A Catenane Assembled through a Single Charge-Assisted Halogen Bond

Interlocked molecules have captured chemists’ imagination owing to their nontrivial topology and the promise of their potential nanotechnological uses as molecular machines[1] and in chemical sensor applications.[2] The synthesis of such sophisticated architectures is a challenge, and as a consequence this has necessitated the implementation of imaginative cation,[3] anion,[4] and neutral[5] templation methodologies, which use a combination of complementary Lewis acid–base, electrostatic, and hydrogen-bonding interactions for component assembly. Halogen bonding (XB) is the attractive highly directional, noncovalent interaction between an electron-deficient halogen atom and a Lewis base.[6] The scope of XB in solid-state crystal engineering has been intensively explored for a number of years,[7] however, in spite of the complementary analogy to ubiquitous hydrogen bonding, it is only in recent times that investigations into the application of solution-phase XB interactions to molecular recognition processes, self-assembly, and catalysis have resulted in this field developing rapidly.[8] Indeed, in conjunction with anion templation, we have used XB to assemble interpenetrated and interlocked molecular frameworks.[8b–d]


Pyridine macrocycle precursor 12:
Compound 11 (2.01 g, 3.5 mmol) was dissolved in MeOH (200 mL), 600 µL of conc. HCl (aq) was added and the reaction mixture stirred at r.t. for 6 hr. The solution was neutralized with NaHCO 3(s) and solvent was removed in vacuo. The solid was washed with Et 2 O (10 x 20 mL) to yield the desired product as a light yellow oil (1.19 g, 2.9 mmol, 78%). 1  an additional 24 hr. After cooling to r.t., the reaction mixture was filtered and the solvent removed in vacuo. The residue was redissolved in hot CHCl 3 (40 mL), filtered and the solvent removed in vacuo. The crude brown oil was purified by column chromatography (SiO 2 , EtOAc:cyclohexane 1:0 to 0:1) to yield the desired product 1 as a white solid.
All volatile components were removed in vacuo to give a yellow oil, which was purified by column chromatography (SiO 2 , 98:2 CH 2 Cl 2 :MeOH) to give compound 17 as a white solid
After cooling to r.t. acetone (20 mL) was added and the solvent removed in vacuo.

4-Iodopyridinium tetrafluoroborate thread, 6·BF 4 .
Compound 33 (28 mg, 37.9 μmol) was dissolved in anhydrous CH 2 Cl 2 (10 mL). Me 3 OBF 4 (7.60 mg, 51.4 μmol) was added and the reaction mixture was stirred at r.  reported by us in references [8] and [9] . The specific force field parameters used to describe these interactions were developed as follows. DFT calculations were performed with Gaussian 09 [10] using the hybrid functional B3LYP [11] on systems A1, and B1 (see Figure S4), models for the I•••N interaction in 7 + . System B1 contains two meta-methoxy substituents as mimetic groups of the ether linkages of the two interlocked macrocycles on the [2]catenane assembly. Figure S4: Optimised B3LYP gas phase structures of A1 and B1 models used in the parameterization of C-I•••N charged assisted interactions.

S25
For C, O, H, and N the standard 6-311+G** basis set was used while I was described by the aug-cc-pVDZ-PP basis set and pseudo potential taken from the EMSL Basis Set Exchange website. [12] The geometries were optimized in the gas phase and, for the larger model B1, solvent effects (CHCl 3 ) were also considered by means of IEFPCM calculation with radii and non-electrostatic terms for Truhlar and co-workers' SMD solvation model. [13] The relevant structural parameters are collect in Table S1. Taking this in consideration, we proceeded with the parameterization of the possible C-I•••N halogen bond interactions present in 7·BF 4 . Parameterizations were performed with the addition of a specific term to the General Amber Force Field (GAFF) [14] parameters used to describe the macrocycle components of the [2]catenane interlocked assembly as suggested in references [15] and [16] . The positive region on the electrostatic potential of the iodopyridinium moiety centred on the iodine atom is represented by a pseudo-atom (DU) with a van de Waals parameter set to zero, a C-I-DU bond angle set to 180º, and a bending force constant of 150 kcal mol −1 Å −2 . The optimal I-DU equilibrium distance was evaluated using several I-DU distances and computing the electrostatic potential and the Restricted Electrostatic Potential (RESP) charges for each new DU position. This was followed by a Molecular Mechanics (MM) minimization of the corresponding assembled catenane structures in Amber 12 [17] .

S26
Subsequently, the MM optimised distances (I•••N) were compared with the DFT data ( Table   2). The electrostatic potential for the two individual macrocycles was calculated at the HF/6-31G* level, with the aug-cc-pVDZ-PP basis set for iodine using previously optimized  Table S2 along with the relevant RESP charges.  Moreover, preliminary gas phase MM calculations on the iodopyridinium macrocycle showed that default potential energy barrier (V n ) of 0.3 kcal for the X-ca-na-X torsion angle (where ca and na are the atom types assigned to the pyridinium ring carbon atoms and nitrogen atom, respectively, and X means any other atom type) led to a slight bending of the N-methyl group relatively to pyridinium ring plane. This structural imprecision was corrected using the torsion force field parameters available in the GAFF for the related X-cd-na-X torsion angle (4 / 6.8 / 180.0 / 2), where cd is a sp 2 carbon in non-pure aromatic systems.

Generation of starting structures for the MD simulations.
Since there is no X-ray structure for catenane 7·BF 4 and given its conformational freedom, the starting structures for the MD simulations were generated by a protocol based on a gas-phase quenched dynamics simulation. [19] The previous MM optimised structure with I-DU = 2.10 Å was subjected to a 5 ns MD run at 500 K saving a trajectory with 10000 structures that should sample most of the conformational space. All these structures were minimized by MM and clustered by RMSD similarity. Two structures corresponding to representative conformations of the most populated clusters were selected. These structures shown in Figure S6 as S1A and S1B present an C-I•••N halogen bond and - donor-acceptor interactions between the S28 iodopyridinium and the hydroquinone rings of the macrocycle 1. These two co-conformations basically differ in the conformation adopted by the iodopyridinium macrocycle leading to slightly different binding scenarios. S1A S1B S2A S2B Figure S6: Starting co-conformations used in the MD simulations, S1A and S1B exhibit an C-I•••N halogen bond and S2A and S2B present an C-I•••O halogen bond.
In order to increase the sampling, the iodopyridinium macrocycle in a structure similar to S1A was rotated affording an interlocked assembly with the C-I bond pointing away from the N of the pyridine ring. This struture was subjected to the same protocol mentioned earlier yielding two starting co-conformations S2A and S2B presented in Figure S6, where the C-I bond is pointing towards an oxygen atom from the polyether loop of the macrocycle 1.

Molecular Dynamics Simulations. The MM minimized conformations shown in Figures S6
were immersed in cubic box of CHCl 3 molecules [20] (~ 43 Å side, after equilibration). One BF 4 anion, with charges and parameters from reference [21] , was used as counter-ion. All systems were simulated using the following multi-stage protocol. The solvent was initially relaxed keeping the solute fixed with a strong positional restraint (500 kcal mol -1 Å -2 ). Then, the restraint was removed and a MM energy minimisation of entire system was performed followed by a heating stage (0 to 300 K) during 50 ps using the Langevin thermostat with a collision frequency of 1 ps -1 in an NVT ensemble and a weak positional restraint on the solute (10 kcal mol -1 Å -2 ). The system was then equilibrated for 5 ns in a NPT ensemble at 1 atm with isotropic pressure scaling using relaxation time of 2 ps and data collection was performed for 50 ns. All simulations were carried out with the pmemd.cuda AMBER executable, able to accelerate explicit solvent Particle Mesh Ewald (PME) calculations [22] trough the use of GPUs. [23] All bonds involving hydrogen atoms were constrained using SHAKE [24] allowing the usage of 2 fs time step. A 10 Å cutoff was used for the non-bonded van der Waals interactions. Trajectory analysis was performed with the cpptraj utility of AmberTools12 while the molecular representations were drawn with PyMOL. [25] Plots were performed with Gnuplot. [26] Representative co-conformations from the MD trajectories in solution were obtained by clustering the structures by RMSD similarity, using the averagelinkage clustering algorithm. [27] A frame from the most populated cluster was chosen as representative of the simulation.

S30
MD simulations extended analysis. Figure S7: Histograms for the distribution of the I···N distances (Å) and the C-I···N angles (º) for simulations with S1A and S1B. Data were collected during 50 ns. Figure S8: Representative co-conformations of 7·BF 4 for simulations with S1A (left) and S1B (right).
Extended discussion on simulations with S2A and S2B. In the S2 scenarios (A and B), an C-I•••O halogen bond is possible between the iodopyridinium and the three available oxygen atoms of the polyether loop numbered O1, O2, and O3 (see Figure S9). In these simulations S32 Figure S9: Histograms for the distribution of the I···O distances (Å) and the C-I···O angles (º) for simulations with S2A and S2B. Data were collected along the 50 ns. The numbering scheme used for oxygen atoms is shown at the top.
Part V: References.