Article No. mb J. Mol. Biol. (1998) 284, 1095±1111

Transcription

1 Article No. mb J. Mol. Biol. (1998) 284, 1095±1111 Structure and Functional Implications of the Polymerase Active Site Region in a Complex of HIV-1 RT with a Double-stranded DNA Template-primer and an Antibody Fab Fragment at 2.8 AÊ Resolution Jianping Ding 1,Kalyan Das 1,Yu Hsiou 1,Stefan G. Sarafianos 1 Arthur D. Clark Jr 1,Alfredo Jacobo-Molina 1,Chris Tantillo 1 Stephen H. Hughes 2 and Edward Arnold 1 * 1 Center for Advanced Biotechnology and Medicine (CABM) and Rutgers University Chemistry Department, 679 Hoes Lane Piscataway, NJ , USA 2 ABL-Basic Research Program NCI-Frederick Cancer Research and Development Center P.O. Box B, Frederick, MD , USA *Corresponding author The structure of human immunode ciency virus type 1 (HIV-1) reverse transcriptase (RT) complexed with a 19-mer/18-mer double-stranded DNA template-primer (dsdna) and the Fab fragment of monoclonal antibody 28 (Fab28) has been re ned at 2.8 AÊ resolution. The structures of the polymerase active site and neighboring regions are described in detail and a number of novel insights into mechanisms of polymerase catalysis and drug inhibition are presented. The three catalytically essential amino acid residues (Asp110, Asp185, and Asp186) are located close to the 3 0 terminus of the primer strand. Observation of a hydrogen bond between the 3 0 -OH of the primer terminus and the side-chain of Asp185 suggests that the carboxylate of Asp185 could act as a general base in initiating the nucleophilic attack during polymerization. Nearly all of the close protein-dna interactions involve atoms of the sugar-phosphate backbone of the nucleic acid. However, the phenoxyl side-chain of Tyr183, which is part of the conserved YMDD motif, has hydrogen-bonding interactions with nucleotide bases of the second duplex base-pair and is predicted to have at least one hydrogen bond with all Watson-Crick base-pairs at this position. Comparison of the structure of the active site region in the HIV-1 RT/dsDNA complex with all other HIV-1 RT structures suggests that template-primer binding is accompanied by signi cant conformational changes of the YMDD motif that may be relevant for mechanisms of both polymerization and inhibition by non-nucleoside inhibitors. Interactions of the ``primer grip'' (the b12-b13 hairpin) with the 3 0 terminus of the primer strand primarily involve the main-chain atoms of Met230 and Gly231 and the primer terminal phosphate. Alternative positions of the primer grip observed in different HIV-1 RT structures may be related to conformational changes that normally occur during DNA polymerization and translocation. In the vicinity of the polymerase active site, there are a number of aromatic residues that are involved in energetically favorable p-p interactions and may be involved in the transitions between different stages of the catalytic process. The protein structural elements primarily responsible for precise positioning of the template-primer (including the primer grip, template grip, and helices ah and ai of the p66 thumb) can be thought of functioning as a ``translocation track'' Present address: A. Jacobo-Molina, Instituto Tecnolo gico de Monterrey, Campus Morelos, Paseo de la Reforma No. 182-A, Lomas de Cuernavaca, Cuernavaca, Morelos, MeÂxico. Abbreviations used: dntp, deoxynucleoside triphosphate; dsdna, double-stranded DNA; Fab28, antigen-binding fragment of monoclonal antibody 28; HIV-1, human immunode ciency virus type 1; MIR, multiple isomorphous replacement; NNIBP, non-nucleoside inhibitor-binding pocket; NNRTI, non-nucleoside reverse transcriptase inhibitor; NRTI, nucleoside reverse transcriptase inhibitor; RT, reverse transcriptase; SAS, saturated ammonium sulfate; MMLV, Moloney murine leukemia virus. address of the corresponding author: arnold@cabm.rutgers.edu 0022±2836/98/491095±17 $30.00/0 # 1998 Academic Press

2 1096 Implications of the Polymerase Active Site Region that guides the relative movement of nucleic acid and protein during polymerization. # 1998 Academic Press Keywords: AIDS; polymerase active site; polymerase structure; protein-nucleic acid interaction; X-ray crystallography Introduction Human immunode ciency virus type 1 (HIV-1) is the causative agent of AIDS. The reverse transcriptase (RT) of retroviruses, including HIV-1, has two enzymatic activities: a DNA polymerase that can copy either RNA or DNA templates and an RNase H (Jacobo-Molina & Arnold, 1991; Whitcomb & Hughes, 1992; Telesnitsky & Goff, 1997; Arts & Le Grice, 1998). HIV-1 RT has an essential role in viral replication and is the target of a number of HIV-1 inhibitors. Most RT inhibitors can be divided into two classes: nucleoside analog RT inhibitors (NRTIs) and non-nucleoside RT inhibitors (NNRTIs; reviewed by Larder, 1993; Tantillo et al., 1994; De Clercq, 1995, 1997; Arnold et al., 1996; Ding et al., 1997a; Emini &Fan 1997). Several of these inhibitors have been approved for the treatment of AIDS. However, the ef cacy of these drugs is limited both by serious side effects (toxicity) and rapid emergence of drug-resistant viral strains (see the reviews by De Clercq, 1994, 1995; Tantillo et al., 1994; Larder et al., 1995; Arnold et al., 1996; Ding et al., 1997a). Recently, treatment of patients with combinations of both HIV-1 RT and protease inhibitors has reduced the viral load in many HIV-infected individuals to undetectable levels (Cavert et al., 1997; Perelson et al., 1997). HIV-1 RT is a heterodimer consisting of p66 and p51, two subunits which share a common amino terminus. The p51 subunit corresponds to the polymerase domain of the p66 subunit. The carboxyl terminus of p66 forms the RNase H domain. A number of crystal structures of HIV-1 RT have been determined (Kohlstaedt et al., 1992; Jacobo-Molina et al., 1993; Ding et al., 1995a,b; Esnouf et al., 1995, 1997; Ren et al., 1995a,b; Rodgers et al., 1995; Das et al., 1996; Hopkins et al., 1996; Hsiou et al., 1996). Integration of biochemical and structural data for HIV-1 RT has yielded insights into many aspects of RT function. However, there is still a great deal to learn about HIV-1 RT structure-function relationships and the mechanisms of drug inhibition and drug resistance. This knowledge can be used to design or develop improved drugs to treat HIV-infected individuals. We report here the structure of the HIV-1 RT/ dsdna/fab28 complex re ned at 2.8 AÊ resolution. This re ned structure of the HIV-1 RT/dsDNA/ Fab28 complex permits detailed analysis of the structure of the polymerase active site and the speci c interactions between HIV-1 RT and the dsdna in the vicinity of the polymerase active site. Detailed analysis of the geometry of the bound dsdna and a summary of the overall protein-nucleic acid interactions has been reported (Ding et al., 1997b). Results and Discussion Overall structure The overall structure of the HIV-1 RT/dsDNA/ Fab28 complex is shown in Figure 1. The structure of HIV-1 RT in this complex is similar to that of unliganded HIV-1 RT and HIV-1 RT in the HIV-1 RT/NNRTI complexes. However, there are signi cant conformational changes that involve entire subdomains (in particular the thumb subdomain of p66), secondary structural elements, and/or individual amino acid residues. These changes are especially apparent in regions involved in interactions with the bound nucleic acid, inhibitors, or Fab28. The dsdna is bound in a cleft formed by the ngers, palm, and thumb subdomains of p66. The connection subdomains of both p66 and p51 form the `` oor'' of the DNA-binding cleft. The interactions between the dsdna and protein involve primarily the sugar-phosphate backbone of the nucleic acid and structural elements of the ngers, palm, thumb, and RNase Hof p66, and are not sequence-speci c (Figure 2 and Table 1). The bound dsdna has ahybrid structure (Ding et al., 1997b). The ve base-pairs near the polymerase active site have a conformation similar to A-form DNA; the nine base-pairs towards the RNase H active site have a conformation similar to B-form DNA. There is a signi cant bend involving the four base-pairs that join the A-form and B-form DNA portions; the helical axes of the A-form and B-form portions make an angle of about 41. The geometry of the four base-pairs in the bent region does not conform to either typical A-form or B-form DNA. The 5 0 -nucleotide overhang of the template strand is positioned near the center of the helix so that its base stacks with the nucleotide bases of the rst duplex base-pair (Pri1 and Tem1; the template-primer numbering scheme is given in Figure 2). The Fab28 fragment binds to a portion of the surface of the p51 palm subdomain and has no contacts with either the p66 subunit or the dsdna. The segment of p51 containing the Fab28 epitope corresponds to the b12-b13-b14 sheet of p66. This segment is disordered in structures in which HIV-1

3 Implications of the Polymerase Active Site Region 1097 Figure 1. Ribbon (Carson, 1987) diagram showing the overall structure of the HIV-1 RT/dsDNA/Fab28 complex. The subdomains of the p66 and p51 subunits of HIV-1 RT are colored as follows: ngers, blue; palm, red; thumb, green; connection, yellow; and RNase H, orange. The bound dsdna is shown with the template strand as a dark gray ribbon and the primer strand as a light gray ribbon; base-pairs are represented by bars. The monoclonal antibody fragment Fab28 is shown with the light chain in light gray and the heavy chain in dark gray. RT is not bound to Fab28, but is well ordered in the present structure. The secondary structure of this segment of p51 is completely rearranged relative to its structure in the p66 subunit (which forms part of the ``primer grip''; Jacobo-Molina et al., 1993). In the complex with Fab28 this segment adopts an extended conformation. Structure of the polymerase active site The polymerase active site, which contains three conserved aspartic acid residues (Asp110, Asp185, and Asp186), is located in the p66 palm and lies at the base of the DNA-binding cleft. Asp185 and Asp186 form part of the YMDD motif of HIV-1 RT, corresponding to the more general YXDD motif (X ˆMet (in HIV), Val, Leu, or Ala), which is highly conserved in retroviral RTs (Johnson et al., 1986). Analogous carboxylate amino acids have been identi ed in sequences of all known RNAdependent and DNA-dependent polymerases (Kamer & Argos, 1984; Argos, 1988; Poch et al., 1989; Delarue et al., 1990). Substitution of any of these three aspartate residues with other amino acids (including glutamate) dramatically impairs catalytic activity, indicating that the polymerization ef ciency of HIV-1 RT is critically dependent on the precise geometry of the polymerase active site (Larder et al., 1989; Le Grice et al., 1991; Boyer et al., 1992; Hostomsky et al., 1992; Kaushik et al., 1996). In the present structure, these three aspartate residues are proximal to the 3 0 -terminal nucleotide of the primer strand (Figure 3). The O d1 atom of Asp185 forms a hydrogen bond with the 3 0 -OH group of Pri1 (d ˆ 3.3 AÊ and the O-H... O d1 angle ˆ 109 ). Although no metal ions were seen in this crystal structure, the polymerase of HIV-1 RT requires divalent metal ions (preferably Mg 2 )as cofactors and it has been proposed that there are two Mg 2 bound to the active site aspartate residues in the active form of HIV-1 RT. The O d2 atom of Asp186 is proximal to the O2P atom of the 3 0 -terminal phosphate (3.8 AÊ ). Interaction between these oxygen atoms may be mediated via either a metal ion or a water molecule. Such a water molecule might also be a ligand for one metal ion at the polymerase active site. Analogous indirect interactions between a carboxylate oxygen atom of Asp186 and a phosphate oxygen atom of the 3 0 -terminal nucleotide of the primer strand appear to exist in other polymerase structures. In the crystal structure of a ternary complex of rat DNA polymerase b with a ddctp substrate and a DNA template-primer, Asp256 at the polymerase active site is close to the phosphate oxygen atom of the 3 0 -terminal nucleotide of the primer strand (3.6 AÊ ; Pelletier et al., 1994). There is room between Asp256 of rat DNA polymerase b and the 3 0 -terminal nucleotide to accommodate a water molecule in a position where it could form hydrogen-bonding interactions with a carboxylate oxygen atom of Asp256, a phosphate oxygen atom of the 3 0 -terminal nucleotide of the primer strand, and a metal ion (Pelletier et al., 1994). In the structure of at7 DNA polymerase ternary complex with a ddntp substrate and a DNA template-primer, the side-

4 1098 Implications of the Polymerase Active Site Region Figure 2. Sequence and numbering scheme of the dsdna template-primer and interactions between the dsdna and nearby amino acid residues of HIV-1 RT. Nucleotides of the template strand and the primer strand are designated Tem and Pri, respectively. The positions of nucleotides are numbered from the polymerase active site toward the RNase H active site; the template overhang is numbered negatively from the polymerase active site. For the amino acid residues, the subdomains and secondary structural locations are given with the abbreviation of subdomains as: p66f, p66 ngers; p66p, p66 palm; p66t, p66 thumb; p66c, p66 connection; and p51c, p51 connection. Interactions with distances shorter than 3.3 AÊ are indicated by thick continuous lines; interactions with distances between 3.3 and 3.6 AÊ, thin continuous lines; interactions with distances between 3.6 and 3.8 AÊ, broken lines. Amino acid residues involved in hydrogenbonding interactions with nucleotides are marked with an asterisk. Due to poor electron density, the side-chains of Arg284 and Thr286 could not be placed with con dence and these residues were modeled as alanine residues. chain of one of the catalytic carboxylate residues (Glu655) is positioned so that it could have a similar indirect interaction with the 3 0 -terminal phosphate of the primer strand (Doublie et al., 1998). However, the same phosphate oxygen atom makes a favorable hydrogen-bonding interaction with the side-chain of His704 in the T7 DNA polymerase complex structure. Table 1. Nucleic acid-protein interactions (d AÊ ) Template Protein Distance (AÊ ) Primer Protein Distance (AÊ ) Tem-1 a Ade C2 0 Gly152 C a 3.5 Pri1 Ade P Met230 C a 3.8 C4 0 C a 3.4 O2P N 3.1 b O4 0 N 3.4 O2P C a 3.3 O4 0 C a 3.5 C2 0 Asp185 O d1 3.5 O4 0 Gln151 C b 3.8 C5 0 Tyr183 C d2 3.7 O4 0 C 3.7 C5 0 Met230 C e 3.5 C1 0 Gly152 N 3.3 C4 0 Tyr183 C d2 3.8 C1 0 C a 3.4 O4 0 C e2 3.7 C3 0 C a 3.7 C1 0 Met184 C g 3.7 O3 0 C a 3.3 O3 0 Asp185 C g 3.8 Tem1 Thy O2P Asp76 O d2 3.7 O3 0 O d1 3.3 b C5 0 Gly152 O 3.2 O3 0 Asp186 C b 3.7 C5 0 C a 3.6 Pri2 Cyt O2P Met230 O 3.6 C5 0 C 3.5 O2P Gly231 N 3.6 C4 0 O 3.6 O2P C a 2.8 C4 0 C a 3.7 O2 Tyr183 O Z 3.5 b C4 0 C 3.4 C2 0 Met230 C e 3.5 O4 0 N 3.7 C5 0 C 3.7 continued

5 Implications of the Polymerase Active Site Region 1099 Table 1ÐContinued Template Protein Distance (AÊ ) Primer Protein Distance (AÊ ) O4 0 C a 3.1 C5 0 O 3.0 O4 0 C 3.4 C4 0 C e 3.1 O3 0 Lys154 C a 3.7 C4 0 C 3.8 Tem2 Gua N2 Tyr183 O Z 3.7 b C4 0 N e2 3.7 C5 0 Lys154 C a 3.8 O4 0 C e 3.5 O4 0 Pro157 C b 3.7 C1 0 C e 3.6 O4 0 C g 3.4 C3 0 C e 3.4 O3 0 Glu89 C d 3.4 O3 0 C e 3.0 O3 0 O e2 3.6 b O3 0 C a 3.0 O3 0 C g 3.4 O3 0 C 3.2 Tem3 Gua P C g 3.6 O3 0 O 3.7 O2P C g 3.4 O3 0 Gly231 N 3.6 O5 0 C g 3.3 Pri3 Cyt O2P Lys263 N z 3.3 b Tem6 Cyt O3 0 Asn265 N d2 3.3 b O2P C e 3.2 O3 0 Lys353 N z 3.5 b C5 0 Trp266 C e2 3.8 O3 0 Asn265 C g 3.7 C5 0 C g 3.6 O3 0 O d1 3.2 b C5 0 C d2 3.7 Tem7 Cyt P Lys353 N z 3.0 C5 0 C d1 3.6 O1P N z 3.0 b C5 0 N e1 3.7 O2P C e 3.5 C4 0 C e2 3.3 O2P N z 2.5 b C4 0 C z2 3.6 C5 0 Asn265 N d2 3.7 C4 0 C d2 3.5 O3 0 Ser280 C b 3.8 C4 0 N e1 3.6 Tem8 Cyt O2P C 3.5 O4 0 C e3 3.8 O2P O 3.2 b O4 0 C z3 3.8 O2P C b 3.6 O4 0 C d2 3.8 O3 0 Gly285 N 3.6 b Pri4 Gua C5 0 Lys259 C a 3.7 O3 0 Arg284 c C a 3.5 C5 0 C 3.7 O3 0 C b 3.8 C5 0 O 3.0 Tem9 Gua P Gly285 N 3.3 C5 0 Lys263 N 3.8 O1P N 3.1 b C5 0 C g 3.5 O1P C a 3.6 C4 0 N 3.4 O2P N 3.1 b C4 0 C g 3.6 O2P C a 3.6 O4 0 Gly262 C a 3.7 O2P Thr286 c O 3.7 O4 0 C 3.8 O4 0 Lys263 N 3.8 O3 0 C g 3.7 O3 0 C e 3.4 Pri5 Cyt C5 0 Gln258 N e2 3.4 C5 0 C b 3.7 C4 0 C b 3.6 O3 0 Lys259 C a 3.7 Pri6 Gua C2 0 Gln258 N e2 3.1 C4 0 N e2 3.7 O4 0 N e2 3.5 b C1 0 N e2 2.9 C3 0 N e2 3.8 O3 0 N e2 3.7 Pri10 Thy O2P Lys395 d C e 3.7 Pri11 Thy O1P Arg358 N E 3.6 b O1P Glu396 d O e1 3.7 O1P O e2 3.3 b Pri12 Gua O1P Ala360 N 3.6 b O2P N 3.8 Pri13 Thy O1P Ala360 C b 3.6 O2P Tyr501 C e1 3.5 O2P O Z 3.4 b O2P His361 N d1 3.8 O2P C e1 3.3 O2P Ile505 C d1 3.2 Pri14 Cyt P Thr473 O g1 3.7 O1P O g1 3.1 b O1P Lys476 C g 3.5 O2P Gln475 C b 2.9 O2P O e1 3.6 b O2P C g 3.8 Pri15 Cyt O3 0 Thr473 O g1 3.6 b O3 0 Gln475 O e1 3.2 b a The nucleotide numbering scheme for the template is given in Figure 2. b These atoms are involved in potential hydrogen-bonding interactions. c Arg284 and Thr286 were modeled as alanine residues due to poor electron density for the side-chains. d Residues of the p51 subunit.

6 1100 Implications of the Polymerase Active Site Region Figure 3. (a) Structure of the polymerase active site region of HIV-1 RT including the primer grip. Secondary structural elements of the p66 palm subdomain are shown as red ribbons. The three catalytically essential aspartic acid residues (Asp110, Asp185, and Asp186) are shown with cyan sidechains. Tyr183 and Met184, which form part of the conserved YMDD motif, are shown with gold sidechains. Amino acid residues at the primer grip are shown in green. The dsdna is shown with the template strand in dark gray and the primer strand in light gray. (b) A schematic diagram showing interactions between the 3 0 -terminal nucleotide of the primer strand (Pri1) and amino acid residues at the polymerase active site, with selected distances given in AÊ. Hydrogen-bonding interactions between the side-chain O d1 atom of Asp185 and the 3 0 -OH of Pri1, and between the amide nitrogen atom of Met230 of the primer grip and the phosphate oxygen atom of Pri1 are indicated by heavy lines. Universal hydrogen bonds between Tyr183 and nucleotide bases in the minor groove Tyr183 and Met184 of the conserved YMDD motif are also involved in interactions with the 3 0 terminus of the primer strand. The side-chains of both Tyr183 and Met184 have hydrophobic interactions with the ribose of Pri1. In addition, the O Z atom of Tyr183 has potential hydrogen-bonding interactions with the cytosine O2 atom of Pri2 (d ˆ 3.5 AÊ and the O Z -H... O2 angle ˆ 130 ) and possibly the guanine N2 atom of Tem2 (d ˆ3.7 AÊ and the O Z -H...N2 angleˆ151 ;Figures 3(a) and 4). These interactions are the only close contacts (43.8 AÊ ) between HIV-1 RT and the template-primer involving atoms of the nucleotide bases. Straightforward extrapolation suggests that analogous hydrogen bonds can be formed between the O Z atom of Tyr183 and the bases of nucleotides in the minor groove of the template-primer with any of the Watson-Crick base-pairs (Figure 4). Figure 4 shows that substitution of the CG base-pair with ata base-pair would permit the Tyr183 O Z atom to form a hydrogen bond with the O2 atom of thymine. Similarly, with a UA base-pair the O Z atom of Tyr183 can also form a hydrogen bond with the O2 atom of uracil. In the case of GC, AT, and A U base-pairs, analogous hydrogen-bonding interactions can be maintained, since the relative geometry of the atoms involved in hydrogen bonds would be similar. Thus, we predict that Tyr183 can form at least one hydrogen bond with any Watson-Crick base-pair (one for A T, T A, AU, or UA base-pairs, and two for CG orgc base-pairs). This model also predicts that loss of the hydrogen-bonding interaction would lead to a reduction in the af nity of HIV-1 RT for any template-primer, which would be expected to directly impact the biochemical properties of the enzyme, including delity and processivity. Replacing Tyr183 with phenylalanine causes a substantial reduction in HIV-1 RT polymerization

7 Implications of the Polymerase Active Site Region 1101 (Figure 3(b)). It is possible that the relatively large tyrosine and phenylalanine side-chains provide a steric barrier that participates in the precise positioning of the primer 3 0 -terminal nucleotides for catalysis. In addition, the ability of Tyr183 to form hydrogen bonds with atoms of the nucleotide base-pair in the minor groove might play an important role in precisely positioning the template-primer relative to the polymerase active site during translocation of the template-primer following incorporation of a dntp substrate. As the template-primer translocates, the hydrogen bonds between Tyr183 and the second nucleotide basepair would be weakened, then broken. During this process, Tyr183 could reorient its phenoxyl sidechain by sterically feasible alterations of the w 1 torsion angle and form new hydrogen bond(s) with the new (second) nucleotide base-pair. This process could facilitate a ratchet-type motion of the template-primer that has been hypothesized for translocation of polymerases (Georgiadis et al., 1995; Patel et al., 1995; Pelletier et al., 1996). Figure 4. Predicted hydrogen-bonding interactions between the O Z atom of Tyr183 of the conserved YMDD motif and the nucleotide base atoms in the minor groove of the template-primer. The side-chain of Tyr183 can form a potential hydrogen bond(s) with the second nucleotide base-pair of the duplex region in the minor groove. Shown are the interactions with a cytosine guanine base-pair observed in this structure (top) and with a thymineadenine base-pair predicted by extrapolation (bottom). With CG and GC base-pairs, the O Z atom of Tyr183 can form two hydrogen-bonding interactions with the O2 atom of cytosine and the N2 atom of guanine. With TA, AT, and UA base-pairs, the O Z atom of Tyr183 is predicted to form one hydrogen bond with the O2 atom of thymine or uracil. ef ciency (HIV-1 RT containing a Tyr183Phe mutation has 20-30% of the activity of wild-type enzyme; Boyer et al., 1992; Chao et al., 1995). The aromatic character of this residue also seems to be critical, since replacing Tyr183 with non-aromatic amino acids causes an even more dramatic impairment of polymerase activity (Larder et al., 1987; Chao et al., 1995; Bakhanashvili et al., 1996). In the present structure, the aromatic ring atoms C d2 and C e2 of Tyr183 (which are present also in phenylalanine) have van der Waals interactions with the ribose ring atoms O4 0, C4 0, and C5 0 of Pri1 Structural changes at the YMDD motif that accompany DNA binding The main-chain of the YMDD motif at the polymerase active site has an unusual b-turn conformation (type II 0 ), which has been seen in other HIV-1 RT structures (Ding et al., 1995a; Esnouf et al., 1995; Ren et al., 1995a; Rodgers et al., 1995; Hsiou et al., 1996) and Moloney murine leukemia virus (MMLV) RT (Georgiadis et al., 1995). It has been suggested that the unusual geometry of the b9-b10 turn might be required to precisely position Asp185 and Asp186 for catalysis (Rodgers et al., 1995). An implication of this twin geometry is that the side-chains of all four amino acid residues in the YMDD motif are exposed on the same side of the b sheet. The same type II 0 b-turn geometry has also been found at the analogous position in the active site structures of a Bacillus DNA polymerase (a member of the pol Iclass; Kiefer et al., 1997), T7 DNA polymerase (Doublie et al., 1998), and the poliovirus RNA-dependent RNA polymerase (Hansen et al., 1997). Met184, which lies in the b9-b10 turn, has an energetically unfavorable main-chain conformation (torsion angles f 54 and j 107 ). In unliganded and NNRTI-bound HIV-1 RT structures, the unusual main-chain conformation of Met184 is stabilized by a hydrogen bond between its carbonyl O atom and either the N e2 atom of Gln182 (Ding et al., 1995b; Ren et al., 1995a; Rodgers et al., 1995; Hsiou et al., 1996) or the N e2 atom of Gln161 (Esnouf et al., 1995). In the current structure, the N e2 atom of Gln182 is repositioned and hydrogen bonded with the O g1 atom of Thr165; there is no hydrogen bond between Met184 and Gln182. The side-chain of Met184 has close contacts with the ribose of the 3 0 -terminal nucleotide of the primer strand (Figure 3), and the side-chains of Tyr115 and Phe160, both of which

8 1102 Implications of the Polymerase Active Site Region are probably involved in dntp binding. Biochemical and clinical data show that HIV-1 RT variants containing Met184Val and Met184Ile mutations are resistant to some but not all of the clinically important dideoxynucleoside analogs (Gu et al., 1992; Gao et al., 1993; Schinazi et al., 1993; Tisdale et al., 1993). The conformation of YMDD at the polymerase active site in the DNA-bound HIV-1 RT structure is signi cantly different from its conformation in the unliganded or NNRTI-bound HIV-1 RT structures (Figure 5). When b9 and b10 (residues and ) of the unliganded or NNRTIbound HIV-1 RT structures were superimposed onto those of the DNA-bound HIV-1 RT structure, the maximum rms deviation was found to be less than 0.8 AÊ. However, the C a positions of the amino acid residues of YMDD (residues ) differed on average by 1.6, 2.1, 1.5, and 1.3 AÊ, respectively (Figure 5). The position of YMDD in the present structure is ``pushed'' down and away from the template-primer when compared with its position in all of the other HIV-1 RT structures, suggesting that conformational changes take place at YMDD upon binding of the template-primer substrate and that some degree of exibility of the structural segments that contain the polymerase active site is required during DNA polymerization. The conformational mobility of the YMDD motif may have implications for the mechanism of inhibition by NNRTIs, which is still not fully understood (reviewed by Sara anos et al., 1997). The amino acid residues involved in NNRTI binding and in NNRTI resistance are located near the polymerase active site. NNRTI binding induces conformational changes in and around the NNIBP, which include repositioning of the primer grip and reorientation of the side-chains of Tyr181 and Tyr188 toward the polymerase active site. The precise stereochemistry of the polymerase active site, metal ions, dntp, and the template-primer is essential for ef cient catalysis; perturbation of the position or restriction of the mobility of any of these elements could severely impair polymerase activity. Even though binding of an NNRTI may cause structural distortion of the YMDD motif that is smaller in an absolute sense than the movements of the p66 thumb or the primer grip, since residues of the YMDD motif are directly involved in DNA polymerization, distortion and/or restriction of the mobility of this motif could explain why NNRTI binding appears to affect primarily the chemical step of DNA polymerization (Rittinger et al., 1995; Spence et al., 1995, 1996). None of the NNRTIbound HIV-1 RT structures described to date contain either template-primer or dntp; a structure of an NNRTI-bound HIV-1 RT in the presence of template-primer and dntp should be most helpful in elucidating the details of inhibition of HIV- 1 RT by NNRTIs. Structure of the primer grip and its interactions with the dsdna The b12-b13 hairpin of the antiparallel threestranded b-sheet (b12-b13-b14) in the p66 palm subdomain is in close proximity to nucleotides at the 3 0 -primer terminus (Figure 3). This structural element is involved in positioning the 3 0 -primer terminus relative to the polymerase active site, and has been called the ``primer grip'' (Jacobo-Molina et al., 1993). Mutagenesis experiments demonstrated that substitution of alanine residues in the b12-b13 hairpin (Trp229, Met230, Gly231, and Tyr232) in p66 alters both DNA polymerase and RNase Hactivities (Jacques et al., 1994; Ghosh et al., 1996; Palaniappan et al., 1997). In the present structure, Met230 and Gly231 of the b12-b13 turn have extensive interactions with nucleotides of the 3 0 -primer terminus (Figure 3and Table 1). There are 18 close contacts between Met230 and the sugar-phosphate backbone of Pri1 and Pri2, and three close contacts between Gly231 and the phosphate group of Pri2 (d AÊ ). The amide nitrogen atom of Met230 forms a hydrogen bond with the O2P atom of the phosphate group of Pri1 (d ˆ 3.1 AÊ and the N-H... O2P angle ˆ 144 ). The extensive interactions of the primer grip with the 3 0 -primer terminus primarily involve the mainchain atoms of Met230 and Gly231 and are relatively non-speci c van der Waals contacts (Figure 3 and Table 1). During DNA polymerization and translocation, exibility of both the primer strand and the primer grip is likely to be important for optimal positioning of the primer terminus relative to the polymerase active site. The fact that most of the contacts are relatively weak van der Waals interactions between the primer grip and the primer strand might allow the enzyme to move relative to the primer strand during DNA translocation. Although many of the amino acid residues in the primer grip are highly conserved among retroviral RTs, hydrophobic residues at positions 229, 232, and 234 seem to be required for activity, with strong preference for Trp, Tyr, and Leu, respectively (Xiong &Eickbush, 1990; Ding et al., 1995b). No mutations have been reported at these positions in either tissue culture or clinical isolates. In fact, substitutions of Trp229 and Tyr232 disrupt enzymatic activity of HIV-1 RT (Jacques et al., 1994; Ghosh et al., 1996). Structural analysis shows that there are a number of aromatic residues in the vicinity of the polymerase active site and the primer grip, including Tyr181, Tyr183, and Tyr188 (in the b9-b10 hairpin), and Phe227, Trp229, Tyr232, and Trp239 (in the b12-b13-b14 sheet), where the aromatic side-chains are involved in energetically favorable p-p interactions (Table 2). The aromatic side-chain of Trp266 (ah) of the p66 thumb subdomain is also involved in p-p interactions with these aromatic residues, especially with Tyr232 (Table 2). Replacing Trp266 of helix ah with alanine has only a modest effect on the rate of polymerase cat-

9 Implications of the Polymerase Active Site Region 1103 Figure 5. Conformational change of the YMDD motif that accompanies template-primer binding to HIV-1 RT. (a) Superpositions of unliganded and NNRTI-bound HIV-1 RT structures onto the DNA-bound HIV-1 RT structure at the b9-b10 hairpin (residues were used as the basis for superposition). The structures shown are: HIV-1 RT/ dsdna/fab28, red; unliganded HIV-1 RT (Hsiou et al., 1996), light gray; unliganded HIV-1 RT (Rodgers et al., 1995), gold; HIV-1 RT/8-Cl TIBO (Ding et al., 1995a), green; HIV-1 RT/a-APA (Ding et al., 1995b), cyan; HIV-1 RT/a-APA (Ren et al., 1995a), dark gray; and HIV-1 RT/nevirapine (Smerdon et al., 1994), blue. The position of YMDD in the DNA-bound HIV-1 RT structure is different from that in all other HIV-1 RT structures and is pushed down relative to the template-primer, suggesting that the conformational change of the YXDD motif is caused by binding of the template-primer. (b) Graph showing the displacements of C a atoms of amino acid residues at the b9-b10 hairpin in unliganded and NNRTI-bound HIV-1 RT structures relative to those in the DNA-bound HIV-1 RT structure. Although the rms deviations between the DNA-bound HIV-1 RT and other HIV-1 RT structures in the b9-b10 hairpin region are less than 0.8 AÊ, displacements of equivalent C a atoms of the amino acid residues of YMDD (residues ) are (on average) 1.6, 2.1, 1.5, and 1.3 AÊ, respectively.

10 1104 Implications of the Polymerase Active Site Region alysis or on dntp binding, but HIV-1 RT with this substitution has greatly increased dissociation rates for DNA, substantially reduced processivity, and lower frameshift delity (Beard et al., 1994; Bebenek et al., 1995). Substitution of Trp266 with threonine affects RNase H activity and strand transfer ef ciency (Gao et al., 1998). In addition to having extensive p-p interactions with surrounding aromatic residues, the side-chain of Trp229 also has close contacts with the side-chain of Asp186 (the shortest distance is 3.4 AÊ ), and the side-chain of Met230 makes a number of hydrophobic contacts with the side-chain of Tyr183 (the shortest distance is 3.3 AÊ ). The network of hydrophobic interactions between these residues appears to link the polymerase active site, the primer grip, and the p66 thumb together, which could play an important role in stabilizing the primer grip and the polymerase active site. These residues may also be involved in communicating the stages of the catalytic cycle from one site to another, and hence, may have the potential to affect enzymatic activities either directly or indirectly. Biological implications of alternative positions of the primer grip Evidence to support the importance of primer grip mobility during DNA polymerization comes from the observation that this structural segment can adopt substantially different positions and conformations in different HIV-1 RT structures. Except for local differences in the conformation of the YMDD motif (discussed above), the p66 palm subdomain of the DNA-bound HIV-1 RT can be superimposed quite well onto that of unliganded HIV-1 RT (Hsiou et al., 1996). Superposition of these two structures based on residues 155 to 215 (the core structural elements of the p66 palm subdomain, ae, b9, b10, and af) gives an rms deviation of 0.7 AÊ for 61 C a atoms of the core elements and a maximum positional difference of 1.0 AÊ for C a atoms of amino acid residues at the primer grip. A similar comparison of structures of HIV-1 RT/ NNRTI complexes with the DNA-bound HIV-1 RT structure indicates that although the position and conformation of the core structural elements of the palm subdomain in HIV-1 RT/NNRTI complexes are similar to those in the HIV-1 RT/dsDNA/ Fab28 complex (an rms deviation of 1.0 AÊ for 61 C a atoms of the core elements, based on HIV-1 RT complexes with a-apa (Ding et al., 1995b) and TIBO (Ding et al., 1995a)), amino acid residues of the b12-b13 hairpin (which contains the primer grip) exhibit signi cantly larger displacements of 3.5 to 5.0 AÊ. Binding of NNRTIs induces conformational changes, including rotations of Tyr181 and Tyr188 side-chains and displacement of the b12-b13-b14 sheet; these movements are required to create the hydrophobic pocket where the inhibitors bind (Ding et al., 1995a,b; Ren et al., 1995a,b; Rodgers et al., 1995). In NNRTI-bound HIV-1 RT structures, reorientation of the Tyr181 side-chain is associated with displacement of Trp229, which is pushed outward relative to the polymerase active site. Displacement of Trp229 is coupled with the conformational change of the b12-b13-b14 sheet, including the primer grip. Repositioning of the b12-b13-b14 sheet (the primer grip) probably alters the positioning of the template-primer, especially the 3 0 -OH group of the primer terminus, relative to the polymerase active site. NNRTI binding also causes the p66 thumb subdomain to adopt a hyper-extended conformation in which the thumb is further away from the p66 ngers than the p66 thumb in the DNA-bound HIV-1 RT structure (Hsiou et al., 1996). Repositioning of the b12-b13- b14 sheet appears to be coupled with the displacement of the p66 thumb subdomain, and these two conformational changes could be largely responsible for the effects of NNRTI binding on template- Table 2. Centroid-to-centroid distances for pairs of residues whose aromatic side-chains have potential p-p interactions in the vicinity of the polymerase active site and the primer grip for three different types of HIV-1 RT structures Centroid-to-centroid/shortest C-C distances (AÊ ) Residue 1 Residue 2 RT/dsDNA Unliganded RT a RT/NNRTI b Tyr181 Tyr183 (10.3/8.2) d (10.3/7.9) 6.4/4.2 Tyr / / /5.0 Trp / / /4.0 Tyr183 Trp229 (8.4/6.7) (8.5/6.7) 6.6/5.2 Tyr188 Phe227 (7.6/6.1) (7.7/6.4) 6.3/4.6 Trp / /4.3 (7.1/5.8) Trp229 c Tyr / / /4.2 Tyr232 Trp / / /3.7 Trp /4.1 (9.4/7.4) 5.3/3.8 a The unliganded HIV-1 RT structure is Hsiou et al. (1996), PDB entry 1DLO. b The NNRTI-bound HIV-1 RT structure is Ding et al. (1995b), PDB entry 1HNI. c Only the six-membered rings of Trp229, Trp239, and Trp266 were used in the calculation. d Where the entry is parenthesized, the centroid-to-centroid distance between the corresponding aromatic side-chains is greater than 7AÊ (Burley &Petsko, 1985) and the shortest C-C distance is greater than 4.8 AÊ (Hunter et al., 1991), therefore, these pairs of residues are not considered to have energetically signi cant p-p interactions.

11 Implications of the Polymerase Active Site Region 1105 primer af nity, processivity of polymerization, and RNase Hcleavage speci city (Hsiou et al., 1996). The displacement of the primer grip could also explain biochemical data obtained from pre-steady state kinetic experiments which show that NNRTI binding blocks the chemical step of HIV-1 RT polymerization (Rittinger et al., 1995; Spence et al., 1995). According to this model (Ding et al., 1995b; Smith et al., 1995; Das et al., 1996; Sara anos et al., 1997), NNRTI binding displaces the primer grip, that disturbs the position and/or mobility of the 3 0 terminus of the primer strand at the polymerase active site and/or alters the stereochemical relationship between the primer terminus, a bound dntp, and associated divalent metal cation(s). In this model, an incoming dntp could still bind to an HIV-1 RT/NNRTI complex without signi cant alteration of af nity, but the presence of the NNRTI would alter the relative positions of the incoming dntp and the 3 0 -OH group of the primer strand. Indirect evidence supporting this model is provided by biochemical experiments showing that mutations across the DNA-binding cleft at amino acid positions believed to interact with the template-primer substrate can affect NNRTI binding (Kew et al., 1996) or restore an impairment in polymerase activity caused by mutations in the NNIBP (Kleim et al., 1996; Boyer et al., 1998). It seems likely that the communication between those distant sites is mediated via interactions with the nucleic acid substrate (Boyer et al., 1998). Other models for NNRTI inhibition mechanisms have been proposed (Kohlstaedt et al., 1992; Esnouf et al., 1995; Sara anos et al., 1997); further structural studies of HIV-1 RT bound with NNRTIs, templateprimer, and dntps are likely to help resolve which mechanism(s) is most likely to be applicable. There is a relatively large movement (4-5 AÊ )of the nucleic acid relative to the protein during translocation following nucleotide incorporation. Although it is not yet known what structural elements of protein and/or nucleic acid participate in this movement, the primer grip is a good candidate. It is possible that some portions of the protein (for example, the primer grip) move in concert with the DNA. The position of the primer grip in the HIV-1 RT/NNRTI complex structures is displaced approximately one nucleotide along the primer strand relative to its position in the HIV-1 RT/ dsdna complex. It is possible that the alternative position of the primer grip observed in the HIV-1 RT/NNRTI complexes may be related to a state that is used in the normal catalytic cycle. Repositioning of the primer grip back to the position observed in the HIV-1 RT/dsDNA complex might permit the enzyme to ``reload'' for the relative movement of nucleic acid and protein that accompanies the catalytic cycle. A translocation track in reverse transcriptases The network of contacts between HIV-1 RT and the dsdna template-primer (Figure 2) in the current structure involves several groups of amino acid residues on the surface of the protein, including residues of the polymerase active site, the primer grip, the template grip, the helices ah and ai of the p66 thumb, some portions of the p66 and p51 connection subdomains, and RNase H. These groups of amino acid residues can be thought of as modular elements for controlling relative movements of polymerase and the template-primer. We propose that a useful name for this entire collection of contacting elements might be ``translocation track'' and that translocation can be accomplished via coordinated relative movements of individual modules. The net result of a complete cycle of polymerase translocation is that a network of contacts must be broken and reformed. Some of the protein elements involved in this process might move relative to each other, and some of the contacts would be broken and remade at different times. For example, as was discussed above, comparisons of different HIV-1 RT structures have indicated that the primer grip (b12-b13) and the p66 thumb can adopt distinctly different positions relative to the essential aspartate residues and the catalytic site. Nucleotide incorporation would drive translocation of the primer strand via repositioning of the primer grip/ thumb contact network, followed by sliding of the template-primer relative to the active site and the template grip. The cycle would be completed by having the primer grip snap back into the original position. An implication of having several cooperating regions is that a ratchet-type mechanism can be envisioned in which template-primer handling can be controlled by the interconnected modules of the translocation track. The energy for the movement would be provided by cleavage of the incoming dntps (Patel et al., 1995). Bebenek et al. (1997) proposed the existence of a ``minor groove binding track'' in HIV-1 RT. Their molecular dynamics and energetic calculations were based on a modeled extension of the RT/ DNA complex described by Jacobo-Molina et al. (1993). Bebenek et al. (1997) suggested that there were several speci c interactions of HIV-1 RT residues with the minor groove of the template-primer; their proposals, however, are not consistent with the experimental structure presented here. In particular, the speci c predictions made by Bebenek et al. (1997) for contacts of residues Ile94, Gln258, Gly262, Trp266, and Gln269 with the template-primer are incorrect and not observed in the crystal structure. For example, the only signi cant interactions of HIV-1 RT residues with nucleotide bases of the template-primer involve Tyr183, and hydrogen-bonding interactions of Gln258 and Trp266 side-chains with nucleotide bases predicted by Bebenek et al. (1997) do not occur in the actual structure. The erroneous predictions based on the dynamics and modeling calculations probably re ect the dif culty in accurately simulating a structure of this complexity. In the molecular dynamics calculation, the C a positions of residues

12 1106 Implications of the Polymerase Active Site Region in the complex were permitted to vary from the experimentally described starting model; the C a positions of residues involved in template-primer interactions deviated 1.6 to 2.8 AÊ from their original positions. We believe that the minor groove binding track is an inappropriate term because it does not accurately re ect the nature of the contact network between HIV-1 RT and template-primer. Although the proposal of the existence of a translocation track is based on analysis of interactions between HIV-1 RT and the template-primer, there will probably be homologous structural elements in all reverse transcriptases. It is likely that most polymerases will have an analogous overall set of cooperating elements to enhance the correct direction of product formation and the ef ciency of nucleic acid translocation during polymerization (Guajardo &Sousa, 1997). Materials and Methods Crystallization and diffraction data collection Preparation and puri cation of HIV-1 RT, 19-mer/18- mer dsdna template-primer, and Fab28 have been described by Jacobo-Molina et al. (1991) and Clark et al. (1995). The sequence of the 19-mer template strand is 5 0 -ATGGCGCCCGAACAGGGAC-3 0 in which the 5 0 -A forms a one base overhang. The sequence of the primer strand was chosen to mimic the primer-binding site of HIV-1 genome and the 3 0 terminus of human t-rna Lys,III, which is used by HIV-1 RT as a primer to initiate reverse transcription in vivo. Crystallization of the HIV-1 RT/ dsdna/fab28 complex was carried out as described by Jacobo-Molina et al. (1991) and Clark et al. (1995). Crystals of the HIV-1 RT/dsDNA/Fab28 complex have the symmetry of space group P with unit cell dimensions a ˆ b ˆ and c ˆ AÊ. An asymmetric unit contains one HIV-1 RT/dsDNA/Fab28 complex and has a molecular mass of 180 kda, corresponding to a V M of 5.03 AÊ 3 /Da and a solvent content of 75.6% (assuming the standard partial speci c volume for protein and DNA is 0.74 ml/g (Matthews, 1968)). Low resolution X-ray diffraction data were collected using the Mark II Multiwire System (Hamlin, 1985; Howard et al., 1985) in our local X-ray laboratory. High resolution X-ray diffraction data were collected at the Cornell High Energy Synchrotron Source (CHESS) F1 beamline. Diffraction data from cooled crystals ( 10 C) were processed and scaled together from images measured from multiple crystals using a modi ed version of the Purdue oscillation lm processing package (G. Kamer & E.A., unpublished results). Diffraction data from frozen crystals ( 165 C) were processed and scaled together from images measured from a single crystal using DENZO (Minor, 1993; Otwinowski &Minor, 1996) and SCALEPACK (Otwinowski &Minor, 1996). For data collection at 165 C (frozen), crystals were soaked in a synthetic cryoprotective solution containing 40% (w/v) SAS, 20% (w/v) glucose, and 10% (v/v) glycerol for 20 minutes, and were then mounted using a nylon loop and frozen directly in the cold gaseous N 2 stream ( 165 C; Teng, 1990; Rodgers, 1994). In some experiments, crystals were soaked in three steps (20 minutes each) with cryoprotective solutions containing 40% SAS, 10% glucose; 40% SAS, 10% glucose, 10% glycerol; and 40% SAS, 20% glucose, 10% glycerol, respectively. The native dataset used in the structure determination and re nement of the HIV-1 RT/dsDNA/Fab28 complex was collected at the CHESS F1 beamline and merged from 186 images measured at 10 C from 36 crystals. This dataset contains 76,187 unique re ections (I 5 2s(I)) to 2.8 AÊ resolution with an R merge (I) of 0.13 and a completeness of 86.6%. Two additional datasets measured from crystals frozen at 165 C at the CHESS F1 beamline were useful in multiple crystal form averaging for phasing improvement (dataset 1: a ˆ b ˆ AÊ, c ˆ AÊ, 3.3 AÊ resolution, R merge (I) ˆ 0.12, 47,494 unique re ections, and 91.0% complete; dataset 2: a ˆ b ˆ AÊ, c ˆ AÊ, 3.0 AÊ resolution, R merge (I) ˆ 0.10, 49,899 unique re ections, and 71.7% complete). Structure determination and refinement The initial solution of heavy-atom positions for high resolution heavy-atom derivatives was obtained by interpretation of isomorphous difference Patterson syntheses (Arnold et al., 1992). Initial multiple isomorphous replacement (MIR) phased electron density maps were calculated using a combination of various diffraction datasets. Solvent attening (Wang, 1985) was employed to improve the MIR phases. Phases beyond 3.5 AÊ were relatively poor compared with those at lower resolution. The solvent- attened MIR maps were suf cient to permit backbone tracing and model building for most of the protein molecule. Nevertheless, electron density was still poor for some portions of the protein, especially in regions of the b3-b4 connecting loop; b11 to b13 of p66; b11 to b14 of p51; and parts of the thumb subdomains in both p66 and p51. In addition, many amino acid residues lacked good density for their side-chains, leading to some ambiguity in amino acid assignments. To improve the quality of the experimental maps, we tried a variety of approaches to incorporate the initial partial model structure into the experimental MIR phases with less model bias (Ding & Arnold, 1994). Phase combination procedures using partial atomic models to de ne molecular envelopes for density modi cation yielded electron density maps with signi cantly improved quality for many side-chains and for regions where electron density was weak in the MIR-phased maps. Structural interpretation of HIV-1 RT was based on a number of electron density maps calculated using MIR phases with resolution limits between 3.0 and 3.5 AÊ (Jacobo-Molina et al., 1993). The atomic model was constructed using graphics program O(Jones et al., 1991). Structure re nement was performed using the program X-PLOR (Brunger, 1993). Molecular modeling was guided by difference Fourier maps with various coef cients and simulated-annealing omit maps. The free R-factor (Brunger, 1992) was calculated in each cycle of model building and re nement to monitor the progress. During the structure re nement of the HIV-1 RT/ dsdna/fab28 complex, structures of HIV-1 RT/NNRTI complexes (Smerdon et al., 1994; Ding et al., 1995a,b) and unliganded HIV-1 RT structures (Rodgers et al., 1995; Hsiou et al., 1996) were used as cross-references to guide model building. For DNA re nement, dihedral restraints for the sugar puckerings were applied to constrain the ve base-pairs near the polymerase active site to A-form geometry and the nine base-pairs near the RNase H active site to B-form geometry. In the nal stages of structure re nement, to improve phase quality and reduce model bias, the multiple crystal

13 Implications of the Polymerase Active Site Region 1107 Figure 6. Portion of a (2F o F c ) difference Fourier electron density map computed at 2.8 AÊ resolution in the region of the polymerase active site of p66 (contour level 1s). The phases were calculated using the nal atomic coordinates of the HIV-1 RT/dsDNA/Fab28 complex, shown as a stick model. form map averaging technique was used to perform real space electron density averaging between diffraction datasets measured from frozen and cooled crystals using the programs RAVE (Kleywegt & Jones, 1994) and dmmulti (Cowtan, 1994). During the map averaging procedure the structure of the HIV-1 RT/dsDNA/Fab28 complex was divided into 18 fragments by subdomain (HIV-1 RT: 13; dsdna: one; and Fab28: four). The averaged electron density maps were of good quality and resolved many ambiguities in backbone tracing and sidechain placement in regions where electron density was weak or lacking in other maps. Structure re nement of the nal atomic model converged to an R-factor of for 60,257 re ections (71.0% completeness) and a free R-factor of for 3206 re ections (3.8% completeness) between 8.0 and 2.8 AÊ resolution with F 5 2s(F). Statistics for the nal model using all data without resolution and amplitude cutoffs were: R-factor ˆ for 72,320 re ections (81.5% complete) and free R-factor ˆ0.366 for 3867 re ections (4.4% complete). Figure 6shows arepresentative portion of a (2F o F c ) difference Fourier map at 2.8 AÊ resolution in the region around the polymerase active site. A summary of the re nement and atomic model statistics is given in Table 3. Due to weak or invisible electron density for the side-chains, the following residues were modeled as alanine residues: p66: 13, 28-30, 32, 36, 40, 43-44, 61, 64-74, 78, , 194, 199, 203, 211, , 238, 242, , 275, 278, , , , , , 331, , 524, 530, 540, , 550, and ; p51: 13, 90-91, 173, , 268, , , 284, , , 300, 305, , 424, and 428; the light chain of Fab28: 61, 147, 149, , 183, , 195, and 211; and the heavy chain of Fab28: 125. The b3-b4 loop (64-74), b7-b8 loop ( ), the region connecting b14 and ah ( ), the ai-aj loop ( ), the carboxyl-terminal ve residues of the p66 subunit, and the connecting loop between b14 and ah ( ) of the p51 subunit, are highly disordered with very poor electron density. The backbone trace in these regions is tentative. Since the original description of the structure (Jacobo-Molina et al., 1993), the availability of additional experimental data and the re nement process permitted iterative model improvement that resulted in modi cations of main- and side-chain conformations throughout the structure. In particular, residue assignment in some regions of the structure led to differences from the original assignments in PDB entry 1HMI, including residues of p66, and , , , , and of p51. At the later stages of structure re nement, some residual electron densities were consistently seen at a Table 3. Statistics of structure re nement and model Resolution range (AÊ ) 8.0±2.8 Sigma cutoff (F o ) F o 52s No. of reflections 63,463 (working set: 60,257; test set: 3,206) Completeness (%) 74.8 (working set: 71.0; test set: 3.8) R-factor Free R-factor All RT p66 p51 Fab Heavy Light DNA No. of residues No. of non-h atoms 11, Average B-factor (AÊ 2 ) Main-chain atoms Side-chain atoms All atoms rms bond lengths (AÊ ) rms bond angles ( ) rms dihedral angles ( ) rms improper angles ( ) Ramachandran plot (%) Most favored regions Allowed regions Generously allowed Disallowed regions R ˆ jjf o j jf c jj/jf o j