Comparison of the Crystal Structures and Intersubunit Interactions of Human Immunodeficiency and Rous Sarcoma Virus

The crystal structures of the proteases (PRs) encoded by the Rous sarcoma virus (RSV) and the human immunodeficiency virus (HIV) have been compared. The crystallographic monomer of HIV PR superimposes on the two crystallographically independent subunits of the RSV PR dimer with root mean square deviations of 1.45 and 1.55 di for 86 and 88 common Ca atoms, respectively. There is a conserved structural core con- sisting of seven P-strands forming two perpendicular layers, a helix, and the amino-and carboxyl-terminal &strands. PRs from related retroviruses fold into sim- ilar structures with surface turns of variable length between the B-strands. Both HIV and RSV PR dimers have significant subunit-subunit interactions in three regions: the

The crystal structures of the proteases (PRs) encoded by the Rous sarcoma virus (RSV) and the human immunodeficiency virus (HIV) have been compared. The crystallographic monomer of HIV PR superimposes on the two crystallographically independent subunits of the RSV PR dimer with root mean square deviations of 1.45 and 1.55 di for 86 and 88 common Ca atoms, respectively.
There is a conserved structural core consisting of seven P-strands forming two perpendicular layers, a helix, and the amino-and carboxyl-terminal &strands.
PRs from related retroviruses fold into similar structures with surface turns of variable length between the B-strands. Both HIV and RSV PR dimers have significant subunit-subunit interactions in three regions: the "firemen's grip" at the active site; the salt bridges involving Arg,, Asp,,, and Arg,, of HIV PR; and the termini of the two subunits, which form a fourstranded antiparallel &sheet.
The specific interactions of the termini differ in the two PRs. The carboxyl termini, residues 96-99 of HIV PR and residues 119-124 of RSV PR, contribute -50% of the intersubunit ionic and hydrogen bond interactions and -45% of the buried surface area involved in dimer formation. This information may be useful in the design of site-directed mutations or inhibitors of dimer formation.
The protease (PR)' encoded by retroviruses is essential for production of infectious viral particles since it is required for processing of the gag and pol polyprotein precursors (Kramer et al., 1986;Krausslich and Wimmer, 1988). Therefore, human immunodeficiency virus (HIV) PR has been proposed as a potential target for antiviral drugs. Other retroviral PRs show similar amino acid sequences and some cross-specificity for peptide cleavage sites, which suggests that the three-dimensional structures will be conserved. They contain the characteristic catalytic triplet Asp-Thr/Ser-Gly (Toh et al., 1985) and are inhibited by typical aspartic protease inhibitors (Kotler et al., 1989;Richards et al., 1989).
Crystal structures have been determined of PRs from Rous sarcoma virus (RSV) (Miller et al., 1989a) and from bacterially expressed (Navia et al., 1989;Lapatto et al., 1989) and chemically synthesized HIV (Wlodawer et al., 1989)   ; and more recently, the crystal structure of HIV PR complexed with a peptide inhibitor has been experimentally determined (Miller et al., 198913). Currently, the RSV PR structure is the most accurately determined. It has been refined to an R-factor of 14.4% at 2.0-A resolution (Jaskblski et al., 1990). The structure o.f chemically synthesized HIV-l PR was determined at 2.8-A resolution and was refined to an R-factor of 18.4% (Wlodawer et al., 1989). Each subunit of PR folds in a similar manner to one domain of the pepsin-like proteases, which may have arisen by a gene dupli-  Lee and Richards (1971). A probe size of 1.4 A was used.

AND DISCUSSION
HIV-l PR crystallizes as a dimer of two identical subunits (imposed by crystal symmetry), whereas the RSV PR dimer contains two crystallographically independent subunits. The HIV PR subunit superimposes with root mean square deviations of 1.55 A for 88 CLY atoms and 1.45 A for 86 common CLY atoms on the two nearly identical subunits of RSV PR ( Fig. 1, upper). The overall topology is very similar, although HIV-l PR of 99 residues is considerably shorter than 124residue RSV PR. All of the deletions in HIV PR relative to RSV PR occur at surface loops. The two largest deletions are between p-strands b and c, and b' and c'. The region leading up to the flap has a differentOconformation, with a root mean square deviation of over 2.3 A for all atoms of residues 35-57. The helical turn between p-strands d and a' in RSV PR is absent in HIV-1 PR. Residues 61-70 are not visible in the RSV PR electron density maps (Miller et al., 1989a), and the flap in HIV-l PR is involved in crystallographic contacts with a symmetry related flap and other residues; and comparison is not straightforward.
The single flap of nonviral aspartic proteases has a different conformation (Blundell et al., 1985;Suguna et al., 1987). The subunit on the left shows the nomenclature for the secondary structure, with P-strands a-d and a'-d' and helix h' (Blundell et al., 1985). &Strands b, c, b', and c' form the A layer, and p-strands c, d, c', and d' form the B layer (Pechik et al., 1989). The amino and carboxyl termini are indicated (N and C, respectively). The corresponding sequence alignment is given by Weber (1989). The subunit on the right is shaded where the structure of HIV PR differs from that of RSV PR, usually due to shorter surface turns. The active site is indicated by an arrow, and the dashed lines represent the residues of the flap which were not visible in the electron density map for RSV PR.
All of the known retroviral PR sequences can be fit into a similar alignment, which has a conserved structural core and surface loops of varying lengths (Weber, 1989). HIV PR has the smallest and most compact structure. The conserved core structure consists of the A layer (p-strands b, c, b' and c'), which is almost perpendicular to the B layer (p-strands c, d, c' and d') (Pechik et al., 1989), the carboxyl-terminal helix, the flexible flap, and P-strands a and q at the termini (Fig. 1,   lower). The 79 residues forming the core structure have 31% sequence identity and 1.63-and 1.57-A root mean square deviations between main chain atoms of HIV compared to the two independent subunits of RSV PR (excluding residues 36-55 in the flap of HIV PR). This is comparable with differences observed between other pairs of protein structures with -30% sequence identity (Chothia and Lesk, 1986  of a long polyprotein and is released by autocatalytic cleavage (Kramer et al., 1986). Since retroviral PRs are active as dimers, one assumption is that two polyproteins must be correctly aligned to form a PR dimer before autocatalysis may occur. It has been suggested that inhibiting formation of the PR dimer may provide a useful treatment for diseases caused hy retroviruses (Wlodawer et ab, 1989). Therefore, the intersubunit interactions of RSV and HIV PRs have been described to understand the molecular basis for dimer formation. Intersubunit interactions occur in several regions: at the active site, between residues from helix h' and .&strands b and d, between the ends of the flaps, and between amino-and carboxyl-terminal P-strands a and q (Fig. 1, lower). In HIV PR, NH and C=O of Gly"' at the end of the flap form hydrogen bonds to C=O and NH of Gly"' in t.he flap of the other subunit. No comparison can be made with RSV PR since the equivalent residues of the flaps are apparently disordered. The active site of the PR dimer includes residues from both subunits. The triplet Asp-Ser/Thr-Gly is followed by Ala in retroviral PRs or by Ser or Thr in almost all nonviral aspartic proteases. These 4 residues interact with the same 4 residues from the other subunit of the PR dimer in an arrangement which has been named the "fireman's grip" (Blundell et al., 1985). The a$ive sites of the two retroviral PRs superimpose with a 0.49-A root mean square deviation for all 44 atoms in the 8 residues (Fig. 2). Nonviral aspartic proteases differ only in the additional interaction formed by Ser or Thr following the conserved triplet (James and Sielecki, 1983;Miller et al., 1989a). The effect of Ser or Thr instead of Ala4' in RSV PR has been tested by site-directed mutagenesis (Leis et al., 1990). Since the catalytic residues of all aspartic proteases are remarkably similar, this region may not provide a good target for a selective inhibitor of HIV-l PR.
Asp20 forms salt bridges with Arg8' from helix h' and with Ar$ at the end of @-strand b', and the side chains of Thr3' in P-strand d and of AsnBR in the helix form a hydrogen bond interaction (Fig. 3, upper). A similar network of interactions between equivalent residues is observed in RSV PR (Fig. 3,  lower), although Ar$ is next to a deletion in HIV PR relative to RSV PR. Mutations of these residues in HIV PR reduce or eliminate protease activity (Louis et Q1., 1989;Loeb et al, 19891. These include the conservative substitutions of Lys for Arg7 or Glu for Asp"', which is consistent with a conserved conformation. Although Thr3', Arg7, and AsnfAsp= are highly conserved in all PRs, the residue corresponding to Asp% in HIV PR is Gln, and the residue corresponding to ArgS is Glu in Moloney murine leukemia virus and feline leukemia virus (Weber, 1989). The conformation would be conserved if the same residues formed a different set of interactions in which Arg at position 87 formed a salt bridge with Glu at position 8 and a hydrogen bond with Gln 29. This could be tested by making multiple substitutions of these residues which were designed to form alternative interactions.
The amino and carboxyl termini of the two PRs form a four-stranded antiparallel ,&sheet, in which the amino terminus of one subunit lies next to the carboxyl terminus of the adjacent subunit (Fig. 4). This arrangement is distinctly different from the interdomain six-stranded p-sheet seen in the pepsin-like proteases (Blundell and Pearl, 1989). In addition to the hydrogen bonds between main chain C=O and NH groups, there are further interactions which differ in the two PR dimers. In HIV PR, there is a salt bridge between the amino terminus of one subunit and the carboxyl terminus of the other subunit. The carboxyl terminus can also form an intersubunit hydrogen bond interaction with Hi@', in @strand c'. There is a network of hydrogen bonds linking the side chains of AsnQ8 and Gln2 from both subunits. Deletion of Phegs inactivates HIV PR (Hostomsky et al., 1989), although several amino acid substitutions can be accommodated (Loeb et al., 1989), which suggests that the length of the carboxyl terminus is important. a This does not include residues 61-70 of the flaps.
In RSV PR, there is no salt bridge between the termini. Instead, the carboxyl terminus forms a salt bridge with Arg"" from the other subunit in the dimer (Fig. 4, lower). The amino terminus forms a network of interactions involving G1ug2, Asnlz3', and Asn". Leu"" at the carboxyl terminus is displaced out of the P-sheet on the surface of the protein. Other PRs, such as Moloney murine leukemia virus PR, have additional carboxyl-terminal residues which may be accommodated by a continuation from the position of Leu'24 in RSV PR.
The terminal P-sheet has hydrophobic side chains facing the interior of the protein close to the active site. Leug7 is adjacent to Leu'"' and Thr*'j' in HIV PR, whereas Leulzl is near Se? of the catalytic triplet in RSV PR. In other retroviruses, Leug7 is replaced by Ile and Leu24 by Val, which suggests that the internal hydrophobic packing is conserved (Weber, 1989). Mutation of Leuz4, Thrz6, or Leug7 reduces the HIV PR activity (Loeb et al., 1989) so that altering the internal packing may perturb the catalytic activity.
In HIV PR, a total of 34 hydrogen bond and four ionic interactions occur between the two subunits: two hydrogen bonds occur between the flaps; five occur between active site residues 25-27 (Fig. 2); eight occur between residues 6-8, 29, and 87 (Fig. 3); and 19 hydrogen bonds and two ion pairs are formed by terminal residues l-4 and 96-99 (Fig. 4, upper). In fact, residues 96-99 contribute 50% of the ionic and 56% of the hydrogen bond intersubunit interactions. Similarly, the subunit-subunit interactions of RSV PR involve 35 hydrogen bond and three ionic interactions: six hydrogen bonds occur between active site residues 37-39; 11 occur between residues 6-10, 41, 111, and 115; and 18 hydrogen bond and two ionic interactions are formed by terminal residues l-5 and 119-124. Again, the carboxyl termini contribute most of the intersubunit interactions. In the case of the RSV PR dimer, these calculations are incomplete since -10 residues of each flap are not visible in the electron density maps (Jaskblski et al., 1990). Calculation of solvent-accessible surface area (Table I) shows that -3000 A* of surface area is buried on dimer formation in both RSV and HIV PRs. This is an estimate of the hydrophobic contribution to dimer formation. Carboxylterminal residues 96-99 of HIV PR or residues 119-124 of RSV PR contribute at least 44% of the buried surface area. Since the two carboxyl termini also contribute at least 50% of the ionic and hydrogen bond intersubunit interactions, these residues are expected to have a significant effect on dimer formation. This suggests that using peptides as competitive inhibitors of P-sheet formation by the amino and carboxyl termini may efficiently inhibit dimer formation and thus reduce PR activity. Interference with subunit-subunit interactions at the carboxyl terminus is the presumed mechanism for the specific inhibition of herpesvirus ribonucleotide reductase by peptides corresponding to the carboxyl terminus of the small subunit (Dutia et al., 1986;Cohen et al., 1986).
The prediction is that a peptide containing residues 96-99 of HIV PR or residues 120-124 of RSV PR should interfere with formation of the PR dimer. Unlike active site inhibitors, this type of inhibitor is likely to be specific for retroviral, rather than nonviral, aspartic proteases. The large contribution of the carboxyl termini to the intersubunit hydrogen bond and ionic interactions, which differ in RSV and HIV PRs, suggests that it may be possible to design an inhibitor of dimer formation that will be specific for one particular retrovirus, for example, HIV.