A halogen bond-mediated highly active artificial chloride channel with high anticancer activity

Modularly tunable monopeptidic scaffold enables rapid and combinatorial evolution of a halogen bond-mediated highly active chloride channel, exhibiting an excellent anticancer activity toward human breast cancer.


Introduction
Tightly regulated by varying families of ion channels, precise maintenance of ion concentration gradients across biological membranes is crucial for many biological and cellular processes. Chloride is the most abundant anion in the human body due to its important roles in controlling the membrane potential, cell volume, cellular pH balance, secretion of transepithelial uid and electrolytes, etc. 1 Misregulation of chloride ions could lead to a variety of life-shortening 'channelopathies' including cystic brosis, Bartter syndrome and Dent's disease. Accordingly, signicant research efforts have been made recently to coherently develop articial chloride carriers 2 or channels 3 as possible replacements for natural chloride channels for channelopathies. 4 On the other hand, articial anion transporters might also disrupt pH gradients 4f,5a,b or ion homeostasis 5c that may induce cell death, and therefore may have medical applications as anticancer agents. 4f,5 Recently, Matile et al. reported self-assembled chloride carriers 2b,c or channels, 3i elegantly exploiting halogen bonding interaction as the driving force for anion transport. Halogen bonds offer good bond strength and also good anion-induced directionality. 6 Therefore, articial anion transporters comprising simple structural motifs based on halogen-anion interactions may indeed offer a new horizon of prospects to comprehend complex chloride transport phenomena, oen manifested in protein channels. Nevertheless, most documented articial chloride channels purposefully harness various non-covalent interactions such as electrostatic, Hbonding, anion-p and halogen bonding forces. Channels that primarily rely on halogen bonds for facilitating anion transport still remain limitedly investigated with just one precedent. 3i Such a rare occurrence presumably results from the lack of suitable molecular scaffolds for aligning electron-decient iodine atoms into a well-dened one-dimensional (1D) array. Moreover, the majority of existing chloride channels 3a-m suffer from relatively complex molecular structures and/or low selectivity in anion recognition, greatly restricting their application in drug discovery.
We recently described a novel class of readily accessible monopeptide-based scaffolds capable of self-assembling into 1D columnar structures through highly directional intermolecular H-bonding forces, subsequently enabling rapid roomtemperature gelation of diverse types of crude oils (Fig. 1a). 7a-c Revealed by the crystal structure of F-Phe-C4 (Fig. 1a), one unique feature of this monopeptide-based scaffold is that the same type of side chains such as Fmoc, R 1 and R 2 are always aligned to the same side as a result of the high directionality of H bonds (Fig. 1a). 7a On this basis, we envisioned that replacing the Fmoc group with a chloride-binding unit such as the tet-rauoroiodobenzyl group may lead to an interesting class of articial anion channels with anion transport mediated by halogen bonds (Fig. 1b-d). 3i,7d A high intrinsic modularity involving R 1 and R 2 groups in the backbone should allow for rapid and combinatorial optimization of transport activity and selectivity of channels. In this report, we demonstrate that this indeed is the case, with the identied chloride channel exhibiting a high activity in chloride transport with respect to many other types of anions. We further demonstrate that human breast cancer cells (BT-474) are sensitive to a sodium chloride concentration gradient across the membrane, and this feature can be utilized by articial chloride channels to inhibit cancer cell growth for their potential uses in cancer chemotherapy. 4f,5

Results and discussion
Combinatorial identication of highly active aniontransporting articial channels We synthesized a total of 15 possible channel molecules (ve amino acids Â three alkyl chains of different lengths, Fig. 1b) with the tetrauoroiodobenzyl group as the chloride-binding and -transporting unit. Corroborated by the crystal structure of F-Phe-C4 (Fig. 1a) and the computationally optimized structure of L8 (Fig. 1c), the ability of these 15 molecules to pile up to form a H-bonded 1D structure can be convincingly demonstrated by the ability of A10, L8 and L10 to efficiently congeal in n-hexane via the formation of a 3D entangled brous network (ESI, Fig. S1 †). The ion transport activities of the synthesized channels were then evaluated using the pH-sensitive HPTS (8-hydroxypyrene-1,3,6-trisulfonic acid) assay ( Fig. 2 and S2 †). 8 In a typical experiment, large unilamellar vesicles (LUVs), encapsulating HPTS (100 mM) and NaCl (100 mM) at pH ¼ 7.0, were diluted into the same buffer at pH 8.0 to generate a pH gradient across LUVs. Aer the addition of channel molecules, changes in the uorescence intensity of HPTS were monitored over 5 min. Fig. 2b shows that, among all 15 channels tested at 10 mM, A10, L8 and L10 display the highest ion transport capacities with fractional activities (R Cl À ) of 60%, 68% and 59%, respectively. The corresponding EC 50 values, which are the   2 (a) Schematic illustration of the HPTS assay for the Cl À transport study conducted using a pH gradient of 7 to 8 and the pH-sensitive dye HPTS entrapped inside LUVs. (b) Normalized transport activities (R Cl À ) obtained over 5 min at 10 mM for all channel molecules. R Cl À ¼ (I Cl À À I 0 )/ (I Triton À I 0 ) wherein I Cl À and I 0 are the ratiometric values of I 460 /I 403 before the addition of triton at t ¼ 300 s, and I Triton is the ratiometric value of I 460 /I 403 at t ¼ 300 s right after the addition of triton with internal/external buffers containing 100 mM NaCl. All the data were averaged over two measurements.
Cl À /OH À as the major transport species Except for the presence of arrayed electron-decient iodine atoms that can bind the Cl À anion, 2b,c,3i all peptide molecules carry no other readily accessible functional groups for strongly binding either cationic or anionic species. Further, varying extravesicular metal chloride salts from LiCl to CsCl produces near-identical changes in uorescence intensity ( Fig. 3a and S4 †), suggesting the inability of L8 or A10 to transport any of the ve alkali metal ions. These combined structural features and experimental evidence led us to believe that, rather than the H + / M + antiport mechanism, either the OH À /Cl À antiport ( Fig. 2a) or H + /Cl À symport mechanism may largely account for the observed increases in the uorescence intensity of HPTS.
To deduce the most likely transport species, we rst carried out the SPQ assay 9 using a chloride-sensitive SPQ dye (6-methoxy-N-(3-sulfopropyl)quinolinium). As illustrated in Fig. 3b and S5, † the addition of L8 or A10 results in a rapid quenching of SPQ uorescence in a concentration-dependent manner. These results suggest that the inux of Cl À increases with increasing channel concentrations from 0.32 to 5 mM for L8 and from 2.5 to 20 mM for A10, thereby establishing the Cl À anion as one of the molecular species transported by both L8 and A10.
Next, carbonyl cyanide 4-(triuoromethoxy)phenylhydrazone (FCCP, a well-known proton carrier) was employed (Fig. 3c). Compared to the transport efficiencies of 11% for FCCP alone and of 57% for L8 alone, an increased transport efficiency of 13% (e.g., 78-57% -(11-3%)) for L8 in the presence of FCCP indicates a cooperative action between L8 and FCCP. In the case of A10, an even larger enhancement of 25% was observed ( Fig. S6 †). These results suggest that the transport rate of Cl À is faster than that of H + .
The HPTS assay was then performed in the presence of valinomycin (VA, a K + -selective carrier, Fig. 3d) to compare the transport rate between OH À and Cl À . For this assay carried out under iso-osmolar K + (extravesicular) vs. Na + (intravesicular) and pH gradient, the VA-mediated inux of K + ions will induce an anion channel-mediated inux of either OH À or Cl À in order to maintain overall charge equality. If the inux of OH À is faster than that of Cl À , an increase in uorescence intensity is anticipated. Experimentally, valinomycin at 25 pM produces a small increment of uorescence intensity (7%) compared to the blank (3%). In a similar way, the ion transport activities of L8 (3.6 mM) in the presence and absence of valinomycin were found to be near-identical (58% vs. 59%), which is indicative of a preferential transport of Cl À over OH À (e.g., Cl À > OH À ). A similar result was also seen for A10 (Fig. S7 †). Consistent with transport activities presented in Fig. 3a, these data also suggest that the H + /M + antiport mechanism is unlikely for L8-mediated enhancement in uorescence intensity.
A mere increase of 1.3% (4.6-3.3%, Fig. 3e) in uorescence intensity at 3.6 mM for molecule 1, carrying a simple benzene group, strongly suggests a lack of readily accessible functional groups in the H-bonded peptidic backbone for interacting with either anions or protons, and thus a minor role played by the backbone in mediating chloride transport. Instead, it is halogen bonds formed between anions (e.g., Cl À or OH À ) and electron-decient iodine atoms that cause efficient exchanges between Cl À and OH À anions across LUVs as observed for L8 (57%), pointing to a high unlikelihood of having H + /Cl À as the transport species. The herein assumed formation of halogen bonds I/Cl À and I/OH À was recently observed in their corresponding crystal structures 2c and can be further supported by 19 F NMR titration experiments involving titrating 0-20 equiv. of tetrabutylammonium chloride into a D 2 O-saturated CDCl 3 solution containing L8 at 1 mM (Fig. S8 †). Relative to the internal standard (1,4-diuorobenzene, À120.50 ppm), increasing additions of up to 20 equiv. of TBACl led to increasing upeld shis of up to 0.61 ppm in the chemical shi of 19 F of L8, a fact that is consistent with earlier observations 2b and indicates binding between the chloride anion and the acidic iodine atom.
Additional comparisons among L8 and 2-4 convincingly demonstrate the importance of (1) the co-existence of two amide bonds that likely allow for a tighter and more directional stacking of channel molecules (L8 vs. 2), (2) the side chain of leucine for more efficient stacking (2 vs. 3) and (3) the amide bond in forming a H-bonded structure for ordered spanning of the hydrophobic membrane region (3 vs. 4). These summative ndings are also in accordance with the molecular dynamics (MD) simulation results (Fig. 4).

Anion selectivity in chloride transport
With high activity exhibited by these chloride transporters, our subsequent investigations focused on evaluating and comparing anion selectivity for the most efficient anion channels formed by A10, L8 and L10. For this purpose, both intraand extravesicular anions were kept the same (100 mM NaX, X À ¼ Cl À , Br À , I À , NO 3 À and ClO 4 À ) with a proton gradient of pH 7 (inside) to pH 8 (outside). Data compiled in Fig. 3f, S9 and S10 † show that A10 with an EC 50 value of 9.4 mM exhibits good anion selectivity in chloride transport, and channels L8 and L10 likely selectively transport non-chloride anions (Fig. S10 †).

Chloride transport through a channel mechanism
Single channel current traces for chloride transport, recorded in a planar lipid bilayer at various voltages including À100 mv in symmetric baths (cis chamber ¼ trans chamber ¼ 1 M KCl, Fig. 3g and S11 †), unambiguously conrm that L8-mediated chloride transport occurs via a channel, rather than a carrier mechanism. On the basis of the tted linear current-voltage (I-V) plot ( Fig. 3h and S10 †), the Cl À conductance (g Cl À ) of L8 was found to be 586 AE 11 fS.
To shed some light on the possible structural features of onedimensionally aligned channel molecules in the lipid membrane, MD simulation using the CHARMM program, 10a-e PME method 10f and SHAKE algorithm 10g was performed on L8. In particular, the H-bonded 1D structure, which consists of eight molecules of L8 (528 atoms), was rst computationally optimized using the COMPASS force eld 10h and then was embedded in a bilayer of 128 POPC molecules (17 152 atoms) solvated on two sides by 2 Â 2397 water molecules (Fig. 4a). This leads to a simulation system of 32 062 atoms with a dimension of 70Å (w) Â 70Å (w) Â 74Å (h). Aer equilibration steps, the production run was carried out for 30 ns. The last 20 ns trajectories with 1000 structural snapshots were used for analysing both the hydrophobic thickness of the POPC membrane and the distribution probability of angle q (e.g., the angle formed by the three immediately adjacent iodine atoms such as iodine atoms 2-4, Fig. 4a and b) using a probability density function. 10i To estimate the hydrophobic membrane thickness, the Z-coordinates along the membrane normal of all 254 ester O-atoms from 64 POPC molecules located either at the top or on the bottom layers were averaged. The separation distance between the two averaged Z-coordinates of O-atoms from the top and bottom layers was calculated with deduction of a van der Waals diameter of 3.1Å for the O-atom. The same calculation was performed for all 1000 structural snapshots to derive the probability distribution of the hydrophobic membrane thickness (Fig. S12 †). From these computations, 633 structures have a hydrophobic thickness of 27 to 29Å, and the average thickness over 1000 structures is 28.1Å, a value that could be corroborated by the experimentally and computationally determined values of 27.1 and 27.8Å, respectively, for the POPC membrane. 11 Embedding eight molecules of L8 in the lipid membrane shows that the rst and eighth molecules are not aligned well with the central six molecules (Fig. 4a and c). The most highly populated angles for q 234 , q 345 , q 456 and q 567 are 159 , 162 , 163 and 157 , respectively, with a representative structure, closely capturing these angles, shown in Fig. 4c. In this highly populated structure, the largest intermolecular separation distance along the Z axis among the eight iodine atoms occurs between the 1 st and 7 th iodine atoms (25Å). Consistent with the intermolecular separation of 5.0Å in Fmoc-Phe-C4 in the solid state (Fig. 1a) 7a and of 4.9Å in the computationally optimized Hbonded L8 (Fig. 1c), this separation distance of 25Å suggests that six or seven molecules might be sufficient to span the hydrophobic core distance of 28.1Å.
Through these MD simulations, the most important point to note is that, regardless of angle q of varying magnitudes, all eight molecules of L8 remain H-bonded to each other in all 1000 structures surveyed (Fig. 4c). This demonstrates the reliability of intermolecular H-bonds in linking six or seven anion-binding molecules together to form a self-assembled 1D pathway, which spans the hydrophobic membrane region to facilitate highly efficient anion transport across the membrane.

High activity in chloride transport
The use of halogen bonds to mediate anion transport was rst explored by Matile, 2b,c,3i with compounds 5 (EC 50 ¼ 3.1 mM) 2c and 8 (EC 50 ¼ 0.88 mM in terms of chloride binding units) 3i as the most active carrier and channel molecules among their own Fig. 4 (a) Embedding computationally optimized H-bonded L8 inside a simulation box of 70Å (w) Â 70Å (w) Â 74Å (h), comprising 128 POPC molecules and 4794 water molecules. (b) Probability distribution patterns of q 234 , q 345 , q 456 and q 567 obtained after analysing 1000 structures. (c) A representative highly populated structure with the four q angles approaching those having the highest probabilities shown in (b). POPC ¼ 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine. categories, respectively. These EC 50 values, including those of 6 (260 mM) and 7 (68 mM) presented in Table 1, were determined using cholesterol-free LUVs with a total lipid concentration of 31.3 mM, while our assay conditions contain lipid and cholesterol at concentrations of 73.9 and 36.9 mM, respectively. For a fair comparison, we have therefore re-determined the EC 50 values of A10, L8 and L10 using Matile's conditions (100 mM NaCl, 10 mM HEPES, pH gradient from 7 to 8, and the total lipid concentration without cholesterol ¼ 31.3 mM). 2b,c,3i To conrm that we are able to produce LUVs with properties similar to those prepared by Matile and his co-workers under the same assay conditions, we have performed Hill analysis on commercially available compound 5, and obtained an EC 50 value of 3.6 mM with a Hill coefficient of 3.3 (Fig. S13 †). These values are similar or identical to the reported values of EC 50 (3.1 mM) and the Hill coefficient (3.3). 2c Under these assay conditions, the EC 50 values of A10, L8 and L10 were determined to be 2.37, 0.39 and 0.93 mM (Table 1 and Fig. S13 †), respectively. The most active L8 is 8.2 and 1.2 times more active than carrier 5 (3.6 mM) and unimolecular channel molecule 8 (0.88 mM in terms of effective chloride-binding units). When normalized based on the molecular weight, the ion transport activity of L8 is about 6.5 and 1.6 times those of 5 and 8, respectively. That L8 is more potent than unimolecular channel 8 is a clear indication of the excellent ability of the noncovalently associated peptidic scaffold in inducing linearly arrayed iodine atoms into a conformation that is more conducible to chloride transport than the rod-like scaffold in 8.
In addition, chloride carrier 5, when evaluated under our assay conditions containing 33 mol% cholesterol that makes membrane less uid, 12 shows a moderate transport activity of 38% at 40 mM, and such moderate activity remains unchanged even when the concentration increases to 100 mM (Fig. S14 †). These values suggest the ion transport activity of 5 to be at least 10 times lower than that of L8 having an EC 50 value of 3.6 mM in the cholesterol-containing environment.

High anticancer activity
The above veried highly efficient chloride transport through a channel mechanism displayed by these articial anion channels prompted us to examine the possibility of their uses in cancer chemotherapy. 4f,5 Human breast cancer cells (BT-474, obtained from the American Type Culture Collection, USA) were cultured in Dulbecco's Modied Eagle Medium in the presence of up to 19.2 g L À1 NaCl (6.4 g L À1 is the typical concentration for cell growth) with concentrations of L8 or A10 varying from 0 to 100 mM ( Fig. 5 and S15 †). The viabilities of cells, determined aer culturing the cells for 1 day at 37 C with 5% CO 2 in the absence of channel molecules, show that BT-474 cells are sensitive to the sodium chloride concentration gradient (Fig. 5a). That is, compared to cell viability at [NaCl] ¼ 6.4 g L À1 , 12, 79 and 99% fewer cells survive at [NaCl] of 12.8, 16.0 and 19.2 g L À1 , respectively. Continued testing shows that A10 is considerably more potent than L8 in inhibiting cell growth across all three different concentrations of NaCl (e.g., 6.4, 9.6 and 12.8 g L À1 ) with an IC 50 value of 20 mM for A10 at [NaCl] ¼ 6.4 g L À1 (>100 mM for L8, Fig. 5b and S15 †). For comparison, highly effective anticancer agent cisplatin has an IC 50 value of 37 mM against the same BT-474 cells. 13 Assuming all A10 molecules associate to form channels with each channel comprising six or more such molecules, the IC 50 value in terms of effective channel concentration is lower than 3.3 mM.

Conclusions
In conclusion, we have developed a class of articial chloride channels with high activity and good selectivity via directional self-assembly of polarized electron-decient iodine atoms. While these linearly arrayed acidic iodine atoms, which form a series of chloride-binding and -transporting sites, are responsible for facilitating chloride transport likely via a multiion jumping mode, the high modularity of the monopeptide backbone enables rapid optimization of the transport activity and selectivity of channels in a combinatorial format. The identied channels A10, L8 and L10 all turn out to be very active with the best EC 50 values for chloride transport reaching 0.39 mM (1.2 mol% relative to lipid) and 3.6 mM (3.2 mol% relative to lipid/cholesterol) in cholesterol-free and -containing LUVs. In particular, the highly active A10 exhibits not only a fractional transport activity for chloride much better than other monovalent anions including bromide, but also an excellent inhibitory activity toward human breast cancer cells with an IC 50 value of 20 mM. Further renement to enhance the anion selectivity and cytotoxicity of channels toward cancer cells is possible and expected.

Conflicts of interest
There are no conicts to declare.