Tilting and Tumbling in Transmembrane Anion Carriers: Activity Tuning through n‐Alkyl Substitution

Abstract Anion transport by synthetic carriers (anionophores) holds promise for medical applications, especially the treatment of cystic fibrosis. Among the factors which determine carrier activity, the size and disposition of alkyl groups is proving remarkably important. Herein we describe a series of dithioureidodecalin anionophores, in which alkyl substituents on one face are varied from C0 to C10 in two‐carbon steps. Activities increase then decrease as the chain length grows, peaking quite sharply at C6. Molecular dynamics simulations showed the transporter chloride complexes releasing chloride as they approach the membrane‐aqueous interface. The free transporter then stays at the interface, adopting an orientation that depends on the alkyl substituent. If chloride release is prevented, the complex is positioned similarly. Longer chains tilt the binding site away from the interface, potentially freeing the transporter or complex to move through the membrane. However, chains which are too long can also slow transport by inhibiting movement, and especially reorientation, within the phospholipid bilayer.


Synthesis and Characterisation
The six new decalin transporters described in this communication were synthesised as detailed below.
The Boc-protected diamine 6 was prepared as previously described. 1 The required isothiocyanates were obtained commercially, with exception of 4-hexylphenylisothiocyanate 8d and 4octylphenylisothiocyanate 8e which are not commercially available, and were prepared as described below. 2

General methods
All reagents were purchased from commercial suppliers and used without further purification, unless otherwise stated. Anhydrous DCM and THF were dried by passing through a modified Grubbs system, with an alumina column manufactured by Anhydrous Engineering. Flash column chromatography was performed using silica gel (Aldrich, pore size 60 Å, particle size 40-63 µm) as the absorbent. Routine monitoring of reactions was performed using precoated silica gel TLC plates (Merck silica gel 60 F254).
Spots were visualised under UV light or by staining with phosphomolybdic acid, potassium permanganate or ninhydrin; Rf values are given under these conditions. 1 H, 13 C and 19 F NMR spectra were recorded using a ECS 300, ECS 400, Varian 400, Varian 500A (carbon sensitive) or Varian 500B (proton sensitive) spectrometer. All spectra are recorded at 298 K unless otherwise stated. Chemical shifts (δ) are quoted in parts per million (ppm), coupling constants (J) are quoted in Hz and spectra are referenced to the appropriate residual solvent peak. Mass spectra were recorded on a Bruker microTOF II (ESI), VG Analytical Quattro (ESI) or VG Analytical Autospec (EI). IR spectra were recorded on a Perkin-Elmer Spectrum 100 FT-IR spectrometer. Elemental analysis was carried out by the microanalysis department at the School of Chemistry, University of Bristol. The carbon numbering system for the decalin-based transporters 7a-f is as shown below for 7f:

Synthesis of isothiocyanates 8d and 8e 4-Hexylphenyl isothiocyanate (8d)
To a solution of 4-hexylaniline (1.0 mL, 5.1 mmol) in absolute EtOH (5 mL) was added carbon disulfide (3.0 mL, 50 mmol) and triethylamine (0.7 mL, 5 mmol). The mixture was stirred at room temperature for 1 h and then cooled to 0 °C using an ice-bath. A solution of di-t-butyl dicarbonate (1.10 g, 5.04 mmol) in absolute EtOH (1 mL) was added to the reaction mixture, followed by a solution of DMAP (12 mg, 0.10 mmol) in absolute EtOH (1 mL), and the mixture was stirred at 0 °C for 5 minutes. The mixture was allowed to warm to room temperature and was stirred for 30 minutes after which the solvent was removed in vacuo yielding 4-hexylphenyl isothiocyanate 8d (1.15 g) as a crude brown residue which was used without further purification: Rf 0.8 (50% EtOAc in hexane); 1

4-Octylphenyl isothiocyanate (8e)
To a solution of 4-octylaniline (1.0 mL, 4.4 mmol) in absolute EtOH (5 mL) was added carbon disulfide (2.7 mL, 45 mmol) and triethylamine (0.6 mL, 4 mmol). The mixture was stirred at room temperature for 1 h and then cooled to 0 °C using an ice-bath. A solution of di-t-butyl dicarbonate (0.94 g, 4.3 mmol) in absolute EtOH (1 mL) was added to the reaction mixture, followed by a solution of DMAP (10 mg, 0.082 mmol) in absolute EtOH (1 mL), and the mixture was stirred at 0 ºC for 5 minutes. The mixture was allowed to warm to room temperature and was stirred for 30 minutes after which the solvent was removed in vacuo yielding 4-octylphenyl isothiocyanate 8e (1.13 g) as a crude brown residue, which was used without further purification: Rf 0.9 (50% EtOAc in hexane); 1

General Procedure I for synthesis of decalin transporters 7a-f
To a solution of Boc-protected diamine 6 (50 mg, 0.11 mmol) in anhydrous DCM (4 mL) was added TFA (1 mL) dropwise over 5 min and the solution was stirred at room temperature, under N2, for 16 h.
The solvent was removed in vacuo and excess TFA was removed by co-evaporation with toluene and

Chloride binding affinities by 1 H NMR titration in DMSO-d6
The binding affinities of receptors 7 for chloride were measured by 1 H NMR titrations against n-Bu4N + Clin DMSO-d6 containing 0.5% H2O. The hygroscopic guest n-Bu4N + Cland the host compounds were dried under high vacuum to remove residual solvents or water prior to solution preparation.
The concentration of the host was around 1 mM in all titrations; the concentration of guest and the volumes of the aliquots of guest solution added to the host solution were varied over the experiments.
The guest solution contained the same concentrations of host as the starting host solution (i.e., n-Bu4N + Clwas dissolved in a solution of ~1 mM host in DMSO-d6 with 0.5% H2O), so that the concentration of host did not decrease over the course of the experiment.
All 1 H NMR titration binding studies were performed using a Varian 500 MHz NMR spectrometer (proton sensitive probe) at 298 K. Spectra were calibrated to the residual solvent peak in DMSO-d6 (δ = 2.50 ppm). Binding constants were determined by fitting the shift of both NH signals to a 1:1 binding model using a least-square fitting procedure in a custom-made Excel spreadsheet.        The value for R obtained from the 1 H NMR spectra was used to calculate the equilibrium constant Ke: We have to take into account the equilibrium between the unbound Et4N + Xguest in the organic phase and the dissociated Et4N + and Xions in the aqueous phase, described by: The Kd value for Et4N + Clhas been determined and reported previously to be 1.269 × 10 -5 M -1 . 3

S24
The binding constant Ka is then calculated from Ke and Kd : Table S1. Extraction data and derived association constants of decalins 7a-f to Et4N + Clin CHCl3.

General experimental description of transport measurements
Chloride ion transport was measured using large unilamellar vesicles (LUVs, 200 nm average diameter) composed of 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) and cholesterol at a ratio of 7:3. POPC was obtained from Avanti® Polar Lipids, Inc. Extrusion apparatus and 200 nm polycarbonate membranes were obtained from GC Technology Ltd. The lipid solutions were prepared using chloroform that had been pH neutralised by passing the chloroform through a flash chromatography column containing basic alumina, and all aqueous solutions were prepared using deionised water that had been passed through a Millipore filtration system.
The following is a typical example of an anion transport experiment: The first 30 seconds and the initial drop (due to quenching of external lucigenin) were removed, the traces were normalised and the resulting F/F0 traces were averaged. The traces thus obtained from the transport measurements at transporter:lipid ratios of 1:1000 are given in Figure 1a of the main text and traces at 1:2500 are shown in Figure S25. Normalised Fluorescence,

Fitting of transport data
Half-lives of fluorescence decay were obtained by fitting the reciprocal transport curves (F0/F) from 0-500 seconds to a single exponential decay function (Eq. 1) using Origin 9.0. The half-life t½ was calculated using fit parameter 'b' (Eq. 2).
(1) All parameters obtained from fitting the transport data are summarised in Table S2. Table S2. Half-lives, initial rates, and specific initial rates of transport for decalin compounds 7a-f.

Test for leaching of decalin 7a from vesicle membranes
Decalin 7a was pre-incorporated into vesicles (7 µmol cholesterol, 3  If the transporter were capable of leaching from the membrane into the aqueous phase, both dilution with water and the additional passage through Sephadex would be expected to lower the concentration in the membrane (in the former case by increasing the volume of the aqueous reservoir, in the latter by removing transporter dissolved in water). Neither treatment caused a significant change in the rate of chloride transport. Therefore, we conclude that 7a does not leach from the membrane into the aqueous phase and that the lower rates of transport by 7a with respect to 7d are not caused by a lower amount of transporter present in the membranes during the measurements. 6

Computational Methods
The MD simulations were carried out with the AMBER 14 software suite, 7 using the LIPID14 force field 8 for the POPC lipids and the GAFF 9 parameters and atomic RESP charges 10 for transporters 7a-f as follows.

a) Calculation of atomic RESP charges for transporters 7a-7f
The initial structure of 7a was generated from a crystal structure of a decalin derivative featuring a carboxylate group at C4a position. 11 Two phenyl thiourea substituents were appended at both C2 and C7 decalin skeleton positions. The remaining five molecules of the diaxial thiourea series were generated from 7a by addition of the appropriate alkyl chain at the para position of both phenyl rings.
Subsequently, all structures were optimized at the HF/6-31G* level with the Gaussian 09 software. 12 The optimization of each single structure of 7a-7f was followed by a single point calculation to generate the electrostatic potential (ESP) at the same theory level, using the Merz-Singh-Kollman scheme with 4 concentric layers per atom and 6 density points in each layer (IOp(6/33=2, 6/41=4, 6/42=6)). The initial atomic charges of each molecule were then calculated by RESP fitting, along with the attribution of GAFF atom types, using the antechamber module, as implemented in the AMBER software suite. 7a To obtain atomic charges less dependent of the molecular conformation or orientation, the calculation of the final RESP charges employed in the MD simulations reported was preceded by conformational analyses on receptors 7a-f. The initial molecular mechanics (MM) energy minimized structures, using the initial RESP charges, were heated in the gas phase at 500 K for 50 ps, followed by a collection run of 5 ns, using a time step of 1 fs. The use of this high temperature allows the stochastic searching of conformational space, since the energetic barriers are easily surmounted. 13 Frames were saved every 0.1 ps leading to a trajectory file containing 50000 structures for each molecule. All these structures were further minimized by MM using a steepest descendent gradient followed by the conjugate gradient algorithm, until the convergence criterion of 0.0001 kcal mol -1 Å -1 was attained. Afterwards, the MM minimised structures were clustered with the UCSF Chimera software. 14 Five representative conformations with substantially different RMSD values with both thiourea binding units adopting a syn configuration but displaying a different spatial disposition were selected and underwent new HF/6-31G* geometry optimizations and ESP calculations as described above. The individual ESP data were extracted from the corresponding Gaussian 09 12 output and then the ESP data of the five conformations were concatenated and subsequently used to generate the input files for the two-stage RESP fitting, using identical weights for all conformations.

b) Generation of chloride complexes
The structures of 7a-7f chloride complexes were obtained in gas phase via conformational analyses as described above for the free molecules. However, the production runs were carried out only for 1 ns, leading at the end the quenched dynamics protocol to 10000 MM minimized structures. The most representative structure of each complex, obtained by cluster analysis with UCSF Chimera software, 8 ,was selected for the subsequent MD simulations ( Figure S27).

c) Simulations in a POPC bilayer i. Simulation of the initial POPC bilayer
Before the simulation of the chloride complexes inserted into a POPC membrane model, a membrane system with a orthorhombic shape (dubbed free membrane) composed of 128 phospholipids, 6500 TIP3P water molecules, 18 Cland 18 Na + ions (0.15 M) was built with a water per lipid ratio of 50.1.
This free membrane was simulated for 150 ns, under periodic boundary conditions at 303 K with force field parameters taken from LIPID14 and the ions described with van der Waals parameters developed to be used along with the TIP3P water model. 15 The simulation protocol followed is equivalent to the one detailed below for the membrane systems with chloride complexes. The structural parameters area per lipid, bilayer thickness, order parameters and electron density profiles estimated for the last 70 ns of sampling, are similar to those originally reported for the POPC bilayer system containing a water per lipid ratio of 31.6. 16 Therefore, increasing the number water molecules had no impact in the bilayer system, and this equilibrated free membrane was subsequently used in the MD simulations carried out with the chloride complexes either inserted into the core of the bilayer (setup Aan approximation to the experimental chloride transmembrane transport studies) or immersed in the water slab (setup B).
This initial positioning of the chloride complexes of 7a-7f within the free membrane was performed with Packmol, 17 and yielded systems A.7a to A.7f, as well as systems B.7a to B.7f, as given in Table   S3.

ii. Simulation of membrane systems with chloride complexes at the bilayer coresetup A
The MD simulations A1.7a to A4.7f were carried out as follows: the initial configuration of each system was submitted to 10000 steps of MM energy minimization with a 500 kcal/mol Å 2 positional restraint on the chloride complex and lipid molecules, through the steepest descent algorithm for 3000 steps plus 7000 steps of the conjugated gradient algorithm, followed by the relaxation of the entire system for another 10000 steps, with the same protocol. The equilibration of the system proceeded by heating it to 303 K in an NVT ensemble for 100 ps with a 10 kcal/mol Å 2 restraint on the chloride complex and lipid molecules. The equilibration stage proceeded with a 5 ns run using an NPT ensemble with a 5 kcal/mol Å 2 restraint on the chloride complex. Then, the positional restraint was removed and the simulation continued for further 150 ns. The long-range electrostatic interactions were described with the Particle Mesh Ewald (PME) algorithm 18 using a real-space cut-off at 10 Å. The cut-off for the Lennard-Jones interactions was also set at 10 Å. The temperature of the system was maintained independently by coupling the system to an external bath temperature of 303 K, using the Langevin thermostat, 19  atoms were constrained using the SHAKE algorithm, 21 allowing the use of a 2 fs time step. This protocol was employed to run four independent replicates of 150 ns of setup A membrane systems (Table S3), with the frames saved every 10.0 ps.

iii. Simulation of membrane systems with chloride complexes at the bilayer coresetup A'
From the end of the equilibration period of simulations A1.7a-A1.7f, a single production run of 150 ns was undertaken with each chloride complex maintained through the use of four distance restraints between the nitrogen atoms of each thiourea binding units and the chloride. Harmonic restraints of 5 kcal/mol were applied to obtain N···Cldistances of 3.5 Å.

iv. Simulation of membrane systems with chloride complexes in the water slabsetup B
The MD simulations B1.7a-B2.7f, listed in Table S3, were carried out using the aforementioned protocol given for simulations of setup A, except for the length of the production runs, which were 200 ns long.

Structural impact of 7a-f on the bilayer model
The structural impact of 7a-f in the membrane model was examined through evaluation of the area per lipid, bilayer thickness, electron density profiles and order parameters |SCD| during the last 50 ns of each MD replicate of setup A with each transporter (see Table S3). The average values for area per lipid and bilayer thickness are gathered in Table S4 and the electron density profiles and order parameters are plotted in Figures Figure S28 to Figure S39. The area per lipid and bilayer thickness average values are quite similar to the ones reported for the membrane system without any transporter. The small variations observed in all these structural parameters can be attributed to the necessary adjustments for the nesting of the transporters at the interface level. Table S4. Comparison between the area per lipid and bilayer thickness of simulations A1.7a-A4.7f (50 ns of sampling) to the free membrane system (70 ns of sampling  Figure S28. Electron density profiles of simulations A1.7a-A4.7a with the full system plotted in black, water in blue, phospholipids in wheat, phosphorus atoms in orange, chloride ions in green, sodium atoms in pink and 7a in purple. The ions and the transporter are scaled 5 times. The z = 0 Å corresponds to the core of the POPC bilayer. The free membrane profile is also shown as a red line. Figure S29. Electron density profiles of simulations A1.7b-A4.7b with the full system plotted in black, water in blue, phospholipids in wheat, phosphorus atoms in orange, chloride ions in green, sodium atoms in pink and 7b in purple. The ions and the transporter are scaled 5 times. The z = 0 Å corresponds to the core of the POPC bilayer. The free membrane profile is also shown as a red line.

Interaction energies between transporters 7a-f and POPC bilayer
The interaction energies between the POPC membrane model and the thiourea molecules were  Table S8 for each system and in Table S9 for each replicate, and visualized in Figure   S49.   The ΔEELEC values are almost independent of the transporter. This term also includes the hydrogen bonding interactions between the phospholipid heads and N-H binding sites, which were not found to differ over the range of transporters (Table S8) in agreement with the similar Pint···decalinCOM and Pint···N-HCOM distances reported in Fig. 4 of the main text. In contrast, the ΔEvdW term is the major contribution for the stabilisation of the transporters inserted into the phospholipid bilayers. As expected, a linear trend between the number of carbon atoms of alkyl chain substituent (and thus the total size of the molecule) and the ΔEvdW was found, with the most negative stabilizing energy between 7f, bearing the longest alkyl substituents, and the phospholipids.