Structural basis for the recognition and cleavage of abasic DNA in Neisseria meningitidis

Abstract

Base excision repair (BER) is a highly conserved DNA repair pathway throughout all kingdoms from bacteria to humans. Whereas several enzymes are required to complete the multistep repair process of damaged bases, apurinic-apyrimidic (AP) endonucleases play an essential role in enabling the repair process by recognizing intermediary abasic sites cleaving the phosphodiester backbone 5′ to the abasic site. Despite extensive study, there is no structure of a bacterial AP endonuclease bound to substrate DNA. Furthermore, the structural mechanism for AP-site cleavage is incomplete. Here we report a detailed structural and biochemical study of the AP endonuclease from Neisseria meningitidis that has allowed us to capture structural intermediates providing more complete snapshots of the catalytic mechanism. Our data reveal subtle differences in AP-site recognition and kinetics between the human and bacterial enzymes that may reflect different evolutionary pressures.

Abasic (apurinic-apyrimidic, AP) sites within DNA pose a serious threat to all living organisms; if unrepaired, they lead to stalled replication, increased mutations, and an overall loss of genomic integrity. For example human cells defective in AP-site processing are nonviable (1). AP sites occur through either the spontaneous loss of nucleotide bases via chemical breakage of the N-glycosylic bond or as intermediates in the repair of DNA lesions such as oxidatively damaged bases, spontaneously deaminated cytosines, or alkylation of bases (2). It is estimated that 10⁴ abasic sites are produced per day per cell in eukaryotes and if left unrepaired will result in significant mutagenic loading on the cell (3, 4). All living organisms have therefore evolved specific enzymes that recognize and repair damaged or missing DNA bases known as the base excision repair (BER) pathway (2).

In BER, DNA glycosylases recognize and remove damaged bases to produce abasic sites. The resultant abasic sites are then recognized by AP endonucleases that bind and catalyze a magnesium-dependent 5′ cleavage of the phosphodiester backbone, producing a 3′-OH. Other bifunctional DNA glycosylases cleave the resulting abasic site via an AP lyase mechanism, which leaves a 3′ unsaturated aldehyde (5). In either case, DNA backbone cleavage marks the AP site for subsequent repair by DNA polymerases and ligases. Two main families of AP endonucleases have been identified: the major human AP endonuclease (APEX1) and Escherichia coli homolog ExoIII belong to the ExoIII family of AP endonucleases and are highly conserved (∼51% sequence homology; ∼29% sequence identity). The other is the EndoIV family, named after the E. coli enzyme, and they cleave abasic sites in a Zn²⁺-dependent reaction. Both E. coli and Saccharomyces cerevisiae possess both ExoIII and EndoIV type enzymes, whereas humans have two ExoIII family enzymes.

Molecular and biochemical details of abasic DNA recognition and backbone cleavage have mainly focused on the human APEX1 enzyme and bacterial homolog ExoIII. Several crystal structures of AP endonucleases have been determined, including the human and several bacterial enzymes (6–11). To date, molecular and mechanistic details of abasic DNA recognition and backbone cleavage have been limited to the human enzyme APEX1, where crystal structures of APEX1–DNA complexes (product and noncleaved complexes) have been determined (7). A comparison of the noncleaved (minus metal) and product (cleaved plus metal) APEX1 complexes suggested a mechanism that involves a catalytic aspartate activating a water molecule to attack the phosphoryl group. However, no structure for a APEX1–DNA precleaved metal complex was reported and therefore more complete structural snapshots for the AP endonuclease catalytic mechanism are not yet available. Furthermore, there is no structural information for a bacterial AP endonuclease bound to its substrate or product DNA.

Neisseria meningitidis is an important human pathogen and is the leading cause of meningitis and sepsis (12). The bacterium is an obligate pathogen and is exquisitely adapted for survival in the human nasopharynx (13). During colonization and disease, the meningococcus is subject to constant attack from reactive oxidative species, generated by respiratory epithelial and phagocytic cells as a component of the human innate immune system. This distinct lifestyle suggests that it is likely to have evolved strategies to avoid oxidative stress that are distinct from those in other model organisms like E. coli, and which promote its adaptation for survival. Previously, we found that N. meningitidis, like its human host, possess two ExoIII family AP endonucleases, NExo and NApe. NExo has greater sequence identity to the E. coli enzyme ExoIII, whereas NApe is more like APEX1 at both the level of sequence identity and function: of the two neisserial enzymes, only NApe posseses Mg²⁺-dependent AP endonuclease activity (8). We showed that NApe is necessary for survival of N. meningitidis under oxidative stress and during bloodstream infection (8). In contrast, we found that the closely related NExo, although structurally similar, does not possess AP endonuclease activity, but is a specialized 3′-phosphatase (14).

Here we describe the high-resolution crystal structure of NApe bound to DNA duplexes containing an AP analog (tetrahydrofuran; THF) in the presence and absence of manganese, representing unique structures of a prokaryotic AP endonuclease bound to substrate DNA. We have captured structural snapshots of the catalytic mechanism, which allows us to propose a detailed structure-based mechanism. Surprisingly, our structures reveal that the neisserial NApe enzyme recognizes substrate DNA in a completely different mode to the human APEX1 enzyme, despite both enzymes having near identical folds.

Results and Discussion

Structure of the NApe AP–DNA Complex and Comparison with the Human APEX1 Complex.

We have determined a number of crystal structures of NApe bound to two 11-bp DNA fragments containing an AP-site analog (tetrahydrofuran, THF; referred throughout text as “abasic”) adjacent to a 3′G (2.0 Å) or 3′C (2.75 Å) (Table S1). The only difference between both structures is in the 3′G structure; the paired cytosine is in two conformations, the reasons for which are unclear. To avoid this complication, we used DNA containing a 3′C adjacent to the AP site for all subsequent structures. We also used a similar DNA sequence and length to previous studies of the human enzyme (7) to allow direct structural comparisons.

NApe folds into the four-layered α/β conformation similar to previous crystal structures and other ExoIII family AP endonucleases including APEX1 (Fig. 1A) (6, 8, 10). A comparison between DNA bound and unbound NApe shows no large-scale conformational changes (rmsd = 0.3 Å for Cα atoms). The most significant changes occur at two loops comprising residues 114–116 and 206–210. These loops “pinch” the abasic site from opposite sides of the duplex DNA (Fig. 1 A and B). Loop 114–116 is positioned near the DNA major groove, making contact with the 5′ phosphate adjacent to the AP site but no direct base interactions (Fig. 1B). Loop 206–210 points toward the DNA minor groove and stabilizes the orphan base opposite the abasic site through direct hydrogen bonding (see below). The abasic containing phosphate backbone is intact and the ring of the abasic analog residue is flipped into a hydrophobic pocket (Fig. 2A). As a result, the DNA backbone is bent ∼42° away from the protein relative to standardized B-form DNA. The DNA backbone of the abasic strand forms direct interactions with NApe, whereas on the opposite strand only the backbone 3′ to the abasic site is contacted (Fig. 2B). Additionally a positive electrostatic surface potential provides a favorable interaction surface for duplex DNA (Fig. 2C). In APEX1, similar loops bind the abasic site from opposite sides, but use a different binding mode to stabilize the abasic site, resulting in a different conformation of the DNA substrate (Fig. 1C). R177 forms base stacking interactions that compensate for the missing base with M270 also forming base stacking interactions that stabilize the orphan base (Fig. 1D). A comparison between the NApe and APEX1–DNA complexes also reveals subtle changes in the relative orientations of the recognition loop helix due to the differing modes of abasic site binding and stabilization (Fig. 1E; rmsd = 1.3Å Cα atoms; DNA recognition loops rmsd = 2.5 Å Cα atoms).

Fig. 1.

Comparison of crystal structures of the N. meningitidis NApe and human APEX1 AP endonucleases bound to AP DNA substrate. (A) NApe–DNA complex with surface representation of DNA (orange) and ribbon representation of NApe (green). The AP site is colored red with the two AP-site recognition loops colored lime. (B) Blow-up of the AP-site recognition loops in the NApe–DNA complex. Key amino acids are shown as labeled and as ball and stick. (C) APEX1–DNA complex (PDB ID code 1DE8) with surface representation of DNA (purple) and ribbon representation of APEX1 (blue). The AP site is colored yellow with the two AP-site recognition loops colored cyan. (D) Blow-up of the AP-site recognition loops in the APEX1–DNA complex. Key amino acids are shown as labeled and as ball and stick. (E) Overlay of DNA-bound complexes of NApe and APEX1. Coloring is as in A and C.

Fig. 2.

Specific features of DNA AP substrate recognition by NApe. (A) The AP deoxyribose is flipped out of the duplex DNA into a hydrophobic pocket comprising W204, I220, and W218. (B) Schematic showing the specific interactions between NApe and DNA AP substrate. Direct DNA interactions by specific side chains are labeled and shown as blue lines. Stacking interaction of R208 is shown as an oval (green). (C) Electrostatic surface potential showing a large positively charged (blue) DNA interaction surface with the active site pocket in the center of the surface.

Structure of NApe–DNA Prehydrolysis Complex.

To capture a prehydrolysis complex without cleavage of the DNA backbone, we carried out cocrystallizations in the absence of divalent cations, which are essential for DNA cleavage activity (15). We also used a catalytically inactive mutant D149N to capture a precleavage DNA substrate with an intact phosphodiester backbone but in the presence of a divalent cation (manganese). In the prehydrolysis complex we observe the scissile phosphate (O5* and O2P) bound to S112 and Y109 with the O1P atom indirectly bound via three water-mediated interactions (Fig. 3A). A water molecule (Wat3) forms hydrogen bonds with N151, D149, and H247 and is located opposite the O3* leaving group of the scissile phosphate (Fig. 3A). By using a catalytically inactive mutant, D149N, we were able to cocrystallize a precleavage DNA substrate in the presence of manganese. In this structure we observe the manganese coordinated through interactions with N10, E36, and the O1P of the scissile phosphate (Fig. 3B). Compared with the metal-free complex, the O1P of the phosphate moves 0.8 Å toward the manganese, and is in a suitable position for in-line (S_N2) attack by the Wat3 water molecule (Fig. 3C). The repositioned phosphate also results in the O5* forming an additional hydrogen bond with N151. The remaining parts of DNA and protein are near identical between both structures (rmsd = 0.7 Å, all atoms). It is notable that there are no direct contacts between the phosphate and side chains of D149 or H247, which is different from the human enzyme.

Fig. 3.

Structural details of the DNA AP substrate complexes and structure-based mechanism for phosphodiester bond cleavage by NApe. Hydrogen bonds and metal ion contacts are shown in dashed lines. Amino acid side chains are labeled in A. Water molecules are shown as spheres colored cyan and numbered, and the manganese ion appears as a purple sphere. (A) Wild-type NApe bound to AP site in absence of metal ion. Scissile phosphate is bound by hydrogen bonding network that involves three bound water molecules. (B) NApeD149N bound to AP site in presence of a manganese (labeled) and intact phosphodiester bond. The N149 side chain is labeled. Manganese is bound by E36, N10, D246, and the O1P of the scissile phosphate. (C) Overlay of the metal-free and metal-bound structures. The scissile phosphate moves 0.8 Å toward the manganese allowing in-line S_N2 attack by an activated water molecule (Wat3). (D) NApe–DNA product complex with a cleaved phosphodiester bond and bound manganese (purple sphere). (E) Overlay of metal-bound NApe DNA substrate (B) and product (D) complexes showing the different conformations for the 3′ and 5′ groups of the cleaved DNA product. (F) Overlay of metal-free NApe DNA substrate (A) and metal-bound product complexes (D). Note the colocation of the proposed attacking water molecule in the metal-free complex and the O3P in product complex. (G) Schematic shows the DNA AP substrate colored red, divalent metal ion as a green sphere, and the NApe side chains as sticks colored black. Curly arrows indicate the flow of electrons in the reaction mechanism. The AP substrate is bound by NApe such that scissile phosphate forms interactions with the bound divalent metal, Y109 and N151. The side chains of D149, H247, and D221 act as a proton transfer network to facilitate proton abstraction and activation of bound water for in-line S_N2 attack of the phosphoryl group. The pentacovalent transition state (Center) collapses and leads to cleavage of the scissile P-O3′ bond and inversion of the 5′ phosphate configuration. The transition state and O3′ leaving group are stabilized by the metal ion and further bound water.

Structure of NApe–DNA Product Complex.

To capture the structure of a cleaved DNA product complex, we performed in situ cleavage by soaking crystals of the metal-free complex with manganese before data collection. The final structure was refined to 2.5 Å resolution and a comparison with the metal-free complex revealed no overall conformational change (rmsd = 0.7 Å, all atoms). The largest changes are around the scissile phosphate where the covalent bond between the phosphor atom and O3* of the adjacent 5′ nucleotide is not observed, suggesting that the DNA substrate has been cleaved within the crystal (Fig. 3D). In the cleaved structure, the O1P is shifted 0.6 Å toward H247, forming interactions with H247 and the manganese (Fig. 3 D and E). The O5* is no longer hydrogen bonded to S112, but retains an interaction with N151. At the cleavage site, the released 3′ hydroxyl shifts 2.7 Å and makes direct contact with the manganese (Fig. 3 D and E). Strikingly, the position of the water (Wat3) in the metal-free structure is replaced by O3P of the new phosphate group, and is hydrogen bonded to D149 and H247 (Fig. 3F).

Structure-Based Catalytic Mechanism for NApe.

By capturing structural snapshots of the prehydrolysis substrate and product complexes in the presence of divalent metal ion, we are able to propose a detailed structure-based mechanism for NApe and general AP endonucleases (Fig. 3G). In our proposed mechanism, we identify the attacking water (Wat3), based on the striking observation that this water is ideally positioned for in-line attack and occupies the same position as the O3P in the newly formed 5′ phosphate of the product (Fig. 3F). We propose that the activation of this water to provide a nucleophile for in-line S_N2 attack is mediated by direct hydrogen bonding to D149 and H247, which is in turn hydrogen bonded to D221. The direct hydrogen bonding to D149 will activate the water, whereas the extended hydrogen-bonding network of H247 and D221 would provide the necessary chemical environment to facilitate protein abstraction via a proton relay system mediated by H247. N151, which is also hydrogen bonded to the attacking water, positions the water for activation.

The role of the metal ion in the precatalytic complex is to orientate and polarize the scissile phosphate group by binding to the O1P atom. By modeling a square bipyramidal coordination around the metal, a water molecule in the first hydration shell is observed in contact with the O3* leaving group. The metal ion is thus well positioned to stabilize the increased negative charge around the phosphate center in the transition state, and also the transfer of this charge onto the O3* leaving group. Upon cleavage, the O3* is seen to move 2.7 Å, and is directly coordinated by the metal ion, supporting the role of the metal ion in leaving group stabilization (Fig. 3G).

It has previously been proposed that under certain conditions (neutral pH), a second metal binding site (Pb²⁺) can be observed in APEX1 (16). A comparison of this structure (in the absence of DNA substrates) with our NApe:DNA structures (crystallized at neutral pH), reveals that the proposed second metal ion binding site colocates with the position of the attacking water (Wat3) in NApe. Whereas it is formally possible that the water we assign as the nucleophile is a bound sodium ion from the crystallization conditions, we think this is unlikely for the following reasons. (i) If the water represented a second metal binding site then it would not be possible to place a water molecule in the correct geometry for scissile phosphate cleavage without a significant repositioning of the phosphodiester backbone. (ii) A water molecule in this position is perfectly oriented for in-line S_N2 attack of the scissile phosphate. (iii) This atomic position is replaced by one of the oxygen atoms in the newly formed phosphate group in the product complex. (iv) We observe an ordered water molecule 1 Å from this position in the non–DNA-bound complex (PDB ID code 2JC5). Taken together we believe that this provides a strong basis for interpreting the bound water (Wat3) as the nucleophile.

To further elucidate the role of the metal ion in the catalytic mechanism, we also performed a metal ion titration with Ca²⁺. Calcium typically acts as an inhibitor of enzymes in phosphodiester hydrolysis, because it has similar coordination chemistry to Mg²⁺, but its larger ionic radius and reduced electronegativity mean that it is unable to support catalysis (17). Titrations of different metal ions can lead to synergistic effects, which have been interpreted as the requirement for two metal ions in catalysis (16, 18, 19). We performed reactions of NApe with varying ratios of Mg²⁺ and Ca²⁺, while maintaining a constant chloride concentration to minimize ionic strength effects (Fig. S1A). It can be seen that in the absence of Mg²⁺ there is no reaction, consistent with Ca²⁺ being unable to support catalysis. As Mg²⁺ concentration increases, there is a linear increase in the observed rate, with no apparent synergism between the metal ions. Even when synergistic effects are observed, their interpretation is problematic because they may be caused by metal ions binding at positions other than the active site, leading to perturbations through ionic or allosteric effects. The importance of ionic strength can be seen from further metal ion titrations, which demonstrate the sensitivity of the enzyme to ionic strength and the inhibitory effect of Ca²⁺, again with no apparent synergy (Fig. S1 B and C). Consequently our metal ion titrations do not provide any evidence to support a two-metal ion mechanism.

Orphan Base Recognition by NApe Is Different from Human APEX1.

In the NApe–DNA bound structures, a loop comprising residues 206–210 intercalates into the double helix from the minor groove to form interactions with the orphan base opposite the abasic site (Fig. 4A). For guanine and adenine R208 stacks between the orphan base and its 5′ adjacent base pair (Fig. 4 A and B). In contrast, N207 forms two hydrogen bonds with the N2 and N3 of a guanine orphan base (Fig. 4A) but does not hydrogen bond to adenine (Fig. 4B). For a cytosine orphan base, R208 forms a direct hydrogen bond with the O2 atom of the cytosine and stacks with the 5′ base pair adjacent to the abasic site (Fig. 4C). N207 forms no interactions with the orphan base (Fig. 4C). To test whether differing orphan bases affect the binding affinities for a NApe–DNA substrate complex, we performed fluorescent anisotropy binding assays (Fig. S2). This clearly demonstrated that the identity of the orphan base did not affect the binding affinity of NApe for substrate DNA.

Fig. 4.

Recognition of different orphan bases by NApe. Recognition loop for NApe (206–210) is colored green with R206 shown as a ball and stick and Van der Waals cage (A–C). The orphan base is labeled and the DNA phosphates are colored orange. Dashed line indicates a stacking interaction and a solid line, a direct hydrogen bonding interaction. (A) NApe–DNA complex with guanine as the orphan base. (B) NApe–DNA complex with adenine as the orphan base. Note that N207 points away from the orphan base and makes nonspecific contacts. (C) NApe–DNA complex with cytosine as the orphan base. Note that R208 forms a specific hydrogen bond with the orphan base as well as a stacking interaction with base 3′ to the AP site. N207 points away from the orphan base in a similar position to that in B. (D) APEX1–DNA complex (PDB ID code 1DE8) showing similar interactions irrespective of the orphan base. Two residues from each recognition loop (R177 and M270) form stacking interactions with the base 3′ to the orphan base and 3′ to the AP site. Y269 is the equivalent residue to N207 and forms interactions with the metal binding residue D70. This results in a different conformation for the DNA substrate and subsequent positioning of the scissile phosphate in the active site.

The subtle amino acid changes in the substrate recognition loops leads to a surprising and unexpected difference in how bacteria and human enzymes recognize AP-containing DNA. Unlike APEX1, NApe recognizes the AP-containing DNA substrate with surprising plasticity (from sequence specific to nonspecific) depending on the nature of the orphan base opposite the AP site. In contrast, APEX1 forms no specific orphan base interactions and accommodates all four bases in a similar manner (7). It is interesting to speculate whether the specific recognition of a guanine orphan base in NApe-like bacterial AP endonuclease is an evolutionary remnant that is related to the frequency of cytosine deamination in prokaryotes, although there is no evidence to support this. Such differences will require further studies to determine whether the subtle difference in AP-site recognition are related to the efficiency of BER in vivo.

Kinetic Comparison of the Human APEX1 with NApe.

Given the high level of structural similarity between the human homolog APEX1 and NApe, we were interested in comparing their enzymatic activities. We used a FRET-based fluorescent reporter assay with labeled DNA hairpin containing an abasic site analog, as well as a gel-based assay (Fig. S3). Surprisingly, APEX1 showed a 100 times greater k_cat than NApe. The value of K_M differed by a similar degree, between the two enzymes, so that the k_cat/K_M remained comparable between the two enzymes (Fig. 5 A and B). This intriguing observation led us to consider what structural elements are responsible for the observed differences in enzyme kinetics. We therefore tested the catalytic role of the AP recognition loops in NApe, focusing on N207 and R208. Unexpectedly, both Ala mutants showed significantly higher activities than wild type with k_cat for N207A ∼3 times higher than wild type, and R208A ∼30 times higher. Similar increases in K_M were also observed for both mutants, which again maintained similar values of k_cat/K_M to wild-type NApe (Fig. 5A and Table S2). Both loop mutants are far removed from the active site, but intercalate into the abasic DNA and form interactions with the orphaned base (Fig. 4). We therefore tested the interactions of both loop mutants with abasic DNA substrates in the absence of magnesium to avoid cleavage. Anisotropy binding experiments demonstrated that neither mutant had a significant difference in binding affinity compared with wild type (Fig. 5C). The observed catalytic differences between NApe and APEX1 are intriguing and it will be interesting to determine the nature of these differences in relation to genome size and complexity, cellular metabolism, enzyme concentration, and DNA base damage rates in vivo.

Fig. 5.

NApe enzyme activity and AP substrate binding. (A and B) Steady-state reactions were performed with an abasic DNA hairpin substrate and cleavage monitored by fluorescence (Materials and Methods). Data are plotted on linear (A) and log (B) scales for comparison and are shown with the best fit to the Michaelis–Menten equation with the values listed in Table S2. (C) DNA binding to abasic substrate DNA was monitored by fluorescence anisotropy in the absence of divalent metal ions under standard binding conditions (Materials and Methods); data are shown with the best fit to the tight binding equation with the parameters listed in Table S2. (D) Dissociation of the enzyme–substrate complex was investigated by mixing a fluorescently labeled enzyme–substrate complex (500 nM APEX1, 400 nM Hex 19-AP) with a high concentration of unlabeled substrate DNA (4 μM 19-AP). Reequilibration of the enzyme to the unlabeled substrate results in a reduction in the anisotropy of the labeled DNA, which was monitored by stopped flow. Data are shown fitted to a single exponential decay with the rates listed in Table S2.

Product Dissociation for NApe Appears Rate-Limiting.

To investigate the interaction of the loop mutants with DNA, we used stopped flow to determine the rate of dissociation from a fluorescently labeled substrate DNA (19-AP). The 19-AP was equilibrated with an excess of wild-type NApe, N207A, or R208A loop mutants. This was done in the absence of Mg²⁺ to avoid cleavage. Addition of MgCl₂ to the preincubated mixture led to rapid and complete cleavage of the substrate, demonstrating saturation of abasic site binding (Fig. S4). The change in anisotropy was observed on mixing the prebound complex with a 10-fold excess of unlabeled competitor substrate DNA (Fig. 5D). Wild-type NApe exhibited an extremely slow dissociation rate, whereas N207A and R208A had dissociation rates that increased by 11-fold and 45-fold, respectively (Fig. 5D). Compared with the dissociation rates for full-length APEX1 and ΔN-APEX1 (missing N-terminal 35 amino acids), it can be seen that the two mutants begin to approach the much faster dissociation rates observed for the human enzyme (Fig. 5D). The higher off-rates and K_Ms observed for N207A and R208A are consistent with these NApe mutants behaving more like the human enzyme than wild-type NApe (Table S2). The high off-rates are also consistent with weaker substrate and product binding, which correlates with the high values of K_M and k_cat observed in steady-state reactions for both loop mutants and APEX1 (Fig. 5A and Table S2). Slow product release has previously been reported for APEX1 and it has been suggested that this relates to protecting the site of damage until the next step in the repair process is initiated (20).

Evolutionary Conservation of AP-Site Recognition in Bacteria.

Based on the unexpected differences in abasic site recognition between NApe and APEX1, which correlated with differences in their catalytic activities, we speculated that they represented evolutionarily distinct classes of AP endonucleases. By using a new probability mixture model sequence-clustering algorithm (details in SI Materials and Methods), we analyzed 1,491 AP endonuclease sequences from all organisms. Sequences were clustered on the basis of the identities of amino acids at structurally equivalent positions to S115, N207, and R208 (termed S-N-R cluster) within the DNA recognition loops, with the consideration of residues in the active site involved in the catalytic reaction and metal binding. From this analysis we identified six clusters that were significantly different from each other (Fig. S5). The NApe-like and APEX1-like sequences clustered separately with APEX1-like sequences representing the largest cluster in eukaryotes (Fig. S5, cluster 3; R/Q-Y-M), whereas the NApe-like sequence is more common in prokaryotes (Fig. S5, cluster 1; S-N/Y-R). On the basis of multiple sequence alignments of the full AP endonuclease sequences, we also constructed phylogenetic trees colored by cluster membership for both eukaryotes and prokaryotes, which showed distinct preferences within subbranches (Fig. S6 A and B).

This detailed analysis shows that prokaryotes tend to contain the sequence pattern X/S-Y/N/Q-R for the DNA recognition loops, whereas in eukaryotes this is generally X/R-Y-M/R. The presence of Y prefixes the conformation of the main recognition loop via a conserved hydrogen-bonding network, whereas the presence of an N or Q can result in orphan base binding (Fig. S7). Another major difference is the preference of an R in prokaryotes versus the preference for M in eukaryotes. This results in a distinctly different AP-site recognition mechanism as described above (Fig. 4). The functional significance of these distinct sequences requires further investigation but when considered with the mutational analysis described above, could relate to the enzymatic capacity of different AP endonucleases within different physiological contexts.

Conclusions

In this study we show that the ExoIII type AP endouncleases from N. meningitides (NApe) binds and recognizes abasic DNA differently from the human enzyme. Crystal structures for NApe–DNA complexes representing different stages in the catalytic mechanism supports a one-metal mechanism with an activated water molecule acting as a nucleophile. Biochemical binding and activity measurements show that the neisserial enzyme although slower than the human enzyme, has a similar k_cat/K_M and that the release of product is rate limiting. By analyzing the structure/sequence determinant for substrate binding and recognition, we have identified distinct phylogenetic groupings that appear to show preferences between prokaryotes and eukaryotes.

Materials and Methods

Protein Mutagenesis, Purification, and Crystallization.

An extended description of protein purification, site-direct mutagenesis, complex formation and purification, crystallization, and in situ DNA cleavage can be found in SI Materials and Methods.

Data Collection and Process, Structure Determination, and Refinement.

Datasets were collected at Daresbury beamline 10.1 (now closed) and the Diamond Light Source beamlines IO3 and IO4. Except the dataset of product complex, all other datasets were processed in iMosflm (21) and scaled and truncated in Scala (22). The product complex data were automatically processed into P42₁2 space group in Xia2 (23) and were changed to space group of P4₃2₁2 using Reindex Reflections in CCP4 suite (22). The R factor flag was copied from the metal-free dataset to avoid possible contamination. Statistics of the datasets are given in Table S1. Structures were solved by molecular replace in Phaser (24), using the crystal structure of NApe alone. The DNA chains were built in manually. Metal ions were located according to the anomalous map. The structures were refined in Phenix (25), with TLS fragmentation from the serving Web site (26). Refinement statistics of the final models are given in Table S1. DNA curvature was analyzed in Curves+. All structure figures were made in Pymol. Crystallographic coordinates have been deposited in the Protein Databank (PDB ID codes: 4B5F, 4B5G, 4B5H, 4B5I, 4B5J, and 4B5M).

Steady-State Kinetics.

All reactions, including metal ion titrations, were performed using oligonucleotide substrates and reaction conditions as previously described (27). Data were fitted to the Michaelis–Menten equation using Grafit 6 (Erithacus software).

Anisotropy Binding Assays.

The 19-bp 5′-HEX DNA (SI Materials and Methods) (50 or 100 nM) was titrated with increasing amounts of the protein. Steady-state fluorescence anisotropy of HEX was recorded (excitation at 535 nm and emission at 550 nm). Data were collected at 25 °C on a Jobin Yvon FluoroMax-3 in 50 mM Tris 7.5, 125 mM NaCl, 1 mM DTT, 1 mM EDTA. Data were fitted to the tight binding equation (28) using Grafit 6 (Erithacus software).

Stopped-Flow Measurements.

The 19-bp 5′-HEX DNA was used as a substrate. Enzyme (500 nM) and labeled DNA (400 nM) were prepared and mixed in the stopped flow with a 10-fold excess of unlabeled substrate. Reaction traces were monitored using the Hi-Tech KinetAsyst (TgK Scientific) in dual channel anisotropy mode using Hg 546 nm excitation and 550 nm long pass filter. Reactions were carried out in 50 mM Tris 7.5, 50 mM NaCl, 10 mM EDTA at 15 °C. Data of three reproducible datasets were averaged and fitted to a single exponential decay using Grafit 6 (Erithacus software).

Bioinformatics Analysis of Loop Sequences.

A new probability mixture model sequence-clustering algorithm was used to cluster AP endonuclease sequences from all organisms based on the identities of amino acids at structurally equivalent positions to S115, N207, and R208. A detailed description of the method is given in SI Materials and Methods.