Structural and functional studies of the metalloregulator Fur identify a promoter-binding mechanism and its role in Francisella tularensis virulence

Francisella tularensis is a Gram-negative bacterium causing tularaemia. Classified as possible bioterrorism agent, it may be transmitted to humans via animal infection or inhalation leading to severe pneumonia. Its virulence is related to iron homeostasis involving siderophore biosynthesis directly controlled at the transcription level by the ferric uptake regulator Fur, as presented here together with the first crystal structure of the tetrameric F. tularensis Fur in the presence of its physiological cofactor, Fe2+. Through structural, biophysical, biochemical and modelling studies, we show that promoter sequences of F. tularensis containing Fur boxes enable this tetrameric protein to bind them by splitting it into two dimers. Furthermore, the critical role of F. tularensis Fur in virulence and pathogenesis is demonstrated with a fur-deleted mutant showing an attenuated virulence in macrophage-like cells and mice. Together, our study suggests that Fur is an attractive target of new antibiotics that attenuate the virulence of F. tularensis.

Panel B. SEC-MALLS analysis experiments are done in the presence of 500 mM NaCl. FtFur-WT presents a MW of 64 kDa corresponding to a tetramer in solution. E63A mutant shows a non-symmetric peak at 48 kDa MW corresponding to a mixture of tetramer and dimer. E76A present two peaks. The major peak corresponds to a MW of 32 kDa for a dimer in solution, the minor peak presents a MW of 64kDa corresponding to a tetramer. Finally, the double mutant E63AE76A presents one single symmetric peak at 32 kDa corresponding to a stable dimer in solution. This experiment has been repeated three times and validates the major role of these two amino acids in the stabilization of the tetramer.
Panel C. Nuclease assay on FtFur, FtFurΔS2, FtFurE63A, FtFurE76A, FtFurE63AE76A performed as previously described 4 . The FtFurΔS2 and the FtFurE63AE76A mutants were not able to bind DNA contrarily to the WT, FtFurE63A and FtFurE76A mutants, in the conditions of the assay. Figure 6. Algorithm for the preparation of initial coordinates for the computation of free energy profiles.

Supplementary
The algorithm consists of a main subprogram doing the translations and pre-equilibrations, which is called twice: 1) 10 first windows corresponding to 0.5 Å translations followed by 2) 15 windows of 1 Å translations. Three variables are defined "dtot", the total translation distance for the "moving" subsystem, "i", the conformation counter and "dist", the current translation distance in Å. Figure 7. Structure of the FtFur tetramer corresponding to the last conformation in the pulling process.

Supplementary
The moving subsystem (Chains A and D) has been translated (10*0.5+15*1.0) =20 Å from the initial tetramer along the X direction. Centers of mass of the "fixed" subsystem and the two subunits of the moving Ft dimer are indicated. The moving subsystem is constituted of two Ft subunits: 1 and 2. Forces are applied on their centers of mass to pull them away or maintain them at a fixed distance of the "fixed" subsystem. Part of the backbone of the "fixed" system highlighted in red is subject to position restraints during the pulling process. Black dashed lines show hydrogen bonds in the secondary structures which are maintained using NOE type distance restraints. Panel A. The five major contributors to the interaction energy between chain A and the "fixed" dimer in the FtFur tetramer simulation. Same legend as in Fig. 2. On average, 5 residues contribute to around 30% of the total interaction energy (with a maximum of 56% at 2.93 nm). The maximum total interaction energy for all residues over all 26 windows is calculated first, then the contribution of each residue is normalized with respect to this maximum. The plot shows the corresponding percentage of the maximum interaction energy for selected residues. The blue surface corresponds to the sum of the contributions of all residues others than those selected. Rmeas, redundancy-independent merging R-factor c Structure of Mn-FtFur has been solved by the MAD method using 1 crystal, and refined with the remote diffraction data only

Molecular Modelling of the FtFur tetramer
The pdb file corresponding to the Xray structure of the FtFur tetramer was prepared for gromacs (GROningen MAchine for Chemical Simulations) version 5.1.2 9 with pdb2gmx using gromos54a7 united atom force field 10 . All histidine residues were given type HISA with protonated Nδ1 atom. In all 4 protein subunits cysteines 93, 96, 133 and 136 were deprotonated and given a total charge of -0.75 (-0.05 for Cα, -0.15 for Cβ, -0.55 for S). This charge can be compared to that of the CYS residue in gromos54a7 force field, with total charge of -0.5. A value of -0.75 was adopted instead in this work, high enough to stabilize a tetracoordinated Zinc ion, providing correct orientation of the cysteines and proper metal environment during further simulations. (Use of a -0.5 charge led to the escape of the metal from its binding site).

Construction of the MgFur dimer DNA complex
The structure of Magnetospirillum gryphiswaldense (4RB1) was used as initial model of a Fur per DNA complex. The Fur dimer includes residues 20 to 133 and the DNA double strand is composed of CGGATAATGATAATCATTATCGC. First, an initial model (Mod1) of the MgFur dimer per DNA complex with 4 iron ions and 2 bound water molecules (included in structure 4RB1) was built and slightly energy minimized with CHARMM 11 using the all-atom "par_all27_prot_na" parameter file for proteins and nucleic acids. Second, another initial model (Mod2) of the protein dimer alone was built starting from the same PDB but with the extended atom "param19" force field for proteins. Then, an initial structure of the protein DNA complex compatible with the Gromacs force field was created from protein dimer from Mod2 and DNA from Mod1. The structure was oriented so that DNA was centered at the origin with its principal axis along Y. We noticed that the MgFur dimer in this structure sits asymmetrically on the DNA fragment with K15 in chain A interacting with CYT21 and K15 in chain B extending over CYT1. It was consequently decided to move the last two GC bases upstream leading to a GCCGGATAATGATAATCATTATC fragment. The new position of the 2 moved nucleotides was controlled with VMD and regularized through 1000 steps of energy minimization with CHARMM. This new DNA fragment offers the advantage of providing good interactions for all residues in the Fur dimer while still including the full length consensus Fur box GATAATGATAATCATTATC.

Construction of the FtFur dimer per wild type DNA complex
Structure of FtFur in the presence of iron (this work) from ASP7 to ARG137 was used for the Fur/wtDNA simulation. In the FtFur+DNA structure, the model built from MgFur was used as template to align FtFur. FtFur was superimposed on MgFur already sitting on the Fur box using least square fit of matching atom positions. Backbone atoms from corresponding structured parts (helices and sheets) of the two protein were selected for this structure superposition for a final rms difference of 4 Å.

Construction of the FtFur dimer per mutated DNA complex
The mutated DNA (mutDNA) sequence GCCGGATACTGATAGTCCTGATC contains four mutations with respect to the Fur-box (A9 to C, A15 to G, A18 to C and T20 to G). Initial coordinates for mutDNA were taken from the previously obtained structure of wild type DNA by simple matching of corresponding heavy atoms and building of missing hydrogens with CHARMM, followed by 1000 steps energy minimization. Initial direct superimposition of the Fur dimer onto this mutated DNA showed that relatively large structure modifications could occur in the region of the mutations. Consequently, a 100 ps NPT simulation of solvated mutDNA was run to release the strains in the structure. This last structure was superimposed on the initial model built in the presence of FtFur to ensure optimal orientation. Finally, initial coordinates for the FtFur dimer per mutated DNA complex (FtFur/mutDNA) were obtained by concatenating coordinates of the protein in the FtFur/wtDNA complex and those of the reoriented mutDNA.

Computational details of the solvation and equilibration
For the tetramer, the maximum dimensions of the protein were calculated, 28 Å were added in the +X direction to allow for its further translation. Then, a simulation box 5 Å larger than these dimensions in all directions was created and the tetramer centered in the simulation box with "editconf". Initial dimensions of the box were 10.53*8.03*6.54 nm. "gmx solvate" was used to add solvent (SPC water) to the protein from a preequilibrated water box (spc216.gro) for a total of 15586 water molecules. 37 sodium and 33 chloride ions were added (replacing 70 water molecules) for a total ionic force of 0.1 mol.L -1 and a zero total charge of the 52114 atom system. For the FtFur/wtDNA complex, initial dimensions of the box were 9.28*8.91*6.35 nm. 15505 water molecules were added of which 77 were replaced by sodium and 32 by chloride ions to ensure neutrality and a total ionic force of 0.1 mol.L -1 for a total of 50117 atoms in the system. For the FtFur/mutDNA complex, initial dimensions of the box were 9.64*9.22*6.39 nm. 16680 water molecules were added of which 79 were replaced by sodium and 34 by chloride ions to ensure neutrality and a total ionic force of 0.1 mol.L -1 for a total of 53642 atoms in the system.
All three systems (FtFurTetra, FtFur/wtDNA, FtFur/mutDNA) were energy minimized until the force on all atoms was less than 1000 kJ.mol -1 .Å -1 . Heavy atoms of the protein were restrained to their initial position to prevent big structural changes before the production run with a force constant of 10 kJ.mol -1 .Å -2 . Then we ran the system under NPT conditions at 310 K and 1 atm for 100 ps, with Berendsen temperature and isotropic pressure coupling 12 with τT = 0.1 ps, τP = 2 ps and compressibility = 4.5 × 10 -5 bar -1 (same conditions in the NPT simulation used to equilibrate mutDNA).

Computation of Free energy profiles
Free energy profiles for the extraction (by translation along a fixed direction: Ox) of one FtFur dimer from the tetramer (dimer of dimers) and of FtFur from DNA were computed. The FtFur tetramer is organized around 4 subunits with chain names B and C for the fixed dimer and A and D for the translated dimer. The simulations will thus include a "moving" subsystem (FtFur dimer, chains A and D) and a "fixed" subsystem (FtFur dimer, chains B and C, wtDNA or mutDNA). The profiles were built using the "umbrella sampling" technique and result from the overlapping of 26 computation windows, one for each translation distance. The simulation protocol for the equilibration of umbrella sampling windows is best described in supplementary Figure 6.
As a result of this initialization protocol, 26 structures "conf_i" were generated yielding initial positions for the translated dimer prior to the potential of mean force calculation. Then the umbrella sampling calculation, itself, consisted of 26 repeats of: -Reading of initial structure "conf_i" and reference structure "conf_0" for harmonic restraints; -Running 100 ps NPT equilibration with position restraints on the "fixed" subsystem and distance restraints on the protein. The "moving" subsystem was subject to two harmonic biasing forces (umbrella potential) applied between the centers of mass of the 2 Fur dimer subunits and the center of mass of the "fixed" subsystem with force constants of 500 kJ mol -1 nm -2 .
-Running 10 to 15 ns NPT production simulations with same restraints and biasing potential. These simulations were concatenated from series of 1 to 2 ns runs allowing us to check the convergence of the PMFs. Harmonic biasing forces were applied along the X direction only (direction of the translation).
Position restraints were applied on backbone atoms of the "fixed" subsystem for residues K28 to S35 (Loop between H1 and H2) and for all residues N83 to E138 (C-Terminal) except V94 to M98 possibly interacting with the "moving" subsystem during the translation. Force constants of 10 kJ mol -1 Å 2 in all directions were used. NOE type distance restraints were added to maintain the secondary structure of the protein (both "fixed" and "moving" parts in the case of the tetramer). All backbone hydrogen bonds between N-H and O=C for residues T16 to K24 (H1), S35 to K44 (H2) and G51 to E63 (H3) plus the short beta-sheet (I67-N83, N69-E81, L71-M79) were restrained between 1.8 and 2.0 Å with force constants of 20 kJ mol -1 Å 2 .
In both equilibration and production umbrella sampling MD simulations, the system was simulated under NPT conditions. Temperature was fixed at 310 K with Temperature coupling using a Nose-Hoover extended ensemble with τT = 0.5 ps, Pressure was controlled at 1 atm with extended-ensemble Parrinello-Rahman isotropic pressure coupling with τP =1 ps and compressibility = 4.5 × 10 -5 bar -1 13 . Again, a time step of 2 fs was used.

Fitting of the free energy profiles
Data points corresponding to the outputs of wham were fitted with a sum of 1, 2 or 3 sigmoid functions with R 14 : S(x) = + b 1 + (− * ( − )) After a first fit with the raw data, the energy offset (a in equation) was subtracted from the final energy profiles.