Domain Mapping of Human Apurinic/Apyrimidinic Endonuclease

We recently described the pre-steady state enzymatic binding kinetics of apurinic/apyrimidinic endonuclease (AP endo). In this report we describe the domain structure of the enzyme in solution determined by mild protease digestion in the presence and absence of substrate, product, and an efficient competitive inhibitor (HDP). AP endo is a 35.5-kDa protein with a high degree of homology to its prokaryotic counterpart, exonuclease III (Exo III), except for the amino terminus, which is lacking in the prokaryotic enzyme. The entire conserved region plus an additional 20 residues unique to the eukaryotic enzyme was inaccessible to trypsin and V8 protease, indicating that it forms a tight globular structure. In contrast, the amino-terminal 35 residues were readily accessible to all the proteases investigated, leading us to conclude that they associate poorly with the rest of the structure and constitute a highly fluid region. When AP endo was boiled with SDS and cooled prior to the addition of V8 protease, several acidic residues within the globular domain became protease-accessible, indicating rapid renaturation except along the nuclease fold with restoration of globular conformation for the carboxyl two-thirds of the molecule. Of all the proteases tested, only chymotrypsin was able to cleave internal to the globular portion without prior denaturation. Although AP endo cleaved with chymotrypsin retained full enzymatic activity, the activity was lost when the digested peptides were recovered after denaturation by heat and/or boiling in SDS, precipitation, and renaturation or when fragments were recovered from an SDS gel and renatured. Thus, the protein is probably held together strongly by noncovalent interactions that maintain enzymatic function after protease nicking. The three major chymotrypsin cleavage sites, Tyr-144, Leu-179, and Leu-205, became strikingly less accessible to protease digestion in the presence of abasic site-containing DNA. Since the three residues form a spherical triangle on the surface of the molecule on one side of the nuclease fold, there must be multiple means by which DNA containing an abasic site associates with the enzyme. The most likely explanation is that substrate and product, both of which were present during proteolysis, bind differently to the enzyme. Finally, the two cysteine residues thought to be involved in the redox reaction of AP endo with Jun protein were entirely inaccessible to proteolysis even after prolonged exposure of AP endo to reducing agents. Consequently, if AP endo plays a role in the physiological function of Jun, it must undergo major conformational changes in the process. Alternatively, the two cysteines could maintain an appropriate conformation such that other residues participate directly in the redox activity.

Apurinic/apyrimidinic endonuclease (AP endo) 1 is a critical enzyme in the repair of abasic sites in DNA. These sites can arise through spontaneous loss of a nucleobase, through the action of N-glycosyl hydrolases, or through oxidative damage of DNA (1,2). The crystallographic structure of human AP endo has been determined at 2.2-Å resolution, revealing a 2-fold symmetric ␣/␤ fold with a single metal ion bound at the active site (3). The three-dimensional structure of the human enzyme is similar to the prokaryotic homologue, exonuclease III (Exo III), from Escherichia coli (4,5). Not only does the amino acid alignment of AP endo with Exo III reveal a high degree of overall sequence similarity, but also the tertiary structure of the putative active site (the nuclease fold) is conserved (5). However, the crystal structures of both AP endo and Exo III were obtained in the absence of DNA so that there is still no clear picture as to the precise interaction of the abasic site substrate or its cleavage product with the enzyme. Furthermore, the amino-terminal 60 residues of AP endo are missing in Exo III, and the first 35 amino acid residues were truncated from AP endo in order to obtain crystallization. Hence, the relationship of the amino terminus to the remainder of the protein and of the protein to its substrate and product is unknown. In this report we present the results of experiments using limiting proteolysis that provide information on the dynamic structure of human AP endo and its interaction with DNA containing an abasic site.

EXPERIMENTAL PROCEDURES
Domain Mapping-AP endo was prepared by expression of the protein from plasmid pXC53 as described (6). The protein was digested at 23°C with several proteases for the time intervals and with the concentrations listed in the text. Trypsin (bovine pancreas, either soluble or covalently attached to acrylic beads) and chymotrypsin (bovine pancreas, soluble) were obtained from Sigma, while V8 protease (Endoproteinase Glu-C) was obtained from Boehringer Mannheim. Trypsin digestion was performed in a buffer consisting of 50 mM Tris-HCl, pH 7.4, 10 mM MgCl 2 , 50 mM KCl, and 1 M urea. V8 protease digestion was performed in the same buffer lacking urea or with the urea substituted by 0.1% SDS. When SDS was included in the buffer, the protein was boiled for 3 min and cooled to room temperature before addition of the protease. Chymotrypsin digestion was performed in the same buffer as trypsin except that urea was omitted and the pH was maintained at 8.0. After each reaction was terminated by the addition of 6ϫ protein loading buffer (7) and boiling, the products, resolved by SDS-polyacryl-* This work was supported in part by funds from National Institutes of Health Grant CA 72702 and Northeastern University (to P. R. S.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ To whom correspondence should be addressed. Tel.: 617-373-3492; Fax: 617-373-3724. amide gel electrophoresis on 12.5% or 16% gels, were visualized with Coomassie Blue (7). When digestions were performed in the presence of DNA, the oligonucleotide was added 5 min before the addition of the protease. The ␤-elimination product of the abasic site containing oligonucleotide, HDP, was prepared by heating as described previously (6).
Recovery of Peptides from SDS-PAGE-Peptides to be eluted from SDS-PAGE were visualized by KCl staining, excised, homogenized, eluted in 50 mM Tris-HCl, pH 7.4, 0.1 mM EDTA, 150 mM NaCl, and acetone-precipitated as described by Hager and Burgess (8). The acetone precipitate was dissolved in 10 l of 6 M guanidine HCl in 50 mM Hepes-NaOH, pH 7.4, 20% glycerol, 0.1 mg/ml bovine serum albumin, 150 mM NaCl, 1 mM dithiothreitol, and 0.1 mM EDTA. The protein solution was shaken for 20 -60 min on a rotary shaker, diluted 50-fold in the same buffer without guanidine HCl, and left at room temperature overnight to allow reformation of secondary and tertiary structure. On two occasions digested and undigested AP endo were boiled in protein loading buffer, acetone-precipitated, renatured, and assayed directly. Recovery of the undigested control protein and the digestion products after boiling in protein sample buffer or after elution from SDS-PAGE was estimated from the enzymatic time course over a 10,000-fold range of dilutions. An untreated control and an undigested control to which renaturation buffer was added were used to determine the effects of renaturation buffer on nicking activity.
Protein Sequencing-The identity of particular fragments was determined from the MW app on denaturing SDS-PAGE and the first 6 -8 amino acids obtained by protein sequencing. Protease digestion products were resolved by SDS-PAGE as described above. Instead of staining, they were transferred to Sequalon AA or polyvinylidene difluoride membranes for amino-terminal sequencing using an Applied Biosystems model 77A sequencer (Sealy Center for Molecular Science, Galveston, TX) or an Applied Biosystems Model 477A sequencer (Tufts University Analytical Core, Boston, MA), and the membrane was stained briefly with Coomassie Blue to locate the appropriate bands.
Enzymatic Assays-Enzymatic assays were performed as described (6). In brief, the substrate was prepared from a double-stranded 45-base pair oligomer containing a single G/U base pair at position 21: To generate the abasic site at position 21, the single-stranded oligomer, labeled at the 5Ј-end of the U-containing strand, was annealed with its homologue. After the double-stranded oligomer was treated with uracil DNA glycosylase (1 unit/100 pmol of U-containing oligonucleotide; Epicentre Technologies, Madison, WI) for 20 -30 min at 37°C in the presence of 0.1 M NaBH 4 , the uracil DNA glycosylase was heat-inactivated at 75°C for 5 min. Steady-state enzyme activities were determined as described earlier (6) using 500 nM substrate and 0.5 nM enzyme over a 10 -60-s time interval. Single-turnover binding assays were performed using 0.4 nM enzyme and 4 nM substrate. Substrate was mixed with enzyme in the presence of 4 mM EDTA, which permits the substrate to bind but prevents enzymatic cleavage, for varying periods of time (0 -30 s). Divalent cation (10 mM MgCl 2 , final concentration) was then added simultaneously with trap, consisting of HDP ϩ heparin (6), so that the enzyme was limited to a single turnover. Cleavage was terminated by the addition of EDTA to a final concentration of 87 mM. Product (20mer) was resolved from substrate (45-mer) by gel electrophoresis on denaturing 15% polyacrylamide gels containing 8 M urea (6,9), followed by quantitation of substrates and products by PhosphorImager analysis (Molecular Dynamics, Sunnyvale, CA). To obtain binding constants, simulations using the HOPKINSIM program (10) were fitted to the data.

Native AP Endo Contains a Flexible Amino-terminal Portion
and a Tight Globular Domain-AP endo was treated with limiting concentrations of trypsin (0.1 g), which resulted in degradation of the amino-terminal 32 residues without cleavage in other parts of the protein (Figs. 1 and 2). The next basic residue, Lys-35, remained inaccessible to cleavage. Increasing the concentration of trypsin to 1 g resulted in degradation of AP endo to small peptides. Similarly, V8 protease (0.02-1.0 g) removed the terminal 36 residues after a 2-h exposure but failed to cleave Glu-39 (data not shown). Exposure to V8 protease (1 g) for as long as 24 h at 37°C failed to cleave the endonuclease further. Several other conditions known to optimize proteolysis by V8 protease (14) failed to promote further digestion. These included: 50 mM potassium phosphate (pH 6.5) with or without added KCl, 50 mM ammonium carbonate (pH 8.0) with 1 mM CaCl 2 instead of MgCl 2 , and 50 mM Hepes-NaOH (pH 7.4) with 1 mM CaCl 2 instead of MgCl 2 . Exposure to chymotrypsin, which is described in detail below, resulted in partial loss of the amino terminus up to the first 31 residues ( Fig. 1) and cleavage at two specific internal residues. Because the first 36 residues of the amino terminus are readily accessible to proteolysis by all the proteases examined, they are flexible and only loosely associated with the remainder of the molecule.
The Flexible Portion Does Not Extend to the Cysteine Residues Thought to be Involved with Jun Interaction-Note that Cys-65 and Cys-93, the two cysteine residues required for AP endo acting as a redox factor for Jun, were not accessible to protease digestion (11,12). Since AP endo replaces dithiothreitol in enabling Jun to bind DNA, we reasoned that AP endo must itself be reduced at the pertinent cysteine residues to perform this function. We also reasoned, as did Walker et al. (13), that full reduction of Cys-65 and Cys-93 residues might change the conformation of AP endo such that more of the amino terminus might become accessible to protease digestion. In particular, we wanted to learn whether Lys-63, Lys-77, Lys-78, Lys-79, Lys-85, or Lys-95 would become cleavable to trypsin. However, when AP endo was incubated with 10 mM dithiothreitol for 24 h on ice and then subjected to proteolysis with trypsin, the region in which the two cysteines were located was still inaccessible to trypsin digestion (data not shown). Therefore, either the reducing agent had no effect on the protein or the protein was already fully reduced and full reduction does not lead to a change in conformation that results in solvent accessibility.
The Globular Domain of AP Endo Can Be Loosened by Boiling in Detergent to Make the Nuclease Fold Accessible-Since V8 protease functions in the presence of SDS, we chose to examine whether AP endo could be degraded after being boiled in the presence of detergent ( Figs. 1 and 2). The results are depicted visually in Fig. 5A. When AP endo was boiled in the presence of 0.1% SDS and cooled before V8 protease was added, it lost its enzymatic activity (Fig. 3) and several subdomains became more accessible to proteolysis (Figs. 1, 2, and 5A). The most prominent proteolytic sites were Glu-96, Glu-216, and Asp-283. Glu-96 and Asp-283 lie along the nuclease fold, whereas Glu-216 is exposed to the exterior surface (3) and is solvent-accessible. The surprise was that Glu-216 was not cleaved in the absence of detergent. Furthermore, Asp-283 is involved in catalysis (13), 2 whereas Glu-96 is the divalent cation binding site (3). One nearby residue that might have been expected to cleave in the presence of detergent but did not cleave was Glu-101 located in the loop between ␤2 and ␣3. Structurally, the enhanced accessibility argues either for greater flexibility along the nuclease fold than other regions of the protein or a failure to renature along the nuclease fold so that key residues extending into the region become sensitive to added V8 protease.
Chymotrypsin Cleaves within the Tight Globular Domain-Chymotrypsin was the only protease that readily cleaved the globular domain without the addition of denaturants (see Figs. 1, 2, 5B, and 6). In fact, to slow the protease, urea was omitted, Mg 2ϩ was substituted for Ca 2ϩ (14), and the pH was adjusted to 8.0. When these alterations in digestion conditions were made, three major digestion products were observed: a 19.9-kDa fragment beginning at Gly-145 and extending to the carboxyl terminus, a 19.5-kDa fragment beginning at the amino were boiled in SDS-PAGE loading buffer, recovered after acetone precipitation as described under "Experimental Procedures," and assayed for enzymatic activity at 1 nM and 5 nM over a 30-s interval. In this experiment 35% of the protein (lane 7) remained intact, as shown by SDS-PAGE followed by scanning densitometry. Columns 8 and 9, AP endo (190 pmol in 75 l) was incubated with buffer or with 19 g of chymotrypsin for 110 min at 22°C, and the products were resolved by SDS-PAGE. Protein from the sham-treated sample was recovered from the 30 -36-kDa region of the gel, while protein from the proteasetreated sample was recovered from the 18 -21-kDa range of the gel. After acetone precipitation and renaturation as described under "Experimental Procedures," the samples were diluted to a final concentration between 0.5 M and 0.05 nM (presuming full recovery) and assayed over a 30-s time interval. Recovery of the undigested 35.5-kDa protein was estimated as 22% of the starting material based on enzymatic activity.
terminus and extending to Leu-179, and a 12.8-kDa fragment beginning with Val-206 and ending at the carboxyl terminus ( Figs. 1 and 2). The two ϳ19-kDa fragments, unresolved by SDS-PAGE, were identified during amino acid sequencing in a ratio of approximately 3:1 (19.9-kDa fragment:19.5-kDa fragment). In keeping with the flexibility of the amino-terminal portion, substantial quantities of digestion products ending in Leu-179 and beginning at Arg-18, Ala-29, and Lys-32 were also observed. Although cleavage on the carboxyl side of aromatic amino acids as well as leucine and alanine is standard for chymotrypsin (14), the cleavage after Lys-31 is likely due to trace contamination with trypsin found in most commercial preparations of chymotrypsin and the extreme sensitivity of the amino-terminal domain to trypsin digestion. Note that Lys-141 and Lys-203, two potential trypsin cleavage sites in the region of Tyr-144 and Val-206, respectively, had failed to cleave when the endonuclease was exposed to trypsin. Thus, these residues were unlikely to be solvent-accessible.
Enzymatic Activity Is Recovered after Proteolysis but Not if the Fragments Are Dissociated before Assay-AP endo was shortened by trypsin digestion using trypsin attached to a solid surface. The endonuclease was separated from protease by microcentrifugation, and the amino-terminal peptides were removed by dialysis against 50 mM Hepes (pH 7.4) containing 0.1 mM EDTA. Thus truncated, AP endo retained its activity fully, both in terms of turnover number and ability to bind to substrate (data not shown). Similarly, AP endo that had been digested with chymotrypsin also retained its enzymatic activity (Fig. 3). Unlike trypsin, which failed to cleave within the globular domain and could even be used to define the globular domain, chymotrypsin cleaved residues within the globular domain. Therefore, we examined the chymotrypsin result further. AP endo, either digested with chymotrypsin or mockdigested, was boiled with SDS-PAGE sample loading buffer, precipitated with acetone, and renatured as described under "Experimental Procedures." The activity recovered in the digested sample (16% of the mock-digested control) could be accounted for by the fact that 35% of the protein remained undigested in this experiment. Enzyme that had been boiled for 5 min without SDS, heated to 78°C for 5 min, or boiled in SDS in protein sample buffer had no activity. At the same time, when the 19.5-kDa and 19.8-kDa fragments were eluted from an SDS-PAGE, renatured, and assayed for enzymatic activity, we recovered 0% of the control activity recovered from undigested enzyme eluted from the same gel. Since the two ϳ19-kDa fragments overlapped and encompassed the entire molecule, it was theoretically possible that activity could have been restored by renaturation. However, this did not occur. In short, the chymotrypsin data indicated that, even if the protein was nicked by proteolysis, hydrogen bonding and other noncovalent interactions maintained sufficient tertiary structure to protect enzymatic activity. However, if the protein was heated or treated with detergent to disrupt noncovalent interactions or if the protease fragments were dissociated after nicking, either by treatment with detergent or gel separation, activity could not be recovered.
Cleavage at Tyr-144,  Is Blocked by the Presence of DNA, Particularly an Oligonucleotide Containing an Abasic Site-We investigated the effect of the presence of 45-mer double-stranded oligonucleotide on digestion by trypsin, V8 protease, and chymotrypsin. Although there was no effect on digestion by the first two proteases, there was a striking effect on chymotrypsin digestion (Fig. 4). The moiety at position 21 of the oligonucleotide differed in each case: a G/U pair, for which the enzyme has a relatively poor binding affinity (6,15); a phosphodeoxyribose (the ␤-elimination product of an abasic site), which is an efficient inhibitor of AP endo (6); or an abasic site. To ensure complete binding, the ratio of oligonucleotide to AP endo during the incubation was 6:1. In the case of the oligonucleotide containing an abasic site, both product and substrate were present because AP endo was active during digestion due to the presence of Mg 2ϩ required for proteolysis. All three oligonucleotides delayed digestion of full-length AP endo. (In Fig. 4, compare the retention of the 35.5-kDa starting material in lanes 2-5 with that in lanes 7-10, 12-15, and 17-20.) However, only the presence of substrate/product delayed the appearance of the doublet band containing the 19.5and 19.8-kDa fragments. (In Fig. 4 compare the rate of disappearance of substrate and the amount of ϳ19-kDa fragments in the same lanes.) Note also the lack of appearance of either a 22.7-kDa fragment or a 15.7-kDa fragment that would have signaled decreased accessibility at either Leu-179 or Leu-205 without affecting the other sites. Although total degradation continues to occur during proteolysis, we feel that the presence of oligonucleotide, particularly substrate/product, has a profound effect on the accessibility of Tyr-144, Leu-179, and Leu-205 to chymotrypsin by blocking all three sites simultaneously. DISCUSSION Limiting proteolysis is an extremely useful tool in describing functional and structural regions of a protein molecule in solu-tion (16,17). Whereas x-ray crystallography provides details of a particularly thermodynamically stable conformation, domain mapping by limiting proteolysis provides information on dynamic structure. We have shown that AP endo contains two major regions, an amino-terminal flexible portion of approximately 36 amino acids, which is readily accessible to protease digestion, and the remainder of the molecule, which forms a tight globular domain that is inaccessible to most proteases. After boiling in 0.1% SDS, the enzyme loses its activity and the tight globular domain becomes somewhat more flexible or re-  chymotrypsin digestion (B). Ribbon diagram in stereo image with relevant amino acid residues shown in three dimensions. A, cleavage sites identified by V8 plus SDS. Note that Glu-216 is on the exterior of the molecule, while Glu-96 and Asp-283 face the nuclease fold. B, cleavage sites identified by chymotrypsin. Tyr-144 and Leu-179 are clearly solvent-accessible. Leu-205, which appears to be somewhat less solvent accessible from visual inspection of the crystal structure, nevertheless has space for a H 2 0 molecule adjacent to it and is also accessible by limiting proteolysis. The three-dimensional structure was visualized from the crystal coordinates by means of RASMOL (27).
natures partially, exposing several residues along the nuclease fold and one on the exterior surface to V8 protease. Even so, it is striking that after boiling in detergent, large regions of the molecule renature so efficiently that they are still relatively inaccessible to digestion. The only protease able to digest the protein within the globular domain is chymotrypsin. In this case digestion provides three major fragments involving cleavage at Tyr-144 or cleavage at Leu-179 and Leu-205. This is the first dynamic structural information available for this enzyme.
Perhaps the most striking result to emerge from this study is the alteration in the accessibility of the three major chymotrypsin cleavage sites in the presence of 45-mer oligonucleotide, particularly the 45-mer containing an abasic site. The three residues form a spherical triangle on one side of the nuclease fold (see space-filling residues in Figs. 5B and 6). The distance (arc length along a great circle 3 ) from Leu-179 to Leu-205 is 29.7 Å, the distance from Leu-179 to Tyr-144 is 34.4 Å, and the distance from Leu-205 to Tyr-144 is 15.8 Å. These distances correspond to approximately one turn of a DNA helix for the first two pairs and one-half a turn of the helix for the third pair (18). Given the distances between the three residues, it is unlikely that a single oligonucleotide could block all three sites simultaneously. Hence, we need to account for at least two modes of interaction. The blockage at Leu-179 is consistent with the model proposed by Gorman et al. (3), whereby the DNA binds across one aspect of the nuclease fold (See Fig. 6, arrows as aligned with Fig. 3 in Ref. 3), Leu-179 being part of ␣ helix ␣5 proposed to bind the major groove of DNA. There were no major protease cleavage fragments originating in the loop on the side of the nuclease fold opposite from Leu-179 in the absence of DNA, and no new fragments appeared during proteolysis in the presence of DNA. Therefore, we could not observe directly whether DNA binds across the nuclease fold, which would occur if ␣ helix ␣1 interacted with the minor groove of DNA in analogy with DNase I (19). The surface of the nuclease between Leu-179 and Leu-205 is occupied by ␣ helices ␣6 and ␣9 (see Figs. 5B and 6). However, the surface of the nuclease between Leu-179 and Tyr-144 is occupied by several regions that are relatively disordered, including Glu-150, His-151, Asp-152, Glu-154, Gly-176, Asn-174, Pro-173, Val-180, and Arg-181 that might provide an alternative binding site for a DNA helix and enable the substrate to swivel about Leu-179. By far the more intriguing distance is the half a helical turn from Tyr-144 to Leu-205. The region between Tyr-144 and Leu-205 is occupied by parts of ␤ sheets ␤5, ␤6, and ␤7 as well as ␣ helix ␣6, containing a number of conserved amino acids facing the exterior surface of the molecule. Furthermore, the straight line distance through the nuclease fold from Tyr-144 to Glu-96, the divalent cation binding site required for enzymatic activity (3), is 16 Å or again approximately one-half of a helical turn of DNA.
How then would an abasic site interact with the active site residues of AP endo? Even in the absence of enzyme, the presence of an abasic site lowers the melting temperature (20,22) and confers conformational changes to the adjacent and opposing base pairs (20,21). An abasic site causes perturbations in the helix for one-half a helical turn on the strand opposite the abasic site in the 5Ј direction (23) as shown by accessibility to nucleases S1 and P1. Addition of a second abasic site on the opposite strand disturbs the nicking reaction, especially when the abasic site is 3 base pairs 5Ј away from the initial site on the opposite strand, or slightly less than one-half a helical turn 3. Compare the locations of cleavage sites that are observed by limiting proteolysis and blocked by the presence of abasic site-containing oligonucleotide (space-filling residues) with the loop regions of AP endo (arrows), which were proposed to interact with an abasic site-containing oligonucleotide (3) based on the interaction of DNase I with DNA from the co-crystal (19). Note the presence of Leu-179 in ␣ helix ␣5 on one edge of the nuclease fold, providing support for the possibility that DNA interacts across the nuclease fold. Note also that Tyr-144 and Leu-205 are oriented at ϳ45°to Leu-179 with regard to the nuclease fold. away (24,25). 4 As a consequence, we propose that substrate and product bind differently to AP endo. Because the abasic site is already prone to deformation, the enzyme could induce a bend in the DNA, enabling the abasic site to become accessible to catalytic residues found in the nuclease fold. One-half a helical turn of DNA could lie along the backside of the enzyme and one-half could lie through the nuclease fold. The interference by abasic sites on the strand opposite to the site to be cleaved could now be viewed as distorting the helix so that it no longer fit into the nuclease fold. Once the abasic site is nicked and the double helix resumes a standard B or A form, it could remain associated with the endonuclease such that the predominant binding mode is across the face of the nuclease fold defined by ␣ helix ␣5. The fact that a double-stranded oligomer blocks cleavage but not to the same degree as one containing an abasic site is consistent with the fact that AP endo binds double-stranded DNA, whether or not there is an abasic site, at about 1% the level of its abasic site-containing substrate (6).
These results leave several questions unanswered. In particular, AP endo has now been shown to interact with DNA polymerase ␤ both by yeast two-hybrid analysis and electrophoretic mobility shift assays (26). Presumably the portion of the nuclease that interfaces with the polymerase must face the exterior of the molecule. However, one cannot presume that the amino-terminal domain fulfills this function, since many portions of the molecule distributed throughout the primary structure are solvent-accessible (3). Finally, Cys-65 and Cys-93, the residues previously thought to be responsible for AP endo acting as a redox factor in terms of binding Jun protein (11,12), are clearly located internally and are not exposed to solvent even after extended incubation under reducing conditions. Although we cannot rule out a major conformational change occurring in the presence of Jun protein, one possible explanation for this observation is that Cys-65 and Cys-93 are not themselves involved in the redox activity. Rather, the two cysteines are responsible for maintaining the molecule in a conformation such that other residues on the surface of AP endo can act as a reductant for Jun protein.