NMR structure of the C-terminal domain of TonB protein from Pseudomonas aeruginosa

NMR sample preparation

The C-terminal 96 residues of a TonB protein from P. aeruginosa (UniProt: Q51368) was cloned from genomic DNA (ATCC-47085) as an N-terminal SUMO fusion protein, which was previously termed PsTonB-96 (Guerrero, Ciragan & Iwaï, 2015). The construct expresses a His-tag SUMO fusion protein that produces a 99-residue protein after the SUMO tag removal, which we termed PaTonB-96. PaTonB-96 contains the last 96 residues of PaTonB (247–342) and three residues (SHM) at the N-terminal end from the cloning site. The plasmid (pFGRSF15) coding the gene of PaTonB-96 was transformed into E. coli ER2566 strain (New England Biolabs, Ipswich, MA, USA) for production of doubly ¹⁵N, ¹³C-labeled samples. The transformed E. coli cells were grown overnight at 30 °C in 50 mL LB medium supplemented with 25 μg/mL kanamycin. The cells were spun down at 900 × g for 15 min and gently re-suspended in two L pre-warmed M9 medium supplemented with 25 μg/mL kanamycin, containing ¹⁵NH₄Cl and ¹³C₆-D-glucose as sole nitrogen and carbon sources, respectively. The cells were grown at 37 °C until an OD₆₀₀ of 0.6. Then, the temperature was lowered to 30 °C for protein expression. The protein expression was induced with a final concentration of one mM isopropyl β-D-1-thiogalactopyranoside (IPTG). The cells were incubated for additional 5 h before harvesting by centrifugation at 4,900 × g at 4 °C for 15 min. The labeled PaTonB-96 was purified following the previously published protocol (Guerrero, Ciragan & Iwaï, 2015). The purified protein was dialyzed against 20 mM sodium phosphate buffer (pH 6.0). The protein solution of PaTonB-96 was concentrated to one mM for NMR analysis using a centrifugal device. The final sample volume was 250 μL containing 10% D₂O.

NMR measurements

NMR measurements for the structure determination were recorded on Varian INOVA 800 MHz equipped with a cryogenically cooled five mm probe head. For the sequential backbone assignment, a standard set of double and triple resonance NMR spectra were recorded at 25 °C, including [¹H, ¹⁵N]-HSQC, HNCO, HNCA, HNCACB, HN(CO)CA, HN(CA)CO, and CBCA(CO)NH (Sattler, Schleucher & Griesinger, 1999). The aliphatic side-chain assignment was carried out using [¹H, ¹³C]-HSQC, HCCH-COSY, ct-[¹H, ¹³C]-HSQC, HBHA(CO)NH, H(CCCO)NH, (H)CC(CO)NH, ¹⁵N-resolved [¹H, ¹H]-TOCSY, and ¹⁵N-edited [¹H-¹H]-NOESY spectra. The assignments for the aromatic side-chains were based on the spectra of (HB)CB(CGCD)HD, (HB)CB(CGCDCE)HE, aromatic region ct-[¹H, ¹³C]-HSQC and ¹³C-edited [¹H,¹H]-NOESY. Data was acquired using VnmrJ (Varian Inc., Palo Alto, CA, USA) and data were processed using the NMRpipe software (Delaglio et al., 1995).

NMR measurements for the backbone dynamics were recorded on a Bruker Avance 850 MHz equipped with a cryogenically cooled probe head. The longitudinal (T₁) and transverse (T₂) relaxation rates and ¹⁵N{¹H}-heteronuclear nuclear Overhauser effects (NOEs) for backbone ¹⁵N atoms were determined at 25 °C using the well-established NMR experiments (Barbato et al., 1992; Kay, Torchia & Bax, 1989). T₁(¹⁵N) and T₂(¹⁵N) relaxation times were determined using the following delay times: 10, 50, 100, 200, 300, 500, 800, 1,000, 1,200, and 2,000 ms for T₁ and 16, 64, 96, 128, 156, 196, 224, and 256 ms for CMPG pulse train with one ms interval for T₂ relaxation rates, respectively. Relaxation times were obtained by fitting a single exponential decay to peak intensity values: I(t) = I₀ × exp (–t/T₁) or I₀ × exp (−t/T₂), where I(t) is the peak volume at a time t. Heteronuclear ¹⁵N{¹H}-NOEs were obtained with a relaxation delay of 5 s with or without saturation of protons. Heteronuclear ¹⁵N{¹H}-NOEs (η) were determined from the volumes of the HSQC signals using the ratio of η = I/I₀. The relaxation data were processed and analyzed using Bruker Dynamic Center (Version 2.1.8; Bruker Inc., Billerica, MA, USA).

Solution NMR structure determination

The sequence-specific resonance assignment was performed with standard methods using triple resonance NMR experiments and CcpNmr Analysis software (version 2.4.1) (Vranken et al., 2005). The chemical shift values from the sequential resonance assignment were used together with NOE peak lists for the structure calculation with the program CYANA 3.0 (Mumenthaler et al., 1997; Güntert & Buchner, 2015). NOE distance restraints were obtained from 3D ¹⁵N- and ¹³C-edited [¹H, ¹H]-NOESY spectra with 80-ms mixing time. The conformations of prolines were checked based on CB-CG chemical shift before the structure calculation (Schubert et al., 2002). All prolines were predicted to be in trans conformation except P299 that was set to cis-conformation. The three-dimensional NMR conformers were generated using CYANA 3.0, based on the automated NOESY cross peaks assignment (Güntert, 2004; Güntert, Mumenthaler & Wüthrich, 1997). The restrained energy minimization of the final 20 best conformers was performed using AMBER 14 (Pearlman et al., 1995), and the structures were validated with PSVS 1.5 (Bhattacharya, Tejero & Montelione, 2007). The structural statistics are summarized in Table 1.

MD simulation

The MD simulations were performed with Gromacs5 (Abraham et al., 2015) by using Amber ff99SB-ILDN force field and tip4p water model (Jorgensen et al., 1983; Lindorff-Larsen et al., 2010). The first 10 ns from total 400-ns simulation was considered to be the equilibrium period by monitoring the protein root-mean-square-deviation, inertia tensor eigenvalues, and rotation angles. The rest was used for the analysis since the first 10 ns of simulation trajectories was sufficiently enough to remove the significant fluctuations in these parameters (Ollila, Heikkinen & Iwaï, 2018). The NMR structure determined in this work was used as the initial structure and the secondary structures remained unchanged during the entire simulation period. The temperature was coupled to 25 °C with v-rescale thermostat (Bussi, Donadio & Parrinello, 2007), and the pressure was isotropically set to one bar using Parrinello-Rahman barostat (Parrinello & Rahman, 1981). The time-step was two fs, Lennard-Jones interactions were cut-off at one nm, PME (Darden, York & Pedersen, 1993; Essmann et al., 1995) was used for electrostatics, and LINCS (Hess, 2008) was used to constrain all bond lengths. The simulation data is available from the Zenodo repository (Ollila, 2018a).

The analysis and interpretation of ¹⁵N spin relaxation times are described in detail elsewhere (Ollila, Heikkinen & Iwaï, 2018). Briefly, rotational correlation functions of the backbone N–H bonds were calculated from the simulation data and the spin relaxation times were calculated using the Redfield equations (Abragam, 1961; Kay, Torchia & Bax, 1989; Ollila, Heikkinen & Iwaï, 2018). Before the spin relaxation time calculation, the overestimated overall rotational diffusion in the MD simulation due to the water model was corrected by dividing the rotational diffusion coefficients around all inertia axes with a factor of 1.2, assuming that the protein rotates as an anisotropic rigid body. This scaling factor was found to be capable of reproducing the ¹⁵N spin relaxation times in good agreement with the experimental data of PaTonB-96 for tip4p water model (Ollila, Heikkinen & Iwaï, 2018). The computer codes used for the analysis and the related data are available (Ollila, 2018b, 2018c).

NMR solution structure of PaTonB-96

The protein consisting of the C-terminal 96 residues of TonB from P. aeruginosa (PaTonB-96) was previously identified as a minimal domain that was soluble when expressed using an E. coli expression system, while other shorter fragments were not soluble even with several different fusion partners (Guerrero, Ciragan & Iwaï, 2015). This fragment has 45% identity (40/88 residues) to EcTonB-92, suggesting a similar structure to EcTonB. [¹H, ¹⁵N]-HSQC spectrum of PaTonB-96 shows the well-dispersed NMR signals for amide groups (Fig. 2), indicating a globular folded conformation. All of the main-chain chemical shifts were assigned except for S244 and H^N of H245. 95.2% of the expected side-chains were assigned. The assigned chemical shifts were used for the automatic analysis of NOE peaks with the CYANA to calculate the NMR structure. Figure 3 shows the lowest energy solution NMR structure of PaTonB-96 with the secondary structure elements and a superposition of the 20 lowest energy conformers. The structural statistics of the 20 NMR conformers are summarized in Table 1. PaTonB-96 adopts a mixed α/β structure with the topology of βαββαββ. The structure consists of a β-sheet composed of three anti-parallel β-strands (β2, β3, and β5), a β-sheet composed of two short β-strands (β1 and β4) and two short α-helices (αI and αII). The structural coordinates and the chemical shifts have been deposited in the Protein Data Bank (http://www.rcsb.org/, PDB code: 6FIP) and BMRB (http://www.bmrb.wisc.edu/, accession number: 34235).

Comparison of the NMR structures between E. coli and P. aeruginosa

The primary and secondary structures of PaTonB-96 were compared with the structure of EcTonB-137 (PDB: 1XX3) (Peacock et al., 2005) (Fig. 4). The length of PaTonB-96 is similar to that of the globular part of EcTonB-137 (88 residues). Notable structural differences are observed around the C-terminal end. The C-terminal end of EcTonB-137 forms an anti-parallel β-strand (β6) with β5-strand to constitute a β sheet. In contrast, the C-terminal end in PaTonB-96 is unstructured and exhibits an extended flexible conformation as also seen from the spin relaxation data (Fig. 5). The difference can be easily explained by the shorter C-terminal end after the β5-sheet in PaTonB-96, being too short to form an additional β-strand. The shorter C-terminal end in PaTonB-96 is compensated by the longer loop between β4 and β5-strands with additional five residues, compared with the structured 88-residue region of EcTonB-137. Higher mobility of this longer loop is also confirmed by the ¹⁵N relaxation measurement and MD simulation (Fig. 5). Interestingly, the extended and disordered conformation of the C-terminal end in PaTonB-96 more closely resembles the crystal structures of EcTonB-92 (PDB: 1U07) or the structures from TonB/TBDT complexes (PDB: 2GRX and 2GSK) (Fig. 1). The interaction between β5-strand of TonB and the TonB box in TBDTs has been proposed to be essential for the TonB-mediated energy transfer (Pawelek et al., 2006; Shultis et al., 2006). Such interaction would be hindered by the additional β6-stand found in EcTonB-137. Thus, our results suggest that β5-strand is more accessible for the proposed interactions with TonB box in PaTonB-96 where the β6-strand is absent.

Structural motions in PaTonB-96

Fast dynamics (picosecond to nanosecond) of proteins has been commonly investigated by measuring the spin relaxation times (T₁, T₂, and heteronuclear NOEs) of backbone ¹⁵N atoms (Jarymowycz & Stone, 2006; Van Den Bedem & Fraser, 2015). Here, we characterize the structural dynamics of PaTonB-96 by combining T₁, T₂, and heteronuclear NOEs for backbone ¹⁵N atoms and MD simulations (Ollila, Heikkinen & Iwaï, 2018). The measured spin relaxation times for PaTonB-96 are shown together with the MD simulation results (Fig. 5A). The good agreement with the experimental data allows us to use the MD simulation trajectory to interpret the timescales for the overall rotational diffusion (τ_c), effective correlation times for internal mobility (τ_eff), and order parameters (S²). The overall rotational diffusion coefficients around inertia axes (see the caption of Fig. 5) show that PaTonB-96 has high axiality (A = 4.4), but with a rather low rhombicity (R = 0.2) suggesting that the protein has an approximately ellipsoidal shape. The MD simulation analysis estimates τ_c = 7.2 ns for the timescale of average overall rotational diffusion. This is in line with 6.9 ns estimated from the T₁/T₂ values using the model-free approach (Kay, Torchia & Bax, 1989; Ollila, Heikkinen & Iwaï, 2018) as well as with the values in the literature for monomeric proteins with the similar molecular weight (Krishnan & Cosman, 1998). We thus conclude that PaTonB-96 is monomeric in solution as was the previously published NMR structure for EcTonB-137 (Peacock et al., 2005).

The results reveal that there are four regions with enhanced internal motions in PaTonB-96 (Fig. 5). The regions with notable conformational fluctuations are colored in yellow and purple in the overlaid snapshots from the MD simulation (Fig. 5). The first few N-terminal residues exhibit low order parameters and long effective correlations times related to the enhanced conformational fluctuations, suggesting that the N-terminal region is already the beginning of the flexible central region of the TonB protein that connects between TM region and CTD. The MD simulation revealed orientational fluctuations in αI helix (purple in Fig. 5B). This fluctuation also influenced the order parameters and effective correlation times of the neighboring residues in the MD simulation (Fig. 5A). However, the changes were not observed in the ¹⁵N spin relaxation analysis, presumably because the ¹⁵N relaxation is not sensitive to the time scale of the fluctuation. Residues 320–326 in the loop between β4 and β5-strands indicate some enhanced conformational fluctuations, which were also detectable by the ¹⁵N spin relaxation analysis and order parameters but does not affect the effective correlation times (Fig. 5A). The last five residues of the C-terminal end (residues 338–342) showed the enhanced conformational fluctuations, characterized with lower order parameters, long effective correlation times, and changes in the ¹⁵N spin relaxation rates, which are consistent with the experimental ¹⁵N spin relaxation data. The flexible residues at the C-terminal end are in close contact with the αI helix, indicating orientational fluctuations as described above.

Modelling of PaTonB and TonB box interactions

It is believed that TonB conveys the CM chemical potential to TBDTs by structural changes via the interaction between CTD of TonB and TonB box, a short stretch of peptide chain in the N-terminus of plug domain of TBDTs (Cadieux & Kadner, 1999; Pawelek et al., 2006; Peacock et al., 2005; Shultis et al., 2006) (Fig. 1A). In the crystal structure of EcTonB-93/BtuB complex, EcTonB orients the αI helix against the plug domain and forms a parallel β-sheet interaction with the TonB box of BtuB (Shultis et al., 2006) (Fig. 6A). We created a hypothetical model of a complex between PaTonB-96 and TBDT by superimposing PaTonB-96 to residues 153–233 of EcTonB in the crystal structure of the complex because no structure of TBDTs from P. aeruginosa is available (Fig. 6B). β5-strand (residues 226–231) of EcTonB-93 and TonB box of BtuB forms the parallel β-strands, of which interactions are mostly hydrophobic (Fig. 6C). The modeled structure of PaTonB-96/BtuB complex indeed resembles the conserved TonB box interactions observed with the EcTonB/BtuB complex, supporting that the modeled structure is plausible. The highly conserved Val10 in the TonB box is located next to phenylalanine in both TonB structures (Phe230 and Phe336 in EcTonB-93 and PaTonB-96, respectively), suggesting that this hydrophobic interaction might be critical for the TonB/TBDT complexes (Fig. 6C).