Peroxiredoxin-6 (PRDX6) is an unusual member of the peroxiredoxin family of antioxidant enzymes that has only one evolutionarily conserved cysteine. It reduces oxidized lipids and reactive oxygen species (ROS) by oxidation of the active-site cysteine (Cys47) to a sulfenic acid, but the mechanism for conversion back to a thiol is not completely understood. Moreover, it has phospholipase A2 (PLA2) activity in addition to its peroxidase activity. Interestingly, some biochemical data are inconsistent with a known high-resolution crystal structure of the catalytic intermediate of the protein, and biophysical data indicate that the protein undergoes conformational changes that affect enzyme activity. In order to further elucidate the solution structure of this important enzyme, we used chemical cross-linking coupled with high-resolution MS (CX–MS), with an emphasis on zero-length cross-links. Distance constraints from high confidence cross-links were used in homology modelling experiments to determine a solution structure of the reduced form of the protein. This structure was further evaluated using chemical cross-links produced by several homo-bifunctional amine-reactive cross-linking reagents, which helped to confirm the solution structure. The results show that several regions of the reduced version of human PRDX6 are in a substantially different conformation from that shown for the crystal structure of the peroxidase catalytic intermediate. The differences between these two structures are likely to reflect catalysis-related conformational changes. These studies also demonstrate that CX–MS using zero-length cross-linking is a powerful strategy for probing protein conformational changes that is complementary to alternative methods such as crystallographic, NMR and biophysical studies.

INTRODUCTION

Regulation of oxidative stress is an important problem in cellular homoeostasis. The presence of reactive oxygen species (ROS) and their by-products can have multiple deleterious effects on cells. The lung is particularly vulnerable to these effects, due to its continual exposure to oxidants via ambient air as well as its extensive capillary networks [1]. Oxidative damage has been notably associated with various disease states in the lung, ranging from acute lung injury (ALI) to chronic obstructive pulmonary disease (COPD) [2,3]. In order to protect against oxidative injury, eukaryotic cells express a number of antioxidative stress proteins. Among these is a family of proteins known as peroxiredoxins. Peroxiredoxins differ from others involved in similar functions by working in conjunction with thiol-containing electron donor molecules, as opposed to using redox cofactors or prosthetic groups [4,5]. Most peroxiredoxins employ two conserved cysteine residues in a disulfide bond as their electron donor group in the presence of a reductant, typically thioredoxin. This is known as a 2-Cys mechanism [4,5].

Peroxiredoxin-6 (PRDX6) is a unique case in the peroxiredoxin family. It is a homodimeric enzyme found in mammals, particularly mammalian lungs [6] that features at least two distinct enzymatic activities, the reducing property common to all peroxiredoxins, as well as phospholipase A2 (PLA2) via a conserved catalytic triad (His26, Ser32, Asp140) [7,8]. It is also the only peroxiredoxin with the reported ability to reduce phospholipid hydroperoxides [9]. Moreover, PRDX6 does not feature the two canonical cysteine residues that are the hallmark of most members of the peroxiredoxin family. Only one cysteine, Cys47, is conserved across species and its peroxidase activity is therefore described as a 1-Cys mechanism that involves oxidation of Cys47 to a sulfenic acid as a peroxidase catalytic intermediate. The enzymatic cycle is completed by reduction in the active site back to a thiol by glutathione in conjunction with GST [10,11]. PRDX6’s PLA2 activity is regulated by phosphorylation of Thr177, indicating that activation of the PLA2 activity involves a conformational change [12,13]. Both the PLA2 as well as the phospholipid hydroperoxide activities require conformational specificity in the binding of the enzyme to the phospholipid substrate [14].

A high-resolution crystal structure for human PRDX6 has been reported but, in order to obtain this structure, it was apparently necessary to mutate a non-conserved cysteine (Cys91) and oxidize the active-site cysteine to the sulfenic form [15]. Perhaps because the active site is in the catalytic intermediate state, the crystal structure does not appear to accurately reflect the solution structure of the reduced form of the protein. For example, a previous study showed that Thr177, which is totally buried in the crystal structure, can be phosphorylated, indicating solvent exposure under conditions that should include the conformation in solution of the protein with Cys47 in the reduced state [12,13]. This phosphorylation induces a further conformational change that increases PLA2 activity and affinity for liposomes. Other biochemical data indicate that the protein undergoes conformational changes that can modulate enzyme activity, such as the conformational change that results in protein activation when binding with GST [11] and the one that occurs upon binding to the phospholipid head group [8]. Considering the substantial conformational flexibility of the protein and importance of conformational changes on both enzyme activities, elucidation of the solution structure of the reduced form of the enzyme is needed.

We combined chemical cross-linking and MS (CX–MS) in order to determine distance restraints for testing and refinement of homology models of PRDX6. CX–MS has been shown to be a powerful method to distinguish between alternative predicted conformations of a protein and to refine predicted structures to highly accurate levels when using homology modelling [16]. The utility of this approach has been demonstrated in many studies including recent analyses of diverse macromolecular systems [1719].

Our group has recently optimized the use of cross-linkers with no spacer arms between the cross-link reactive sites (also known as zero-length cross-linkers) because such cross-links define the most stringent distance constraints [2023] Typically, these reactions use the zero-length cross-linking reagent 1-ethyl-3-(-3-dimethylaminopropyl) carbodi-imide hydrochloride (EDC), which is used in conjunction with sulfo-N-hydroxysuccinimide (sulfo-NHS). Cross-links are formed between carboxylic acids (aspartic acid and glutamic acid side chains and the protein C-terminus) and primary amines (lysine side chains and the protein N-terminus). Because no extra atoms are contributed by the cross- linking reagents, the atoms that contribute to the cross-link need to be within salt bridge distance [24]. Zero-length cross-linked peptides can be especially difficult to identify because incorporation of stable isotope tags into the cross-link, a strategy frequently used with longer cross-linking reagents, is not feasible.

In the present study, we probed the solution structure of recombinant PRDX6 using CX–MS. A substantial number of high-confidence zero-length cross-links clearly indicated differences between the solution structure of the enzyme in its reduced state [15] and the previously reported crystal structure of the catalytic intermediate (PDB code 1PRX). The distance constraints defined by these cross-links were used in homology modelling experiments using the MODELLER program [25] to determine the solution structure of the reduced form of PRDX6. Comparison of the solution structure of the reduced form of PRDX6 determined using CX–MS and the crystal structure of the catalytic intermediate provides insights into conformational changes associated with the catalytic activity of this 1-Cys peroxiredoxin.

EXPERIMENTAL

Expression and purification of human PRDX6

A plasmid expressing human Prdx6 that had been codon-optimized for Escherichia coli was purchased from DNA 2.0. The codon-optimized sequence of human Prdx6 was inserted into the plasmid pJExpress414 between an NdeI site and an XhoI site. The protein was then expressed as described previously [8,10] and purified using ion-exchange chromatography, resulting in a homogeneous product as determined by SDS/PAGE [6,8,10].

Sedimentation equilibrium

The oligomeric state of human PRDX6 was determined using sedimentation equilibrium in an Optima XLI analytical ultracentrifuge. Immediately before the sedimentation experiment, samples were gel filtered in 20 mM Tris/HCl, 130 mM sodium chloride, 0.1 mM EDTA and 5% glycerol, pH 7.4, using two Superdex75 (GE Healthcare) columns connected in series and maintained at 4°C to remove any aggregates. For each experiment, a minimum of three different initial sample concentrations and two rotor speeds were analysed. In some cases samples were analysed at both 4°C and 30°C. Sample volumes of 110 μl were loaded into Epon double sector centrepieces. A water blank scan was acquired immediately prior to sample analysis and was used to correct for window distortion in the fringe displacement data [26]. Fringe displacement data were collected every 4 h until equilibrium was obtained as determined from comparison of successive scans using the WinMatch v 0.99 program. The blank scan correction was edited using the WinReed v0.999 program. Global data analysis was performed using the WinNonlin v0.99 program. The programs WinMatch, WinReed and WinNonlin are available from the www.rasmb.org website.

Cross-linking reactions

Cross-linking reactions were performed using previously described methods [20,22]. Briefly, zero-length cross-linking reactions involving EDC and sulfo-NHS were performed at 0°C and 37°C, with final reagent concentrations of 20 mM and 10 mM respectively at 0°C, and 2.5 mM and 1.25 mM respectively at 37°C. Aliquots of each reaction were removed and quenched at 15, 30 and 60 min for reactions performed at 0°C and 15, 30, 60 and 120 min for reactions performed at 37°C. Reactions involving disuccinimidyl glutarate (DSG) and disuccinimidyl suberate (DSS) were performed with a 1:1 ratio of deuterium compared with hydrogen-labelled cross-linkers, specifically DSG-2H6:H6 and DSS-2H12:H12 and final reagent concentrations of 0.125 mM for reactions at 37°C and 1 mM for reactions at 0°C. Aliquots of each reaction were quenched at 15, 30, 60 and 120 min.

SDS/PAGE and trypsin digestion

The SDS/PAGE and trypsin digestion methods have been previously described [21]. Briefly, samples for the reactions described above were run in SDS/PAGE gels and bands of interest were excised and digested with trypsin.

MS analysis and identification of cross-linked peptides

For identification of zero-length cross-links, the cross-linked and control human PRDX6 tryptic digests were analysed and cross-linked sites were identified as previously described using an LTQ-Orbitrap XL™ mass spectrometer (Thermo Scientific) [22]. Briefly, cross-linked and control samples were analysed in parallel using LC–MS/MS and a label-free comparison was used to identify ions specific to the cross-linked samples. Data analysis and cross-linked peptide identification was performed using the ZXMiner program [22]. Cross-links for DSG and DSS were analysed in a similar fashion using xQuest/xProphet [27].

Homology modelling

The PRDX6 protein sequence and the 1PRX crystal structure were submitted to MODELLER 9v11 [25] to generate and refine human PRDX6 solution structures. All modelling experiments were run as 50-model trials using the ‘very slow’ refinement algorithm and discrete optimized protein energy (DOPE) score as an output. Homology modelling and refinement were performed simultaneously by including known intra- and inter-subunit cross-links as distance restraints between α-carbons imposed at 11.0±0.1 Å (1 Å=0.1 nm). Each model was subject to 1000 iterations and ten optimization repeats. The completed models were then analysed according to their DOPE score and the highest-scoring model under this criterion was chosen for further analysis. Molecular graphics were created using Open-Source PyMOL Version 1.3 (http:s://www.pymol.org/), which also was used to calculate distances between α-carbons of cross-linked glutamate, aspartate and lysine. No cross-links were observed involving the N-terminal amine group.

RESULTS

Human PRDX6 is in the reduced state

The recombinant human PRDX6 used in the present study was shown to be in the reduced state by MS analysis. That is, neither Cys47 nor Cys91 were significantly oxidized to the sulfenic or higher oxidation states when the reported purification method [13] was used. The oxidation state was assessed by two independent methods. First, LC–MS/MS was used to analyse tryptic digests of representative purified preparations of the protein. Data were analysed by considering all possible oxidation states of cysteine in the database search and all identified forms of cysteine-containing peptides were quantified by integrating peak areas after extracting ion chromatographs for pertinent precursor ions. Potential oxidation was also assessed by analysing the intact protein by ESI–MS analysis. In all cases, oxidation of the cysteines to sulfenic acid or higher oxidation states was less than 5%, unless the samples were deliberately oxidized or were stored for extended periods of time in solution at 0–4°C.

Human PRDX6 is a high-affinity homodimer

Separation of human PRDX6 by HPLC gel filtration showed a single prominent peak with an apparent mass of 45 kDa based on comparison with a standard curve using globular proteins (Figure 1A). This Stokes' radius is within reasonable error of the molecular mass expected of a dimeric PRDX6 protein. The oligomeric state of the recombinant PRDX6 was evaluated by sedimentation equilibrium. These experiments were conducted by parallel analysis of three protein concentrations (0.2, 0.4 and 0.8 mg/ml). After equilibrium was reached, the resulting protein concentration curves were globally fitted to various oligomeric models. As shown in Figure 1(B), the data best fitted a dimeric model with no evidence of significant monomer, indicating that PRDX6 forms high-affinity homodimers in solution with a Kd < 20 nM.

Oligomer states and conformations of human PRDX6 protein preparations

Figure 1
Oligomer states and conformations of human PRDX6 protein preparations

(A) HPLC gel filtration of recombinant human PRDX6. (B) Sedimentation equilibrium of the human PRDX6 peak at three different initial loading concentrations (0.8 mg/ml in black, 0.4 mg/ml in red and 0.2 mg/ml in green) at 55,256 × g and 30°C. All samples were prepared and analysed in 20 mM Tris/HCl, 130 mM NaCl, 0.1 mM EDTA and 5% glycerol, pH 7.4.

Figure 1
Oligomer states and conformations of human PRDX6 protein preparations

(A) HPLC gel filtration of recombinant human PRDX6. (B) Sedimentation equilibrium of the human PRDX6 peak at three different initial loading concentrations (0.8 mg/ml in black, 0.4 mg/ml in red and 0.2 mg/ml in green) at 55,256 × g and 30°C. All samples were prepared and analysed in 20 mM Tris/HCl, 130 mM NaCl, 0.1 mM EDTA and 5% glycerol, pH 7.4.

Zero-length CX–MS indicates that the solution structure diverges significantly from the crystal structure

Figure 2 shows the locations of the peroxidase and PLA2 active sites in the crystal structure of human PRDX6, which as noted above was prepared using a protein preparation where Cys91 was mutated to a serine and Cys47 was oxidized to a sulfenic acid [15]. In this structure of the peroxidase catalytic intermediate, the active-site residues for PLA2 activity are on the surface and the oxidized Cys47 residue is solvent-accessible at the bottom of a deep hydrophobic cavity. However, Thr177, which can be phosphorylated and modulates PLA2 activity [12], is not solvent-exposed in this structure. Chemical cross-linking was used to determine whether the solution structure of the reduced form of PRDX6 matched the crystal structure. Zero-length cross-linking was chosen for initial studies because these most precise distance constraints are the most valuable for evaluating and refining protein structures [16].

Crystal structure of human PRDX6 with subunit A in blue and subunit B in cyan

Figure 2
Crystal structure of human PRDX6 with subunit A in blue and subunit B in cyan

(A) Cartoon representation of the dimer. Cys47 is shown as yellow sticks; Cys91 is shown as orange sticks. (B) Solid surface representation of the dimer. Phospholipase catalytic triad (Ser26, His32, Asp140) is shown as red spheres; Cys47 is shown as yellow spheres.

Figure 2
Crystal structure of human PRDX6 with subunit A in blue and subunit B in cyan

(A) Cartoon representation of the dimer. Cys47 is shown as yellow sticks; Cys91 is shown as orange sticks. (B) Solid surface representation of the dimer. Phospholipase catalytic triad (Ser26, His32, Asp140) is shown as red spheres; Cys47 is shown as yellow spheres.

The protein was cross-linked at either 0°C or 37°C and reagent concentrations were adjusted to roughly compensate for the increased reaction rate of the higher temperature. Aliquots from each reaction were quenched over a range of time points to monitor for potential conformational distortion that could result from over-cross-linking (Figure 3). Moderate yields of cross-linked dimers were observed at longer reaction times, consistent with a high-affinity homodimer oligomeric state, as indicated by sedimentation equilibrium. No higher-order oligomers were detected. In addition, cross-linking of standard proteins confirmed that incidental cross-linking of molecules due to random collisions of non-associated molecules was not detectable under the conditions used in the present study. For MS analysis, larger protein loads and multiple lanes of each reaction condition were separated on gels and the monomer and dimer bands across all EDC/NHS reaction time points were excised and digested with trypsin.

Chemical cross-linking of the codon-optimized human PRDX6

Figure 3
Chemical cross-linking of the codon-optimized human PRDX6

SDS/PAGE of the human PRDX6 after incubation with the indicated chemical cross-linkers. Control (C) and time points (minutes) are indicated above each lane. The position of molecular mass markers are shown on the left and the migration of monomer and dimer bands are indicated.

Figure 3
Chemical cross-linking of the codon-optimized human PRDX6

SDS/PAGE of the human PRDX6 after incubation with the indicated chemical cross-linkers. Control (C) and time points (minutes) are indicated above each lane. The position of molecular mass markers are shown on the left and the migration of monomer and dimer bands are indicated.

Discovery MS analyses were performed by parallel LC–MS/MS analysis of the control sample (un-cross-linked) and all time points of cross-linked monomer and dimer bands. The LC–MS patterns for the control and cross-linked sample were compared in a label-free mode to identify MS precursor ions unique to the cross-linked sample that were putative cross-linked peptides. Targeted LC–MS/MS was then conducted on the longest time point using a list of m/z and retention times for the putative cross-linked precursors. The longest time point was used because it was expected to have the highest yield of most cross-linked peptides. This targeted LC–MS/MS reanalysis was used to obtain high-resolution high-mass-accuracy spectra for both MS1 and MS2.

As part of the cross-link identification process, the final output obtained from the ZXMiner program [22] was examined and only positively identified cross-links with a quality score [geometric mean (GM)] higher than that of the highest-scoring false positive cross-link were considered as high-confidence cross-links. Only high-confidence cross-links were used in order to ensure that the dataset had a false discovery rate (FDR) of 0%. Finally, any ambiguities between multiple possible cross-linkable sites on identified cross-linked peptide complexes were evaluated using ZXMiner.

PyMOL was used to visualize zero-length cross-links on the crystal structure and determine distances between cross-linked residues. EDC cross-links are formed between amine and carboxy groups that are within salt bridge distance [24] and these groups are primarily lysine and aspartate or glutamate side chains. However, α-carbon distances are typically used to evaluate cross-links because locations of protein backbones are better defined and less variable than side chains in either crystal structures or predicted models or actual structures in solution. Considering side chain lengths and uncertainty of structures, most zero-length cross-links involve residues with α-carbons within 12 Å. However, we previously showed that a few cross-links in well-characterized proteins have α-carbon distances up to about 16 Å, particularly for cross-links involving coil regions or subunit–subunit linkages [22]. As such, we considered 12 Å as the typical α-carbon limit for cross-linked residue distances with an upper limit of 16 Å.

High-confidence cross-links were assigned as intra-subunit, inter-subunit or ambiguous using several criteria. Cross-links confidently identified in the monomer band were considered to be intra-subunit. Cross-links observed in the dimer band were considered to be either intra-subunit or inter-subunit and both distances were calculated. A difference between these two distances of more than 11 Å resulted in assignment to the shorter distance possibility. Finally, for all ambiguous assignments, the crystal structure was visually examined using PyMOL for the presence of any major intervening structural elements, such as critical dimer interface contacts or an intervening helix. Using these criteria, at least 12 of 15 high confidence zero-length cross-links could be assigned as either intra-subunit or inter-subunit. Surprisingly, eight of these 12 observed distances were substantially larger than the expected 12 or 16 Å limits (Table 1). This indicated that the solution structure of the reduced form of the enzyme deviated substantially from the crystal structure.

Table 1
Human PRDX6 cross-links identified using EDC
Cross-link group z1 MH+ (Da)2 Mass error (ppm) Sequence A3 Sequence B3 Distance (Å)4 Cross-linked residues 
Inter-chain cross-links 
E01 1656.93474 4.6 AA[K]LAPEFAK [E]LPSGK 19.8 56–210 
E02 2640.42255 4.1 [E]LAILLGM#LDPAEKDEK ELPSG[K]K 27.5 109–215 
E03 2726.30843 5.7 DINAYNC(EE)PTEK GVFT[K]ELPSGK 15.8 92/93–209 
E04 3040.64670 5.3 LIALSIDSV[E]DHLAWSK GVFT[K]ELPSGK 22.4 77–209 
E05 4, 5 3044.62645 2.9 [E]LAILLGM#LDPAEKDEK GVFT[K]ELPSGK 33.0 109–209 
E06 3238.80547 4.2 ELAILLGM#LDPAEK[D]EK VVISLQLTAE[K]R 28.4 123–173 
E07 3, 4 3727.85644 1.8 ELAILLGM#LDPAE[K]DEK (D)G(D)SVM#VLPTIPEEEAK 37.9 122–183/185 
Intra-chain cross-links 
E08 1243.74389 6.2 [K]LFPK [E]LPSGK 21.0 200–210 
E09 1748.92572 5.0 GVFT[K]ELPSGK YTPQ[P] 18.9 209–224 
E10 1750.99291 4.9 [K]LFPKGVFTK YTPQ[P] 12.6 200–224 
E11 1943.10067 4.8 VVISLQLTAE[K]R YTPQ[P] 7.0 173–224 
E12 3, 4 2609.26208 4.5 AA[K]LAPEFAK DINAYNC(EE)PTEK 10.3 56–92/93 
Ambiguous cross-links5 
E13 2514.39782 5.5 [E]LAILLGM#LDPAEKDEK [K]LFPK 49.9/40.0 109–200 
E14 2770.40894 1.3 FHDFLG[D]SWGILFSHPR ELPSG[K]K 39.3/38.9 31–215 
E15 2872.47512 5.2 AA[K]LAPEFAK (D)G(D)SVMVLPTIPEEEAK 20.3/26.3 56–183/185 
Cross-link group z1 MH+ (Da)2 Mass error (ppm) Sequence A3 Sequence B3 Distance (Å)4 Cross-linked residues 
Inter-chain cross-links 
E01 1656.93474 4.6 AA[K]LAPEFAK [E]LPSGK 19.8 56–210 
E02 2640.42255 4.1 [E]LAILLGM#LDPAEKDEK ELPSG[K]K 27.5 109–215 
E03 2726.30843 5.7 DINAYNC(EE)PTEK GVFT[K]ELPSGK 15.8 92/93–209 
E04 3040.64670 5.3 LIALSIDSV[E]DHLAWSK GVFT[K]ELPSGK 22.4 77–209 
E05 4, 5 3044.62645 2.9 [E]LAILLGM#LDPAEKDEK GVFT[K]ELPSGK 33.0 109–209 
E06 3238.80547 4.2 ELAILLGM#LDPAEK[D]EK VVISLQLTAE[K]R 28.4 123–173 
E07 3, 4 3727.85644 1.8 ELAILLGM#LDPAE[K]DEK (D)G(D)SVM#VLPTIPEEEAK 37.9 122–183/185 
Intra-chain cross-links 
E08 1243.74389 6.2 [K]LFPK [E]LPSGK 21.0 200–210 
E09 1748.92572 5.0 GVFT[K]ELPSGK YTPQ[P] 18.9 209–224 
E10 1750.99291 4.9 [K]LFPKGVFTK YTPQ[P] 12.6 200–224 
E11 1943.10067 4.8 VVISLQLTAE[K]R YTPQ[P] 7.0 173–224 
E12 3, 4 2609.26208 4.5 AA[K]LAPEFAK DINAYNC(EE)PTEK 10.3 56–92/93 
Ambiguous cross-links5 
E13 2514.39782 5.5 [E]LAILLGM#LDPAEKDEK [K]LFPK 49.9/40.0 109–200 
E14 2770.40894 1.3 FHDFLG[D]SWGILFSHPR ELPSG[K]K 39.3/38.9 31–215 
E15 2872.47512 5.2 AA[K]LAPEFAK (D)G(D)SVMVLPTIPEEEAK 20.3/26.3 56–183/185 

*Observed charge states of the cross-linked peptide.

†Observed mono-isotopic mass of observed peptide.

‡[]: Cross-linked residue; (): potential cross-linked residue (ambiguous location); #: methionine oxidation.

§All distances are between α-carbons of cross-linked residues in the human PRDX6 crystal structure.

║Ambiguous cross-links are cross-links that could not be definitively assigned as inter- or intra-molecular; both distances are described in the table entry, separated by a slash. The inter-chain distance is before the slash and the intra-chain distance is after the slash.

Determination of a solution structure for human PRDX6

MODELLER was used to refine the 1PRX crystal structure by imposing α-carbon distance constraints for the 12 unambiguously assigned cross-links. The resulting preliminary model, hereafter referred to as the EDC1 model, was then compared with the crystal structure via a superimposition (Figure 4A), which showed that most major structural features in the PRDX6 homodimer remained unchanged, with an overall RMSD of 1.4 Å. Moreover, an analysis of the Ramachandran angles using Coot [28] for this model also revealed that over 90% were found in preferred or allowed conformations, reinforcing the validity of the model. The top five models generated, as ranked by DOPE score, were also very similar to one another, indicating convergence on a best model.

Initial molecular model of human PRDX6 using EDC cross-links

Figure 4
Initial molecular model of human PRDX6 using EDC cross-links

(A) Superimposition of the crystal structure (blue) and EDC1 model (red; RMSD: 1.4 Å). For simplicity, only the A chain on both dimers is highlighted; both B chains are in grey. Cys47 is shown as yellow sticks, Cys91 is shown as orange sticks and Thr177 is shown as light green sticks. (B) Close-up of the C-terminal region on chain A, which shows the largest difference relative to the crystal structure (residues 190–224). (C) α-Carbon distances between residues identified using EDC cross-links for the crystal structure and EDC1 model. Expected distance cut-offs are 12 Å for well-ordered regions (lower black dashed line) and 16 Å for disordered regions (upper black dashed line). (D) α-Carbon distances between residues with ambiguous assignments as to whether they are inter-chain or intra-chain.

Figure 4
Initial molecular model of human PRDX6 using EDC cross-links

(A) Superimposition of the crystal structure (blue) and EDC1 model (red; RMSD: 1.4 Å). For simplicity, only the A chain on both dimers is highlighted; both B chains are in grey. Cys47 is shown as yellow sticks, Cys91 is shown as orange sticks and Thr177 is shown as light green sticks. (B) Close-up of the C-terminal region on chain A, which shows the largest difference relative to the crystal structure (residues 190–224). (C) α-Carbon distances between residues identified using EDC cross-links for the crystal structure and EDC1 model. Expected distance cut-offs are 12 Å for well-ordered regions (lower black dashed line) and 16 Å for disordered regions (upper black dashed line). (D) α-Carbon distances between residues with ambiguous assignments as to whether they are inter-chain or intra-chain.

This model also pinpointed the area of maximum variability between the two structures as involving residues 190–224 (Figure 4B), with an RMSD of 2.1 Å. The distances for the cross-links used as distance restraints were all below the 16 Å cut-off point (Figure 4C). In addition, the α-carbon distances for the three ambiguous cross-links that were not used in this structural refinement showed smaller inter-chain distances compared with the intra-chain distances in the EDC1 model (Figure 4D). This trend, coupled with the observation that all three cross-links had impeding intervening structures between their side chains for a putative intra-chain cross-link, enabled us to assign them as inter-chain cross-links. However, they remained larger than the expected 16 Å cut-off, indicating that further refinement was needed.

A second model, hereafter referred to as the EDC2 model, was produced using MODELLER to refine the 1PRX crystal structure by imposing expected α-carbon distance constraints for all 15 EDC cross-links. Similar to the EDC1 model, most major structural features in the PRDX6 homodimer remained unchanged relative to the crystal structure with an overall RMSD of 1.7 Å (Figure 5A) and the area of maximum variability between the two structures was residues 190–224 with an RMSD of 2.9 Å (Figure 5B). This model was also analysed via Ramachandran plot and found to have over 90% of its bond angles in allowed or preferred conformations. As before, the top five models by DOPE score were also structurally convergent. In this model, all 15 EDC cross-links were now within the 16 Å maximum distance (Figures 5C and 5D) indicating that all zero-length cross-link data supported this structure.

Molecular model of human PRDX6 using all EDC cross-links

Figure 5
Molecular model of human PRDX6 using all EDC cross-links

(A) Superimposition of the crystal structure (blue) and EDC2 model (purple; RMSD: 1.7 Å). For details of colour schemes and highlighted residues, see Figure 4. (B) Close-up of the C-terminal region on chain A, which shows the largest difference relative to the crystal structure (residues 190–224). (C) α-Carbon distances between residues identified using EDC cross-links for the crystal structure and EDC2 model. Expected distance cut-offs are 12Å for well-ordered regions (lower black dashed line) and 16Å for disordered regions (upper black dashed line). (D) α-Carbon distances between residues with initially ambiguous assignments as to whether they are inter-chain or intra-chain.

Figure 5
Molecular model of human PRDX6 using all EDC cross-links

(A) Superimposition of the crystal structure (blue) and EDC2 model (purple; RMSD: 1.7 Å). For details of colour schemes and highlighted residues, see Figure 4. (B) Close-up of the C-terminal region on chain A, which shows the largest difference relative to the crystal structure (residues 190–224). (C) α-Carbon distances between residues identified using EDC cross-links for the crystal structure and EDC2 model. Expected distance cut-offs are 12Å for well-ordered regions (lower black dashed line) and 16Å for disordered regions (upper black dashed line). (D) α-Carbon distances between residues with initially ambiguous assignments as to whether they are inter-chain or intra-chain.

Independent evaluation of the PRDX6 solution structure using homo-bifunctional cross-linkers

We performed independent CX–MS experiments on human PRDX6 using the homo-bifunctional cross-linkers DSG (2H6:H6 mixture; 7.7 Å linker) and DSS (2H12:H12 mixture; 11.4 Å linker) to further evaluate the PRDX6 solution structure (Figure 3). As with the EDC experiments, the monomer and cross-linked dimer bands were excised, digested with trypsin and analysed by LC–MS/MS. These cross-linked peptides were identified using xQuest/xProphet [27]. The resulting high-confidence cross-links (Tables 2 and 3) were mapped on to both the crystal structure and the EDC2 structure (Figures 6A and 6B). Due to the lengths of the spacer arms between the reactive groups of these cross-linkers [24], maximum α-carbon distances of 22 Å for DSG and 26 Å for DSS were used, consistent with previous studies using these cross-linkers [21]. When the crystal structure was analysed, four of ten DSG cross-links exceeded the maximum expected distance threshold, further confirming that the solution structure of the reduced protein significantly differed from that of the catalytic intermediate. However, three of these four cross-links were within expected limits when the EDC2 model was considered. For DSS, which has an even longer spacer arm, none of the cross-links exceeded the threshold for that cross-linker on any of the structures, indicating the inability of this longer cross-link to distinguish between alternative conformations. The single DSG cross-link (D10) that continued to exceed the expected distance threshold in the EDC2 structure was further evaluated. Interestingly, this cross-link involved Lys204, which is located within the most variable region of the protein as described above. This cross-link (Table 2) was also notable because its assignment as an inter-chain or intra-chain cross-link was ambiguous when compared with the 1PRX crystal structure due to α-carbon distances of 30.3 Å and 36.3 Å for inter-chain and intra-chain interactions respectively. There also were intervening structural elements for both possible types of cross-link. However, this controversy was resolved when the EDC2 model was examined, as the inter-chain interaction no longer had major intervening structural elements and displayed an α-carbon distance of 26.1 Å, as opposed to 42.5 Å for the intra-chain interaction.

Table 2
Human PRDX6 cross-links identified using DSG
Cross-link group z1 MH+ (Da)2 Mass error (ppm) Sequence A3 Sequence-B3 Cross-linked residues 
Inter-chain cross-links 
D01 1898.042 −4.8 AA[K]LAPEFAK ELPSG[K]K 56–215 
D024 1955.988 −5.2 DE[K]GMPVTAR ELPSG[K]K 125–215 
D03 2302.258 0.5 GVFT[K]ELPSGK AA[K]LAPEFAK 209–56 
D04 2376.195 −1.9 GVFT[K]ELPSGK DE[K]GM#PVTAR 209–125 
D05 2570.369 −1.5 VVISLQLTAE[K]R DE[K]GM#PVTAR 173–125 
Intra-chain cross-links 
D06 2630.409 −0.1 L[K]LSILYPATTGR DE[K]GMPVTAR 144–125 
D07 2826.431 3.8 DGDSVM#VLPTIPEEEA[K]K ELPSG[K]K 199–215 
D08 2883.675 −1.5 L[K]LSILYPATTGR VVISLQLTAE[K]R 144–173 
D09 3263.753 −0.6 NV[K]LIALSIDSVEDHLAWSK LAPEFA[K]R 67–63 
Ambiguous cross-links 
D10 2176.226 −1.4 AA[K]LAPEFAK LFP[K]GVFTK 56–204 
Cross-link group z1 MH+ (Da)2 Mass error (ppm) Sequence A3 Sequence-B3 Cross-linked residues 
Inter-chain cross-links 
D01 1898.042 −4.8 AA[K]LAPEFAK ELPSG[K]K 56–215 
D024 1955.988 −5.2 DE[K]GMPVTAR ELPSG[K]K 125–215 
D03 2302.258 0.5 GVFT[K]ELPSGK AA[K]LAPEFAK 209–56 
D04 2376.195 −1.9 GVFT[K]ELPSGK DE[K]GM#PVTAR 209–125 
D05 2570.369 −1.5 VVISLQLTAE[K]R DE[K]GM#PVTAR 173–125 
Intra-chain cross-links 
D06 2630.409 −0.1 L[K]LSILYPATTGR DE[K]GMPVTAR 144–125 
D07 2826.431 3.8 DGDSVM#VLPTIPEEEA[K]K ELPSG[K]K 199–215 
D08 2883.675 −1.5 L[K]LSILYPATTGR VVISLQLTAE[K]R 144–173 
D09 3263.753 −0.6 NV[K]LIALSIDSVEDHLAWSK LAPEFA[K]R 67–63 
Ambiguous cross-links 
D10 2176.226 −1.4 AA[K]LAPEFAK LFP[K]GVFTK 56–204 

*Observed charge states of the cross-linked peptide.

†Observed mono-isotopic mass of observed peptide.

‡[]: Cross-linked residue; (): potential cross-linked residue (ambiguous location); #: methionine oxidation.

§Multiple variants of this cross-link have been identified in this dataset.

Table 3
Human PRDX6 cross-links identified using DSS
Cross-link group z1 MH+ (Da)2 Mass error (ppm) Sequence A3 Sequence B3 Cross-linked residues 
Inter-chain cross-links 
S014 1940.09 −4.2 AA[K]LAPEFAK ELPSG[K]K 56–215 
S024 1998.038 −3.7 DE[K]GMPVTAR ELPSG[K]K 125–215 
S034 2344.296 −3.6 GVFT[K]ELPSGK AA[K]LAPEFAK 209–56 
S044 3469.896 −0.8 ELAILLGM#LDPAE[K]DEK L[K]LSILYPATTGR 122–144 
S05 3523.839 −3.7 ELAILLGM#LDPAE[K]DEKGM#PVTAR ELPSG[K]K 122–215 
Intra-chain cross-links 
S064 2688.448 −1.6 L[K]LSILYPATTGR DE[K]GM#PVTAR 144–125 
S07 2925.727 L[K]LSILYPATTGR VVISLQLTAE[K]R 144–173 
S08 3830.905 −0.8 DINAYNCEEPTE[K]LPFPIIDDR AA[K]LAPEFAK 97–56 
Cross-link group z1 MH+ (Da)2 Mass error (ppm) Sequence A3 Sequence B3 Cross-linked residues 
Inter-chain cross-links 
S014 1940.09 −4.2 AA[K]LAPEFAK ELPSG[K]K 56–215 
S024 1998.038 −3.7 DE[K]GMPVTAR ELPSG[K]K 125–215 
S034 2344.296 −3.6 GVFT[K]ELPSGK AA[K]LAPEFAK 209–56 
S044 3469.896 −0.8 ELAILLGM#LDPAE[K]DEK L[K]LSILYPATTGR 122–144 
S05 3523.839 −3.7 ELAILLGM#LDPAE[K]DEKGM#PVTAR ELPSG[K]K 122–215 
Intra-chain cross-links 
S064 2688.448 −1.6 L[K]LSILYPATTGR DE[K]GM#PVTAR 144–125 
S07 2925.727 L[K]LSILYPATTGR VVISLQLTAE[K]R 144–173 
S08 3830.905 −0.8 DINAYNCEEPTE[K]LPFPIIDDR AA[K]LAPEFAK 97–56 

*Observed charge states of the cross-linked peptide.

†Observed mono-isotopic mass of observed peptide.

‡[]: Cross-linked residue; (): potential cross-linked residue (ambiguous location); #: methionine oxidation.

§Multiple variants of this cross-link have been identified in this dataset.

Evaluation of solution structure using non-zero-length cross-linkers

Figure 6
Evaluation of solution structure using non-zero-length cross-linkers

(A) Histogram analysis of α-carbon distances between residues identified using DSG cross-links for the crystal structure (blue), EDC2 model (purple) and final model (forest green). Expected distance cut-off is 22 Å (dashed black line). (B) Histogram analysis of α-carbon distances between residues identified using DSS cross-links for the crystal structure, EDC2 model and final model. Expected distance cut-off is 26 Å. (C) Superimposition of the crystal structure (blue) and final model (forest green; RMSD: 1.6 Å). For details of colour schemes and highlighted residues, see Figure 4. (D) Superimposition of the EDC2 model (purple) and final model (forest green; RMSD: 0.7 Å).

Figure 6
Evaluation of solution structure using non-zero-length cross-linkers

(A) Histogram analysis of α-carbon distances between residues identified using DSG cross-links for the crystal structure (blue), EDC2 model (purple) and final model (forest green). Expected distance cut-off is 22 Å (dashed black line). (B) Histogram analysis of α-carbon distances between residues identified using DSS cross-links for the crystal structure, EDC2 model and final model. Expected distance cut-off is 26 Å. (C) Superimposition of the crystal structure (blue) and final model (forest green; RMSD: 1.6 Å). For details of colour schemes and highlighted residues, see Figure 4. (D) Superimposition of the EDC2 model (purple) and final model (forest green; RMSD: 0.7 Å).

In order to further explore the impact of the D10 DSG cross-link on the protein structure, we generated a final solution structure using MODELLER to refine the 1PRX crystal structure by imposing expected α-carbon distance constraints for all 15 EDC cross-links plus the D10 distance constraint. All DSG and DSS cross-links were within expected limits in this model (Figures 6A and 6B). We then compared the final structure to the crystal structure (Figure 6C). As with the above models, overall backbone structure was similar to an RMSD of 1.6 Å and the RMSD of the C-terminal tail (residues 190–224) was 2.4 Å, which indicated that this region was actually slightly closer to the crystal structure than in the EDC2 model. This model was also analysed via Ramachandran plot and once again we found that over 90% of its bond angles were in preferred or allowed regions. The top five models by DOPE score also converged structurally. The final model was also directly compared with the EDC2 model (Figure 6D). As expected, the differences between the EDC2 model and the final solution structure were minor with an overall RMSD of 0.74 Å and an RMSD of 1.0 for the C-terminal region. Their distributions of Ramachandran bond angles were quite similar overall.

DISCUSSION

We utilized CX–MS and structural refinement with cross-link distance constraints to determine an experimentally validated solution structure of the reduced form of PRDX6. The final structure fits distance constraints of all high-confidence cross-links using reagents with three different cross-link spacer arms (0, 7.7 and 11.4 Å). Consistent with prior studies, the zero-length cross-links were the most useful for identifying conformational differences between the solution structure of the reduced protein and the crystal structure of the catalytic intermediate. In contrast, the longest cross-linker was too imprecise and could not unambiguously identify any discrepancies between the solution structure and the crystal structure. The differences between the solution structure and crystal structure were mostly, but not entirely, located within the C-terminal tail of the protein (residues 190–224). This region primarily consist of coil, which probably contributes to its plasticity. However, the fact that all 33 distance constraints can be satisfied by a single structure suggests, but does not prove, that the solution structure of the protein as isolated herein is primarily a single conformation rather than an ensemble of inter-converting structures.

Importantly, significant changes in protein structure are not limited to residues 190–224. Another functionally important change is the solvent accessibility of Thr177. As noted above, this residue is phosphorylated in response to certain stimuli and increases PLA2 activity and affinity for liposomes [12], but it is completely buried in the crystal structure. In contrast, this residue is substantially solvent-exposed in our EDC2 model and the final solution structure (Figure 7). Taken together, these data suggest that the reduced form, but not the oxidized form, of the enzyme, should be capable of being phosphorylated.

Surface accessibility of Thr177 in PRDX6 crystal and solution structures

Figure 7
Surface accessibility of Thr177 in PRDX6 crystal and solution structures

Close-up of protein surface display for Thr177 (green spheres) in (A) 1PRX crystal structure and (B) EDC2 solution structure.

Figure 7
Surface accessibility of Thr177 in PRDX6 crystal and solution structures

Close-up of protein surface display for Thr177 (green spheres) in (A) 1PRX crystal structure and (B) EDC2 solution structure.

We also evaluated whether the conformational differences between the solution and crystal structures correlated with the most flexible regions in the crystal structure. An analysis of the Debye–Waller temperature factors (B-factors) in the crystal structure (Figure 8A), shows that residues 92–93 and 121–126 exhibited the highest B-factors. The remainder of the protein had B-factors less than 40, which are indicative of relatively rigid structures. Comparison of residues 92–93 in the crystal structure and the final solution structure show that this coil region is not any more variable than its flanking regions (Figure 8B). Residues 121–126 encompass another coil region that also does not exhibit substantially greater variability between structures than other adjacent coil regions. Interestingly, however, this region is significantly more compact in the final solution model compared with the crystal structure (Figure 8C). Finally, as noted above, the C-terminal region encompassed by residues 190–224 shows the largest variation between the two structures (Figure 8D) but this region does not have high B-factors. Interestingly, an analogous region in 2-Cys peroxiredoxins also exhibits conformational change as part of the catalysis process [4]. However, there is one key difference in that the 2-Cys C-terminal region experiences localized unfolding in response to the breaking of a disulfide bond, whereas the analogous region in PRDX6 becomes more compact. Overall, there does not appear to be a strong correlation between regions of high flexibility in the crystal structure and regions involved in conformational changes associated with transition from the reduced form to the peroxidase catalytic intermediate form of the enzyme. This suggests that these conformational changes play an important role in the catalytic mechanism and do not simply reflect variations in intrinsically mobile regions of the protein. It also suggests that these conformational changes associated with the peroxidase mechanism could help to regulate PLA2 activity. That is, when the enzyme is in the reduced form, but not the catalytic intermediate state, Thr177 is solvent-exposed and can be phosphorylated thereby increasing PLA2 activity and affinity for liposomes.

Relationship of B-factors and variations between the PRDX6 crystal and solution structures

Figure 8
Relationship of B-factors and variations between the PRDX6 crystal and solution structures

(A) Analysis of B-factors for 1PRX crystal structure. (B) Superimposition of 1PRX crystal structure (blue) and final model (forest green) highlighting residues Glu92 and Glu93 (orange and yellow for 1PRX and final respectively). (C) Superimposition of 1PRX crystal structure (blue) and final model (forest green) highlighting residues Glu92 and Glu93 (orange and yellow for 1PRX and final respectively). (D) Superimposition of 1PRX crystal structure (blue) and final model (forest green) highlighting residues 190–224 (wheat and magenta for 1PRX and final respectively).

Figure 8
Relationship of B-factors and variations between the PRDX6 crystal and solution structures

(A) Analysis of B-factors for 1PRX crystal structure. (B) Superimposition of 1PRX crystal structure (blue) and final model (forest green) highlighting residues Glu92 and Glu93 (orange and yellow for 1PRX and final respectively). (C) Superimposition of 1PRX crystal structure (blue) and final model (forest green) highlighting residues Glu92 and Glu93 (orange and yellow for 1PRX and final respectively). (D) Superimposition of 1PRX crystal structure (blue) and final model (forest green) highlighting residues 190–224 (wheat and magenta for 1PRX and final respectively).

The present study highlights the advantages of using zero-length cross-linkers for probing fairly subtle but biologically important conformational changes. Zero-length cross-links have been previously shown to be the most useful type of cross-link for refining protein structures and experimentally verifying homology models [16]. However, this approach has been under-utilized, due to the increased difficulty of identifying such cross-links. These difficulties are due to the fact that stable isotope tags or other features that aid in MS-based identification of cross-linked peptides cannot be exploited with a cross-linker that does not incorporate external atoms into the cross-link bridge. This hurdle has recently been addressed by development of an analysis strategy using label-free LC–MS comparisons, targeted acquisition of high-mass-accuracy MS/MS spectra of putative cross-links and development of the ZXMiner software that has been optimized for in-depth identification of zero-length cross-links [22]. The power of this approach for in-depth analysis was previously demonstrated by comparing alternative programs on the same dataset [21]. It is further demonstrated in the present study by the identification of a greater number of high-confidence cross-links EDC (15 unique cross-link sites) when compared with DSG (ten cross-link sites) or DSS (eight cross-link sites). This larger number of zero-length cross-links was identified despite less extensive overall apparent cross-linking relative to the DSG and DSS reactions as indicated by the less extensive formation of covalently linked homodimers on SDS gels (Figure 3). More importantly, EDC cross-linking represented the most sensitive test for conformational changes. When observed cross-links were compared with the crystal structure, eight of 12 EDC cross-links (67%) involved residues that were further apart than the maximum likely distance, whereas only four of ten DSG (40%) and zero of eight DSS (0%) cross-links were outside expected maximum distances.

In-depth identification of sub-stoichiometric levels of cross-links is important for structural studies because extensive chemical modification of a protein may perturb its structure. To minimize the risk of this complication, cross-link reaction time courses were always analysed (Figure 3) and evidence of precursor ions corresponding to identified cross-linked complexes was required at the early time points, as previously described [23]. Also, it is likely that zero-length cross-linking may be less sensitive to over-cross-linking and conformational perturbation because the reaction is reversible and dead-end products are typically not formed. In contrast, amine-specific homo-bifunctional cross-linkers such as DSG and DSS will readily modify most or all solvent-exposed surface amines, resulting in a large change of net surface charge. In addition, if the second site on the reagent does not react with another amine on the protein, a dead-end product is formed. This irreversible reactivity of most exposed amines makes over-reaction and conformational perturbation a substantial concern for such cross-linkers.

In summary, the solution structure of the reduced form of PRDX6 is of particular interest because multiple biochemical studies have identified conformational changes of the protein that are associated with changes in enzyme activity and the only reported high-resolution crystal structure is of the peroxidase catalytic intermediate form of the protein [15]. Our development of an experimentally supported solution structure is consistent with multiple biochemical studies that indicate that PRDX6 is a protein with substantial plasticity that can affect both known enzyme activities of this protein. It also yields novel insights regarding the nature of the changes that occur during the peroxidase catalysis and the likely interplay between regulation of the two enzyme activities, as indicated by the solvent accessibility of Thr177. The reported structure also provides an important structural reference for this protein, as the reduced form of PRDX6 is much more commonly encountered than the peroxidase catalytic intermediate. These insights combine to give us a greater understanding of this important antioxidant enzyme.

We gratefully acknowledge the assistance of The Wistar Institute Proteomics Core and the expert assistance of Peter Hembach and Elena Sorokina.

FUNDING

This work was supported by the National Institutes of Health [grant numbers HL10216 (to A.B. F.), CA10815 (NCI core grant to The Wistar Institute) and T32 GM008275 (to R.R.S.)].

Abbreviations

     
  • CX–MS

    chemical cross-linking coupled with high-resolution MS

  •  
  • DOPE

    discrete optimized protein energy

  •  
  • DSG

    disuccinimidyl glutarate

  •  
  • DSS

    disuccinimidyl suberate

  •  
  • EDC

    1-ethyl-3-(-3-dimethylaminopropyl) carbodi-imide hydrochloride

  •  
  • PLA2

    phospholipase A2

  •  
  • PRDX6

    peroxiredoxin-6

  •  
  • ROS

    reactive oxygen species

  •  
  • sulfo-NHS

    sulfo-N-hydroxysuccinimide

References

References
1
Fisher
A.B.
Forman
H.J.
Glass
M.
Mechanisms of pulmonary oxygen toxicity
Lung
1984
, vol. 
162
 (pg. 
255
-
259
)
[PubMed]
2
Lang
J.D.
McArdle
P.J.
O'Reilly
P.J.
Matalon
S.
Oxidant-antioxidant balance in acute lung injury
Chest
2002
, vol. 
122
 (pg. 
314S
-
320S
)
[PubMed]
3
Rahman
I.
Adcock
I.M.
Oxidative stress and redox regulation of lung inflammation in COPD
Eur. Respir. J.
2006
, vol. 
28
 (pg. 
219
-
242
)
[PubMed]
4
Wood
Z.A.
Schroder
E.
Robin Harris
J.
Poole
L.B.
Structure, mechanism and regulation of peroxiredoxins
Trends Biochem. Sci.
2003
, vol. 
28
 (pg. 
32
-
40
)
[PubMed]
5
Rhee
S.G.
Chae
H.Z.
Kim
K.
Peroxiredoxins: a historical overview and speculative preview of novel mechanisms and emerging concepts in cell signaling
Free Radic. Biol. Med.
2005
, vol. 
38
 (pg. 
1543
-
1552
)
[PubMed]
6
Manevich
Y.
Fisher
A.B.
Peroxiredoxin 6, a 1-Cys peroxiredoxin, functions in antioxidant defense and lung phospholipid metabolism
Free Radic. Biol. Med.
2005
, vol. 
38
 (pg. 
1422
-
1432
)
[PubMed]
7
Kang
S.W.
Baines
I.C.
Rhee
S.G.
Characterization of a mammalian peroxiredoxin that contains one conserved cysteine
J. Biol. Chem.
1998
, vol. 
273
 (pg. 
6303
-
6311
)
[PubMed]
8
Manevich
Y.
Reddy
K.S.
Shuvaeva
T.
Feinstein
S.I.
Fisher
A.B.
Structure and phospholipase function of peroxiredoxin 6: identification of the catalytic triad and its role in phospholipid substrate binding
J. Lipid Res.
2007
, vol. 
48
 (pg. 
2306
-
2318
)
[PubMed]
9
Fisher
A.B.
Dodia
C.
Manevich
Y.
Chen
J.W.
Feinstein
S.I.
Phospholipid hydroperoxides are substrates for non-selenium glutathione peroxidase
J. Biol. Chem.
1999
, vol. 
274
 (pg. 
21326
-
21334
)
[PubMed]
10
Chen
J.W.
Dodia
C.
Feinstein
S.I.
Jain
M.K.
Fisher
A.B.
1-Cys peroxiredoxin, a bifunctional enzyme with glutathione peroxidase and phospholipase A2 activities
J. Biol. Chem.
2000
, vol. 
275
 (pg. 
28421
-
28427
)
[PubMed]
11
Ralat
L.A.
Manevich
Y.
Fisher
A.B.
Colman
R.F.
Direct evidence for the formation of a complex between 1-cysteine peroxiredoxin and glutathione S-transferase pi with activity changes in both enzymes
Biochemistry
2006
, vol. 
45
 (pg. 
360
-
372
)
[PubMed]
12
Wu
Y.
Feinstein
S.I.
Manevich
Y.
Chowdhury
I.
Pak
J.H.
Kazi
A.
Dodia
C.
Speicher
D.W.
Fisher
A.B.
Mitogen-activated protein kinase-mediated phosphorylation of peroxiredoxin 6 regulates its phospholipase A(2) activity
Biochem. J.
2009
, vol. 
419
 (pg. 
669
-
679
)
[PubMed]
13
Rahaman
H.
Zhou
S.
Dodia
C.
Feinstein
S.I.
Huang
S.
Speicher
D.
Fisher
A.B.
Increased phospholipase A2 activity with phosphorylation of peroxiredoxin 6 requires a conformational change in the protein
Biochemistry
2012
, vol. 
51
 (pg. 
5521
-
5530
)
[PubMed]
14
Manevich
Y.
Shuvaeva
T.
Dodia
C.
Kazi
A.
Feinstein
S.I.
Fisher
A.B.
Binding of peroxiredoxin 6 to substrate determines differential phospholipid hydroperoxide peroxidase and phospholipase A(2) activities
Arch. Biochem. Biophys.
2009
, vol. 
485
 (pg. 
139
-
149
)
[PubMed]
15
Choi
H.J.
Kang
S.W.
Yang
C.H.
Rhee
S.G.
Ryu
S.E.
Crystal structure of a novel human peroxidase enzyme at 2.0 Å resolution
Nat. Struct. Biol.
1998
, vol. 
5
 (pg. 
400
-
406
)
[PubMed]
16
Leitner
A.
Walzthoeni
T.
Kahraman
A.
Herzog
F.
Rinner
O.
Beck
M.
Aebersold
R.
Probing native protein structures by chemical cross-linking, mass spectrometry, and bioinformatics
Mol. Cell Proteomics
2010
, vol. 
9
 (pg. 
1634
-
1649
)
[PubMed]
17
Greber
B.J.
Boehringer
D.
Leibundgut
M.
Bieri
P.
Leitner
A.
Schmitz
N.
Aebersold
R.
Ban
N.
The complete structure of the large subunit of the mammalian mitochondrial ribosome
Nature
2014
, vol. 
515
 (pg. 
283
-
286
)
[PubMed]
18
Kaake
R.M.
Wang
X.
Burke
A.
Yu
C.
Kandur
W.
Yang
Y.
Novtisky
E.J.
Second
T.
Duan
J.
Kao
A.
, et al. 
A new in vivo cross-linking mass spectrometry platform to define protein-protein interactions in living cells
Mol. Cell. Proteomics
2014
, vol. 
13
 (pg. 
3533
-
3543
)
[PubMed]
19
Olson
A.L.
Tucker
A.T.
Bobay
B.G.
Soderblom
E.J.
Moseley
M.A.
Thompson
R.J.
Cavanagh
J.
Structure and DNA-binding traits of the transition state regulator AbrB
Structure
2014
, vol. 
22
 (pg. 
1650
-
1656
)
[PubMed]
20
Li
D.
Harper
S.L.
Tang
H.Y.
Maksimova
Y.
Gallagher
P.G.
Speicher
D.W.
A comprehensive model of the spectrin divalent tetramer binding region deduced using homology modeling and chemical cross-linking of a mini-spectrin
J. Biol. Chem.
2010
, vol. 
285
 (pg. 
29535
-
29545
)
[PubMed]
21
Sriswasdi
S.
Harper
S.L.
Tang
H.Y.
Gallagher
P.G.
Speicher
D.W.
Probing large conformational rearrangements in wild-type and mutant spectrin using structural mass spectrometry
Proc. Natl. Acad. Sci. U.S.A.
2014
, vol. 
111
 (pg. 
1801
-
1806
)
[PubMed]
22
Sriswasdi
S.
Harper
S.L.
Tang
H.Y.
Speicher
D.W.
Enhanced identification of zero-length chemical cross-links using label-free quantitation and high-resolution fragment ion spectra
J. Proteome Res.
2014
, vol. 
13
 (pg. 
898
-
914
)
[PubMed]
23
Harper
S.L.
Sriswasdi
S.
Tang
H.Y.
Gaetani
M.
Gallagher
P.G.
Speicher
D.W.
The common hereditary elliptocytosis-associated alpha-spectrin L260P mutation perturbs erythrocyte membranes by stabilizing spectrin in the closed dimer conformation
Blood
2013
, vol. 
122
 (pg. 
3045
-
3053
)
[PubMed]
24
Paramelle
D.
Miralles
G.
Subra
G.
Martinez
J.
Chemical cross-linkers for protein structure studies by mass spectrometry
Proteomics
2013
, vol. 
13
 (pg. 
438
-
456
)
[PubMed]
25
Sali
A.
Blundell
T.L.
Comparative protein modelling by satisfaction of spatial restraints
J. Mol. Biol.
1993
, vol. 
234
 (pg. 
779
-
815
)
[PubMed]
26
Yphantis
D.
Equilibrium centrifugation of dilute solutions
Biochemistry
1964
, vol. 
3
 (pg. 
297
-
317
)
[PubMed]
27
Rinner
O.
Seebacher
J.
Walzthoeni
T.
Mueller
L.N.
Beck
M.
Schmidt
A.
Mueller
M.
Aebersold
R.
Identification of cross-linked peptides from large sequence databases
Nat. Methods
2008
, vol. 
5
 (pg. 
315
-
318
)
[PubMed]
28
Emsley
P.
Lohkamp
B.
Scott
W.G.
Cowtan
K.
Features and development of Coot
Acta Crystallogr. D Biol. Crystallogr.
2010
, vol. 
66
 (pg. 
486
-
501
)
[PubMed]

Author notes

The crystal structure of human peroxidase enzyme at 2.0 Å resolution appears in the PDB under accession code 1PRX.