Solution NMR structures of oxidized and reduced Ehrlichia chaffeensis thioredoxin: NMR-invisible structure owing to backbone dynamics

Thioredoxins are small ubiquitous proteins that participate in a diverse variety of redox reactions via the reversible oxidation of two cysteine thiol groups in a structurally conserved active site. Here, the NMR solution structures of a reduced and oxidized thioredoxin from Ehrlichia chaffeensis (Ec-Trx, ECH_0218), the etiological agent responsible for human monocytic ehrlichiosis, are described. The overall topology of the calculated structures is similar in both redox states and is similar to those of other thioredoxins: a five-stranded, mixed β-sheet (β1–β3–β2–β4–β5) surrounded by four α-helices. Unlike other thioredoxins studied by NMR in both redox states, the ¹H–¹⁵N HSQC spectrum of reduced Ec-Trx was missing eight additional amide cross peaks relative to the spectrum of oxidized Ec-Trx. These missing amides correspond to residues Cys35–Glu39 in the active-site-containing helix (α2) and Ser72–Ile75 in a loop near the active site, and suggest a change in backbone dynamics on the millisecond-to-microsecond timescale associated with the breakage of an intramolecular Cys32–Cys35 disulfide bond in a thioredoxin. A consequence of the missing amide resonances is the absence of observable or unambiguous NOEs to provide the distance restraints necessary to define the N-terminal end of the α-helix containing the CPGC active site in the reduced state. This region adopts a well defined α-helical structure in other reported reduced thioredoxin structures, is mostly helical in oxidized Ec-Trx and CD studies of Ec-Trx in both redox states suggests there is no significant difference in the secondary structure of the protein. The NMR solution structure of reduced Ec-Trx illustrates that the absence of canonical structure in a region of a protein may be owing to unfavorable dynamics prohibiting NOE observations or unambiguous NOE assignments.

1. Introduction

Thioredoxins are small proteins of approximately 100 residues that are found in organisms ranging from archaea to mammals and participate in a diverse variety of redox reactions via the reversible oxidation of two cysteine thiol groups to a disulfide (Holmgren, 1985, 1995). In the reducing environment inside the cell the primary function of Trxs is to prevent the formation of unwanted disulfide bonds (Collet & Messens, 2010), especially in metabolically essential enzymes such as ribo­nucleotide reductase (Laurent et al., 1964). Thioredoxins protect proteins from oxidative aggregation and inactivation (Holmgren, 1985, 1995), are a critical component of cellular antioxidant defense strategies against reactive oxygen and nitrogen species (Lu & Holmgren, 2014), play a role in regulation of apoptosis (Ravi et al., 2005), act as growth factors and modulate inflammatory responses (Powis et al., 2000). Once oxidized, Trxs are reduced by thioredoxin reductase via an NADPH-dependent reaction to reactivate the Trxs (Mustacich & Powis, 2000).

Human monocytic ehrlichiosis is an important emerging infectious disease in the United States (Paddock & Childs, 2003). The pathogen responsible, Ehrlichia chaffeensis, is an obligate intracelluar bacterium in the Anaplasmataceae family (Rikihisa, 2010) that is transmitted between mammals by the blood-sucking arthropod Amblyomma americanum (also known as the lone star tick; Ewing et al., 1995). Clinical symptoms of human monocytic ehrlichiosis include headache, fever, myalgia, anorexia and chills. While the severity of the disease may be limited to asymptomatic seroconversion, because E. chaffeensis replicates inside mammalian monocyte and macrophage cells, infection may lead to leukopenia and life-threatening complications such as kidney or respiratory failure. Consequently, human monocytic ehrlichiosis may represent the most prevalent life-threatening tick-borne infection in the United States (Olano & Walker, 2002). No vaccines against human monocytic ehrlichiosis exist and the choice of antibiotics available to treat infection is currently limited to doxycycline and rifampin, with the effectiveness of these treatments reduced if initiation of treatment is delayed. Since reactive oxygen species are abundantly produced by host monocytes and macrophages, it is expected that a robust Trx system is vital for intracellular survival of E. chaffeensis (Lu & Holmgren, 2014), and thus Trx inhibition may represent an intervention strategy for human monocytic ehrlichiosis. The 107-residue protein ECH_0218 (Ec-Trx) has been identified as a Trx in E. chaffeensis. To assist structure-based drug design (Staker et al., 2015) targeting Ec-Trx and to better understand cellular redox mechanisms, we describe the NMR solution structure and backbone dynamic features of Ec-Trx in the oxidized and reduced states.

2. Materials and methods

2.1. Cloning, expression and purification

The Ec-Trx gene (ECH_0218, YP_507041.1) was PCR-amplified using the genomic DNA of E. chaffeensis (strain Arkansas) and the oligonucleotide primers 5′-GGGTCCTGGTTCGATGATTGAGCAAATTGGAGACAGTG-3′ (forward) and 5′-CTTGTTCGTGCTGTTTATTAGTTTATATTGTTATTTATTTCACTAATTAT-3′ (reverse) (Invitrogen, Carlsbad, California, USA) containing LIC primers (bold). To enable protein purification by metal-chelation chromatography (Choi et al., 2011), the amplified Ec-Trx gene was inserted into the NruI/PmeI-digested expression vector AVA0421 by ligation-independent cloning such that the N-terminus of the expressed gene product contained a 21-residue extension: MAHHHHHHMGTLEAQTGPGS–. A heat-shock method was then used to transform the recombinant plasmid into Escherichia coli BL21(DE3)-R3-pRARE2 cells (a gift from SGC Toronto, Ontario, Canada). 750 ml of minimal medium (Miller) containing ¹⁵NH₄Cl (1 mg ml⁻¹), D-[¹³C₆]-glucose (2.0 mg ml⁻¹), NaCl (50 µg ml⁻¹), MgSO₄ (120 µg ml⁻¹), CaCl₂ (11 µg ml⁻¹), Fe₂Cl₃ (10 ng ml⁻¹) chloramphenicol (35 µg ml⁻¹) and ampicillin (100 µg ml⁻¹) was used to obtain uniformly ¹⁵N,¹³C-labeled Ec-Trx. The transformed cells were grown at 37°C to an OD₆₀₀ of ∼0.8, transferred to a 25°C incubator and protein expression was induced with isopropyl β-D-1-thiogalactopyranoside (0.026 µg ml⁻¹). Approximately 4 h later the cell culture was harvested by mild centrifugation and frozen (−80°C). Ec-Trx was purified from the thawed pellet using a conventional two-step protocol: metal-chelate affinity chromatography on a 20 ml Ni-Agarose 6 FastFlow column (GE Healthcare, Piscataway, New Jersey, USA) followed by gel-filtration chromatography on a Superdex 75 HiLoad 26/60 column (GE Healthcare). The only difference between the protocols used to prepare oxidized and reduced Ec-Trx was the addition of 1 mM dithiothreitol (DTT) to the gel-filtration buffer (100 mM NaCl, 20 mM Tris pH 7.1) to obtain the reduced state.

2.2. Circular dichroism spectroscopy

An Aviv Model 410 spectropolarimeter (Lakewood, New Jersey, USA) was used to collect circular dichroism data from a 400 µl sample of ∼0.07 mM Ec-Trx in a quartz cell of 0.1 cm path length. A steady-state wavelength spectrum was first collected for oxidized Ec-Trx at 20°C in buffer containing no DTT (100 mM NaCl, 20 mM Tris pH 7.0). The same sample was then made 1.0 mM in DTT and the spectrum of reduced Ec-Trx was recorded 1 h later. Thermal denaturation curves were obtained for Ec-Trx in both states by recording the ellipticity at 220 nm in 2.0°C intervals from 10 to 90°C. The reported steady-state wavelength spectrum is the average of two consecutive scans, collected with a bandwidth of 1.0 nm and a time constant of 1.0 s, processed by subtracting a blank spectrum from the protein spectrum and automatic line smoothing using the AVIV software.

2.3. Nuclear magnetic resonance spectroscopy

All NMR data were collected from double-labeled (¹³C,¹⁵N) samples of Ec-Trx (∼1.0 mM) using Agilent spectrometers operating at ¹H resonance field strengths of approximately 600, 750 or 800 MHz. All spectrometers were equipped with an HCN probe, pulse-field gradients and Varian Biopack software. Because the ¹H–¹⁵N HSQC spectrum of reduced Ec-Trx contained the best amide cross peak line shape and chemical shift dispersion at 20°C, NMR data for both Ec-Trx redox states were collected at this temperature. For reduced Ec-Trx, the ¹H, ¹³C and ¹⁵N chemical shifts of the backbone and side-chain resonances were assigned from analysis of two-dimensional ¹H–¹⁵N HSQC, ¹H–¹³C HSQC, HBCBCGCDHD and HBCBCGCDCHE spectra and three-dimensional HNCACB, CBCA(CO)NH, HCC-TOCSY-NNH, CC-TOCSY-NNH and HNCO spectra. Because the ¹H–¹⁵N HSQC spectrum of oxidized Ec-Trx was similar to the ¹H–¹⁵N HSQC spectrum of reduced Ec-Trx, fewer NMR experiments were required to assign the ¹H, ¹³C and ¹⁵N chemical shifts for oxidized Ec-Trx: two-dimensional ¹H-¹⁵N HSQC and ¹H–¹³C HSQC spectra and three-dimensional CC-TOCSY-NNH and HNCO spectra. For samples in both redox states, three-dimensional ¹⁵N-edited NOESY-HSQC and ¹³C-edited NOESY-HSQC (aliphatic and aromatic) spectra, collected with a mixing time of 85 ms, were analyzed to obtain the side-chain ¹H assignments and the NOE-based distance restraints required for the structure calculations. Slowly exchanging amides were identified by lyophilizing the double-labeled (¹³C,¹⁵N) samples, re-dissolving them in 99.8% D₂O and immediately collecting a ¹H–¹⁵N HSQC spectrum (∼10 min later). Steady-state {¹H}–¹⁵N heteronuclear NOE values (NOE = I_sat/I_unsat) were measured (20°C, 600 MHz ¹H resonance field strength) in triplicate by taking the ratio of ¹H–¹⁵N HSQC cross-peak heights in spectra recorded in the presence (I_sat) and absence (I_unsat) of 3.0 s of proton presaturation prior to the ¹⁵N excitation pulse (Farrow et al., 1994). Felix2007 (MSI, San Diego, California, USA) and Sparky (v.3.115) were used to process and analyze, respectively, all of the NMR data. The ¹H, ¹³C and ¹⁵N chemical shifts were referenced using indirect methods (DSS = 0 p.p.m.) and deposited in the Biological Magnetic Resonance Bank database (https://www.bmrb.wisc.edu) under accession numbers BMRB-19452 and BRMB-19938 for reduced and oxidized Ec-Trx, respectively.

2.4. Structure calculations

Using the ¹H, ¹³C and ¹⁵N chemical shift assignments and peak-picked NOESY data as initial experimental inputs, solution structures for Ec-Trx in the oxidized and reduced states were determined with the programs CYANA (v.2.1) and CNS (v.1.1) following previously described protocols (Buchko et al., 2015; Buchko, Hewitt et al., 2013). On the basis of preliminary structure calculations and chemical shift trends, six restraints between the side-chain S atoms of Cys32 and Cys35 (2.0–2.1, 3.0–3.1 and 3.0–3.1 Å for the S^γ—S^γ, S^γ—C^β and C^β—S^γ distances, respectively) were introduced into the structure calculations to fix this disulfide bond in oxidized Ec-Trx (Sagaram et al., 2013). It was not possible to assign the side-chain atoms (¹³C and ¹H) of the third conserved proline, Pro76, in Ec-Trx and unambiguously identify the conformational state of Pro76 based on the intra-residue NOE pattern between proline side-chain protons. In all reported Trx structures the third conserved proline residue is exclusively observed in the cis conformation (Collet & Messens, 2010), and therefore, this assumption was made in our structure calculations and Pro76 was constrained to the cis conformation in the structure calculations. The final ensemble of 20 CYANA-derived structures was refined with explicit waters with CNS (v.1.1) using the PARAM19 force field and force constants of 500, 500 and 1000 kcal for the NOE, hydrogen-bond and dihedral restraints, respectively. For these water refinement calculations, the upper boundary of the CYANA distance restraints was increased by 10% for reduced Ec-Trx and 5% for oxidized Ec-Trx, with the lower boundary set to the van der Waals limit for both. Structure quality was assessed with the online Protein Structure Validation Suite (PSVS v.1.5; Bhattacharya et al., 2007) and these values are included in the summary of structure statistics in Table 1. Atomic coordinates for the final ensemble of 20 structures for reduced and oxidized Ec-Trx were deposited in the Research Collaboratory for Structural Bioinformatics (RCSB) Protein Data Bank (PDB) under PDB codes 6amr and 6ali, respectively. Note that while the protein sequence in the BMRB and PDB is numbered continuously, Met1–Asn107, here the first native residue (Met22 in the BMRB and PDB sequence) is numbered Met1.

Table 1
Summary of the structural statistics for Ec-Trx in the reduced and oxidized states

All statistics are for the 20-structure ensemble deposited in the Protein Data Bank.

Ec-Trx	Reduced	Oxidized
PDB code	6amr	6ali
BMRB code	BRMB-19452	BRMB-19938
Restraints for structure calculations
Total NOEs	1454	1724
Intraresidue NOEs	378	487
Sequential (i, i + 1) NOEs	407	486
Medium-range (i, i + j; 1 < j ≤ 4) NOEs	294	332
Long-range (i, i + j; j > 4) NOEs	375	419
φ angles	75	77
ψ angles	75	77
Hydrogen bonds	32	44
Disulfide bonds	—	1
Structure calculations
No. of structures calculated	100	100
No of structures used in ensemble	20	20
Structures with restraint violations
Distance restraint violations > 0.05 Å	0	0
Dihedral restraint violations > 6°	0	1
R.m.s.d. to mean (Å) (residues Met1–Asn107)
Backbone N—C^α—C=O atoms	1.54 ± 0.24	0.82 ± 0.18
All heavy atoms	2.09 ± 0.24	1.31 ± 0.21
R.m.s.d. to mean (Å) (PSVS ordered residues)
Backbone N—C^α—C=O atoms	0.75 ± 0.15	0.71 ± 0.16
All heavy atoms	1.20 ± 0.17	1.07 ± 0.14
Ramachandran plot summary†
Most favored regions (%)	95.8	96.7
Additionally allowed regions (%)	3.7	3.1
Generously favored regions (%)	0.4	0.2
Disallowed (%)	0	0
Global quality scores (Met1–Asn107)
PROCHECK Z-score (raw) (all)	−2.54 (−0.43)	−1.06 (−0.18)
PROCHECK Z-score (raw) (φ, ψ)	−0.97 (−0.30)	−0.24 (−0.14)
MolProbity clash score	−3.99 (32.14)	−3.44 (28.91)

†Calculated using MolProbity (Chen et al., 2010) with residues Met1–Ile107.

3. Results and discussion

3.1. Solution structure of reduced and oxidized Ec-Trx

Fig. 1 shows a superposition of the ordered regions of the ensemble structure closest to the average for reduced (PDB entry 6amr, purple) and oxidized (PDB entry 6ali, wheat) Ec-Trx generated using the program SuperPose (Maiti et al., 2004). As observed in both NMR and X-ray crystallographic structures of other oxidized and reduced Trxs (Holmgren, 1985, 1995; Jeng et al., 1994; Amorim et al., 2007; Olson et al., 2013; Peterson et al., 2005; Qin et al., 1994; Stefanková et al., 2005; Collet & Messens, 2010), Ec-Trx is composed of a five-stranded, twisted, mixed β-sheet (↑↑↑↓↑; β1–β3–β2–β4–β5) surrounded by four α-helices. Helices α1 and α3 are located on one face of the β-sheet and helices α2 and α4 on the opposite face, with α3 approximately perpendicular to α2 and α4. The catalytic CGPC motif, Cys32–Cys35, is located at the N-terminal end of α2, exposed to the surface of the protein. The reduced and oxidized structures are generally similar, with a backbone r.m.s.d. of 2.0 Å between ordered regions. The most significant difference between the two structures obtained from the available NMR data is the length of α2, which extends between residues Pro34 and Tyr49 in the oxidized state but only between residues Gln41 and Glu51 in the reduced state (Fig. 1b). This difference appears to contrast with the crystal and NMR structures of other Trxs in both redox states, where no such dramatic change in α2 was observed (Holmgren, 1985, 1995; Jeng et al., 1994; Olson et al., 2013; Peterson et al., 2005; Qin et al., 1994; Stefanková et al., 2005; Collet & Messens, 2010). For example, in the NMR solution structures of human Trx and Mycobacterium tuberculosis TrxC in both redox states, the backbones of the region containing the active-site cysteine residues form part of an α-helix and superimpose well (Qin et al., 1994; Olson et al., 2013), with only the side-chain S–S distance increasing (1–1.5 Å) on going from the oxidized to the reduced state. Was the difference observed between the two NMR-based structures of Ec-Trx real or a consequence of the available NMR data for the structure calculations?

Figure 1 (a) Superposition of cartoon representations of the structures of oxidized (PDB entry 6ali, wheat) and reduced (PDB entry 6amr, purple) Ec-Trx closest to the average structure in the calculation ensembles. The 21-­residue N-terminal tag has been removed for clarity, with the side chains of the active-site cysteine residues, Cys32 and Cys35, colored yellow (oxidized) and magenta (reduced). (b) The primary structure of Ec-Trx with the STRIDE elements of secondary structure identified in the reduced (re) and oxidized (ox) states (α-helices, red; β-strands, blue). Residues lacking an assigned amide cross peak are identified by hollow letters.

3.2. Chemical shift verification of cysteine oxidation states

Foremost in assessing the NMR structural differences in α2 is to ascertain the oxidation state and disulfide-bond pairings of the cysteine residues under both conditions used in the structure calculations. The two cysteine residues in the active-site CGPC motif are conserved in all Trxs and are essential for catalytic activity (Collet & Messens, 2010). In addition to the two cysteine residues, Cys32 and Cys35, in the CGPC motif of Ec-Trx, a third cysteine, Cys17, is also present in the primary amino-acid sequence of Ec-Trx. Because the chemical shift of the cysteine C^β atom is sensitive to the redox state of the cysteine S^γ atom, it can be used to infer the redox states of cysteine residues in a protein (Sharma & Rajarathnam, 2000). Generally, the cysteine ¹³C^β chemical shift in the reduced state is <32 p.p.m. and increases to >35 p.p.m. in the oxidized state. The Cys17 ¹³C^β chemical shift changes little in the presence (27.8 p.p.m.) and absence (27.9 p.p.m.) of the reductant DTT, unambiguously verifying that Cys17 in Ec-Trx is a free thiol under both conditions. While this is expected in the presence of DTT, in the absence of DTT this observation indicates that Cys17 is not forming intermolecular or intramolecular disulfide bonds. On the other hand, out of the two remaining cysteine residues, only the ¹³C^β chemical shift for Cys35 in the absence of DTT could be assigned because it was the only other cysteine amide that was detected in the ¹H–¹⁵N HSQC spectrum under both conditions (Fig. 1b). At 35.7 p.p.m., the Cys35 ¹³C^β chemical shift is consistent with the formation of an oxidized cysteine participating in a disulfide bond. The possibility of Cys35–Cys35 intermolecular disulfide-bond formation is excluded as both oxidized and reduced Ec-Trx eluted from a size-exclusion column with a retention time characteristic of a monomer (data not shown). Because the ¹³C^β chemical shift for Cys17 is characteristic of a reduced cysteine, by elimination Cys35 must be forming a disulfide bond with Cys32. As described in §2.4, these chemical shift trends, along with proximity in preliminary structure calculations, were used to justify the introduction of six restraints between the side-chain S atoms of Cys32 and Cys35 in the structure calculations.

3.3. Data and structure quality

The ¹H–¹⁵N HSQC spectrum for reduced Ec-Trx shown in Fig. 2(a) contains amide cross peaks with good chemical shift dispersion in both the ¹H and ¹⁵N dimensions characteristic of a structured protein (Yee et al., 2002). In the absence of the reductant DTT the ¹H–¹⁵N HSQC spectrum for oxidized Ec-Trx, shown in Fig. 2(b), was generally similar to the spectrum for reduced Ec-Trx except for the appearance of eight amide cross peaks (circled in orange in Fig. 2b) and more uniform amide cross-peak intensities. Cross peaks were also missing for ten backbone amides in Ec-Trx in both redox states. Missing amide resonances in ¹H–¹⁵N HSQC spectra of Trxs appear to be uncommon in either redox state. One missing amide resonance at the second cysteine in the CGPC motif was reported in the ¹H–¹⁵N HSQC spectrum of reduced Saccharomyces cerevisiae Trx2 (Amorim et al., 2007). Instead of disappearing amide resonances in ¹H–¹⁵N HSQC spectra, a subset of amide cross peaks are usually observed to shift markedly in ¹H–¹⁵N HSQC spectra of reduced and oxidized Trxs, as shown for human (Qin et al., 1994), E. coli (Jeng et al., 1994), Arabidopsis thaliana (Peterson et al., 2005) and M. tuberculosis (Olson et al., 2013) Trxs. In these examples, the overall backbone of solution structures for α2 determined in both redox states was similar except for slight differences in the region of the active site corresponding to amides with the observed chemical shift perturbations. Missing amide resonances have been observed at a dimer interface (Buchko et al., 2017); however, size exclusion column chromatography showed that Ec-Trx was monomeric in both redox states (§3.2) as is common for Trxs (Eklund et al., 1991). Missing or non-uniform amide resonances may also be indicative of transient association (Buchko, Lin et al., 2013); however, no significant changes in the ¹H–¹⁵N HSQC spectra were observed following a tenfold dilution of the NMR samples used in our studies (data not shown). Instead, the missing amide cross peaks in the ¹H–¹⁵N HSQC spectrum of reduced Ec-Trx observed here are likely to signify that this region of the protein is undergoing chemical exchange on an intermediate timescale (milliseconds to microseconds; Buchko et al., 1999; Ådén et al., 2011). For example, amide resonances in ¹H–¹⁵N HSQC spectra have been observed to disappear in peroxiredoxins as a function of the redox state of a single pair of intramolecular cysteine residues, with the disappearance attributed to substantial changes in backbone dynamics (Ådén et al., 2011; Buchko et al., 2016). Here, the eight additional amide residues that are missing cross peaks in reduced Ec-Trx, Cys35–Glu39 and Ser73–Ile75, are at the N-terminal end of a helix containing the active site and a loop near this site, respectively. Note that in S. cerevisiae Trx2 the residues equivalent to Ec-Trx residues Ser73–Ile75 were reported to be considerably line-broadened in the reduced state (Amorim et al., 2007), indicating that this region in S. cerevisiae Trx2 was likely to be undergoing chemical exchange on a nearly similar intermediate timescale (milliseconds to microseconds). The amide resonances that are missing in both Ec-Trx redox states are for residues N-terminal to the residues that disappear in reduced Ec-Trx (Phe27–Gly33) and in a loop connecting β3 and α3 that is also near the active site (Ile58–Glu61; Fig. 1b). As illustrated in Fig. 3, all the missing amide resonances in both redox states (blue) or for reduced Ec-Trx (cyan) are clustered at or near the active site of the CGPC motif, implying intermediate timescale (milliseconds to microseconds) dynamics that expand upon breaking the Cys32–Cys35 disulfide bond.

Figure 2 (a) The ^1H–^15N HSQC spectrum of ^13C- and ^15N-labeled reduced Ec-Trx (∼1.0 mM) collected at a proton resonance frequency of 750 MHz at 20°C in 100 mM NaCl, 20 mM Tris, 1 mM DTT pH 7.0, with the assigned amide cross peaks labeled. Gly84 (underlined) is folded into the spectrum and Gly6, at 9.20 (^1H) and 107.4 (^15N) p.p.m., is not shown. (b) The ^1H–^15N HSQC spectrum of ^13C- and ^15N-labeled oxidized Ec-Trx (∼1.0 mM) collected at a proton resonance frequency of 600 MHz at 20°C in 100 mM NaCl, 20 mM Tris pH 7.0, with the assigned amide cross peaks labeled. Not shown is the cross peak for Gly84 at 9.85 (^1H) and 103.9 (^15N) p.p.m. In both spectra the backbone amide cross peaks from the non-native N-terminal tag are identified with red labels, cross peaks below the displayed contour levels are identified with an `x', side-chain amide resonance cross-peak pairs are connected by a red dashed line and the eight amide cross peaks that are absent in Fig. 2(a) are circled in orange in Fig. 2(b).

Figure 3 Both faces of the surface representation of the structure closest to the average for oxidized Ec-Trx, illustrating the locations of residues with missing amide cross peaks in ^1H–^15N HSQC spectra. Residues with assigned amide ^1H–^15N HSQC cross peaks are colored magenta. Residues with missing amide cross peaks in both redox states are colored blue and those missing after reduction of the Cys32–Cys35 disulfide bond are colored cyan. The active-site cysteine residues are colored yellow. The representation on the left is in approximately the same orientation as shown in Figs. 1 and 4, with the representation on the right a 180° rotation about the y axis.

Owing to the eight fewer amide cross peaks in the ¹H–¹⁵N HSQC spectrum of reduced Ec-Trx, only 69% of the backbone and side-chain ¹H, ¹³C and ¹⁵N chemical shifts were assigned for reduced Ec-Trx, compared with the 79% assigned for oxidized Ec-Trx (excluding the N-terminal tag). One consequence of this assignment difference, as summarized in the structure statistics in Table 1, was 270 fewer NOEs and 12 fewer hydrogen bonds in the structure calculations for reduced Ec-Trx. Despite these differences in restraints, the ordered regions, as defined by the PSVS program, of the final ensembles of 20 structures both converged well on the average structure, with backbone (N—C^α—C=O) and all-heavy-atom r.m.s.d.s of less than 0.8 and 1.2 Å, respectively. On the other hand, when using the complete native sequence (residues Met1–Asn107) for the r.m.s.d. calculations, the convergence is poorer for reduced Ec-Trx (1.5 and 2.1 Å) than for oxidized Ec-Trx (0.8 and 1.3 Å), reflecting the fewer restraints in the former calculations. The good convergence of the ordered regions of both structures is visually apparent in Figs. 4(a) and 4(b), which show a superposition of the ordered regions of the final ensemble of 20 structures upon the average structure. In both structure ensembles the region with the least convergence is also the region where the missing amides in the ¹H–¹⁵N HSQC spectra cluster (Fig. 3): around the CGPC active site. Ensemble analysis with the PSVS validation software package further confirmed a quality set of final structures (Bhattacharya et al., 2007). The Ramachandran statistics for the φ/ψ pairs of all of the residues in both ensembles were overwhelmingly in acceptable space (>95.8% in most favored regions and >3.1% in additionally allowed regions) and all of the structure-quality Z-scores were acceptable (>−2). Hence, given the available NMR data as input for the structure calculations, the final ensemble of calculated structures meets the generally accepted criteria for acceptable solution structures.

Figure 4 (a, b) Backbone superposition of the ensemble of 20 structures calculated for reduced (a) and oxidized (b) Ec-Trx on the structure closest to the average. (c, d) Cartoon representation of the reduced (c) and oxidized (d) Ec-Trx structure closest to the average structure from the ensembles in (a) and (b), respectively, to assist reference to the structure ensembles. In (a)–(d) the 21-residue N-terminal tag has been removed for clarity; the α-helices and β-­strands are colored red and blue, respectively, and the side chains of the active-site cysteines are colored yellow.

Amide cross peaks were not observed for residues Cys35–Glu39 in the reduced state of Ec-Trx owing to chemical exchange on the intermediate timescale (milliseconds to microseconds) that broadens these resonances beyond detection (Ådén et al., 2011; Buchko et al., 2016). Consequently, the difference in the length of α2 in the calculated NMR structure for reduced Ec-Trx is likely to be owing to the absence of NOEs and backbone hydrogen bonds to define the structure in this region of the protein and does not represent the complete unfolding of approximately half an α-helix. This conclusion is supported by the steady-state wavelength CD spectra for reduced (purple) and oxidized (red) Ec-Trx shown in Fig. 5(a). The data were obtained by first measuring the CD spectrum of oxidizing Ec-Trx, making the sample in the CD cell 1 mM in DTT using a 1 M stock solution and collecting a spectrum for reduced Ec-Trx 1 h later. The spectra superimpose well and are characterized by a single minimum at ∼218 nm and an extrapolated maximum of >200 nm, which are features of a structured protein with a mixture of α-helical and β-strand components (Holzwarth & Doty, 1965). The superimposed CD profiles suggest that the structures of reduced and oxidized Ec-Trx are similar, and this is corroborated by similar (<1% difference) deconvolution values calculated using the AVIV software. More significantly, the absence of an increase in the ellipticity of the ∼218 nm CD band upon the addition of DTT to oxidized Ec-Trx suggests that half an α-helix does not disappear in reduced Ec-Trx. More likely, the full α-helix is still largely present out to Pro34, but only the backbone dynamics of the helix around the active site have changed to make it invisible in the ¹H–¹⁵N HSQC spectrum. This illustrates that care should be taken when viewing `disordered' regions in ensembles of NMR-based structures because the absence of canonical secondary-structure elements may be a result of the absence of NOEs owing to intermediate (millisecond to microsecond) backbone dynamics that make structure in these regions NMR-invisible.

Figure 5 (a) Circular dichroism steady-state wavelength spectra for oxidized (red) and reduced (purple) Ec-Trx (∼0.07 mM) in NMR buffer at 20°C. (b) The CD thermal melts for oxidized (red) and reduced (purple) Ec-Trx obtained by recording the ellipticity at 218 nm at 2°C intervals between 10 and 90°C.

3.4. Steady-state {¹H}–¹⁵N heteronuclear NOEs

The missing and broadened amide resonances in the ¹H–¹⁵N HSQC spectra of Ec-Trx are likely to be the result of intermediate (millisecond to microsecond) backbone dynamics in the protein backbone localized around the active site, as summarized in Fig. 3. This active site is localized on the surface of the protein to facilitate interaction with substrate (binding and reduction; Gleason & Holmgren, 1988; Collet & Messens, 2010). Localized protein dynamics are known to play a role in the binding and catalysis properties at, and around, active sites (Ishima & Torchia, 2000) and have been proposed to play a role in the activity of Trxs (Stefanková et al., 2005; Amorim et al., 2007). To further probe the backbone dynamics of Ec-Trx, this time on the fast picosecond-to-nanosecond timescale, steady-state {¹H}–¹⁵N heteronuclear NOE ratios were collected in triplicate for Ec-Trx and the mean results are presented in Fig. 6. In both redox states most of the ratios are above 0.8, an approximate value typical of well structured residues. The few exceptions are the termini, where ratios below 0.8 are often observed (Peterson et al., 2005), and a few isolated residues located at, or next to, loops between regions of ordered secondary structure. While these exceptions indicate localized points of thermal motion on the picosecond-to-nanosecond timescale and have been proposed to possibly play a role in switching between Trx redox states (Amorim et al., 2007), the near-uniform distribution of the steady-state {¹H}–¹⁵N heteronuclear NOE ratios above 0.8 suggest that the Ec-Trx structure is otherwise relatively rigid in both redox states (Peterson et al., 2005).

Figure 6 Backbone amide {^1H}–^15N heteronuclear NOEs for oxidized (a) and reduced (b) Ec-Trx collected at 20°C. On top of the plots is a schematic representation of the elements of α-helical (red) and β-strand (blue) secondary structure observed for oxidized Ec-Trx and the residues missing an assigned amino-acid cross peak in ^1H–^15N HSQC spectra (oxidized and reduced state colored white, reduced state only colored gray).

3.5. Ec-Trx and conserved proline residues in the Trx scaffold

Highly conserved residues have been identified in Trxs, as illustrated for Ec-Trx and six other thioredoxins in Fig. 7. However, these conserved residues are all not essential for catalytic activity, but instead play roles in dictating the structure, dynamics and redox properties of the protein (Collet & Messens, 2010). Included in this group of conserved residues that are not essential for catalysis are three proline residues. In Ec-Trx, the first conserved proline is Pro34 located in the CGPC motif that plays a large role in determining the reducing power of Trxs (Collet & Messens, 2010). The second proline, Pro40, is six residues away from the first and serves to introduce a kink into α2 that isolates the CGPC motif from the rest of the helix (Collet & Messens, 2010), as shown here for oxidized Ec-Trx (Figs. 1 and 4). Proline residues in the middle of α-helices are known to introduce a kink into the helix (Woolfson & Williams, 1990). In reduced Ec-Trx Pro40 separates α2 into regions with amide cross peaks that are absent or present in the ¹H–¹⁵N HSQC spectrum. As illustrated in Fig. 7, the four amino acids between the conserved last cysteine in the CGPC motif and the conserved proline appear to be semi-conserved (similar) in Trxs. However, in Ec-Trx only two of these four residues are semi-conserved. It may be that the SM and AE differences at the second and fourth positions, respectively, in Ec-Trx destabilize the N-terminus of α2 enough to introduce backbone dynamics on the millisecond-to-microsecond timescale after breaking the intramolecular Cys32–Cys35 disulfide bond. The third conserved proline residue, Pro76, is located just before β4. As described in §2.4, this proline is always observed in the cis conformation (Collet & Messens, 2010), and therefore, was constrained to the cis conformation in our structure calculations. This proline is in a loop spatially near the CGPC motif, and amide cross peaks for the three residues N-terminal to Pro76, Ser73–Ile75, are not observed in the ¹H–¹⁵N HSQC spectrum of reduced Ec-Trx (Figs. 1 and 3), suggesting that the dynamics of this proximal region are also affected by the reduction of the Cys32–Cys35 disulfide bond.

Figure 7 Clustal Omega multiple primary amino-acid sequence alignment, illustrated with ESPript (Robert & Gouet, 2014), of selected Trx proteins with structures in the PDB: E. chaffeensis (PDB entries 6amr and 6ali), Saccharomyces cerevisiae Trx1 (PDB entry 2i9h; Pinheiro et al., 2008), S. cerevisiae Trx2 (PDB entry 2hsy; Amorim et al., 2007), Homo sapiens (PDB entry 3trx; Forman-Kay et al., 1991), Plasmodium falciparum (PDB entry 2mmn; C. Munte, H. Kalbitzer & R. Schirmer, unpublished work), Bacillus subtilis (PDB entry 2gzy; Li et al., 2007) and Mycobacterium tuberculosis (PDB entry 2l59; Olson et al., 2013). Above the alignment are the elements of secondary structure observed for oxidized Ec-Trx (α-helices colored red, β-strands colored blue). Below the alignment the residues missing an assigned amino-acid cross peak in the ^1H–^15N HSQC spectra are highlighted (oxidized and reduced state colored white, reduced state only colored gray).

3.6. Thermal stability

As described earlier, the major negative CD band for reduced and oxidized Ec-Trx with a minimum at ∼218 nm is a spectral feature of a structured protein with a mixture of α-helical and β-strand components (Fig. 5a). The thermal stability of Ec-Trx in both redox states can be assessed by measuring the ellipticity at 218 nm at 2.0°C intervals between 10 and 90°C. In this spectral region, an increase in ellipticity with increasing temperature is generally associated with the transition from a structured to a denatured state (Karantzeni et al., 2003). As shown in Fig. 5(b), the ellipticity for oxidized Ec-Trx (red) largely increases in a regular fashion from 10 to 88°C. On the other hand, the ellipticity for reduced Ec-Trx (purple) essentially follows the profile for oxidized Ec-Trx (red) until ∼80°C and then increases rapidly, with precipitate observable after cooling (irreversible denaturation). This difference suggests that in the presence of an intramolecular disulfide bond the temperature of thermal denaturation increases, making the oxidized protein more stable. In general, oxidized Trxs are more stable than the reduced form, with this difference in stability being a driving force for the catalytic reaction (Collet & Messens, 2010). For example, a 12°C increase in thermal stability was observed for E. coli Trx between the reduced and oxidized states (Jeng et al., 1994; Qin et al., 1994). Because the small structural changes observed between E. coli Trx in the reduced and oxidized states cannot account for the significant thermal and chemical stability differences between the two redox states, it was proposed that these stability differences may be owing, at least in part, to changes in the dynamic properties of E. coli Trx in the reduced and oxidized states (Stefanková et al., 2005). The dynamic differences implied by the disappearance of eight amide cross peaks in the ¹H–¹⁵N HSQC spectrum of reduced Ec-Trx (Fig. 2) may play a role in the decreased thermal stability of reduced Ec-Trx despite the general similarities of the structures in both redox states (Fig. 1).

4. Conclusions

The overall topology of the solution structures determined for Ec-Trx in the oxidized and reduced states is similar to the topology reported in other thioredoxin structures. Using the available NMR data, the length of the α-helix containing the CGPC active site (α2) was observed to extend between residues Gly33 and Tyr49 in the oxidized state but only between residues Gln40 and Gln48 in the reduced state. However, these structural differences in the catalytic helix are very likely to be owing to the absence of proton–proton NOEs that are essential to define this region during the structure calculations, making the structure NMR-invisible in this region. These NOEs are missing because eight amide cross peaks are absent in the ¹H–¹⁵N HSQC spectrum of reduced Ec-Trx relative to that of oxidized Ec-Trx. While the absence of these cross peaks is detrimental to the structure calculations, they do identify a change in the backbone dynamics on the intermediate millisecond-to-microsecond timescale associated with the breakage of an intramolecular Cys32–Cys35 disulfide bond. Dynamics have been suggested to play a contributing role in the catalytic function of thioredoxins and their roles in the thioredoxin system. If the dynamics observed here for Ec-Trx are unique to the intracellular bacterium E. chaffeenis, this may represent a potential pathway for new intervention strategies for human monocytic ehrlichiosis.