A universal entropy-driven mechanism for thioredoxin–target recognition

Significance

Thioredoxin (Trx) is universally conserved thiol-oxidoreductase that regulates numerous cellular pathways under thiol-based redox control, and its activity is often upregulated in malignant cancer cells. Despite its central importance, the mechanism by which Trx recognizes its target proteins remains unknown. Herein, we address this longstanding question by investigating the noncovalent forces involved in Trx–target interactions. Using a combination of biochemical and quantitative biophysical methods, we identify favorable entropy as the major force in molecular recognition that drives target specificity. These findings, and others reported herein, afford considerable new insight into Trx–target recognition, which is critical to understanding its function in normal metabolism and represents a fundamental step toward the development of new pharmacological strategies to address redox-related disorders.

Abstract

Cysteine residues in cytosolic proteins are maintained in their reduced state, but can undergo oxidation owing to posttranslational modification during redox signaling or under conditions of oxidative stress. In large part, the reduction of oxidized protein cysteines is mediated by a small 12-kDa thiol oxidoreductase, thioredoxin (Trx). Trx provides reducing equivalents for central metabolic enzymes and is implicated in redox regulation of a wide number of target proteins, including transcription factors. Despite its importance in cellular redox homeostasis, the precise mechanism by which Trx recognizes target proteins, especially in the absence of any apparent signature binding sequence or motif, remains unknown. Knowledge of the forces associated with the molecular recognition that governs Trx–protein interactions is fundamental to our understanding of target specificity. To gain insight into Trx–target recognition, we have thermodynamically characterized the noncovalent interactions between Trx and target proteins before S-S reduction using isothermal titration calorimetry (ITC). Our findings indicate that Trx recognizes the oxidized form of its target proteins with exquisite selectivity, compared with their reduced counterparts. Furthermore, we show that recognition is dependent on the conformational restriction inherent to oxidized targets. Significantly, the thermodynamic signatures for multiple Trx targets reveal favorable entropic contributions as the major recognition force dictating these protein–protein interactions. Taken together, our data afford significant new insight into the molecular forces responsible for Trx–target recognition and should aid the design of new strategies for thiol oxidoreductase inhibition.

Thioredoxin (Trx) is a small 12-kDa redox-active protein ubiquitously expressed from bacteria to mammals with a universally conserved dithiol (-Cys-Gly-Pro-Cys-) active site. Together with NADPH and the flavoprotein, thioredoxin reductase (TrxR), it constitutes one of the central antioxidant systems in the cell (1, 2). Collectively, this redox couple functions by relaying reducing equivalents from NADPH to the dithiol active site of Trx, which subsequently uses them to reduce disulfide bonds in a wide variety of target proteins (Fig. 1). The essentiality of Trx in cell function and viability has been demonstrated by severe developmental disorders in the early mouse embryo and embryonic lethality in Trx KO mice (3). Additionally, Trx has been found to be expressed at relatively high levels in many human malignancies, including lung, pancreatic, colon, gastric, and breast cancer (4).

Fig. 1.

Physiological function and pathological significance of Trx. (A) Mechanism for general disulfide reduction catalyzed by Trx. (B) Trx–protein interactions listed in conjunction with their physiological function and pathological implications. Trx reduces RNR with essential function in DNA synthesis and repair (5), MSrs linked to protein repair and aging (6) and Prxs, implicated in multiple redox-related disorders (7). Human Trx (Trx1) activates transcription factor Ref-1 critical in cell growth and survival (6) and inhibits proapoptotic protein ASK1 (9). Extracellular Trx1 reduces HIV envelop protein gp120 (10) and TG2, an enzyme implicated in celiac disease (11). Bacterial Trx reduces SRs with a crucial role in sulfur assimilation (8).

Trx was first identified by virtue of its ability to reduce ribonucleotide reductase (RNR), an enzyme involved in DNA synthesis and repair (5). Since this discovery, Trx has been found to recognize and regulate the activity of a myriad of proteins in both prokaryotic and eukaryotic organisms (Fig. 1B). For example, Trx activates methionine sulfoxide reductases (MSrs), enzymes involved in protein repair and aging (6), and peroxidoxins (Prxs), a family of enzymes that converts H₂O₂ to water (7). Bacterial Trx is essential to the catalytic cycle of sulfonucleotide reductases (SRs), enzymes involved in reductive sulfur assimilation and validated targets for antibacterial drugs (8). In humans, cytosolic Trx (Trx1) inhibits the activity of proapoptotic protein, ASK-1 and the tumor suppressor protein, PTEN (4, 9). On translocation to the nucleus, Trx1 activates a number of transcription factors regulating various aspects of cell growth and survival, such as Ref-1, p53, NF-κB, and Hif-1α (6). Recently, extracellular Trx1 was found to interact with HIV envelop glycoprotein, gp120, and transglutaminase 2 (TG2), an enzyme associated with celiac disease (10, 11). The above examples, and other reports, clearly show that Trx regulates a wide range of biological processes and also implicate the oxidoreductase in many human pathologies, such as cancer, inflammatory and neurodegenerative diseases, as well as cardiovascular disorders (4). Modulation of Trx function has also emerged as a potential therapeutic strategy in treatment of various diseases (12).

Numerous proteomic surveys have documented the specificity of Trx in target recognition (13–15). However, the molecular mechanism underlying this process, particularly in the absence of a signature Trx-binding sequence or structural motif, has remained largely unknown. Periodic efforts have been made to study individual Trx–target interactions through X-ray crystallography (16–18). In these studies, the Trx–protein complex has been trapped via the mixed disulfide that forms between the target and a Trx mutant, in which the resolving cysteine is mutated to an alanine or serine. Although the mixed disulfide represents an important step during the redox cycle of Trx, it depicts a state well past the initial step of noncovalent association. Interestingly, single-molecule force microscopy studies on disulfide reduction have demonstrated Trx-dependent subangstrom level changes in the disulfide geometry and lengthening of the disulfide before its reduction, thereby underscoring the importance of noncovalent association in Trx–target interaction (19). To our knowledge, there are only two prior studies investigating the noncovalent interaction between Trx and its target proteins (20, 21). However, to date, there has been no comprehensive study to characterize and compare the thermodynamics of noncovalent Trx–target interactions, thus leaving a large gap in our fundamental understanding of this important class of protein–protein interactions.

Herein, we study the molecular features that underlie noncovalent Trx–target recognition using ITC. Our data reveal that Trx recognizes the oxidized form of target proteins with high selectivity and that this interaction is primarily governed by the favorable entropy of binding. Importantly, our results establish that such entropy-driven recognition is a common mechanism used by Trx to interact with targets from other functional protein classes evaluated in this study. Furthermore, our findings suggest that recognition depends, in large part, on the conformational restriction inherent to the oxidized target. Taken together, our findings significantly advance our understanding of Trx-mediated recognition and regulation of target proteins.

Results

To investigate the noncovalent recognition forces involved in Trx–target interactions, we selected Escherichia coli 3′-phosphoadenosine-5′-phosphosulfate (PAPS) reductase or EcPAPR as our first model protein. EcPAPR functions as a facile model for Trx targets owing to its stability and ability to be purified in large quantities to meet the demands of ITC studies (SI Appendix, Fig. S1). PAPR catalyzes the reduction of PAPS to 3′-phosphoadenosine-5′-phosphate (PAP) and sulfite (SO₃^-2). In a well established two-step mechanism, the sulfate group of PAPS undergoes nucleophilic attack by the reactive thiolate of PAPR to give the enzyme-S-sulfocysteine (E-Cys-S-SO₃⁻) intermediate, followed by Trx-dependent cleavage of the S-S bond to release SO₃^-2 (Fig. 2) (8).

Fig. 2.

PAPR catalytic mechanism. The sulfur atom of PAPS undergoes nucleophilic attack by C239 of PAPR resulting in formation of E-Cys-S-SO₃⁻ intermediate, followed by release of SO₃^2- vis-à-vis Trx-mediated thiol-disulfide exchange (8).

Trx Selectively Binds to Oxidized PAPR.

In our initial studies, we investigated the binding of E. coli Trx to the reduced form of wild-type (WT) EcPAPR (i.e., PAPR-SH). Interestingly, at midmicromolar concentrations of both proteins, no measurable binding was observed. Indeed, only weak binding was observed even at millimolar concentrations of Trx, yielding a lower limit for the intrinsic binding affinity of ∼110 µM (Fig. 3). A comparison of the crystal structures of apo-PAPR [Protein Data Bank (PDB) ID 1SUR], the noncovalent PAPR•PAP complex (PDB ID 2OQ2) and the covalent PAPR-Trx complex (PDB ID 2O8V), reveals significant differences in the active site architecture (17, 22, 23). For example, the otherwise disordered C-terminal tail residues (∼30 amino acids) including the catalytic cysteine (C239) fold over the substrate-binding pocket upon ligand binding (SI Appendix, Fig. S2). Such structural rearrangements induced by PAPS binding are essential to establish the full complement of enzyme–substrate interactions as well as to position C239 for nucleophilic attack. The orderliness of C-terminal tail residues in the substrate-bound state is further substantiated by our limited trypsin proteolysis studies in solution in presence or absence of the substrate PAPS (SI Appendix, Fig. S3). Given the conformational changes that accompany substrate binding, we next tested Trx binding to oxidized PAPR or PAPR-S-SO₃⁻. In the absence of Trx, the addition of saturating PAPS to reduced PAPR led to full conversion to the stable PAPR-S-SO₃⁻ intermediate as verified by intact mass analysis (SI Appendix, Fig. S4). However, measurement of purely noncovalent interactions between PAPR-S-SO₃⁻ and Trx is not possible in this state because Trx reduces the S-S bond through covalent thiol-disulfide exchange reaction (Fig. 2). As Trx shows only very weak binding to PAPR-SH, this essentially irreversible process would perturb the equilibrium between PAPR-S-SO₃⁻, Trx and the noncovalent Trx•PAPR-S-SO₃⁻ complex, thereby introducing errors in the measured binding constant and stoichiometry. To address this issue and isolate the noncovalent binding interaction between PAPR-S-SO₃⁻ and Trx, we used a variant of Trx in which the catalytic cysteine is changed to alanine (C32A). If Trx recognizes PAPR-S-SO₃⁻, then C32A Trx is predicted to bind, but not reduce, the target protein. Consistent with this model, we observed that C32A Trx bound to PAPR-S-SO₃⁻ with a binding affinity (K_d) of 1.5 µM in 1:1 molar complex (Fig. 3B). By contrast, C32A Trx displayed very weak binding to PAPR-SH, analogous to WT Trx. NMR analyses have reported negligible structural variation between the WT and the active-site cysteine mutant of Trx (24). Additionally, C32S Trx displayed equivalent affinity for PAPR-S-SO₃⁻, compared with C32A, indicating similar binding behavior for WT and C32A/S Trx variants. Interestingly, interaction with its target proteins like gene-5-protein (g5p) of T7-DNA polymerase and ASK1 has been suggested to be dependent on Trx redox state, with oxidized Trx (Trx-S₂) showing no binding whatsoever to the target proteins (9, 20). To evaluate this proposal, we next tested the binding of Trx-S₂ to PAPR-S-SO₃⁻. The resulting interaction data indicates that Trx-S₂ retains its affinity for oxidized PAPR (SI Appendix, Fig. S5). In hindsight, this observation can be rationalized on the basis of essentially identical 3D structures of oxidized and reduced Trx (25). Together, these data show that Trx preferentially recognizes oxidized PAPR and that target binding is independent of Trx redox state.

Fig. 3.

ITC determination of Trx binding to reduced and oxidized PAPR. (A) Pictorial representation of conformational change in PAPR going from reduced to oxidized form after the addition of saturating PAPS. (B) (Upper) Time-dependent deflection of heat signal after each injection of Trx in the microcalorimetric cell containing 70 µM/monomer of homodimeric PAPR in oxidized or reduced form. Reduced PAPR titrated with 2 mM C32A Trx (green). Oxidized PAPR titrated with 0.75 mM C32A Trx (black). ITC experiments were performed in 50 mM bis-Tris propane buffer containing 0.02% Brij-35 (pH 7.4) at 25 °C. (Lower) Integrated calorimetric data corrected for Trx dilution fit to independent binding model (NanoAnalyze software) to obtain binding and thermodynamic parameters. (C) Graphical representation of thermodynamic parameters for the C32ATrx•oxidized PAPR interaction.

Favorable Entropy Drives Noncovalent Association of Trx and PAPR.

The thermodynamic parameters measured for the Trx•PAPR interaction indicate that favorable entropy contributed the most to the free energy of binding (ΔG), whereas favorable enthalpy was only a minor component (Fig. 3C). Interestingly, the 3D structure of the PAPR-Trx covalent complex (PDB ID 2O8V) determined previously by our group predicts an array of noncovalent interactions between PAPR and Trx (Fig. 4 A and B) (17). Trx was found to interact with PAPR accessing two distinct surfaces, an 800 Å site involving PAPR residues 188–199 and 202–212, and another 783 Å site constituting the C-terminal tail residues 235–244. Among the key interactions predicted were a possible H-bonding interaction between W205 of PAPR and E30 of Trx, a salt bridge interaction between D206 of PAPR and K36-E30 of Trx, aromatic π-stacking interactions between W201 and W205 of PAPR as well as Y199 of PAPR and W31 of Trx. Among the C-terminal chain residues, C238 of PAPR was predicted to be involved in interdigitated network of salt bridge interactions with R73 of Trx. To probe the thermodynamic contribution of these individual residues to the noncovalent association of oxidized PAPR and Trx, we performed site-directed mutagenesis. The results from comparative binding analysis of site-directed mutants may often be obscured by enthalpy–entropy compensatory effects (26). Indeed, our initial efforts using alanine mutants of critical residues displayed enthalpy–entropy compensation. However, use of nonalanine mutants, selected based on the polarity of substituted amino acid side chains (27), effectively minimized this issue providing useful thermodynamic data devoid of compensatory effects. All variants of PAPR retained the ability to bind PAPS and form the PAPR-S-SO₃⁻ intermediate (SI Appendix, Fig. S6). Next, we investigated the association of C32A Trx with each PAPR-S-SO₃⁻ variant by ITC (Table 1). Notably, all oxidized PAPR variants were able to associate at some measurable level with Trx and, consistent with our results for WT PAPR, favorable entropy remained as the dominant force for recognition. The W205F and D206E variants showed nearly 10-fold decrease in K_d, compared with WT (Fig. 4C and Table 1). In both variants, ΔG decreased as a direct result of the decrease in binding enthalpy, compared with WT PAPR. By contrast, L210A PAPR, a residue likely participating in hydrophobic interactions with Trx, showed a decrease in ΔG due to a decrease in favorable entropy. The G208A PAPR variant showed a negligible change in ΔG and K_d within the error limits. Among the variants of PAPR residues within the C-terminal tail, E238A showed an 11.5-fold decrease in the binding affinity, resulting from a decrease in favorable enthalpy and entropy components. Overall, the site-directed mutagenesis analyses indicate that the minor enthalpic contribution to the Trx•PAPR interaction can be mapped to individual residues depicted in the X-ray structure. However, Trx remains effective in recognizing PAPR variants like W205F where the favorable enthalpy contribution is drastically reduced by ∼80%. These findings highlight the contribution of favorable entropy as the principle driving force for Trx•PAPR interaction.

Fig. 4.

Mapping molecular determinants of binding enthalpy. Crystal structure of Trx-EcPAPR covalent complex (PDB ID 208V) showing interactions between Trx (green) and PAPR (blue) at two adjacent sites (17). (A) First 800-Å binding hotspot burying PAPR residues 188–199 (α-helix 7) and 202–212 (ω-loop). Based on distance measurements, the noncovalent interactions predicted between PAPR (nonitalicized) and Trx1 residues (italicized) include: H-bonding between W205 (PAPR) and E30 (Trx); salt bridge interactions between D206 (PAPR) and K36–E30 (Trx); π-stacking interactions between Y191 (PAPR) and W31 (Trx); hydrophobic contacts of L210 (PAPR) with M37 and G33 (Trx1). (B) Second 783-Å binding hotspot burying PAPR residues 235–244 (C-terminal tail) including the catalytic cysteine, C239. An interdigitated network of salt bridges are predicted between R237, E238, and E243 (PAPR) and R73 (Trx). (C) Plot of difference in the enthalpy change (ΔH_mutant − ΔH_WT) obtained from two independent ITC measurements for each PAPR variant.

Table 1.

ITC determination of C32A Trx binding to WT and mutant PAPR in oxidized form (PAPR-S-SO₃⁻) at 25 °C

PAPR	ΔH, kJ/mol	TΔS, kJ/mol	ΔG, kJ/mol	Kd, μM	Fold > Kd
WT	−9.3 ± 0.4	24.0 ± 1.0	−33.2 ± 0.6	1.5 ± 0.3	1.0
Y191F	−5.8 ± 1.1	22.0 ± 0.6	−27.6 ± 0.5	13.9 ± 2.6	9.6
Y201F	−7.6 ± 1.0	21.2 ± 0.4	−28.9 ± 0.6	8.4 ± 2.1	5.8
W205F	−1.6 ± 0.2	25.8 ± 0.2	−27.4 ± 0.4	15.2 ± 2.5	10.4
D206E	−2.6 ± 0.8	24.9 ± 2.2	−27.5 ± 0.5	13.7 ± 0.4	10.2
G208A	−9.3 ± 1.3	23.0 ± 0.2	−32.3 ± 1.1	2.1 ± 0.9	1.4
L210A	−8.7 ± 0.5	18.8 ± 0.1	−27.5 ± 0.4	14.5 ± 2.1	10.0
E238A	−6.2 ± 0.4	20.9 ± 0.6	−27.1 ± 0.2	16.8 ± 1.0	11.5

Experiments performed in 50 mM bis-Tris propane buffer with 0.02% Brij-35 (pH 7.4). Data reported as average ± SD from n = 2 measurements.

Conformational Restriction of Oxidized PAPR is Essential for Trx Recognition.

Because PAPR reduction is accompanied by conformational change, regions that undergo structural rearrangements, like the C-terminal tail, may contribute toward the favorable entropy that drives Trx•PAPR interaction. To test this hypothesis, we generated a PAPR truncation variant lacking 10 of the C-terminal residues, including C239. This variant showed no measurable binding to Trx in the presence of saturating PAPS. Given the absence of C239 in truncated PAPR, no PAPR-S-SO₃⁻ intermediate can form. Computational modeling based on the X-ray structure of the PAPR-Trx mixed disulfide complex indicates formation of a putative sulfite-binding pocket at the protein–protein interface (17). Therefore, we wondered whether the absence of the sulfite group might affect the strength of Trx•PAPR interaction. To test this possibility, we generated C239A PAPR, which binds to PAPS (SI Appendix, Fig. S7), but lacks the ability to form the PAPR-S-SO₃⁻ intermediate. Trx bound to C239A PAPR in presence of saturating PAPS, albeit with approximately fourfold reduced affinity compared with WT PAPR-S-SO₃⁻, suggesting that the sulfite group strengthens the interaction between PAPR and Trx (Table 2). Importantly, the reduction in affinity results from the loss in favorable entropy. It is likely that the sulfite group, by virtue of its ability to form intramolecular hydrogen bonds with other PAPR residues, helps tether the C-terminal tail and facilitate formation of the closed conformer. The conformational restrictions of the C-terminal tail in oxidized PAPR decreases the entropy, compared with the reduced open form. Consequently, the selectivity of Trx toward the oxidized closed PAPR conformer correlates with a low-entropy state. Previous results and independent verification by us shows that oxidized PAPR has a very low affinity for PAPS (SI Appendix, Fig. S7) (22). As expected, the presence of PAPS did not affect Trx binding to oxidized PAPR. However, the product PAP was capable of rebinding to oxidized PAPR with a measurable K_d of 21 µM (SI Appendix, Fig. S8). In presence of saturating PAP, C32A Trx bound to oxidized PAPR with a fivefold reduced affinity (Table 2 and SI Appendix, Fig. S9), indicating that Trx preferably binds to oxidized PAPR alone rather than as a complex with PAP. It is possible that PAP release may result from charge–charge repulsion with the nascent sulfite group and also contribute toward a lowering of entropy in the oxidized PAPR conformer.

Table 2.

ITC determination of C32A Trx binding to different PAPR forms in absence or presence of saturating PAPS/PAP at 25 °C

Entry	PAPR form	Kd, μM	Fold > Kd
1	PAPR-S-SO3−	1.5 ± 0.3	—
2	PAPR-SH	∼110	73.3
3	Truncated PAPR	No binding	—
4	C239A PAPR + PAPS	6.3 ± 0.5	4.2
5	PAPR-S-SO3− + PAP	7.6 ± 0.7	5.0

Experiments performed in 50 mM bis-Tris propane buffer with 0.02% Brij-35 (pH 7.4). Data reported as average ± SD from n = 2 measurements.

Changes in Solvation Contribute Favorably to the Trx•PAPR Interaction.

Although our data identify the C-terminal segment of PAPR as the primary molecular determinant for favorable binding entropy, our next goal was to better define the nature of this contribution. Total binding entropy (ΔS_total) obtained from ITC measurement of protein–protein interactions is a complex term, but can be simplified as a sum of changes in the solvation entropy (ΔS_sol), protein translational and rotational entropy (ΔS_RT) and protein conformational entropy (ΔS_conf) (28). Although accurate quantitative dissection of ΔS_tot into contributions from ΔS_sol, ΔS_RT, and ΔS_conf is difficult and often controversial, some initial insight can be obtained by combining inputs from experimental measurements and theoretical models. Negative heat capacity changes (ΔC_p) are usually indicative of involvement of hydrophobic interactions or water of solvation. We measured a ΔC_p value of −920 ± 29 J•K⁻¹•mol⁻¹ for the Trx•PAPR interaction, indicating involvement of hydrophobic interactions or water of solvation (SI Appendix, Fig. S10). The experimentally measured ΔC_p value was further used to calculate a ΔS_sol of 235 J•K⁻¹•mol⁻¹ at 25 °C (SI Appendix, Table S2) (29). As suggested by Murphy et al., “cratic” entropy was considered as the best estimate of ΔS_RT with a calculated value of −33.3 J•K⁻¹•mol⁻¹ (30). Because the experimental value of ΔS_tot measured by ITC was 80.6 ± 4 J•K⁻¹•mol⁻¹, ΔS_conf was calculated to be −121.1 J•K⁻¹•mol⁻¹. In short, these data reveal ΔS_sol as a significant contributor to the favorable entropy underlying the Trx•PAPR interaction.

Model for Trx•PAPR Recognition.

Based on above findings, we propose a model for PAPR recognition by Trx (SI Appendix, Fig. S11). In the open and reduced form, PAPR binds to PAPS in an enthalpy-driven interaction, accompanied by conformational change in PAPR restricting the flexible C-terminal tail on top of the catalytic domain. Next, nucleophilic attack by C239 leads to formation of the PAPR-S-SO₃⁻•PAP complex. Covalent transfer of sulfite group allow additional noncovalent interactions to form between the modified C-terminal tail and active site residues, thereby increasing the propensity of PAPR-S-SO₃⁻ to exist in a closed conformation. Thus, in this first half of PAPR catalytic cycle, PAPR goes from a high-entropy native or open conformation to a low-entropy intermediate or closed conformation. As indicated earlier, PAP exits the binding pocket before Trx binding, further lowering the entropy of the PAPR-S-SO₃⁻ intermediate. Driven by the favorable increase in solvation entropy, Trx preferentially interacts with this low-entropy form of oxidized PAPR. In subsequent steps, PAPR-S-SO₃⁻ intermediate is reduced by Trx and PAPR returns to its high-entropy native state.

Entropy Gain is the Main Driver of Trx–Target Interactions.

Next, we asked whether the entropy-driven recognition mechanism is unique to Trx•PAPR interaction or whether it might be extended to other Trx•SR interactions. We explored this possibility with adenosine-5′-phosphosulfate reductases (APSRs), which are members of the SR family and are known to undergo substrate-induced conformational changes akin to PAPR (8, 31). We first evaluated C32A Trx binding to Mycobacterium tuberculosis APSR (MtAPSR). Similar to EcPAPR, C32A Trx selectively interacted with the oxidized form (i.e., MtAPSR-S-SO₃^–) with a K_d of 7.9 ± 0.6 µM (Table 3). No measurable binding was observed with reduced MtAPSR, suggesting that the conformational changes concomitant with substrate binding and APSR oxidation are a prerequisite for binding to MtAPSR, similar to EcPAPR. Importantly, the binding interaction between Trx and MtAPSR-S-SO₃^– was endothermic and entirely driven by favorable entropy (SI Appendix, Fig. S12). Similarly, Trx selectively bound to the oxidized form of Pseudomonas aeruginosa APSR (PaAPSR), an enzyme with ∼27% homology to MtAPSR (31), in an entropy-driven recognition process (Table 3). Collectively, these data indicate that favorable entropy drives Trx binding to SRs.

Table 3.

ITC determination of C32A Trx binding to various protein targets in their oxidized form at 25 °C

Trx Target	ΔH, kJ/mol	TΔS, kJ/mol	ΔG, kJ/mol	Kd, μM
EcPAPR-S-SO3−	−9.3 ± 0.4	24.0 ± 1.0	−33.2 ± 0.6	1.2 ± 0.3
MtPAPR-S-SO3−	2.0 ± 0.7	31.0 ± 0.9	−29.0 ± 0.2	7.9 ± 0.6
PaPAPR-S-SO3−	−6.3 ± 0.6	22.3 ± 0.1	−28.6 ± 0.5	9.4 ± 1.5
ScGPx3-S2	−6.2 ± 0.9	20.3 ± 1.3	−26.4 ± 0.3	22.3 ± 3.0
hPrx1-S-S-hPrx1*	−4.2 ± 1.1	19.8 ± 0.8	−24.0 ± 0.3	58.8 ± 8.0
ScGPx3-S-S-G	-4.9	5.6	9.9	∼135

Experiments performed in 50 mM bis-Tris propane buffer with 0.02% Brij-35 (pH 7.4) unless stated. Data reported as average ± SD from n = 2 measurements except for C82S ScGpx3-S-S-G.

^*Experiment performed in 50 mM phosphate buffer with 0.02% Brij-25 (pH 7.4).

To test the generality of the entropy-driven recognition mechanism, we extended our studies to other Trx targets. We selected Saccharomyces cerevisiae glutathione peroxidase (ScGpx3) and human Peroxiredoxin 1 (hPrx1) as models for Trx-mediated reduction of intra- and intermolecular disulfide bond, respectively. Homology modeling of ScGpx3 using the template structure of bovine Gpx3 indicates that the catalytic cysteines of ScGpx3 are separated by 13 Å and that significant conformational restriction occurs upon intramolecular disulfide bond formation (32). Identical postoxidation conformational changes have also been reported for other known glutathione peroxidases (33). As expected, C32A Trx readily bound to ScGpx3 upon oxidation (ScGpx3-S₂) with H₂O₂ in an entropy-driven recognition process, whereas only very weak binding was detectable with the double C38S/C82S variant used as a surrogate for reduced Gpx3 (Table 3 and SI Appendix, Fig. S13). hPrx1 belongs to the class of typical 2-cys Prxs, which are well known to adopt a homodecameric state in its reduced form and a homodimeric state in the intermolecular disulfide-linked or oxidized form (7). Interestingly, 2-cys Prxs also undergo conformational rearrangement after oxidation, with the conserved loop-helix motif bearing peroxidatic cysteine undergoing local unfolding to facilitate intermolecular disulfide bond formation (7). Consistent with our findings with other Trx targets, hPrx1 in a fully reduced state showed no measurable binding to C32A Trx. Next, using H₂O₂, we generated the oxidized hPrx1 homodimer linked by two intermolecule disulfide bonds (SI Appendix, Fig. S14). Trx bound to oxidized hPrx1 and this interaction was also driven by a favorable entropy (Table 3). The observed stoichiometry of this protein–protein complex was 1:1, indicating that each oxidized hPrx1 dimer interacted with two molecules of Trx. Recently, some glutathionylated and mycothiolated proteins have also been found to be targets of Trx oxidoreductase activity (34, 35). Therefore, we tested binding of C32A Trx to C82S variant of glutathionylated ScGpx3 (ScGpx3-S-S-G). Trx binds to ScGpx3-S-S-G in an entropy-driven process, albeit with approximately sixfold reduced affinity compared with ScGpx3-S₂ (Table 3 and SI Appendix, Fig. S15). The possible explanation for reduced affinity could be the fact that ScGpx3-S-S-G is conformationally less restricted compared with ScGpx3-S₂. Overall, our data indicate that Trx recognizes the oxidized form of target proteins with exquisite selectively via a universal entropy-driven mechanism.

Discussion

Trx recognizes many S-S–containing target proteins, thereby functioning as a redox “hub” that links cellular redox state to protein function. There has been a long-standing interest in identifying the primary forces involved in recognition of multiple target proteins by Trx. Several X-ray structures for covalent Trx–protein (or peptide) complexes have been reported (16–18). Although the static structures afforded by X-ray crystallography can provide details about enthalpy of interaction at atomic resolution, they seldom provide information regarding entropic forces, thus necessitating the use of alternative biophysical methods for complete thermodynamic characterization of protein–protein interactions (28). To our knowledge, this work represents the first comprehensive, solution-based thermodynamic analysis of noncovalent Trx–target interactions. One of the key findings of this work is that Trx displays exquisite selectivity in binding to oxidized target proteins, and this preferential recognition, in large part, is derived from the conformational restriction inherent to oxidized target proteins. The comparatively weak affinity of Trx for reduced targets suggests that such proteins are otherwise free to execute their biological function and that Trx forms an association only after oxidation. Next, we show that the Trx•PAPR interaction does not depend on the redox-state of Trx as both Trx-(SH)₂ and Trx-S₂ bind to oxidized PAPR with identical affinities (SI Appendix, Fig. S5). This finding is consistent with the remarkably identical 3D solution structures of Trx-(SH)₂ and Trx-S₂, with only very subtle differences in the dihedral angles and side-chain orientations of active-site residues (overall RMSD between the Cα atoms: 0.17 Å) (25). Previously, it has been suggested that the interaction of Trx with g5p and ASK1 target proteins is critically dependent on the redox state of Trx, despite being independent of Trx oxido-reductase function (9, 20). However, later studies have proposed alternative models possibly involving the oxidoreductase function of Trx in regulation of these proteins (36, 37). In light of these studies and others, the absolute dependence of these interactions on the Trx redox-state may need to be reevaluated using a direct solution-based measurement of noncovalent interaction between Trx and these target proteins. Therefore, combined with the available biophysical and biochemical evidence, our results indicate that the recognition of target proteins by Trx is likely to be independent of the Trx redox-state. Given the overall low steady-state levels of Trx-S₂ (38), nonproductive association with oxidized targets would not be favored under normal conditions. However, under oxidative stress, Trx-S₂ may shield the oxidized target protein from reduction by Trx or another redox system, such as glutaredoxin. Overall, our data reveal target redox-state as a critical feature for recognition.

In another significant result from this work, we show that Trx associates with the oxidized target proteins using favorable entropy as a universal driving force for recognition (Table 3). Along these lines, entropy is also used as a principle recognition force by other “hub” proteins like protein kinase A (PKA), which interacts with multiple A-kinase anchoring proteins (39). Further dissection of the entropy term for Trx•PAPR interaction indicates solvation entropy (ΔS_sol) as the major contributing factor in the favorable entropy of interaction (SI Appendix, Table S1). On the basis of the myriad of apolar residues surrounding the Trx dithiol active site, hydrophobic interactions have been proposed to play an important role in Trx•target recognition (1, 2). In the experimental first test of this proposal, our data implicate solvation entropy, which dominates the hydrophobic effect, as the major driving force underlying Trx•target protein interactions. In addition, our data shows that the conformational restriction inherent to oxidized target proteins is a prerequisite for noncovalent recognition by Trx. Such conformational restrictions essentially reduce conformational entropy, one of the major barriers in protein–protein interactions (40). They may also expose buried or solvent-shielded hydrophobic residues, associated with a significant increase in the water of solvation. Trx preferentially binds to the conformationally restricted, low-entropy states by displacing the water molecules into bulk solvent. Interestingly, such conformational restriction around the Trx-binding site has been reported for the majority of, if not all, targets (7, 17, 23, 32). For instance, proteins with nonnative disulfide bonds constitute one class of target proteins for Trx (41). The transient disulfides that form between distant cysteines entail conformational change, which is likely to decrease their overall entropy and make them suitable targets for Trx recognition. Based on these observations, a binding model identical to PAPR can be proposed for the interaction of Trx with disulfide-linked target proteins (Fig. 5).

Fig. 5.

A proposed model for recognition of targets with disulfide bonds by Trx. Formation of disulfide bond in target proteins after oxidation induces conformational change, lowering the entropy. Trx preferentially binds to this low-entropy conformer in an interaction driven by favorable entropy, subsequently reducing it to the high-entropy native form.

In summary, our findings significantly advance our limited understanding of forces that drive Trx•target interactions, which are often central to numerous life-threatening conditions. The universal entropy-driven mechanism for target recognition outlined herein also rationalizes the ability of Trx to recognize a diverse set of target proteins. Given the role of Trx–protein interactions in pathogenesis of numerous redox-related disorders, there is a widespread interest in pharmacological inhibition of specific Trx–protein interactions for therapeutic benefit (12). However, the currently available inhibitors of cytoplasmic Trx including PX-12 and AW464 are generally nonselective and do not target specific Trx–protein interfaces (4). Our findings here pave the way for the development of new pharmacological strategies to address redox-related disorders, ostensibly through the design of inhibitors that selectively target the Trx-oxidized protein interface.

Materials and Methods

Preparation of Proteins for ITC Studies.

Expression and purification of WT Trx (17), WT EcPAPR (8), WT MtAPSR (8), WT PaAPSR (8), and WT and mutant ScGPx3 (32) was carried out as reported earlier. Cloning, expression and purification of hPrx1 was performed as detailed in the SI Appendix. Trx and EcPAPR mutants were constructed using QuikChange PCR mutagenesis kit (Agilent) and oligonucleotide primers listed in SI Appendix, Table S2. Proteins were oxidized using the procedures detailed in SI Appendix and their oxidation to the desired oxidized forms was confirmed by intact mass analysis using LTQ XL Linear Ion Trap Mass Spectrometer (Thermo Scientific).

ITC Binding Experiments.

Nano-ITC Low-Volume (TA Instrument) sample cell was rinsed with 5 × 0.3 mL of freshly degassed ITC buffer 1 or 2 (SI Appendix). It was then loaded with 220–250 µL of reduced, oxidized or mutant target protein samples. The 50 µL Nano-ITC syringe was rinsed with ITC buffer 1 or 2 and loaded with 45–50 µL of Trx-(SH)₂, Trx-S₂, or C32A/S Trx samples. The instrument temperature was set to 25 °C or the value specified with a stirring speed of 300 rpm. A total of 14 injections of Trx (first injection = 2 µL; remaining injections = 3 µL each, separation between injections = 300 or 500 s) were made into the sample cell containing target proteins. The dilution control for Trx was performed by substituting target proteins in the sample cell with ITC buffer 1 or 2. The raw ITC data were analyzed using NanoAnalyze software. The integrated data were corrected for Trx dilution and fit to the independent model (NanoAnalyze software) to determine the binding and thermodynamic parameters. ITC studies of PAPS/PAP with the WT or mutant EcPAPR were performed using the same protocol with PAPS/PAP in the ITC syringe and EcPAPR in sample cell.