedge into the base pairs at the single/double-strand junction to form favorable hydrophobic interactions. The terminal 5′ phosphate of the DNA binds to a positively charged pocket buried at the end of the active site, while the scissile phosphate bridges two active site Mg²⁺ ions. Our data suggest a mechanism for processivity in which wedging of Leu78 and other apolar residues into the base pairs of the DNA restricts backward movement, whereas attraction of the 5′ phosphate to the positively charged pocket drives forward movement of the enzyme along the DNA at each cycle of the reaction. Thus, processivity of λ exonuclease operates not only at the level of the trimer, but also at the level of the monomer.

The resection of DNA ends to form a long 3′ overhang is a fundamental step in the repair of dsDNA breaks by homologous recombination. Whereas in eukaryotes DNA end-resection involves a complex network of proteins (1), bacteriophage and other DNA viruses encode a simple two-component “SynExo” recombination system that consists of a processive 5′-3′ exonuclease to resect DNA ends, and a single-strand annealing (SSA) protein to bind the resulting 3′ overhang and promote annealing of complementary strands (2–4). The two proteins interact with one another to form a complex known as a “synaptasome” (4–6), which may serve to physically load the SSA protein onto the 3′ overhang as it is generated by the exonuclease. The Redαβ system of bacteriophage λ, which consists of λ exonuclease and β-protein, is particularly well studied and is currently being employed in powerful methods for genetic engineering and nanopore DNA sequencing (7–9).

The λ exonuclease is a Mg²⁺-dependent enzyme that binds to dsDNA ends and processively digests the 5′-ended strand to form 5′ mononucleotides and a long 3′-ended ssDNA tail (6, 10). The crystal structure of λ exonuclease, determined in the absence of DNA, revealed a ring-shaped homotrimer with a central channel that is wide enough to bind dsDNA at one end, but only allow passage of ssDNA at the other (11). Because the DNA could remain topologically linked to the trimer via threading of the 3′ overhang through the central channel, the structure provided a compelling basis for the highly processive nature of the reaction. The core fold of λ exonuclease is conserved in type II restriction endonucleases (12), and in several nucleases involved in DNA repair (13–15), viral replication (16–18), and RNA processing (19, 20). These enzymes all exhibit a conserved PD-(D/E)XK motif that binds one or more metal ions, usually Mg²⁺, to form the active site for cleavage.

The crystal structure of λ exonuclease provided a compelling model for its processive mode of action, but in the absence of a structure with DNA substrate, several questions remain. If the DNA binds within the central channel of the trimer as predicted, how does the 5′ end of the DNA access one of the three active sites, which lie approximately 15 Å from the central channel? Does the enzyme use all three of its active sites in a sequential, cyclical manner, or does the DNA engage with a single active site on the trimer for multiple rounds of cleavage? Is processivity based solely on topological linkage of the trimer to the DNA, or does the enzyme possess a “motor” activity that couples energy released from cleavage of the phosphodiester bond to forward movement along the DNA? Here, we present crystal structures of λ exonuclease in complex with DNA that show the detailed interactions between the terminal nucleotides of the DNA and the active site residues of the protein. Based on these structures, and on accompanying biochemical data, we propose an “electrostatic ratchet” mechanism for processivity in which forward movement of the enzyme is driven by attraction of the terminal 5′ phosphate of the DNA to a positively charged pocket at the end of the active site cleft.

Results

Crystal Structures of λ Exonuclease Bound to DNA.

We have determined two crystal structures of λ exonuclease in complex with DNA. The first is a complex with a symmetric 12 bp duplex with 5′-OH ends in the presence of Ca²⁺, which replaces Mg²⁺ but inhibits nuclease activity (10). In this structure, the DNA is bound within the central channel of the trimer with the 5′ end of one strand projecting toward one of the three active sites (Fig. S1A). However, the scissile phosphate of the terminal nucleotide is still 11 Å from the active site Ca²⁺, and thus the DNA is not fully inserted into the active site. This structure likely represents a nonproductive complex that has been detected in enzymatic studies on DNA substrates with 5′-OH ends (21, 22). Based on this structure, we designed a DNA with a 12-bp duplex and a 5′-dinucleotide overhang, which was intended to extend fully into the active site. Because previous data indicated the importance of a 5′ phosphate on the DNA for maximal activity (10, 21, 22), we also added a 5′ phosphate to the 14-mer strand of the duplex. Finally, we used the inactive K131A variant of the enzyme, so that the complex could be crystallized with Mg²⁺ instead of Ca²⁺. The resulting structure, refined at 1.88-Å resolution (Table S1), reveals a catalytically relevant complex that will now be described. Aside from the noted differences in the active site cleft, the two structures are very similar to one another.

The DNA binds within the central channel of the trimer as expected, but is asymmetrically tilted so that the 5′-dinucleotide overhang binds to the active site of subunit B, the green subunit as shown in Fig. 1. The duplex portion of the DNA is essentially B form (Fig. S2), and there are no large conformational changes that occur in the protein upon DNA binding: The bound and unbound trimers superimpose to an rmsd of 1.2 Å for all C^α atoms. The loop connecting helices C and D of each subunit (CD loop, residues 42–50) extends out from rim of the central channel to contact the downstream portion of the DNA in three different ways (Fig. 1C). As expected for a non-sequence-specific DNA-binding protein, most of the interactions involve hydrogen bonds or ion pairs to the sugar-phosphate backbone, and there are few contacts with the bases (Fig. S3). A notable interaction is the insertion of Arg45 of subunit C deep into the minor groove of the downstream portion of the DNA, with its positively charged guanidino group lying midway between the two sugar-phosphate backbones (Fig. 1 C and D). This type of interaction, seen prominently in the nucleosome, favors A-tract sequences in which a narrowing of the minor groove leads to an increased negative electrostatic potential (23). Although the sequence is not an A tract, the minor groove at this region of the DNA narrows by about 1–2 Å to close in on the side chain of Arg45. Because this interaction does not involve sequence-specific hydrogen bonds, Arg45 could act as a rudder to help keep the enzyme on track as it moves along the DNA.

Fig. 1.

Crystal structure of λ exonuclease in complex with DNA. (A) The DNA sequence used for cocrystallization. (B) Stereo ribbon view with subunit A, cyan; B, green; and C, wheat. The 5′-dinucleotide overhang of the DNA binds to the active site of subunit B. The two Mg²⁺ ions are shown as magenta spheres. Phosphate ions of subunits A and C are shown as orange and red spheres, and the two chloride ions of subunit A are shown as gray spheres. (C) Side view showing the binding of the CD loops to the downstream portion of the DNA. The side chain of Arg45 of subunit C inserts into the minor groove. The 1.88-Å F_o - F_c electron density map is contoured at 2.0σ around the DNA. (D) Electrostatic surface view generated in PyMOL with negative red and positive blue. The three Arg45 residues, which contact the DNA in the central channel, are labeled. (E) Surface view from the back of B, showing how the DE loop splits the two strands of the duplex. A rear portal through which cleaved mononucleotides could be released is also indicated.

The structure reveals that the enzyme unwinds exactly two base pairs from the end of the DNA prior to cleavage. The two nucleotides of the 5′ overhang twist away from the duplex to insert into the active site of subunit B, while the last nucleotide of the 3′-ended strand pokes through the central channel to emerge out the back of the trimer (Fig. 1E). The loop between helices D and E (DE loop, residues 69–75) extends across the central channel at the back of the trimer, with the 3′ end of the DNA on one side and the 5′ end on the other. Four apolar residues of the DE loop, Val73, Ala75, Ala77, and Leu78, pack against the base pair exposed at the end of the duplex to form a hydrophobic wedge that is positioned to split apart the base pairs as the enzyme moves along the DNA (Fig. 2C). In particular, the side chain of Leu78 inserts between the second and third bases of the 5′-ended strand to separate the two terminal nucleotides in the active site from the duplex portion of the DNA in the central channel. The bases of the two nucleotides in the active site stack on top of one another and are sandwiched by Leu78 on one side and Trp24 on the other. Trp24 does not form stacking interactions, as might be expected, but instead packs against the face of the terminal base with the edge of its indole ring (Fig. 2A).

Fig. 2.

Active site interactions. (A) Close-up view of the active site cleft. The two Mg²⁺ ions are shown as magenta spheres with coordinating interactions in dashed lines. The arrow indicates the proposed attack of the hydrolytic water molecule on the scissile phosphate. (B) Stereo view of electron density around the Mg²⁺ ions. The interactions of three additional conserved residues, Glu85, Gln157, and Tyr154, can be seen in this view. (C) Hydrophobic interactions between residues of the DE loop and the single/double-strand junction of the DNA. The side chains of Val73, Ala75, Ala77, and Leu78 bury 55, 26, 36, and 80 Å² of solvent accessible surface area in the complex, respectively. (D) Interactions of the 5′ phosphate with the positively charged pocket. (E). Electrostatic surface view of the 5′-phosphate-binding pocket. The view is from the back of A, looking into the active site through the rear portal. The surface of the protein is transparent so that the terminal nucleotides of the DNA can be seen. The positive charge of the 5′-phosphate-binding pocket, which is deeply buried, can be seen behind the 5′ phosphate. The positive charge in front of the bases is from Lys82, a residue that lines the rear portal.

In the active site to which the DNA is bound, there are now two Mg²⁺ ions, both of which exhibit octahedral coordination (Fig. 2). Thus, the structure reveals that λ exonuclease uses a classic two-metal mechanism (24) seen for the proofreading domain of Escherichia coli DNA polymerase 1 (25), RNaseH (26), MutH protein (13), and several type II restriction endonucleases (27). The two Mg²⁺ ions, Mg^A and Mg^B, are spaced by 4.0 Å and bridged by Asp119 on one side and the scissile phosphate of the DNA on the other, which coordinates Mg^B twice. Mg^A binds to the site occupied by Mn²⁺ or Ca²⁺ in the absence of DNA (11) (Fig. S1A), and is also coordinated by Glu129, the backbone carbonyl of Leu130, and two water molecules, one of which is positioned for in-line attack on the scissile phosphate (Fig. 2A). Mg^B does not bind in the absence of DNA and is also coordinated by three water molecules. Glu85, a residue that is conserved as Glu or His in related enzymes (Fig. S4), does not contact either Mg²⁺ ion directly, but forms an outer sphere ligand interaction to each of them, through bound water molecules. Two other highly conserved residues, Gln157 and Tyr154, form hydrogen bonds with the scissile phosphate and the neighboring 3′ phosphate, respectively. By analogy with MutH and other two-metal nucleases (13, 24), the structure supports a mechanism in which Mg^A, together with the neighboring 3′ phosphate and Lys131, activates a water molecule for nucleophilic attack, whereas Mg^B stabilizes the trigonal bipyramidal geometry of the transition state through its bidentate coordination to the scissile phosphate.

The 5′ phosphate of the DNA binds to a positively charged pocket buried at the end of the active site, where it contacts the guanidino group of Arg28, three backbone amides, two of which are at the N terminus of helix C, three Ser/Thr hydroxyls, and the edge of the indole ring of Trp24 (Fig. 2 D and E). Binding of the 5′ phosphate to this site was predicted from the binding of free phosphate in the uncomplexed structure (11), and the R28A protein has substantially reduced activity and processivity (22). Based on these observations and on the fact that the duplex with 5′-OH ends binds in a nonproductive manner, attraction of the 5′ phosphate to this positively charged pocket appears to be essential for pulling the DNA substrate fully into the active site. All of the residues that contact the 5′ phosphate are conserved in alkaline nucleases of Kaposi’s sarcoma-associated herpesvirus (KSHV; ref. 17) and EBV (18), suggesting that 5′-phosphate recognition is an important feature of this group of 5′-3′ exonucleases (Fig. S4).

The occupancy of the active sites of subunits A and C on the trimer, which are not bound by DNA, is intriguing (Fig. S5). Neither active site contains Mg²⁺, which is present at 5 mM in the crystal, suggesting that both Mg^A and Mg^B require DNA for tight binding. The active site of subunit C contains a phosphate ion at the same site as the 5′ phosphate of the DNA, as was seen in the uncomplexed structure (11). The active site of the third subunit, subunit A, also contains a phosphate, but at a position that is shifted by 8.5 Å toward the back of the trimer, near a small rear portal through which cleaved mononucleotides could be released (Fig. 1E and Fig. S5C). Movement of the phosphate is accompanied by a reorientation of the side chain of Arg28, which remains bound to it. At the expected phosphate-binding site on subunit A, near the N terminus of helix C, is a pair of chloride ions, spaced by 2.6 Å from one another. The presence of these chloride ions highlights the positively charged nature of this pocket on the enzyme, even when the guanidino group of Arg28 is displaced. Although it is not clear from the structure why the liganded states of subunits A and C are different, it is interesting to speculate that movement of Arg28 and the phosphate of subunit A could reflect conformational changes, possibly coupled to the catalytic cycle, that are involved in ejection of cleaved mononucleotides out of the active site during processive digestion.

Structure-Activity Analysis.

To probe the roles of key residues seen in the structure, the effects of alanine substitutions at 11 different residues on exonuclease activity were determined (Fig. 3). The targeted sites include residues within the central channel that contact the downstream portion of the DNA (R45, K49, M53, K76, R137), the hydrophobic wedge (L78), the active site (E85, D119, K131), and the 5′-phosphate-binding site (W24, R28). To measure exonuclease activity, digestion of a linear 2.7 kB pUC19 DNA by a limiting amount of enzyme was monitored by agarose gel electrophoresis with SYBR gold staining. This method allows for simultaneous visualization of the dsDNA substrate and ssDNA product of the reaction. At a concentration of 0.05 nM trimer, wild-type λ exonuclease digests one strand of 1.2 nM of duplex within 40 min to release the ssDNA product. Based on this result, we estimate a rate of digestion of approximately 30 nucleotides per second for wild-type enzyme under these conditions, which is comparable to previous measurements (21, 22, 28, 29). The reaction appears to be completely processive because there are no intermediate bands, and the full amount of ssDNA product is released.

Fig. 3.

Structure-activity analysis. (A) Structure of the trimer showing the mutated residues, which are colored red for DNA binding, blue for hydrophobic wedge, orange for active site, and black for 5′ phosphate. (B) Exonuclease assays. The indicated concentration of WT or mutant enzyme was incubated with 1.2 nM of linear pUC19 DNA at 37 °C. Aliquots removed at the indicated times were analyzed by agarose gel electrophoresis and SYBR gold staining. Control lanes labeled “ds” and “ss” mark the substrate and product of the reaction, respectively. The reactions are sorted by decreasing activity.

The E85A, D119A, and K131A proteins had no detectable activity (or barely detectable for E85A), consistent with the prominent roles of these highly conserved active site residues in the two-metal cleavage mechanism. Mutation of two residues that contact the 5′ phosphate of the DNA resulted in no activity (R28A) or very little activity (W24A), indicating that these interactions are also important for catalysis. As for the residues contacting the downstream portion of the DNA, the R45A protein had no detectable activity, consistent with the prominent role of this residue in binding to the minor groove as seen in the structure. By contrast, the M53A, K49A, and K76A substitutions, which occur at residues that make more subtle contacts with the DNA, have only modest effects on exonuclease activity. Curiously, the R137A protein had no detectable activity. The importance of Arg137, a nonconserved residue that contacts the fifth phosphate from the 5′ end of the DNA (ca. 16 Å from Mg^A), is not apparent from the structure. Finally, the L78A protein exhibited a significantly diminished level of activity, consistent with the prominent role of this residue in forming the hydrophobic wedge of the DE loop. Reactions of L78A also resulted in the appearance of intermediate bands, evident as reproducible streaking, apparently due to the production of partially processed DNA molecules. Such streaking was also seen for K76A, but not for WT or any of the other mutants.

We also examined the exonuclease activities of the 11 protein variants using a continuous assay in which excess enzyme (10 nM trimer) was added to 0.05 nM of linear pUC19 DNA, and digestion was monitored by the decrease in fluorescence of PicoGreen (Fig. S6). An apparent rate of the reaction can be calculated from the initial slope of the decrease in fluorescence. Although the rates are lower in this assay due to the lower pH used for PicoGreen (7.5 vs. 9.4), the relative rates of the protein variants were similar to what was observed in the gel-based assay.

Discussion

We report here two crystal structures of λ exonuclease in complex with DNA. A question from the previous structure of the uncomplexed enzyme was, if the DNA binds within the central channel as proposed, how does the 5′ end of the DNA access one of the three active sites, which lie approximately 15 Å from the central channel? The structure of the complex shows that binding of the 5' end to the active site is accomplished by tilting of the DNA relative to the central axis of the trimer, combined with unwinding to thread two nucleotides of the 5′-ended strand into the active site cleft. Although the DNA in the crystal contains a 5′-dinucleotide overhang and was not actually unwound to form the complex, because the 5′ and 3′ termini are on opposite sides of the DE loop, it is clear from the structure that the enzyme must unwind exactly two base pairs prior to cleavage. Due to the DNA used for crystallization, the structure does not show interactions that might occur with the extruded 3′-ended strand. However, as the rear surface of the trimer is rather flat, such interactions are not likely to be extensive. Unwinding of the DNA is facilitated by four apolar residues of the DE loop that wedge into the base pairs to form favorable hydrophobic interactions. A similar topological feature, which is often flexible or disordered, is found in related enzymes that resect dsDNA, including RecE protein and the alkaline nucleases of KSHV and EBV, but not in RecB protein, which acts on ssDNA. In addition, the key Leu78 residue of the DE loop that wedges into the bases appears to be conserved as leucine or valine in the other three proteins (Fig. S4).

The previous structure of λ exonuclease suggested a mechanism for processivity based on topological linkage of the trimer to the DNA, but how or if the enzyme couples energy from phosphodiester bond cleavage to forward movement along the DNA remained an open question. Myers and coworkers noted that a phosphate bound to Arg28 at a pocket near the active site could mark the binding site for the terminal 5′ phosphate of the DNA, and observed that the R28A substitution disrupted processivity of the enzyme (22). Because a 5′ phosphate is required for normal levels of activity, these observations suggested that binding of the 5′ phosphate to the pocket formed by Arg28 is required for correct positioning of the DNA within the active site. The structure with the 5′-phosphorylated duplex confirms that the 5′ phosphate binds to this pocket on the enzyme precisely as predicted. Moreover, the structure with the 5′-OH terminated duplex reveals that without a 5′ phosphate, a nonproductive complex is formed in which the DNA is bound within the central channel, but not unwound and inserted into the active site cleft. All together, these data suggest that attraction of the 5′ phosphate to the positively charged pocket is required to drive unwinding of the DNA to pull the 5′-ended strand into the active site to form the productive complex. Interestingly, from a recent structure of Xrn1, an unrelated enzyme that digests messenger RNA in the 5′ to 3′ direction, it was proposed that attraction of the 5′ phosphate to a similar positively charged pocket is critical for its mechanism of processivity (30).

Based on these data, we propose an electrostatic ratchet mechanism for processivity of λ exonuclease (Fig. 4). In step 1, initial capture of the DNA end results in formation of the complex seen in the structure with the 5′-OH terminated duplex, in which the DNA is bound within the central channel but not fully inserted into the active site. In step 2, attraction of the 5′ phosphate to the positively charged pocket, along with insertion of the hydrophobic wedge into the base pairs, drives unwinding of the DNA to form the productive complex seen in the structure with the 5′-phosphorylated duplex. Although our data do not establish that steps 1 and 2 are truly distinct from one another, the availability of structures for each provides a framework to test this possibility. Upon binding of the Mg²⁺ ions and correct alignment of the hydrolytic water molecule, the scissile bond is cleaved (step 3), and the resulting 5′ mononucleotide, together with one or both Mg²⁺ ions, released through the small rear portal (step 4). Release of the cleaved mononucleotide would result in an intermediate in which the first nucleotide-binding site is empty, and the second site is occupied by the next nucleotide to be cleaved, which now has a newly exposed 5′ phosphate. At this point, the enzyme could in principle move forward or backward along the DNA. The observation that it is two nucleotides of the DNA that are unwound by the enzyme, instead of just the one that is reacted on, is suggestive, because after cleavage of the terminal nucleotide, the DNA would still be unwound by one base pair, such that interactions between the hydrophobic wedge of the DE loop and the bases of the DNA would be maintained. These interactions, in particular the insertion of Leu78 between the second and third bases of the 5′-ended strand, could help to keep a tight grip on the DNA, restricting backward movement. Although attraction of the 5′ phosphate of the next nucleotide to be cleaved to the positively charged pocket at the end of the active site could provide a significant force to pull the DNA forward, there is no such force apparent for backward movement. After translocation of the enzyme forward along the DNA to position the next nucleotide to be cleaved within the active site (step 5), repeated cycles of steps 3–5 would lead to processive digestion.

Fig. 4.

Electrostatic ratchet mechanism for processivity of λ exonuclease. See Discussion for a description of the different steps. The 5′-phosphate group on the DNA is represented by an asterisk (*). The positively charged pocket at the end of the active site cleft is represented by the two plus (+) signs.

The structure-activity analysis largely supports this model. The key roles of Arg45 and Arg28 in binding to the minor groove and the 5′ phosphate, respectively, are confirmed, because the R45A and R28A substitutions essentially abolish exonuclease activity. The L78A protein also has significantly diminished activity, consistent with the key role proposed for this residue in forming the hydrophobic wedge that unwinds the DNA and restricts backward movement. Although the L78A substitution does not totally abolish activity, alanine at this position does retain some apolar character, and there are three other apolar residues of the DE loop that also help to form the hydrophobic wedge (Val73, Ala75, and Ala77). The presence of intermediate bands for the L78A protein suggests a possible defect in processivity, although further studies will be necessary to establish this definitively. That Leu78 appears to be conserved in RecE and alkaline nucleases of KSHV and EBV further suggests a key role for this residue in the reaction. Although Arg28 is also conserved in the latter two enzymes (but not in RecE), Arg45, and the extended CD loop on which it resides, is apparently unique to λ exonuclease (Fig. S4).

According to this model, forward movement of the enzyme along the DNA is driven, at least in part, by electrostatic attraction of the 5′ phosphate to a positively charged pocket at the end of the active site cleft. How substantial is the electrostatic force between the positively charged pocket and the 5′ phosphate likely to be? In the “intermediate” complex depicted in Fig. 4, the distance between the positively charged pocket and the 5′ phosphate of the next nucleotide to be cleaved is approximately 7 Å, which may seem too large to be significant. However, as the attraction involves at least two units of opposite charge, and the active site pocket is largely buried, the electrostatic binding energy is likely to be significant. Although electrostatic fields in proteins are difficult to compute, the presence of two adjacent chloride ions and one phosphate in the active site of subunit A suggests that this pocket on the enzyme could carry as much as four units of positive charge. If one or both of the Mg²⁺ ions dissociate from the active site after cleavage, as suggested by their absence in subunits A and C, electrostatic repulsion of the 5′ phosphate with the three active site carboxylates could destabilize the intermediate to further propel the DNA forward. Binding of the hydrophobic wedge to the single/double-strand junction, or interactions of the terminal bases with one another or with Trp24, could also play a role in forward movement.

It is conceivable that the enzyme could also possess a motor activity in which energy released from cleavage of the high-energy phosphodiester bond, estimated to be -5.3 kcal/mol (31), drives conformational changes that are coupled to movement of the enzyme along the DNA. However, there are no large-scale conformational changes in the trimer, such as subunit reorientations, that occur upon DNA-binding. Thus, the available data do not support a motor activity for λ exonuclease that would be analogous to other ring-shaped enzymes, such as hexameric helicases, in which binding and hydrolysis of ATP induces large reorientations of the subunits (32). There are, however, some small local changes in the subunit to which the DNA is bound, most notably in the DE loop that wedges into the DNA (Fig. S5D), and in Lys51, which flips up to contact the fourth phosphate on the 5′-ended strand (Fig. S5A). We also observe a significant reorientation of the side chain of Arg28 of subunit A, together with movement of its bound phosphate ion. It is interesting to speculate that these changes could be coupled to the catalytic cycle, to eject cleaved mononucleotides out of the active site through the rear portal. We also note that the present structure does not explain the complete inactivity of the R137A protein, as Arg137 makes only a loose contact (4.3 Å) to the fifth phosphate of the 5′-ended strand. Thus, there may be as yet unobserved states of the enzyme that could be involved in the catalytic cycle.

Single molecule studies of λ exonuclease measured similar average rates of digestion, but led to different conclusions regarding the sequence dependence of the digestion rate (28, 29). One study observed that the local rate of digestion correlated with the energy required to unwind the duplex and concluded that melting of the terminal nucleotide to position it within the active site is the rate-limiting step of the catalytic cycle (28). Our structure supports this result and reveals that it is in fact two base pairs that are melted from the duplex prior to cleavage. A second study observed pausing of the enzyme at particular GGCGA sequences on the DNA, and it was suggested that this sequence could be recognized as ssDNA, possibly through stacking interactions of the guanine bases with aromatic residues of the active site cleft (29). Such a mechanism for pause-site recognition is not explained by the current structure, because only two nucleotides of the DNA are unwound and there are no interactions with the terminal nucleotides that would appear to favor the GGCGA sequence. However, the insertion of Arg45 into the minor groove of the downstream portion of the DNA could provide a means for sequence-dependent pausing, as this interaction favors A-tract sequences in which the minor groove is narrowed (23). The AT-rich sequence of the right half of the most prominent pause site (GGCGATTCT), coupled with the higher energetic barrier to unwinding the GC-rich sequence on the left half, could provide a structural basis for pausing. It is conceivable, however, that pause-site recognition occurs through an off-pathway structure that is substantially different from the one reported here. Along these lines, a structure of the enzyme in complex with DNA containing the pause sequence would be illuminating.

Given that λ exonuclease and the related RecE exonuclease (15) both form toroidal oligomers with multiple active sites symmetrically disposed about a central channel, an interesting question emerges: Do these enzymes use a sequential mechanism in which the DNA moves cyclically from one active site to the next for each round of cleavage, or does the DNA engage with a single active site on the oligomer for multiple rounds of cleavage? Our electrostatic ratchet mechanism posits that processivity operates not only at the level of the trimer, but also at the level of the monomer. Thus, we predict that despite its symmetric architecture, the enzyme does not use a sequential type of mechanism such as that proposed for hexameric helicases (32, 33), F₁-ATPase (34), and other toroidal oligomers.

Materials and Methods

The λ exonuclease proteins were expressed in E. coli as N-terminal 6His fusions and purified by nickel affinity and anion exchange chromatography. Site directed mutagenesis was performed by the QuikChange procedure (Stratagene) and mutations confirmed by DNA sequencing and mass spectrometry of the purified proteins. Oligonucleotides used for crystallization were purchased HPLC-purified from Integrated DNA Technologies. For crystallization, 10 mg/mL λ exonuclease (wild-type or K131A variant) was mixed with a 1.1 M excess of DNA duplex in the presence of either 5 mM CaCl₂ (Form 1) or 5 mM MgCl₂ (Form 2), and crystallized by hanging-drop vapor diffusion. X-ray diffraction data were collected on flash-frozen crystals at the Advanced Photon Source. Structures were determined by molecular replacement. Exonuclease activities were determined by incubating a limiting amount of enzyme at 37 °C with linear pUC19 dsDNA and analyzing the reaction products by agarose gel electrophoresis with SYBR gold staining, or by incubating excess enzyme with linear pUC19 and PicoGreen and monitoring the decrease in fluorescence.

Additional methodological details are reported as SI Text.