Mutational and Structural Studies Aimed at Characterizing the Monomer of HIV-1 Protease and Its Precursor*

An experimental protocol for folding the mature human immunodeficiency virus-1 (HIV-1) protease is presented that facilitates NMR studies at a low protein concentration of ∼20 μm. Under these conditions, NMR spectra show that the mature protease lacking its terminal β-sheet residues 1-4 and 96-99 (PR5-95) exhibits a stable monomer fold spanning the region 10-90 that is similar to that of the single subunit of the wild-type dimer and the dimer bearing a D25N mutation (PRD25N). Urea-induced unfolding monitored both by changes in 1H-15N heteronuclear single quantum correlation spectra and by protein fluorescence indicates that although PR5-95 monomer displays a transition profile similar to that of the PRD25N dimer (50% unfolded (U50) = ∼1.9 m), extending the protease with 4 residues (SFNF) of its N-terminally flanking sequence in the Gag-Pol precursor (SFNFPRD25N) decreases the stability of the fold (U50 = ∼1.5 m). Assigned backbone chemical shifts were used to elucidate differences in the stability of the PRT26A (U50 = 2.5 m) and SFNFPRD25N monomers and compared with PRD25N/T26A monomer. Discernible differences in the backbone chemical shifts were observed for N-terminal protease residues 3-6 of SFNFPRD25N that may relate to the increase in the equilibrium dissociation constant (Kd) and the very low catalytic activity of the protease prior to its autoprocessing at its N terminus from the Gag-Pol precursor.

The HIV-1 2 genome encodes a protease as part of the large Gag-Pol precursor. Like all retroviral proteases, HIV-1 protease is active only as a homodimer. In addition to catalyzing its own release from the Gag-Pol polyprotein via cleavage at its N and C termini to produce the mature protease (PR) domain, the protease is responsible for cleaving the Gag and Gag-Pol polyproteins at specific sites to produce the mature structural (matrix, capsid, and nucleocapsid) and functional proteins (reverse transcriptase, RNase H, and integrase) crucial for virus matu-ration and propagation (1,2). This indispensable role played by the protease in the viral replication cycle motivated extensive studies of structure-based drug design of active site inhibitors of the mature protease (3,4). Various protease inhibitors are currently being used and developed for the treatment of HIV/ AIDS. However, effective long term treatment of AIDS patients has been hampered by the rapid emergence of drug-resistant variants that are less susceptible to inhibition even under highly active antiviral therapy (5,6). Recent second generation inhibitors of the PR, designed to overcome resistance (i) by reducing the size of their hydrophobic groups so that mutation of the active site residues will have less effect on inhibitor affinity and (ii) by optimizing favorable polar interactions with main chain atoms and to conserved residues, have been shown to curtail drug resistance (7).
The initial steps in the maturation of the protease (Gag-Pol precursor) involve the folding and dimerization of the protease domain when it is in the form of the large Gag-Pol precursor. Previous studies using a mini-precursor, in which the protease is flanked by sequences corresponding to the cleavage sites at its termini, showed that the processing at the N terminus of the protease, which is concomitant with the appearance of catalytic activity, precedes the C-terminal cleavage (8 -10). Further examination of protease precursor containing the flanking transframe region sequences ( Fig. 1) by kinetics and NMR revealed that the very low catalytic activity of the protease precursor prior to the cleavage at the N terminus is due to a much higher dimer dissociation constant (K d ) as compared with the mature protease (11). Upon intramolecular cleavage at the N terminus, the protease forms a stable dimer and exhibits a very low K d (Ͻ10 ϫ 10 Ϫ9 M in 50 mM sodium acetate, pH 5, at 25°C (8,13)). This suggests that inhibition of the protease function either by preventing or by disrupting dimer formation prior to its maturation provides an attractive avenue for inhibitor design.
For the mature wild-type protease, protein folding appears to be concomitant with dimerization at a concentration of 5 nM and above (11, 14 -16). Therefore, a systematic understanding of the folding and dimerization events and the properties of the protease both as a monomer and as a dimer in its precursor and mature forms is essential for the rational design of inhibitors of dimerization. There have been several studies reporting the development of dimerization inhibitors toward the mature HIV-1 protease based on the terminal interface conformation, but to date, the structures of these complexes have not been verified. The low K d of the mature protease had also precluded detailed studies of the protease monomer until it was observed that subtle mutations of conserved residues involved in interface contacts increase the K d by several orders of magnitude, thus allowing NMR studies of the monomer to be performed at 0.2-0.3 mM concentrations. The conserved intra-monomer contact between Asp-29 and Arg-87 residues and the active site inter-monomer (interface) contact of Asp-25 and Thr-26 residues play a key role in stabilizing dimer formation with T26A exhibiting the largest increase in K d (Table 1 and Fig. 2) (17)(18)(19). In addition, interface contacts between the terminal 1-4 and 96 -99 residues are also critical for dimerization (20). The inter-subunit ␤-sheet contacts between the C-terminal strands (96 -99/96Ј-99Ј, the prime indicates the second subunit) have a larger influence on dimerization than those between the Nand C-terminal ␤-strands (1-4/96Ј-99Ј (17)). Thus, the construct bearing the 4-residue deletion at the C terminus of the mature protease, which does not significantly dimerize even up to 1 mM concentration, facilitated the determination of the first three-dimensional structure of the protease monomer by NMR (Fig. 2) (21).
Sequences flanking the protease domain also influence dimer formation, although the exact molecular mechanism of their action on dimerization is not understood. Unlike the flanking C-terminal residues (reverse transcriptase region), which do not appear to affect the catalytic activity or the K d in fusion with the protease (9,22,23), the native transframe region, which flanks the N terminus of the protease in the Gag-Pol precursor, drastically influences the dimer stability of the protease indicated by the very low catalytic activity of the protease prior to its maturation at its N terminus ( Fig. 1) (10,11). Even a two to four residue extension at the N terminus of the protease significantly impairs dimerization similar to the 4-residue N-terminal deletion construct PR   (21). Although the N-terminal extensions destabilize the dimer, our NMR investigation of the protease precursor flanked by 56 amino acids of the native transframe region at the N terminus revealed that although the transframe region is mainly unstructured, the protease domain adopts a monomer fold that is similar to that of the mature protease monomer PR   (21).
Despite our extensive studies to characterize the monomer, the propensity for these monomers to aggregate at 0.2-0.3 mM concentrations had precluded further studies to determine the stability of the monomer fold as compared with the dimer, and importantly, to use the monomer to reliably access the binding of potential inhibitors of dimerization. Herein to facilitate these studies, we initially optimized conditions for protein folding and for acquiring 1 H-15 N HSQC spectra at ϳ20 M concentration using the high sensitivity cryogenic NMR probe. Optimizing the NMR experiments at low protein concentration notably allowed studies of the urea-induced unfolding of the monomer and dimer by monitoring protein fluorescence under similar equilibrium conditions. Under these optimized conditions, the minimal sequence required for monomer folding was accessed by successive terminal deletions. Subsequently, urea-induced unfolding was monitored both by fluorescence and by NMR, and backbone chemical shifts of the monomers were assigned to characterize the effect of the mutation on the stability of the fold and the monomer-dimer equilibrium. Finally, 1 H-15 N HSQC spectra of the monomer and the dimer were acquired in the presence of excess putative dimerization inhibitors of the mature protease. Contrary to the results that suggest that these peptides dissociate the dimer (6,24), there was no change in the chemical shifts of residues at the termini or the active site. These observations indicate the lack of interaction between either the monomer or the dimer with the inhibitor under the conditions tested.

EXPERIMENTAL PROCEDURES
Protease Constructs-The mature PR (11) domain in all constructs, optimized for NMR and kinetic studies, bears 5 mutations, Q7K, L33I, L63I to minimize autoproteolysis and C67A  All constructs bear five mutations, three mutations, Q7K, L33I, and L63I, that restrict autoproteolysis and two mutations, C65A and C95A, to avoid Cys-thiol oxidation. The site of mutation(s) is subscripted. SFNF PR D25N denote four C-terminal residues SFNF of p6 pol fused to the N-terminus of PR D25N . Constructs marked with an asterisk were extensively used in this study. Proteins were prepared using both the dialysis and quench protocols of protein folding at a final concentration of ϳ20 M and 200 M, respectively, in 50 mM sodium acetate buffer, pH 5. K d values were estimated based on 1 H-15 N HSQC spectra recorded at 20°C. m, u, and ua denote monomer, unfolded, unfolded aggregate, respectively. At ϳ20 M concentration, PR and PR D25N are mostly dimeric, and PR T26A , PR D25N/T26A , PR 1-95 , PR  , and PR 5-95 are mostly monomeric. Further deletions into the protease (PR 1-90 , PR 5-90 , PR 10 -90 ) lead to destabilization of the monomer fold and aggregation. Addition of DMP323, a potent active site inhibitor, promotes ternary complex formation (dimer ϩ DMP323) of PR T26A , PR D25N/T26A , PR 5-99 , and SFNF PR D25N but not of PR 1-95 and PR  .
Deletion mutants *PR 1-95 Ͼ1000 *PR 5-99 500 *PR 5-95 Ͼ1000 *PR  m/u/ua *PR  m/u/ua *PR  u/ua Extension mutant *SFNF PR D25N Ͼ500 and C95A to prevent cysteine-thiol oxidation. Plasmid DNA (pET11a, Novagen, Madison, WI) encoding PR (11) was used with the appropriate oligonucleotide primers to generate the constructs PR D25N and PR D25N/T26A . The PR encoding plasmid was sequentially extended one codon at a time to produce SFNF PR. SFNF PR template was then used to introduce a D25N mutation. Construction of PR 5-99 and PR 1-95 has been described before (11,17). A stop codon was engineered into constructs PR 5-99 to produce PR 5-90 and PR 5-95 and in PR 1-95 to produce PR 1-90 . All constructs were generated using the QuikChange mutagenesis protocol (Stratagene, La Jolla, CA). To express PR 10 -90 , the region encoding residues 10 -90 was amplified using PR template and appropriate forward and reverse primers and cloned into pET11a vector between NdeI and BamH1 sites. A Gly residue was included at the N terminus of PR 10 -90 to allow nearly complete excision of the N-terminal methionine and attain N-terminal homogeneity. All constructs were verified both by DNA sequencing and by mass spectrometry.
Protein Preparation-Escherichia coli BL21(DE3) were grown in minimal media containing 15 N ammonium chloride with or without 13 C glucose as the sole nitrogen and carbon sources, respectively, at 37°C and induced for expression. Proteins were purified from inclusion bodies using an established protocol as described previously involving size-exclusion chromatography under denaturing conditions followed by reversephase high pressure liquid chromatography. Peak fractions (ϳ0.5 mg/ml) were stored in aliquots at Ϫ70°C. Proteins were dialyzed extensively against 30 mM formic acid and concentrated to ϳ2 mg/ml using Millipore YM-10 Centriprep and Centricon concentrators (Millipore, Beford, MA) and stored at 4°C. Samples were folded for NMR, kinetics, and fluorescence studies freshly when needed using the protocol described in Fig.  3. Concentrations of PR and its mutants are given for a monomer, unless noted otherwise.
Fluorescence and Kinetics-Intrinsic fluorescence was recorded using FluoroMax-3 fitted with a Peltier temperature controller (Horiba Jobin Yvon, Edison, NJ) over a period of 30 min at 20°C with an excitation and emission wavelength of 285 and 350 nm, respectively, and a bandwidth of 2 nm. Protein maintained in 30 mM formic acid was diluted to a volume of 1.5 ml in 5 mM sodium acetate, pH 6, and the resulting solution was then 2-fold diluted with either 0.1 M sodium acetate, pH 5 (solution A: folded) or 10 M urea, 0.1 M sodium acetate, pH 5 (solution B: unfolded) to achieve a final protein concentration of ϳ10 M in 3 ml. Solution A was transferred into a 4-ml cuvette and allowed to equilibrate in the instrument for 5-10 min, and the fluorescence reading was acquired over a period of 30 min. Normally a stable reading was achieved in about 15 min. For urea denaturation studies, fluorescence was measured in incre-  (21)). The two chains of the dimer are shown in green and red. Active site interface residues Asn-25 and Thr-26 are displayed as ball-and-stick models. Conserved residues Asp-29 and Arg-87, which form the intra-monomer contact critical to the stability of the dimer, are labeled. The region spanning 10 -90 of the monomer adopts a fold that is similar to that of the single subunit of the dimer. Residues 1-5 and 5-9 shown in C in white and yellow, respectively, exhibit no specific structure in the monomer. The minimal region required for a stable monomer fold spans the region 5-95 as described in this study. The dialysis protocol consumes ϳ6 h of total process time to prepare folded protease at a high protein concentration of ϳ0.3 mM. In the quench protocol of folding, unfolded protein is maintained as a stock solution of ϳ2 mg/ml in 30 -50 mM formic acid at 4°C. When needed, the folded protein is prepared within 5 min to a final concentration of ϳ20 M as described. In combination with the CryoProbe, the latter method of protein folding is highly suitable for acquiring 1 H-15 N HSQC spectra to probe potential inhibitors of dimerization and compounds that interact with the monomer. TFA, trifluoroacetic acid. ments of 0.25 M urea by removing solution A from the cuvette and adding an equal volume of solution B up to a final concentration of 4.75 M urea.
Protease activity was monitored using the substrate Lys-Ala-Arg-Val-Nle-(4-nitrophenylalanine)-Glu-Ala-Nle-NH 2 (California Peptide Research, Napa, CA) in 50 mM sodium acetate buffer, pH 5, at 25°C. Proteins were folded in the same manner as prepared either for fluorescence or for NMR experiments. Protein concentrations were determined both spectrophotometrically (absorbance at 280 nm) and by Bio-Rad assay (Bio-Rad Laboratories). Specific cleavages giving rise to products due to the autoprocessing of active precursor proteases and autoproteolysis of active mature proteases were accessed both by SDS-PAGE and by mass spectrometry.
NMR Spectroscopy-Samples for acquiring NMR spectra were prepared by mixing 40 -45 l of protein with 105-110 l of 5 mM sodium acetate buffer, pH 6, and subsequently either with 150 l of 0.1 M acetate buffer, pH 5, or with the same buffer containing urea. All 1 H-15 N-correlation spectra were recorded using 20 -25 M protein (in monomer) in 50 mM acetate buffer at pH 5 in 95% H 2 O/5% D 2 O and a sample volume of ϳ321 l in a 5-mm Shigemi tube (Shigemi, Inc., Allison Park, PA). Putative dimerization inhibitors (California Peptide Research) were either mixed with 5 mM acetate buffer (pH adjusted to 6) prior to the protein folding step or mixed with 0.1 M acetate buffer (pH adjusted to 5) when the pH is set (Fig. 3, Quench protocol). Backbone chemical shifts of C ␣ and N were determined using HNCA experiments (33) for 0.2-0.3 mM proteins in the same buffer condition as above. Spectra were acquired on DMX500 spectrometers with a CryoProbe (Bruker Instruments, Billerica, MA) at 20°C. NMR data were processed and analyzed using the nmrPipe, nmrDraw, and PIPP software (34,35).

RESULTS AND DISCUSSION
Rationale and Experimental Design-The wild-type mature protease catalyzes its own cleavage rapidly (termed autoproteolysis), leading to severe loss in catalytic activity even at low nM concentrations. Thus, solution NMR studies of the uninhibited dimeric protease are feasible for a limited period of time only by using an optimized PR construct, to limit autoproteolysis, and by conducting experiments at pH 5.8, which is above the optimal pH for catalytic activity. The optimized PR construct for expression bears the mutations Q7K, L33I, L63I, C67A, and C95A, the first 3 to limit autoproteolysis and the latter 2 to avoid cysteine-thiol oxidation, leading to protein aggregation (11,25,26). Because these mutations do not discernibly affect either the K d or the kinetics of PR as compared with the wildtype protease, PR has been used as a pseudo wild-type for successive mutational studies and for comparison of the fold, stability, and catalytic properties of the mature protease and its precursor forms (11,26,27). An active site D25N mutation, to abolish catalytic activity, was introduced into PR to permit long term studies of either the free or the substrate-bound protease dimer in solution (28). This mutation, however, increases the K d by Ͼ50-fold as compared with PR, which is ϳ1000-fold less than other interface and non-interface mutations ( Table 1).
Studies of protease monomers at 0.2-0.3 mM concentration require suppressing both the autoproteolysis in mutants, where there is significant dimer formation in NMR experimental conditions, as well as the propensity of the folded monomers to aggregate. In addition, previous monomer studies required a starting protein amount of ϳ2 mg, two dialysis steps, and concentrating the protein to ϳ0.3 mM; together these steps consume about 6 h of total preparation time (Fig. 3, Dialysis  protocol).
To minimize both aggregation and autoproteolysis, we have performed NMR studies of mutant proteases at ϳ20 M in monomer concentration using high sensitivity cryogenic probes. We have also devised a simpler folding scheme that is accomplished within 5 min (quench protocol) with reproducible monomer folding efficiency of Ͼ90% and negligible signal loss, even 3 months after sample preparation. Thus, 1 mg of protein is sufficient for at least seven NMR samples at a final monomer concentration of ϳ20 M. Specific extension, substitution, and deletion mutations were created in PR (Table 1) to access folding and stability of these mutants monitored both by NMR and by fluorescence measurements. Autoproteolysis was almost non-existent in all of the mutants studied because of the high K d (Ͼ2 orders of magnitude as compared with PR) except for PR  , which exhibits minor amounts of degradation products even at 20 M concentration. The autoproteolysis observed for PR 5-99 is consistent with our earlier studies showing that it exhibits significant dimer formation and catalytic activity assayed using the chromogenic substrate (11). Additionally, a D25N mutation was introduced in some constructs to totally abolish catalytic activity to allow acquiring data for extended periods of time. Protein folding, stabilities, and K d values were accessed and compared for three substitution, six deletion, and one extension mutants (Table 1, constructs marked with an asterisk).

PR Lacking the Terminal Residues 1-4 and 96 -99 Exhibits a Stable Monomer
Fold-In previous studies, we had shown that a 4-residue deletion at either the N terminus (PR 5-99 ) or the C terminus (PR 1-95 ) of the protease shifts the monomer-dimer equilibrium toward the monomer. In 1 H-15 N HSQC spectra recorded using 0.2-0.3 mM protein, PR 5-99 exhibited signals characteristic of a mixture of monomers and dimers, whereas PR 1-95 was almost exclusively monomer (17). However, previous assessment of the mutant bearing 4-residue deletions at both termini (PR 5-95 ) of the mature PR was deemed unreliable due to a large decrease in the observed signal intensities of the HSQC spectrum immediately after protein preparation using the dialysis protocol (18). Now we have re-examined PR 5-95 that was prepared using the quench protocol at a final concentration of ϳ20 M, thus avoiding the step for concentrating the protein after protein folding as in the case of the dialysis protocol. Comparison of the 1 H-15 N HSQC spectra of PR 5-99 and PR   (Fig. 4, A and B) indicates that PR 5-95 exhibits a monomer fold similar to that of PR 5-99 with negligible aggregation. These two spectra are nearly identical to that of other monomer constructs described previously (17,18). Although peaks that correspond to the dimer were not observed in the PR 5-99 spectrum at ϳ20 M, the minor signals in the center of the spectrum of PR   (Fig. 4A, 8.0 -8.5 ppm for 1 H and 108 -128 ppm for 15 N) most likely arise from unstructured fragments of the protease due to autoproteolysis and a very small fraction of unfolded protein. Subjecting PR 5-99 to SDS-PAGE analysis after acquiring the HSQC spectrum reveals fragments of the protease resulting from autoproteolysis. Based on the observations that no catalytic activity was detectable for PR 5-95 at 1-2 M protein using a sensitive chromogenic substrate (29) and no fragments resulting from autoproteolysis were detectable by SDS-PAGE after acquiring NMR spectra at 20 M concentration, very minor signals observed in the PR 5-95 spectrum corresponding to the narrow region of 8 -8.5 ppm for 1 H and 108 -128 ppm for 15 N most likely represent the unfolded fraction of the protein (Ͻ5%). Although an increase in the K d of some mutants (e.g. PR R87K , PR T26A , Table 1) is offset by enhancing the dimer interface contacts through the addition of an inhibitor, the K d for PR  is expected to be the same or higher than PR 1-95 , which does not significantly dimerize at concentrations up to 1 mM even in the presence of the potent inhibitor DMP323 (21,30).
In contrast to PR 5-99 and PR 5-95 , the 1 H-15 N HSQC spectrum recorded using PR 10 -90 (Fig. 4C) exhibited weak signals that are characteristic of an unstructured protein. The loss in signal intensities suggests that the majority of the protein undergoes aggregation even at a low protein concentration of 20 M. Analysis of two other constructs, PR 1-90 and PR  , showed significant portions of both proteins existing both in aggregated and in folded monomer forms. These results indicate that the region spanning 5-95 of PR is sufficient to maintain a stable fold in 50 mM sodium acetate buffer, pH 5, and that further deletion of residues spanning the regions 5-9 and/or 91-95 significantly reduces the stability of the monomer fold. The observation that nearly all of the PR 5-95 exhibits a monomer fold at these low concentrations, unlike at higher concentrations (Ͼ0.2 mM), similar to the mutants PR 1-95 or PR T26A , suggests that the aggregation of PR 5-95 at a higher concentra-tion is most likely caused by the instability of the fold rather than by the misfolded protein that induces aggregation. Apparently, although the regions corresponding to residues 1-9 and 91-99 of the monomer are unstructured, they contribute to maintaining the monomer fold at a higher concentration, supported by the observation that a major fraction of the T26A monomer (PR T26A ) and other monomers exhibit a monomer fold at ϳ0.2 mM concentration, whereas at the same concentration, a majority of the PR 5-95 aggregates.

Comparison of Urea-induced Unfolding of Protease Monomers and Dimers-Our recent mutational and NMR studies
showing a stable fold for the protease monomer are consistent with a dimer dissociation model in which dimer dissociation and PR unfolding can occur independently. PR dimer contains 4 tryptophanyl residues, Trp-6 and 6Ј near the dimer interface and Trp-42 and 42Ј located externally at the base of the flap region (Fig. 2). Urea-induced unfolding as monitored by associated protein fluorescence change was used to determine the stability of the monomer fold. Folded monomers were prepared at a final concentration of ϳ10 M by the quench protocol of protein folding using the same protein stock solutions, maintained in 25-50 mM formic acid, that were used to make the NMR samples.
The addition of urea to a final concentration of 5 M to the folded monomer, which lacks a defined structure of the terminal residues 1-9 and 91-99 (21), causes a 33-37% quenching of the fluorescence emission as compared with PR or PR D25N dimeric proteases, which show approximately a 50% decrease (13,15). Plots of the intrinsic fluorescence monitored as a function of increasing urea concentration are shown in Fig. 5. The observed midpoints (U 50 ) for the transition from the folded to the unfolded state of the monomers PR 5-95 , SFNF PR D25N , PR T26A , PR D25N/T26A , and the dimer PR D25N are 1.98 Ϯ 0.14, 1.52 Ϯ 0.07, 2.51 Ϯ 0.22, 1.75 Ϯ 0.17 and 1.88 Ϯ 0.07 M, respectively. To determine whether the urea-mediated decrease in fluorescence truly reflects the unfolding of the protease, 1 H-15 N HSQC spectra were also acquired for all of the above proteins using uniformly 15 N-labeled protein at ϳ20 M concentration at increasing urea concentrations in 50 mM sodium acetate buffer, pH 5 (Fig. 6). Spectra shown in Fig. 6, A-C, clearly indicate that the peaks assigned to unfolded protein (marked by f u ) increase with increasing urea concentration, whereas the signals corresponding to the folded monomer decrease, consistent with a simple two-state transition. All of the spectra acquired indicated transitions from a folded to an unfolded state similar to those derived from fluorescence changes, and three representative spectra recorded in 2 M urea for PR D25N , PR T26A , and PR 5-95 are shown in Fig. 6, D-F. The plot of the decrease in the intrinsic fluorescence of the terminal deletion mutant, PR 5-95 , as a function of increasing urea concentration is similar to that of the protease dimer PR D25N (Fig. 5A). Although the absence of structure of the terminal residues 1-9 and 91-99 as well as the existence of an exposed active site of the monomer most likely contribute to the lower fluorescence quenching observed upon unfolding of the monomer as compared with the dimer, the denaturation profiles of the monomer and dimer are similar. Earlier studies (11) have shown that denaturation curves obtained by monitoring enzymatic activity as a function of urea concentration superimpose well onto those obtained by monitoring protein fluorescence. The similarity in the denaturation profiles of the dimer PR D25N and monomer PR  suggests that the region spanning 5-95 is sufficient to maintain a native-like stable fold and that the core unfolding of the monomer is not drastically influenced by the terminal residues.
Comparison of the profiles also indicate that the midpoint for the transition from a folded to an unfolded state of the PR T26A monomer (Fig. 5B, U 50 ϭ 2.51 M) is distinctly higher than all other mutants examined (U 50 range ϭ 1.5-2 M). This increased stability of PR T26A mutant is also clearly evident when comparing the 1 H-15 N HSQC spectra at 2 M urea. Signals corresponding to a random-coil region observed at 1 H 8 -8.5 ppm are more significant in the spectra of PR D25N and PR 5-95 (Fig. 6, D and F), whereas folded monomer signals are still strongly observed in the spectrum for PR T26A (Fig. 6E).
In contrast to PR T26A , SFNF PR D25N exhibits a slightly lower stability against urea as compared with PR 5-95 and PR D25N . If the N-terminal residues (SFNF plus 1-9 of PR) of SFNF PR D25N were entirely flexibly similar to that of PR 1-95 , the urea denaturation profile would be similar to that of PR  and PR D25N . This difference in the urea-induced unfolding profile suggests that the SFNF may influence the monomer stability. The denaturation curves exhibit different shapes, with SFNF PR D25N showing the sharpest transition and PR T26A and PR D25N/T26A showing very broad transitions, possibly indicative of the presence of multiple intermediate species. Additional experiments will be required to characterize these processes in more detail.
Urea-induced Unfolding of PR D25N Dimer-Although fluorescence measurements permitted the urea-induced unfolding profiles of the protease monomers to be compared with that of PR D25N dimer, further NMR measurements were required to determine the monomer and dimer populations throughout the transition from the folded to the unfolded state. These populations were obtained from 1 H-15 N HSQC spectra of PR D25N recorded at various urea concentrations.
In earlier studies, we showed that the D25N mutation increases the K d by ϳ2 orders of magnitude as compared with PR (28) ( Table 1). At ϳ20 M monomer concentration, achieved using the same quench protocol of protein folding as used for the monomers above, the majority of PR D25N is dimeric with ϳ10% of the signal intensity corresponding to the folded monomer (28). The average intensities of four signals of residues Gly-16, Gly-52, Ala-67, and Gly-68 in both PR D25N dimer and PR D25N monomer are plotted in Fig. 7 as a function of increasing urea concentration. The initial ratio of the average dimer signal to the average monomer signal is given by the volume (or area) ratio of the peaks (Fig. 7).
The midpoint of the denaturation curve for PR D25N dimer as measured by NMR (Fig. 7, upper curve) is ϳ1.8 M urea, in excellent agreement with the fluorescence results (Fig. 5A). The denaturation curves for the monomer and dimer (Fig. 7) are very similar in shape, suggesting that, for this construct, the dissociation of the dimer is not significantly more sensitive to urea than the denaturation of the protein fold. Thus, no increase in the relative amount of folded monomer (lower curve) at the low urea concentration range is observed, as predicted if the dimer were to dissociate to a stably folded monomer at these urea concentrations. The decrease in signal observed for the dimer must correspond to unfolding (with or without concomitant dissociation) rather than to conversion to folded monomer since this species does not increase at low urea concentrations. The results are consistent either with unfolding and dissociation having approximately equal sensitivities to urea or with unfolding of the monomer being more urea-sensitive than dissociation.
Conformational Characteristics of the Monomers Accessed from Chemical Shifts-To relate the structural features to differences observed in the denaturation profile among the monomer proteases, we assigned and compared the backbone chemical shifts of SFNF PR D25N , PR D25N/T26A , and PR T26A monomers. For these experiments, samples at a concentration of ϳ0.3 mM were prepared using the dialysis protocol. The results, as expected, showed changes in chemical shifts around the mutation sites. Comparison of C ␣ and 15 N chemical shifts of PR D25N/T26A with that of PR T26A indicates significant differences for amides around residue 25 (Fig. 8A). In addition to these changes, residues 10 -12 and 84 -95 also exhibit small differences in C ␣ and 15 N chemical shifts between PR D25N/T26A and PR T26A . The comparison of PR D25N/T26A and SFNF PR D25N chemical shifts also indicates small changes in shifts at residues 84 -95. The differences in shifts in this region could be either due to the effect of small differences in pH or due to conformational change. Overall, the difference in chemical shifts between PR D25N/T26A and PR T26A is relatively small. The higher stability against urea of PR T26A as compared with PR D25N/T26A cannot be clearly explained by structural differences as manifested by differences in backbone chemical shifts. Thus, the significantly higher stability of PR T26A may be related to alteration in side chain interactions rather than in backbone changes. Based on the proximity of some hydrophobic residues, e.g. Leu-24 and Leu-90, to T26A, we postulate that the methyl group of the substituted Ala-26 enhances hydrophobic contacts. The altered packing around Leu-24 may in turn affect the side chain orientation of residue 11, which is consistent with the observed backbone chemical shift change of this residue. The increase in stability attained through the T26A mutation is offset by the D25N mutation.
In earlier studies (11,21), we had shown that even two to four residues flanking the N terminus of the protease impairs dimer formation and that the hydrolytic cleavage at the N terminus of the protease is crucial for formation of a native-like stable dimer with catalytic activity and a K d in the low nM range. However, the molecular mechanism by which the flanking region sequences influence the K d is not understood. For this reason, the C ␣ and 15 N chemical shifts of SFNF PR D25N construct were compared with the control construct PR D25N/T26A . In addition to the differences in chemical shifts around the active site, the chemical shifts of the N-terminal protease residues 3-6 of SFNF PR D25N monomer significantly differ from those of PR D25N/T26A monomer (Fig. 8B). As described in our earlier reports (21), N-terminal residues (1-10) of the monomer are disordered in solution because of the absence of the terminal interface ␤-sheet. Thus, the difference in the shifts in the N-terminal region among these two monomers suggests that SFNF (C-terminal residues of p6 pol ), while in fusion with the protease, may interact with the N-terminal region of the folded monomer. SFNF PR D25N , which exhibits an increase in K d by Ͼ3-orders of magnitude over PR D25N , also exhibits a lower transition mid-point for urea-induced unfolding than other monomers. One interpretation that accounts for both the change in the chemical shifts and increased sensitivity to urea concentration is that the SFNF residues interact weakly with the N-terminal residues 3-6 of PR. Future NMR structural studies of SFNF PR D25N may allow a more complete   black). Differences in chemical shifts between PR T26A and PR D25N/T26A are observed in the active site region and in the helical region. Differences in the chemical shifts between SFNF PR D25N and PR D25N/T26A are observed for the N-terminal protease residues 3-6 in addition to the active site region. The secondary structure as determined for PR 1-95 is indicated on the top. The arrows and cylinder indicate ␤-sheet and ␣-helical structures, respectively. A direct comparison of the SFNF PR D25N and PR D25N constructs is not feasible because under the conditions of the NMR experiments, SFNF PR D25N is a monomer, whereas PR D25N is mainly a dimer. For this reason, a T26A mutation was introduced in PR D25N to increase the K d to enable comparison of the SFNF PR D25N monomer with the mature protease monomer having the same D25N mutation. PR T26A monomer serves as a control for PR D25N/T26A monomer.