Thioltransferase (Glutaredoxin) Is Detected Within HIV-1 and Can Regulate the Activity of Glutathionylated HIV-1 Protease in Vitro *

Previous studies have suggested that the two conserved cysteines of the HIV-1 protease may be involved in regulating protease activity. Here, we examined diglutathionylated wild type protease (Cys-67-SSG, Cys-95-SSG) and the monoglutathionylated protease mutants (C67A, Cys-95-SSG and C95A, Cys-67-SSG) as potential substrates for thioltransferase (glutaredoxin). Time-dependent changes in the extent of deglutathionylation of each protein were assayed by reverse phase-high performance liquid chromatography. Glutathione alone was not an effective reductant, whereas thioltransferase displayed differential catalysis toward the Cys-95-SSG and Cys-67-SSG sites. At low thioltransferase concentrations (5 nm), deglutathionylation occurred almost exclusively at Cys-95-SSG. With substantially more thioltransferase (100 nm) Cys-67-SSG was partially deglutathionylated but only at 20% of the rate of Cys-95-SSG reduction. Treatment of the diglutathionylated protease with thioltransferase not only restored protease activity but generated an enzyme preparation that had a 3- to 5-fold greater specific activity relative to the fully reduced form. Immunoblot analysis of HIV-1MN virus with an antibody to thioltransferase detected a band co-migrating with recombinant thioltransferase that persisted following subtilisin treatment, indicating the presence of thioltransferase within HIV-1. Our results implicate thioltransferase in the regulation and/or maintenance of protease activity in HIV-1 infected cells.

pression of the protease can lead to premature cell death (4 -6). Studies using a highly active single-chain form of the HIV-1 protease in combination with protease inhibitors (7), as well as studies of HIV-1 containing protease mutants (8), have demonstrated the requirement for a defined range of protease activity for optimal viral production. Similar findings have been obtained for the avian sarcoma leukemia virus protease (9). Thus, retroviral polyprotein processing is a sensitive and tightly controlled process which, if disrupted, could be deleterious to viral maturation.
Recent studies have indicated that the two cysteines of the HIV-1 protease may be involved in redox regulation of protease activity (10). The two cysteine residues are highly conserved among HIV-1 isolates from patients. Of 31 different HIV-1 isolates spanning 6 different major subtypes, all contain cysteine 95, whereas 29 contain cysteine 67 (11). By contrast, the HIV-2 protease does not contain conserved cysteine residues, suggesting that the cysteine residues of the HIV-1 protease are not absolutely required for retroviral polyprotein processing but rather may offer a selective advantage for HIV-1 in vivo. Under native conditions, both cysteines of the HIV-1 protease are unusually susceptible to oxidation (12). Modification of either cysteine residue of the HIV-1 protease with sulfhydryl modifying reagents, such as 5,5Ј-dithiobis(2-nitrobenzoic acid), leads to a decrease or loss in protease activity (10,(12)(13)(14)(15). Additional studies, utilizing protease mutants produced by site-directed mutagenesis, demonstrated that mixed disulfides between the cysteine residues and glutathione (glutathionylation) have dramatic effects on protease activity. Glutathionylation of cysteine 67, a solvent exposed residue, increased activity severalfold and also stabilized the activity in vitro. However, glutathionylation of cysteine 95, located at the dimer interface, abolished protease activity (10). In vivo, cysteine modification of proteins with glutathione increases in cells under conditions of oxidative stress (16 -18), which would make the cysteines of the HIV-1 protease likely candidates for glutathionylation in HIV-1-infected cells.
The reversible nature of glutathionylation and its effects on HIV-1 protease activity led us to investigate the possible involvement of human thioltransferase in regulating the redox status of these cysteine residues. Thioltransferase (EC 1.8.4.2) (also known as glutaredoxin) belongs to the class of enzymes termed thiol-disulfide oxidoreductases, which include thioredoxin and protein disulfide isomerase (19 -21). The prevalence of S-glutathionylation of proteins under conditions of oxidative stress has focused attention on the role of thioltransferase in maintaining the activities of important cellular proteins that are altered by glutathionylation (20 -24). In this light, we hypothesized that thioltransferase may play a similar role for the HIV-1 protease under conditions favoring glutathionylation during viral replication. In the current study, we discovered that thioltransferase preferentially deglutathionylates cysteine 95 of the HIV-1 protease leading to modulation of protease activity. In addition, we report, for the first time, the detection of this mammalian protein within HIV-1 virions.

EXPERIMENTAL PROCEDURES
HIV-1 Proteases and Thioltransferase-Recombinant HIV-1 protease (strain HXB2) and two HIV-1 protease mutants (C67A and C95A) were expressed and purified from Escherichia coli as described previously (10). Protease activity was assayed using a 9 amino acid peptide spanning the p17/24 junction in the HIV-1 Gag protein as the substrate, as described previously (10). Assays for protease activity were carried out in 100 -200 mM sodium phosphate buffer, pH 6.2, containing 1 mM EDTA, 10% glycerol, 5% ethylene glycol for 1-5 min, as indicated, with a final substrate concentration of 2 mM. Recombinant human thioltransferase (57 units/mg) was expressed in E. coli, purified and assayed as described previously (25), and stored at Ϫ70°C in 20% glycerol.
In Vitro Glutathionylation of the HIV-1 Proteases-The HIV-1 wild type protease and mutant proteases (C67A and C95A) were refolded and stored in 20 mM HCl at pH 1.6 at 0.2-1.5 mg ml Ϫ1 as described previously (10). To glutathionylate the cysteines, the proteases were incubated at 37°C for 2 h in a buffer containing 250 mM Tris-HCl at pH 7.8, 6.0 M guanidine, 1 mM disodium EDTA and 40 mM glutathione disulfide. The reaction was stopped by acidification to pH Ͻ 2.0 with trifluoroacetic acid. More than 80% of both cysteines were glutathionylated using this procedure. The glutathionylated proteases were purified (Ͼ95%) from residual unmodified protease by RP-HPLC by exploiting the change in retention time imparted by each glutathionyl moiety. Each of the 3 glutathionylated proteases were separated on a Vydac C 18 column using a gradient method with solvent A, consisting of deionized water containing 0.05% trifluoroacetic acid, and solvent B, consisting of acetonitrile containing 0.05% trifluoroacetic acid. The column was eluted at 1 ml min Ϫ1 and the percent of solvent B was increased linearly from 5 to 35% over 5 min followed by an increase to 57.5% solvent B over the next 15 min; solvent B was then increased to 95% over 2 min and then ramped back to the initial starting conditions over the final 5 min of the gradient program. Protease elution was monitored at both 205 and 276 nm. The location of the peaks from RP-HPLC corresponding to the glutathionylated forms of the C67A and C95A mutant proteases had been identified previously by electrospray mass spectrometry (10). Following purification by RP-HPLC, all proteases were refolded as described previously (14), except that DTT was omitted to prevent reduction of the glutathionyl disulfides. Protein concentrations were calculated using the molar absorptivity (⑀ 280 nm ϭ 12,800 M Ϫ1 cm Ϫ1 ). The final stock solutions of the various proteases were stored in 20 mM HCl, pH 1.6, at Ϫ70°C.
Thioltransferase Assays Using Glutathionylated Forms of the HIV-1 Protease-To test the proteases as substrates for human thioltransferase, the glutathionylated forms of the protease (at 1-4 M) were incubated in 200 mM sodium phosphate buffer, pH 6.2, containing 1 mM EDTA, 10% glycerol, 5% ethylene glycol, 60 g ml Ϫ1 of bovine serum albumin and 0.5 mM reduced glutathione (unless otherwise indicated). Thioltransferase was added, and the solution was incubated at 37°C for various times, and where indicated in text aliquots were assayed for protease activity. The reaction was stopped by acidification (pH Ͻ 2) with trifluoroacetic acid. The samples were then analyzed by RP-HPLC, and the areas of absorbance peaks (205 nm) corresponding to the glutathionylated and deglutathionylated proteases were used to determine the extent of deglutathionylation at each cysteine residue.
Immunoblot Analysis of Viral Preparations-HIV-1 MN samples were obtained from the Clone 4 cell line (26) and digested with subtilisin as described previously (27). The treated viral preparation was then repurified by sucrose banding. Viral samples were treated with 50 mM DTT and 20 mM TCEP (Calbiochem) to maintain the cysteine residues of thioltransferase in their reduced form. Samples were electrophoresed on a 10% Bis-Tris polyacrylamide gel with MES running buffer using the NuPage system from Novex (San Diego, CA). Proteins were electroblotted onto nitrocellulose, and thioltransferase was detected using an anti-glutaredoxin antibody obtained from American Diagnostica, Inc. (Greenwich, CT). Monoclonal antibodies to gp120 were provided by the AIDS Vaccine Program, NCI-FCRDC (Frederick, MD), and the monoclonal antibodies to p24 were obtained from Intracel (Cambridge, MA). For competition experiments, 20 g of glutaredoxin antibody were preincubated at room temperature in the presence or absence of 6 g of purified thioltransferase for 2 h at a dilution of 1:50, followed by centrifugation at 10,000 rpm for 10 min to eliminate thioltransferasebound antibodies.

HPLC Analysis of Glutathionylated and Unmodified HIV-1
Proteases-The wild type HIV-1 protease and the C67A and C95A mutants were each glutathionylated at their respective cysteine residues, purified by RP-HPLC, and refolded as described previously (14) for use as substrates for human thioltransferase. The glutathionylated form of each protease monomer had a distinct retention time on the C 18 column and could be separated easily from the corresponding unmodified form by RP-HPLC analysis, thus providing an effective assay to assess the extent of deglutathionylation (Fig. 1). Of the 3 different glutathionylated HIV-1 proteases, the diglutathionylated wild type, denoted as WT-(SSG) 2 , eluted earliest. Further characterization showed that the eluted fraction corresponding to this peak gave the expected mass by electrospray mass spectrometry (expected 11,388 Da; obtained 11,389 Da) for the WT-(SSG) 2  and C67A protease glutathionylated at cysteine 95 (solid line). Each tracing was obtained with 1-10 g of purified protease and all were normalized to the same absorbance at 205 nm to aid comparison. This analysis reveals that each protease has a distinct retention time on the C 18 column. The difference in retention times possibly reflects varying degrees of interaction with the C 18 column of the surface exposed area around cysteine 67 of the protease versus that for the dimer interface region around cysteine 95.
protease could be readily converted to the reduced form (Ͼ90%) in the presence of 20 mM TCEP, a potent reducing agent (28) (data not shown).
Thioltransferase Preferentially Deglutathionylates Cysteine 95 of the HIV-1 Protease-Thioltransferase readily catalyzed the GSH-dependent deglutathionylation of the C67A,Cys-95-SSG (glutathionylated at cysteine 95) protease as determined by RP-HPLC analysis (Fig. 2). Further experiments demonstrated the requirement for thioltransferase catalysis, because glutathione alone (0.5 mM) showed no measurable reduction of cysteine 95 during a 10 min time course, whereas nearly 30% was deglutathionylated when 5 nM thioltransferase was added with GSH ( Fig. 2). As reported previously, the C95-SSG protease is inactivated by the glutathionyl moiety (10). However, treatment of the inactivated protease with thioltransferase resulted in the restoration of protease activity in a time-dependent manner reflecting the rate of deglutathionylation (Fig.  2). This observation is also consistent with previous results showing that reduction of C67A,Cys-95-SSG with 10 mM DTT could restore protease activity.
In contrast to the results with the C67A,Cys-95-SSG protease, the C95A,Cys-67-SSG (glutathionylated at position 67) protease was a poor substrate for thioltransferase. Treatment with 5 nM thioltransferase did not lead to measurable deglutathionylation of cysteine 67 over a 10 min assay (Fig. 2), nor did it significantly alter protease activity (Fig. 2). C95A,Cys-67-SSG, however, was not completely resistant to thioltransferase treatment. Exposure of the C95A,Cys-67-SSG protease to a much higher concentration of thioltransferase (100 nM) for 5 min deglutathionylated up to 20% of the C67-SSG protease compared with 60% for the C67A, Cys-95-SSG protease. These studies indicated that thioltransferase preferentially catalyzes the deglutathionylation of cysteine 95 of the HIV-1 protease.
The wild type protease, WT-(SSG) 2 (glutathionylated at both cysteine 67 and cysteine 95) (so that the dimeric protease is tetraglutathionylated), was then tested as a substrate for thio-ltransferase. Based on the results with the mutant proteases, one would predict that the WT-(SSG) 2 protease would be selectively deglutathionylated at cysteine 95 in the presence of thioltransferase. Treatment of the WT-(SSG) 2 protease with 20 nM thioltransferase and 0.5 mM GSH for 5 min resulted in a decrease in the area for the WT-(SSG) 2 peak and the generation of an additional peak in the RP-HPLC chromatogram along with two minor peaks eluting later (Fig. 3, top panel). The new peak eluted approximately 0.5 min later than that assigned to the WT-(SSG) 2 protease, and the mass obtained by electrospray mass spectrometry was consistent with a monoglutathionylated form of the protease monomer (expected 11,083 Da; obtained 11,084 Da). The retention time for this new peak corresponded closely to the retention time for the C95A,Cys-67-SSG protease mutant (retention time of 15.6 min compared with 15.8 min for the mutant, see Fig. 1), suggesting that the new peak represented the conversion of part of the WT-(SSG) 2 protease to the WT-Cys-95-SH,Cys-67-SSG protease generated by deglutathionylation at position 95 but not at position 67. If so, we would hypothesize that thioltransferase treatment of the WT-(SSG) 2 protease would restore protease activity.
Thioltransferase Activates Protease Activity of the WT-(SSG) 2 Protease-To determine if thioltransferase restored protease activity for the WT-(SSG) 2 protease, we measured protease activity following thioltransferase treatment of the inactive WT-(SSG) 2 protease, expecting that selective deglutathionylation at cysteine 95 would restore protease activity. As predicted, thioltransferase treatment of the WT-(SSG) 2 protease did reactivate the HIV-1 protease (Fig. 3, top panel, inset). The specific activity was 2.3 M/min/mg based on total protein (Fig.  3, top panel, inset) but it was 8.1 M/min/mg when calculated based on the amount of protein associated with that of the new peak generated by thioltransferase treatment (Table I). The change in protease activity for the WT-(SSG) 2 protease resulting from thioltransferase treatment correlated closely with the extent of deglutathionylation as assessed by RP-HPLC (Table  I). Calculating specific activity based on the percentage of protease deglutathionylated to active forms by thioltransferase (which includes those forms deglutathionylated at cysteine 95 as determined by RP-HPLC analysis) yielded a specific activity approximately 3-to 5-fold greater than that expected for the TCEP-treated enzyme (Table I). These data are consistent with our previous studies, which showed that glutathionylation of C95A protease at cysteine 67 significantly increased protease activity as compared with wild type (10). Thus, we conclude that selective glutathionylation of the wild type enzyme at cysteine 67 may result in a form of the enzyme with significantly greater protease activity than unmodified wild type protease.
Deglutathionylation of the WT-(SSG) 2 protease was also studied as a function of GSH concentration. In the absence of GSH, no deglutathionylation was observed (Fig. 4). Measurable deglutathionylation of the WT-(SSG) 2 protease by GSH alone was not detected until more than 1 mM GSH was present, whereas the thioltransferase-catalyzed reaction was already half-maximal at less than 0.1 mM GSH (Fig. 4). Based on these data, thioltransferase (20 nM), in the presence of glutathione, was more than 50,000 times more potent than GSH alone (Ͼ1 mM) at deglutathionylating cysteine 95.
GSH alone at concentrations as high as 10 mM did not result in measurable deglutathionylation of the cysteine 67 residue of the WT-(SSG) 2 protease. In contrast, treatment with 10 mM DTT resulted in partial deglutathionylation at both cysteine 67 and cysteine 95 as shown by the presence of four peaks in the RP-HPLC chromatogram representing the four different redox forms of the wild type protease monomer (not shown). The same four peaks were obtained when the WT-(SSG) 2 protease was treated with 100 nM thioltransferase (Fig. 3, bottom panel). In this case, the protease specific activity was 5.2 M/min/mg based on total protein but is 8.3 M/min/mg when calculated based on the amount of active forms of the enzyme present in the mixture. This value is still significantly higher than that for the fully reduced wild type protease and may indicate progressive deglutathionylation of the tetraglutathionylated wild type protease dimer. This could generate partially glutathionylated heterodimers as intermediates that are more active than the fully reduced dimeric protease. These studies show that the enzymatic deglutathionylation by thioltransferase is much more efficient than chemical deglutathionylation by DTT on a molar basis (100 nM thioltransferase versus 10 mM DTT, i.e. 100,000 times), and the thioltransferase-catalyzed reaction displayed a clear preference for cysteine 95, which is consistent with the data for the individual mutant proteases.
Thioltransferase Is Detected Within HIV-1 Virions-The activation of protease activity from the selective deglutathionylation of cysteine 95 of the HIV-1 protease by human thioltransferase suggested a possible role for this enzyme in HIV-1 replication. It has been shown that vaccinia virus and T4 phage both encode and package their own glutaredoxins, which are functional as thioltransferases (29 -31). Thus, it seemed possible that HIV-1 might package human thioltransferase in a similar manner. To test this hypothesis, we utilized viral preparations from an H9 cell clone (26) that produces virions in the absence of significant levels of cellular microvesicles that would contaminate the preparation with cellular proteins as indicated by others (32). Such preparations had previously been employed to identify the presence of cyclophilin A in HIV-1 (27). When such a virion preparation is treated with subtilisin, proteins found on the outside of HIV-1 virions are digested, TABLE I Activation of HIV-1 protease activity following thioltransferase treatment of the WT-(SSG) 2 protease WT-(SSG) 2 enzyme at 1.8 M (20 g ml Ϫ1 ) (Ͼ95% glutathionylated at Cys-67 and Cys-95 as determined by HPLC) was incubated at 37°C in a total volume of 75 l containing 200 mM sodium phosphate, pH 6.2, 10% glycerol, 5% ethylene glycol, 1 mM EDTA, 60 g ml Ϫ1 bovine serum albumin, and 0.5 mM reduced glutathione. The reactions were started by the addition of 0 -20 nM of thioltransferase. A 2.5-l aliquot was assayed for protease activity 4 min after treatment by incubating the protease in the presence of 3 mM protease substrate for 5 min. The remainder of the sample (72.5 l) was acidified with 5 l of 10% trifluoroacetic acid 5 min after the initiation of the reaction, and the extent of deglutathionylation of the WT-(SSG) 2 protease at Cys-95 was determined by RP-HPLC as described under "Experimental Procedures." Treatment of WT-(SSG) 2  a Specific activity was calculated based on the percent of the total enzyme that was converted to active forms as assessed by RP-HPLC (the monoglutathionylated form of the protease and the wild type unmodified form). In the case of TCEP treatment, Ͼ90% of the total enzyme was converted to the fully reduced form, and the specific activity obtained can be used to compare with that of the preparations treated with thioltransferase. b NA, not applicable; trace levels of deglutathionylated protease present cannot be accurately assessed by RP-HPLC. The data correspond to one representative experiment; other experiments gave similar results. whereas proteins within HIV-1 remain intact. Following subtilisin digestion, the virus preparation is purified via a sucrose gradient that eliminates any remaining microvesicle contamination (27). As shown in Fig. 5A, a band co-migrating with pure human thioltransferase was detected in purified HIV-1 virion preparations. Furthermore, following competition of the thioltransferase antibody with purified thioltransferase, the band co-migrating with thioltransferase was no longer detected, confirming this band as human thioltransferase (Fig. 5A). After subtilisin treatment, the band intensity decreased somewhat, indicating that some of the thioltransferase was located externally to the virus. However, the remaining band was about 50% of the intensity of the untreated virus preparation. By contrast, gp120, a membrane protein located on the outside of the virion, was almost completely eliminated by subtilisin treatment (Fig.  5B), and as expected, p24, the capsid protein, was unaffected by the subtilisin treatment (Fig. 5B). Overall, these results indicated that a significant portion of the thioltransferase band in the viral preparations represented thioltransferase within viral particles. DISCUSSION Cysteine residues in proteins serve a number of important roles, including catalysis (33), protein folding (34), and DNA binding (35). Cysteines can also play an important role in the regulation of enzyme activity like that described for citrate synthase and adenovirus protease (33, 36 -38). The highly conserved nature of the two cysteine residues of the HIV-1 protease, which are neither directly involved in the catalytic mechanism nor involved in intramolecular disulfide bond formation, suggests a role for these residues in the regulation or maintenance of protease activity. In support of this, Davis et al. (10) found that modification of cysteine 67 with the common cellular thiol glutathione leads to increased protease activity and enzyme stabilization, whereas the same modification at cysteine 95 leads to inhibition of activity. The concept of regulation is now further supported by the current studies that demonstrate thioltransferase selectivity. The data demonstrate that thioltransferase preferentially catalyzes the deglutathionylation of cysteine 95 from each of the diglutathionylated subunits of the dimeric HIV-1 wild type protease. This may then lead to a more stable form(s) of the protease (WT-Cys-95-SH,Cys-67-SSG) with greater protease activity than the fully reduced enzyme based upon studies of the glutathionylated C95A protease mutant (10). This is supported by the apparent increase in specific protease activity noted for the fully diglutathionylated wild type protease mixture following thioltransferase treatment. The differential effects of glutathionylation at Cys-95 and Cys-67 and the differential catalysis by thioltransferase suggest that the protease may undergo multiphasic changes in activity during the infection cycle, dependent on the redox status of the host cell and virion.
Mixed disulfide formation in proteins (S-thiolation) is significantly increased in cells placed under oxidative stress (16 -18, 39). Recently, evidence has accumulated indicating that HIV-1 infection leads to oxidative stress and that this is enhanced by agents that induce oxidative stress, such as tumor necrosis factor ␣ or hydrogen peroxide (40 -45). Also, several studies have shown that HIV-1 infection leads to significant changes in the levels of reduced and oxidized glutathione in certain cells (40,46,47). Thus, it is likely that modification of the cysteine residues of cellular and viral proteins would be favored in cells that undergo oxidative stress when infected with HIV-1. In particular, cysteine 67 may be especially susceptible to S-thiolation in HIV-1 infected cells since this residue is solventexposed and unusually reactive under native conditions (12,48). In support of the concept that viral infection can lead to oxidative stress is the recent demonstration that sendai virus leads to the production of significant amounts of glutathionemixed disulfides in cells 24 h after infection (49). Thioltransferase, which has been found to be a ubiquitous enzyme in mammalian cells, is abundant in human macrophages (50) and therefore may play a significant role in maintaining the activity of the HIV-1 protease during oxidative stress.
We found that thioltransferase was detectable in virion preparations following subtilisin digestion, providing evidence that this enzyme is present within virions. Other host proteins have been identified previously within HIV-1 virions including cyclophilin A, which is essential for viral infectivity (11,27,(51)(52)(53). Immunoblot analysis indicated very low levels of thioltransferase, suggesting passive acquisition of cellular thioltransferase by the virus. It would be of interest to determine if thioltransferase within HIV-1 virions offers an advantage in regards to viral infectivity. Interestingly, vaccinia virus and T4 bacteriophage each encode two different functional glutaredoxins that have thioltransferase activity (29 -31, 54). Although the role of the glutaredoxins (thioltransferases) in the viral life cycle is uncertain, its presence in the virus supports the argument that thioltransferase may play a fundamental role in optimizing HIV-1 replication.
Several roles for thioltransferase in viral replication are conceivable, including homeostatic or regulatory control of the glutathionylation status of sulfhydryl groups on vital proteins, including metabolic enzymes and transcription factors (20). Thioltransferase could also affect the glutathione-dependent folding of viral proteins containing cysteine residues, including the HIV-1 protease (34). Finally, thioltransferase could protect important viral proteins that are oxidatively damaged during oxidative stress. In particular, inactivation of the protease through disulfide bond formation may explain the inability to completely restore polyprotein processing in immature virions once they have been released from cells (55). Our studies suggest that the cysteine residues of the HIV-1 protease coupled with thioltransferase and GSH may optimize HIV-1 protease activity particularly under conditions of oxidative stress.