Human Immunodeficiency Virus*

: Because human immunodeficiency virus (HIV) infection often is detected through prenatal and sexually transmitted disease testing, an obstetrician–gynecologist may be the first health professional to provide care for a woman infected with HIV. Universal testing with patient notification and right of refusal (“opt-out” testing) is recommended by most national organizations and federal agencies. Although opt-out and “opt-in” testing (but not mandatory testing) are both ethically acceptable, the former approach may identify more women who are eligible for therapy and may have public health advantages. It is unethical for an obstetrician–gynecologist to refuse to accept a patient or to refuse to continue providing health care for a patient solely because she is, or is thought to be, seropositive for HIV. Health care professionals who are infected with HIV should adhere to the fundamental professional obligation to avoid harm to patients. Physicians who believe that they have been at significant risk of being infected should be tested voluntarily for HIV.

The human immunodeficiency virus (HIV) 1 protease is an ideal target for the chemotherapeutic treatment of HIV disease (1)(2)(3). Three protease inhibitors, saquinavir (Ro-31,8959), ritonavir (ABT-538), and indinavir (L-735,524) are effective in clinical trials in treating HIV disease (4 -6) and recently were approved by the Food and Drug Administration for the chemotherapeutic treatment of HIV infections. Other protease inhibitors in earlier stage clinical trials are VX-478 and AG1343. Notwithstanding these early successes, viral resistance to individual inhibitors and cross-resistance to multiple inhibitors occurs in vivo (7,8). However, recent experiments in vitro indicate that the appropriate use of inhibitors in combination and sequence may result in effective HIV therapy (9).
Under selection pressure, viral resistance develops in vitro to HIV protease inhibitors, such as saquinavir (10 -13), indinavir (9), ritonavir (14), XM323 (9), A77003 (15,16), AG1343 (17), and VX-478 (9,18). The mutations that confer viral resistance to protease inhibitors are located in the HIV protease gene. In many cases, the mutations generated from selection pressure in vitro are the same as those observed clinically (19). These results suggest that resistance studies in vitro could provide insight in what to expect in clinical trials.
Studies to understand the molecular mechanism of HIV resistance to protease inhibitors have begun (20). In the cases of saquinavir and A77003, mutant proteases that confer viral resistance to the inhibitor have reduced affinity for that inhibitor (12,15). Structural studies indicate that the reduced affinity of these mutant proteases for some inhibitors is, in part, due to the disruption of van der Waals interactions (21,22). In some cases, the decreased affinity of the mutant protease for the inhibitor correlates with reduced catalytic efficiency of the mutant protease (23,24). Changes in the substrate sequence can compensate for loss of catalytic efficiency of mutant proteases (23) and, in fact, processing sites of viruses resistant to BILA1906/2011BS and BILA 2185BS have compensatory amino acid substitutions (25). However, these compensatory mutations in HIV protease substrates have not been observed in vivo. Herein, we define the kinetic mechanism of viral resistance to saquinavir (Ro31-8959) by determining the rate constants for association and dissociation of saquinavir with WT and proteases with mutations that confer resistance to saquinavir. We propose a molecular model to explain these kinetic results. Furthermore, we examined the effect of these mutations on the catalytic efficiency with a variety of substrates that mimic the protease processing sites in vivo.

EXPERIMENTAL PROCEDURES
Materials-2-Aminobenzoyl-Thr-Ile-Nle-Phe(NO 2 )-Gln-Arg-NH 2 (1) was from Cambridge Research Biochemicals; VX-478, AG1343 (nelfinavir), ABT-538 (ritonavir), and Ro31-8959 (saquinavir) were provided by Dr. Roger Tung (Vertex Pharmaceuticals, Cambridge, MA); L-735,524 (indinavir) was provided by Hans Leban (Wellcome Research Laboratories), octapeptide protease substrates were provided by Dick Campbell and Paul Doyle (Wellcome Research Laboratories, Beckenham, * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. ʈ To whom correspondence should be addressed: Dept. of Molecular Biochemistry, Glaxo Wellcome, 5 1 The abbreviations used are: HIV, human immunodeficiency virus; MES, 2-(N-morpholino)ethanesulfonic acid; DTT, dithiothreitol; AIDS, acquired immunodeficiency syndrome; amino acids were designated by the standard one or three-letter codes; t-butyl, tert-butyl; WT, wild type. Kent, UK), and MES and DTT were from Sigma.
Preparation of HIV-1 Protease-HIV-1 protease and its mutants were cloned and expressed in Escherichia coli as described (12). The enzyme was extracted from the insoluble fraction of a cell lysate, chromatographed by gel filtration, and refolded as described (26). The dilute refolded enzyme was bound to a HR10/10 Mono S fast protein liquid chromatography column (Pharmacia Biotech Inc.) that was equilibrated in 100 mM sodium MES, pH 6.5, 10% glycerol, 5% ethylene glycol (Buffer A). The enzyme was eluted with a 20-ml linear gradient of 0 -0.5 M NaCl in Buffer A. DTT (50 mM) was added to the sample. The protease was concentrated to 3-6 M with a Centricon-10 (Amicon, Beverly, MA), then stored at Ϫ70°C. The protease typically comprised greater than 90% of the protein in the sample, as estimated by SDS-polyacrylamide gel electrophoresis analysis. 2 Enzyme Assays-Protease (1-15 nM) was assayed fluorometrically (27) with 10 -40 M 1 and inhibitor (0 -20 nM) in Buffer B (100 mM MES at pH 5.5, 400 mM NaCl, 0.2% PEG-8000, with 5% dimethyl sulfoxide) at 25°C. Fluorescence increase due to hydrolysis of the substrate was monitored on a PerSeptive Biosystems 96-well fluorescence platereader with an excitation filter of wavelength 340 Ϯ 20 nm and an emission filter of wavelength 420 Ϯ 20 nm. Protease activity was also determined with octapeptide substrates (0.1-5 mM) in 100 mM MES, pH 5.5, 400 mM NaCl, and 0.1% Triton X-100 in a total volume of 21 l. After incubation at 37°C for 5-60 min (cleavages were limited to less than 20% product formation), the reaction was quenched with 25 l of 2% trifluoroacetic acid, and the sample was subjected to HPLC analysis. Samples (40 l) were injected on to a C 18 column (Brownlee, spheri-5, 4.6 ϫ 100 mm, 5 m), and compounds were eluted with a 15-ml linear gradient of acetonitrile 5-45% in 0.1% trifluoroacetic acid at a flow rate of 1 ml/min. The elution of substrates and products was monitored by absorbance at 214 nm.
Enzyme Fluorescence Measurements-Rapid protein fluorescence measurements were made with an SF.17MV stopped-flow spectrofluorometer (Applied Photophysics Limited, Leatherhead, UK). Protease (0.05-0.1 M) was mixed with saquinavir (1-5 M) in Buffer B, at 25°C. The excitation wavelength was 280 nm, and emission scattering was detected with a 305 nm filter. Single exponential time courses (400 data points) of changes in the intrinsic protein fluorescence were fitted with the software provided with the instrument.
Viral Resistance to Protease Inhibitors-IC 50 values for WT and mutant virus inhibition by protease inhibitors were determined as described (9).
Equations for Interpretation of Pre-steady-state Kinetic Data-A onestep binding model (Equation 1a) requires that the approach to equilibrium be a pseudo first order process when the concentration of inhibitor is much greater than the enzyme concentration (equation 1b).

Equation for
Steady-state Kinetic Analysis-Initial velocity data from the inhibition of substrate hydrolysis were fitted to Equation 2 to obtain IC 50 values.
K i values were calculated from IC 50 values (Equation 3).
Titration of HIV Protease Active Sites and Determination of K i Values-Enzymatic activity was determined with fluorogenic substrate (1) in the presence and absence of inhibitor, as described above. Equation 4 was fit to the data, where F is the fraction of enzyme bound to inhibitor (F ϭ 1 Ϫ (V inh /V 0 ), where V inh is the steady-state velocity in the presence of inhibitor and V 0 is the velocity of uninhibited enzyme), [E] is the total concentration of enzyme, [I] is the total concentration of inhibitor, and IC 50 is the concentration of inhibitor that binds 50% of the enzyme if [E] is always much less than [I].
K i values were calculated from IC 50 values using Equation 3. Statistical Analysis-The constants defined by the linear equations were fit by standard linear regression analysis, and the constants defined by the nonlinear equations were estimated by an iterative nonlinear least-squares fitting routine using Sigma Plot (Jandel Scientific, Corte Madera, CA). Error estimates were taken from the error matrix generated by the fitting routine. Error estimates for values calculated from fitted values (for example, the calculated K d values) were determined by the propagation of error analysis (28).

Saquinavir Binding to WT and Mutant HIV Proteases-
Saquinavir is a potent inhibitor of HIV protease and HIV replication (29). Mutations, in the HIV protease, L90M, G48V, and L90M/G48V, were selected for by passage of virus in the presence of saquinavir. These mutations resulted in decreased sensitivity of HIV protease and HIV proliferation to saquinavir (10 -13). We determined the affinity of WT and the three mutant proteases for saquinavir (Table I). The K i values of WT, L90M, G48V, and L90M/G48V proteases for saquinavir are 0.033 Ϯ 0.005, 0.68 Ϯ 0.04, 5.4 Ϯ 0.6, and 33 Ϯ 1 nM, respectively. These K i values indicate that the affinity of saquinavir for L90M, G48V, and L90M/G48V proteases are 1/20, 1/160, and 1/1000 that of the WT affinity, respectively.
We determined the association and dissociation rate constants for saquinavir binding to WT and mutant proteases to define the kinetic mechanism for the decrease in saquinavir affinity for mutant proteases. The association rate constants were determined by monitoring the intrinsic protein fluorescence. Inhibitors that have significant absorbance at 325 nm, such as those containing a quinolinic acid residue, quench the protease fluorescence upon binding (30). Saquinavir, which contains a quinolinic acid residue, also quenched the protease fluorescence. For example, when 1 M saquinavir was mixed with 50 nM WT protease, the intrinsic protein fluorescence was quenched in a pseudo-first order process with a rate constant (k obs ) of 46 Ϯ 1 s Ϫ1 (Fig. 1A). k obs increased linearly with saquinavir concentration (Fig. 1A, inset). The association rate constant (k on , Equation 1b) for WT protease and saquinavir (the slope of the line from the plot of k obs versus [saquinavir]) was (4.2 Ϯ 0.2) ϫ 10 7 M Ϫ1 s Ϫ1 . k on values for the three mutant proteases, which, varied between 2 and 4 ϫ 10 7 M Ϫ1 s Ϫ1 (Table  I), were not significantly different from k on values for WT protease.
Typically, dissociation rate constants (k off ) for saquinavir ⅐ protease complexes were determined by monitoring the increase in intrinsic protein fluorescence that resulted from trapping free protease released from protease⅐saquinavir with an excess of a second potent inhibitor, L-735,524, that does not 2 The protease autoprocesses both the C and N terminus within this expression system. The C-terminal autoprocessing (removing a 10amino acid peptide) with G48V and G48V/L90M proteases was not as efficient as with WT protease. Therefore, preparations of these mutant enzymes contained as much as 30% protease with unprocessed C termini. This result was not a surprise because the G48V and G48V/L90M proteases have k cat /K m values that are 1/10 that of WT protease for an octapeptide substrate with the sequence of the processing site. We assumed that the kinetic results reported herein are the result of the G48V and L90M mutations and not a result of incomplete processing. quench the enzyme fluorescence. For example, when 100 nM G48V/L90M protease⅐saquinavir was mixed with 20 M L-735,524, the intrinsic protein fluorescence increased with a first-order rate constant of 0.528 Ϯ 0.003 s Ϫ1 (Fig. 1B) (Table I). However, in the case of WT protease, the protein fluorescence was not stable enough to obtain an estimate of k off . Instead, k off was calculated to be (0.0014 Ϯ 0.0002) s Ϫ1 (Table I, t1 ⁄2 ϭ 12 min) from the K i (which approximates the K d ) and k on (k off ϭ K d ⅐k on ). The k on values for all the proteases and saquinavir were similar, while k off values increased proportionally with the K i values (Table I).

Inhibition of WT and Mutant Protease/Virus by Other Inhibitors under Clinical Evaluation-
We also determined the affinity of WT and the three mutant proteases for four other protease inhibitors that are under clinical evaluation (VX-478, ABT-538, AG1343, and L-735,524, Table II). In contrast to the results with saquinavir, these mutant proteases and WT protease had similar affinity for VX-478, AG1343, or L-735,524. However, G48V and G48V/L90M protease had significantly lower affinity for ABT-538 than WT. In general, a decrease in affinity of the protease for an inhibitor resulted in a rank order decrease in the sensitivity of the respective virus to the drug. For example, L90M, G48V, and L90M/G48V viruses had, respectively, 2-, 6-, and 9-to 40-fold increased IC 50 values compared with WT virus (9,12). Typically, when the K i value for mutant protease did not increase greatly compared to WT, IC 50 values for the respective virus did not increase greatly. For example, IC 50 values for L90M/G48V virus and VX-478 (12), ABT-538, AG1343, or L-735,524 (12) increased 0.7-, 2.3-, 1-, and 2-fold, respectively, compared with WT.
Relative Catalytic Efficiency of WT and Mutant Proteases-We examined the kinetic parameters of nine protease substrates (Table III) with WT and the three mutant proteases. Eight substrates were octapeptides that mimic the consensus sequences of viral cleavage sites. The fluorogenic substrate was used to determine the inhibition constants herein. The L90M protease had similar catalytic efficiency to WT protease with all the substrates tested (K m , k cat , and k cat /K m values were all within a factor of 2.5) (Table III). In contrast, G48V and G48V/ L90M protease had significantly reduced catalytic efficiencies for TLNF-PISP, RKIL-FLDG, AETF-YVDG, and the fluorogenic substrate, where k cat /K m values were 1/10 to 1/60 that of WT protease. These decreased catalytic efficiencies were primarily due to increased K m values. The sequences of these four substrates have little homology; in fact, none of the substrates have the same P1Ј, P3, or P3Ј residues, although PR/RT and RT/RN have the same P1 residue. DISCUSSION Three mutations in HIV protease that confer HIV resistance to saquinavir are L90M, G48V, and G48V/L90M (10 -13). As the rank order of affinity for saquinavir to these three mutant proteases decreases, the rank order of the respective viral sensitivity to saquinavir decreases (12). Herein, we have determined that these mutant proteases have 1/20, 1/160, and 1/1000 the affinity for saquinavir compared with WT protease, respectively.
To further understand the kinetic and molecular mechanism of HIV resistance to protease inhibitors, we determined the association and dissociation rate constants for binding of saquinavir to WT-, L90M-, G48V-, and G48V/L90M protease. There was little difference among the association rate constants for saquinavir and the proteases, which ranged from 2 to 4 ϫ 10 7 M Ϫ1 s Ϫ1 . In contrast, the dissociation rate constants varied up to 540-fold and accounted for the reduction in affinity of HIV protease mutants for saquinavir. HIV protease binds peptidomimetic inhibitors in a two-step mechanism of Equation 5 (30).
FIG. 1. Association of saquinavir to WT HIV protease and dissociation of saquinavir from G48V/L90M HIV protease. A, quenching of intrinsic protein fluorescence after 1 M saquinavir was mixed with 50 nM WT protease, as described under "Experimental Procedures." Superimposed on the data is the best fit of a pseudo-first order process with a rate constant (k obs ) of 46 Ϯ 1 s Ϫ1 to the data. Inset, k obs plotted versus saquinavir concentration. The solid line was the best fit of a linear equation with slope (4.2 Ϯ 0.2) ϫ 10 7 M Ϫ1 s Ϫ1 and an intercept of 0 to the data. B, enhancement of the intrinsic protein fluorescence after 100 nM G48V/L90M protease⅐saquinavir was mixed with 20 M L-735,524, as described under "Experimental Procedures." Superimposed on the data is the best fit of a pseudo-first order process with a rate constant (k obs ) of 0.528 Ϯ 0.003 s Ϫ1 to the data.
In this model, k 2 and k Ϫ2 refer to the flap closing and opening, respectively. However, the binding of saquinavir to WT and mutant proteases appears to be the one-step process of Equation 1a. If k 2 was much greater than k Ϫ1 , such that most encounters to form E⅐I result in the formation of E⅐I*, the two-step process would appear as a one-step process (30). It was estimated by molecular modeling that flap closing could occur within 12 ps, which suggests that k 2 could be as large as 10 11 s Ϫ1 (33). Furthermore, in a separate case, with an analogue of saquinavir, k Ϫ1 was estimated to be 10 s Ϫ1 (30). Thus, it is possible that k 2 could be very much greater than k Ϫ1 . If k 2 was much greater than k Ϫ1 , then the observed association rate constant (k on ) is defined by the single parameter k 1 . Furthermore, k 1 and k Ϫ1 are association and dissociation from a "loose complex" that has limited molecular interactions; therefore, it is likely that mutations would not significantly affect these rate constants. Because the decrease in affinity of saquinavir and mutant proteases is mediated primarily by the single parameter k off (the dissociation rate constant), it is attractive to attribute the increase in k off to a single step in the mechanism. k 2 and k Ϫ2 define the equilibrium (K 2 ϭ k 2 /k Ϫ2 ) between the "tight and loose complex" (flaps closed or open, respectively). In the "tight complex," intricate molecular interactions between enzyme and inhibitor occur. Therefore, these active site mutations would likely significantly affect k 2 and k Ϫ2 (or the equilibrium con-stant K 2 ). Furthermore, since we proposed that K 2 is a rapid equilibrium (k 2 ϾϾϾ k Ϫ1 ) that greatly favors the tight complex, then k off would increase proportionally with a decrease in K 2 (k off ϭ k Ϫ1 /K 2 ). Finally, in molecular modeling studies, mutation M46I, which, in combination with other mutations, confers viral resistance to VX-478, alters K 2 in the absence of any inhibitor (33). Therefore, we propose that the reduction in affinity of L90M, G48V, and G48V/L90M proteases for saquinavir is completely the result of decreasing K 2 , the equilibrium for flap closing of E⅐I.
Catalytic efficiencies of WT and mutant proteases were determined to test for additional effects of the L90M and G48V mutations. Eight octapeptides that were analogues of the eight cleavage sites that occur during HIV proliferation were used as substrates. The catalytic efficiencies measured for WT protease (Table III) were approximately 1/10 the analogous values reported by Tozser et al. (34) and 10 times the analogous values reported by Moore et al. (35) or Tomasselli et al. (36). HIV protease activity is highly dependent upon assay conditions, such as salt concentration (data not shown); therefore, we were not surprised to see this variation in catalytic efficiency values. L90M and WT protease had similar catalytic efficiencies with the eight substrates. This result was not surprising because some mutations increase catalytic efficiency and reduce affinity for inhibitors (31). In contrast, G48V and G48V/L90M protease had  greatly (90 to 95%) reduced catalytic efficiency for TLNF-PISP, RKIL-FLDG, and AETF-YVDG (Table III). However, these mutant enzymes did not have greatly reduced catalytic efficiencies compared to WT enzyme with four of the eight substrates ((k cat / K m ) mutant /(k cat /K m ) wt Ն 1/3, Table III). The decreases in catalytic efficiency was primarily due to increases in the K m for these substrates. Gulnik et al. (24) determined that mutant proteases with reduced affinity for A-77003 also had a reduced catalytic efficiency as a result of increased K m values for Lys-Ala-Arg-Val-Tyr-Phe(NO 2 )-Glu-Ala-Nle-NH 2 . They have suggested vitality factors ((K i ⅐k cat /K m ) mutant /(K i ⅐k cat /K m ) wt ) as better predictors of viral resistance to drugs than K i values because the vitality factor normalizes the decrease in enzyme affinity for the compensatory loss in catalytic efficiency. While the vitality factor is potentially a useful tool to indicate drug resistance, our studies indicate a more sophisticated tool may be required because mutant enzymes do not show the same reduction in catalytic efficiency for all substrates. Finally, since mutations that reduce protease affinity for inhibitors often reduce the protease catalytic efficiency, it is possible that these mutations ultimately hinder viral proliferation and may self-limit viral resistance to protease inhibitors. In fact, some mutations that confer resistance to A-77003 result in catalytically hampered enzyme and virus with impaired growth (32). Whereas L90M, G48V, and G48V/L90M proteases had substantial reductions in affinity for saquinavir, they did not have similar relative decreases in affinity for three other protease inhibitors presently under clinical evaluation (L-735,524, AG-1343, and VX-478). ABT-538 had significantly less affinity for the G48V and G48V/L90M protease than WT protease, but the decreases were not as large as for saquinavir. As predicted from the affinity of G48V/L90M protease for inhibitors, G48V/L90M virus was nearly as sensitive to L-735,524, AG1343, ABT-538, and VX-478 as was WT virus. In a few cases, with other mutant enzymes, changes in affinity for inhibitors does not correlate well with changes in respective virus inhibition. For example, a relatively small increase in K i of ABT-538 for I50V protease (26-fold) correlates with a relatively large increase (6-fold) in respective virus IC 50 . 3 The vitality factor (data not available) might correlate with antiviral activity better than the K i value. Nonetheless, in general, K i values determined herein for mutant proteases are reasonable predictors of the respective virus susceptibility to inhibitors. Therefore, the kinetic parameters of mutant proteases are useful tools to test "next generation" compounds for their cross-resistance profiles.