Advertisement

Unveiling the activation dynamics of a fold-switch bacterial glycosyltransferase by 19F NMR

      Fold-switch pathways remodel the secondary structure topology of proteins in response to the cellular environment. It is a major challenge to understand the dynamics of these folding processes. Here, we conducted an in-depth analysis of the α-helix–to–β-strand and β-strand–to–α-helix transitions and domain motions displayed by the essential mannosyltransferase PimA from mycobacteria. Using 19F NMR, we identified four functionally relevant states of PimA that coexist in dynamic equilibria on millisecond-to-second timescales in solution. We discovered that fold-switching is a slow process, on the order of seconds, whereas domain motions occur simultaneously but are substantially faster, on the order of milliseconds. Strikingly, the addition of substrate accelerated the fold-switching dynamics of PimA. We propose a model in which the fold-switching dynamics constitute a mechanism for PimA activation.
      Proteins do not just occupy a single state but are in dynamic exchange between a set of conformations, and some of these states may at times be populated at very low levels, so-called invisible or dark states (
      • Wolynes P.G.
      Recent successes of the energy landscape theory of protein folding and function.
      ,
      • Clore G.M.
      Visualizing lowly-populated regions of the free energy landscape of macromolecular complexes by paramagnetic relaxation enhancement.
      ,
      • Sekhar A.
      • Kay L.E.
      NMR paves the way for atomic level descriptions of sparsely populated, transiently formed biomolecular conformers.
      ,
      • Alderson T.R.
      • Kay L.E.
      Unveiling invisible protein states with NMR spectroscopy.
      ). Motions occur on a broad range of timescales, and growing evidence suggests that dynamics are instrumental for the function of a multitude of proteins. In enzymatically mediated reactions, protein dynamics enable the recruitment and binding of substrates, stabilization of transition states, and adaptation to a dynamic membrane environment or binding partners (
      • Eisenmesser E.Z.
      • Bosco D.A.
      • Akke M.
      • Kern D.
      Enzyme dynamics during catalysis.
      ,
      • Olsson M.H.
      • Parson W.W.
      • Warshel A.
      Dynamical contributions to enzyme catalysis: critical tests of a popular hypothesis.
      ). Fold-switching, i.e. the interconversion of secondary structure elements, is a functionally essential dynamic event for a certain class of proteins. Such motions modify the topology of a protein, including α-helix–to–β-strand and/or β-strand–to–α-helix transitions (
      • Bryan P.N.
      • Orban J.
      Proteins that switch folds.
      ,
      • Goodchild S.C.
      • Curmi P.M.
      • Brown L.J.
      Structural gymnastics of multifunctional metamorphic proteins.
      ,
      • Lella M.
      • Mahalakshmi R.
      Metamorphic proteins: emergence of dual protein folds from one primary sequence.
      ). Alterations of such dynamic processes are essential for and causative of human disease, as in the case of the prion protein PrP and a variety of proteins associated with neurodegenerative diseases, including α-synuclein, amyloid β-peptide, and huntingtin (
      • Shrivastava A.N.
      • Aperia A.
      • Melki R.
      • Triller A.
      Physico-pathologic mechanisms involved in neurodegeneration: misfolded protein-plasma membrane interactions.
      ,
      • Michaels T.C.T.
      • Šarić A.
      • Habchi J.
      • Chia S.
      • Meisl G.
      • Vendruscolo M.
      • Dobson C.M.
      • Knowles T.P.J.
      Chemical kinetics for bridging molecular mechanisms and macroscopic measurements of amyloid fibril formation.
      ). Although it has been proposed that such major structural changes are widespread in proteins, structures of fold-switching proteins are underrepresented in the PDB (
      • Porter L.L.
      • Looger L.L.
      Extant fold-switching proteins are widespread.
      ), and studies of associated dynamics are scarce, remaining a major challenge (
      • Luo X.
      • Tang Z.
      • Xia G.
      • Wassmann K.
      • Matsumoto T.
      • Rizo J.
      • Yu H.
      The Mad2 spindle checkpoint protein has two distinct natively folded states.
      ,
      • Volkman B.F.
      • Liu T.Y.
      • Peterson F.C.
      Lymphotactin structural dynamics.
      ,
      • Chang Y.-G.
      • Cohen S.E.
      • Phong C.
      • Myers W.K.
      • Kim Y.-I.
      • Tseng R.
      • Lin J.
      • Zhang L.
      • Boyd J.S.
      • Lee Y.
      • Kang S.
      • Lee D.
      • Li S.
      • Britt R.D.
      • Rust M.J.
      • et al.
      Circadian rhythms. A protein fold switch joins the circadian oscillator to clock output in cyanobacteria.
      ).
      Here, we report on the dynamics of the phosphatidyl-myo-inositol mannosyltransferase PimA, an essential enzyme in Mycobacterium tuberculosis (
      • Korduláková J.
      • Gilleron M.
      • Mikusova K.
      • Puzo G.
      • Brennan P.J.
      • Gicquel B.
      • Jackson M.
      Definition of the first mannosylation step in phosphatidylinositol mannoside synthesis PimA is essential for growth of mycobacteria.
      ,
      • Boldrin F.
      • Ventura M.
      • Degiacomi G.
      • Ravishankar S.
      • Sala C.
      • Svetlikova Z.
      • Ambady A.
      • Dhar N.
      • Kordulakova J.
      • Zhang M.
      • Serafini A.
      • Vishwas K.G.
      • Kolly G.S.
      • Kumar N.
      • Palù G.
      • et al.
      The phosphatidyl-myo-inositol mannosyltransferase PimA is essential for Mycobacterium tuberculosis growth in vitro in vivo.
      ). PimA catalyzes the first step in the biosynthesis of phosphatidyl-myo-inositol mannosides (PIMs) by adding a mannose residue donated by the nucleotide sugar GDP-mannose (GDP-Man) to a phosphatidyl-myo-inositol lipid, which is anchored into the cytoplasmic phase of the plasma membrane (
      • Korduláková J.
      • Gilleron M.
      • Mikusova K.
      • Puzo G.
      • Brennan P.J.
      • Gicquel B.
      • Jackson M.
      Definition of the first mannosylation step in phosphatidylinositol mannoside synthesis PimA is essential for growth of mycobacteria.
      ,
      • Guerin M.E.
      • Kaur D.
      • Somashekar B.S.
      • Gibbs S.
      • Gest P.
      • Chatterjee D.
      • Brennan P.
      • Jackson M.
      Insight into the molecular mechanism of the early steps of phosphatidylinositol mannosides biosynthesis in mycobacteria.
      ). PIMs are critical structural components of the mycobacterial cell envelope and precursors of two cell envelope lipoglycans, lipomannan and lipoarabinomannan (
      • Guerin M.E.
      • Korduláková J.
      • Alzari P.M.
      • Brennan P.J.
      • Jackson M.
      Molecular basis of phosphatidyl-myo-inositol mannoside biosynthesis and regulation in mycobacteria.
      ,
      • Sancho-Vaello E.
      • Albesa-Jové D.
      • Rodrigo-Unzueta A.
      • Guerin M.E.
      Structural basis of phosphatidyl-myo-inositol mannosides biosynthesis in mycobacteria.
      ), both of which are virulence factors during tuberculosis (
      • Jankute M.
      • Cox J.A.
      • Harrison J.
      • Besra G.S.
      Assembly of the mycobacterial cell wall.
      ), one of the deadliest infections worldwide (
      • Horsburgh Jr, C.R.
      • Barry III, C.E.
      • Lange C.
      Treatment of tuberculosis.
      ). Crystal structures were previously obtained of PimA in the unliganded state (
      • Giganti D.
      • Albesa-Jové D.
      • Urresti S.
      • Rodrigo-Unzueta A.
      • Martínez M.A.
      • Comino N.
      • Barilone N.
      • Bellinzoni M.
      • Chenal A.
      • Guerin M.E.
      • Alzari P.M.
      Secondary structure reshuffling modulates glycosyltransferase function at the membrane.
      ) and in the presence of GDP or GDP-Man (Fig. 1A) (
      • Giganti D.
      • Alegre-Cebollada J.
      • Urresti S.
      • Albesa-Jové D.
      • Rodrigo-Unzueta A.
      • Comino N.
      • Kachala M.
      • López-Fernández S.
      • Svergun D.I.
      • Fernández J.M.
      • Guerin M.E.
      Conformational plasticity of the essential membrane-associated mannosyltransferase PimA from mycobacteria.
      ). In the crystal structure of unliganded PimA, two distinct conformations of a region in the N-terminal domain, termed the reshuffling region (residues 118–163), are seen, suggesting that this region can undergo motions between a compact and an extended conformation, even in the absence of substrate (Fig. 1A) (
      • Giganti D.
      • Albesa-Jové D.
      • Urresti S.
      • Rodrigo-Unzueta A.
      • Martínez M.A.
      • Comino N.
      • Barilone N.
      • Bellinzoni M.
      • Chenal A.
      • Guerin M.E.
      • Alzari P.M.
      Secondary structure reshuffling modulates glycosyltransferase function at the membrane.
      ,
      • Guerin M.E.
      • Schaeffer F.
      • Chaffotte A.
      • Gest P.
      • Giganti D.
      • Korduláková J.
      • van der Woerd M.
      • Jackson M.
      • Alzari P.M.
      Substrate-induced conformational changes in the essential peripheral-membrane binding mannosyltransferase PimA from mycobacteria. Implications for catalysis.
      ). Moreover, the crystal structures of PimA in the presence of GDP or GDP-Man support the hypothesis that the reshuffling region displays fold-switching of secondary structure elements with both α-helix–to–β-strand and β-strand–to–α-helix transitions upon addition of substrate (Fig. 1A) (
      • Giganti D.
      • Albesa-Jové D.
      • Urresti S.
      • Rodrigo-Unzueta A.
      • Martínez M.A.
      • Comino N.
      • Barilone N.
      • Bellinzoni M.
      • Chenal A.
      • Guerin M.E.
      • Alzari P.M.
      Secondary structure reshuffling modulates glycosyltransferase function at the membrane.
      ). Fold-switching activates PimA, since PimA mutants locked in the unliganded conformation are inactive, while mutants locked in the substrate-bound conformation are active (
      • Giganti D.
      • Albesa-Jové D.
      • Urresti S.
      • Rodrigo-Unzueta A.
      • Martínez M.A.
      • Comino N.
      • Barilone N.
      • Bellinzoni M.
      • Chenal A.
      • Guerin M.E.
      • Alzari P.M.
      Secondary structure reshuffling modulates glycosyltransferase function at the membrane.
      ). Finally, the occurrence of an open-to-closed motion between the N- and C-terminal Rossmann fold domains has been predicted and experimentally demonstrated to occur in PimA and other GT-B glycosyltransferases (
      • Giganti D.
      • Alegre-Cebollada J.
      • Urresti S.
      • Albesa-Jové D.
      • Rodrigo-Unzueta A.
      • Comino N.
      • Kachala M.
      • López-Fernández S.
      • Svergun D.I.
      • Fernández J.M.
      • Guerin M.E.
      Conformational plasticity of the essential membrane-associated mannosyltransferase PimA from mycobacteria.
      ,
      • Guerin M.E.
      • Schaeffer F.
      • Chaffotte A.
      • Gest P.
      • Giganti D.
      • Korduláková J.
      • van der Woerd M.
      • Jackson M.
      • Alzari P.M.
      Substrate-induced conformational changes in the essential peripheral-membrane binding mannosyltransferase PimA from mycobacteria. Implications for catalysis.
      ,
      • Albesa-Jové D.
      • Giganti D.
      • Jackson M.
      • Alzari P.M.
      • Guerin M.E.
      Structure-function relationships of membrane-associated GT-B glycosyltransferases.
      ,
      • Albesa-Jové D.
      • Guerin M.E.
      The conformational plasticity of glycosyltransferases.
      ,
      • Wang S.
      • Zhao Y.
      • Yi L.
      • Shen M.
      • Wang C.
      • Zhang X.
      • Yang J.
      • Peng Y.L.
      • Wang D.
      • Liu J.
      Crystal structures of Magnaporthe oryzae trehalose-6-phosphate synthase (MoTps1) suggest a model for catalytic process of Tps1.
      ).
      Figure thumbnail gr1
      Figure 1Crystal structures of PimA. (A) In its inactive state, PimA adopts two conformations (PDB entry 4NC9, apo state of PimA) with different tertiary structures in the reshuffling region (displayed in rainbow color) but identical topology. In the inactive, extended form (orange), two helices (α4 and α5) point away from the N-terminal domain and fold back along hinge loops onto the N-terminal domain in the inactive, compact conformation (yellow). Note that the loop connecting α4 and α5 is only partially resolved in the crystal structure of the inactive, extended state. Transition to the active state (PDB entry 2GEJ, PimA in the presence of GDP-Man) is accompanied by a fold-switch in the reshuffling region (red). In the crystal structures, R144 is shown/modeled as spheres in green. Selected secondary structure elements of PimA, including the reshuffling region (residues 118–163), as observed in the unliganded (inactive) compact and extended conformations and the liganded (active) PimA-GDP-Man complex, are shown. β5, β6, and β7 are shown in blue, light blue, and orange, respectively. (B) 19F spectra of PimAT126C-V359C-R144C-TFA in 10% D2O (top), PimAR144C-TFA in 100% D2O (center), and PimAR144C-TFA in 10% D2O (bottom). (C) 19F spectra of PimAR144C-TFA in the absence (blue) and at increasing concentrations of GDP-Man. Intensities are corrected for dilution effects.
      Here, we demonstrate by using 19F NMR methods that motions of PimA occur on a second-to-millisecond timescale and include interconversions between several conformational states. Our observations provide a basis to suggest an activation mechanism for PimA in which the addition of substrate accelerates dynamics between active and inactive conformations.

      Results

       19F labeling is a sensitive approach to study dynamics of PimA by NMR

      19F nuclei are NMR active with a gyromagnetic ratio that is 0.94 times that of protons, have 100% natural abundance, and display high sensitivity to their magnetic environment (
      • Gerig J.
      Fluorine NMR of proteins.
      ,
      • Kitevski-LeBlanc J.L.
      • Prosser R.S.
      Current applications of 19F NMR to studies of protein structure and dynamics.
      ,
      • Didenko T.
      • Liu J.J.
      • Horst R.
      • Stevens R.C.
      • Wüthrich K.
      Fluorine-19NMR of integral membrane proteins illustrated with studies of GPCRs.
      ,
      • Manglik A.
      • Kim T.H.
      • Masureel M.
      • Altenbach C.
      • Yang Z.
      • Hilger D.
      • Lerch M.T.
      • Kobilka T.S.
      • Thian F.S.
      • Hubbell W.L.
      • Prosser R.S.
      • Kobilka B.K.
      Structural insights into the dynamic process of β2-adrenergic receptor signaling.
      ). At the same time, fluorine is nearly absent from biological molecules, reducing the number of resonances observed in a spectrum to those that stem from artificially introduced 19F labels. Therefore, 19F nuclei are suitable probes that allow employing NMR experiments for dynamic studies even for large proteins for which more common labeling approaches are not successful (
      • Didenko T.
      • Liu J.J.
      • Horst R.
      • Stevens R.C.
      • Wüthrich K.
      Fluorine-19NMR of integral membrane proteins illustrated with studies of GPCRs.
      ). Based on analyses of the different crystallographic structures of PimA (Fig. 1A), we introduced a Cys mutation in position Arg144 (here termed PimAR144C) (Fig. 1A) and labeled it with 1,1,1-trifluoroacetone (TFA), giving rise to a 19F labeled mutant PimAR144C-TFA (see Experimental procedures). The introduced 19F label of the mutant PimAR144C-TFA is located in different chemical environments in the different crystallographic conformations of PimA, as highlighted in Fig. 1A. In the active state, the residue is buried inside the protein and, thus, not exposed to solvent, while it is solvent-exposed in the inactive states. 19F nuclei have been shown to be sensitive to such differences in the hydrophobicity of their immediate environment (
      • Manglik A.
      • Kim T.H.
      • Masureel M.
      • Altenbach C.
      • Yang Z.
      • Hilger D.
      • Lerch M.T.
      • Kobilka T.S.
      • Thian F.S.
      • Hubbell W.L.
      • Prosser R.S.
      • Kobilka B.K.
      Structural insights into the dynamic process of β2-adrenergic receptor signaling.
      ,
      • Evanics F.
      • Kitevski J.L.
      • Bezsonova I.
      • Forman-Kay J.
      • Prosser R.S.
      19F NMR studies of solvent exposure and peptide binding to an SH3 domain.
      ). Thus, it can be expected that the 19F probe located at Cys144 can sense motions that we hypothesize to occur in the reshuffling region, which are 1) extended to compact domain motion and 2) fold-switching dynamics between the inactive and active states.
      Given that the probe experiences sufficiently distinct chemical environments, domain motions that are typically on the order of milliseconds can be probed by NMR relaxation dispersion experiments (
      • Carver J.
      • Richards R.
      A general two-site solution for the chemical exchange produced dependence of T2 upon the Carr-Purcell pulse separation.
      ). In addition, fold-switching dynamics are expected to occur on a second timescale and can be probed by saturation transfer experiments (
      • Spoerner M.
      • Hozsa C.
      • Poetzl J.A.
      • Reiss K.
      • Ganser P.
      • Geyer M.
      • Kalbitzer H.R.
      Conformational states of human rat sarcoma (Ras) protein complexed with its natural ligand GTP and their role for effector interaction and GTP hydrolysis.
      ).
      We first determined that the introduced label does not alter the three-dimensional structure of PimA. Supporting this notion, wild-type PimA, PimAR144C, and labeled PimAR144C-TFA mutants displayed similar far-UV CD spectra (Fig. S1A) (
      • Guerin M.E.
      • Schaeffer F.
      • Chaffotte A.
      • Gest P.
      • Giganti D.
      • Korduláková J.
      • van der Woerd M.
      • Jackson M.
      • Alzari P.M.
      Substrate-induced conformational changes in the essential peripheral-membrane binding mannosyltransferase PimA from mycobacteria. Implications for catalysis.
      ). In addition, the use of thermal unfolding experiments followed by CD has been extensively used to measure protein-ligand binding interaction (
      • Waldron T.T.
      • Murphy K.P.
      Stabilization of proteins by ligand binding: application to drug screening and determination of unfolding energetics.
      ,
      • Greenfield N.J.
      Using circular dichroism collected as a function of temperature to determine the thermodynamics of protein unfolding and binding interactions.
      ,
      • Létourneau D.
      • LeHoux J.G.
      • Lavigne P.
      Determination of ligand binding affinity and specificity of purified START domains by thermal shift assays using circular dichroism.
      ,
      • Mayhood T.W.
      • Windsor W.T.
      Ligand binding affinity determined by temperature-dependent circular dichroism: cyclin-dependent kinase 2 inhibitors.
      ). The midpoint temperature of unfolding for the unliganded protein is referred to as Tm. It is expected that the addition of ligands to the protein sample will increase the Tm. The higher the affinity of a ligand, the higher the apparent Tm value, because it requires more energy to dissociate higher affinity compounds (
      • Mayhood T.W.
      • Windsor W.T.
      Ligand binding affinity determined by temperature-dependent circular dichroism: cyclin-dependent kinase 2 inhibitors.
      ). We have previously described that the addition of GDP or GDP-Man to wild-type PimA increases the Tm of the enzyme, with the β-phosphate of the nucleotide group playing a prominent role (
      • Guerin M.E.
      • Schaeffer F.
      • Chaffotte A.
      • Gest P.
      • Giganti D.
      • Korduláková J.
      • van der Woerd M.
      • Jackson M.
      • Alzari P.M.
      Substrate-induced conformational changes in the essential peripheral-membrane binding mannosyltransferase PimA from mycobacteria. Implications for catalysis.
      ). As depicted in Fig. S1B, thermal unfolding followed by the far-UV CD signal at 222 nm indicated slight differences in protein stability between unliganded WT PimA, PimAR144C, and PimAR144C-TFA. Interestingly, after the addition of GDP, wild-type PimA, PimAR144C, and PimAR144C-TFA displayed a clear increase in the Tm values (Table S1), supporting the notion that the mutation and labeling do not significantly perturb the enzymatic structure and function (Fig. S1).

       Active and inactive states of PimA show distinct resonances

      PimA contains two Cys residues buried inside the protein core that are not expected to be labeled during the labeling procedure (see Experimental procedures). To verify this, the labeling scheme was applied to WT PimA. Only resonances that are less than 5 Hz wide were observed in a 19F spectrum (Fig. S2), showing that they stem from small molecules. The resonances could be assigned to residual 3-bromo-1,1,1-trifluoroacetone (BTFA) label and 1,1,1-trifluoroacetone (TFA), the latter being the product of an incomplete labeling reaction. The resonance at −119.8 ppm stems from an impurity in the BTFA solution. Therefore, no labeling was observed for WT PimA. We conclude that PimAR144C-TFA is only labeled at the introduced Cys residue.
      Despite being singly labeled, two resonances are observed at −84.1 ppm and −82.7 ppm (Fig. 1B, bottom). As discussed below, these two peaks display saturation transfer, further corroborating that they stem from the same 19F label. To assign the resonances to conformations of PimA, we investigated the 19F spectrum of the triple mutant PimAT126C-V359C-R144C-TFA, previously shown to be locked in the inactive conformation. This variant cannot undergo reshuffling due to a disulfide bond formed between Cys126 and Cys359; these residues are in close spatial proximity in the inactive state but far apart in the active state (
      • Giganti D.
      • Albesa-Jové D.
      • Urresti S.
      • Rodrigo-Unzueta A.
      • Martínez M.A.
      • Comino N.
      • Barilone N.
      • Bellinzoni M.
      • Chenal A.
      • Guerin M.E.
      • Alzari P.M.
      Secondary structure reshuffling modulates glycosyltransferase function at the membrane.
      ). PimAT126C-V359C-R144C-TFA displays only a single resonance at −84.1 ppm (Fig. 1B, top), which therefore can be assigned to the inactive state. Consequently, the resonance at −82.7 ppm is assigned to the active state. This assignment is corroborated by the fact that the resonance at −84.1 ppm shifts 0.1 ppm upfield in 100% D2O, whereas the resonance at −82.7 ppm does not shift (Fig. 1B, center). For fully solvent-exposed 19F labels, upfield solvent isotope shifts of 0.2–0.3 ppm have been observed when comparing shifts in a 90% H2O/10% D2O to 100% D2O solvent (
      • Evanics F.
      • Kitevski J.L.
      • Bezsonova I.
      • Forman-Kay J.
      • Prosser R.S.
      19F NMR studies of solvent exposure and peptide binding to an SH3 domain.
      ,
      • Hull W.E.
      • Sykes B.D.
      Fluorine-19 nuclear magnetic resonance study of fluorotyrosine alkaline phosphatase: the influence of zinc on protein structure and a conformational change induced by phosphate binding.
      ), indicating that the resonance at −84.1 ppm stems from partially exposed 19F label. In the crystal structure, Arg144 is shielded from solvent in the active state, whereas this is not the case for the inactive conformations (Fig. 1A), supporting the assignment. The assignment is further supported by extensive molecular dynamics (MD) simulations performed on the different states of PimA labeled with TFA. Based on these simulations, we calculated the solvent-accessible surface area (SASA) of the TFA label in the different states. A higher SASA value indicates that the label is more solvent-exposed. As shown in Fig. S3, the SASA of the TFA label in the active state is 19.3 Å2, which differs markedly from the values obtained for the protein in the inactive forms (96.7 Å2 and 60.1 Å2 for the extended and compact structures, respectively). Thus, MD simulations revealed that the TFA label in the active state is less surface exposed than that in the inactive state, in agreement with the NMR experiments.

       PimA undergoes slow conformational exchange between the inactive and active states

      To test whether the two resonances in the 19F spectrum of PimAR144C-TFA are in slow exchange, we conducted saturation transfer experiments at increasing saturation times. Exchange rates were obtained from a fit to the model given in (Eq. 4), (Eq. 5), (Eq. 6), (Eq. 7). Experimental data to obtain R1 longitudinal relaxation rates that are required for the analysis are shown in Fig. S4 and Table S2. Figure 2 shows resonance intensity attenuation due to slow conformational exchange for the inactive and active state resonances upon saturation of the active or inactive state resonance, respectively.
      Figure thumbnail gr2
      Figure 2Slow conformational exchange of PimAR144C-TFA. (A) Intensity attenuation of the active state resonance in absence (black) and presence (orange) of saturating concentrations of GDP-Man upon saturation of the inactive state resonance. (B) Intensity attenuation of the inactive state resonance in the absence (black) and presence (orange) of saturating concentrations of GDP-Man upon the saturation of the active state resonance. Experiments were conducted with 100 μm PimAR144C-TFA and 500 μm GDP-Man, where applicable. (C) Secondary structure representation of the reshuffling region. The position of the mutation (R144C) is indicated as a green dot. Parameters (p, populations; ks, slow exchange rates) are obtained from the fit to data in panels A and B.
      In the absence of GDP-Man, the exchange rate was determined to be 2.3 s−1. These dynamics correspond to fold-switching between the active and inactive states of the protein, as demonstrated by the assignments (Fig. 1B). The second timescale of the motion is in accordance with studies of the chemokine lymphotactin, for which fold-switching dynamics were shown to be on the order of seconds (
      • Volkman B.F.
      • Liu T.Y.
      • Peterson F.C.
      Lymphotactin structural dynamics.
      ). In the absence of GDP-Man, the active and inactive states of PimA are nearly equally populated (Fig. 2C and Table 1).
      Table 1Parameters obtained from saturation transfer experiments for (A) the active state resonance and (B) for the inactive state resonance of PimAR144C-TFA
      Condition and resonance stateChemical shift (ppm)kIA
      I, the inactive state; A, the active state, i.e. kIA is the exchange rate from the inactive to the active state, and ΔGIA is the free energy difference between the inactive and active state and vice versa.
      (s−1)
      pIpAks (s−1)ΔGIA (kcal mol−1)
      Active state
      Apo−82.71.08 ± 0.050.47 ± 0.042.31 ± 0.09−0.070 ± 0.090
      GDP-Man−82.64.2 ± 0.20.34 ± 0.0413 ± 1−0.390 ± 0.180
      Inactive state
      Apo−84.11.23 ± 0.040.53 ± 0.042.31 ± 0.090.070 ± 0.090
      GDP-Man−84.38.3 ± 0.90.66 ± 0.1213 ± 10.390 ± 0.180
      a I, the inactive state; A, the active state, i.e. kIA is the exchange rate from the inactive to the active state, and ΔGIA is the free energy difference between the inactive and active state and vice versa.
      To test how GDP-Man affects these dynamics, we first conducted a titration experiment to corroborate that the sugar donor saturates the binding site. As shown in Fig. 1C, titration of GDP-Man to 100 μm PimAR144C-TFA leads to a gradual decrease in intensities of both resonances up to a nucleotide sugar concentration of 200 μm. At this concentration, the inactive state resonance shifts 0.2 ppm upfield, and further addition of GDP-Man does not alter the spectrum anymore. At the field strength employed in Fig. 1C (16.4 T), the active state resonance is too broad to clearly observe a resonance shift. Since the line broadening due to the field-dependent chemical shift anisotropy of 19F is reduced at lower field strength, we compared the spectra of PimAR144C-TFA in the presence and absence of GDP-Man at 14.1 T. At this lower field, a downfield shift of the active state resonance of 0.1 ppm was observable at 500 μm GDP-Man and 100 μm PimAR144C-TFA (Fig. S5). Moreover, both peaks broaden upon addition of GDP-Man, indicating that the presence of GDP-Man alters dynamics and/or populations of PimA. This process may be related to GDP-Man binding itself or to the dynamic processes observed in the reshuffling region of PimA, as discussed below. GDP-Man and GDP were previously seen to bind PimA in enthalpy-driven reactions with dissociation constants (KD) of 0.23 μmH = –15.6 kcal·mol−1) and 0.03 μm (ΔH= –14.0 kcal·mol−1), respectively (
      • Guerin M.E.
      • Kordulakova J.
      • Schaeffer F.
      • Svetlikova Z.
      • Buschiazzo A.
      • Giganti D.
      • Gicquel B.
      • Mikusova K.
      • Jackson M.
      • Alzari P.M.
      Molecular recognition and interfacial catalysis by the essential phosphatidylinositol mannosyltransferase PimA from mycobacteria.
      ). Thus, the submicromolar KD indicates that under the experimental conditions employed here, 500 μm GDP-Man and 100 μm PimAR144C-TFA binding of GDP-Man to PimA is fully saturated. Thus, we can exclude any contribution of GDP-Man binding to the observed dynamics discussed subsequently.
      Under the condition that GDP-Man is present at saturating concentrations, the fold-switching dynamics of PimAR144C-TFA are strongly enhanced, and the exchange rate was observed to increase to 13 s−1 (Fig. 2 and Table 1). A concurrent shift in the population equilibrium toward the inactive states was observed (by around 10%), as evidenced by analyzing peak integrals as well as the saturation transfer experiments (Fig. 2, Fig. S5, and Table S3).

       The inactive state of PimA undergoes millisecond motions between a compact and an extended state

      To capture possible dynamics on the millisecond timescale, we conducted Carr-Purcell-Meiboom-Gill (CPMG) relaxation dispersion experiments on PimAR144C-TFA at two magnetic field strengths in the absence and presence of saturating amounts of GDP-Man. Assuming a two-state exchange, we fitted the data to Equation 3 (see Experimental procedures) to obtain an estimate of the size of the populations, the exchange rate between them, and the chemical shift difference due to the conformational exchange. Data that were acquired at one magnetic field strength only were fitted to Equation 2 to obtain an apparent exchange rate.
      The apparent exchange rate of the resonance representing the inactive state of PimAR144C-TFA and the triple mutant PimAT126C-V359C-R144C-TFA locked into the inactive state are identical (Fig. S6). Although the triple mutant cannot undergo reshuffling to the active state, the faster domain dynamics of the inactive state itself are not affected. This is expected from inspecting the structure of PimA. While the disulfide bond between Cys126 and Cys359 prevents secondary structure reshuffling, it does not impede the extended-to-compact motion of the inactive states (Fig. 1A). Thus, the relaxation dispersion displayed by the inactive state resonance monitors the extended-to-compact dynamics of the reshuffling region.
      As shown in Fig. 3A, the resonance in the 19F spectrum of PimAR144C-TFA monitoring the inactive state displays relaxation dispersion between two conformations with different populations, and the profile is affected by the addition of saturating amounts of GDP-Man, reflecting changes in dynamics and population equilibria. In the absence of GDP-Man, the exchange rate is 8000 s−1, which decreases to 5100 s−1 in the presence of GDP-Man (Table 2). Similar exchange rates have previously been observed for motions of a helix of the β2-adrenergic receptor (
      • Manglik A.
      • Kim T.H.
      • Masureel M.
      • Altenbach C.
      • Yang Z.
      • Hilger D.
      • Lerch M.T.
      • Kobilka T.S.
      • Thian F.S.
      • Hubbell W.L.
      • Prosser R.S.
      • Kobilka B.K.
      Structural insights into the dynamic process of β2-adrenergic receptor signaling.
      ). In the absence of GDP-Man, one of the states is highly populated and the other one is not (1% versus 99%). Upon addition of GDP-Man, the less populated of the inactive states becomes slightly more populated (3% versus 97%), which is accompanied by an upfield resonance shift of the inactive state resonance (Fig. 1C), indicating that the observed change in populations and dynamics results in a more polar environment, on average, for the 19F label. While the population change upon addition of GDP-Man (1.7%) as determined from the relaxation dispersion data is significant, given the fitted errors in the analysis of the relaxation data (Table 2), it might not be possible to deconvolute model parameters entirely, given that experiments were acquired at two magnetic field strengths only; however, changes in dynamics and changes in populations toward the less populated state are also manifest in the observed upshift of the inactive state resonance in Fig. 1C, i.e. to the state in which the 19F label is located in a more polar environment. Resonances of 19F nuclei in a more polar environment are located upfield of 19F nuclei buried inside a protein (
      • Manglik A.
      • Kim T.H.
      • Masureel M.
      • Altenbach C.
      • Yang Z.
      • Hilger D.
      • Lerch M.T.
      • Kobilka T.S.
      • Thian F.S.
      • Hubbell W.L.
      • Prosser R.S.
      • Kobilka B.K.
      Structural insights into the dynamic process of β2-adrenergic receptor signaling.
      ,
      • Evanics F.
      • Kitevski J.L.
      • Bezsonova I.
      • Forman-Kay J.
      • Prosser R.S.
      19F NMR studies of solvent exposure and peptide binding to an SH3 domain.
      ). This is in agreement with our observation that the inactive state resonance, for which the 19F label is more exposed to the solvent than for the active state resonance (Fig. 1B), is upfield of the active state resonance. Based on the crystal structures (Fig. 1A) and our MD simulations (Fig. S3), the 19F label is most solvent-exposed in the inactive, extended conformation. Thus, we identify the sparsely populated state with the inactive, extended conformation. This conclusion is further corroborated by structural features of PimA.
      Figure thumbnail gr3
      Figure 3Intermediate conformational exchange of the inactive and the active states of PimAR144C-TFA. (A) Relaxation dispersion of the inactive state resonance in the absence (black) and presence (orange) of saturating concentrations of GDP-Man. (B) Extended-to-compact dynamics in the reshuffling region (extended conformation in orange; compact conformation in yellow). Parameters (p, populations; Δδ, chemical shift difference) obtained from the fit to data in panel A are indicated (black, no ligand; orange, with GDP-Man). (C) Relaxation dispersion of the active state resonance in the absence (black) and presence (orange) of saturating concentrations of GDP-Man. (D) Extended-to-compact dynamics that occur in the reshuffling region (extended conformation in gray; compact conformation in red). Only the compact, active conformation is structurally characterized. Parameters (p, populations; Δδ, chemical shift difference) obtained from the fit to the data in panel C are indicated (black, no ligand; orange, with GDP-Man). Experiments were conducted with 100 μm PimAR144C-TFA and 500 μm GDP-Man where applicable. All observed exchange rates are on the order of 4000–8000 s−1 (see for details).
      Table 2Parameters obtained from CPMG relaxation dispersion experiments of PimAR144C-TFA fitted to Equation 3
      ConditionConformationChemical shift (ppm)ki (103 s−1)pextendedpcompact|Δδ|
      |Δδ| is a global variable for each peak.
      (ppm)
      ApoActive−82.74.3 ± 1.30.81 ± 0.030.19 ± 0.031.2 ± 0.1
      GDP-ManActive−82.67.3 ± 1.50.70 ± 0.130.30 ± 0.131.2 ± 0.1
      ApoInactive−84.18.0 ± 8.00.009 ± 0.0020.991 ± 0.0022.9 ± 0.3
      GDP-ManInactive−84.35.1 ± 2.70.026 ± 0.0080.974 ± 0.0082.9 ± 0.3
      a |Δδ| is a global variable for each peak.
      The conformational transition from the compact to extended inactive state of PimA, in the absence of GDP-Man, was reproduced by steered-MD simulations, as shown in Fig. S7. This type of calculation forces the system to evolve away from its initial equilibrium condition (compact, inactive) to a final state (extended, inactive), thereby accelerating transitions between different energy minima (
      • Park S.
      • Schulten K.
      Calculating potentials of mean force from steered molecular dynamics simulations.
      ,
      • Adcock S.A.
      • McCammon J.A.
      Molecular dynamics: survey of methods for simulating the activity of proteins.
      ,
      • Li M.S.
      • Mai B.K.
      Steered molecular dynamics—a promising tool for drug design.
      ). This method is generally applied in studying many biophysical processes, including the (un)folding mechanism of proteins (
      • Lu H.
      • Isralewitz B.
      • Krammer A.
      • Vogel V.
      • Schulten K.
      Unfolding of titin immunoglobulin domains by steered molecular dynamics simulation.
      ,
      • Isralewitz B.
      • Baudry J.
      • Gullingsrud J.
      • Kosztin D.
      • Schulten K.
      Steered molecular dynamics investigations of protein function.
      ). As revealed by the steered-MD simulations, the inactive, extended state’s large hydrophobic patches are exposed, making this state potentially aggregation-prone, whereas the hydrophobic patches are shielded from solvent in the inactive, compact state (
      • Giganti D.
      • Albesa-Jové D.
      • Urresti S.
      • Rodrigo-Unzueta A.
      • Martínez M.A.
      • Comino N.
      • Barilone N.
      • Bellinzoni M.
      • Chenal A.
      • Guerin M.E.
      • Alzari P.M.
      Secondary structure reshuffling modulates glycosyltransferase function at the membrane.
      ). The fact that the addition of GDP-Man leads to an increase in population of the inactive, extended state suggests that the slower reshuffling event occurs in that state.

       The active state of PimA undergoes millisecond motions that may reflect open-to-closed dynamics

      As shown in Fig. 3C, the resonance in the 19F spectrum of PimAR144C-TFA monitoring the active states also displays relaxation dispersion, and the profile is again affected by the addition of saturating amounts of GDP-Man, reflecting changes in dynamics and population equilibria. In contrast to the inactive state resonance, the active state resonance is downshifted upon addition of GDP-Man (Fig. S5). This indicates that a population shift toward a more hydrophobic environment for the 19F label occurs, which might be identified with a population shift from a structurally uncharacterized active, extended state to the active, compact state of PimA. Based on this interpretation, most of PimA is found in the active, extended state (81% extended versus 19% compact). This holds true even after addition of GDP-Man (70% extended versus 30% compact) (Fig. 3D). The active, extended state might be represented by the open version of the already structurally characterized active, closed state, suggesting that these dynamics correspond to the open-to-closed motion of PimA (Fig. 1A). The exchange rate between the extended and compact forms of the active states in this case increases from 4300 s−1 to 8000 s−1 upon addition of saturating concentrations of GDP-Man.
      Interestingly, accelerated MD simulations performed on PimA labeled with TFA and starting from the compact, active state indicated that the protein undergoes a conformational change to reach a state in which β7 dissociates from β6, which indicates the initiation of the transition to α-helix (Fig. S8). As steered-MD simulations, this computational approach is useful for improving conformational space sampling. The method does not require any predefined coordinates and reduces the energy barrier between the different low-energy states (
      • Hamelberg D.
      • Mongan J.
      • McCammon J.A.
      Accelerated molecular dynamics: a promising and efficient simulation method for biomolecules.
      ).

       The active and inactive states are structurally distinct

      The relaxation dispersion data indicate that the inactive, compact and inactive, extended states are structurally distinct, as their resonances are separated by 1.2 ppm (Table 2). As discussed above, the resonance of the inactive, extended state is expected to be upfield of the resonance of the inactive, compact state, since the 19F label is more solvent-exposed in the latter (Fig. 1C). Similarly, the relaxation dispersion data indicate that the active, compact and active, extended state are structurally distinct, as their resonances are separated by 2.9 ppm (Table 2). The downfield shift of the resonance of the active, compact state compared with the active, extended state (Fig. S5) indicates that in the active, compact state the 19F label is located in a more hydrophobic environment. Importantly, from the relaxation dispersion data (Table 2) and the 1D spectra (Fig. 1C, Fig. S5), it can be concluded that all conformational states are structurally distinct, indicating that fold-switching is accompanied by a dramatic structural rearrangement near the 19F label. This is consistent with the structural data that show that residue Arg144 is reshuffled from an α-helix (potentially via a solvent-exposed random coil structure) to a β-sheet environment.

      Discussion

      The ability of an enzyme to adopt many different conformations during the catalytic cycle is due to its great flexibility, playing a critical role in the recruitment of substrates, release of products, stabilization of transition states, and adaptation to the cell environment. Employing 19F NMR, we identified four functionally relevant states of PimA that coexist in dynamic equilibria in solution undergoing conformational exchange on timescales from milliseconds to seconds. Specifically, fold-switching, a hitherto poorly characterized process, occurs in the reshuffling region of the N-terminal domain of PimA on a second timescale. Domain motions, on the other hand, occur on a millisecond timescale.
      The observations presented here allow us to propose that the activation of PimA occurs along a sequence of structural rearrangements displayed in Fig. 4. In Fig. S9, a free-energy diagram of the activation process is depicted that is derived from thermodynamic data obtained from the saturation transfer and relaxation dispersion experiments (Table 1, Table 2). Initially, PimA is inactive and in a compact state that shields hydrophobic patches in the reshuffling region. Almost all inactive PimA molecules are in that conformation but are in millisecond exchange with the inactive, extended state. This state exposes hydrophobic patches and, thus, is sparsely populated. However, it offers sufficient degrees of freedom for a rearrangement of secondary structure elements on a second timescale. A drastic change in chemical shift indicates that the inactive, extended state is structurally distinct from the uncharacterized active, extended state, which might correspond to the open state of the structurally characterized closed, GDP-Man-bound conformation of PimA. Following the fold-switch, which we speculate to occur from the inactive, extended state, PimA is in the active, extended state and in millisecond exchange with the active, compact state, which is the conformation in which catalysis occurs. We propose that fold-switching is an activation mechanism of PimA, and since it is a slow process, it may represent the rate-limiting step of the glycosylation reaction catalyzed by PimA. The substrate, GDP-Man, acts as an activator and, thus, functional regulator of PimA, as the binding of GDP-Man enhances the fold-switching dynamics. Intriguingly, a significant population of the inactive state is always present. This may point to a more significant role of the inactive states in the catalytic cycle of PimA than previously anticipated, for instance, in membrane binding of PimA or to facilitate efficient recruitment of phosphatidyl-myo-inositol lipids, which are acceptor substrates of PimA.
      Figure thumbnail gr4
      Figure 4Activation mechanism of PimA. Structural changes of PimA upon activation. Initially, almost all inactive PimA molecules are in the compact conformation but are in millisecond exchange with the inactive, extended state. Following the fold-switch, PimA is in the structurally uncharacterized active, extended state and in millisecond exchange with the active, compact state, which is the conformation in which catalysis occurs. The location of R144 is shown/modeled for clarity.

       Conclusions

      This study demonstrates that the 19F-based NMR approach presented here allows for insights into structures and dynamics for proteins that cannot be studied with more conventional NMR methods. The methods employed here, a combination of saturation transfer methods and relaxation dispersion experiments, allow investigations of dynamics occurring simultaneously and on a wide range of timescales in solution. Here, this approach revealed a remarkable situation in which PimA exists in at least four different conformations that are all present in solution. The dynamics between and the relative populations of these conformations are affected by the addition of substrate, which demonstrates that the equilibria are finely tuned to optimize the catalytic cycle.

      Materials and Methods

       Materials

      1,1,1-Trifluoroacetone (TFA), 3-bromo-1,1,1-trifluoroacetone (BTFA), guanosine 5′-diphospho-d-mannose (GDP-Man), and tris(2-carboxyethyl)phosphine (TCEP) were acquired from Merck (Darmstadt, Germany).

       Cloning and purification of WT PimA and PimA mutants

      PimAR144C and PimAT126C-V359C-R144C mutants of PimA from Mycobacterium smegmatis were obtained from GenScript using pET29a-pimA as the DNA template (
      • Guerin M.E.
      • Schaeffer F.
      • Chaffotte A.
      • Gest P.
      • Giganti D.
      • Korduláková J.
      • van der Woerd M.
      • Jackson M.
      • Alzari P.M.
      Substrate-induced conformational changes in the essential peripheral-membrane binding mannosyltransferase PimA from mycobacteria. Implications for catalysis.
      ). Recombinant WT PimA and the PimAR144C mutant were expressed in Escherichia coli BL21(DE3) pLysS and purified as described previously (
      • Guerin M.E.
      • Schaeffer F.
      • Chaffotte A.
      • Gest P.
      • Giganti D.
      • Korduláková J.
      • van der Woerd M.
      • Jackson M.
      • Alzari P.M.
      Substrate-induced conformational changes in the essential peripheral-membrane binding mannosyltransferase PimA from mycobacteria. Implications for catalysis.
      ). PimAT126C-V359C-R144C mutant was expressed in E. coli Rosetta Gami 2(DE3) pLysS and purified as described previously (
      • Giganti D.
      • Albesa-Jové D.
      • Urresti S.
      • Rodrigo-Unzueta A.
      • Martínez M.A.
      • Comino N.
      • Barilone N.
      • Bellinzoni M.
      • Chenal A.
      • Guerin M.E.
      • Alzari P.M.
      Secondary structure reshuffling modulates glycosyltransferase function at the membrane.
      ).

       Labeling of PimA

      An established protocol was adapted to label PimA with TFA (
      • Chung K.Y.
      • Kim T.H.
      • Manglik A.
      • Alvares R.
      • Kobilka B.K.
      • Prosser R.S.
      Role of detergents in conformational exchange of a G protein-coupled receptor.
      ). 1 mm TCEP was added to 100–200 μm PimA and incubated for 1 h on ice, except for PimAT126C-V359C-R144C-TFA, to which no TCEP was added. Next, BTFA was added in 10-fold molar excess (1–2 mm), and the solution was incubated overnight at 4 °C under gentle agitation. Finally, the solution was passed through a desalting column equilibrated with 50 mm Tris-HCl, pH 7.5, and 150 mm NaCl. The sample conditions for NMR experiments were 100 μm PimA, 50 mm Tris-HCl, pH 7.5, 150 mm NaCl, and 10% D2O. 500 μm GDP-Man was added where appropriate and if not stated otherwise.
      To exchange to 100% D2O solvent, 100 μm PimA was concentrated roughly five times and diluted back to a concentration of 100 μm using 50 mm Tris-HCl, 150 mm NaCl in 100% D2O. This procedure was repeated 5 times.

       NMR spectroscopy

      Experiments were performed at 25 °C on a Bruker Avance 600 MHz III HD spectrometer (19F Larmor frequency of 564 MHz). Relaxation dispersion and titration data for PimAR144C-TFA were additionally acquired on a Bruker Avance 700 MHz spectrometer (19F Larmor frequency of 659 MHz). Both spectrometers were equipped with a 19F cryoprobe. Typical π/2 pulse lengths were 15–19 µs, and the acquisition time was 0.03 s. 10% D2O was added to the samples for frequency locking. For 1D spectra, 400 transients were acquired with a relaxation delay of 1 s. Data were processed in Topspin (Bruker, Billerica, MA) and analyzed using in-house Matlab routines (MathsWorks, Natick, MA) and Igor Pro (WaveMetrics, Lake Oswego, OR), employing an in-house script for nonlinear fitting of relaxation dispersion data (
      • Abelein A.
      • Gräslund A.
      • Danielsson J.
      Zinc as chaperone-mimicking agent for retardation of amyloid β peptide fibril formation.
      ). Overlapping resonances were deconvoluted, employing iNMR software (MestreLab Research, Spain) by fitting spectra to double-Lorentzian functions.

       Far-UV CD analysis

      Spectra were acquired in a J-810 CD spectropolarimeter (Jasco Corp., Tokyo, Japan) by using Hellma 105.200-QS quartz cuvettes with a 1 cm optical path. Spectra were recorded in a continuous mode with 1 nm bandwidth, 1 s response, and a scan speed of 100 nm/min−1. Samples were 2.5 μm PimA, PimAR144C, or PimAR144C-TFA in 10 mm Tris-HCl, pH 7.5. GDP was added at a 1:15 ratio, and 25 scans were accumulated to obtain the final spectra, which were further corrected for the baseline signal. Spectra were recorded in the 200–250 nm range at 20 °C.

       Temperature scans

      Spectra were acquired in a J-810 CD spectropolarimeter (Jasco Corp., Tokyo, Japan) by using Hellma 110-QS quartz cuvettes with a 1 mm optical path by using a Peltier thermal device, allowing temperature control during the experiments. Spectra were recorded in a continuous mode with 1 nm bandwidth, 1 s response, and a scan speed of 100 nm/min−1. Samples were 10 μm PimA, PimAR144C, or PimAR144C-TFA in 10 mm Tris-HCl, pH 7.5. Substrates were added at a 1:15 ratio. In this case, thermal dependencies of the ellipticity were monitored in the range from 20 °C to 90 °C at 222 nm. Temperature was increased stepwise by 1 °C/min.

       Relaxation dispersion experiments

      1D 19F CPMG relaxation dispersion experiments (
      • Carver J.
      • Richards R.
      A general two-site solution for the chemical exchange produced dependence of T2 upon the Carr-Purcell pulse separation.
      ,
      • Luz Z.
      • Meiboom S.
      Nuclear magnetic resonance study of the protolysis of trimethylammonium ion in aqueous solution—order of the reaction with respect to solvent.
      ) were performed analogously to previous studies (
      • Manglik A.
      • Kim T.H.
      • Masureel M.
      • Altenbach C.
      • Yang Z.
      • Hilger D.
      • Lerch M.T.
      • Kobilka T.S.
      • Thian F.S.
      • Hubbell W.L.
      • Prosser R.S.
      • Kobilka B.K.
      Structural insights into the dynamic process of β2-adrenergic receptor signaling.
      ). 2400 transients were acquired during a constant time delay, TCPMG, of 3.84 ms for 14 refocusing frequencies, νCPMG, ranging between ∼130 s−1 and ∼6000 s−1. Effective transverse relaxation rates, R2,eff, were computed from peak intensities ICPMG) relative to the peak intensity, I0, in the absence of a constant time delay as follows
      R2,eff=-TCPMG-1lnIvCPMGI0
      (Eq. 1)


      For data measured at one magnetic field strength, relaxation dispersion data were fitted to Luz-Meiboom's two-state model to obtain an apparent exchange rate, kapp, of the dynamics (
      • Luz Z.
      • Meiboom S.
      Nuclear magnetic resonance study of the protolysis of trimethylammonium ion in aqueous solution—order of the reaction with respect to solvent.
      ):
      R2,effνCPMG=ckapp1-4νCPMGkapptanhkapp4νCPMG+R2,0
      (Eq. 2)


      where νCPMG=0.5τCPMG-1, τCPMG is the separation between two 180° pulses and kapp, R2,0, and c=pApBΔω2are parameters of the fit. The populations, pA and pB, of the two states and the chemical shift difference between the resonances, Δω, cannot be deconvoluted with this model.
      Relaxation dispersion data measured at two magnetic field strengths were fitted to Carver-Richards model assuming two states to obtain the exchange rate of the process, ki (i indicates intermediate exchange), populations, pA and pB, and the chemical shift difference, Δω, between the states (
      • Carver J.
      • Richards R.
      A general two-site solution for the chemical exchange produced dependence of T2 upon the Carr-Purcell pulse separation.
      ):
      pA=1-pB


      ξ=2ΔωpB-pAki


      ψ=pB-pA2ki2-Δω2+4pBpAki2


      η±=18νCPMG±ψ+ψ2+ξ2


      D±=12±1+ψ+2Δω2ψ2+ξ2


      R2,eff=R2,0+12ki-2νCPMGarcoshD+coshη+-D-cosη-
      (Eq. 3)


      To reduce the number of parameters in the fit, it was assumed that the addition of GDP-Man did not alter the conformations of PimA present in solution, i.e. the chemical shift Δδ=Δω/2πf, where f is the spectrometer frequency, was treated as a global constant. Moreover, the magnetic field strength does not affect protein kinetics, i.e. ki, pA, and pB were identical for data obtained at different magnetic field strengths but could vary between samples that contained or did not contain GDP-Man.

       T1 experiments

      Inversion recovery experiments with at least 10 delays, τ, ranging between 25 ms and 800 ms were performed to obtain the longitudinal relaxation time, T1. 1200 transients were acquired, and the relaxation delay was 2 s. The relaxation rate, R1 = T1−1, was obtained from a monoexponential fit of the intensities in dependence on delay τ.

       Saturation transfer experiments

      Saturation transfer experiments were conducted to quantify slow exchange dynamics of PimA (
      • Manglik A.
      • Kim T.H.
      • Masureel M.
      • Altenbach C.
      • Yang Z.
      • Hilger D.
      • Lerch M.T.
      • Kobilka T.S.
      • Thian F.S.
      • Hubbell W.L.
      • Prosser R.S.
      • Kobilka B.K.
      Structural insights into the dynamic process of β2-adrenergic receptor signaling.
      ,
      • Forsén S.
      • Hoffman R.A.
      Study of moderately rapid chemical exchange reactions by means of nuclear magnetic double resonance.
      ). Resonances were saturated with continuous wave irradiation (B1 field of 28 Hz). The resonance associated with the active state was irradiated at −82.3 ppm, and the intensity decay of the inactive state resonance was monitored at −84.1 ppm in the absence of substrate and at −84.3 ppm in the presence of GDP-Man. Conversely, the resonance associated with the inactive state was irradiated at −84.3 ppm, and the intensity decay of the active state resonance was monitored at −82.7 ppm in the absence of substrate and at −82.6 ppm in the presence of GDP-Man. Spectra were acquired for 8 saturation times, varying between 25 ms and 1000 ms. 1200 transients were recorded, and the relaxation delay was 1.5 s. Exchange rates were obtained by fitting residual intensities, Iτ, at increasing irradiation times τ to a two-state model (
      • Spoerner M.
      • Hozsa C.
      • Poetzl J.A.
      • Reiss K.
      • Ganser P.
      • Geyer M.
      • Kalbitzer H.R.
      Conformational states of human rat sarcoma (Ras) protein complexed with its natural ligand GTP and their role for effector interaction and GTP hydrolysis.
      ):
      Imτ=Im0kmn+R1,mkmnexp-τkmn+R1,m+R1,m
      (Eq. 4)


      Where m is inactive and n is active, representing the exchange rate, kmn, from the inactive to the active state and vice versa.
      The exchange rate for slow dynamics, ks (s indicates slow exchange), is the sum of kmn and knm:
      ks=kmn+knm
      (Eq. 5)


      Populations, p, of the states and the free-energy difference, ΔG, are given by
      pn=kmnks
      (Eq. 6)


      ΔGnm=-RTlnpnpm
      (Eq. 7)


      where T is the absolute temperature and R is the universal gas constant.

       Conventional MD simulations

      The MD simulations were performed with the AMBER 18 package (
      • Case D.A.
      • Ben-Shalom I.Y.
      • Brozell S.R.
      • Cerutti D.S.
      • Cheatham III, T.E.
      • Cruzeiro V.W.D.
      • Darden T.A.
      • Duke R.E.
      • Ghoreishi D.
      • Gilson M.K.
      • Gohlke H.
      • Goetz A.W.
      • Greene D.
      • Harris R.
      • Homeyer N.
      • et al.
      ). Three different simulations, starting from the active-compact, inactive-compact, and inactive-extended conformations, were performed. Parameters for the TFA label were generated with the antechamber module of AMBER implemented with the general Amber force field (GAFF) (
      • Wang J.
      • Wolf R.M.
      • Caldwell J.W.
      • Kollman P.A.
      • Case D.A.
      Development and testing of a general amber force field.
      ), with partial charges set to fit the electrostatic potential generated with HF/6-31G(d) by RESP (
      • Bayly C.I.
      • Cieplak P.
      • Cornell W.
      • Kollman P.A.
      A well-behaved electrostatic potential based method using charge restraints for deriving atomic charges: the RESP model.
      ). The charges were calculated according to the Merz-Singh-Kollman scheme using Gaussian 09 (
      • Frisch M.J.
      • Trucks G.W.
      • Schlegel H.B.
      • Scuseria G.E.
      • Robb M.A.
      • Cheeseman J.R.
      • Scalmani G.
      • Barone V.
      • Mennucci B.
      • Petersson G.A.
      • Nakatsuji H.
      • Li X.
      • Caricato M.
      • Marenich A.
      • Bloino J.
      • et al.
      ). Each protein was immersed in a water box with a 10 Å buffer of TIP3P (
      • Jorgensen W.L.
      • Chandrasekhar J.
      • Madura J.D.
      • Impey R.W.
      • Klein M.L.
      Comparison of simple potential functions for simulating liquid water.
      ) water molecules. The system was neutralized by adding explicit counter ions (Na+). A two-stage geometry optimization approach was performed. The first stage minimizes only the positions of solvent molecules and ions, and the second stage is an unrestrained minimization of all the atoms in the simulation cell. The systems then were gently heated by incrementally increasing the temperature from 0 to 300 K under a constant pressure of 1 atm and periodic boundary conditions. Harmonic restraints of 30 kcal·mol−1 were applied to the solute, and the Andersen temperature coupling scheme (
      • Andrea T.A.
      • Swope W.C.
      • Andersen H.C.
      The role of long ranged forces in determining the structure and properties of liquid water.
      ) was used to control and equalize the temperature. The time step was kept at 1 fs during the heating stages, allowing potential inhomogeneities to self-adjust. Water molecules are treated with the SHAKE algorithm such that the angle between the hydrogen atoms is kept fixed. Long-range electrostatic effects are modelled using the particle-mesh-Ewald method (
      • Darden T.
      • York D.
      • Pedersen L.
      Particle mesh Ewald—an Nlog(N) method for Ewald sums in large systems.
      ). An 8 Å cutoff was applied to Lennard-Jones and electrostatic interactions. Each system was equilibrated for 2 ns with a 2 fs time step at a constant volume and temperature of 300 K. Production trajectories were then run for an additional 200 ns under the same simulation conditions. SASA for fluorine atoms of the TFA label was calculated using the keyword “SURF” in the cpptraj module (
      • Roe D.R.
      • Cheatham III, T.E.
      PTRAJ and CPPTRAJ: software for processing and analysis of molecular dynamics trajectory data.
      ) of AMBER.

       Accelerated MD simulations

      With the accelerated MD implemented in AMBER 18, 500 ns accelerated MD simulations were performed on PimA labeled with TFA and starting from the active-compact conformation. These simulations were restarted from the final structure of the 200 ns conventional MD simulations with random atomic velocity initializations at 300 K. Boost potential with an extra boost to the torsions was applied (iamd = 3).

       Steered-MD simulations

      The protocol employed above for conventional MD simulations was also used in steered-MD calculations (
      • Park S.
      • Khalili-Araghi F.
      • Tajkhorshid E.
      • Schulten K.
      Free energy calculation from steered molecular dynamics simulations using Jarzynski’s equality.
      ). The compact-inactive conformation was used as the starting structure. For the final production trajectory, the distance between C-alpha of Ala148 and Ala366 was varied from 13.5 (close, inactive state, 0 ns) to 51.3 Å (extended, inactive state, 200 ns), using an rk2 of 8 kcal·mol−1·Å−1.

      Data availability

      All data described in the manuscript are contained within the manuscript or the supporting information.

      Acknowledgments

      The Swedish NMR Centre at the University of Gothenburg is acknowledged for support. We thank Axel Abelein for providing a script to analyze relaxation dispersion data.

      Supplementary Material

      References

        • Wolynes P.G.
        Recent successes of the energy landscape theory of protein folding and function.
        Q. Rev. Biophys. 2005; 38 (16934172): 405-410
        • Clore G.M.
        Visualizing lowly-populated regions of the free energy landscape of macromolecular complexes by paramagnetic relaxation enhancement.
        Mol. Biosyst. 2008; 4 (18931781): 1058-1069
        • Sekhar A.
        • Kay L.E.
        NMR paves the way for atomic level descriptions of sparsely populated, transiently formed biomolecular conformers.
        Proc. Natl. Acad. Sci. U S A. 2013; 110 (23868852): 12867-12874
        • Alderson T.R.
        • Kay L.E.
        Unveiling invisible protein states with NMR spectroscopy.
        Curr. Opin. Struct. Biol. 2020; 60 (31835059): 39-49
        • Eisenmesser E.Z.
        • Bosco D.A.
        • Akke M.
        • Kern D.
        Enzyme dynamics during catalysis.
        Science. 2002; 295 (11859194): 1520-1523
        • Olsson M.H.
        • Parson W.W.
        • Warshel A.
        Dynamical contributions to enzyme catalysis: critical tests of a popular hypothesis.
        Chem. Rev. 2006; 106 (16683752): 1737-1756
        • Bryan P.N.
        • Orban J.
        Proteins that switch folds.
        Curr. Opin. Struct. Biol. 2010; 20 (20591649): 482-488
        • Goodchild S.C.
        • Curmi P.M.
        • Brown L.J.
        Structural gymnastics of multifunctional metamorphic proteins.
        Biophys. Rev. 2011; 3 (28510063): 143-153
        • Lella M.
        • Mahalakshmi R.
        Metamorphic proteins: emergence of dual protein folds from one primary sequence.
        Biochemistry. 2017; 56 (28570055): 2971-2984
        • Shrivastava A.N.
        • Aperia A.
        • Melki R.
        • Triller A.
        Physico-pathologic mechanisms involved in neurodegeneration: misfolded protein-plasma membrane interactions.
        Neuron. 2017; 95 (28683268): 33-50
        • Michaels T.C.T.
        • Šarić A.
        • Habchi J.
        • Chia S.
        • Meisl G.
        • Vendruscolo M.
        • Dobson C.M.
        • Knowles T.P.J.
        Chemical kinetics for bridging molecular mechanisms and macroscopic measurements of amyloid fibril formation.
        Annu. Rev. Phys. Chem. 2018; 69 (29490200): 273-298
        • Porter L.L.
        • Looger L.L.
        Extant fold-switching proteins are widespread.
        Proc. Natl. Acad. Sci. U S A. 2018; 115 (29784778): 5968-5973
        • Luo X.
        • Tang Z.
        • Xia G.
        • Wassmann K.
        • Matsumoto T.
        • Rizo J.
        • Yu H.
        The Mad2 spindle checkpoint protein has two distinct natively folded states.
        Nat. Struct. Mol. Biol. 2004; 11 (15024386): 338-345
        • Volkman B.F.
        • Liu T.Y.
        • Peterson F.C.
        Lymphotactin structural dynamics.
        Methods Enzymol. 2009; 461 (19480914): 51-70
        • Chang Y.-G.
        • Cohen S.E.
        • Phong C.
        • Myers W.K.
        • Kim Y.-I.
        • Tseng R.
        • Lin J.
        • Zhang L.
        • Boyd J.S.
        • Lee Y.
        • Kang S.
        • Lee D.
        • Li S.
        • Britt R.D.
        • Rust M.J.
        • et al.
        Circadian rhythms. A protein fold switch joins the circadian oscillator to clock output in cyanobacteria.
        Science. 2015; 349 (26113641): 324-328
        • Korduláková J.
        • Gilleron M.
        • Mikusova K.
        • Puzo G.
        • Brennan P.J.
        • Gicquel B.
        • Jackson M.
        Definition of the first mannosylation step in phosphatidylinositol mannoside synthesis PimA is essential for growth of mycobacteria.
        J. Biol. Chem. 2002; 277 (12068013): 31335-31344
        • Boldrin F.
        • Ventura M.
        • Degiacomi G.
        • Ravishankar S.
        • Sala C.
        • Svetlikova Z.
        • Ambady A.
        • Dhar N.
        • Kordulakova J.
        • Zhang M.
        • Serafini A.
        • Vishwas K.G.
        • Kolly G.S.
        • Kumar N.
        • Palù G.
        • et al.
        The phosphatidyl-myo-inositol mannosyltransferase PimA is essential for Mycobacterium tuberculosis growth in vitro in vivo.
        J. Bacteriol. 2014; 196 (25049093): 3441-3451
        • Guerin M.E.
        • Kaur D.
        • Somashekar B.S.
        • Gibbs S.
        • Gest P.
        • Chatterjee D.
        • Brennan P.
        • Jackson M.
        Insight into the molecular mechanism of the early steps of phosphatidylinositol mannosides biosynthesis in mycobacteria.
        J. Biol. Chem. 2009; 284 (19638342): 25687-25696
        • Guerin M.E.
        • Korduláková J.
        • Alzari P.M.
        • Brennan P.J.
        • Jackson M.
        Molecular basis of phosphatidyl-myo-inositol mannoside biosynthesis and regulation in mycobacteria.
        J. Biol. Chem. 2010; 285 (20801880): 33577-33583
        • Sancho-Vaello E.
        • Albesa-Jové D.
        • Rodrigo-Unzueta A.
        • Guerin M.E.
        Structural basis of phosphatidyl-myo-inositol mannosides biosynthesis in mycobacteria.
        Biochim. Biophys. Acta Mol. Cell Biol. Lipids. 2017; 1862 (27826050): 1355-1367
        • Jankute M.
        • Cox J.A.
        • Harrison J.
        • Besra G.S.
        Assembly of the mycobacterial cell wall.
        Annu. Rev. Microbiol. 2015; 69 (26488279): 405-423
        • Horsburgh Jr, C.R.
        • Barry III, C.E.
        • Lange C.
        Treatment of tuberculosis.
        N. Engl. J. Med. 2015; 373 (26605929): 2149-2160
        • Giganti D.
        • Albesa-Jové D.
        • Urresti S.
        • Rodrigo-Unzueta A.
        • Martínez M.A.
        • Comino N.
        • Barilone N.
        • Bellinzoni M.
        • Chenal A.
        • Guerin M.E.
        • Alzari P.M.
        Secondary structure reshuffling modulates glycosyltransferase function at the membrane.
        Nat. Chem. Biol. 2015; 11 (25402770): 16-18
        • Giganti D.
        • Alegre-Cebollada J.
        • Urresti S.
        • Albesa-Jové D.
        • Rodrigo-Unzueta A.
        • Comino N.
        • Kachala M.
        • López-Fernández S.
        • Svergun D.I.
        • Fernández J.M.
        • Guerin M.E.
        Conformational plasticity of the essential membrane-associated mannosyltransferase PimA from mycobacteria.
        J. Biol. Chem. 2013; 288 (23963451): 29797-29808
        • Guerin M.E.
        • Schaeffer F.
        • Chaffotte A.
        • Gest P.
        • Giganti D.
        • Korduláková J.
        • van der Woerd M.
        • Jackson M.
        • Alzari P.M.
        Substrate-induced conformational changes in the essential peripheral-membrane binding mannosyltransferase PimA from mycobacteria. Implications for catalysis.
        J. Biol. Chem. 2009; 284 (19520856): 21613-21625
        • Albesa-Jové D.
        • Giganti D.
        • Jackson M.
        • Alzari P.M.
        • Guerin M.E.
        Structure-function relationships of membrane-associated GT-B glycosyltransferases.
        Glycobiology. 2014; 24 (24253765): 108-124
        • Albesa-Jové D.
        • Guerin M.E.
        The conformational plasticity of glycosyltransferases.
        Curr. Opin. Struct. Biol. 2016; 40 (27450114): 23-32
        • Wang S.
        • Zhao Y.
        • Yi L.
        • Shen M.
        • Wang C.
        • Zhang X.
        • Yang J.
        • Peng Y.L.
        • Wang D.
        • Liu J.
        Crystal structures of Magnaporthe oryzae trehalose-6-phosphate synthase (MoTps1) suggest a model for catalytic process of Tps1.
        Biochem. J. 2019; 476 (31455720): 3227-3240
        • Gerig J.
        Fluorine NMR of proteins.
        Prog. Nucl. Magn. Res. Sp. 1994; 26: 293-370
        • Kitevski-LeBlanc J.L.
        • Prosser R.S.
        Current applications of 19F NMR to studies of protein structure and dynamics.
        Prog. Nucl. Magn. Res. Sp. 2012; 62 (22364614): 1-33
        • Didenko T.
        • Liu J.J.
        • Horst R.
        • Stevens R.C.
        • Wüthrich K.
        Fluorine-19NMR of integral membrane proteins illustrated with studies of GPCRs.
        Curr. Opin. Struct. Biol. 2013; 23 (23932201): 740-747
        • Manglik A.
        • Kim T.H.
        • Masureel M.
        • Altenbach C.
        • Yang Z.
        • Hilger D.
        • Lerch M.T.
        • Kobilka T.S.
        • Thian F.S.
        • Hubbell W.L.
        • Prosser R.S.
        • Kobilka B.K.
        Structural insights into the dynamic process of β2-adrenergic receptor signaling.
        Cell. 2015; 161 (25981665): 1101-1111
        • Evanics F.
        • Kitevski J.L.
        • Bezsonova I.
        • Forman-Kay J.
        • Prosser R.S.
        19F NMR studies of solvent exposure and peptide binding to an SH3 domain.
        BBA-Gen. Subjects. 2007; 1770 (17182189): 221-230
        • Carver J.
        • Richards R.
        A general two-site solution for the chemical exchange produced dependence of T2 upon the Carr-Purcell pulse separation.
        J. Magn. Reson. 1972; 6: 89-105
        • Spoerner M.
        • Hozsa C.
        • Poetzl J.A.
        • Reiss K.
        • Ganser P.
        • Geyer M.
        • Kalbitzer H.R.
        Conformational states of human rat sarcoma (Ras) protein complexed with its natural ligand GTP and their role for effector interaction and GTP hydrolysis.
        J. Biol. Chem. 2010; 285 (20937837): 39768-39778
        • Waldron T.T.
        • Murphy K.P.
        Stabilization of proteins by ligand binding: application to drug screening and determination of unfolding energetics.
        Biochemistry. 2003; 42 (12718549): 5058-5064
        • Greenfield N.J.
        Using circular dichroism collected as a function of temperature to determine the thermodynamics of protein unfolding and binding interactions.
        Nat. Protoc. 2006; 1 (17406506): 2527-2535
        • Létourneau D.
        • LeHoux J.G.
        • Lavigne P.
        Determination of ligand binding affinity and specificity of purified START domains by thermal shift assays using circular dichroism.
        Methods Mol. Biol. 2019; 1949: 293-306
        • Mayhood T.W.
        • Windsor W.T.
        Ligand binding affinity determined by temperature-dependent circular dichroism: cyclin-dependent kinase 2 inhibitors.
        Anal. Biochem. 2005; 345 (16140252): 187-197
        • Hull W.E.
        • Sykes B.D.
        Fluorine-19 nuclear magnetic resonance study of fluorotyrosine alkaline phosphatase: the influence of zinc on protein structure and a conformational change induced by phosphate binding.
        Biochemistry. 1976; 15 (4091): 1535-1546
        • Guerin M.E.
        • Kordulakova J.
        • Schaeffer F.
        • Svetlikova Z.
        • Buschiazzo A.
        • Giganti D.
        • Gicquel B.
        • Mikusova K.
        • Jackson M.
        • Alzari P.M.
        Molecular recognition and interfacial catalysis by the essential phosphatidylinositol mannosyltransferase PimA from mycobacteria.
        J. Biol. Chem. 2007; 282 (17510062): 20705-20714
        • Park S.
        • Schulten K.
        Calculating potentials of mean force from steered molecular dynamics simulations.
        J. Chem. Phys. 2004; 120 (15267476): 5946-5961
        • Adcock S.A.
        • McCammon J.A.
        Molecular dynamics: survey of methods for simulating the activity of proteins.
        Chem. Rev. 2006; 106 (16683746): 1589-1615
        • Li M.S.
        • Mai B.K.
        Steered molecular dynamics—a promising tool for drug design.
        Curr. Bioinform. 2012; 7: 342-351
        • Lu H.
        • Isralewitz B.
        • Krammer A.
        • Vogel V.
        • Schulten K.
        Unfolding of titin immunoglobulin domains by steered molecular dynamics simulation.
        Biophys. J. 1998; 75 (9675168): 662-671
        • Isralewitz B.
        • Baudry J.
        • Gullingsrud J.
        • Kosztin D.
        • Schulten K.
        Steered molecular dynamics investigations of protein function.
        J. Mol. Graph. Model. 2001; 19: 13-25
        • Hamelberg D.
        • Mongan J.
        • McCammon J.A.
        Accelerated molecular dynamics: a promising and efficient simulation method for biomolecules.
        J. Chem. Phys. 2004; 120 (15268227): 11919-11929
        • Chung K.Y.
        • Kim T.H.
        • Manglik A.
        • Alvares R.
        • Kobilka B.K.
        • Prosser R.S.
        Role of detergents in conformational exchange of a G protein-coupled receptor.
        J. Biol. Chem. 2012; 287 (22893704): 36305-36311
        • Abelein A.
        • Gräslund A.
        • Danielsson J.
        Zinc as chaperone-mimicking agent for retardation of amyloid β peptide fibril formation.
        Proc. Natl. Acad. Sci. U S A. 2015; 112 (25825723): 5407-5412
        • Luz Z.
        • Meiboom S.
        Nuclear magnetic resonance study of the protolysis of trimethylammonium ion in aqueous solution—order of the reaction with respect to solvent.
        J. Chem. Phys. 1963; 39: 366-370
        • Forsén S.
        • Hoffman R.A.
        Study of moderately rapid chemical exchange reactions by means of nuclear magnetic double resonance.
        J. Chem. Phys. 1963; 39: 2892-2901
        • Case D.A.
        • Ben-Shalom I.Y.
        • Brozell S.R.
        • Cerutti D.S.
        • Cheatham III, T.E.
        • Cruzeiro V.W.D.
        • Darden T.A.
        • Duke R.E.
        • Ghoreishi D.
        • Gilson M.K.
        • Gohlke H.
        • Goetz A.W.
        • Greene D.
        • Harris R.
        • Homeyer N.
        • et al.
        AMBER 2018. University of California, San Francisco, CA2018
        • Wang J.
        • Wolf R.M.
        • Caldwell J.W.
        • Kollman P.A.
        • Case D.A.
        Development and testing of a general amber force field.
        J. Comput. Chem. 2004; 25 (15116359): 1157-1174
        • Bayly C.I.
        • Cieplak P.
        • Cornell W.
        • Kollman P.A.
        A well-behaved electrostatic potential based method using charge restraints for deriving atomic charges: the RESP model.
        J. Phys. Chem. 1993; 97: 10269-10280
        • Frisch M.J.
        • Trucks G.W.
        • Schlegel H.B.
        • Scuseria G.E.
        • Robb M.A.
        • Cheeseman J.R.
        • Scalmani G.
        • Barone V.
        • Mennucci B.
        • Petersson G.A.
        • Nakatsuji H.
        • Li X.
        • Caricato M.
        • Marenich A.
        • Bloino J.
        • et al.
        Gaussian 09. Gaussian, Inc., Wallingford, CT2009 (Revision E.01)
        • Jorgensen W.L.
        • Chandrasekhar J.
        • Madura J.D.
        • Impey R.W.
        • Klein M.L.
        Comparison of simple potential functions for simulating liquid water.
        J. Chem. Phys. 1983; 79: 926-935
        • Andrea T.A.
        • Swope W.C.
        • Andersen H.C.
        The role of long ranged forces in determining the structure and properties of liquid water.
        J. Chem. Phys. 1983; 79: 4576-4584
        • Darden T.
        • York D.
        • Pedersen L.
        Particle mesh Ewald—an Nlog(N) method for Ewald sums in large systems.
        J. Chem. Phys. 1993; 98: 10089-10092
        • Roe D.R.
        • Cheatham III, T.E.
        PTRAJ and CPPTRAJ: software for processing and analysis of molecular dynamics trajectory data.
        J. Chem. Theory Comput. 2013; 9 (26583988): 3084-3095
        • Park S.
        • Khalili-Araghi F.
        • Tajkhorshid E.
        • Schulten K.
        Free energy calculation from steered molecular dynamics simulations using Jarzynski’s equality.
        J. Chem. Phys. 2003; 119: 3559-3566