Importance of the N Terminus of Rous Sarcoma Virus Protease for Structure and Enzymatic Function

Abstract

All retrovirus proteases (PRs) are homodimers, and dimerization is essential for enzymatic function. The dimer is held together largely by a short four-stranded antiparallel beta sheet composed of the four or five N-terminal amino acid residues and a similar stretch of residues from the C terminus. We have found that the enzymatic and structural properties of Rous sarcoma virus (RSV) PR are exquisitely sensitive to mutations at the N terminus. Deletion of one or three residues, addition of one residue, or substitution of alanine for the N-terminal leucine reduced enzymatic activity on peptide and protein substrates 100- to 1,000-fold. The purified mutant proteins remained monomeric up to a concentration of about 2 mg/ml, as determined by dynamic light scattering. At higher concentrations, dimerization was observed, but the dimer lacked or was deficient in enzymatic activity and thus was inferred to be structurally distinct from a wild-type dimer. The mutant protein lacking three N-terminal residues (ΔLAM), a form of PR occurring naturally in virions, was examined by nuclear magnetic resonance spectroscopy and found to be folded at concentrations where it was monomeric. This result stands in contrast to the report that a similarly engineered monomeric PR of human immunodeficiency virus type 1 is unstructured. Heteronuclear single quantum coherence spectra of the mutant at concentrations where either monomers or dimers prevail were nearly identical. However, these spectra differed from that of the dimeric wild-type RSV PR. These results imply that the chemical environment of many of the amide protons differed and thus that the three-dimensional structure of the ΔLAM PR mutant is different from that of the wild-type PR. The structure of this mutant protein may serve as a model for the structure of the PR domain of the Gag polyprotein and may thus give clues to the initiation of proteolytic maturation in retroviruses.

All retroviruses encode a protease (PR) that functions to cleave the structural polyprotein Gag and the enzymatic polyprotein Pol into their constituent mature proteins found in infectious virions. PR is translated as a domain of the Gag, Gag-Pro, or Gag-Pro-Pol polyprotein, depending on the retrovirus genus (reviewed in references 42 and 43). Cleavage of the viral polyproteins is initiated late in assembly, during or after budding of the nascent virion from the plasma membrane, and leads to the morphological maturation that gives rise to infectious virus. The first cleavage in maturation is believed to be an autocatalytic event in which the PR domain excises itself from the polyprotein (6, 10, 12, 24, 35, 48). What triggers this process remains unknown.

Retrovirus PRs are aspartate PRs, with the catalytic site positioned between the two identical subunits that make up the enzymatically active PR dimer. Thus, dimerization is a requisite for enzyme function. It has been suggested elsewhere that dimerization of the PR domains of two neighboring molecules of the polyprotein is the initial and rate-limiting step of a proteolytic cascade that results in maturation (26; reviewed in reference 43). That dimerization cannot be triggered simply by mass action, by the PR domains becoming concentrated in the nascent virion, is best illustrated by the type B and type D retroviruses (prototypes, mouse mammary tumor virus and Mason-Pfizer monkey virus, respectively). In these viruses, the immature viral core becomes fully assembled in the cytoplasm, and only after transport to the plasma membrane and budding is proteolysis initiated.

High-resolution structures of numerous retrovirus PRs are known, and in all cases a key feature of the dimer is a short, four-stranded antiparallel beta sheet formed by the first and last few residues at the N and C termini (reviewed in reference 45). A large fraction of the dimer interface surface is provided by this structural feature, and preventing its formation blocks enzymatic activity, as shown previously for Rous sarcoma virus (RSV) (18) and human immunodeficiency virus type 1 (HIV-1) (3, 36, 46). By contrast with the highly detailed knowledge of the structural and enzymatic features of mature dimeric PRs, little is known about PR domains as part of the polyproteins in which they are initially embedded. In some studies, PR precursors of this type have been reported to be inactive or poorly active in vitro, but these findings depend on the virus and on the details of the conditions used (10, 24, 29, 37, 47, 48). Assessing the enzymatic activity of a PR domain embedded in a larger segment of polypeptide can be difficult because the PR domain may excise itself during the assay, thereby generating fully active dimeric PR. Even a small amount of mature PR will initiate a feedback loop in which more PR is generated by cleavage in trans, leading ultimately to complete processing of the precursor. Structural studies of retrovirus PRs and larger PR-containing polypeptides are complicated by their limited solubility and by the invariable need to renature these proteins from inclusion bodies after expression in Escherichia coli.

The avian sarcoma and leukemia viruses (ASLV; prototype, RSV) have been an important model system to study PR structure and function. Indeed, the first crystal structure of a retrovirus PR was determined for avian myeloblastosis virus, a strain of ASLV (16). Next to HIV-1 PR, ASLV PR probably has been the best-studied retrovirus PR from an enzymatic point of view (1, 7, 13, 20, 21, 34, 38). Among retroviruses, the ASLV are unusual in that PR is translated as the C-terminal domain of Gag and thus is present in virions in amounts equimolar with those of the other Gag proteins. We showed previously that purified RSV NC-PR protein, consisting of the joined NC and PR domains as they are found in RSV Gag, has very little enzymatic activity and at low concentrations behaves as a monomer in solution (37). These observations suggested that amino acid sequences upstream of PR interfere with dimerization and thus that the RSV Gag protein is analogous to a zymogen. Consistent with this notion, we found in transfection experiments that a cleavage site mutation at the NC-PR junction prevented proper excision of PR from Gag and completely blocked the appearance of mature Gag and also Pol proteins in virions (6, 35, 39). However, in these mutant virions about 25% of the Gag molecules were cleaved near the NC-PR junction, generating an inactive PR species truncated by three amino acid residues at its N terminus. The same truncated species also is found in wild-type virions (32). This result implies that the only activity of the PR domain as part of the polyprotein is to cleave itself out and that all other cleavages in Gag and Pol are mediated by the mature PR.

The PR domain of a single Gag molecule must exist in a monomeric state, at least transiently and perhaps until the time when proteolysis is initiated late in assembly. Thus, knowledge of the three-dimensional structure of a monomeric PR might provide insight into the process of dimerization. Based on initial results with the NC-PR protein in vitro and the ΔLAM protein in mutant virions (35, 37), we decided to analyze the effects of N-terminal mutations on RSV PR activity, dimerization, and three-dimensional structure. The results presented here show that even minor changes at the N terminus block enzymatic activity, and they do so by preventing proper dimerization. From nuclear magnetic resonance (NMR) spectroscopy of the mutant ΔLAM, we infer that this monomeric form of PR is folded, unlike a similar mutant HIV-1 protein (26), and that its structure differs significantly from that of wild-type PR analyzed in parallel.

MATERIALS AND METHODS

Purification of recombinant PRs.

Eight PR mutations were constructed by PCR mutagenesis and subcloned into the pET11 expression plasmid (Novagen) by standard techniques, and the sequence was confirmed. Plasmids were propagated in E. coli DH5α and transformed into BL21 cells for protein expression. E. coli BL21 cells were grown in 2YT medium containing ampicillin and chloramphenicol. For isotope labeling with ¹⁵N and ¹³C, cells were grown in MT-9 CN medium (Martek) and with ¹⁵N alone in M9 medium containing [¹⁵N]ammonium chloride (Isotec). After induction with 1 mM isopropyl thiogalactoside (IPTG) at A₆₀₀ = 0.4, cultures were incubated for another 4 h in rich medium or overnight in minimal medium. Inclusion bodies were collected from sonicated cell lysates, washed two times in 20 mM Tris-HCl (pH 7.5)–5 mM EDTA–2 M urea–2% Triton X-100–10 mM dithiothreitol (DTT)–1 mM phenylmethylsulfonyl fluoride, and then washed once in the same solution without urea and Triton X-100. The washed inclusion bodies were dissolved in 20 mM Tris-HCl (pH 7.5)–7 M urea–10% glycerol–5 mM EDTA by stirring at room temperature for 30 min and diluted to 1 M urea with 20 mM Tris-HCl (pH 8)–20 mM NaCl–5 mM EDTA–10% glycerol, and insoluble material was removed by centrifugation. The supernatant was passed over a DEAE column in the same buffer, and the flowthrough containing the PR was collected and dialyzed into 20 mM Na phosphate (pH 6)–100 mM NaCl–1% glycerol–0.4 M urea–10 mM DTT. After removal of any precipitate, the supernatant was concentrated with an Ultrafree centrifugal filter unit (molecular mass cutoff, 3.5 kDa; Millipore), and the protein concentration and purity were assessed by A₂₈₀ and A₂₆₀, using the molar extinction coefficient for PR, and by sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis (PAGE) and Coomassie blue staining. Molar concentrations of PR as used in this paper refer to the total concentration of the PR polypeptide as if it were monomeric.

The viral PR (vPR) serving as a control was purified from avian myeloblastosis virus (Life Sciences, Inc., St. Petersburg, Fla.) by chloroform-methanol extraction (34). The E. coli-derived wild-type PR, dPR, was obtained by self-cleavage of His-ePR, a protein carrying the sequence MAHHHHHHAPAVS N-terminal to the PR coding region. PAVS is the sequence of the C terminus of NC, i.e., it occurs naturally at this position in Gag. His-ePR was purified either as done for the other PR mutants or by metal chelate chromatography on Ni-nitrilotriacetic acid (Qiagen). In either case, the final step was dialysis into 20 mM NaPO₄ (pH 7.0)–100 mM NaCl–1% glycerol–0.4 M urea. The protein was incubated in this condition at 37°C for 5 h to allow self-cleavage. We found that self-cleavage was more efficient at this pH and low salt than at more acid pH values and higher salt, the conditions normally used for retrovirus PRs. The resulting dPR was dialyzed into the same buffer but with a pH of 6.0, concentrated, and then assayed for activity on protein and peptide substrates as described below.

Enzyme assays.

Activity of wild-type and recombinant PRs was measured on protein and peptide substrates. Standard reaction mixtures contained 100 mM 3-(N-morpholino)propanesulfonic acid (MOPS; pH 6), 0.8 M NaCl, 5 mM EDTA, and 6% glycerol, to which was added 5 μg of the substrate protein ΔMBDΔPR, in a final reaction volume of 30 μl. The substrate is the RSV Gag protein missing the 83-amino-acid membrane binding region of MA at the N terminus and missing the entire 124 amino acid residues comprising PR at the C terminus. From 1 to 10 μg (approximately 2 to 20 μM PR) of each PR was added to the reaction mixture, and incubation was carried out at 39°C for 0.25 to 60 min for vPR or dPR and 10 min to 10 h for the mutants. The reaction products were separated by SDS-PAGE and visualized by Coomassie blue staining. Peptide assays were based on cleavage of the decapeptide PFQAY/PLREA, which corresponds to the cleavage site between the reverse transcriptase and integrase domains in the RSV Pol protein. The same conditions as for the protein substrate assays were used, except that glycerol was omitted. The concentration of peptide was 0.4 mM, and the concentration of PR was 0.07 to 7 μM for vPR and dPR and 14 μM for the mutant PRs. The reactions were stopped by addition of trifluoroacetic acid to 10% and stored at −20°C until they were analyzed by reverse-phase high-pressure liquid chromatography (HPLC) on a C₁₈ Vidac column with absorbance at 210 nm recorded. Reaction products were eluted at a flow rate of 1 ml/min with a 0 to 60% acetonitrile gradient in 0.1% trifluoroacetic acid.

DLS and cross-linking analysis.

Mutant and wild-type PRs were tested for dimerization status by dynamic light scattering (DLS) on a DynaPro MSTC molecular sizing instrument (Protein Solutions, Inc.). Samples at concentrations ranging from 0.5 to 5 mg/ml were passed through a 20-nm-pore-size Whatman Anotop Plus filter disk, and then 15 μl was loaded into the DLS cuvette and illuminated by a laser with a wavelength of 780 to 830 nm. The time scale of the scattered light intensity fluctuations at 25°C was analyzed using the autocorrelation function of the company's software. From the translational diffusion coefficient, the hydrodynamic radius, R_h, can be derived from the Stokes-Einstein relationship. A standard curve of globular proteins of known mass allows estimation of the molecular weight from the R_h of the sample protein with an accuracy of ±10% for globular molecules.

Cross-linking was carried out on vPR or ΔLAM at 1 mg/ml. Samples were incubated with the homobifunctional cross-linking reagent dimethylsuberimidate (DMS) at pH 8.0 and 2 mM or with 1,6-bismaleimidohexane (BMH) at pH 6.0 and 2 or 10 mM (both reagents from Pierce Chemical Company). DMS has a spacer arm length of 11 Å and is reactive toward primary amino groups, whereas BMH has a spacer arm length of 16 Å and reacts with sulfhydryl groups. DMS was allowed to react at room temperature for 2 h in 50 mM triethanolamine buffer (pH 8.1) and was quenched with 0.1 M glycine for 20 min before addition of sample buffer for SDS-PAGE. BMH was allowed to react at room temperature for 30 min in 20 mM NaPO₄ buffer (pH 6) and was quenched by addition of 50 mM DTT.

NMR.

ΔLAM and dPR protein samples for NMR analysis were prepared as described above with [¹⁵N]ammonium chloride (Isotec) as the sole nitrogen source. Sterile D₂O (Martek Biosciences) was added to the sample to a final concentration of 10% (vol/vol), and the samples were placed in microcell NMR tubes (Shigemi, Inc.). NMR experiments were performed at 298 K using a Varian Inova 600-MHz spectrometer equipped with a triple resonance probe with a z-axis pulsed-field gradient. The ¹⁵N-¹H heteronuclear single quantum coherence (HSQC) pulse sequence from Lewis Kay's library (University of Toronto) was employed, which includes water-flip-back pulses to avoid saturation transfer (14) as well as sensitivity enhancement in conjunction with gradient coherence selection (19, 27).

RESULTS

Activity of N-terminally mutated versions of PR.

In earlier work, we showed that an RSV PR truncated by three amino acid residues at its N terminus lacked detectable activity, as evidenced by the complete absence of mature Gag and Pol proteins in virions (6, 35, 39). Also, it was reported elsewhere that the specific activity of wild-type PR expressed in E. coli was only approximately 10% of that of the wild-type PR when isolated from virions (4, 34, 37), the attenuation of activity being attributed to the lack of removal of the initiating methionine residue in E. coli. Consistent with this interpretation, in RSV PR-Pol fusion proteins created with an initiating Met residue in front of the PR domain, the Met residue was not removed and the proteins were enzymatically inactive when highly expressed in insect cells (39). These several observations suggested that the proper N terminus is critical for PR function. In order to more carefully characterize the role of the N terminus, we constructed several additional mutant PRs (Fig. 1). ΔLAM lacks the three N-terminal amino acid residues and thus is equivalent to the truncated PR formed in virions when the cleavage site at the NC-PR junction is mutated (35). The same truncated species also is found in wild-type virions, representing about 7% of the total PR (32). ΔL deletes only Leu1, the first residue of PR. L1A, L1V, and L1I change the N-terminal leucine residue as indicated. As controls for these mutants, three different forms of wild-type PR were used. PR purified from avian myeloblastosis virus virions, designated vPR, was the standard in all assays. The PR polypeptide expressed in E. coli with an initiating methionine residue proximal to Leu1 is designated rPR. This form of PR is poorly active, by inference because of failure of methionine removal in E. coli (see below). We engineered a version of PR, called His-ePR, that processes itself slowly in vitro to liberate wild-type PR with the correct N-terminal Leu1. In this protein, the four natural residues preceding the PR sequence, PAVS, as well as a histidine tag are positioned proximal to the coding sequence. His-ePR was purified in the same manner as the other mutant PRs and incubated under conditions found to maximize the weak autocatalytic processing activity, and the resulting derived PR (dPR) was repurified. As found for other retrovirus PRs, all of the PRs studied here were located almost exclusively in inclusion bodies and thus had to be solubilized in 7 M urea before refolding. Control experiments with vPR showed that denaturation in urea followed by renaturation upon removal of urea did not compromise the specific activity (data not shown).

FIG. 1.

Schematic structure of PR mutants. The rectangle at the top depicts the RSV Gag protein and its major domains, with the amino acid sequence at the NC-PR cleavage site shown above. The horizontal bars represent the several wild-type and mutant PRs studied. The wild-type amino acid residues at the N terminus are indicated, with dots indicating identity and the absence of any symbol indicating a deletion of those residues. The N-terminal sequence of all proteins was determined experimentally.

To estimate the activity of the mutant PRs, two assays were used. The first is semiquantitative and visualizes cleavage of the substrate protein ΔMBDΔPR, a truncated RSV Gag protein missing 83 residues from the N terminus and the 124 PR residues from the C terminus (8, 17). This protein, which contains several natural cleavage sites, was purified after expression in E. coli, and its processing after incubation with PR was monitored by SDS-PAGE and staining. In typical experiments, incubation of 1 μg of vPR with 10 μg of substrate led to 50% cleavage in about 2 min and complete cleavage in 10 to 20 min (Fig. 2A). Complete cleavage was defined by a gel profile in which the mature CA polypeptide was the largest prominent species visible on the stained gel. In the same time period, 1 μg of the mutant protein ΔLAM, L1A, L1V, L1I, or ΔL failed to process the substrate protein (data not shown). However, with a 10-fold increase in PR and longer incubation times, limited cleavage was observed for L1V, with intermediate cleavage products barely evident at 12 min and substantial amounts of CA evident by 10 h (Fig. 2B). Somewhat higher levels of activity were observed for L1I (data not shown). No activity was detectable for the mutant proteins L1A (Fig. 2B) and ΔL and ΔLAM (data not shown) in similar assays. However, in a more sensitive assay in which Western blotting with anti-CA serum was used instead of staining, ΔLAM did manifest detectable activity, estimated at ∼0.1% of the activity of vPR (data not shown). For rPR, 10 μg of enzyme gave complete cleavage by 30 min (Fig. 2C). We attempted to quantitate these results by visually gauging the intensity of bands from gels like those shown in Fig. 2 and estimating at what time and enzyme concentration 50% of the substrate protein had been converted into smaller polypeptides. Based on the assumptions that enzyme activity is directly proportional to the amount of PR in the reaction and that the enzyme remains stable during the incubation period, we derived semiquantitative estimates of the residual activities of the different PRs: dPR was as active as vPR, rPR was 5 to 10% as active, and the mutants ranged from 2 to 3% for L1I to undetectable by the staining assay for L1A and ΔLAM (Table 1).

FIG. 2.

Protein cleavage assay. Wild-type or mutant PRs were incubated with the truncated Gag protein ΔMBDΔPR, and the time course of proteolytic processing was examined by SDS-PAGE and Coomassie blue staining. The reaction mixtures contained 100 mM MOPS (pH 6.0), 0.8 M NaCl, 1 mM EDTA, 6% glycerol, 5 μg of substrate, and either 1 μg of vPR or 10 μg of mutant PRs. The positions of the substrate and the major product, CA, as well as of PR are indicated on the left. The assay times are shown at the top of each panel. (A) vPR; (B) L1A (lanes 1 to 4) and L1V (lanes 5 to 8); (C) rPR.

TABLE 1.

In vitro activities of wild-type and mutant PRs

PR constant	N terminusa	% Activity on protein substrateb	% Activity on peptide substratec	kcatd (min−1)
rPR	MLAMTM	5–10	NDf	ND
L1I	MIAMTM	2–3	ND	ND
L1V	VAMTM	<1	ND	ND
L1A	AAMTM	0e	0.3	0.009
ΔL	-AMTM	<1	0.9	0.02
ΔLAM	---TM	0e	0.1	0.003
dPR	LAMTM	100	65	2.5
vPR	LAMTM	100	100	4

^aN-terminal amino acid sequence was determined for each purified protein. 

^bActivity was estimated as described in the legend to Fig. 2. 

^cActivity was estimated as described in the legend to Fig. 3. 

^dEstimated from amount of N-terminal product peak with assumptions described in the text. 

^eNo product was detectable by SDS-PAGE and Coomassie blue staining. 

^fND, not determined. 

Knowledge of the actual N-terminal sequence of the several PR species is clearly critical for interpreting their enzymatic activities. In E. coli, the second amino acid residue determines if the initiating methionine will be removed (15). Loss of Met generally occurs if the side chain is small but not if it is large. We determined the N-terminal sequence for the E. coli-derived PR proteins and found them to obey this rule. Thus, the initiating Met was removed when the second residue was Ala (L1A and ΔL), Val (L1V), or Thr (ΔLAM), but Met was retained when the second residue was Leu (rPR) or Ile (L1I) (Table 1). N-terminal sequence analysis is rarely quantitative, however, reflecting only the predominant polypeptide species. Thus, a minor species of PR comprising less than about 5% of the total protein would not have been evident in these analyses. Therefore, for the proteins that retained the Met, rPR and L1I, it is unclear to what extent the low activity observed reflects intrinsic activity of the major species bearing an N-terminal Met or alternatively reflects a low level of a fully active species with the Met removed.

As an independent and more quantitative assay for activity, we incubated the mutant PRs with a synthetic 10-amino-acid peptide. This peptide, which includes the cleavage site between the reverse transcriptase and integrase domains of the RSV Pol protein, was reported to be the most efficiently cleaved of several Gag and Pol peptides processed by PR in vitro (41). Each PR was incubated at a final concentration of 0.1 or 0.2 mg/ml with 0.4 mM peptide, and the products and substrate were separated on a C₁₈ column by reverse-phase HPLC. The peak height of the N-terminal product eluting at 17 min was used to estimate rate of cleavage, under conditions where less than 25% of the substrate had been used up. Representative portions of tracings showing incubations with vPR, with ΔLAM, or without any PR are shown in Fig. 3. While 5 μg of vPR incubated for 13 min resulted in the appearance of a robust product peak (arrow), 10 μg of ΔLAM incubated for 630 min resulted in the reproducible appearance of a barely detectable peak just above the baseline. From similar assays, we calculated approximate k_cat values for three mutant versions of PR (Table 1). The results are consonant with those of the polypeptide cleavage assays, showing that ΔL, L1A, and ΔLAM are reduced 100- to 1,000-fold in specific activity. The k_cat value obtained for vPR is in the same range as reported previously for this enzyme, 1 to 11 min⁻¹ (13, 21). We did not seek to determine if the slight difference in k_cat determined for vPR and dPR represents an intrinsic property of these proteins or is due to experimental error.

FIG. 3.

Peptide cleavage assay. Samples of mutant or wild-type PR were incubated with 0.4 mM decapeptide corresponding to the reverse transcriptase-integrase cleavage site, under the same conditions as described for Fig. 2, except for the omission of glycerol. The reaction products were separated by reverse-phase HPLC. Representative A₂₁₀ traces are shown. The elution times of the decapeptide substrate and the N-terminal cleavage product were 20 and 17 min, respectively. The concentration of PR and the assay time for each reaction are shown on top.

Dimerization status of mutant PRs.

In retrovirus PRs, the four-stranded beta sheet comprising the N-terminal four to five and the C-terminal four to five residues (except for the ultimate residue) is known to be critical for dimerization (44). For example, short peptides including the terminal residues of PR will compete at high concentrations with intermolecular beta-sheet formation, causing dissociation of the dimer and loss of activity (3, 18, 36, 46). From these results, we hypothesized that the loss or attenuation of enzyme activity in the mutant RSV PRs was due to lack of dimer formation. To address this hypothesis, we made extensive use of DLS and also carried out corroborative experiments with rate zonal sedimentation in glycerol gradients and with chemical cross-linking. Most experiments focused on the ΔLAM and L1A mutants and on vPR or dPR as controls. A limitation of DLS is that the concentration of a protein of this size must be at least ∼0.5 mg of monomer/ml. Another limitation particular to PRs is their poor solubility. We found that the mutant PRs were stably soluble to only about 3 mg/ml. vPR and dPR were less soluble than the mutant PRs, consistent with the original report that vPR could be crystallized from solutions of 1.3 mg/ml (16, 33). However, addition of 1% glycerol and 0.4 M urea, which had no effect on activity (data not shown), increased this limit to about 5 mg/ml for the mutant proteins and 2 to 3 mg/ml for wild-type PR. Most of the DLS experiments were carried out in the presence of these cosolvents, because their addition allowed the concentration of dPR to be increased to the level where NMR experiments became feasible (see below).

At pH 8.0 and 3 mg/ml, ΔLAM (Fig. 4A) appeared monomeric by DLS, with an inferred molecular mass of ∼15 kDa. As expected, vPR appeared dimeric, with an inferred mass of ∼24 kDa (data not shown). (The actual masses of these proteins are 13.3 and 27.4 kDa, respectively, and the accuracy of mass measurements by DLS is about ±10% for globular proteins.) Similar results were obtained by glycerol gradient sedimentation. These data support the hypothesis that the mutations hindered dimer formation. Since retrovirus PRs in general and RSV PR in particular are active at acid pH but show little activity at pH 8, DLS measurements also were carried out near pH 6. We were surprised to find that under these conditions the predominant form of ΔLAM was dimeric at 4 mg/ml (Fig. 4B). However, reducing the protein concentration to 0.5 mg/ml shifted the protein to a monomeric molecular mass (Fig. 4C). These results suggested that the mutant PR is in a monomer-dimer equilibrium governed by a K_d in the range of 50 μM for these solvent conditions. By definition, at a 50 μM monomer concentration there is an equal concentration of dimer, and thus the total concentration of PR would be 150 μM, corresponding to about 2 mg/ml. We found that the presence of 1% glycerol plus 0.4 M urea shifted the equilibrium slightly toward monomers, so that ΔLAM was largely monomeric at concentrations as high as 3 mg/ml but became predominantly dimeric with a further increase in concentration to 5 mg/ml near the solubility limit (Fig. 4E and F). A parallel control sample of dPR at 2 mg/ml in the same solvent was dimeric as expected (Fig. 4D). The hydrodynamic radius values determined by DynaLS software are summarized in Table 2.

FIG. 4.

Size distribution by DLS. Samples of wild-type and mutant PR were passed through a 20-nm-pore-size filter and then submitted to DLS analysis. Representative tracings compiled by the DynaLS software of the DynaPro molecular sizing instrument are shown. The vertical axis is scattering intensity, and the (logarithmic) horizontal axis is the hydrodynamic radius (in nanometers) of the PR in the sample. Monomers and dimers of PR are not resolved, but the position of the peak is determined by a weighted average of monomers and dimers present, and the width of the main peak is indicative of the homogeneity of the sample.

TABLE 2.

Monomer or dimer status of wild-type and mutant PRs by DLS analysis

Lane	PR construct	Concn (mg/ml)	pHa	Peak value		Mean value		Statusc
				Rh (nm)b	MWg estimate	Rh (nm)b	MWg estimate
1	dPR	2	6.0	2.38	23	2.51	27	Dimer
2	PR-PRd	0.5	5.5	2.42	25	2.57	28	Dimer
3	ΔLAM	3.5	8.0	1.95	15	2.11	18	Monomer
4	ΔLAM	3.5	6.4	2.42	25	2.52	27	Dimer
5	ΔLAM	2	6.4	2.41	25	2.40	25	Dimer
6	ΔLAM	0.5	6.4	1.94	14	1.89	13	Monomer
7	ΔLAMe	3	6.0	1.90	14	2.01	15	Monomer
8	ΔLAMf	4.3	6.0	2.38	24	2.36	24	Mostly dimer
9	L1A	3.5	6.0	1.93	14	2.14	18	Monomer

^aBuffers were Tris-HCl (pH 8.0)–100 mM NaCl, sodium phosphate (pH 6.4)–100 mM NaCl, and sodium phosphate (pH 6.0)–100 mM NaCl–0.4 M urea–1% glycerol. 

^bHydrodynamic radius, R_h, as calculated by manufacturer's software. 

^cMonomer or dimer status derived from the estimated molecular weight. This is calculated by DynaLS software from the mean hydrodynamic radius measured, using a standard curve of molecular weight versus R_h of globular molecules. 

^dCovalently linked, single-chain PR dimer (5) expressed in E. coli. 

^eAverage of four experiments. 

^fAverage of two experiments. 

^gMW, molecular mass in kilodaltons. 

The ΔLAM dimers observed at pH 6.0 and high protein concentrations must be different in specific activity and therefore in structure from wild-type PR dimers. The logic for this conclusion is based on consideration of any monomer-dimer equilibrium and on the estimated K_d for this reaction for ΔLAM. In the HPLC assay, the highest concentration of ΔLAM tested for enzymatic activity was 0.2 mg/ml (equivalent to 15 μM monomer), and this amount of mutant PR showed about a 1,000-fold reduction in activity compared with vPR. Assuming a K_d of 50 μM, at 0.2 mg of total protein/ml about 30% of the protein would be in dimers, vastly more than the 0.1% of activity observed. Even if the K_d were 500 μM, about 5% of the protein would be dimeric at 0.2 mg/ml, far too high to explain the feeble activity observed at this concentration. Therefore, we conclude that the mutant dimers observed by DLS are severely compromised in their function and thus that they differ somehow in structure from wild-type dimers. Possibly, the mutant dimers observed by DLS, like monomers, lack enzymatic activity altogether. In that case, the tiny amount of enzymatic activity manifested by ΔLAM would be attributed to a very minor fraction of mutant protein that successfully forms a “wild-type” dimer, which would be present at much too low a level to be detectable by any physical measurements.

We used chemical cross-linking to provide supporting evidence for the existence of mutant PR monomers at alkaline pH and for structural differences between mutant and wild-type dimers at acid pH. Incubation of 1 mg of vPR/ml with 2 mM DMS, a bifunctional amino-reactive reagent, at pH 8.1 led to the appearance of a DMS-dependent dimer band by SDS-PAGE (Fig. 5B, lanes 1 and 2), as reported many years previously in cross-linking experiments with intact virions (31). Under identical conditions, no specific cross-linking was observed for ΔLAM (lanes 3 to 5) or L1A, L1V, or L1I (data not shown), consistent with the conclusion from DLS and sedimentation analysis that at this pH these proteins are monomeric at the concentration tested. The chemical reaction between imidate and amino groups does not proceed well at acid pH, and hence DMS cannot be used effectively at pH 6.0. As an alternative, we chose a bifunctional reagent that reacts with sulfhydryl groups, BMH, to test for cross-linking at a pH where retrovirus PRs are active. The RSV PR polypeptide has a single sulfhydryl group, Cys113. In the crystal structure of the wild-type dimer, the SH groups of Cys113 and Cys113′ are 26 Å apart, a distance that is too great to allow cross-linking by BMH, which spans only 16 Å. Consistent with this prediction, incubation of 1 mg of vPR/ml with BMH did not lead to the appearance of any BMH-dependent bands by SDS-PAGE (Fig. 5A, lanes 5 to 8). By contrast, at the same protein concentration incubation of ΔLAM yielded a distinct species migrating at the position of a dimer (lanes 1 to 4). As a specificity control, no BMH-dependent cross-linking was observed for cytochrome c, which has two free SH groups (lanes 9 to 12). The fact that no dimers were observed for vPR, even though the SH group is on the surface, further supports the conclusion that the cross-linked dimer species in ΔLAM reflects real protein-protein interactions. In summary, these results support the conclusion that the dimeric ΔLAM protein found at high protein concentrations differs from the wild-type dimer. We interpret the results to mean that the two Cys residues are within 16 Å of each other in the mutant dimer. However, the data do not exclude the possibility that the mutant dimer in question represents a population of different dimeric species.

FIG. 5.

Cross-linking analysis. (A) BMH cross-linking. Samples of 30 μl at 1.0 mg of protein/ml were treated for 30 min with BMH in Na-phosphate buffer (pH 6.0)–100 mM NaCl and then were quenched with 50 mM DTT. Lanes 1 to 4, PRΔLAM; lanes 5 to 8, vPR; lanes 9 to 11, cytochrome c. (B) DMS cross-linking. Samples of 20 μl at 1.0 mg of protein/ml were treated for 2 h with DMS in triethanolamine buffer (pH 8.1)–100 mM NaCl and then were quenched with 100 mM glycine. Lanes 1 and 2, vPR; lanes 3 to 5, PRΔLAM. The positions of PR and molecular size markers of 30 and 21 kDa on Coomassie blue-stained SDS gels are indicated on the side, and the concentration of cross-linker in each reaction is shown at the bottom. The arrows indicate the positions of the dimers observed.

Comparison of ΔLAM and dPR by NMR.

Our long-term goal is to determine a high-resolution solution structure of a monomeric form of RSV PR, in order to gain insight into the structure of the PR domain of Gag before dimerization occurs. As a first step toward this goal, we obtained HSQC spectra for ΔLAM and dPR proteins that had been labeled with ¹⁵N. The ¹H-¹⁵N HSQC experiment provides a two-dimensional spectrum in which the cross peak positions reflect the resonance frequencies, or chemical shifts, of both the protons and ¹⁵N nuclei that share a covalent bond. Although resonance peaks in an HSQC can arise from NH and NH₂ correlations found in some amino acid side chains, the spectrum is dominated by backbone amide correlations. Because this experiment results in a single unique cross peak for most amino acid residues, reflecting their chemical environment, the HSQC spectrum is often viewed as a “fingerprint” of the protein. Chemical shift perturbation mapping, also known as structure-activity relationship by NMR, is a facile experiment wherein HSQC spectra are compared upon alteration of the system. Movement of peaks between HSQC spectra indicates that the residue assigned to that peak has experienced a change in chemical environment, often indicative of a structural rearrangement or altered interaction, if there has been no change in the solvent system.

Peaks within the HSQC spectrum of the ΔLAM protein were found to be widely dispersed in both dimensions (Fig. 6A), indicative of a stably folded protein. Resonance assignments for the backbone at a concentration of 3 mg/ml have been reported elsewhere (33a). At concentrations necessary for solution structure determination (3 to 5 mg/ml), DLS measurements indicated the presence of both monomeric and dimeric PRs. Since a single set of peaks is apparent in the HSQC spectrum, the monomer-dimer equilibrium is in fast exchange with respect to the NMR time scale. Hence, the observed resonances are a population-weighted ensemble average. Increasing the concentration of PR from 1 to 5 mg/ml adjusted the equilibrium from predominantly monomer to predominantly dimer according to DLS data. This observation was corroborated by ¹⁵N relaxation measurements at 3 and 5 mg/ml, which yielded a statistically significant increase of average T1/T1ρ ratios with an increase in concentration (data not shown) indicative of increased molecular correlation, or “tumbling” time. Increased average size is manifested by subtle changes in the HSQC spectrum (Fig. 6B). No peaks move more than 0.03 ppm in the ¹H dimension or 0.3 ppm in the ¹⁵N dimension, but peak movement can be detected for several peaks as shown in insets of Fig. 6B. The residues that do change do not map to the wild-type PR interface but rather are widely dispersed throughout the protein.

FIG. 6.

¹H-¹⁵N HSQC spectra of ΔLAM PR at 3 mg/ml with backbone assignments labeled (A), ΔLAM PR at 1 and 5 mg/ml (black and red contours, respectively) (B), and ΔLAM PR at 5 mg/ml and dPR at 2 mg/ml (red and black contours, respectively) (C). Insets in panel B show magnified cross peaks for residues that experience the largest peak shifts upon increase in concentration. Inset boxes have dimensions of 0.2 ppm of ¹H by 0.5 ppm of ¹⁵N. In panel C, solid arrows denote residues that map to alpha-helix residues 112 to 118. Hollow arrows denote residues in the mobile flap region (residues 58 to 70), and the asterisk indicates the C terminus.

Comparison of the HSQC spectrum of dPR with that of the predominantly dimeric ΔLAM at 5 mg/ml (Fig. 6C) revealed discernible changes for approximately a quarter of the residues. Since the dPR spectrum has not been assigned, definitive assessment of whether the chemical shift of a particular residue is unchanged is not possible. Many residues appear virtually unchanged in location and intensity, comprised primarily of the sharp, solvent-exposed residues of the mobile flap region (residues 58 to 70) and the C terminus (L124). Residues 112 to 118, which comprise most of the alpha helix found in the wild-type PR dimer, do not appear to shift significantly and are all of similar intensities. Most of the loops, with the exception of the a-b loop (residues 6 to 10) and the b-c loop (residues 21 to 27) (Fig. 7), also appear to be unchanged. Although many individual residues within the beta strands do not appear to move substantially, large stretches of beta-strand secondary structure are perturbed between the two HSQCs. This does not preclude the possibility of similar beta strands in both systems, since backbone chemical shifts of extended secondary structure are exquisitely sensitive to relatively small changes of torsion angles and hydrogen bond lengths. In summary, there are real differences between the structures of dimeric ΔLAM and dPR, but the two species may share many common features.

FIG. 7.

Ribbon diagram of RSV PR derived from X-ray crystal structure (16). Dashed lines denote mobile regions where no electron density was found. Amino acid residues of ΔLAM that shifted upon an increase in concentration from 1 to 5 mg/ml (Fig. 6B) are projected onto the diagram. The ribbon was created using Molscript (22).

DISCUSSION

Proteolytic processing of Gag and Pol proteins occurs late in retrovirus maturation, but the mechanism by which processing is retarded until that time is unknown. It has been assumed that a proteolytic cascade is initiated by the dimerization of some of the PR domains of the polyprotein, followed by the autocatalytic excision of PR from the polyprotein in which it is embedded. The notion that dimerization is the rate-limiting step in initiation of proteolysis stems in part from analysis of engineered linked dimers of HIV-1 and RSV PR (2, 4, 5, 9, 11, 23, 40). Several arguments suggest that the initial cleavage in the polyprotein occurs in cis, i.e., that the dimerized PR domains act on the same polypeptides in which the PR domains are embedded (6, 10, 25). However, there are no definitive experiments that exclude cleavage in trans, i.e., the two dimerized PR domains attacking a neighboring third polyprotein molecule (12). In either case, once a mature dimeric PR is formed, it would act in trans to excise other PR domains, leading to a positive feedback loop, and also to cleave the polyproteins at other sites. To gain insight into the role of dimerization in activation of proteolysis, we have studied the enzymatic and structural properties of several RSV PR mutants. Mutations at the N terminus were found to render PR grossly defective in enzymatic activity in vitro, in assays both of proteolytic processing and of peptide hydrolysis. These observations extend those made previously in vivo in the RSV system (6, 28, 35, 39). The enzymatic defect is accounted for by the DLS observations that in the mutants the monomer-dimer equilibrium is dramatically shifted toward monomers.

More than one explanation is required to account for the observation that the several mutants do not dimerize properly. The four-stranded beta sheet comprising the N- and C-terminal five to six residues was known from previous studies to be of critical importance in dimer stability. For example, the beta sheet accounts for about 50% of the intersubunit ionic and hydrogen bond interactions between the two subunits (44). In the HIV-1 system, high concentrations of a peptide mimicking the N-terminal or C-terminal residues of PR inhibit enzyme activity, presumably by prying the dimer apart by competition (3, 36, 46). Thus, the extremely poor ability of ΔLAM to dimerize probably is due to the loss of backbone and side chain contacts between the N-terminal sequence of one subunit and the C-terminal sequence of the other. The ΔLAM dimer that apparently does form at high protein concentrations is some 1,000-fold reduced in specific activity, implying that it differs structurally from wild-type PR dimer. Comparison of the 1- and 5-mg/ml ΔLAM HSQC spectra revealed peak shifts that are not clustered about the wild-type dimer interface but rather are dispersed through other regions. If the mutant dimer is a single species, this result would imply a global folding difference compared with the wild-type dimer that would bring these dispersed regions together into a dimer interface. Alternatively, there may be multiple dimeric configurations, which would indicate a loss of specificity in the dimerization process. In at least one of these putative dimers, the Cys residues must be within 16 Å to account for the cross-linking by BMH. We speculate that the mutant dimer may represent an intermediate folding state for the wild-type enzyme.

In contrast to ΔLAM, the PR mutant L1A should be able to form the same backbone contacts of the beta sheet as wild-type PR, and therefore it must be the difference in side chains that accounts for the 1,000-fold reduction in its activity. The crystal structure of RSV PR (16) shows the Leu1 side chain to be buried in a hydrophobic pocket formed by the side chains of residues Val13, Leu36, Val84, Leu117, and Leu119 of the same subunit, plus Leu121′ of the other subunit. The carbon-to-carbon distances between relevant parts of these six side chains and the side chain of Leu1 are all less than 5.5 Å. Apparently, insertion of the Leu1 side chain into this pocket is critical for stabilization of the dimer. Either the empty pocket is unstable, causing a conformational change in the rest of the molecule that alters dimerization, or the loss of free energy associated with packing of the L1 side chain into the cavity destabilizes the dimer. It had been suggested previously that mutations of the residues involved in the conserved hydrophobic packing of retrovirus PRs would perturb enzymatic activity (44). We have not examined the L1V mutant for dimerization, but its low activity presumably reflects a very limited ability of the slightly shorter side chain of Val to fit properly into the pocket.

The weak activities of rPR and L1I, both of which have the initiating Met as an extra N-terminal residue, cannot be accounted for by either of the above arguments. We speculate that the enzymatic defect of these mutants also is due to lack of dimerization. The crystal structure of wild-type PR shows the amino group of Leu1 interacting with the oxygen atom of the amide of Asn123′. Any N-terminal extension would prevent this interaction. However, direct comparison between wild-type PR and these mutants is complicated due to the additional hydrophobic side chain of the Met residue. Because a minor species with a different N terminus would not be detected by N-terminal sequence analysis, it is not possible to assign the enzymatic activity observed for the rPR and L1I proteins to the major species carrying the initiating Met. Possibly, the major forms of these PRs are as diminished in activity as L1A, and the residual activity observed is due to limited removal of the initiating Met, thereby resulting in a small amount of wild-type PR in the case of rPR.

In summary, we suggest that the three types of mutations that cause loss of activity of RSV PR affect dimerization through different interactions. However, they all underscore the critical role played by the first few residues in stabilizing the dimeric form. Deletions at the N terminus prevent beta-sheet formation. Changes in the Leu1 residue prevent proper interaction of the side chain with a hydrophobic pocket created by six other leucine and valine residues. N-terminal extensions prevent formation of a hydrogen bond with the penultimate Asn residue of the protein. We did not quantitate the extent of the enzymatic defect of the N-terminally extended PRs because of the difficulty in distinguishing between activity of the extended PR itself and activity of any wild-type PR that had been generated therefrom by autocleavage. Based on recent results in the HIV-1 system (40), a fourth type of dimer-destabilizing mutation probably will need to be added to this list. The conserved Thr or Ser residue in the active-site sequence DTG or DSG apparently is critical for dimer stability. The hydroxyls of these two symmetrically positioned residues are a key part of the so-called “fireman's grip,” a hydrogen bonding network that was believed to maintain the geometry of the active site (30). Strisovsky et al. (40) showed that loss of these hydroxyl groups was enough to cause dissociation of the HIV-1 PR dimer. However, activity could be maintained in the context of a linked dimer, implying that the fireman's grip in fact is not essential for enzymatic activity. These results further demonstrate the delicate balance of forces that stabilize the dimer.

For HIV-1 PR, folding and dimer formation are hypothesized to be concomitant (12). Further evidence for this hypothesis was presented recently by Louis et al. (26). A four-residue N-terminal deletion of HIV-1 PR was found to be largely unfolded unless a dimer-stabilizing inhibitor was added. This is in stark contrast to results reported here for the similar RSV ΔLAM mutant, which is clearly well folded at the predominantly monomeric concentration of 1 mg/ml. As the concentration is increased to 5 mg/ml where dimers are predominant, the HSQC remains largely unchanged. This is not consistent with a large-scale unfolded-to-folded structural reorganization. Therefore, we believe that the concomitant folding and dimerization model of retrovirus PRs does not extend to RSV PR.

Comparison of the HSQCs of ΔLAM and dPR reveals a number of residues that experience substantially different chemical environments. However, many structural features appear to be conserved in the mutant protein, specifically including the mobile flap, alpha helix, and many of the loops and turns. Therefore, the general fold may well be similar to that of the wild-type PR. Since the N-terminally extended proteins His-ePR and the longer NC-PR (37) have a low autoprocessing activity and only barely detectable enzyme activity when measured on substrates in trans, and since these proteins also are compromised in dimerization, we predict that the PR domain in these extended proteins will be found to fold like ΔLAM. This prediction remains to be tested. Furthermore, we hypothesize that the structure of ΔLAM represents the prefolded status of the PR domain of the Gag polyprotein. This hypothesis can be addressed by comparing the ΔLAM HSQC with one obtained for an N-terminally extended construct such as His-ePR. Insight into the active-site geometry of precursor PR domains and the basis of low activity compared to mature PR may be important in understanding how proteolytic cleavage of RSV is initiated. Toward this end, we are pursuing high-resolution solution structures of ΔLAM PR and of extended PR forms.