Structure of HIV-1 protease in complex with potent inhibitor KNI-272 determined by high-resolution X-ray and neutron crystallography

Abstract

HIV-1 protease is a dimeric aspartic protease that plays an essential role in viral replication. To further understand the catalytic mechanism and inhibitor recognition of HIV-1 protease, we need to determine the locations of key hydrogen atoms in the catalytic aspartates Asp-25 and Asp-125. The structure of HIV-1 protease in complex with transition-state analog KNI-272 was determined by combined neutron crystallography at 1.9-Å resolution and X-ray crystallography at 1.4-Å resolution. The resulting structural data show that the catalytic residue Asp-25 is protonated and that Asp-125 (the catalytic residue from the corresponding diad-related molecule) is deprotonated. The proton on Asp-25 makes a hydrogen bond with the carbonyl group of the allophenylnorstatine (Apns) group in KNI-272. The deprotonated Asp-125 bonds to the hydroxyl proton of Apns. The results provide direct experimental evidence for proposed aspects of the catalytic mechanism of HIV-1 protease and can therefore contribute substantially to the development of specific inhibitors for therapeutic application.

The HIV-1 protease (EC 3.4.23.16) is a dimeric aspartic protease that cleaves the nascent polyproteins of HIV-1 and plays an essential role in viral replication (1–3). Currently, the development of HIV-1 protease inhibitors is regarded as a major success of structure-based drug design (4), and the inhibitors of HIV-1 protease are important compounds for establishing highly-active antiretroviral therapy for AIDS (5). Despite this success, adverse effects linked to the use of HIV-1 protease inhibitors and the emergence of HIV-1 mutants resistant to inhibitor action remain critical factors in the clinical failure of antiviral therapy. The emergence of drug resistance based on the rapid rate of viral replication (6) and the high error rate of reverse transcriptase (7) have become the most urgent concern in HIV-1 treatment.

In considering the goal of effective inhibition of HIV-1 protease, the catalytic mechanism must be understood. The catalytic mechanism of HIV-1 protease has been inferred from the structurally analogous protease pepsin. However, it has been a matter of some debate, with various alternative catalytic mechanisms proposed (8–16). One approach to investigate the catalytic mechanism of HIV-1 protease is to determine the structure of a transition-state analog complex, by high-resolution X-ray and neutron crystallography to identify the location of key hydrogen atoms. As a transition-state mimetic inhibitor, we chose KNI-272 containing allophenylnorstatine (Apns) with hydroxymethylcarbonyl (HMC) isostere and thioproline (Thp), which is highly selective and potent against HIV-1 protease with a picomolar inhibitory constant (10, 17, 18). The previous report of crystal structure showed that the KNI-272 bound to the enzyme exhibited a low energy conformation at an Apns-Thp linkage and 3 water molecules bridged between the inhibitor and enzyme (19). The thermodynamic change upon inhibitor binding indicates that the process is enthalpy driven, presumably because of the burial of a hydrophobic region of the inhibitor; furthermore, the contribution of buried water molecules has been suggested from NMR studies (20, 21).

The HIV-1 protease uses the characteristic aspartyl dyad, which we distinguish by Asp-25 and Asp-125 for consistency with previous reports (19). Interactions between HIV-1 protease and its inhibitor should strongly depend on the ionization state of the catalytic active site; indeed, KNI-272 affinity to HIV-1 protease depends on pH (21). Although the active-site Asp residues are related by a structural dyad in the unbound form, the catalytic mechanism may proceed via asymmetric protonation states. Therefore, determination of the protonation state of the aspartyl groups of enzyme/inhibitor complexes should elucidate the enzymatic mechanism of HIV-1 protease, and subsequently provide knowledge key to improve inhibitor design (22).

To identify the locations of the hydrogen atoms that are important for the catalytic action of HIV-1 protease, we used neutron crystallography. Neutrons strongly interact with hydrogen and deuterium atoms, and neutron-scattering lengths of hydrogen and deuterium atoms are very similar to those of carbon, nitrogen, and oxygen atoms (23). Detection of hydrogen (deuterium) atoms in water molecules also provides useful information about the dynamic feature of protein-bound water molecules; there are at least 4 states of protein-bound water molecules that can be determined by neutron crystallography (24). Neutron diffraction experiments require a relatively large crystal because of the low flux of neutron beams (10⁶ to 10⁹ neutrons cm⁻²s⁻¹) (23). We have already succeeded in preparing a large crystal of HIV-1 protease in complex with a transition-state analog inhibitor KNI-272 (25); thus, we performed neutron structure analysis of HIV-1 protease in complex with KNI-272 to observe key hydrogen atoms involved in the catalytic reaction. Here, we show experimental evidence for the protonation states of the active-site Asp residues of HIV-1 protease obtained through the technique of combining neutron and X-ray crystallography using crystals grown under identical conditions and identical data collection conditions (thus permitting the use of higher-resolution phase information with the lower-resolution neutron-scattering factors).

Results

Structure of HIV-1 Protease Determined by Neutron Diffraction.

The tertiary structure of HIV-1 protease in complex with inhibitor KNI-272 was determined by a joint X-ray/neutron refinement using 1.9-Å neutron diffraction data and 1.4-Å X-ray diffraction data collected at room temperature (Fig. 1). A total of 1,591 hydrogen and 520 deuterium atoms and 143 hydration water molecules were included in the model by comparing the results of neutron and X-ray diffraction. Because the data collection was performed in a buffer prepared in D₂O, 153 side-chain exchangeable hydrogen atoms including catalytic aspartate Asp-25 and KNI-272 were replaced mostly with deuterium atom. The average exchange ratio to deuterium atom was calculated to be 98.6%. There are also 186 exchangeable hydrogen atoms of backbone amides in an asymmetric unit, and average exchange ratio to deuterium atom was calculated to be 53.1% as a result of occupancy refinement (Table S1). This occupancy value is comparable with that in the case of dihydrofolate reductase (26). A total of 797 hydrogen bonding interactions including 149 hydrogen bonds with hydrated waters were determined.

Fig. 1.

Tertiary structure of HIV-1 protease determined by neutron diffraction. The HIV protease dimer is shown by a ball and stick model; water molecules and bound inhibitor are shown by space-filling representation. Hydrogen and deuterium atoms are colored gray. Carbon (green), oxygen (red), nitrogen (blue), and sulfur (yellow) atoms in protease are indicated. Carbon atoms in KNI-272 are colored dark gray. Figs. 1, 2, and 4 were made by using the program Pymol (www.pymol.org).

The overall tertiary structure determined by neutron diffraction at room temperature was closely similar to that of the 0.93-Å resolution X-ray crystal structure determined at 100K: the rmsd for Cα atoms was 0.35 Å.

Interactions Between HIV-1 Protease and KNI-272 at the Active Site.

To further confirm the locations of hydrogen and deuterium atoms in the vicinity of the catalytic residues Asp-25 and Asp-125, the 2F_o − F_c nuclear density map was calculated and contoured at 1.8 σ. The map shows a bulky density near the positions of the Oδ2 atom of Asp-25 and O2 atom of KNI-272 although the deuterium atom (Dδ2) from the carboxyl group in Asp-25 and the deuterium atom (DO2) of the hydroxyl group in the HMC isostere in KNI-272 were omitted (Fig. 2A). The F_o −F_c nuclear density map also shows a strong density belonging to deuterium atoms Dδ2 and DO2 (Fig. 2B). These 2 deuterium atoms bind to the carboxyl oxygen (Oδ2) of Asp-25 and the hydroxyl oxygen atom (O2) of the HMC isostere in KNI-272, respectively. The occupancies of hydrogen and deuterium atoms for both positions were refined to 0.0 and 1.0, respectively. In contrast, no nuclear density for the Dδ2 atom of the carboxyl group in Asp-125 was observed, and no nuclear density between the 2 catalytic aspartic acids was observed despite their short distance (3.1 Å) (Fig. 2B).

Fig. 2.

Neutron and X-ray maps of the active site in HIV-1 protease. (A) 2F_o − F_c nuclear density map contoured at 1.8 σ. The deuterium atoms on Asp-25 (labeled Dδ2) and KNI-272 (labeled DO2) are shown, but were omitted for map calculation. (B) F_o − F_c omit nuclear density map calculated without the contribution of the Dδ2 and DO2 atoms. Maps are shown at 4.5 σ (red) and 5.5 σ (blue) levels. The 2 omitted deuterium atoms are colored cyan. (C) 2F_o − F_c (cyan) and F_o − F_c (red) electron density maps contoured at 1.5- and 2.5-σ values were drawn at 0.93-Å resolution.

Although no significant additional electron density corresponding to these hydrogen atoms could be seen in the 0.93-Å resolution X-ray structure (Fig. 2C), the bond lengths between the carbon (Cγ) and the 2 oxygen atoms (Oδ1 or Oδ2) in Asp-125 are nearly equivalent (1.27 and 1.25 Å, respectively), whereas the bond lengths between Cγ and Oδ1 or Oδ2 in the protonated Asp-25 are 1.20 and 1.32 Å, respectively (Table 1). For comparison, the bond lengths between Cγ and Oδ1 or Oδ2 in Asp-29 and Asp-129, which are responsible for inhibitor binding, are listed in Table 1. These results show that only the Oδ2 atom of Asp-25 is protonated, and Asp-125, Asp-29, and Asp-129 are deprotonated.

Table 1.

Bond distances between carbon and oxygen atoms in the carbonyl group of Asp residues responsible for interaction with KNI-272

Residue	Carbon atom	Oxygen atom	Distance, Å	Estimated SD, Å
Asp-25	Cγ	Oδ1	1.203	0.007
Asp-25	Cγ	Oδ2	1.322	0.007
Asp-29	Cγ	Oδ1	1.242	0.008
Asp-29	Cγ	Oδ2	1.261	0.009
Asp-125	Cγ	Oδ1	1.265	0.008
Asp-125	Cγ	Oδ2	1.251	0.008
Asp-129	Cγ	Oδ1	1.244	0.008
Asp-129	Cγ	Oδ2	1.250	0.008

The bond distances were calculated by using the structure determined by 0.93-Å resolution X-ray crystallography.

Both side chains of Asp-25 and Asp-125 formed 2 strong hydrogen bonds (≈2.7-Å distance) with HMC of KNI-272. One is between the Dδ2 of Asp-25 and the O4 of HMC and the other is between DO2 of hydroxyl group in HMC and Oδ1 of carboxyl group in deprotonated Asp-125. The distances between the deuterium and acceptor atoms are relatively shorter (1.8–1.9 Å) than other hydrogen bonds with KNI-272 (Table 2).

Table 2.

Direct hydrogen bonds and water molecule-mediated hydrogen bonds between HIV-1 protease and KNI-272 determined by neutron crystallography

Donor atom	Deuterium atom	Acceptor atom	Distance*, Å	Distance†, Å	Angle‡, °
Asp-25 Oδ2	Asp-25 Dδ2	KNI O2	3.0 [3.1]	2.3	124
Asp-25 Oδ2	Asp-25 Dδ2	KNI O4	2.7 [2.6]	1.9	144
Asp-29 N	Asp-29 D	HOH608 O	3.0 [2.9]	2.1	160
Ile-50 N	Ile-50 D	HOH301 O	3.1 [3.0]	2.3	157
Asp-129 N	Asp-129 D	KNI O3	2.9 [2.9]	2.0	158
Ile-150 N	Ile-150 D	HOH301 O	3.1 [2.9]	2.2	165
KNI N2	KNI DN2	Gly-148 O	2.9 [2.8]	2.2	131
KNI N3	KNI DN3	Gly-127 O	3.1 [3.1]	2.2	156
KNI N5	KNI DN5	HOH608 O	3.1 [3.1]	2.1	167
KNI O2	KNI DO2	Asp-125 Oδ1	2.7 [2.6]	1.8	154
KNI O2	KNI DO2	Asp-125 Oδ2	3.2 [3.2]	2.4	144
HOH301 O	HOH301 D1	KNI O6	2.8 [2.9]	1.9	146
HOH301 O	HOH301 D2	KNI O5	2.8 [2.8]	1.8	168
HOH566 O	HOH566 D1	Asp-129 Oδ1	3.0 [2.9]	2.0	163
HOH566 O	HOH566 D2	Gly-127 O	2.8 [2.8]	2.1	128
HOH566 O	HOH566 D2	KNI O3	3.2 [3.0]	2.6	117

The numbers in square brackets indicate the value obtained from X-ray structure at 100 K.

*Distance between donor and acceptor atoms.

^†Distance between deuterium and acceptor atoms.

^‡Angles of donor–deuterium–acceptor atoms.

Water Molecules Localized at the Interface Between KNI-272 and HIV-1 Protease.

There are 6 water molecules (HOH301, HOH322, HOH354, HOH566, HOH607, and HOH608) within a distance of 3.5 Å from KNI-272 as shown in Fig. 3 and Table S2). Deuterium atoms of these water molecules were confirmed in the 2F_o − F_c and F_o −F_c omit maps. Three water molecules (HOH301, HOH566, and HOH608) mediating hydrogen-bonding interactions between HIV-1 protease and KNI-272 show clear nuclear densities for deuterium as shown in Fig. 4. The water molecule, HOH301, located at the symmetric position of the dimer interface forms hydrogen bonding interactions with KNI-272 and the main chain nitrogen (N) atoms of Ile-50 and Ile-150 in the “flap” structure of HIV-1 protease (Fig. 4A). HOH607 participates as a proton donor in hydrogen bonding interactions with Gly-27 and the catalytic residue Asp-125. Collectively, a total of 16 hydrogen bonds were identified in the interaction between HIV-1 protease and KNI-272, and 9 of 16 hydrogen bonding interactions were water mediated.

Fig. 3.

Schematic diagram of the interaction between HIV-1 protease and KNI-272 (bold lines). Hydrogen bonds are shown by broken lines. Asterisks indicate the hydrogen atoms replaced with deuterium atom (occupancies of deuterium atom are >0.5).

Fig. 4.

Neutron and X-ray maps of water molecules involved in the water-mediated hydrogen bonds between HIV-1 protease and KNI-272. Shown are the structural environment of HOH301 (A), HOH566 (B), and HOH608 (C). 2F_o − F_c X-ray maps drawn in blue were contoured at 3.0 σ. F_o − F_c nuclear density maps drawn in red were calculated after omitting D1 or D2 atoms of water molecules and are contoured at the 3.0-σ level. Hydrogen bonds are represented by dotted lines in pink.

Discussion

KNI-272 is a unique inhibitor designed as a transition-state analog for peptide bond hydrolysis by HIV-1 protease (17, 18). Because the locations of the hydrogen atoms of catalytic aspartates in HIV-1 protease have not previously been experimentally determined, neutron crystallography of HIV-1 protease in complex with KNI-272 was performed to permit elucidation of the enzymatic mechanism of HIV-1 protease.

The neutron structure determination of HIV-1 protease demonstrated that the carboxyl group of Asp-25 is protonated, whereas that of Asp-125 is deprotonated. In previous reports, the ionization states of the catalytic residues of HIV-1 protease in complex of KNI-272 or pepstatin A were investigated by pD (or pH) dependence of chemical shift and the H/D isotope effect by using ¹³C NMR (27, 28). Although the chemical shift of both Asp-25 and Asp-125 had not been changed through pD 2.5–6.2, the results of the isotope effect indicated protonation of Asp-25. In contrast, ab initio molecular dynamics-based assignment suggested that both aspartic groups are protonated, and a hydrogen bond is formed between the 2 aspartic acids (29). When considering these 2 alternatives, the neutron diffraction data support protonation of Asp-25 and the isotope effect results.

It has been suggested that a low-barrier hydrogen bonding interaction, a unique hydrogen bond with 2 acceptors sharing a hydrogen atom (30, 31), occurs between the carboxyl groups of Asp-25 and Asp-125 in the absence of bound inhibitor (11) and in product complex (14). However, we could not detect any significant nuclear density between Asp-25 and Asp-125 in our analysis, probably because the side-chain conformations of these catalytic aspartates are fixed by hydrogen-bonding interaction with the main-chain N atoms of Gly-27 and Gly-127.

The catalytic mechanism of HIV-1 protease has been extensively investigated by several approaches (12). Our determination of the protonation state and location of deuterium atoms in the enzyme/inhibitor complex will help further understanding of the enzymatic mechanism of HIV-1 protease. As shown in Fig. 5, we summarize a possible model for the catalytic mechanism of HIV-1 protease consistent with our current neutron diffraction data and previous literature (8, 10, 15, 32, 33). The structure of the HIV-1 protease/KNI-272 complex displays the structural characteristics of a tetrahedral transition-state complex (state *1 in Fig. 5). Whereas the hydrogen bond between the carbonyl group in HMC and Asp-25 appears to be the primary interaction with substrate, the location of the hydroxyl group in HMC appears ideal to mimic the location of an attacking water molecule in catalysis. Taken together, our results demonstrate that Asp-25 provides a proton to the carbonyl group of the substrate and Asp-125 contributes to activate the attacking water molecule as a nucleophile. Because the inhibitor does not allow hydrolysis to proceed, it is concluded that the protonation states of both the catalytic residues in the KNI-272 complex retains those within the ES complex.

Fig. 5.

A proposed energy profile for the catalysis by HIV-1 protease. E, P, and S denote enzyme, products, and substrate, respectively. The reaction coordinate of the separated states is referred from the results of Bjelic and Aquvist (15).

There are also several reports describing the X-ray structure of HIV-1 protease in complex with not only product (14, 34, 35) but also a tetrahedral intermediate (16, 33, 36). It appears that the 2 oxygen atoms of hydroxyl (O2) and carbonyl oxygen (O4) in HMC of KNI-272 are located at similar positions to the oxygen atoms in the gem-diol of hydrated substrate (16) and hydrated KVS-1 inhibitor (33). However, Kumar et al. (36) reported an important observation of the interaction between the catalytic residues and the intermediate. They observed hydrogen bond formation between one catalytic residue and nitrogen atom of the intermediate, which is distinct from our observation; and each observation may suggest a different step in catalytic reaction.

Although current inhibitor design is focused on improvements in affinity and specificity for target proteins, the inhibitor design for HIV-1 protease is more focused on its drug resistance; one of the efficient approaches is to improve interaction with the structurally-essential regions in HIV-1 protease, such as catalytic residues or key regions of the protein backbone. According to calorimetric binding experiments performed as a function of pH, efficient KNI-272 binding requires protonation of a catalytic residue because the affinity of KNI-272 decreases above pH 6 (21) (likely caused by the deprotonation of Asp-25 from our neutron data). This insight leads us to hypothesize that an efficient inhibitor would retain key interactions seen in KNI-272 binding and neutralize the charged state of the catalytic residues. Consequently, introduction of a positive charge in an inhibitor at the active site would be favorable because of the charge–charge interaction with a deprotonated catalytic residue isolated from solvent. With regard to the use of backbone interactions with an inhibitor, there is an example of darunavir containing bis-THF forming hydrogen-bonding interactions with main-chain atoms of Asp-29 and Asp-30, which might account for potent inhibition against highly drug-resistant mutants of HIV-1 protease (37).

Displacement of a water molecule interacting with an inhibitor might lead to increased affinity to HIV-1 protease mutants from an entropic point of view. Six water molecules (HOH301, HOH322, HOH354, HOH566, HOH607, and HOH608) directly bound to KNI-272 may be candidates for displacement (Table S2 and Fig. 3). Indeed, displacing a water molecule corresponding to HOH301 was used for designing cyclic urea-based inhibitors (38). The results of NMR (water/NOESY and water/rotating-frame Overhauser effect spectroscopy) showed that HOH301, HOH566, and HOH608 are long-lived water molecules responsible for binding of KNI-272 to HIV-1 protease (20). It was reported that the favorable binding enthalpy is caused by interactions with long-lived water molecules around KNI-272 (21). Molecular dynamics simulation indicated that HOH301 and HOH607 significantly contribute to the binding free energy (39). These water molecules have relatively small B factors in the structures determined by 1.9-Å neutron and 0.93-Å X-ray structure analyses. Overall, therefore, the structural analyses including protonation status of the catalytic residues and the information for the bridging water molecules determined in this study provide important information applicable to the design of potentially novel and specific HIV-protease inhibitors.

Methods

Preparation of HIV-1 Protease.

Preparation of HIV-1 protease was performed as reported (25). In brief, the chemically-synthesized DNA encoded the gene for the initial methionine and the 99-aa HIV-1 protease including 5 mutations of Q7K, L33I, L63I, C67A, and C95A (to prevent autoproteolysis and cysteine thiol oxidation) (40) were used for expression in Escherichia. coli. The expressed protein was refolded and purified by cation exchange chromatography followed by reversed-phase chromatography. The final yield of HIV protease was ≈30 mg from 6-L cultivation determined by using UV absorption at 280 nm with a molecular extinction coefficient of 8,600 cm⁻¹M⁻¹.

Crystallization of HIV-1 Protease.

Crystallization and crystal growth were performed as reported (25). Briefly, a crystal of HIV-1 protease complexed with KNI-272 (≈0.3 × 0.3 × 0.1 mm) was grown and used as a seed to grow a larger crystal suitable for neutron diffraction studies. Approximately 2 mg of protein solution in 0.125 M citrate/0.25 M phosphate buffer (pH 5.5) was used for seed crystal growth; this crystal was then transferred to a protein solution containing 1.5 mg/mL of HIV-1 protease. A large crystal was subsequently obtained with the dimension of 3.6 × 2.0 × 0.5 mm. For neutron diffraction data collection, the crystal was soaked for 2 weeks at 293 K in the crystallization solution (pD 5.0), containing D₂O. This pD 5.0 is near the optimum in enzymatic activity of HIV-1 protease (11). A crystal obtained at the same crystallization condition was used for X-ray diffraction data collection at room temperature. Before data collection, the crystal was also soaked into D₂O based crystallization solution (pD 5.0) at 293 K.

Data Collection.

Neutron diffraction data were collected to 1.9-Å resolution by using a crystal (3.6 × 2.0 × 0.5 mm) at room temperature on the BIX-4 diffractometer (41) installed at the 1G-A site of the JRR-3 reactor in the Japan Atomic Research Agency. Data collection was carried out by using the step-scan method with an interval angle of 0.3° and exposure times of 360 min per frame. The total time required to collect the total of 181 frames was 46 days. The diffraction data were integrated, scaled, and merged by using the programs DENZO and SCALEPACK (42). The detailed statistics of data collection are listed in Table S3. Additional X-ray diffraction data were collected to 1.4-Å resolution at room temperature for the use of joint refinement. The data collection was performed by using the oscillation method with rotation angle of 1° and exposure times of 5 s per frame at BL6A at the Photon Factory (Ibaraki, Japan). The diffraction data were integrated, scaled, and merged by using the program HKL2000 (42).

Ultra high-resolution X-ray diffraction data were collected to 0.93 Å with a crystal (0.6 × 0.2 × 0.1 mm) at 100 K at the BL41XU beamline at SPring-8 (Hyogo, Japan). The crystal was soaked into the precipitant solution containing 45% (wt/vol) glycerol, then flash-frozen under a N₂ gas cryostream (100 K). The high-resolution diffraction dataset containing 180 frames was collected with exposure times of 10 s per frame and by changing the X-ray beam position along the crystal every 36 frames. Subsequently, a dataset to accurately collect the lower-resolution diffraction data was collected with an exposure time of 1 s per frame. The diffraction data containing a total of 360 frames were integrated, scaled, and merged by using the programs HKL2000 (42). The detailed statistics of data collection are listed in Table S3.

Structure Determination.

Refinement of the 0.93-Å-resolution X-ray structure was carried out by using the program CNS (43) followed by SHELX-97 (44) with manual model adjustment using XtalView (45). X-ray structure [Protein Data Bank ID code 1HPX (19)] was used as the starting model for refinement. Each crystallographic asymmetric unit contained a HIV-1 protease dimer comprising 2 equivalent monomers residues distinguished by the residue numbers 1–99 and 101–199. Although HIV-1 protease was expressed with an extra Met at the N terminus, any electron density maps of the Met residue were not observed. Hydrogen atoms were included in the model by SHELXL-97 as riding hydrogens. Restraints (DFIX and DANG) of carboxyl atoms in Asp residues were finally released. SHELX-97 was used to calculate the estimated standard deviation for carboxyl bond length. The final model contained alternative conformations of 36 amino acid residues. Refinement statistics are also given in Table S3.

A coordinate set for neutron structure was uniquely obtained by using a joint refinement method (46) with the computer program PHENIX (47) and the 1.9-Å neutron and 1.4-Å X-ray diffraction datasets collected at room temperature with a crystal prepared under identical conditions. X-ray structure determined at 0.93-Å resolution was used as a starting model of refinement. The hydrogen atoms in the protein and water were initially placed by using programs PHENIX (47) and CNS (43). These locations were manually adjusted with the program XtalView (45). Solvent-accessible and -exchangeable hydrogen atoms in side chains of Arg, Asn, Gln, His, Lys, Ser, Thr, Trp, and Tyr were replaced with deuterium atoms. Both hydrogen and deuterium atoms were added to the buried side chains (Thr-26, Thr-31, and Thr-80) and the side chains of catalytic residues and KNI-272 at exchangeable sites, and the occupancies of deuterium/hydrogen atoms were subsequently evaluated. There are 3 asparagines and 5 glutamines in the HIV-1 protease structure; the side-chain amide conformations for Gln-2, Gln-18, Gln-58, Gln-102, and Asn-198 were corrected based on the nuclear density map from neutron crystallography. D₂O molecules were placed according to the locations of oxygen atoms determined by using the 1.4-Å-resolution X-ray data, and then deuterium atoms were placed into the positive densities observed in neutron 2F_o − F_c maps. Finally, 143 water molecules were included for the joint refinement by using neutron and X-ray diffraction data. The R and R_free values for the final model were 19.3% and 22.2%, respectively, as summarized in Table S3.