Unequivocal Determination of Site-Specific Protein Disulfide Bond

Sep 16, 2013 - Jenna Scotcher,. †,§. Benjamin J. Bythell,. † and Alan G. Marshall*. ,†,‡. †. National High Magnetic Field Laboratory, Flori...
0 downloads 0 Views 1MB Size
Article pubs.acs.org/ac

Unequivocal Determination of Site-Specific Protein Disulfide Bond Reduction Potentials by Top-Down FTICR MS: Characterization of the N- and C‑Terminal Redox-Active Sites in Human Thioredoxin 1 Jenna Scotcher,†,§ Benjamin J. Bythell,† and Alan G. Marshall*,†,‡ †

National High Magnetic Field Laboratory, Florida State University, 1800 E. Paul Dirac Drive, Tallahassee, Florida 32310, United States ‡ Department of Chemistry and Biochemistry, Florida State University, 95 Chieftain Way, Tallahassee, Florida 32306, United States S Supporting Information *

ABSTRACT: We report the reliable determination of equilibrium protein disulfide bond reduction potentials (E°′) by isotope-coded cysteine alkylation coupled with top-down Fourier transform ion cyclotron resonance mass spectrometry (FTICR MS). This technique enables multiple redox-active sites to be characterized simultaneously and unambiguously without the need for proteolysis or site-directed mutagenesis. Our model system was E. coli thioredoxin, and we determined E°′ for its CGPC active-site disulfide as −280 mV in accord with literature values. E°′ for the homologous disulfide in human thioredoxin 1 (Trx1) was determined as −281 mV, a value considerably more negative than the previously reported −230 mV. We also observed S-glutathionylation of Trx1 and localized that redox modification to Cys72; E°′ for the intermolecular disulfide was determined as −186 mV. Intriguingly, that value corresponds to the intracellular glutathione/ glutathione disulfide (GSH/GSSG) potential at the redox boundary between cellular differentiation and apoptosis.

D

The reduction potential of a protein that contains redoxactive metal center(s) or organic cofactor(s) is often determined by electrochemical methods such as protein film voltammetry (PFV) or mediated redox potentiometry. Recently, reduction potentials for disulfide bonds in archaeal thioredoxins were measured directly by PFV.8 However, the electrochemistry of protein dithiol/disulfide reactions is challenging to investigate directly due to inefficient electron transfer between the protein and the electrode, and the reversible 2 e−/2 H+ reaction is thus not usually observed at the electrode surface.8−10 The reduction potential of a protein disulfide bond is normally determined indirectly from the position of its equilibrium with an appropriate reference redox couple whose standard reduction potential is already known: e.g., NADPH/NADP+ or glutathione/glutathione disulfide (GSH/ GSSG).9,11 Typically, the analyte protein is incubated in a series of “redox buffers” with defined applied reduction potentials (E′) established by varying the ratio of the reduced to the oxidized form of the reference redox couple, which is present in excess and at a constant total concentration. At equilibrium, the fraction of the dithiol or disulfide form of the analyte protein in each redox buffer is measured, e.g., by fluorescence spectroscopy or HPLC, and the reduction potential at which the

isulfide bonds are oxidative post-translational modifications (PTMs) that are increasingly recognized for their role as “redox switches” that control protein activity or function.1,2 Reversible disulfide bond formation is implicated in a diverse range of physiological processes, including visual phototransduction, muscle contraction, the innate immune response, coagulation, and regulation of blood pressure.3−7 It is imperative that these crucial redox modifications, often overlooked due to the long-held notion that disulfide bonds exist solely for stabilization of tertiary or quaternary protein structure, are fully characterized so that cellular processes can be understood definitively at the molecular level. The thermodynamic stability of a disulfide bond is described by its biochemical standard reduction potential, E°′. The biochemical standard reduction potential is a measure of the tendency of a chemical species to gain electrons under standard thermodynamic conditions and at neutral pH (298 K, 1 atm, and all reactants at a concentration of 1 M except H+ which is at pH 7.0). The reduction potential of a redox couple under nonstandard conditions, E′, is related to the standard potential by the Nernst equation, in which R is the gas constant, T is absolute temperature, n is the number of electrons transferred, F is the Faraday constant, and x and y are the stoichiometric coefficients: E′ = E°′ −

RT [electron donor]x ln nF [electron acceptor]y © 2013 American Chemical Society

Received: June 19, 2013 Accepted: August 29, 2013 Published: September 16, 2013

(1) 9164

dx.doi.org/10.1021/ac401850p | Anal. Chem. 2013, 85, 9164−9172

Analytical Chemistry

Article

distributions are highly complex, and FTICR MS provides the superior resolution and mass accuracy necessary to resolve and identify the many multiply charged product ions that result from fragmentation of an intact protein.22,23

reduced and oxidized forms are present at equal concentration is calculated; that reduction potential corresponds to the standard potential, E°′ (also known as the midpoint or equilibrium redox potential) according to eq 1. Determination of the reduction potential of a simple redox protein that contains only one redox-active site is straightforward because no interfering signals arise from the oxidation/ reduction of additional redox-sensitive cysteine residues. However, many proteins that are subject to redox regulation or are implicated in redox-signaling pathways potentially harbor multiple redox-sensitive cysteine residues: e.g., metallothionein, sarcoplasmic reticulum calcium pumps (SERCA), tumorsuppressor p53, and the transcription factors FOXO4 and Yap1p.1,12−15 To measure E°′ for each active site without resorting to site-directed mutagenesis, one must be able to distinguish between two or more redox-active sites and allow their oxidation states to be quantified independently. Recently, reduction potentials for the two active-site disulfides in human protein disulfide bond isomerase (PDI) were determined by a bottom-up MS approach.16 The reduced and oxidized cysteine residues were alkylated by iodoacetic acid and 13C2-bromoacetic acid (differentially labeling the reduced and oxidized cysteines with stable isotope-coded reagents negates any difference in the ionization efficiency or liquid chromatography retention time for the different peptide/ protein redox states) followed by proteolytic digestion and MALDI-TOF MS. The dithiol/disulfide ratio for each active site was determined from the light/heavy ratio of the corresponding tryptic peptides, and E°′ for each active site was subsequently determined. Measurement of protein reduction potentials by bottom-up MS requires that the redox-active sites be contained within detectable peptides. Furthermore, the combinatorial occurrence of the redox modifications cannot be investigated reliably by this method. In contrast, the top-down MS approach to protein characterization, in which proteolytic digestion is omitted and sequence/structure/PTMs are investigated directly by tandem mass spectrometry (MS/MS) of the intact protein, provides information across the entire length of the protein in every experiment.17−19 Thus, multiple redox-active sites within a single protein, together with any other PTMs, can be characterized simultaneously, and the combinatorial relationship of redox modifications is preserved. Sample processing is also reduced significantly because proteolysis and separation of the resulting peptides are not required. Thioredoxins are ubiquitous redox proteins involved in numerous cellular processes, including cell growth, regulation of gene expression, and apoptosis.20,21 They contain a characteristic CXXC motif that facilitates the reduction or oxidation of disulfide bonds via a thiol−disulfide exchange mechanism. The equilibrium redox potentials for thioredoxins from many lower species have been determined by various techniques and are well documented in the literature (see ref 1 for a comprehensive list of thioredoxin standard reduction potentials). However, the redox properties of mammalian thioredoxin 1 (Trx1), which contains additional redox-active sites, are not well characterized. Here, we determine E°′ for the CGPC active-site in E. coli Trx and human Trx1 by isotopecoded alkylation of cysteine residues coupled with top-down Fourier transform ion cyclotron resonance mass spectrometry (FTICR MS). Application of this method also yielded sitespecific determination of E°′ for a second redox modification in Trx1 for the first time. Top-down MS/MS product ion



EXPERIMENTAL METHODS Preparation of Reduced Thioredoxin. Recombinant E. coli Trx and recombinant 6xHis-tagged human Trx1 (both obtained from Sigma-Aldrich, St. Louis, MO, USA) were resuspended in 200 mM sodium phosphate buffer, pH 7.0, 2 mM EDTA, and 10 mM dithiothreitol (DTT), and incubated at room temperature for 1 h. Excess DTT was removed by buffer exchange by use of PD MiniTrap G-25 columns (GE Healthcare Life Sciences, Pittsburgh, PA, USA). Protein was eluted in 50 mM ammonium acetate, pH 5.5, and stored at −20 °C prior to redox titrations. Redox Titrations. Dihydrolipoic acid/lipoic acid (Lip[SH]2/Lip[S2]; E°′ = −290 mV)24 redox titrations were performed with both E. coli Trx and human Trx1, and GSH/ GSSG (E°′GSSG/2GSH = −240 mV)25 redox titrations were performed with human Trx1 only. Redox buffers with applied redox potentials (E′) that ranged from −222 to −349 mV or from −86 to −270 mV were prepared by varying the ratio of Lip(SH)2 to Lip(S2) or GSH to GSSG in 100 mM sodium phosphate buffer, pH 7.0, and 1 mM EDTA, while maintaining the total concentration of the redox couple at 10 mM (concentrations of Lip(SH)2/Lip(S2) and GSH/GSSG and the corresponding E′ values are listed in Supporting Information Table S1). All solvents were purged with argon, and redox buffers were prepared under a constant flow of argon. Thioredoxin (∼20 μM) was incubated at each E′ for 2 h at 25 °C. Each redox titration was repeated three times with fresh redox buffers. Fifteen min and 1 and 4 h incubations were also performed at E′Lip(S2)/Lip(SH)2 = −285 mV and E′GSSG/2GSH = −187 mV to ensure that equilibrium was reached (see Supporting Information Table S2). Reactions were quenched by the addition of an equal volume of ice cold 20% (w/v) trichloroacetic acid (TCA). Protein precipitates were collected by centrifugation and washed with 1% TCA. Human Trx1 precipitates that were not intended for cysteine alkylation were resuspended in 6 M guanidine hydrochloride and 100 mM sodium phosphate buffer, pH 7.0, and immediately desalted with C18 ZipTip pipet tips (Millipore, Billerica, MA, USA). Isotope-Coded Alkylation of Reduced and Oxidized Cysteine Residues. Free thiol groups (reduced cysteines) were alkylated by resuspending protein precipitates in 6 M guanidine hydrochloride and 100 mM sodium phosphate buffer, pH 7.0, supplemented with 1 mM natural isotopic abundance “light” N-ethylmaleimide (NEM). Oxidized cysteine residues were reduced by the addition of DTT to a final concentration of 5 mM. The newly free thiol groups were alkylated by the addition of a perdeuterated “heavy” form of NEM, N-ethylmaleimide-d5 (D5-NEM; Cambridge Isotope Laboratories, Andover, MA, USA) to a final concentration of 20 mM. Each alkylation and reduction step was performed at room temperature for 30 min. Samples were desalted with C18 ZipTip pipet tips prior to positive electrospray ionization (ESI) FTICR MS. ESI FTICR MS and Data Analysis. E. coli Trx eluates from C18 ZipTips were diluted 10 times with denaturing ESI solution (50% methanol, 50% water, and 0.1% formic acid), and human Trx1 eluates were diluted 5 times; the final E. coli and human 9165

dx.doi.org/10.1021/ac401850p | Anal. Chem. 2013, 85, 9164−9172

Analytical Chemistry

Article

protein concentrations were estimated to be ∼4 and ∼7 μM. Samples were microelectrosprayed into the mass spectrometer at 400 nL/min as previously described.26 E. coli Trx mass spectra were acquired with a modified 14.5 T linear ion trap (LTQ, Thermo Fisher Scientific, San Jose, CA, USA) FTICR mass spectrometer.27 For top-down collisioninduced dissociation (CID), 500,000 target ions of the charge state of interest were isolated in the LTQ, followed by fragmentation with a normalized collision energy of 20%. Ions resulting from three cycles of CID were accumulated in an adjacent storage octopole before transfer to the ICR cell (external ion accumulation increases the number of ions that are transferred to the ICR cell).28 FTICR data was acquired from m/z 400 to 2000 at a mass resolving power (m/Δm50%, in which Δm50% is the full width of a mass spectral peak at halfmaximum peak height) of 200,000 at m/z 400. 100 acquisitions were signal-averaged for each spectrum. Spectra were analyzed with Xcalibur 2.1 software (Thermo Fisher Scientific), and monoisotopic molecular masses were obtained with the “Xtract All” deconvolution tool. Human Trx1 mass spectra were acquired with a custom-built 9.4 T hybrid FTICR mass spectrometer.29 For CID of intact Trx1, ions of a specific charge state were isolated in a massselective quadrupole followed by acceleration into a nitrogencontaining octopole maintained at a typical pressure of 4.6 × 10−6 Torr. Product ions were subsequently transferred to the ICR cell for detection over an m/z range of 300 to 2000 with a time-domain transient length of 1.02 s. Each mass spectrum was the sum of 100 acquisitions. Spectra were analyzed with Predator Analysis software, and application of the THRASH algorithm yielded monoisotopic molecular masses for detected ions.30,31 For assignment of b and y product ions, the monoisotopic molecular mass lists were searched against theoretical mass lists for E. coli Trx or human Trx1 with ProSight PTM software (https://prosightptm.northwestern.edu).32 The mass measurement error (MME) tolerance was 10 ppm, and product ions with a mass measurement error greater than 2 standard deviations of the mean were omitted from the results. The assignments of diagnostic product ions that were important for localization of modifications were manually verified. The reduced/oxidized ratio for a redox-active Cys residue was based on the magnitudes of the most abundant isotopologues of the corresponding “light” and “heavy” product ions.

MS to maximize fragmentation efficiency, i.e., to observe many b and y product ions in high abundance.) The top-down FTICR mass spectrum and corresponding fragment map illustrating the b and y product ions resulting from CID of Trx equilibrated at −349 mV are shown in Figure 1. Of the 108 peptide bonds, 37 are cleaved and most of the

Figure 1. Top-down FTICR MS/MS for alkylated E. coli thioredoxin (Trx). Top: CID mass spectrum of Trx equilibrated at −349 mV followed by alkylation of the reduced and oxidized cysteine residues with NEM and (after disulfide reduction) D5-NEM. The most abundant product ions are annotated. Bottom: Corresponding fragment map illustrating the b and y product ions resulting from CID of the 11+ charge state of intact NEM-alkylated E. coli Trx. Crucially, the redox-active site residues Cys32 and Cys35 are contained within a series of b and y ions, so that the NEM/D5NEM ratio, and therefore the dithiol/disulfide ratio, can be determined. Each of the Cys-containing product ions was predominantly modified with natural isotope abundance NEM, indicating that most of E. coli Trx is in the dithiol form at the relatively reducing potential of −349 mV.

assigned product ions are from the N-terminal region of the protein. Note that extensive fragmentation across the entire length of the protein sequence is not necessary to quantify cysteine oxidation; all that is required is that one or more product ions encompass the cysteine residue(s) of interest. CID of dialkylated E. coli Trx resulted in 12 b ions and 12 y ions (not including multiple charge states that are typically observed in top-down MS/MS); all of the assigned product ions that contained one or both of the active-site Cys residues are listed (along with their mass measurement errors) in Supporting Information Table S3. Thus, a straightforward top-down CID experiment based on only one charge state of alkylated E. coli Trx results in numerous available product ions to quantify Cys32 and Cys35 oxidation. Two N-terminal product ions (b323+ and b393+) corresponding to three different applied redox potentials are shown in Figure 2 (top). An advantage of the present top-down MS/MS method is that the number of redox-sensitive thiols within a segment of the protein sequence is immediately apparent from the mass difference between the light and heavy versions of the Cys-containing product ions. For example, the 5 Da mass difference between light and heavy b323+ indicates that this product ion contains one redox-sensitive cysteine (Cys32), and the 10 Da mass difference between light and heavy b393+



RESULTS AND DISCUSSION Reduction Potential of the Active-Site Disulfide in E. coli Thioredoxin. We selected E. coli Trx as a model system because measurement of its standard redox potential is well documented in the literature.33−35 Recombinant protein was equilibrated at a range of solution redox potentials, from E′ = −349 to −222 mV, set by the lipoic acid redox couple. Reactions were quenched after 2 h by TCA precipitation, followed by alkylation of the reduced cysteine residues with “light” NEM, reduction of the active-site disulfide by DTT, and alkylation of the newly free thiol groups with “heavy” D5-NEM (resulting in a readily resolved 5 Da mass difference per NEM relative to D5-NEM alkylation). For each value of E′, intact differentially alkylated Trx was desalted and analyzed by ESI FTICR MS and the [M + 11H]11+ charge state precursor was fragmented by CID. (For each CID experiment described in this paper, ions with a combination of relatively high charge state and large signal-to-noise ratio (S/N) were chosen for MS/ 9166

dx.doi.org/10.1021/ac401850p | Anal. Chem. 2013, 85, 9164−9172

Analytical Chemistry

Article

−284 mV.33−35 Furthermore, the small standard deviations observed for dithiol/disulfide quantitation over three experiments demonstrate that the measurements are highly reproducible. Reduction Potential of the Active-Site Disulfide in Human Thioredoxin 1. We selected Trx1 to demonstrate that top-down FTICR MS permits the redox properties of multiple redox-active sites to be characterized unequivocally and simultaneously. In addition to its N-terminal CGPC motif, Trx1 contains three extra cysteine residues in the C-terminal half of its primary sequence, Cys61, 68, and 72, each of which has been reported to display redox activity.36 The redox titration and cysteine alkylation procedure described for E. coli Trx was repeated with recombinant human Trx1. The [M + 13H]13+ charge states of differentially alkylated Trx1 were subsequently fragmented by CID, and the resulting product ions were analyzed by FTICR MS. (As expected, we did not observe any notable differences in the light/heavy ratios of product ions resulting from CID of precursor ions with a different charge.) The top-down MS/MS and corresponding fragment map displaying the b and y product ions resulting from CID of Trx1 equilibrated at −349 mV are shown in Figure 3 (assigned product ions are listed in Supporting Information Table S4). Numerous b ions are observed, from b56 to b80, spanning the N-terminal active-site residues Cys31 and Cys34 (numbering of the Trx1 primary sequence begins at the Val residue immediately following the His-tag, which is designated by italics in Figure 3b). These product ions distinguish the N-

Figure 2. Determination of the standard biochemical reduction potential, E°′, for the active site disulfide of E. coli thioredoxin by topdown FTICR MS/MS. Top: Oxidation of Cys32 and Cys35 at increasing applied redox potentials, E′ (set by the Lip[SH]2/Lip[S2] redox couple), illustrated by the abundances of the “light” (NEMlabeled) and “heavy” (D5-NEM-labeled) N-terminal product ions b323+ and b393+ (which encompass one and both of the active-site Cys residues, hence the 5 and 10 Da mass differences). An average light/ heavy ratio (corresponding to the dithiol/disulfide ratio) was determined from multiple product ions for each E′. Bottom: Plot of the percentage of reduced Trx (dithiol form) versus E′. Error bars show the standard deviation for three experiments. E°′ is calculated as −280.1 ± 0.7 mV from nonlinear regression, in excellent agreement with literature values.33−35

indicates that this product ion encompasses two redox-sensitive cysteines (Cys32 and 35). At the relatively reducing potential of −318 mV, the light product ions are in excess, illustrating that most of E. coli Trx is in the dithiol form. As E′ increases (i.e., less negative potential), the magnitudes of the heavy product ions increase, reflecting the oxidation of the active-site dithiol to a disulfide. The light/heavy ratio for a product ion is a direct measure of the reduced/oxidized ratio for the cysteine residue(s) in that ion. Oxidation of the E. coli Trx N-terminal active site was quantified from the magnitudes of the most abundant isotopologues of four product ion (b363+, b373+, b383+, and b393+) light/heavy ratios. Ions were selected for quantitation on the basis of high S/N ratio and location in an uncrowded region of the mass spectrum. An average light/heavy ratio for each applied reduction potential was calculated from the specified b ions (variation in the light/heavy ratio among the four selected b ions in each CID spectrum was small, with standard deviation ranging from 0.4% to 2.3% for 39 spectra), and the entire experiment was performed in triplicate. The resulting plot of E′ versus the mean percentage of protein in the dithiol form is shown in Figure 2 (bottom). E°′ for the E. coli Trx disulfide bond was thereby determined to be −280.1 ± 0.7 mV, in excellent agreement with the literature values of −270 and

Figure 3. Top-down FTICR MS/MS of alkylated human thioredoxin 1 (Trx1). Top: CID mass spectrum of 6XHis-tagged Trx1 (the His tag is indicated in italics) equilibrated at −349 mV followed by labeling of the reduced and oxidized cysteine residues with NEM and D5-NEM. The most abundant product ions are annotated. Bottom: Corresponding fragment map illustrating the b and y product ions that resulted from CID of the 13+ charge state of intact alkylated Trx1. As for E. coli Trx, human Trx1 Cys residues are predominantly alkylated with “light” NEM at −349 mV. Note that oxidation of the N-terminal active-site residues Cys31 and Cys34 (Trx1 numbering begins after the His tag) can be distinguished from any redox activity of the C-terminal cysteines by the b and y product ions that are observed between Cys34 and Cys61. Furthermore, the product ions observed between Cys61, 68, and 72 permit any potential oxidation of each of these three residues to be quantified. Thus, a simple CID experiment enables the redox properties of multiple redox-active sites to be measured simultaneously. 9167

dx.doi.org/10.1021/ac401850p | Anal. Chem. 2013, 85, 9164−9172

Analytical Chemistry

Article

terminal active site from the C-terminal cysteines, enabling oxidation of the two sets of cysteines to be quantified independently without the need for proteolytic digestion or site-directed mutagenesis (to mutate any interfering Cys residues to a redox-inert amino acid such as serine). Furthermore, a succession of y ions, from y33 to y53, distinguishes Cys61, 68, and 72 from each other, permitting unambiguous localization of any C-terminal redox modifications and measurement of their reduction potentials. Thus, a simple CID experiment of one charge state of alkylated Trx1 yields a plethora of product ions from which oxidation of the N-terminal active site and C-terminal cysteines can be quantified. It is important to note that thioredoxins are small proteins, and quantifying oxidation of multiple cysteine residues in significantly larger, more complex redox proteins by the topdown MS/MS method we describe here will likely be more challenging. A limitation of top-down MS/MS for PTM characterization is protein size; dissociation efficiency typically decreases as protein mass increases, and it becomes progressively difficult to achieve fragmentation across the full length of the peptide backbone.17−19 However, a variety of dissociation techniques can be implemented, each with inherent advantages and disadvantages, to optimize fragmentation of the redox protein and observe the important Cys-containing diagnostic ions. Furthermore, limitations of top-down MS/ MS can be overcome with the “middle-down” approach, whereby the analyte protein is cleaved into large peptides (e.g., at low abundance residues by an appropriate protease or chemical reagent) that are more amendable to MS/MS than the intact protein, yet contain multiple PTM sites and provide extensive sequence information.18,19 An N-terminal product ion of Trx1 containing the active-site residues Cys31 and 34 (b758+) and a C-terminal product ion containing Cys61, 68, and 72 (y484+) are shown for three different applied reduction potentials in Figure 4 (top). Oxidation of the N-terminal active site is made apparent by the increasing magnitudes of the heavy versions of the b ions that contain Cys31 and Cys34. In contrast, the y ions that contain Cys61, 68, and 72 are present solely in the light form, illustrating that the C-terminal thiols are resistant to oxidation by Lip(SH)2/Lip(S2). Despite the lack of oxidation of Cys61, 68, and 72 under these redox conditions, it is clear that gasphase fragmentation of intact Trx1 enables any potential oxidation of the C-terminal thiols to be quantified without interference from the redox activity of the N-terminal activesite. Oxidation of the N-terminal active-site was quantified from the light/heavy product ions b728+, b738+, b748+, b757+, b758+, and b777+. The experiment was performed in triplicate, and the standard reduction potential for the active-site disulfide in Trx1 was determined as −280.8 ± 0.6 mV (see Figure 4 (bottom)). That value is similar to the E°′ that we determined for E. coli Trx but in notable contrast to E°′ reported for the Trx1 activesite disulfide by Watson et al. based on its equilibrium with the glutathione redox couple: −230 mV.37 We also performed a GSH/GSSG redox titration with Trx1 (to investigate Cys61, 68, and 72 oxidation; discussed below), and on the basis of the average light/heavy ratio of b758+ at −240 mV (0.0874), we estimate E°′ for the N-terminal active site disulfide to be −271 mV (calculated from the Nernst equation), in good agreement with the value we obtained from the Lip(SH)2/Lip(S2) redox titration (the ∼10 mV discrepancy is probably due to a slightly

Figure 4. Determination of E°′ for the N-terminal active site disulfide in human Trx1 by top-down FTICR MS/MS. Top: Relative quantitation of Cys31 and Cys34 oxidation at increasing applied redox potential (set by Lip[SH]2/Lip[S2]) from the light and heavy product ion, b758+. The C-terminal product ion, y484+, encompasses Cys61, 68, and 72; the absence of heavy y484+ reveals that oxidation of the C-terminal Cys residues is not favorable under the listed redox conditions. Bottom: Plot of the percentage of Trx1 Cys31 and Cys34 in the dithiol form versus E′. Error bars show the standard deviation for three experiments. E°′ was calculated as −280.8 ± 0.6 mV from nonlinear regression and differs from the previously reported −230 mV.37

inaccurate value of E°′ for the glutathione or lipoic acid redox couples) but still substantially lower than the literature value of −230 mV. Substantial variation in the redox properties of members of the thioredoxin superfamily exists; an important determinant of E°′ is the identity of the two amino acids between the Cys residues in the CXXC active-site motif.38 Standard redox potentials in the region of −230 mV are usually associated with glutaredoxins (GrX), which possess a characteristic CXYC motif, whereas thioredoxins, which possess a conserved CGPC motif, tend to be stronger reductants with redox potentials close to −270 mV.1,39 However, the interjacent dipeptide is not the only determinant of redox potential; peripheral residues can profoundly influence E°′. For example, there is a 35 mV difference in E°′ for E. coli GrX1 and GrX3 even though they possess the same CPYC active-site motif within a similar three-dimensional fold.33,40 Furthermore, mutation of both active-site CGHC motifs in human PDI to thioredoxin CGPC motifs results in redox potentials of ∼−230 mV.16 Thus, despite the presence of the typical CGPC motif, it is possible that human Trx1 is a less potent reductant than its homologues from other species; however, a Grx-like E°′ for human Trx1 is not supported by our data. Watson et al. utilized a “redox Western blot” approach combined with MS analysis of proteolytic peptides to measure E°′ for the Cys31−Cys34 disulfide bond in human Trx1.37 Briefly, recombinant protein was equilibrated in GSH/GSSG 9168

dx.doi.org/10.1021/ac401850p | Anal. Chem. 2013, 85, 9164−9172

Analytical Chemistry

Article

Figure 5. Identification of a C-terminal redox-active site in human Trx1 by top-down FTICR MS/MS. Left: Mass spectra of unalkylated Trx1 equilibrated at GSH/GSSG solution redox potentials of −240 and −155 mV; the 12+ charge states are shown. At −240 mV, the molecular mass of the dominant species, calculated from the m/z value of the most abundant isotopologue (1148.2340), is consistent with Trx1 containing one intramolecular disulfide bond (Trx1−S2; empirical formula: C610H936N164O184S8; calculated most abundant molecular mass =13,766.689 Da; mass measurement error [MME] = 2.3 ppm). At −155 mV, a mass increase of 305 Da is observed, consistent with formation of an intermolecular disulfide bond between Trx1 and a molecule of glutathione (GSH Mw = 307 Da). Upper right: Fragment map illustrating the b and y product ions that resulted from CID of the 14+ charge state of intact unalkylated Trx1 equilibrated at −155 mV. Bottom right: N-terminal Cys31- and Cys34containing product ion, b738+, and C-terminal Cys61-, 68-, and 72-containing product ion, y464+, corresponding to E′ = −155 mV. Comparison of the experimental isotopologue distributions with the calculated isotopologue distribution for b738+ containing an intramolecular disulfide bond (blue circles; empirical formula: [C362H552N102O101S4]8+; MME = 0.89 ppm) and y464+ modified with a GSH adduct (red circles; empirical formula: [C233H370N60O80S5]4+; MME = 0.44 ppm) confirms that the intramolecular disulfide is located at the N-terminal active site and that either Cys61, 68, or 72 is S-glutathionylated. The product ions y404+, y405+, y414+, y424+, and y434+ localize the GSH adduct even further: to Cys68 or Cys72. CID FTICR MS/MS of NEM/D5-NEM-alkylated Trx1 localizes the GSH adduct to Cys72 (see Supporting Information Figure S1).

cycle and subcellular location.42 The GSH/GSSG redox couple is 50 mV more oxidizing than Lip(SH)2/Lip(S2) and is reported to induce oxidation of the C-terminal Cys residues in Trx1.37,43 We performed a GSH/GSSG redox titration with Trx1, from E′ = −270 to −86 mV, to investigate oxidation of Cys61, 68, and 72. Cysteine alkylation was omitted initially so that any GSH/GSSG-induced redox modifications of Trx1 could be identified. The mass spectra of Trx1 equilibrated at −240 and −155 mV are shown in Figure 5 (left). The calculated molecular mass of 6XHis-tagged Trx1 is 13,768.7 Da (taken from the most abundant isotopologue for the empirical formula C610H938N164O184S8). At −240 mV, the measured mass of the dominant species is 13,766.7 Da, consistent with Trx1 that contains one intramolecular disulfide bond, i.e., Trx1−2H. At −155 mV, a mass shift of +305 Da is observed, consistent with the formation of an intermolecular disulfide bond between Trx1 and a molecule of glutathione (GSH, 307 Da). Trx1 dimers are not observed in the mass spectra, suggesting that intermolecular disulfides between molecules of Trx1 are not formed under these redox conditions. Indeed, Trx1 homodimers are known to form at high protein concentration.44 The [M + 14H]14+ charge states of Trx1 equilibrated at −240 and −155 mV, that did not undergo cysteine alkylation, were fragmented by CID. On the basis of the E°′ that we determined for the N-terminal active-site of Trx1 (−281 mV), we expected that most Trx1 molecules would contain an intramolecular disulfide at −240 mV. Indeed, top-down MS revealed a disulfide bond between Cys31 and Cys34 (made apparent by a −2 Da mass shift for b ions that encompassed these two residues), whereas the C-terminal Cys residues were in the thiol form at −240 mV (data not shown). The fragment map displaying the b and y product ions resulting from CID of Trx1 equilibrated at −155 mV is shown in Figure 5 (upper right),

buffer; free thiol groups were labeled with iodoacetic acid, and the reduced and oxidized forms of Trx1 were separated by native polyacrylamide gel electrophoresis. Oxidation of Cys31 and Cys34 was quantified from the resulting immunoblot by densitometry. Prior to redox Western blotting, the redox states of Trx1 in each of the separated gel bands were analyzed by MALDI-TOF MS following in-gel proteolytic digestion with trypsin. A major drawback of this bottom-up MS approach to PTM characterization is that, due to large variations in ionization efficiency, the observed peptides often do not reflect the true identities and composition of the proteoforms present in the sample. (The term “proteoform” designates all of the different molecular forms in which the protein product of a single gene can be found, i.e., different isoforms, polymorphisms, mutations, post-translational modifications, etc.41) Thus, the surprisingly high Trx1 redox potential reported by Watson et al. may have resulted from an inaccurate or incomplete characterization of the redox states of Trx1 in the separated gel bands. Another source of error leading to the high literature value of the Cys31−Cys34 disulfide redox potential may have been the use of GSH/GSSG redox buffer. Because of the substantial disparity in GSH and GSSG concentrations that is required to achieve solution redox potentials more negative than ∼−270 mV (giving rise to potentially large errors in the assumed E′), an alternative redox couple should be used to access more reducing potentials.9 For the most accurate determination of equilibrium standard redox potentials, the reference redox couple should have a value of E°′ as close to that of the analyte as possible. Identification of a Second Redox-Active Site in Human Thioredoxin 1. GSH/GSSG is a dominant cellular redox buffer, responsible for maintaining redox potentials from ∼−300 to −140 mV depending on the stage of the cell life 9169

dx.doi.org/10.1021/ac401850p | Anal. Chem. 2013, 85, 9164−9172

Analytical Chemistry

Article

and N- and C-terminal product ions that encompass Cys31 and Cys34 and Cys61, 68, and 72 (b738+ and y464+) are shown in Figure 5 (lower right) (all of the assigned product ions are listed in Supporting Information Table S5). The calculated isotopologue distribution of b738+ with an intramolecular disulfide bond and the calculated isotopologue distribution of y464+ with an S-glutathione adduct are consistent with the experimental isotopologue distributions (MME = 0.89 and 0.44 ppm), confirming that the N-terminal active site is in the oxidized, intramolecular disulfide form (as expected) and Cys61, 68, or 72 is glutathionylated. The product ions y404+, y405+, y414+, y424+, and y434+ localize the GSH modification even further: to Cys68 or Cys72. However, a diagnostic product ion that distinguishes Cys68 from Cys72 was not observed; therefore, we cannot identify the site of Trx1 S-glutathionylation unambiguously from the CID mass spectrum of Trx1 equilibrated at −155 mV. No product ions were assigned that are consistent with intramolecular disulfide formation between Cys61, 68, or 72 or S-glutathionylation of the N-terminal active site. Reduction Potential of the Second Redox-Active Site in Human Thioredoxin 1. To determine E°′ for the Trx1− GSH mixed disulfide on Cys68 or Cys72, we differentially alkylated the reduced and oxidized Cys residues following equilibration with GSH/GSSG and performed top-down MS/ MS (as described above for the N-terminal active-site disulfide of Trx1). The N-terminal product ion b758+ and C-terminal product ion y484+ are shown for three different GSH/GSSG potentials in Figure 6 (top). It is clear from the absence of light b758+ that the N-terminal active-site is ∼100% oxidized in GSH/ GSSG redox buffers with E′ ≥ −218 mV, as expected from E°′Cys31−Cys34 = −281 mV. It is also clear from the 5 Da difference between the light and heavy analogues of y484+ that one Cys residue is oxidized in the C-terminal half of Txr1; topdown MS/MS revealed that redox modification to be Sglutathionylation as described above. CID of glutathionylated unalkylated Trx1 did not yield a product ion between Cys68 and Cys72, precluding absolute localization of the GSH adduct. However, CID of NEM/D5-NEM-alkylated Trx1 results in slightly increased sequence coverage between Cys61 and Cys72 (7 peptide bonds are cleaved compared to 4 for unalkylated Trx1; see Figures 3 (bottom) and 5 (upper right)), and a product ion that distinguishes Cys68 from Cys72 and encompasses Cys72 only (y333+) is observed. That product ion is labeled by natural isotope abundance NEM at the most negative GSH/GSSG applied redox potentials and by D5-NEM at the most oxidizing potentials, indicating that Cys72 is the site of S-glutathionylation in Trx1 (see Supporting Information Figure S1). At E′ ≥ −123 mV, we also detected C-terminal Cys-containing y ions consistent with the presence of an additional D5-NEM adduct (data not shown), indicating that Cys61 and/or Cys68 may be oxidized under relatively strong oxidizing conditions. However, the abundance of those ions was too low to enable characterization of this third Trx1 redox modification. S-Glutathionylation of Cys72 in Trx1 has been observed previously, induced by GSSG or S-nitrosoglutathione (GSNO) treatment, and that redox modification was shown to inhibit its disulfide-reductase activity.43 An intramolecular disulfide bond between Cys61 and Cys68, induced by the strong thioloxidizing agent diamide, has also been shown to attenuate the disulfide-reductase activity of Trx1.45 On the basis of their bottom-up MS/MS data and redox Western blots, Watson et al.

Figure 6. Determination of E°′ for the C-terminal redox-active site of human Trx1 by top-down FTICR MS/MS. Top: Relative quantitation of Cys72 S-glutathionylation at various E′ (set by GSH/GSSG) from the light and heavy product ion, y484+. (The contribution of the overlapping light y484+ isotopologue peaks to the magnitude of the heavy y484+ isotopologue peaks was taken into account in determining the thiol/disulfide ratio.) Bottom: Plot of the percentage of reduced Cterminal active site (thiol form) versus E′. Error bars show the standard deviation for three experiments. E°′ was calculated as −185.7 ± 0.7 mV from nonlinear regression. Interestingly, S-glutathionylation of Cys72 occurs over a range of E′ that corresponds closely with the intracellular GSH/GSSG redox potentials at different stages of the cell life cycle: 1, Proliferation (−280 to −230 mV); 2, Differentiation (−230 to −190 mV); 3, Apoptosis (−190 to −150 mV); 4, Necrosis (>−150 mV).42

inferred that a Cys61−Cys68 intramolecular disulfide formed when Trx1 was incubated in GSH/GSSG buffers with solution potentials as low as −270 mV.37 However, we detected only a minimal amount of Cys61 and/or Cys68 oxidation at GSH/ GSSG potentials ≥−123 mV; the Cys72-GSH mixed disulfide was by far the most dominant C-terminal redox modification in Trx1 at physiological GSH/GSSG redox potentials. It remains to be determined if the Cys61−Cys68 disulfide is biologically relevant. Nonetheless, our data suggests that a potent intracellular oxidizing molecule such as a reactive oxygen or reactive nitrogen species would be required to induce this redox modification in vivo. Interestingly, glutathionylation of Cys72 is considered to serve as a reversible switch that regulates the function of Trx1 by inhibiting thioredoxin reductase (TrxR)mediated restoration of the CGPC active site to its reduced form.43 Oxidation of Cys72 was quantified from the product ions y464+, y484+, y495+, y515+, and y525+ (contributions of the overlapping light isotopologue peaks to the magnitudes of the heavy isotopologue peaks were accounted for in determining the thiol/disulfide ratios). The experiment was performed in triplicate, and E°′ for Cys72 was determined as −185.7 ± 0.7 mV (see Figure 6 (bottom)), corresponding closely to the 9170

dx.doi.org/10.1021/ac401850p | Anal. Chem. 2013, 85, 9164−9172

Analytical Chemistry



intracellular GSH/GSSG potential for a cell entering apoptosis: approximately −190 mV.42 Interestingly, oxidation of the CGPC active site in Trx1 is known to be involved in apoptosis via disruption of the Trx1−ASK1 interaction (apoptosis signal regulating kinase 1; involved in tumor necrosis factor [TNF] αinduced apoptosis).46 Thus, a shift in the intracellular redox potential to ∼−190 mV accompanied by a substantial increase in the extent of Cys72 oxidation may represent a mechanism for inhibition of TrxR-mediated reduction of the Trx1 catalytic site and the subsequent onset of pro-apoptotic signaling, induced by dissociation of the Trx1−ASK1 complex. The E°′ that we determined for the Cys72−GSH mixed disulfide also indicates that oxidation of Cys72 is not favorable in proliferating cells (E′ = ∼−280 to −230 mV)42 in accord with knowledge that the reducing activity of the CGPC active site is required for Trx1-induced stimulation of cell growth.47

AUTHOR INFORMATION

Corresponding Author

*Phone: +1 850 644 0529. Fax: +1 850 644 1366. E-mail: [email protected]. Present Address §

J.S.: Cardiovascular Division, King’s College London, The Rayne Institute, St. Thomas’ Hospital, London, SE1 7EH, United Kingdom. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS The authors thank Santosh G. Valeja and Nicolas L. Young for valuable discussions and assistance with operation of the mass spectrometers. This work was supported by NSF DMR-1157490 and the State of Florida.





CONCLUSION We demonstrate for the first time that top-down FTICR MS coupled with differential isotope-coded cysteine alkylation is an effective method for reliable and accurate determination of the equilibrium reduction potentials for protein disulfide bonds. A powerful advantage of this top-down MS approach for measurement of E°′ is the ability to characterize multiple redox active sites within a single protein simultaneously. In addition, the identity and site-specific location of each redox modification is easily revealed; hence, a wealth of information regarding a protein’s redox properties can be garnered from a single set of top-down MS experiments. Gas-phase fragmentation of intact NEM/D5-NEM-alkylated E. coli Trx equilibrated with the lipoic acid redox couple yielded E°′ for the CGPC active-site disulfide as −280.1 ± 0.7 mV, in excellent agreement with the literature values. Application of the present method to human Trx1 resulted in a wealth of product ions that distinguished the N-terminal active-site from the three C-terminal cysteine residues; a simple MS/MS experiment therefore enabled oxidation of the two sets of cysteine residues to be quantified independently. E°′ for the CGPC catalytic site in human Trx1 was thus determined as −280.8 ± 0.6 mV on the basis of its equilibrium with the lipoic acid redox couple and calculated as −271 mV on the basis of its equilibrium with the glutathione redox couple. These values are notably more reducing than the literature value of −230 mV.37 E°′ for the Trx1 catalytic disulfide should be verified by other investigators to prevent confusion and the erroneous prediction of, e.g., the redox state of endogenous intracellular Trx1. In addition, top-down FTICR MS revealed Cys72 in Trx1 to be S-glutathionylated at physiologically relevant GSH/GSSG redox potentials. E°′ for this redox modification was determined as −185.7 ± 0.7 mV, which is, i.e., intriguingly close to the intracellular GSH/GSSG potential for a cell entering apoptosis. Our data, together with previous findings, strongly indicate that glutathionylation of Cys72 in mammalian Trx1 is an important component of a pro-apoptotic redoxsignaling pathway.



Article

REFERENCES

(1) Wouters, M. A.; Fan, S. W.; Haworth, N. L. Antioxid. Redox Signaling 2010, 12, 53−91. (2) Ryu, S. E. J. Biochem. 2012, 151, 579−588. (3) Liu, W.; Wen, W. Y.; Wei, Z. Y.; Yu, J.; Ye, F.; Liu, C. H.; Hardie, R. C.; Zhang, M. J. Cell 2011, 145, 1088−1101. (4) Jackson, M. J. IUBMB Life 2008, 60, 497−501. (5) Schroeder, B. O.; Wu, Z. H.; Nuding, S.; Groscurth, S.; Marcinowski, M.; Beisner, J.; Buchner, J.; Schaller, M.; Stange, E. F.; Wehkamp, J. Nature 2011, 469, 419−423. (6) Ahamed, J.; Versteeg, H. H.; Kerver, M.; Chen, V. M.; Mueller, B. M.; Hogg, P. J.; Ruf, W. Proc. Natl. Acad. Sci. U.S.A. 2006, 103, 13932− 13937. (7) Zhou, A. W.; Carrell, R. W.; Murphy, M. P.; Wei, Z. Q.; Yan, Y. H.; Stanley, P. L. D.; Stein, P. E.; Pipkin, F. B.; Read, R. J. Nature 2010, 468, 108−111. (8) Chobot, S. E.; Hernandez, H. H.; Drennan, C. L.; Elliott, S. J. Angew. Chem., Int. Ed. 2007, 46, 4145−4147. (9) Gilbert, H. F. Methods Enzymology 1995, 251, 8−28. (10) Becker, D. F. In Redox Biochemistry; Banerjee, R., Ed.; John Wiley & Sons: Hoboken, 2008; pp 247−251. (11) Rabenstein, D. L. In Oxidative Folding of Peptides and Proteins; Buchner, J., Moroder, L., Eds.; The Royal Society of Chemistry: Cambridge, 2009; pp 220−235. (12) Maret, W.; Vallee, B. L. Proc. Natl. Acad. Sci. U.S.A. 1998, 95, 3478−3482. (13) Scotcher, J.; Clarke, D. J.; Mackay, C. L.; Hupp, T.; Sadler, P. J.; Langridge-Smith, P. R. R. Chem. Sci. 2013, 4, 1257−1269. (14) Putker, M.; Madl, T.; Vos, H. R.; de Ruiter, H.; Visscher, M.; van den Berg, M. C. W.; Kaplan, M.; Korswagen, H. C.; Boelens, R.; Vermeulen, M.; Burgering, B. M. T.; Dansen, T. B. Mol. Cell 2013, 49, 730−742. (15) Brandes, N.; Schmitt, S.; Jakob, U. Antioxid. Redox Signaling 2009, 11, 997−1014. (16) Chambers, J. E.; Tavender, T. J.; Oka, O. B. V.; Warwood, S.; Knight, D.; Bulleid, N. J. J. Biol. Chem. 2010, 285, 29200−29207. (17) Reid, G. E.; McLuckey, S. A. J. Mass Spectrom. 2002, 37, 663− 675. (18) Han, X. M.; Aslanian, A.; Yates, J. R. Curr. Opin. Chem. Biol. 2008, 12, 483−490. (19) Cui, W. D.; Rohrs, H. W.; Gross, M. L. Analyst 2011, 136, 3854−3864. (20) Arner, E. S. J.; Holmgren, A. Eur. J. Biochem. 2000, 267, 6102− 6109. (21) Powis, G.; Montfort, W. R. Annu. Rev. Biophys. Biomol. Struct. 2001, 30, 421−455. (22) Marshall, A. G.; Hendrickson, C. L.; Jackson, G. S. Mass Spectrom. Rev. 1998, 17, 1−35.

ASSOCIATED CONTENT

* Supporting Information S

Additional information as noted in the text. This material is available free of charge via the Internet at http://pubs.acs.org. 9171

dx.doi.org/10.1021/ac401850p | Anal. Chem. 2013, 85, 9164−9172

Analytical Chemistry

Article

(23) Meng, F. Y.; Forbes, A. J.; Miller, L. M.; Kelleher, N. L. Mass Spectrom. Rev. 2005, 24, 126−134. (24) Loach, P. A. In Handbook of Biochemistry and Molecular Biology; Fasman, G. D., Ed.; CRC Press: Cleveland, 1976; pp 122−130. (25) Rost, J.; Rapoport, S. Nature 1964, 201, 185. (26) Emmett, M. R.; White, F. M.; Hendrickson, C. L.; Shi, S. D. H.; Marshall, A. G. J. Am. Soc. Mass Spectrom. 1998, 9, 333−340. (27) Schaub, T. M.; Hendrickson, C. L.; Horning, S.; Quinn, J. P.; Senko, M. W.; Marshall, A. G. Anal. Chem. 2008, 80, 3985−3990. (28) Senko, M. W.; Hendrickson, C. L.; Emmett, M. R.; Shi, S. D. H.; Marshall, A. G. J. Am. Soc. Mass Spectrom. 1997, 8, 970−976. (29) Kaiser, N. K.; Quinn, J. P.; Blakney, G. T.; Hendrickson, C. L.; Marshall, A. G. J. Am. Soc. Mass Spectrom. 2011, 22, 1343−1351. (30) Blakney, G. T.; Hendrickson, C. L.; Marshall, A. G. Int. J. Mass Spectrom. 2011, 306, 246−252. (31) Horn, D. M.; Zubarev, R. A.; McLafferty, F. W. J. Am. Soc. Mass Spectrom. 2000, 11, 320−332. (32) LeDuc, R. D.; Taylor, G. K.; Kim, Y. B.; Januszyk, T. E.; Bynum, L. H.; Sola, J. V.; Garavelli, J. S.; Kelleher, N. L. Nucleic Acids Res. 2004, 32, W340−W345. (33) Aslund, F.; Berndt, K. D.; Holmgren, A. J. Biol. Chem. 1997, 272, 30780−30786. (34) Mossner, E.; Huber-Wunderlich, M.; Glockshuber, R. Protein Sci. 1998, 7, 1233−1244. (35) Cheng, Z. Y.; Arscott, L. D.; Ballou, D. P.; Williams, C. H. Biochemistry 2007, 46, 7875−7885. (36) Wu, C. G.; Parrott, A. M.; Fu, C. X.; Liu, T.; Marino, S. M.; Gladyshev, V. N.; Jain, M. R.; Baykal, A. T.; Li, Q.; Oka, S.; Sadoshima, J.; Beuve, A.; Simmons, W. J.; Li, H. Antioxid. Redox Signaling 2011, 15, 2565−2604. (37) Watson, W. H.; Pohl, J.; Montfort, W. R.; Stuchlik, O.; Reed, M. S.; Powis, G.; Jones, D. P. J. Biol. Chem. 2003, 278, 33408−33415. (38) Huber-Wunderlich, M.; Glockshuber, R. Folding Des. 1998, 3, 161−171. (39) Cheng, Z. Y.; Zhang, J. F.; Ballou, D. P.; Williams, C. H. Chem. Rev. 2011, 111, 5768−5783. (40) Foloppe, N.; Nilsson, L. Structure 2004, 12, 289−300. (41) Smith, L. M.; Kelleher, N. L. Nat. Methods 2013, 10, 186−187. (42) Jones, D. P. J. Intern. Med. 2010, 268, 432−448. (43) Casagrande, S.; Bonetto, V.; Fratelli, M.; Gianazza, E.; Eberini, I.; Massignan, T.; Salmona, M.; Chang, G.; Holmgren, A.; Ghezzi, P. Proc. Natl. Acad. Sci. U.S.A. 2002, 99, 9745−9749. (44) Weichsel, A.; Gasdaska, J. R.; Powis, G.; Montfort, W. R. Structure 1996, 4, 735−751. (45) Hashemy, S. I.; Holmgren, A. J. Biol. Chem. 2008, 283, 21890− 21898. (46) Lu, J.; Holmgren, A. Antioxid. Redox Signaling 2012, 17, 1738− 1747. (47) Oblong, J. E.; Berggren, M.; Gasdaska, P. Y.; Powis, G. J. Biol. Chem. 1994, 269, 11714−11720.

9172

dx.doi.org/10.1021/ac401850p | Anal. Chem. 2013, 85, 9164−9172