Structural basis of cellular redox regulation by human TRP14.

Thioredoxin-related protein 14 (TRP14) is involved in regulating tumor necrosis factor-alpha-induced signaling pathways in a different manner from human thioredoxin 1 (Trx1). Here, we report the crystal structure of human TRP14 determined at 1.8-A resolutions. The structure reveals a typical thioredoxin fold with characteristic structural features that account for the substrate specificity of the protein. The surface of TRP14 in the vicinity of the active site includes an extended loop and an additional alpha-helix, and the distribution of charged residues in the surface is different from Trx1. The distinctive dipeptide between the redox-active cysteines contributes to stabilizing the thiolate anion of the active site cysteine 43, increasing reactivity of the cysteine toward substrates. These structural differences in the active site suggest that TRP14 has evolved to regulate cellular redox signaling by recognizing a distinctive group of substrates that would complement the group of proteins regulated by Trx1.

Thioredoxin (Trx) 1 is a small redox protein that is ubiquitously distributed from Archaes to human (1). In diverse organisms, it plays various physiological roles, acting as an electron donor and as a regulator of transcription and apoptosis as well as antioxidants (1). Trx functions as a carrier that transfers electrons to enzymes involved in DNA synthesis and protein disulfide reduction: ribonucleotide reductase, methionine sulfoxide reductase (2,3). Transcription factors such as NF-B, p53, PEBP32, and AP-1 have been shown to be regulated by Trx (4 -6). Trx is also involved in the defense against oxidative stress through peroxiredoxins (Prxs) that directly remove cellular reactive oxygen species. Other functions of Trx include the regulation of apoptosis signaling kinase 1 (ASK1) to inhibit its activity (7,8).
Sequences of Trxs in various species are 27-69% identical to that of Escherichia coli, and all Trxs have the same threedimensional fold, which consists of central five ␤-strands surrounded by four ␣-helices (9,10). The N-terminal cysteine of the Cys-Gly-Pro-Cys motif in Trx is redox-sensitive, and the motif is highly conserved. When compared with the general cysteine, the N-terminal cysteine has a lowered pK a value (11). Various proteins sharing the Trx-like active site sequence have been found and classified as part of the Trx superfamily (12). Among them, thioredoxin-related protein (TRP32) with the N-terminal thioredoxin domain has been found to bind to the catalytic fragment of mammalian STE-20-like kinase (13). The active site sequence Cys-Gly-Pro-Cys of TRP32 is identical to that of thioredoxin. In contrast, nucleoredoxin (14) and thioredoxin-related transmembrane protein (15), which seem to be involved in the various redox regulation, have modified sequences Cys-Pro-Pro-Cys and Cys-Pro-Ala-Cys, respectively.
Recently, a novel thioredoxin-related protein (TRP14) was found from rat brain (16,17). It is a 14-kDa cytosolic protein that contains a Cys-Pro-Asp-Cys motif with a sequence identity to human Trx (Trx1) of about 20%. TRP14 shows activity as disulfide reductase and takes up electrons from cytosolic thioredoxin reductase (TrxR1), like Trx1. However, TRP14 does not donate electrons to ribonucleotide reductase, methionine sulfoxide reductase, and Prxs, which are well known substrates of Trx1. TRP14 regulates TNF-␣-induced signaling pathways in a different manner from Trx1. In RNA interference experiments, the depletion of TRP14 increased TNF-␣-induced phosphorylation and degradation of IB␣ more than the depletion of Trx1 did. TRP14 also facilitated activation of JNK and p38 MAP kinase induced by TNF-␣. Unlike Trx1, TRP14 shows neither interaction nor interference with ASK1.
Here, we determined the crystal structure of human TRP14 to understand the structural mechanisms for the substrate specificity of the protein and the biological role. The structure of TRP14 reveals a special surface topology with unique electrostatic surface properties near the active site that is quite different from Trx1. The distinctive surface properties of TRP14 explain the mechanism by which the target specificity is different between TRP14 and Trx1. In addition, the structural study of TRP14 provides us with information on the diversity and specificity of cellular redox regulation in the cytosol that are achieved by two distinctive Trx isotypes, TRP14 and Trx1. 6 FF were dialyzed against buffer B (10 mM Tris (pH 7.5), 2 mM dithiothreitol, and 1 mM EDTA) and applied to the Mono-Q column. The purified protein was dialyzed against 10 mM Hepes (pH 7.00) and concentrated to 20 mg/ml for crystallization.
Crystallization and Data Collection-Crystallization of TRP14 was carried out by using the hanging drop vapor diffusion method at 18°C. The reservoir solution consisted of 10%(w/v) polyethylene glycol 8000, 10%(w/v) polyethylene glycol monomethylether 2000, 0.1 M MES (pH 6.5). 1.8 l of protein solution was mixed with an equal volume of the reservoir solution. Plate-shaped crystals grew at the full size of 1.0 ϫ 0.5 ϫ 0.2 mm in 1-2 days. The TRP14 crystals belonged to the P2 1 2 1 2 1 space group with unit cell dimensions of a ϭ 26.86 Å, b ϭ 48.39 Å, c ϭ 81.84 Å. The crystals contained one monomer in the asymmetric unit with the Matthews coefficient and the solvent content of 1.83 Å/Da and 34%, respectively. The multiwavelength anomalous dispersion data for the structure determination were collected in the beamline 18B at the Photon Factory by using an ADSC CCD detector. Before the data collection, a single crystal was soaked in stabilizing solution supplemented with 0.5 M sodium bromide for about 30 s and flash-frozen. Data for five different wavelengths including two peak wavelengths ( 1 and 2 ), two edge wavelengths ( 1 and 2 ), and a remote wavelength ( 5 ) were collected (Table I). Measurements for multiple peak and edge wavelengths were performed to take account of differences in the absorption edge of bromine between solution and crystalline states. All data were integrated using the program MOSFLM (18) and scaled by using the program SCALA in CCP4 package (19).
Structure Solution, Model Building, and Refinement-The heavy atom search by using the program SnB (20) located two bromide sites, and the heavy atom parameters were refined by using the program SHARP (21). The phases were further improved by solvent flattening by using the program DM (19). A summary of phasing and refinement statistics is in Table I. The resulting electron density map was of sufficient quality for the initial model building. The program O (22) was used for the model building. When most residues in the molecule were built into the model, the model coordinates were transferred to the high resolution data and refined by using the program CNS (23). The randomly selected 5% of diffraction data was set aside for R free calculation. Water molecules were gradually added by the water_pick routine implemented in CNS. In the refined model, 94.2% of all residues in the molecule are in the most favored regions, and only one residue (Glu 107 ) remains in a disallowed region of Ramachandran plot drawn by the program PROCHECK (24). The residue Glu 107 has strong interaction with the residue Asn 81 in the symmetry-related molecule, and the quality of electron density map around Glu 107 is excellent. The final model includes residues 4 -122 of TRP14 and 43 water molecules.
Activity Assay of Redissolved Crystals-TRP14 crystals were redissolved in 1 mM EDTA and 50 mM Hepes-NaOH (pH 7.0) after rinsing the crystals several times with the reservoir solution. The activity assay of TRP14 was started by the addition of oxytocin (Sigma) at a final concentration of 100 M to a reaction mixture containing 50 mM Hepes-NaOH (pH 7.0), 1 mM EDTA, 0.2 mM NADPH, 100 nM TrxR1, and 6 M TRP14. The reaction was performed at 30°C, and oxytocin reduction was monitored on the basis of A 340 .

RESULTS AND DISCUSSION
Overall Structure-The structure of TRP14 was determined by the multiwavelength anomalous dispersion method with a sodium bromide (0.5 M) soaked crystal. TRP14 folds into a compact and globular domain with the approximate dimensions of 25 ϫ 30 ϫ 35 Å (Fig. 1). Despite low sequence identity (20%) between TRP14 and Trx, the topological structure of TRP14 is similar to that of Trx. When we superimposed TRP14 and Trx1 (Protein Data Bank code 1ERU) (10), 81 out of 119 C␣ atoms could be aligned with a root mean square deviation of 2.0 Å (Fig. 2). The redox-active Cys-Xaa 1 -Xaa 2 -Cys motif (Xaa is any amino acid) of TRP14 (Cys 43 -Pro 44 -Asp 45 -Cys 46 ) is located at the N terminus helix ␣2. A disulfide bond between the two cysteines was evident in the electron density map, indicating that the structure is in an oxidized state.
There are several important structural differences between TRP14 and Trx1. First, the region of TRP14 corresponding to helix ␣3 of Trx1 is divided into two separated helices (␣3a and ␣3b) with the insertion of 4 residues (Figs. 1b and 2). In addition, helix ␣3a is shifted toward the face of the redox-active site significantly. A structure-based sequence alignment (Fig. 2) shows that residues 73-87 including helix ␣3a and a part of helix ␣3b in TRP14 cannot be aligned with Trx1. Helix ␣3a in TRP14 interfering with the surface near the redox-active site may play a role in determining target specificity by limiting its interaction with Trx target proteins (see below).
where F PH , F P , and F H are derivative, native and heavy-atom structure factor amplitudes, respectively.
where F o and F c are the observed and calculated structure factor amplitudes, respectively. e The R free value was calculated from 5% of all data that were not used in the refinement.
Other structural differences are found in loop ␤2-␣2 (residues 33-42) that have an insertion of 7 residues in comparison with Trx1 (Fig. 2). The extended loop of TRP14 is located in front of the redox-active Cys-Xaa 1 -Xaa 2 -Cys motif and is likely to play a role in target recognition or specificity as the helices ␣3a and ␣3b. The extended loop appears to be stabilized by Active Site-The active site Cys 43 -Pro 44 -Asp 45 -Cys 46 motif in TRP14 is located at the starting position of helix ␣2. The N-terminal Cys 43 is exposed to solvent, indicating that the thiol group of the residue could be easily accessible for the reaction with substrates. The reactivity and redox potential of Trxrelated proteins are highly dependent on the dipeptide sequence within the Cys-Xaa 1 -Xaa 2 -Cys motif. The dipeptide sequence (Pro-Asp) of TRP14 is unique in comparison with that of Trx1 (Gly-Pro). The difference of the dipeptide sequences between TRP14 and Trx1 seems to affect the pK a of the first cysteine that is an important indicator of the reactivity of the redox-active cysteines. The cysteine pK a of the Trx-related proteins ranges between 3.0 and 7.0, being significantly lower than that of the natural cysteine thiolate (ϳ8.5). The lowered cysteine pK a is essential for the redox reaction because the ionization of the cysteine sulfide increases reactivity toward substrates by increasing the nucleophilicity of the cysteine.
In TRP14, the Cys 43 thiolate anion is stabilized by the partial positive charge of the helix dipole of helix ␣2 as in Trx1. However, the strength of stabilization seems to be different between TRP14 and Trx1. In particular, the unique dipeptide (Pro-Asp) sequence between the two cysteines seems to contribute to the low pK a of the Cys 43 thiolate by strengthening the helix dipole that is mainly governed by main chain hydrogenbonding network of the first two turns of the helix. In Trx1, the dipeptide sequence is Gly-Pro, in which the position of proline is different from that of TRP14. The proline residue of Trx1 in the Xaa 2 position results in the disruption of a hydrogen bond between the amide nitrogen and the N-terminal cysteine sulfur. In comparison, the Asp in the Xaa 2 position of TRP14 does not disrupt the hydrogen bond (Fig. 3). The proline of TRP14 in the Xaa 1 position does not affect the hydrogen-bonding network because the main chain amide nitrogen of Xaa 1 is not in the geometry to make a strong hydrogen bond with the N-terminal cysteine sulfur. The strengthening of the helix dipole in TPR14 is consistent with the biochemical measurement of the pK a value of the N-terminal Cys 43 to be 6.1 (16), which is lower than that of the corresponding cysteine of Trx (ϳ7.0). The structure of an E. coli disulfide bond exchange protein DsbA exhibits a similar phenomenon (25). DsbA is known for having an extremely low pK a value of 3.4. The dipeptide sequence within the Cys-Xaa 1 -Xaa 2 -Cys motif of DsbA (Cys-Pro- His-Cys) is similar to those in TRP14, in which the position of Pro is the same with TPR14 and a non-proline residue occupies the Xaa 2 position. In the crystal structure of reduced DsbA, the thiolate in Cys 30 (corresponding to Cys 43 in TRP14) forms hydrogen bonds with amide groups of Xaa 1 and Xaa 2 residues. Although the side chain of His 32 plays a major role in lowering the pK a value by forming a charge interaction with the thiolate, mutations of His 32 to non-charged residues such as Tyr and Ser still represent low pK a values ranging from 3.8 to 4.9, indicating that the non-proline residue in the Xaa 2 position plays a critical role in lowering the pK a of the first cysteine in the CXXC motif (26).
Surface Characteristics-The surface surrounding the active site of TRP14 includes the protruding helix ␣3a and loop ␤2-␣2 (Fig. 4a), which is highly distinguishable from the wide and flat surface of Trx1 and Grx (Fig. 4, b and c). The wide and flat surface of Trx1 and Grx would be advantageous for interacting with a variety of substrates that need to be reduced by Trx1 and Grx. In comparison, the specialized molecular surface of TRP14 is likely to restrict the interaction of many proteins that do not have a molecular surface complementary to the specialized TRP14 surface. In the complex structures between Trx1 and peptide substrates that originated from Ref-1 (27) and NF-B (28), the target peptides bind to the crescent-shaped groove on the surface of the Trx1. In the peptide binding interactions, the side chain of Tyr 60 in the NFB peptide corresponding to the P-2 binding site (P0 denotes the active cysteine of the peptide, and the negative number indicates a residue N-terminal to the cysteine residue) has many hydrophobic interactions with various residues (Phe 27 , Val 59 , Ala 66 , Val 71 , Ala 73 , and Thr 74 ) in Trx1. Although the Ref-1 peptide is complexed with Trx in a reversed chain direction (27), the P-2 site is replaced by bulky tryptophan in the Ref-1 peptide, and its hydrophobic interactions are still similar to those of the NFB complexed structure. However, in TRP14, the protruding helix ␣3a completely covers the P-2 binding site and thus blocks the access of Trx substrates.
In addition to the molecular surface differences, the charge distribution on the TRP14 active site surface is completely different from that of Trx1. Although the Trx1 surface is mainly made of hydrophobic patches, that of TRP14 has highly charged characters with both positively charged regions (Lys 40 , Lys 74 , Lys 78 , and Arg 85 ) and negatively charged regions (Asp 36 , Asp 45 , Glu 48 , and Glu 107 ). Among the charged residues, the Xaa 2 residue (Asp 45 ) of the Cys-Xaa 1 -Xaa 2 -Cys motif plays an important role to form a negatively charged surface patch. Several hydrophobic residues (Trp 42 , Val 71 , Trp 77 , Pro 94 , Val 93 , and Leu 111 ) are spread out in between the charged regions. The highly charged surface of TRP14 should act as a determinant for the specific recognition of potential substrates. The surface of Grx also has a small patch of positively charged residues (Fig. 4c), but the extent of the charged region is significantly less than that of TRP14, and the distribution of the charged residues is completely different between the two proteins. Thus, TRP14 seems to have evolved to recognize a completely different set of substrates that are not reduced by Trx1 or Grx.
The significant structural and surface differences between TRP14 and other cellular reductants led us to test whether the crystallized TRP14 retains its biological activity. We redissolved the TRP14 crystals and measured the reductase activity of the protein by using oxytocin as the substrate (Fig. 5). Oxytocin, which comprises 9 amino acids with an intramolec- ular disulfide, was shown to be a substrate of the soluble TRP14 (16). In Fig. 5, the TRP14 protein redissolved from crystals exhibits almost the same activity as the soluble TRP14, indicating that the TRP14 conformation observed in the crystal represents the biologically active state.
Substrate Specificity-Trx-related proteins function as electron shuttles by transferring electrons (or reducing power) from donors to acceptors. In the process, the proteins must interact with both electron donors and acceptors in a sequential manner. TRP14 accepts electrons from Trx reductase 1 (TR1) but not from Trx reductase 2 (TR2), whereas Trx1 does from both TR1 and TR2 (16). The differences in specificity for the electron donation of TRP14 and Trx1 are more pronounced. That is, there seem to be no common substrates between TRP14 and Trx1. For example, Trx transfers electrons to ribonucleotide reductase, methionine sulfoxide reductase, and Prxs, whereas TRP14 reduces none of the proteins. Instead, TRP14 was shown to interact with LC18 (17). These distinctive substrate specificities should be accounted for by the surface differences described in the previous section.
Of interest is the ability of TRP14 to accept electrons from TR1, as does Trx1, although the surfaces near the redox-active site of the two proteins are completely different. The reactivity toward TR1 may be due to the C-terminal extension of TR1 with selenocysteine, which would transfer electrons from the TR1 active site to substrates (29). The flexibility of the Cterminal extension would enable TR1 to accommodate both TRP14 and Trx1 as substrates (29). Although TR2 also has the selenocysteine-containing C-terminal extension, the cytosolic protein TRP14 seems not to have evolved to interact with TR2, which is a mitochondrial protein.
There were previous examples in which the substrate specificity of redox proteins was attributed to differences in the interacting surface characteristics. The Grx-homologous protein NrdH exhibits Trx1-like activity because the active site surface of NrdH has hydrophobic characteristics like Trx1 (30). The crystal structure of Haemophilus influenzae hybrid-Prx5 showed that the specificity of electron donors was determined by the binding surfaces of Prx isotypes (31). Although sequences of human Prx5 and H. influenzae hybrid-Prx5 are highly homologous, the active site surfaces are different, and the two proteins have different substrate specificities in which human Prx5 and H. influenzae hybrid-Prx5 use Trx and Grx as electron donors, respectively. These results support that the active surface of redox proteins affects the substrate specificity of the proteins and the specialized surface of TRP14 provide the protein with a highly specific selectivity in binding with its potential substrates.
Implication for Cellular Redox Regulation-Redox homeostasis is important in cellular functions including signal transduction, transcription, immune responses, and apoptosis (32). The functional regulations by cellular redox status are mediated by oxidation of exposed cysteine residues by reactive oxygen species, and the reduction of the oxidized cysteines is done by Trx-related proteins. Although the conventionally known reducing agent Trx1 has broad substrate specificity, it would be an impossible task for a protein to recognize all of the redoxregulated proteins in cells. Thus, mammalian cells should have auxiliary factors that could complement the function of Trx1. Until today, there were only three known Trx-related proteins (Trx1, TRP32, and TRP14) in cytosol, where most cellular signaling occurs. TRP14 is a good candidate to play an active role as a redox regulator because the protein accepts electrons from TR, whereas TRP32 does not (33). The structure of the cytosolic protein TRP14 reveals a charged and uneven active site that is very different from that of Trx1. However, the completely different active site surface may be complementary in the function of two proteins as redox regulators by extending the range of substrates recognized by the redox regulators. In addition to the surface differences, TRP14 and Trx1 also exhibit differences in hydrogen-bonding interactions of the redoxactive cysteines due to a different dipeptide sequence between the two cysteines, which explains the different redox properties of TPR14, such as cysteine reactivity (pK a ) and redox potential. Thus, the TRP14 structure shows that the protein is designed to act as both an auxiliary factor for Trx1 and a specific redox regulator for cellular redox signaling pathways.