Comparison of the Substrate-binding Pockets of the Rous Sarcoma Virus and Human Immunodeficiency Virus Type 1 Proteases*

A steady state kinetic analysis of the avian myeloblastosis virus/Rous sarcoma virus (AMV/RSV) and human immunodeficiency virus Type 1 (HIV- 1) retroviral proteases (PRs) was carried out using a series of 40 peptide substrates that are derivatives of the AMVI RSV nucleocapsid-PR cleavage site. These peptides contain single amino acid substitutions in each of the seven positions of the minimum length substrate re- quired by the PR for specific and efficient cleavage. These peptide substrates are distinguished by the in- dividual enzyme subsites of the AMV/RSV and HIV-1 PRs. The molecular basis for similarities and differ- ences of the individual subsites for both proteases is discussed using steady state kinetic data and modeling based on crystal structures. terminus (14). The concentration of this peptide was determined by amino acid composition analysis. The peptides used in this study were designed with prolines at their amino termini so that the relative fluorescence intensity represents only the newly formed amino termini produced as a result of proteolytic cleavage. Arginine residues were added to the carboxyl terminus to improve solubility of the peptides. The addition of these amino acids did not influence their kinetic parameters.

ll To whom all correspondence should be addressed. known to be specifically and efficiently processed by PR does not reveal a simple consensus sequence.
Structural analysis of inhibitor complexes of HIV-1 PR (5-11) indicates that the substrate most likely binds to PR in an extended , 6 conformation, maintained by a network of hydrogen bonds between the main chain amide and carbonyl groups of the substrate and PR (12,13). In addition, subsites (S4-S3') of the substrate-binding pocket have been identified from structural data. The inhibitors used are generally hydrophobic and are capable of interacting with the P3-P3' residues of the substrate. Shown schematically in Fig. 1 is an AMV/RSV peptide representing the cleavage site found between the AMV/RSV nucleocapsid (NC) and PR proteins (14) and referred to as the NC-PR peptide substrate. This peptide substrate is depicted as being bound to the substrate-binding pocket. This diagram also emphasizes the relative size and position of the S4-S3' subsites that interact with each substrate position and lists amino acids that differ between the RSV and HIV-1 PR subsites. Interestingly, essentially identical subsites have been described for the various HIV-1 PR complexes despite large differences in the primary structure of the substrate-based inhibitors (5)(6)(7)(8)(9)(10)(11). This suggests that the subsites are constrained by the structure of PR and to a large extent are independent of substrate sequence. This hypothesis is supported by data obtained from the analysis of AMV/RSV PRs with mutations in the substrate-binding pocket that demonstrate that PR subsites are capable of acting independently in the selection of substrates (12). Note that the RSV and AMV PRs, which differ in sequence by two amino acids, are biochemically indistinguishable and have been used interchangeably in these studies. In order to avoid confusion, we will use AMV/RSV throughout to refer to both PRs. However, the viral PR was purified from AMV, whereas the RSV PR was bacterially produced.
Previously, we described the activities of AMV/RSV and HIV-1 PRs on an AMV/RSV NC-PR peptide substrate that was cleaved specifically and efficiently by both enzymes. In a subsequent report, we characterized the activities of both AMV/RSV and HIV-1 PRs on analogs of the NC-PR peptide substrate that contained single amino acid substitutions beginning in the P4 position of the substrate and extending through the P3' position (12). This set of systematically modified peptides was capable of biochemically distinguishing the subsites of these PRs. In this report, we describe results from kinetic analyses of AMV/RSV and HIV-1 PRs on the 40 modified NC-PR peptide substrates and discuss these data 11711 Enzyme Subsites ofthe AMVIRSV and HIV-1 PRs in terms of the molecular basis of PR substrate specificity and catalytic efficiency. Furthermore, this study has identified amino acids in each of the enzyme subsites that may be important determinants for specific recognition of substrates.

EXPERIMENTAL PROCEDURES
Purification of Retroviral Proteases-AMV PR, purified from virus as described (15), was obtained from Molecular Genetic Resources, Tampa, FL. HIV-1 PR expressed in and purified from bacteria was a generous gift of Dr. J. Giam, Case Western Reserve University.
Peptides-The peptides used in this study were synthesized and purified by Advanced Chemtech (Louisville, KY). Peptides were solubilized in 1 mM 2-mercaptoethanol, and their concentrations were determined by quantitative amino acid composition analysis. Specificity of cleavage was established by direct amino-terminal analysis of product peptides (16).
Assay of PR Activity-The reaction mixture contained 100 mM sodium phosphate, pH 5.9,2.4 M sodium chloride (1 M sodium chloride was used for HIV-1 and HIV-2 PR reactions), 10-400 pM peptide as indicated, and 0.5-5 pg/ml PR. Reaction volumes were 25 pl. Reactions were initiated by the addition of PR, incubated for various periods of time, and stopped by the addition of 175 pl of 0.5 M sodium borate, pH 8.5. Twenty p1 of 0.05% (w/v) fluorescamine was then added. HIV-1 PR was never incubated more than 8 min due to its instability, presumably a result of autodegradation. The relative fluorescence, after reaction with fluorescamine, was determined on a Perkin-Elmer LSdB spectrofluorometer using an excitation wavelength of 386 nm and an emission wavelength of 477 nm. Excitation and emission slit widths were 10 nm. Relative fluorescence intensity was converted to nanomoles of product using a standard curve described by the following equation: nmol of product = relative fluorescence intensity/313. The standard curve was obtained using a hexapeptide with a free amino terminus (14). The concentration of this peptide was determined by amino acid composition analysis. The peptides used in this study were designed with prolines at their amino termini so that the relative fluorescence intensity represents only the newly formed amino termini produced as a result of proteolytic cleavage. Arginine residues were added to the carboxyl terminus to improve solubility of the peptides. The addition of these amino acids did not influence their kinetic parameters.
Steady State Kinetic Amlysis-Kinetic constants were determined using the assay described above. Concentrations of peptide ranged from 0.25-4 times the K , value. No more than 20% of the substrate was allowed to be consumed during the course of any given experiment. Initial velocity data used to calculate kinetic constants were obtained from at least three experiments performed in duplicate. Kinetic constants were determined by a nonlinear fit of the data to the Michaelis-Menten equation using the NFIT program (14). Correlation coefficients of the fit were greater than 0.98, and the standard deviation of the constants reported was 520%.
Molecular Modeling-The crystal structures of the native RSV PR (17) and HIV-1 PR, as well as the complexes of the latter with inhibitors (5)(6)(7)(8)(9) were superimposed using C, atoms and examined on an Evans and Sutherland ESVlO computer graphics system using the program FRODO (18). The residues forming the subsites were determined directly for the HIV-1 PR complexed with inhibitors, and the corresponding residues in RSV PR were obtained from the structural and sequence alignments (12). The NC-PR substrate was modeled as described previously (12), and different amino acid side chains were substituted at positions P4 to P3' in order to provide a structural explanation for the kinetic data.

Rationale for Choice of Substitutions in Peptide Substrates-
We have synthesized 40 different peptide substrates, based on the AMV/RSV NC-PR cleavage site (PPAVS-LAMTM), which contain single amino acid substitutions in each of seven positions, P4-P3'. The amino acids present in these seven positions of the substrate have been shown by biochemical and crystallographic analyses to be necessary for specific and efficient processing by the retroviral PR. The amino acids substituted into the NC-PR peptide were chosen to explore predictions of size and polarity of each of the seven known subsites of the PR substrate-binding pocket. For instance, Gly, which has no side chain, was substituted at each position to test the importance of substrate side chain interactions with each PR subsite. A number of general features of the AMV/RSV subsites were taken into account in designing the NC-PR analogues. For example, subsite S4 is near the protein surface SO that it may accommodate a variety of different types of substrate side chains. Subsite S3 is formed partially by the charged residues, Arg-10 and Arg-105. Subsite S2 is small and hydrophobic, whereas subsite S1 is large and hydrophobic. Subsite S1' is medium-sized and hydrophobic, and the natural substrates do not show a residue larger than Leu a t Pl'. Subsite S2', like S2, is small and hydrophobic. Subsite S3' is near the surface of the enzyme and is partially exposed to solvent.
The studies described here utilized peptides as substrates in place of the multidomain polyproteins that are substrates in uiuo. This strategy appears to be valid since peptides based upon the naturally occurring cleavage sites are specifically and efficiently cleaved by the PR of the virus from which the cleavage sites were derived (12,14,(19)(20)(21). Furthermore, the relative efficiencies of processing of various sites were similar for both polyprotein and peptide substrates (22, 23).*

Comparison of HIV-1 PR kinetic data for AMVIRSV NC-PR substrates with substitutions in the P4-P3' positions
Varying concentrations of the NC-PR (PPAVS LAMTMRR) or NC-PR substrates with amino acid substitutions in the P4-P3' positions as indicated were incubated with purified HIV-1 PR (6-24 ng), and the extent of cleavage was determined using the fluorescamine assay as described under "Experimental Procedures." The wild type amino acid in the NC-PR peptide is indicated in the parentheses in the first column. ND, cleavage products were not detected. This would indicate that the relative catalytic efficiency (kat/&) for these reactions was ~0.001. AMV/RSV PR subsites, which highlights the differences between the two enzymes, is presented in Table IV and in Fig. 1. Note that because the peptide substrate binds in an extended , f 3 conformation, alternate subsites, such as S1 and 53, are adjacent to one another ( Fig. 1).

Relative amino acid
Subsite S4"The S4 subsite of HIV-1 PR is located near the enzyme surface such that amino acids in the P4 substrate position may be partially exposed to surrounding water molecules ( Fig. 2a). This is consistent with the observation that amino acids found in the P4 position of HIV-1 PR gag and pol polyprotein cleavage sites are mostly polar (Table 111). Similarly, the S4 subsite of AMV/RSV PR is near the surface. However, modeling of this site predicts that Pro-62 and Gln-63, derived from the flaps, may interact with the side chain

Comparison of AMVIRSV PR kinetic data for AMVIRSV NC-PR substrates with substitutions in the P4-P3' positions
Varying concentrations of the same NC-PR (PPAVS LAMTMRR) or NC-PR substrates with amino acid substitutions in the P4-P3' positions as indicated were incubated with purified AMVIRSV PR (25-100 ng). and steady state kinetic data were analyzed. of the P4 amino acid (12) (Fig. 2a). These additional amino acid residues in AMV/RSV PR provide for a more enclosed S4 subsite, capable of accommodating hydrophobic residues. This is consistent with the observation that mostly small hydrophobic amino acids residues are found in the AMV/ RSV gag and pol polyprotein cleavage sites at this position (Table 111). When residues 61-63 are deleted from AMV/RSV PR, the avian enzyme utilizes substrates with polar residues (Asn or His) in P4 similarly to HIV-1 PR.3 This conclusion that the AMV/RSV PR S4 subsite is more enclosed than the HIV-1 PR S4 subsite is further supported by the kinetic data summarized in Tables I and 11, which demonstrate that HIV-1 PR is less sensitive to changes in the P4 position than

Amino acid residues in gag and pol AMVIRSV and HIV-I protease cleavage sites
The amino acid sequence in bold letters represents the AMV/RSV NC-PR cleavage site. Numbers in parentheses indicate the multiple of times the amino acid is found in the natural gag and pol cleavage sites.  (4) Ala (2) Val Phe Thr Arg (2) Ala (2) Ser (2) Thr Pro Ala (2) Leu (2) (3) Ala (3) Phe CYS Ile (2) Val Ala Asn (4) Met (3) Ser Tyr (2) Leu Ala GlY Leu (2) Met TYr Phe (3)  Asn Ala Leu Thr GlY Pro (2) Ser (2) His Ala Met Pro (3) Phe (2)  Leu Val (3) Ala Leu Ile Ser Gly CYS Gln (3) Leu (2) Ile (2) Glu  ' The amino acid residues that are not conserved or unique to the AMV/RSV PR (not present in the HIV-1 PR) or are conserved between *Amino acid residues in the second subunit of the dimer are indicated by a prime. The first amino acid number refers to AMV/RSV, the AMV/RSV and HIV-1 PR subsites as indicated.
whereas the second refers to HIV-1 PR. The underlined (italicized) residues are involved in substrate main chain hydrogen bond formation.
AMV/RSV PR. For example, the NC-PR analogue with Gly at P4 is a reasonable substrate for HIV-1 PR but not for AMV/RSV PR, where the kcat/& is reduced to 10% of that with Pro. Substitution of Leu or Phe for Pro in the P4 position of the NC-PR substrate resulted in a substantial loss in binding and a small decrease in the catalytic rate by HIV-1 P R (Table  I). These substitutions also resulted in a substantial decrease in cleavage by AMV/RSV PR. This result suggests that large hydrophobic residues are not tolerated in the P4 position by either enzyme, which is consistent with the presence of several polar residues in the S4 subsite (Fig. 2a, Table IV). In contrast, when polar residues were substituted into the P4 position, all of the peptides were substrates, although the kccat/& values varied depending upon the amino acid substituted (Tables I  and 11). It is worth noting that introduction of His in the P4 position of the NC-PR substrate increased its relative KM for HIV-1 PR but decreased it for AMV/RSV P R (Tables I and   11). This result can be explained by the presence in S4 of Ile-42 in AMV/RSV PR, as compared with Asp-30 in HIV-1 PR. Molecular modeling suggests that Asp-30 of HIV-1 P R can form an ionic interaction with His in the P4 substrate position, which would improve its KM. This has been tested experimentally by substituting Asp for Ile-42 in the AMV/ RSV PR. The resulting enzyme showed a 2-fold increase in efficiency of cleavage of the His-P4 substituted NC-PR substrate,* similar to that observed for HIV-1 PR (Table I). Tozser et al. (13) found comparable results in their analysis of HIV-1 and HIV-2 PRs, i.e. most P4 analogues tested were utilized as substrates with the exception of those with larger hydrophobic residues. These studies used a peptide with a totally different sequence from NC-PR. Also, they observed that amino acids frequently found in turns in the threedimensional structure of proteins (Gly, Pro, Asn), when substituted in the P4 position, produced good substrates for HIV-1 P R (13).  (Fig. 2b). This allows side chains of amino acids in the P3 position freedom of movement to either interact with hydrophobic residues near the S1 subsite or polar residues at the surface of the enzyme. Thus, the S3 subsite can accommodate both polar and nonpolar residues. This is consistent with the presence of both polar and hydrophobic amino acids in the P3 position of the known AMV/RSV and HIV-1 gag and pol cleavage sites (Table 111). This ability of the S3 subsite to accommodate a variety of residues is also seen with substitutions in the P3 position of the NC-PR peptide (Tables I  and 11). Replacing Ala with larger amino acids that are charged (His, Arg), polar (Asn), or nonpolar (Phe), resulted in a 2-7.5-fold increase in efficiency of cleavage by AMV/ RSV PR, depending upon the amino acid substituted. The increased efficiency was due to both an improvement in binding and an increase in catalytic rate (Table 11). In contrast, substitutions of smaller residues, such as glycine or proline, were not tolerated as well.
Although AMV/RSV PR exhibited increased activity with many of the P3-modified peptides, HIV-1 PR retained about the same level of activity with all the substitutions (Table I). These results may be explained, in part, by the fact that the HIV-1 S3 subsite is more open than that of AMV/RSV due to the presence of Pro-81 in HIV-1 PR, as compared with Arg-105 at the structurally identical position in the avian enzyme (Fig. 2b). This difference would reduce the possibility for additional van der Waal interactions in the HIV-1 S3 subsite, thus precluding changes in catalytic efficiency. The &fold preference for Phe in P3 by AMV/RSV PR relative to HIV-1 PR can be explained by the presence of Gly-106 in AMV/RSV PR, as compared with Val-82 in the analogous position of HIV-1 PR. Val is at the top of the HIV-1 S3 subsite pointing into the pocket so that it could sterically interfere with the binding of a large amino acid residue such as Phe (12). This steric hindrance would not be a problem with Gly in the AMV/RSV subsite. The AMV/RSV PR S3 subsite also contains 2 basic residues, His-65 and Arg-105, not found in the HIV-1 S3 subsite. These residues confer on portions ofthe avian subsite a greater degree of hydrophilicity. However, the charged groups of these amino acids are probably not providing strong ionic interactions with substrate. This is inferred from the fact that calculation of the relative energy of interaction from the steady state kinetic data for the P3-substituted NC-PR peptide substrates indicates that there is only a 1-1.3 kcal/mol improvement in the energy of interaction for the best substrates. Ionic interactions would contribute an additional 4-5 kcal/mol of energy to the interaction of this substrate with PR, which would correspond to increases in the catalytic efficiency of 40-50-fold (26, 27). Such large changes in the catalytic efficiencies are not observed (Table 11). The magnitude of change observed here is more consistent with small perturbations in side chain van der Waal or weak hydrogen bond interactions. This suggests that the aliphatic side chain of Arg-105 in AMV/RSV PR may have an important hydrophobic role in defining the S3 subsite. In agreement with our results, Konvalinka et al. (28) found that HIV-1 PR will accept a variety of residues in the P3 position, whereas Strop et al. (29) report that the AMV enzyme has a preference for large polar or nonpolar residues at this position.
It is worth noting that in our kinetic studies of PR, we have not discussed the differential free energy of binding as described by Fersht (27). These calculations assume that there is no significant conformational change of the enzyme upon substrate binding. This is not the case for the retroviral PR, where comparison of the crystal structures of HIV-1 PR in the absence and presence of inhibitor (5) shows that the flaps are very flexible. They apparently must open to allow the polyprotein substrate to enter the binding site and then close to dock the substrate and restrain the scissile bond prior to catalysis (30). After cleavage, the flaps again move to release the products. In addition, we have exchanged small residues in the natural NC-PR substrate, in many cases, for larger ones. This may cause a sufficient rearrangement of the side chains in the PR substrate-binding pocket that it invalidates assumptions made by Fersht (27) in derivation of the relationship between catalytic efficiency and free energy of binding.
Subsite S2"The S2 subsites of both PRs (Fig. 2c) are sterically restricted, as compared with S4 and S3 and are predicted to accommodate hydrophobic residues (Table IV). Mostly small or intermediate sized hydrophobic residues are found in the P2 position of natural gag and pol cleavage sites (Table 111), with the exception of Phe in one AMV/RSV site and Asn in four of the HIV-1 sites. This ability of HIV-1 PR to accommodate a more polar residue may be related to the presence of Asp-30 in the S2 subsite of HIV-1, as compared with Ile-42 at the analogous location in AMV/RSV PR. Also, the presence of Val-32 in HIV-1 PR, as compared with Ile-44 in AMV/RSV PR, allows for a larger side chain at P2.
The requirement in S2 for small hydrophobic substrate residues is consistent with the results of replacement of Val at the P2 position of the NC-PR peptide substrate with larger hydrophobic residues (Leu or Trp); there is a significant decrease in the catalytic efficiency of AMV/RSV PR (Table  11)   Waal interactions in the individual enzyme subsites is a (1993) Biochemistry 32, 3354-3362. have reported kinetic measurements for HIV-2 PR, which has Ile at position 32 instead of the Val present in HIV-1 PR. A similar compensation in substrate for the gain of a methyl group in the amino acid at position 32 in HIV-2 was not observed. This was attributed to an unfavorable steric interaction of Pro in P1' with the 8-branched side chains such as Val in P2.
AS with S3 and S4, the S2 subsite of AMV/RSV PR is predicted to be smaller than that of HIV-1 PR due to the presence of Ile-44 and Ile-42 in AMV/RSV PR, as compared with Val-32 and Asp-30 in HIV-1 PR (Fig. 2c). Thus, although neither enzyme preferred polar or large hydrophobic residues in P2, the HIV-1 PR was more tolerant of substitutions than AMV/RSV PR. Table I1 shows lower relative bak/KM values for AMV/RSV PR than HIV-1 PR with the same analogues ( Table I). The ability of HIV-1 PR to accommodate various amino acids in the S2 subsite, as well as in the S3 and S4 subsites, may explain why previous studies of HIV-1 PR demonstrated an expanded substrate selectivity range, as compared with AMV/RSV PR (20).
Konvalinka et al. (28), using a different peptide as a starting point, found a relatively strong preference for Val or Ile in the P2 position for HIV-1 PR. Using the same enzyme and pol precursor proteins as substrate, Jupp et al. (31) found a preference for Val, as compared with Gly in this position. Strop et al. (29), using myeloblastosis-associated virus PR, found a preference for small hydrophobic residues in S2, which is consistent with the preference for Val in P2 of NC-

PR.
Subsite SI-The S1 subsites of both HIV-1 and AMV/RSV PR, like the S3 subsites, consist of relatively large and deep pockets (Fig. 2d). They also include the catalytic aspartic acid residues. However, the S1 subsites are mainly hydrophobic and embedded inside the protein. It is therefore not surprising to find hydrophobic residues at this position in the natural gag andpol polyprotein cleavage sites (Table 111). Also, similar to the more open S3 subsite, substitution of large hydrophobic amino acids in the P1 position increases the turnover rate of substrate. For instance, exchanging Trp, Phe, or Tyr for Ser in the P1 position of the NC-PR substrate resulted in 26, 17, and 19-fold increases, respectively, in efficiency of cleavage of these substrates by AMV/RSV PR (Table 11). These increases are due to favorable changes in both binding and rate of catalysis. Among the different NC-PR analogues analyzed, substitution of amino acids in P1 produced the largest increases in catalytic efficiency. This probably reflects the large size of the S1 subsite and its closeness to the scissile bond. The finding that substitution of AMV/RSV Val-104 with HIV-1 Thr or AMV/RSV Ser-107 with HIV-1 Asn significantly increased the rate of catalysis by the AMV/RSV PR without changing the substrate specificity' is consistent with the above argument. The HIV-1 PR also showed an increase in activity with Trp substituted in P1, but only 8-f0ld, whereas Leu in this substrate position resulted in a 13-fold increase in the rate of catalysis. Thus the HIV-1 enzyme prefers mediumsized hydrophobic residues in P1. This preference can be explained by the presence of Val-82 in HIV-1 PR, as compared with Gly-106 in AMV/RSV PR (12). The Val side chain points down into the subsite pocket. Thus, in contrast to the 534, S3, and S2 subsites, the S1 subsite of HIV-1 PR is smaller than that of AMV/RSV PR (see Fig. 2 4 . Several reports have noted the importance of having a large hydrophobic residue, particularly a Leu, in P1 (21,(29)(30)(31). This feature is also reflected in the choice of naturally occurring P1 residues in the case of HIV-1 (Table 111). Interestingly, polar or charged residues substituted at the P1 position (His, Arg, and Glu) resulted in better substrates for AMV/RSV PR than for the HIV-1 enzyme. This may reflect the fact that the AMV/RSV S1 subsite is more polar than that of HIV-1 PR due to the presence of Arg-105 and His-65. In contrast, Gly substitution was tolerated far better Subsites SI '-53'"Because of the symmetrical structure of the PR homodimer, the Sl'-S3' subsites utilize many of the same amino acids that form the Sl-S3 subsites. Thus, one would expect the properties of the "prime" subsites to mimic those described for Sl-S3. However, the peptide substrate is asymmetric and binds in an asymmetric manner. Therefore, the properties of the prime subunit of the enzyme are different from those of the "nonprime" subunit. One difference appears to be that the Sl'-S3' subsites are smaller. This can be seen directly by comparing the stereo picture of the subsite S3 to that of S3' and of subsite S2 to that of S2' (Fig. 2, b-g and c-f, respectively). In the case of S1 and Sl', measurements were made of the distance between the C, atoms of residues 105 and 106 in AMV/RSV PR and 81 and 82 in HIV-1 PR to the corresponding substrate positions.* It was found that these distances were 0.7-1.5 and 1.3-1.6 A closer to P1' than to P1, respectively. Thus, changes in the Pl'-P3' amino acids to larger residues usually resulted in decreased catalytic efficiencies for both PRs (Tables I and 11). The effect of the size difference between the related prime and nonprime subsites was also observed in the analysis of a AMV/RSV p2-p10 peptide substrate/inhibitor (20). Its ability to inhibit the PR activity was related, in part, to the steric properties of amino acid side chains in the P2' and P3' positions and their ability to fit into the relatively small S2' and S3' subsites (20).
None of the amino acids substituted in the P1' position of the NC-PR substrate improved the efficiency of cleavage by either PR, as compared with wild type. This is in contrast to the large increases in the activity observed for both PRs when the same substitutions were placed in the P1 position. The amino acids tested included polar (Arg, Gln, Glu), large hydrophobic (Phe), or no side chain (Gly). AMV/RSV PR tolerated polar residues in the P1' position of the NC-PR substrate (such as Arg) better than HIV-1 PR (Tables I and  11). This is also seen in the amino acids found in the natural gag and pol cleavage sites listed in Table 111, in which three of the avian sites contain either Ser or Thr. Like the structurally related S1 subsite, the avian S1' subsite is more polar due to the presence of His-65' and Arg-105, as compared with  In referring to PR, the prime amino acids are derived from the second subunit of the PR dimer. Substitution of His-65 and Arg-105 with Gly and Pro, respectively, produces AMV/RSV PRs that no longer tolerate polar amino acid residues in the P1' position of the NC-PR peptide substrate (12).
The more hydrophobic nature of the HIV-1 S1' subsite presumably accounts for the activity of the NC-PR peptide with Phe substituted in P1' (Table I). However, HIV-1 PR is less active with this substrate than with wild type, which contains Leu in this position. The reduced activity of the HIV-1 PR may be due to steric hindrance between Val-82 and Phe in PI' (Fig. 2e). Substitution of Gly in P1' resulted in a 20-fold and at least 1000-fold decrease in efficiency of cleavage of the NC-PR substrate by HIV-1 and AMV/RSV PRs, respectively. This result verifies the importance of side chain interactions at Pl'.
The S1' subsites for both enzymes are predicted to accommodate medium-sized hydrophobic residues best. Although few such residues were tested in the context of the NC-PR peptide, they were tested in the context of the AMV/RSV p2-by HIV-1 PR. p10 peptide substrate (PYVG-SGLY) (20). The substitution of Ser in PI' with Ala produced an active substrate with an efficiency of cleavage equivalent to the wild type NC-PR peptide substrate (20). Substitution of the larger Leu in P1' improved the efficiency of specific cleavage by the AMV/RSV PR an additional 2-fold, due primarily to an increase in the rate of catalysis.
Changes in kcat seem to be more evident for substitutions on the prime side relative to the nonprime side of both the NC-PR and the p2-p10 (20) peptide substrates. The kinetic mechanism for HIV-1 PR described by Hyland et al. (32,33) proposes the existence of several potentially rate-limiting steps after substrate binding. This includes conformational changes, such as opening and closing of the PR flaps, formation of the amide-hydrate intermediate, or product release itself. Our studies cannot identify the rate-limiting step.
Therefore, when the kat varies, it is not clear if this reflects modulation of the rate of the existing rate-limiting step or a change to a different rate-limiting step in the reaction. The S2' subsite, like S2, is predicted to accommodate small hydrophobic residues (Fig. 2f). The same series of amino acids tested in the P2 position of the NC-PR substrate were introduced into the P2' position. Every substitution listed in Tables I and I1 resulted in a significant loss in the efficiency of cleavage by both AMV/RSV and HIV-1 PRs. Although AMV/RSV PR preferred the wild type Ala over any other substitution, HIV-1 PR was somewhat more tolerant than AMV/RSV of changes in P2' to Ser, Gly, or Leu. Trp, however, is too large for either PR to fit in the S2' subsite. Tozser et al. (24) observed that Val and Leu, medium-sized hydrophobic residues, produced the best substrates for HIV-1 and -2 PRs when substituted in P2', in qualitative agreement with the results shown here.
Substitution of the small hydrophobic residues Ala or Val for Gly in the P2' position of the AMV/RSV p2-p10 peptide, produced substrates that were 2-3-fold more efficiently cleaved by AMV/RSV PR than a P2' Leu-substituted p2-plO peptide (20). This is consistent with the hypothesis that the S2' subsite prefers small hydrophobic residues. Similar results were obtained by Margolin et al. (34) when substitutions were made in the P2' position of the HIV-1 MA-CA processing site, Le. a preference for hydrophobic residues, and low or no cleavage when substitution of Gly or Trp was made in this position. With their substrate, however, Ile had a higher kcat than Ala or Leu, which further demonstrates the effect of context on substrate selection by PR. In studies with the myeloblastosis-associated virus PR, Strop et al. (29) reported an apparently symmetrical requirement for small hydrophobic residues in both P2 and P2' and large hydrophobic residues in P1 and Pl'. Although our results indicate similar preferences, the substrate side chain specificities are not identical in either the 52 subsite, as compared with S2' or in the S1 subsite, as compared with SI'. The AMV/RSV PR, but not the HIV-1 PR, was stimulated when Tyr replaced Met in the P3' position of the NC-PR substrate. As presented in the discussion of the S3 subsite, this difference is probably related to the presence of Gly-106 in AMV/RSV PR and Val-82 in HIV-1 PR (Fig. 2g). The other substitutions tested at P3', Leu, Asn, Gly and Asp, caused large reductions in activity for both enzymes.
The binding pockets of HIV-1 and HIV-2 PRs are predicted to contain four conservative differences (25). Crystallographic data for the HIV-2 enzyme are not publicly available at this time. However, it is reasonable to assume that given the higher degree of homology between the HIV-1 and HIV-2 PRs, as compared with that of HIV-1 and AMV/RSV PR, the residues comprising the substrate-binding pocket would be closely related. The few reports that exist (13,24), as well as our own unpublished data, indicate that the HIV-1 and HIV-2 PRs have very similar substrate specificities.

Contribution of Amino Acid Side C h i n Interactions to Bind-
ing of Substrate-The peptides with Gly inserted into the different positions were generally poor substrates with either PR. This was the case even for substitutions at the P3 or P1 positions, where the broadest range of side chains were observed. When comparing the two enzymes, however, HIV-1 PR was more tolerant of glycine substitutions than AMV/ RSV PR. Apparently HIV-1 PR is somewhat less dependent upon side chain interactions than the larger, and perhaps more flexible, AMV/RSV PR. This difference in dependence on side chain interactions can be seen by comparing the catalytic efficiency of each substrate relative to the catalytic efficiency of the comparable Gly-modified substrate. For example, Leu at P1 is 19.3-fold better than Gly for HIV-1 PR, whereas it is 85-fold better for AMV/RSV PR. Overall, the side chains have a larger contribution to the catalytic efficiency for AMV/RSV than for HIV-1 PR.
General Comments-There is interaction between the adjacent subsites of the enzyme-binding cleft. Residues that produce good substrates at certain positions in one peptide sequence may not work well in other sequences. For example, although Leu and Ser occur in the P2' position of natural AMV/RSV PR substrates (Table 111), the same substitutions are not favored in the context of the NC-PR substrate ( Tables   I and 11). Tritch et al. (22) tried to mix features from various efficiently cleaved substrates and found that the combinations tended to be very poor substrates. Thus substrate selection by PR must be based on recognition of all 7 amino acids of the substrate simultaneously.
Drug Design and Viral Resistame-Rational design of inhibitors of HIV-1 PR is aided by structural data from the large number of co-crystal structures that have been determined. In some cases, new inhibitors have been designed based upon structural information (7,11). Analysis of HIV-1 PR-inhibitor structures has suggested that substrate or inhibitor binding has two components (25). 1) The main chain C=O and NH groups of peptides or peptidic inhibitors have hydrogen bond interactions with PR residues. These are highly conserved among the different crystal structures' and are independent of amino acid sequence. Molecular energy minimization calculations suggest that main chain atoms contribute 56 to 68% of the total PR-inhibitor interaction energy? This suggests that PR residues that form hydrogen bond interactions with main chain atoms of the inhibitor will be conserved and are less likely to mutate in selection of drug resistant viruses. Thus, it is important to include potential hydrogen bond donors and acceptors for these interactions in the design of inhibitors. 2) Each side chain of the substrate or peptidic inhibitor lies in successive subsites from S4 to S3' formed by PR residues. This component is dependent upon the amino acid sequence of a substrate (or inhibitor) and is being studied by the steady state kinetic measurements presented here and by other laboratories. Amino acids that provide tight binding at a particular position and form good PR substrates are predicted to also form effective inhibitors when these amino acids are placed at the same position in an inhibitor. This has been demonstrated with modifications of the AMV/RSV p2-p10 inhibitory peptide (20), in which a 40fold improvement in Ki was observed when Gly at P1 was replaced by Trp and Tyr at P3 was replaced by Phe. These Enzyme Subsites of the AMVIRSVand HIV-1 PRs specific substitutions were chosen based on the kinetic data presented herein.
Finally, alterations in residues forming the PR subsites have been shown to effect the binding of substrates (12,20). Even a small change in a particular subsite might be sufficient to reduce the affinity for a specific inhibitor. We would thus predict that the residues that we have proposed as being important determinants in substrate selection in the various PR subsites would be potential sites for selection of natural mutations that will produce drug-resistant phenotypes in AIDS patients treated with inhibitors of HIV PR. These residues would include HIV Val-32, Pro-81, and Val-82 and perhaps Asp-30 and/or Gly-48.