Mode Analysis of a Fatty Acid Molecule Binding to the N-terminal 8-kDa Domain of DNA Polymerase β

We reported previously that long-chain fatty acids are potent inhibitors of mammalian DNA polymerase β. At present, based on information available from the NMR structure of the N-terminal 8-kDa domain, we examined the structural interaction with the 8-kDa domain using two species, C18-linoleic acid (LA) or C24-nervonic acid (NA). In the 8-kDa domain with LA or NA, the structure that forms the interaction interface included helix-1, helix-2, helix-4, the three turns (residues 1–13, 48–51, and 79–87) and residues adjacent to an Ω-type loop connecting helix-1 and helix-2 of the same face. No significant shifts were observed for any of the residues on the opposite side of the 8-kDa domain. The NA interaction interface on the amino acid residues of the 8-kDa domain fragment was mostly the same as that of LA, except that the shifted cross-peaks of Leu-11 and Thr-79 were significantly changed between LA and NA. The 8-kDa domain bound to LA or NA as a 1:1 complex with a dissociation constant (K D ) of 1.02 or 2.64 mm, respectively.

We reported previously that long-chain fatty acids strongly inhibited the activities of mammalian DNA polymerase ␣ (pol ␣) 1 and DNA polymerase ␤ (pol ␤) in vitro and plant DNA polymerases, albeit less potently, but that at the concentrations used, the fatty acids hardly influenced the activities of prokaryotic DNA polymerases or other DNA metabolic enzymes such as DNase I (1,2). The most potent inhibitors were fatty acids, which have the following characteristics: hydrocarbon chain containing 18 or more carbons, a free carboxyl end, and the cis-configuration is preferred to the trans-configuration. Fatty acids in the trans-configuration have a much weaker inhibitory effect on pol ␤, and those in which the carboxyl end is chemically modified can lose the inhibitory effect on both pol ␣ and pol ␤. The mode of inhibition by longer chain fatty acids showed the same characteristics, except that the minimum inhibitory doses of these longer chain fatty acids were much lower (2,3). Lineweaver-Burk plots of the fatty acids indicated that both the substrate (i.e. deoxynucleotide)-binding and the template DNA-binding sites of pol ␣ were nonantagonistically inhibited by the fatty acids, but they were effective as antagonists against the sites of pol ␤. For pol ␤, fatty acids acted by competing with not only the substrate but also the template-primer DNA. In screening inhibitors of eukaryotic DNA polymerase, we also found several natural compounds which inhibited pol ␤ in the same manner as fatty acids (4 -10).
Pol ␤ is the smallest known DNA polymerase in animal cells with a molecular mass of 39 kDa, and its structure is highly conserved among mammals (11). This protein has a modular two-domain structure, with apparent flexibility within a protease-sensitive region between residues 82 and 86, which separates the two domains. Treatment with trypsin yields an N-terminal domain fragment (8 kDa), which retains binding affinity for single-stranded DNA (ssDNA), and a C-terminal domain fragment (31 kDa) with reduced DNA polymerase activity (12,13). We reported previously the mode of biochemical inhibition by fatty acids using two of the pol ␤ fragments that were proteolytically separated (3). The fatty acids were found to bind to the 8-kDa DNA-binding domain fragment and to suppress binding to the template-primer DNA. A 10,000-fold higher level of fatty acid was required for binding to the 31-kDa catalytic domain or to inhibit the DNA polymerase activity, suggesting that it directly disturbs the template-primer incorporation into the template-primer-binding domain and indirectly competes with the substrate on its binding site in the catalytic domain (3). The binding between the enzymes and fatty acids can be released by detergents without any permanent damage to the structure (2, 3). These results suggested that the binding is physiologically specific and has some roles in vivo: for example, to maintain the enzymes in the inactive state on the internal surface of the membranes. In this study, we analyzed the structural interactions of C 18 and C 24 fatty acids with pol ␤, especially the N-terminal 8-kDa domain, in cross-linking studies using a 5Ј-end-labeled photoprobe (dT 14 D) instead of template-primer DNA and the binding surface of the 8-kDa domain in contact with the fatty acids by NMR.
The crystal and NMR structures of pol ␤ and the N-terminal 8-kDa domain of pol ␤ have been determined recently by Wilson and his co-workers (14 -22). Based on the information available from their studies of crystal and NMR analyses of pol ␤ and the 8-kDa domain (14 -22), we compared the interaction interface for ssDNA template with that for the fatty acids. The study of the relationship between fatty acids and pol ␤ may reveal why C 16 or shorter fatty acids cannot inhibit the polymerase activity (2), why the minimum inhibitory doses of longer chain fatty acids are much lower than those of shorter chain species, although the biochemical mode of inhibition is the same, and how the fatty acids bind to the N-terminal 8-kDa domain. These studies may help to further clarify the structure and function of pol ␤ and subsequently may allow us to speculate on the in vivo role of DNA polymerase inhibition by fatty acids.

EXPERIMENTAL PROCEDURES
Sample Preparation-The N-terminal 8-kDa fragment of rat DNA pol ␤ (residues 2-87) was overexpressed in Escherichia coli strain BL21 harboring the expression plasmid "Lys-87" constructed in our laboratory. Overproduction of the N-terminal 8-kDa domain and the purification procedure have principally been described in our previous report (23). For 15 N-correlated NMR experiments, the N-terminal 8-kDa domain was expressed from BL21/Lys-87 grown on minimal medium containing 15 NH 4 Cl as the sole nitrogen source (24). In preparing the NMR sample, the purified N-terminal domain was concentrated using a Centricon-3 (Amicon) and exchanged into 5 mM potassium phosphate buffer (pH 7.0) and 20% D 2 O. Two samples for NMR experiments contained 1.25 mM 15 N-labeled N-terminal 8-kDa domain after addition of fatty acids (linoleic acid (LA) and nervonic acid (NA)). The fatty acids (12.5 mM each) were dissolved in Me 2 SO-d 6 . Fragments of 8 and 31 kDa of pol ␤ were prepared and purified as described previously (3).
Circular Dichroism Spectroscopy-For CD analysis, 6.25 M (50 g/ ml) purified 8-kDa domain fragment of pol ␤ and 62.5 M fatty acid mixture (molecular ratio, enzyme:inhibitor ϭ 1:10) were dissolved in 5 mM potassium phosphate buffer (pH 7.0) containing 5% methanol. The chromatograms for the mixtures were compared with that of the 8-kDa domain alone. The concentrations of the proteins were determined by UV adsorption at 280 nm (⑀ 280 ϭ 5440 M Ϫ1 cm Ϫ1 ). CD measurements were performed on a Jasco J-720 spectropolarimeter in a 1-cm cell at 25°C. The CD spectra were collected from 260 to 200 nm at a resolution of 1 nm using up to eight scans. The per residue molar ellipticity (degree⅐cm 2 dmol Ϫ1 ) was calculated from the concentration for the 87residue polypeptide.
NMR Experiments-NMR spectra were measured at 750 MHz on a Varian Unity-Plus 750 spectrometer. 1 H-15 N HMQC spectra were recorded at a temperature of 30°C. Each spectrum size was 1024 complex points in the t2 dimension and 96 complex points in t1. The data were zero filled in both dimensions, and a shifted sine-bell was applied as a window function for resolution enhancement. A total of 32 scans per FID was accumulated, leading to a measuring time of 90 min per HMQC spectrum.

RESULTS AND DISCUSSION
The in vitro relationship between mammalian DNA polymerases and fatty acids has been investigated (2, 3). As described in the Introduction, longer chain fatty acids (over 18 carbons) strongly inhibited DNA polymerase activities in vitro. Fatty acids in which they are of the trans-configuration have much weaker inhibitory effects on especially pol ␤, and the fatty acids in which the carboxyl end is chemically modified can lose the inhibitory effect. In this study, we analyzed the structure of pol ␤ and its relationship to the long-chain fatty acids in more detail. Lineweaver-Burk plots of fatty acids indicated that both the substrate-binding and the template DNA-binding sites of pol ␤ were antagonistically inhibited by fatty acids (2,3). We tested fatty acids from C 18 to the longest commercially available, C 24 , previously (3). Among the fatty acids examined, the strongest inhibitor was a C 24 fatty acid, nervonic acid (NA), and the weakest was a C 18 fatty acid, linoleic acid (LA) (3). We therefore analyzed the mode of binding to pol ␤ using the longest and the shortest fatty acids in the present study.
The effects of fatty acid were analyzed using proteolytic methods. Rat pol ␤, which was also used in this study, can be divided into two domain fragments (8-and 31-kDa polypeptides) using controlled proteolysis (12,13). The 8-kDa domain is the DNA template-binding domain, and the 31-kDa domain is the catalytic part involved in DNA polymerization. According to the methods described by Kumar et al. (12), we purified both of these fragments by fast protein liquid chromatography Superose 12 (lanes 3 and 4 in Fig. 4, Ref. 3) and used them in this experiment. We also suggested previously that the fatty acids bind to pol ␤ and the 8-kDa domain fragment, but not to the 31-kDa domain fragment (3). Both NA and LA appear to interact with the enzyme or the 8-kDa domain fragment in the same way, but the longer chain binds to the domain fragment more tightly and inhibits DNA polymerase activity much more strongly.
Analysis of the Binding between Fatty Acids and Fragments of 8 and 31 kDa of Pol ␤ by Using Cross-linking-The labeling abilities of the synthesized 32 P 5Ј-end-labeled photoprobes (dT 14 D) were investigated using the N-terminal 8-kDa and C-terminal 31-kDa domain fragments of the recombinant rat pol ␤ (39 kDa). The 32 P 5Ј-end-labeled photoprobe (dT 14 D) had a molecular mass of 5 kDa. dT 14 D was mixed with 8-and 31-kDa fragments of pol ␤ and irradiated with near-UV light as described under "Experimental Procedures," and then analyzed by gel mobility shift assay (Fig. 1). As shown in Fig. 1A, autoradiography of the radioactive products of the 8-kDa domain fragment resolved by SDS-PAGE showed a shift from the original 8-kDa to the 13 (8 ϩ 5)-kDa position as the labeled protein complex (lane 1). Poly (dT), which is a DNA template, competed with the photoprobe for binding to the 8-kDa domain fragment (lanes 1-3 in Fig. 1A). The 8-kDa band observed in  Fig. 1A). The 8-kDa domain fragment of pol ␤ could be shifted with the cross-linked dT 14 D, but the 31-kDa fragment, the catalytic domain without a DNA-binding site, could not, because this domain has no DNA binding capacity (lane 1 in Fig. 1A). Crosslinking of the fragments of pol ␤ with photolabile dT 14 D was inhibited by addition of the natural template, poly(dT) (lanes 1-3 in Fig. 1A), showing that the 8-kDa domain fragment has contained the template DNA-binding site.
LA and NA interfered with complex formation of the 8-kDa fragment and ssDNA template (Fig. 1,B and C). pT 14 1C) were 0, 0.25, 0.5, and 1, respectively. NA interfered with binding of the DNA to the 8-kDa fragment, and at an I/E ratio of 1, the interference became nearly complete (Fig. 1C). We, therefore, concluded that one molecule of fatty acids competes with oligonucleotide, i.e. dT 14 D, suggesting that a fatty acid molecule interferes with the binding of template DNA to one molecule of the 8-kDa domain fragment. Similar results were obtained using LA instead of NA (Fig. 1B). LA also interfered with DNA binding to the fragment, but the interference was not complete at an I/E ratio of 1. Thus, the C 24 fatty acid, NA, showed stronger interference than the C 18 fatty acid, LA, at the I/E ratio of 1. This may explain the observation that the inhibition of pol ␤ activity by NA was 10-fold stronger than that by LA although the biochemical mode of inhibition was the same (2, 3).
To explain why the minimum inhibitory dose of the longer chain fatty acid was much lower than that of short chain species, the dissociation constants (K D ) between each of the fatty acids and the domain fragment were also analyzed as described in the later part of this report (Fig. 3). To investigate the binding mode including K D in detail, NMR structures of the N-terminal 8-kDa domain with or without the fatty acids were determined.
CD Spectra of 8-kDa Domain and Mixture of 8-kDa Domain and Fatty Acids-The CD spectra of complexes of the 8-kDa domain fragment and the fatty acids were very similar to the CD spectrum of the 8-kDa domain fragment alone (Fig. 2). The comparable maximal negative ellipticities at 208 and 220 nm indicated that the overall helical structure in the mixture of the 8-kDa domain, and the fatty acids were similar to that of the 8-kDa domain fragment alone (Fig. 2). The spectra of the protein-LA complex and the protein-NA complex were similar to each other, but the maximal negative ellipticity of 208 nm and the maximal positive ellipticity of 235-260 nm of the protein-NA complex were higher than those of the protein LA complex (Fig. 2). The unchanged ratio of the maximal negative ellipticity at 222 nm versus 208 nm in the mixture of the 8-kDa domain fragment and the fatty acids suggested no increase in helical structure in comparison with the 8-kDa domain fragment alone. On the basis of these results, we concluded that the fatty acids do not adversely affect the overall structure of the 8-kDa domain fragment.

Analysis of the Binding of Fatty Acids to the N-Terminal 8-kDa Domain by NMR-
The NMR structures of the N-terminal 8-kDa domain have recently been determined by Wilson, Mullen, and their co-workers (19). According to their results, the 8-kDa domain (residues 1-87) is formed by four ␣-helices, packed as two antiparallel pairs. The pairs of ␣-helices cross one another at 50°giving them a V-like shape. The 8-kDa domain contains a motif termed the "helix-hairpin-helix." The protein residues involved in template DNA binding have been identified by NMR using chemical shift changes (16). The helix-3-hairpin-helix-4 motif and residues in an adjacent ⍀-type loop connecting helix-1 and helix-2 form the ssDNA interaction surface (16). Furthermore, they also found that several mutants of the 8-kDa domain (F25W, K35A, K60A, and K68A) showed impaired template DNA binding activity (14). In a biochemical study, using the purified recombinant 8-kDa domain, photochemical cross-linking studies showed that residues Ser-30 and His-34 cross-linked to p(dT) 16 (27).
In studying the effects of fatty acid binding, the recombinant 8-kDa domain fragment was titrated with a 12.5 mM stock solution of LA or NA. Two-dimensional 1 H- 15 4. Expanded 1 H-15 N HMQC spectra of the 15 N-labeled N-terminal 8-kDa domain-fatty acid complex. A, the N- Leu-11 resonance being averaged into a single resonance (␦av) (i.e. (␦F Ϫ ␦B)Ͻ Ͻk off for the complex (28)). Fitting of the titration curve for the amide proton resonance of Leu-11 indicated that the 8-kDa domain binds to LA or NA as a 1:1 complex with a K D of 1.02 or 2.64 mM, respectively (Fig. 3), indicating that the longer fatty acid could bind to the fragment more tightly. This probably explains why the minimum inhibitory dose of the longer chain fatty acid was much lower, although the biochemical mode of inhibition was the same. Since C 16 or shorter fatty acids are expected to have higher K D values than the K D of the C 18 fatty acid, this may also explain why the shorter chain fatty acids cannot inhibit polymerase activity. Fig. 4A shows the 1 H- 15  In the presence of either LA or NA, the cross-peaks were shifted as follows: Lys-5, Ala-6, Gln-8, Glu-9, and Leu-11 were in the unstructured segment; Glu-26 was in helix-1, which is adjacent to the ⍀-type loop; Asn-28, Val-29, Ser-30, Ile-33, and Lys-35 were in the ⍀-type loop; Asn-37 and Tyr-39 were in helix-2, which is adjacent to the ⍀-type loop; His-51 and Lys-52 were in a turn; Ile-73, Asp-74, Phe-76, Leu-77, and Ala-78 were in helix-4, which is adjacent to the 48 -55 turn and 79 -87 unstructured linker segment; Thr-79, Gly-80, Leu-82, and Lys-84 were in the unstructured linker segment that connects to the 31-kDa catalytic domain in the full-length enzyme. These chemical shift changes can be explained in terms of the fatty acid contact and perturbation in the electrostatic charge distribution at the surface. Surface residues displaying chemical shift changes were predominantly, although not entirely, clustered on one side of the domain (Fig. 6A). Furthermore, the fatty acid-binding interface of the 8-kDa domain consists of two regions: one consisting of Leu-11 in the 1-13 unstructured segment, His-51 in the 45-55 turn, and Thr-79 in the 79 -87 unstructured linker segment ("I" in Fig. 6A), while the other consists of an ⍀-type loop, including helix-1 and helix-2 ("II" in Fig. 6A). Fig. 6B shows the mapping in the solution structure of the 8-kDa domain with ssDNA. The data determined by Wilson, Mullen, and their co-workers, the NMR structures (16,19) and the results of site-directed mutagenesis of the 8-kDa domain (14), were used to illustrate the map. According to Prasad et al. (14), the site-directed mutants of Phe-25, Lys-35, Lys-60, or Lys-68 were impaired template DNA binding activity. Since the fatty acids bind to the ssDNA-binding region of the 8-kDa domain and compete for binding with  (14,16,19). The amino acid residues Phe-25, Lys-35, Lys-60, and Lys-68, which were shown to be necessary for ssDNA binding activity by site-directed mutagenesis (14), are indicated in blue. The amino acid residues of the major shifted cross-peaks from HMQC NMR experiments (16) are depicted in pink.
template DNA as shown in Fig. 1, two of the maps were compared. In Fig. 6B, the residues (Gln-31, Asn-37, Arg-40, Lys-41, Ala-57, Glu-58, Lys-60, Gly-66, Lys-68, Glu-71, Lys-72, Glu-75, Ala-78, Leu-82, Arg-83, and Leu-85) of the p(dT) 8 interaction interface from the HMQC NMR experiment (16) are shown. As shown in Fig. 6, A and B, the only site shifted not only by fatty acid binding, but by ssDNA (i.e. p(dT) 8 ) binding was Lys-35 in the ⍀-type loop including helix-1 and helix-2. The region II shown in Fig. 6A appears to have an important role in the fatty acid effect. The fatty acids probably compete with template DNA at the residue Lys-35 and bind to the site, which subsequently inhibits the ssDNA binding activity on the 8-kDa domain. In the region I shown in Fig. 6A, Leu-11, His-51, and Thr-79 are different from the other DNA binding sites (Phe-25, Lys-60, and Lys-68), suggesting that the methyl end of fatty acids disturb the binding of the template DNA at region I. Lys-35 in region II is a hydrophilic amino acid, and Leu-11 and His-51 in region I are hydrophobic amino acids. The carboxyl ends of the fatty acids may, therefore, show a preference for binding to the hydrophilic site, and the other side, the methyl end, may be absorbed to the hydrophobic site. We reported previously (2, 3) that longer chain fatty acids inhibited the binding activity more strongly and that C 16 or shorter fatty acids have no inhibitory effect. The distance between regions I and II may be a key to explain these characteristics of the inhibition by fatty acids. As shown in Figs. 3, 4C, and 5, the shifted cross-peaks of Leu-11 and Thr-79 were significantly changed between LA and NA. The longer chain fatty acids (over C 24 ) are expected to more tightly bind to these residues of amino acids in the 8-kDa domain.
Modeling of the Fatty Acid Interaction Interface--To confirm the above assumption, we performed modeling analysis using the results of NMR experiments. The results of computer simulation of the binding mode between the N-terminal 8-kDa domain and the fatty acids are shown in Fig. 7. NA (yellow line) on the 8-kDa domain (blue-white line) in Fig. 7, A and B, was bridged from Lys-35 (red line) to Leu-11 (red line), His-51 (red line), and Thr-79 (red line) and intercalated smoothly into the pocket between helix-1 and helix-2 in the ⍀-type loop. The distance between the Lys-35 hydrophilic region and Leu-11 and His-51 hydrophobic regions fit the length of U-shaped NA. On the other hand, LA (yellow line) was trapped more deeply in the pocket, although the LA binding model is basically the same as that of NA, and was further away from regions I and II than in the NA binding model (Fig. 7, C and D). The methyl end of LA may not be able to bind firmly to region I, and subsequently, a steeper U-shaped model was postulated for LA (Fig. 7, C and  D). In this simulation, the fatty acid structures were modeled, but the 8-kDa domain structure was fixed. Although the residues shown by red lines seemed to be separated from the fatty acid structures (yellow lines) (Fig. 7), the fatty acid ends are thought to bind to the respective amino acid residues, and at least the Lys-35 binding area in the 8-kDa domain peptide must be strained. The unstructured segment of the 1-13 turn is comprised of the N-terminal residues and is flexible. The 79 -87 turn, i.e. the unstructured linker segment, must also be structurally flexible. Therefore, when the fatty acids bound to the domain at His-51 and Lys-35, the N-terminal turn including Leu-11 and the unstructured linker segment including Thr-79 appear to be adjacent to His-51 in the 45-55 turn. The longer chain fatty acids, since they can more likely gain access to region I, can affect the tighter binding to region I. LA or shorter fatty acids are located at some distance from the site and may hardly induce the movement. We also reported previously that the saturated forms of fatty acids, for example C 18stearic acid, had no inhibitory effect on pol ␤, although then could suppress the activity of pol ␣ (2). The carbon chain in the saturated form fatty acid molecule is linear and does not form a U-shaped curve as seen in the unsaturated form fatty acids such as LA and NA. The linear chain may not be able to intercalate between helix-1 and helix-2 in the ⍀-type loop and thus cannot inhibit pol ␤ activity.
In conclusion, the lack of an effect of shorter chain fatty acids, the positive relationship between longer carbon chain length and tighter binding, and the configuration effectiveness on pol ␤ can be explained by our model.
The inhibitory effects of fatty acids on DNA polymerase activity occur by binding between the 8-kDa domain and the fatty acid as a 1:1 complex, and this binding can be released by nonionic detergents (2,3). The fatty acids are present on the internal surface of the cytoplasmic membranes. These observations suggest that the inhibitory effect of fatty acids on DNA polymerase activity occurs in vivo and is reversibly controlled by binding to or release of the DNA polymerase from the fatty acids, perhaps on the membranes. These observations may help in determining the mechanisms of control of these enzymes in vivo.