Site-specific and redox-controlled S-nitrosation of thioredoxin

Abstract

Protein S-nitrosation on cysteine residues has emerged as an important posttranslational modification in mammalian cells. Previous studies have suggested a primary role for thioredoxin (Trx) in controlling protein S-nitrosation reactions. Human Trx contains five conserved Cys, including two redox-active catalytic Cys (Cys32 and Cys35) and three non-active-site Cys (Cys62, Cys69, and Cys73), all of which have been reported as targets of S-nitrosation. Prior reports have studied thermodynamic end points of nitrosation reactions; however, the kinetics of Trx nitrosation has not previously been investigated. Using the transnitrosation agent, S-nitrosoglutathione, a kinetic analysis of the selectivity and redox dependence of Trx nitrosation at physiologically relevant concentrations and times was performed, utilizing a mass spectrometry-based method for the direct analysis of the nitrosated Trx. Reduced Trx (rTrx) was nitrosated 2.7-times faster than oxidized Trx (oTrx), and rTrx was nitrosated selectively on Cys62, whereas oTrx was nitrosated only on Cys73. These sites of nitrosation were confirmed at the peptide level using a novel modification of the biotin-switch technique called the reductive switch. These results suggest separate signaling pathways for Trx-SNO under different cellular redox states.

Protein S-nitrosation, the modification of thiols to nitrosothiols (SNOs), has emerged as an important posttranslational modification (1). S-nitrosation of a thiol that is essential for structure or function (e.g., catalytic Cys, metal-binding Cys, and Cys involved in key structural disulfide bonds) can lead to dramatic functional and cellular outcomes. A prime example is the S-nitrosation of the catalytic cysteine in caspase-3 (Casp3) leading to inhibition of apoptosis (5). S-Nitrosation of Casp3 involves a cascade initiated by thioredoxin (Trx) (2–8), a ubiquitous, 12-kDa cytosolic protein that contains five Cys (Cys32, Cys35, Cys62, Cys69, and Cys73). We previously reported that human Trx1 is selectively nitrosated on Cys73 both in vitro and in Jurkat cells and that nitrosated Trx can form a complex with proapoptotic Casp3 and selectively transfer the nitroso moiety to the active-site Cys of Casp3 (4). This transnitrosation event blocks Casp3 catalytic activity and thus inhibits apoptosis as mentioned above (5). These data represented an example of highly specific transnitrosation signal transduction mediated by a protein–protein interaction occurring in cells. Because Trx interacts with numerous proteins and also maintains cellular redox homeostasis (9), Trx may have a central role in transnitrosation reactions in cells.

Biological nitrosation begins with the formation of nitric oxide (NO). NO reacts at diffusion-controlled limits in a radical–radical recombination reaction with thiyl radicals of glutathione (γ-Glu-Cys^•-Gly, GS^•), yielding S-nitrosoglutathione (GSNO) (10). Additionally, aerobic solution decomposition of NO involves the formation of N₂O₃, an efficient nitrosating agent that reacts avidly with nucleophiles such as water (to form nitrite) (11), and with glutathione (GSH) to form GSNO (12, 13). GSH is present at millimolar concentrations in all mammalian cells and is important for maintaining a reduced intracellular environment (14). GSNO has been detected both intra- and extracellularly in mammals and is the most abundant SNO in cells, with concentrations estimated at 1–5 μM (15–17). Once formed, SNOs, like GSNO, can undergo a transnitrosation reaction with another thiol leading to the transfer of NO. Although GSNO reacts promiscuously with protein thiols in vitro (18), the process is more selective in vivo, with nitrosation often limited to a specific Cys within the protein and modulation of GSNO levels controlling protein nitrosation (19). Formation of a specific protein–protein or protein–ligand complex may align the SNO and thiol for intermolecular transfer, suggesting a possible mechanism for transnitrosation signaling specificity. Cascading transnitrosation reactions between different, successive interacting partners is an intriguing paradigm for SNO-mediated signal transduction in cells.

The two redox-active, catalytic Cys (Cys32 and Cys35) of Trx reduce Cys disulfides on substrate proteins, producing a Cys32-Cys35 disulfide on the now oxidized Trx (oTrx) (9). This disulfide can then be reduced in vivo by the NADPH-dependent Trx reductase (TrxR) to regenerate reduced Trx (rTrx). The relative levels of rTrx and oTrx are likely controlled by the redox state of the cell and are related to both NADPH and GSH concentrations. The three non-active-site cysteines (Cys62, Cys69, and Cys73) are highly conserved in mammals but without a specific function. Cys73 is the site for dimerization between Trx molecules, but this appears to be an in vitro artifact.

There are inconsistent results pertaining to the specificity of GSNO-dependent nitrosation of Trx: Nitrosation has been reported on all three of the non-active-site Cys. A crystal structure of Trx was reported with a nitrosated Cys62 (20, 21), whereas other studies indicate that Cys69 (3, 22, 23) or Cys73 (3–5, 7, 24) are primary sites of nitrosation. Several groups reported that the redox state of the active-site influences nitrosation of non-active-site Cys. However, the reported redox-dependent, Cys specificity on Trx differs considerably under different experimental conditions (3, 7). Wu et al. reported no detectable nitrosation on rTrx and abundant nitrosation of oTrx on Cys73 (7), whereas Hashemy and Holmgren reported that rTrx is nitrosated at Cys69 and Cys73, with oTrx nitrosating only on Cys73 (3). Trx also catalyzes the removal of nitroso groups (denitrosation) from other protein thiols via the catalytic Cys (Cys 32 and Cys 35) (2, 6, 8, 25). Given the reducing activity of Trx/TrxR/NADPH (26), it is not surprising that Trx can act to denitrosate at high concentrations. Despite these reports, the role of Trx in controlling cellular nitrosation/denitrosation, and the mechanistic contribution of each of the Trx Cys, remains unclear.

Detection of protein SNOs is difficult. The commonly used methods suffer from low sensitivity and high background. The biotin-switch method and variations on the original procedure (19, 25, 27–34) involve denaturation of the protein, alkylation of free Cys, and precipitation of the protein. SNOs are then reduced with ascorbate followed by the reaction of the newly liberated Cys residues with a reactive group containing a tag for later identification. The biotin-switch method is limited by high background, low sensitivity, and nonspecific loss of protein during sample workup (35). Interpretation of results is complicated by differential ascorbate reduction and labeling rates of different Cys-SNOs and Cys, respectively (35, 36). Furthermore, the sites of nitrosation may be scrambled or lost upon denaturation. The biotin switch and alternate methods (e.g., MS or UV detection) (3, 7), are typically performed at supraphysiological concentrations of protein (10–100 μM) and GSNO (100 μM–1 mM) to aid detection. Additionally, long reaction times (> 10 min) are used, which likely drives the reaction to a thermodynamic equilibrium; thus, kinetic information on nitrosation is lost.

To gain a better understanding of the Trx-GSNO reaction under biologically relevant conditions, we developed an in vitro, MS-based methodology for the rapid and direct detection of protein-S-nitrosation, allowing kinetic analysis of the transnitrosation reaction between GSNO and Trx while simultaneously monitoring Trx redox status. We studied the selectivity and redox dependence of this reaction under relevant cellular concentrations of Trx and GSNO. rTrx and oTrx not only show different rates of nitrosation with GSNO, but also strikingly different reaction selectivity, with nitrosation of oTrx on Cys73 and rTrx on C62. We confirmed these findings with a peptide-level analysis, using a modification of the biotin-switch technique that we call the reductive switch, in which labeling is performed on native protein to prevent loss or scrambling of nitrosation upon denaturation. Together, these techniques allowed a thorough analysis of both the kinetics and the selectivity of Trx nitrosation by GSNO. These results may have implications regarding selective signaling under oxidative stress/hypoxic conditions.

Results

Direct Detection of Trx Nitrosation and Redox State with TOF-MS.

The redox state of the Trx active-site Cys may affect nitrosation (3, 7). To explore the effect of the Trx redox state on the GSNO-dependent rate of nitrosation, conditions were first developed to fully oxidize and reduce the Trx active-site Cys prior to reaction with GSNO. The oxidation state of recombinant human Trx1 was determined by TOF-MS analysis by deconvolution of the protein envelope to determine protein mass. Treatment of humanTrx (recombinant his₁₀-tagged protein) with Tris(2-carboxyethyl) phosphine (TCEP) produced rTrx with a mass of 14,126.6 (Fig. 1 A and D). Oxidized Trx is often prepared by treatment with H₂O₂ (3, 7); however, incubation of Trx with 1 mM H₂O₂ resulted in apparent thiol oxidation beyond the active-site Cys (m/z of 14,124.1) as well as in intermolecular disulfide bonds, yielding a dimeric oTrx (Fig. 1 B and E). Therefore, more moderately oxidizing conditions were investigated. Treatment of Trx with 1.2 equivalents of insulin (a natural Trx substrate) yielded oTrx (m/z of 14,124.7) that remained monomeric (Fig. 1 C and F). Insulin treatment is a physiologically relevant way of generating oTrx in vitro. This 2-Da shift between rTrx and oTrx (loss of two protons upon formation of the disulfide) was readily observed by TOF-MS analysis. When uniformly [¹⁵N]-labeled oTrx [molecular weight (MW) of 14,297.4)] was mixed in a 1∶1 ratio with [¹⁴N]-rTrx (MW of 14,126.6), both [¹⁵N]-oTrx and [¹⁴N]-rTrx underwent 1-Da mass shifts to 14,298.3 and 14,125.6, respectively, after 1 min of reaction (Fig. 1G). This result demonstrates that the average mass of the deconvoluted Trx peak represents the relative percentage of rTrx and oTrx in a mixture. Furthermore, oTrx is a substrate for rTrx, and in solution these species are in rapid equilibrium.

Fig. 1.

Direct detection of rTrx, oTrx, and Trx-SNO. (A) rTrx (TCEP treated) protein envelope. (B) oTrx (H₂O₂ treated) showing evidence for dimeric protein in the protein envelope (dimer exhibits double the number of peaks within a given interval due to higher charge states). (C) oTrx (insulin treated) protein envelope. (D) Deconvoluted spectrum of rTrx. (E) Deconvoluted spectrum of oTrx (H₂O₂ treated). (F) Deconvoluted spectrum of oTrx (insulin treated). (G) Deconvoluted spectrum of [¹⁴N]-rTrx treated 1∶1 with [¹⁵N]-oTrx, showing the shift of both [¹⁴N]-rTrx (Left) and [¹⁵N]-oTrx (Right) to a 50% reduced state (MW of 14,125.6 and 14,298.3, respectively). Black lines represent 50% oxidized, dashed gray lines represent fully oxidized, and solid gray lines represent fully reduced. (H) Deconvoluted spectra for a representative time course of GSNO (10 μM) incubation with rTrx (1 μM) showing the singly nitrosated (+29) peak and lack of double nitrosation (+58). The peak at 14,167 (+41) did not increase over time upon GSNO treatment; it is likely the Trx:acetonitrile adduct. (I) Deconvoluted spectrum of polynitrosated Trx (treatment of 100 μM Trx with 1 mM GSNO for 15 min).

Having determined that oTrx and rTrx could be quantitatively differentiated, the transnitrosation of Trx by GSNO and the redox state of Trx over the course of reaction were simultaneously measured. The RapidFire DS Module™ (Agilent Technologies), a dual-valve, desalting system (Fig. S1) was used for rapid removal of buffer salts, GSH, GSH disulfide (GSSG), and GSNO from the reaction mixtures, followed by direct injection for analysis by TOF-MS. The rapid and in-line desalting allowed the monitoring of the reaction at the anticipated physiological concentrations of both Trx (1 μM) and GSNO (10 μM) at short time points (1–4 min). The resolution and sensitivity of the TOF-MS system were sufficient to detect nitrosated proteins above 2% of total Trx. During the reaction time course, the deconvoluted base peak for rTrx (14,126.6) decreased in intensity, while a +29 peak emerged (MW of 14,155.4, indicating the gain of NO and loss of hydrogen) (Fig. 1H). The peak at 14,167 (+41) did not increase over time upon GSNO treatment; it is likely the acetonitrile adduct of Trx. Under these conditions, there was no evidence for glutathionylation [+306, which has been detected by others using aged GSNO stocks that contain decomposition products (37)] or for doubly (+58) or triply (+87) nitrosated Trx, which is consistent with the specificity implied by selective GSNO binding and transfer.

Reactions carried out at higher concentrations of GSNO (1 mM) and rTrx (100 μM) for longer periods of time (15 min) produced increased amounts of polynitrosation (Fig. 1I) and led to complete disulfide formation at the active site (rTrx to oTrx). The dominant Trx species was doubly nitrosated, and significant proportions of protein with three, four, and five NO additions were also observed. Because Trx contains only five Cys, even the active-site Cys (thought to be involved in denitrosation) became nitrosated under these conditions, possibly due to the simultaneous nitrosation of both Cys32 and Cys35 before the disulfide could form. At shorter times (4 min, Fig. S2), significant amounts of single, double, and triple nitrosation were observed.

Treatment of oTrx and rTrx with GSNO.

Having determined that Trx nitrosation could be directly detected without derivatization, separate kinetic analyses on the rate of nitrosation for rTrx and oTrx by GSNO were carried out. Both rTrx and oTrx (1 μM) were readily nitrosated upon treatment with 10 μM GSNO. The rate of nitrosation for rTrx was greater than that for oTrx, with pseudo-first-order rate constants of 3.53 ± 0.15 and 1.33 ± 0.05 s^-1, respectively (Fig. 2A). No evidence of polynitrosation was observed. oTrx-SNO was 2 Da lower in mass from rTrx-SNO (Fig. 2C), indicating that the initial redox state of Trx was reflected in the nitrosated species, without interconversion between rTrx to oTrx (as would be expected if rTrx first oxidized to oTrx and was then nitrosated). Rather, oTrx-SNO and rTrx-SNO are two distinct species that form at different rates.

Fig. 2.

Nitrosation of rTrx and oTrx. (A) Extent of nitrosation of rTrx and oTrx reactions of 1 μM Trx with 10 μM GSNO as detected by TOF-MS. Data shown are the average (with standard deviation) of the percent nitrosation determined by the ratio of the +29 (singly nitrosated) peak to total Trx. Data were collected as two independent runs with n = 5 for each time point. (B) Oxidation of rTrx at the active site (Cys32-Cys35) over the course of the reaction with GSNO. Percent oxidation for each time point was calculated as the mass shift from fully reduced (with a 2-Da shift for fully oxidized). (C) Average masses (4-min time point, n = 5 with SD) of rTrx, oTrx, rTrx-SNO, and oTrx-SNO.

Over the course of the reaction, rTrx partially oxidized, forming low amounts of oTrx (< 10% of total unmodified Trx, Fig. 2B). Oxidation of the rTrx active site following reaction with GSNO has been reported previously (3). On examination of the rTrx active-site oxidation with GSNO, GSSG, and GSH, it appears that GSSG, a contaminant in commercially available GSNO, is the relevant oxidant in vitro (Fig. 3 and Figs. S3 and S4).

Fig. 3.

Nitrosation rates of Cys mutants of Trx. (A) Plot of calculated rates of nitrosation for rTrx (black) and oTrx (gray). Rates were calculated assuming pseudo-first-order kinetics for Trx with an excess of GSNO, as the negative of the linear slopes of the plot of ln (Trx consumption) versus time. Trx consumption was calculated as the average (with standard deviation) of 100-% nitrosation (determined by the ratio of the +29 (singly nitrosated) peak to total Trx). Rates were calculated from a single experiment with n = 5 at each data point (C69S, C73S/C69S, C32S, C32S/C35S), or as the average from injections performed over multiple days (wild-type, C62S, C73S). (B) Plot of calculated nitrosation rates for the active-site mutant C32S/C35S, which served as a control for the reduction/oxidation conditions used in A. (C) Table of the rates shown in A and B.

Nitrosation Specificity of rTrx Versus oTrx.

Given the striking differences in nitrosation rate between rTrx and oTrx, the redox-dependent specificity of Cys nitrosation was determined. A library of Trx Cys mutants was designed to remove possible sites of nitrosation. The mutants made included C62S, C69S, C73S, C69S/C73S, and C32S/C35S and a series of kinetic experiments with GSNO were performed. Pseudo-first-order rate constants (Fig. 3C) were calculated based on the percent of nitrosated Trx. As previously described, mutant Trx (1 μM) was reacted with GSNO (10 μM) for 1, 2, 3, and 4 min, desalted using the RapidFire DS Module™, and analyzed by TOF-MS. For all analyzed proteins, only a single NO addition was observed under these conditions.

The rTrx mutants C73S (3.26/s) and C69S (3.59/s), both nitrosated at rates similar to wild type (3.53/s), with the double mutant C73S/C69S (2.92/s) slightly slower (Fig. 3A, black bars; Fig. 3C contains rate constants with error for each). Strikingly, rTrx-C62S (0.56/s) nitrosated significantly more slowly (> 6-fold), suggesting that the primary site of nitrosation for rTrx is Cys62. The oTrx mutant C73S demonstrated the slowest nitrosation rate (0.73/s), with C69S (1.13/s) also nitrosating slightly slower than wild type (1.33/s) (Fig. 3A, gray bars). A C69S/C73S double mutant showed no detectable nitrosation, and C62S nitrosated slightly faster than wild type (1.92/s). Based on these observations, Cys73 is the primary site of nitrosation for oTrx, with a secondary site at Cys69. Cys62, the presumed site of nitrosation for rTrx, does not appear to be nitrosated on oTrx. The specificity of oTrx nitrosation is markedly different from rTrx. The active-site mutant C32S/C35S was used to ensure that the use of insulin and TCEP did not influence the rate of nitrosation; indeed, there was no significant difference between rates of the insulin treated (1.53/s) and TCEP treated (1.64/s) C32S/C35S mutant (Fig. 3B)

Peptide-Level Confirmation of Cys Selectivity for Nitrosation of oTrx and rTrx.

To confirm the nitrosation selectivity suggested by the TOF-MS analysis of intact Trx and Trx mutants, a modified version of the biotin-switch assay, dubbed the reductive switch, was employed to identify the individual sites of nitrosation (Fig. S5). Briefly, the reaction was initiated by incubating Trx (1 μM) with GSNO (10 μM), and quenched over time (3, 4, 5 min) with deuterium₅-labeled N-ethyl maleimide (d₅-NEM, heavy label). Nonheavy labeled Cys were subsequently reduced with DTT and counterlabeled with nonisotopic NEM (light label). Samples were then trypsinized and analyzed by liquid chromatography (LC)-TOF-MS to determine ratios of heavy to light labeling for each Cys-containing peptide. Using DTT, which both quenched the d₅-NEM and reduced SNOs and other oxidized Cys, in place of ascorbate, increased the reproducibility of the assay by eliminating precipitation or buffer exchange steps that can lower protein recovery. Because Trx was labeled in the native (nondenatured) state, local protein structure improved SNO stability (as it is thought to in vivo) during the initial alkylation (heavy label) step; even in the folded state, nearly complete d₅-NEM labeling was seen in control experiments with rTrx (> 90%).

A potential limitation of the analysis is that DTT may reduce other oxidation species (e.g., disulfides), potentially leading to false positives. However, control experiments substituting GSH or GSSG for GSNO showed no time-dependent increase in oxidation on the peptides containing the non-active-site Cys (Cys62, Cys69, or Cys73), though the active-site Cys-containing peptide oxidized upon treatment with GSSG. GSNO treatment showed a clear, time-dependent increase of the NEM to d₅-NEM ratio on the peptides containing Cys62, Cys69, and Cys73 (Fig. S6), indicating the analysis accurately tracks the nitrosation of Trx. Furthermore, the nitrosation rates calculated with this method were comparable with the direct analysis of the intact, nitrosated protein.

Using the reductive-switch method, independent measurements of the rates of nitrosation on the Cys73- and on Cys62/Cys69-containing peptides (these two Cys were on a single tryptic peptide) (Fig. 4B) could be made. For the wild-type protein, rTrx showed the highest rate of nitrosation on the Cys62/Cys69-containing peptide (2.77/s, Fig. 4A, white bars), whereas oTrx had the fastest rate on the Cys73-containing peptide (1.47/s, Fig. 4A, gray bars), consistent with the intact protein results. The Cys62/Cys69-containing peptide showed little nitrosation in any of the oxidized proteins tested (wild type, C62S, or C73S all had rates < 0.9/s), whereas the reduced proteins (wild type, C73S) nitrosated readily on the Cys62/Cys69-containing peptide (2.77/s and 3.58/s respectively, Fig. 4, white bars). Considerable attenuation of the nitrosation rate on the Cys62/Cys69-containing peptide of rTrx-C62S mutant (0.73/s) supports the finding that Cys62, and not Cys69, is the target of nitrosation on rTrx.

Fig. 4.

Reductive-switch assay for nitrosation. (A) Rates of nitrosation for rTrx and oTrx on the peptides containing Cys73 (gray) and Cys62/Cys69 (white). C62S+C73S (mix) is a 1∶1 mixture of the two mutant proteins (final concentration 1 μM). All data points (0.1, 3, 4, and 5 min) were collected in triplicate from independently prepared samples. All samples were subjected to the reductive-switch protocol, and nitrosation at each data point was determined as the ratio of heavy labeled to light-labeled peptide. Rates were calculated assuming pseudo-first-order kinetics, as the negative of the linear slopes of the plot of ln (Trx consumption) versus time. (B) Table of the rates shown in A.

Unexpectedly, the rate of nitrosation on the Cys73-containing peptide of rTrx (2.23/s) was faster overall than the nitrosation of the Cys73-containing peptide of oTrx (1.47/s) (Fig. 4B). For the C62S mutant, this selectivity was reversed, and nitrosation on the Cys73-containing peptide was faster for oTrx-C62S (2.8/s) than for rTrx-C62S (1.29/s). Furthermore, the rTrx-C73S mutant protein nitrosated slightly faster than wild-type rTrx on the Cys62/Cys69-containing peptide. Taken together with the intact protein results, these data indicate that Cys62 is key to the nitrosation of rTrx, but that downstream reactions of rTrx-Cys62-SNO can occur, possibly via inter- or intra-Trx transnitrosation from Cys62 to Cys73, explaining the nitrosation observed on Cys73 of wild-type rTrx. If intermolecular transnitrosation was occurring, we would expect that Cys62 on one molecule of Trx would compensate for the mutation in rTrx-C62S, allowing faster nitrosation on Cys73; whereas if intramolecular transnitrosation from Cys62 to Cys73 was occurring, the addition of Cys62 on another Trx molecule would not change the rate of nitrosation on Cys73. To determine which of these possibilities were most likely, a 1∶1 mixture of the C62S and C73S Trx proteins was assayed. Both the reduced and oxidized forms of this mixture (rC62S + rC73S and oC62 + oC73S, respectively) had rates identical to wild type on both the Cys73-containing peptide and the Cys62/Cys69-containing peptides (Fig. 4), consistent with intermolecular transnitrosation, where rTrx-Cys62-SNO transfers NO to Cys73 on another rTrx molecule (Fig. 5, Lower Left). The in vivo relevance of this SNO-Trx mediated transnitrosation of Trx remains unclear.

Fig. 5.

Model of nitrosation selectivity for oTrx and rTrx. (A) rTrx structure (based on PDB ID code 1ERT) (39) showing the location of the 5 Cys. (B) Proposed model for interconversion and nitrosation of rTrx and oTrx.

Discussion

In the studies reported here we have undertaken a kinetic analysis of the redox dependence on Cys selectivity of Trx following exposure to GSNO. To this end, two previously undescribed methods for the analysis of Trx nitrosation were developed. The first, using a dual-valve rapid desalting device, followed by TOF-MS, allowed for the direct detection of Trx nitrosation and redox state at short time intervals. By analyzing the intact protein without derivatization, SNO loss and intra- or intermolecular scrambling of the nitrosation site was minimized. This technique allowed for analysis at time points as short as 1 min, allowed many replicates in a single experiment, was highly sensitive (approximately 1 pmol protein could be detected), and was highly reproducible. The second methodology, the reductive switch, was complementary to the online-desalting/TOF-MS intact protein experiments. Analysis of peptides following the reductive-switch procedure allowed the assignment of nitrosation sites to single peptides via TOF-MS. Taken together, these two techniques revealed the kinetics and selectivity of Trx nitrosation by GSNO.

Of the five cysteine residues in human Trx1 (Fig. 5A), only two were selectively nitrosated by GSNO, Cys62 and Cys73. rTrx nitrosated rapidly and primarily at Cys62, whereas oTrx nitrosated more slowly overall with selectivity toward Cys73 (Fig. 5B). Cys69, Cys32, and Cys35 did not appear to be selective targets for nitrosation; nitrosation occurred only at high concentrations of GSNO (1 mM) . At biologically relevant concentrations, Trx was nitrosated at only a single Cys (Cys62 or Cys73). This striking specificity preference in Cys reactivity between oTrx and rTrx is what would be expected for a modification involved in signal transduction. The differences in rates of nitrosation (with rTrx nitrosation > 2-fold faster than oTrx) imply that, in the case of a mixture of rTrx and oTrx, rTrx nitrosation would predominate under kinetically controlled conditions.

The concentrations of protein and GSNO had dramatic effects not only on the rates but also on the selectivity of Trx nitrosation. At the high concentrations of GSNO often used for in vitro experiments (100 μM–1 mM), nucleophile–electrophile reactivity drives the reaction leading to full nitrosation of all cysteines in the protein. This intrinsic reactivity, and the high concentrations needed for clear signal when using many standard nitrosation assays, provides a basis for the conflicting data in the literature regarding Trx nitrosation and highlights the importance of physiological conditions and concentrations for studying the selectivity of transnitrosation reactions. In sharp contrast, the results when the reaction is carried out at concentrations comparable to measured physiological concentrations of Trx [2–12 μM (38) and GSNO (1–10 μM (16, 17, 39)] showed clear selectivity for specific Cys and was redox-dependent.

We hypothesize that the selectivity of nitrosation is controlled by the conformation of the protein, either by conformational changes affecting GSNO binding (binding near Cys73 in oTrx and Cys62 in rTrx) or by different protein microenvironments in rTrx and oTrx that either activate the nucleophilicity of the respective Cys (Cys62 and Cys73) or that stabilize the corresponding SNOs. Although rTrx and oTrx are structurally similar (40) according to crystallographic measurements, the local dynamics of the protein may well be different in the two redox states, because the active-site disulfide bond of oTrx is predicted to reduce movement of the loops surrounding the active site. Cys73, the Cys nearest the active site (Fig. 5A), would be predicted to be particularly susceptible to conformational changes around the active site.

The peptide-level analysis using the reductive-switch assay supports the nitrosation of rTrx on Cys62 and of oTrx on Cys73; however, these data revealed an additional mechanistic insight: Cys62 is required for rTrx nitrosation on Cys73, but not for oTrx nitrosation on Cys73. oTrx appears to react with GSNO directly at Cys73, whereas rTrx appears to react with GSNO directly at Cys62, making rTrx-Cys62-SNO the direct kinetic product of the reaction of rTrx with GSNO. However, nitrosation was observed on Cys73 of wild-type rTrx, but not C62S-rTrx, and the reaction of a mixture of C62S-rTrx and C73S-rTrx showed wild-type nitrosation rates on Cys73, implying that intermolecular transnitrosation is possible. It appears that rTrx-Cys62-SNO may subsequently react with another molecule of rTrx to form rTrx-Cys73-SNO (Fig. 5B, Lower Left), and thus rTrx itself can be considered a target for transnitrosation reactions with rTrx-Cys62S. The in vivo significance of this in unknown, as it would be dependent on transient Trx-Trx interactions in areas of high localized Trx concentration. It is possible that for rTrx, Cys62-SNO dominates at shorter time points (at 5 min or less) and is the kinetically favored product, whereas Cys73-SNO is the more stable, thermodynamically favored product. rTrx-Cys62-SNO could then be considered a signaling intermediate for transnitrosation reactions, whereas Trx-C73-SNO could serve a dual role as both a transnitrosation agent (for example, with Casp3) and a thermodynamic sink for “NO storage.”

The potential stability of Trx-C73-SNO is intriguing in light of data indicating that GSNO is unstable in the presence of GSH (41). Attempted nitrosation of Trx by a mixture of GSNO and GSH showed no nitrosation, likely due to the reaction of GSNO with 1 mM GSH. In contrast, Trx-SNO is fairly stable to GSH at concentrations of 1 mM (60% SNO-rTrx remained, Fig. S7). Thus, if the Trx-GSNO reaction were to take place in areas of high local concentrations of GSNO, the stable Trx-SNO species could be available for downstream transnitrosation reactions in the presence of cytoplasmic GSH. Trx-SNO is thus an ideal candidate as an intermediate for redox and nitrosation-dependent signal transduction.

In vivo, cellular redox conditions control the ratio of rTrx to oTrx. Glutathione reductase (GR) (42, 43) likely minimizes concentrations of GSSG, at least in healthy (nonoxidatively stressed) cells, and TrxR (26) also drives oTrx toward rTrx. High intracellular GSH concentrations (1–5 mM) would suggest that rTrx (and rTrx-SNO) likely predominate in vivo under normal conditions. Because both GR and TrxR are NADPH-dependent, it is reasonable to assume that oTrx and oTrx-SNO may accumulate under conditions of oxidative stress or during depletion of NADPH. The differences in Cys selectivity for nitrosation between rTrx and oTrx may represent a redox-dependent “switch” for signal transduction via cascading transnitrosation reactions. Because Cys73 and Cys62 are on opposite faces of the protein (Fig. 5A), it is possible that oTrx-SNO and rTrx-SNO have different binding partners and thus different downstream transnitrosation targets (Fig. 6). This would provide an alternate mechanism of signal transduction during periods of nitrosative or oxidative stress.

Fig. 6.

Hypothetical model for redox-dependent signal transduction via Trx nitrosation. Under cellular reducing conditions, rTrx (Left) nitrosates on Cys62 and transfers that nitroso group to protein partners (blue). Under conditions of oxidative stress, oTrx (Right) nitrosates on Cys73 and transfers that nitroso group to different protein partners (red), such as Casp3.

Materials and Methods

Plasmids, Protein, and Reagents.

All reagents and chemicals were purchased from Sigma unless otherwise noted. Human Trx1 was expressed as an N-terminal his₁₀-tagged construct from the expression vector pET16b, in Escherichia coli BL21(DE3) as described previously (4). The [¹⁵N]-labeled Trx was expressed in BL21(DE3) cells grown in minimal media (M9) supplemented with 14 g/L [¹⁵N]-ammonium sulfate (Cambridge Isotopes) and purified as previously described. Cys mutants were generated by QuikChange (Stratagene) mutagenesis using the manufacturer’s protocol. Trx Cys variants were confirmed using TOF-MS of the intact proteins and the tryptic-digested peptides. Protein purity was assayed by SDS-PAGE, and concentration was determined by the Bradford assay using BSA as the standard. To generate rTrx, 50 μM Trx was treated with 1 mM TCEP or DTT for 5 min at room temperature, then desalted over a Bio-Spin 6 column (Bio-rad). To generate oTrx, 50 μM Trx was treated with TCEP or DTT and desalted as above, followed by treatment with 60 μM (1.2 equivalents) bovine insulin, and desalted over a Bio-Spin 6 column. Removal of insulin was confirmed by TOF-MS.

Direct Analysis of Protein Nitrosation.

rTrx and oTrx were prepared as described above, then diluted to 2 μM in 50 mM potassium phosphate, pH 7.4 (PB). A fresh solution of GSNO (Cayman Chemicals) was prepared in PB and the concentration determined spectrophotometrically (ϵ = 0.92 at 335 nm). GSNO stocks were diluted to 20 μM and stored protected from light for no longer than 6 h. All reactions were carried out at room temperature. Trx and GSNO were loaded into separate 1-mL syringes and connected to a dual syringe pump attached to a 3-μL mixing tee. After mixing, Trx and GSNO were allowed to react in the tubing for a set time (based on flow rate and dead volume) before application to the RapidFire™ DS Module (Agilent Technologies). The system was equilibrated with water (containing 0.1% formic acid, mobile phase A) in the first valve state. During the second valve state, the sample loop (10 μL) was filled with the reaction mixture. In the third valve state, the sample was applied to the trapping cartridge (C18) and washed with mobile phase A. In the fourth valve state, the sample was eluted from the trapping cartridge with mobile base B (acetonitrile containing 0.1% formic acid) and injected directly onto an Agilent TOF mass spectrometer (MSD TOF, Model 1969A). Following injection the device returned to valve state 1. The cycle (equilibration, loop filling, washing/desalting, and elution) was repeated 5 times for each time point. Syringe pump flow rates were changed for each time point to increase reaction time. The following instrument parameters were used for the TOF-MS: gas temperature, 350 °C; gas flow, 10 L/ min; nebulizer, 30 psi; capillary voltage, 3,500 V; fragmentor, 90 V. The total ion chromatogram consisted of periodic elutions for each replicate. The extracted protein envelopes were manually examined to ensure no dimer or insulin remained. Deconvolution of the protein envelopes was performed using MassHunter Software (Agilent Technologies) in the 13,000- to 16,000-MW range, using the extraction ion m/z range of 600–1,600. The resulting deconvoluted spectra intensities were used to calculate the relative ratio of Trx to SNO-Trx for quantitative analysis of reaction rates.

For denitrosation experiments, 1 μM oTrx was treated with 10 μM GSNO or 1 mm GSH for 2 min, after which 10 μM GSNO or 1 mm GSH was added via a second syringe pump and mixing tee, and the mixture incubated for 1 min before rapid desalting and TOF-MS analysis as above.

Reductive-Switch Assay.

rTrx and oTrx were prepared as described above, then diluted to 1 μM in 50 mM potassium phosphate, pH 7.4, with 200 mM NaCl (PBS). Fresh GSNO stocks were prepared as above in PBS. All reactions were performed in triplicate. For each reaction, the final volume was 500 μL containing 1 μM Trx and 10 μM GSNO. Reactions were quenched (Cys alkylated) by the addition of 1 mM (final concentration) d₅-labeled N-methyl-maleimide (d₅-NEM, CDN Isotopes) at 37 °C for 1 h. For the 0.1-min reactions, GSNO was added, tubes vortexed, then d₅-NEM was immediately added. Samples were then treated with 2 mM DTT for 10 min at room temperature to reduce nitrosothiols. NEM (5 mM) was added as a counterlabel and samples were reacted for 1 h at 37 °C. Samples were concentrated over 5-kDa cutoff spin concentrators (Vivaspin) to 50 μL, followed by addition of trypsin (4 μg/mL) (Trypsin Gold, Promega). Samples were digested for 1 h at 37 °C, then analyzed via LC-MS using a Zorbax C18 (2.1 × 50 mm) column and a Waters LCT Premier XE TOF-MS. The heavy- and light-labeled peptides containing Cys32/Cys35, Cys62/Cys69, and Cys73 were quantified using MassLynx software (Waters).