Transmembrane Ion Channels Formed by a Star of David [2]Catenane and a Molecular Pentafoil Knot

A (FeII)6-coordinated triply interlocked (“Star of David”) [2]catenane (612 link) and a (FeII)5-coordinated pentafoil (51) knot are found to selectively transport anions across phospholipid bilayers. Allostery, topology, and building block stoichiometry all play important roles in the efficacy of the ionophoric activity. Multiple FeII cation coordination by the interlocked molecules is crucial: the demetalated catenane exhibits no anion binding in solution nor any transmembrane ion transport properties. However, the topologically trivial, Lehn-type cyclic hexameric FeII helicates—which have similar anion binding affinities to the metalated Star of David catenane in solution—also display no ion transport properties. The unanticipated difference in behavior between the open- and closed-loop structures may arise from conformational restrictions in the linking groups that likely enhances the rigidity of the channel-forming topologically complex molecules. The (FeII)6-coordinated Star of David catenane, derived from a hexameric cyclic helicate, is 2 orders of magnitude more potent in terms of ion transport than the (FeII)5-coordinated pentafoil knot, derived from a cyclic pentamer of the same building block. The reduced efficacy is reminiscent of multisubunit protein ion channels assembled with incorrect monomer stoichiometries.


■ INTRODUCTION
The orderly entanglement of molecular strands within knots 1 and links 2 (catenanes) can induce properties and characteristics 1a2a that are beginning to be explored in areas as diverse as anion binding, 3 catalysis, 4 materials, 5 health care, 6 and the kinetic stabilization of supramolecular structures. 7 One strategy to synthesize examples of complex molecular topologies is to join the ends of building blocks that assemble into cyclic metal double helicates. 1a,8,9 In a typical example, by tuning of the assembly conditions, the same set of components can be enticed to selectively form either a five-membered or a sixmembered Lehn-type cyclic helicate. 8 Subsequent macrocyclization of the building block end groups by ring-closing metathesis gives the corresponding pentafoil (5 1 ) knot 4a or Star of David triply interlocked [2]catenane (6 1 2 link), 9c respectively (Scheme 1). The metalated knot, link, and the parent open cyclic helicates display good to very strong halide binding affinities in their central cavities (K a ≈ 10 5 −10 10 M −1 in MeCN). 3a The anion-binding properties of the interiors of these metal-coordinated molecular structures led us to investigate their potential as transmembrane ion channels or transporters. 10 Protein ion channels typically have complex secondary, tertiary, and quaternary structures that determine their transport properties. 11 For example, membrane-spanning proteins often consist of α-helix bundles or β-barrels, which confer rigidity and anchor the protein within the membrane. In contrast, small-molecule systems are generally much less complex 12 and rely on intermolecular interactions, such as hydrogen bonding 13 and aromatic stacking, 13f,14 to promote their assembly into rigid membrane-spanning ion channels. Metal ion-ligand coordination can also be used to assemble ionophoric supramolecular structures, and a number of metallo-organic ion transporters have also been described. 15 We reasoned that the high anion affinities 3a and the rigid shape with internal cavity imposed by the topology of the metalated pentafoil knot 2 4a and Star of David link 3 9c could make them suitable candidates for ion transportation (Scheme 1). Furthermore, the two molecular structures are assembled from the same repeat unit, 1, allowing the direct comparison of assemblies of the same building block in two different stoichiometries.
A series of assays in phospholipid vesicles were used to examine the ion transport behavior and selectivity of 2 and 3 ( Figure 1). The metalated pentafoil knot was used as the monochloride salt (other anions PF 6 − ), 2, because the fully hexafluorophosphate salt rapidly exchanges one PF 6 − for a Cl − ion by sequestering traces of chloride from glassware or solvent. 3a The somewhat less halophilic metalated Star of David catenane was used as the fully hexafluorophosphate salt, 3. The ion transport abilities of the metalated knot and link were compared with that of the demetalated Star of David catenane, 4, which without transition-metal coordination exhibits no anion binding in solution, and Fe II -coordinated Lehn-type open cyclic hexameric helicates (5 and 6), to determine the influence of (i) metal-binding allostery and (ii) molecular topology ( Figure 2). The results of the vesicle experiments were corroborated by single-channel planar bilayer conductance measurements ( Figure 3).

■ EXPERIMENTAL DETAILS
The Fe II -coordinated pentafoil knot 2 4a and the Star of David [2]catenane 3 9c were prepared from ligand 1 in the presence of different Fe II salts followed by ring-closing olefin metathesis (Scheme 1). The ionophoric activity of the compounds in the vesicles was initially assessed in 8-hydroxypyrene-1,3,6-trisulfonate (HPTS) assays using 4:1 v/v egg yolk phosphatidylcholine (EYPC)/cholesterol vesicles in 3-(N-morpholino)propanesulfonic acid (MOPS) buffer (15.2 mM lipid, [MOPS] = 20 mM, pH 7.4). 16 In these experiments, vesicles containing HPTS and different salts were prepared ([salt] = 100 mM, see Section S3.1, Supporting Information). Either knot 2 or link 3 (2.5 μM, 0.26 mol % relative to EYPC/cholesterol) was added to the solution containing the vesicles, followed by a pulse of NaOH (6.5 μM). Ionophoric activity was indicated by dissipation of the resultant pH gradient between the interior and exterior of the vesicles, which was monitored through the change in the relative fluorescence of HPTS ( Figure 1A). Finally, the vesicles were lysed by the addition of a detergent (Triton X-100) to allow normalization of the fluorescence response. The resulting kinetic data were fitted to pseudo-first-order rate equations as an approximation 13f,17 to compare the effectiveness of the analyzed compounds (see Tables S1−S3). Background rates for ion leakage across the membranes for a range of salts were measured (see Figures S6 and S7) and the values subtracted from the knot/link-mediated ion transport rates.

■ RESULTS AND DISCUSSION
We first examined the ionophoric activity of the knot and link with respect to KBr ( Figure 1B). The (Fe II ) 5 -coordinated pentafoil knot 2 displayed weak ion transport activity (k obs = 4.05 × 10 −4 s −1 , see Figure S9), while (Fe II ) 6 -coordinated Star of David catenane 3 was >50-fold more active (k obs = 2.30 × 10 −2 s −1 , see Figure S8). The pentafoil knot and Star of David link both showed significant ion transport and differ only in the number of units of 1 included in their assembly (five and six, respectively). This difference alters both the size of the central hole and the number of cations associated with each structure. It is somewhat reminiscent of the outcome of subunit assembly with the wrong stoichiometry in some biological channels and pores; misassembly with respect to the number of building blocks still results in a functional ion channel but with greatly reduced efficacy of transport. 18 The Fe II -coordinated knot and link are significantly different from the metal-free Star of David catenane ligand 4 in that the bound metal ions confer structural rigidity, add multiple positive charges, and inductively withdraw electron density from the pyridine rings, all of which create the strong anion binding site in the central cavity (Figures S1 and S3). As expected from its lack of anion binding affinity in solution, the demetalated Star of David link 4 showed no activity in the HPTS assay. Allostery, in the form of the coordination of Fe II cations, is an essential feature to enable both anion binding in solution and anion transmembrane transport by 2 and 3. Nonetheless, the charge provided by complexation to Fe II is not in itself sufficient to cause ion transport; simple Fe II (bipy) 3 complexes, such as FeCl 2 (bipy) 3 , showed no ion transport in analogous experiments ( Figure S12).
To determine the significance of the interlocked molecular topology in the ion transport experiments, we also tested two Lehn-type circular metal helicates: methyl-derivative 5 and 6, Scheme 1. Synthesis of (Fe II ) 5  Journal of the American Chemical Society pubs.acs.org/JACS Article the immediate precursor to 3 (Figure 2). These metalcoordinated complexes have anion affinities in solution that are very similar to the (Fe II ) 6 -coordinated Star of David link 3. 3a,8 Remarkably, given the activity of the topologically complex compounds 2 and 3, neither 5 nor 6 showed any activity in HPTS assays with a range of salts (for HPTS assays with 5, see    Figure S10; for HPTS assays with 6, see Figure 3B). The inactivity of 6 is particularly unexpected as it differs from the highly active (Fe II ) 6 -coordinated Star of David catenane 3 only in that it has not undergone ring-closing olefin metathesis. It appears that the conformational constraints imposed by joining of the strand ends is crucial for the metalated Star of David catenane 3 to be able to form transmembrane ion channels (although the knot is less effective in doing so). The less ordered ligand exterior is apparent in molecular modeling of helicate 6 ( Figure S2). Having established the strong ionophoric activity of the (Fe II ) 6 -coordinated Star of David link 3, we examined its anion selectivity ( Figure 1C, Figure S8). The metalated Star of David link 3 elicited ion transport decreasing in the order KSCN ≈ KI > KBr > KCl > KClO 4 ≈ KNO 3 > KF. The halides and pseudohalides follow the Hofmeister series, 19 suggesting that desolvation 20 may be the rate-limiting factor for ion transport. However, the transport rates of perchlorate and nitrate differ from that expected from their positions in the Hofmeister series (perchlorate sits between iodide and thiocyanate; nitrate sits between bromide and chloride). Haynes et al. observed a similar effect for Zn 10 L 15 cages, which was attributed to size/ shape exclusion 15c,21 hindering the passage of the larger perchlorate and nitrate anions. 19,22 Changing the cation showed much less variation in transport rates ( Figure 1C, Figure S8d, Table S2), demonstrating that the ionophores are selective only between anions. 15a,23 Anion transport was confirmed by high levels of chloride transport in lucigenin assays (see Section S3.2, Supporting Information).
The mechanism of transport was probed to determine whether the (Fe II ) 6 -coordinated pentafoil knot 2 and Star of David catenane 3 were merely disrupting the phospholipid bilayer, acting as an ion carrier or forming ion channels. 5(6)-Carboxyfluorescein assays ruled out vesicle lysis or the formation of large, nonspecific channels (see Section S3.3), 24 while U-tube experiments 25 demonstrated that both the metalated knot 2 and link 3 did not act as carriers (see Section S3.4). These observations suggest that the HPTS assay data arise from anion-selective channels formed in vesicle membranes by the metalated knot and link.
We further investigated the ion channel activity of the molecular knot and link by way of single-channel planar bilayer conductance (PBC) experiments (see Section S4 and Figure  3). 26 A planar bilayer was formed between two wells containing buffer (20 mM MOPS,1 M KCl,pH 7.4) and a potential applied across the membrane. The channel-forming compound was then added to one well ( Figure S19), resulting in step changes in the ion current, indicating the formation of ion channels. While we did not observe current steps with the metalated pentafoil knot 2 within the 2 h duration of the measurement, the addition of the metalated Star of David link 3 led to reproducible, quantized, square-topped "flicker" events 26 ( Figure 3A, Figures S20 and S21), indicative of short-lived ion channels. Increased channel formation was observed under negative applied potentials compared to when a positive potential was applied ( Figure 3B). This is consistent with the positively charged link 3 being driven toward the membrane. 27 The channels have a high conductance around 0.22 nS (measured at +100 mV, Figure 3C), mirroring the high ionophoric activity observed in the HPTS assays (Figure 1). Hille analysis (Section S4.4) of the PBC data allowed the diameter of the channels formed by 3 to be estimated as 5.0− 7.2 Å.
A Hill plot analysis 28 of the HPTS assays for KCl transport by (Fe II ) 6 -coordinated Star of David catenane 3 (see Figure  S11) gave an EC 50 value of 1.3 μM (17.4 μg mL −1 ) and a channel stoichiometry of 1 (i.e., a channel is made by a single molecule of 3). While Hill plot analysis has many underlying assumptions, 29 in combination with the observed ion selectivity, it suggests that the formation of single catenane ion channels able to transport anions (e.g., perhaps through a "relay-race" mechanism 30 ) is a reasonable explanation for the observed ionophoric activity.
The overall picture that emerges from the experimental studies of the (Fe II ) 6 -coordinated Star of David 3 is that the high ionophoric activity arises from membrane insertion to form channels, which are likely to involve a single molecule of 3 (Hill plot analysis, n = 1, Figure S11). Given its size and relatively hydrophobic edges ( Figure S3), we speculate that the catenane could embed in the hydrophobic membrane with its faces parallel to either side of the otherwise intact membrane. Although the thickness of the catenane is significantly less than the thickness of the bilayer, the bilayer may thin around the channel former, as observed for the shorter antibiotics in the peptaibol family. 31 The markedly better ionophoric activity of the Star of David catenane 3 (constructed from six molecules of building block 1) over pentafoil knot 2 (made from five molecules of 1) may be a consequence of the weaker binding of anions (such as Cl − , Br − and I − ) in the larger diameter (∼0.46 nm) central hole of 3 (Figure 4). 3a The X-ray crystal structure 4a of pentafoil knot 2 shows the tightly bound chloride anion is displaced above the central cavity ( Figure S1), with the hole diameter at its narrowest being smaller (∼0.24 nm) than the ionic diameter of a chloride ion (∼0.36 nm), 22 suggesting that the anion might not easily pass through to the other side. This very tightly bound anion may also have a very slow rate of dissociation, effectively blocking the hole. The anion selectivity of the . Solvent-accessible iso-surface overlaid on X-ray crystal structures of (Fe II ) 5 -coordinated pentafoil knot 2 (left) 4a and (Fe II ) 6coordinated Star of David link 3 (right) 9c shown from above (top) and from the side (bottom). The structure dimensions (yellow double-headed arrows) and the diameters of the cavity (red) are shown. A 1 H radius of 1 Å, a typical value for hydrogen-bonding H atoms, 20 was assumed for distance measurements (see Section S2.2).
Journal of the American Chemical Society pubs.acs.org/JACS Article channel formed by 3 broadly follows the relative anion binding affinity of the Star of David catenane in MeCN solution. 3a The calculated Hille diameter is also similar to the hole diameter in the solid state. 9c The low rate of perchlorate and nitrate transport relative to their place in the Hofmeister series is also consistent with anion flow through rigid channels of welldefined size and shape, with a partial size-exclusion mechanism being the cause of the observed selectivity. 19,21,22 Nonetheless, we do not rule out an alternative transport mode not involving the central cavity. If the compounds insert perpendicularly into the membrane, the ions may flow around the molecule, as previously observed for certain DNA duplexes, 32 although it is not clear how to rationalize such a mechanism with the observed size/shape departure from the Hofmeister series in anion transport.

■ CONCLUSIONS
Our results show that a (Fe II ) 6