Peroxiredoxin Evolution and the Regulation of Hydrogen Peroxide Signaling

See allHide authors and affiliations

Science  25 Apr 2003:
Vol. 300, Issue 5619, pp. 650-653
DOI: 10.1126/science.1080405


Eukaryotic 2-Cys peroxiredoxins (2-Cys Prxs) not only act as antioxidants, but also appear to regulate hydrogen peroxide–mediated signal transduction. We showthat bacterial 2-Cys Prxs are much less sensitive to oxidative inactivation than are eukaryotic 2-Cys Prxs. By identifying two sequence motifs unique to the sensitive 2-Cys Prxs and comparing the crystal structure of a bacterial 2-Cys Prx at 2.2 angstrom resolution with other Prx structures, we define the structural origins of sensitivity. We suggest this adaptation allows 2-Cys Prxs to act as floodgates, keeping resting levels of hydrogen peroxide low, while permitting higher levels during signal transduction.

All aerobic organisms have evolved efficient and specific defense systems that rapidly detoxify dangerous oxidants like hydrogen peroxide and superoxide. Though hydrogen peroxide is a source of oxidative stress, it also acts as a second messenger in signal transduction [e.g., the nuclear factor κB (NF-κB) and mitogen-activated protein (MAP) kinase pathways (1)] at least in part by reacting with thiols on proteins involved in signaling (2). Hydrogen peroxide is an effective signaling molecule because it is rapidly produced, is reactive, and is easily controlled by antioxidant enzymes (3, 4). But what determines whether hydrogen peroxide acts as a deleterious oxidant or a beneficial signal? We propose that the cellular switch is found in the ubiquitous peroxidases, 2-Cys peroxiredoxins (2-Cys Prxs).

The 2-Cys Prxs were originally identified as antioxidant enzymes requiring two redox-active cysteines (a peroxidatic and a resolving cysteine) for peroxide decomposition [recently reviewed in (5, 6)]. In addition to reducing hydrogen peroxide and alkyl hydroperoxides, 2-Cys Prxs are also efficient peroxynitrite reductases (7), and recent evidence suggests that they regulate peroxide-mediated signaling cascades [reviewed in (8)]. In vivo, 2-Cys Prxs are required for Myc-mediated transformation and apoptosis (9, 10). They can regulate NF-κB activation (11, 12), and their overexpression can reduce intracellular hydrogen peroxide generated in response to tumor necrosis factor-α (11), p53 (13), epidermal growth factor (13), and thyrotropin (14). Also, overexpression of Prxs is observed in several cancers (1519) and is correlated with resistance to apoptosis induced by radiation therapy (20) or the anticancer drug cisplatin (21).

Typical 2-Cys Prxs are obligate dimers (Fig. 1A) with a mechanism that has been chemically and structurally well defined (5, 6). During catalysis, the peroxidatic cysteine (CP-SH) is oxidized to a sulfenic acid (CP-SOH), which then condenses with the resolving cysteine (CR-SH from the other subunit of the dimer) to form a disulfide, which is in turn reduced by thioredoxin or another enzyme (Fig. 1B) (5, 6, 2224). Insight into how 2-Cys Prxs distinguish oxidative stress from signal transduction comes from recent in vivo studies showing that human PrxI and PrxII are inactivated under conditions of peroxide signaling (25, 26) by irreversible over-oxidation of the peroxidatic cysteine to sulfinic acid (CP-SO2H) (shown as the “inactivation shunt” in Fig. 1B). Similar inactivation has been seen in all in vitro characterizations of 2-Cys Prxs from mammals (2729), plants (30, 31), and yeast (32, 33). In contrast, we have inferred that bacterial 2-Cys Prxs must be much more robust, as they are typically assayed without problem at peroxide concentrations as high as 1 mM (3436). Proving this point, the inactivation of the bacterial 2-Cys Prx from Salmonella typhimurium, AhpC (StAhpC), is a simple exponential process (Fig. 1C) requiring about 100-fold more hydrogen peroxide than does human PrxI (Fig. 1D). Thus, the exquisite sensitivity to inactivation by peroxide is a feature of Prxs that has been selected for in organisms that use peroxide for signaling.

Fig. 1.

Sensitive and robust 2-Cys Prxs. (A) Ribbon diagram of the AhpCC46S homodimer adopting the fully folded conformation. Subunits are colored green and blue, and side chains (red, oxygen; yellow, sulfur) are shown for Ser46 (replacing CP in this mutant) and Cys165 (CR). The dimer is required for activity, as a local unfolding allows CP from one subunit to form a disulfide bond with CR from the other subunit. (B) The catalytic cycle of 2-Cys Prxs. The peroxidatic cysteine is depicted as a thiol (SPH), sulfenic acid (SPOH), sulfinic acid (SPO2H), or in a disulfide with the resolving cysteine (SRH). Colors distinguish the cysteines from different subunits of the dimer, and the striped bar is the intersubunit disulfide bond. The disulfide reductase varies with organism, often being a thioredoxin or flavoenzyme containing a thioredoxin-like domain. (C) Resistance of bacterial AhpC to inactivation by H2O2. The time course of AhpC-dependent NADPH oxidation is shown for H2O2 concentrations of 0 mM (⚫), 0.1 mM (▢),1 mM (◯), 5 mM (Δ), 10 mM (♢), and 30 mM (◼). The inset shows that the enzyme activities decrease according to a simple exponential law. (D) Peroxide dependence of inactivation of robust (AhpC, ◼) and sensitive (PrxI, ⚫) Prxs. Rates of inactivation at each concentration of hydrogen peroxide were combined with the turnover number (0.08 ± 0.02 s-1 for StAhpC, from three independent experiments) to compute the fraction of enzyme inactivated during each catalytic cycle. For PrxI, the data and a turnover number of 0.35 s-1 were derived from fig. 1 of Yang et al. (26). The slopes of the best-fit lines (corresponding to kSO2H/kSSKLU) are 1.4 ± 0.3 M-1 and 162 M-1 for StAhpC and PrxI, respectively. The slope of the line for StAhpC should be considered an upper limit, because at the higher concentrations of peroxide other factors like inactivation of thioredoxin or thioredoxin reductase may also contribute to the loss of activity.

To discover the origins of this sensitivity, we have compared sequences and structures of sensitive versus robust 2-Cys Prxs (Fig. 2A). The sequence alignment reveals two features uniquely present in all the sensitive 2-Cys Prxs: (i) in the loop between α4 and β5, there is a three-residue insertion associated with a conserved Gly-Gly-Leu-Gly sequence (the GGLG motif) and (ii) there is an additional helix (α7) present as a C-terminal extension associated with a conserved Tyr-Phe sequence (the YF motif). A database search revealed that both features are common to typical 2-Cys Prxs from most eukaryotes, with some parasitic protozoans as exceptions, such as kinetoplastids, plasmodia, and giardia. These parasites and certain proteobacteria and cyanobacteria have a variety of intermediate sequences, for which the susceptibility to over-oxidation is not known.

Fig. 2.

Structural differences between robust and sensitive 2-Cys Prxs. (A) Structure-based sequence alignment of seven 2-Cys Prxs that have been documented to be sensitive (top group of sequences) versus three 2-Cys Prxs that have been documented to be robust (lower group of sequences) (41). Secondary structure elements are depicted as helices (cylinders) or strands (arrows), with structural differences as disconnected elements (yellow or cyan). The shifted 310 helix and loop (yellow) and regions that undergo conformational rearrangements during catalysis (cyan) are identified. The GLGG and YF motifs are highlighted in red, and the peroxidatic (CP) and resolving (CR) cysteines are depicted. Sensitive 2-Cys Prx sequences shown are human PrxI (28), PrxII (28, 29), PrxIII (27, 28), and PrxIV (28), yeast TSA1 (33), yeast TPxY (32), and plant BAS1 (30), and robust 2-Cys Prx sequences shown are AhpC from S. typhimurium (S.t.) (34), Streptococcus mutans (S.m.) (35) and Escherichia coli (E.c.) (36). (B) Ribbon diagrams of the peroxidatic active site region of sensitive 2-Cys Prx structures for the FF and the LU conformations. Elements are colored as in (A), with side chains for CP,CR, and the YF motif included. (C) Same as (B), but for a robust enzyme. Structures shown are as follows: (B) FF, SO2H form of human PrxII (PDB entry 1QMV); (B) LU, disulfide form of human PrxI (pdb entry 1QQ2); (C) FF, C46S form of StAhpC (this work); and (C) LU, the disulfide form of StAhpC (PDB code 1KYG). Asterisks marking the ends of the visible C-termini of the LU conformations indicate that many additional C-terminal residues are disordered. Although PrxIISO2H is an inactive enzyme, it is known to adopt the same conformation as active reduced 2-Cys Prxs (22). Comparison of the FF panels of (B) and (C) provides the most dramatic image of how the two sequence features common to sensitive enzymes result in a major difference in the burial of the Cp-containing helix.

Structural studies of 2-Cys Prxs have revealed two stable conformations that are correlated with the catalytic cycle (5, 6). All the states that do not contain a disulfide are present in the “fully folded” (FF) conformation (22, 23), with CP located on a helix in a highly conserved peroxidatic active site and CR buried 14 Å away (Fig. 1A). In contrast, the disulfide state adopts a “locally unfolded” (LU) conformation (24, 37), in which local unfolding of the C-terminus and the active site helix allow CP-SOH and CR-SH to react and form a disulfide. In fully folded human PrxII (22), the loop containing the GGLG motif and the helix containing the YF motif pack next to each other and bury the active site helix containing CP (Fig. 2B). For comparison, we determined the structure for the fully folded conformation of the robust bacterial enzyme StAhpC. No structural features equivalent to those in the eukaryotic enzyme cover the active site helix (Fig. 2, B and C). An analysis of the B-factors shows that this structural difference changes the dynamics, with StAhpC having a much higher intrinsic disorder at the C-terminus and in the peroxidatic active site pocket (Fig. 3). In the eukaryotic enzyme, the peroxidatic active site (i.e., CP), the loop containing the GGLG motif, strand β8, and strand β9 (containing CR) are all covered by the C-terminal helix and have lower and correlated B-factors (fig. S1). The much lower mobility (lower B-factors) of PrxIISO2H shows that the presence of the C-terminal helix (with its conserved, buried YF motif) results in a tighter structure, indicating that it serves as a keystone that strongly stabilizes this whole region of the structure.

Fig. 3.

The insertions in sensitive 2-Cys Prxs create a more rigid structure. Relative B-factors of AhpCC46S (red) and PrxSO2H (green) are plotted against residue number. Helices (cylinders) and strands (arrows) are depicted, with elements not structurally conserved shown above (AhpCC46S) or below (PrxSO2H) the midline. The 310 helix in the α4-β5 loop is shown as a striped cylinder, and the C-terminus is in cyan. CP and CR cysteines and the four regions (A to D) that differ in mobility due to the packing of the C-terminus are indicated. The packing of regions A to D can be seen in Fig. 2B and fig. S1.

The stabilization of the fully folded conformation in sensitive 2-Cys Prxs can explain their sensitivity to over-oxidation (Fig. 4). A conformational change from fully folded (FF) to locally unfolded (LU) is required for disulfide bond formation, and this has an equilibrium constant KLU. The rate constants for disulfide bond formation and for over-oxidation by peroxide are kSS and kSO2H, respectively. Then the rate of catalytically productive disulfide formation is kSS[ELU] (where [ELU] is the concentration of locally unfolded enzyme) and the rate of inactivating over-oxidation is kSO2H[H2O2][EFF]. (The reverse reactions for these two steps are negligible under physiological conditions.) Putting these together, the fraction of protein that will be inactivated each catalytic cycle is given by Math1 which for finact < 0.1 can be approximated as the linear relation Math2 For both human PrxI and StAhpC, the dependence of finact on [H2O2] in the ranges studied is indeed linear, with the enzymes differing in sensitivity by a factor of ∼110 (Fig. 1D). This difference could be due to changes in kSO2H, kSS, and/or KLU. Evidence that the two rate constants are similar for sensitive and robust 2-Cys Prxs is that the enzymes have highly conserved active sites (implying kSO2H is similar) and also have conserved positions of CP and CR (implying kSS changes little). On the basis of the structural evidence, we suggest that stabilizing the fully folded conformation (i.e., optimizing KLU) is the most important evolutionary variable in this case. This model predicts that any changes that disrupt the packing of the C-terminus of a sensitive enzyme will increase KLU and decrease the rate of inactivation. Biochemical support for the model comes from a mutagenesis study on the role of the C-terminal extension of the yeast 2-Cys Prx, TpxY (32). That work showed that the C-terminus could be highly altered with little effect on enzyme activity but that almost any change caused the normally sensitive enzyme to become robust to inactivation by over-oxidation (see fig. S2).

Fig. 4.

Model for quantitative correlation of the structure and sensitivity of mammalian 2-Cys Prxs. Using cartoon images of a sensitive 2-Cys Prx structure, the catalytic cycle is shown including an explicit step for the local unfolding required for disulfide formation. The FF and LU conformations are labeled, with dotted lines representing chain segments in their more mobile, locally unfolded conformations. The peroxidatic cysteine is shown in one of three possible redox states, as an SH, SOH, or SO2H. The C-terminal helix present only in sensitive 2-Cys Prxs is shown as a blue circle in the FF conformation. Relevant chemical rate constants (kSO2H and kSS) and the equilibrium constant for local unfolding (KLU) are labeled. For reactions that are effectively irreversible under physiological conditions, only a forward arrow is shown. Although this scheme only shows hydrogen peroxide as a substrate, alternate substrates like alkyl hydroperoxides or peroxynitrite may also contribute to enzyme inactivation in vivo.

Thus, the sensitivity of 2-Cys Prxs to inactivation by over-oxidation is not a limitation of the 2-Cys Prx catalytic mechanism, but it is a feature that has been selected for during the evolution of eukaryotes. Recent in vivo observations are consistent with the following model of 2-Cys Prx function in eukaryotes. In a resting cell, 2-Cys Prxs are present in large amounts (38) and keep ambient levels of peroxides very low to ensure that no signaling is triggered. In contrast, a transient intracellular peroxide burst (1) would produce peroxide at levels sufficient to rapidly inactivate 2-Cys Prxs, as has been observed in vivo during peroxide signaling in cells treated with tumor necrosis factor (25). Then peroxide could act as a messenger by interacting with other proteins, like the recently characterized yeast oxidant receptor peroxidase 1 (39) and peroxide-sensitive phosphatases (2). At this stage, the activities of the less abundant antioxidant enzymes, glutathione peroxidase and catalase, would be crucial to “mop up” excess peroxide and mitigate against its toxic effects. When too much 2-Cys Prx is present, the peroxide burst would not be sufficient to inactivate all of it, and signaling would be blocked as is seen in cells that overexpress 2-Cys Prxs and block peroxide activation of NF-κB(11, 12). This provides a rationale for the correlation in some cancer cells between defective apoptotic pathways and overexpression of 2-Cys Prxs (20, 21).

In this model, 2-Cys Prxs act as a peroxide floodgate, keeping peroxides away from susceptible targets until the floodgate is opened. The sensitivity to inactivation by substrate provides an irreversible way to open the floodgate, but this is undoubtedly not the only way in which 2-Cys Prxs are regulated. For example, a recent report indicates that cyclin-dependent kinases inactivate human PrxI by phosphorylation on Thr90 at specific stages of the cell cyle (40). This provides a rapid, reversible way in which the floodgate may be opened without the inefficient generation of irreversibly inactivated protein. It may be that opening the floodgate through over-oxidation only occurs during signaling events leading to apoptosis, and in that context the loss of the peroxiredoxin would not be wasteful. Thus, it appears that eukaryotic evolution has resulted in a peroxidase that is less robust than that found in many prokaryotes but has gained a regulatory feature that facilitates peroxide signaling.

Supporting Online Material

Materials and Methods

Figs. S1 to S3

Table S1

References and notes

View Abstract

Stay Connected to Science

Navigate This Article