Porous organic cages as synthetic water channels

Nature has protein channels (e.g., aquaporins) that preferentially transport water molecules while rejecting even the smallest hydrated ions. Aspirations to create robust synthetic counterparts have led to the development of a few one-dimensional channels. However, replicating the performance of the protein channels in these synthetic water channels remains a challenge. In addition, the dimensionality of the synthetic water channels also imposes engineering difficulties to align them in membranes. Here we show that zero-dimensional porous organic cages (POCs) with nanoscale pores can effectively reject small cations and anions while allowing fast water permeation (ca. 109 water molecules per second) on the same magnitude as that of aquaporins. Water molecules are found to preferentially flow in single-file, branched chains within the POCs. This work widens the choice of water channel morphologies for water desalination applications.

gradually grew on the wall of the vial. The crystalline product was removed by centrifugation and washed once with dichloromethane, followed by three times with ethanol and diethyl ether each. The crystals were then further dried under vacuum overnight. MS (ES+) 1117.8 ([M+H] + ).

Synthesis of CC3 in polar solvents.
Instead of using dichloromethane, CC3 crystals can be successfully grown in polar mixture solvent consisting of tetrahydrofuran/water or ethanol/methanol/water. Generally, 1,3,5-triformylbenzene (40 mg, 0.25 mmol) was dissolved in tetrahydrofuran (2 mL) or ethanol/methanol in equal parts (2 mL), and (R,R)-1,2diaminocyclohexane (40 mg, 0.35 mmol) was dissolved in tetrahydrofuran/water or methanol/water in equal parts (2 mL). The (R,R)-1,2-diaminocyclohexane solution was then added slowly to 1,3,5-triformylbenzene solution. The mixture was sealed in a vial and let stand for 1-2 days. Crystals were found on the wall of the vial. CC3 tends to form small and uniform crystals. The usage of methanol in the solvent generally results in twinned crystals.
Synthesis of RCC3. CC3 (250 mg, 0.22 mmol) was dissolved in a dichloromethane/methanol mixture (1:1 v/v, 50 mL) by stirring. When this solution became clear, excess sodium borohydride (1.00 g, 26.5 mmol) was added and the reaction was stirred for a further 24 h at room temperature. The solvent was removed under reduced vacuum using a rotary evaporator until a gel-like product was obtained. The off-white solid was extracted with dichloromethane (20 mL) and washed with water (2 × 100 mL). The dichloromethane phase was evaporated under vacuum. RCC3 was obtained as an off-white solid. MS (ES+) 1155.9 ([M+H] + ).

Synthesis of FT-RCC3.
Paraformaldehyde (20 mg) dissolved in methanol (10 mL) was stirred at 70 °C. RCC3 (30 mg, 0.026 mmol) dissolved in methanol (10 mL) was then added to the paraformaldehyde solution. A white precipitate appeared upon addition of RCC3. The reaction was stirred for a further 2 h at 70 °C before allowing to cool down to room temperature. The FT-RCC3 precipitate was washed with methanol (3 × 10 mL) and dried under vacuum. MS (ES+) 1214.0 ([M+H] + ).
Synthesis of CC19. Dichloromethane (3 mL) was used to dissolve 2-hydroxy-1,3,5benzenetricarbaldehyde (40.3 mg, 0.25 mmol). A solution of (R,R)-1,2-diaminocyclohexane (40 mg, 0.35 mmol) in dichloromethane (3 mL) was then added slowly. Orange powder was obtained almost immediately. The powder was dissolved by the further addition of dichloromethane (1 mL). The vial containing the mixture was sealed and left to stand at room temperature for 1-2 days. Orange crystals were found on the wall of the vial. The crystals were S4 washed with aliquots of diethyl ether (3 × 10 mL) and dried under vacuum. MS (ES+) 1181.7 ([M+H] + ).

Synthesis of CC19 in polar solvents.
Instead of using dichloromethane, CC19 crystals were grown in polar mixture solvent consisting of ethanol/methanol/water. Generally, 2-hydroxy-1,3,5-benzenetricarbaldehyde (40.3 mg, 0.25 mmol) was suspended in ethanol/methanol in equal parts (2 mL), and (R,R)-1,2-diaminocyclohexane (40 mg, 0.35 mmol) was dissolved in methanol/water in equal parts (2 mL). The (R,R)-1,2-diaminocyclohexane solution was then added slowly to 1,3,5-triformylbenzene solution. The mixture was sealed in a vial and let stand for overnight. Crystals were found on the wall of the vial. CC19 tends to form small and uniform yellow crystals.
Synthesis of Pd@RCC3. RCC3 (10 mg, 0.0087 mmol) was dispersed in deionised water (5 mL) by stirring for 2 h. Then palladium acetate (26 μL, 0.25 mmol) dissolved in dichloromethane (10 mgmL -1 ) was added to RCC3 solution. The mixture was stirred for another 2 h before sodium borohydride (30 μL, 0.0079 mmol) dissolved in water (10 mgmL -1 ) was added. The mixture was centrifuged and washed with water 5 times. The wet Pd@RCC3 crystals were dried under a reduced pressure at 75 °C overnight. The dried Pd@RCC3 was then re-dissolved in dichloromethane. Pd@RCC3 was recrystallized by slow evaporation under room temperature.

Synthesis of ASPOC.
(R,R)-1,2-diaminocyclohexane (52.8 mg, 0.46 mmol) and ethylene diamine (27.8 mg, 0.46 mmol) were dissolved in dichloromethane (3 mL). This solution was then slowly added to 1,3,5-triformylbenzene (100 mg, 0.62 mmol) dissolved in dichloromethane (10 mL). The mixture was sealed in a vial and left to stand at room temperature for 7 days. The mixture was dried by solvent evaporation using rotary evaporator under reduced pressure and washed five times with aliquots of ethyl acetate.
Estimation of embedding efficiency of RCC3 in liposome. The embedding efficiency of RCC3 in liposome was measured using UV-Vis spectroscopy. The embedding efficiency of the channels was calculated based on the calibration curves of the UV-Vis absorbance of blank liposomes and RCC3 dissolved in methanol. Briefly, stock RCC3 (1 mgmL -1 ) in methanol was added to plain liposome solution (200 mL of 1 mgmL -1 liposome solution in HEPES buffer) at various fmCLRs (0.01 to 0.05). The calibration curve was constructed by matching the absorbance intensity with the fmCLR of the samples. A wavelength of 298.5 nm was chosen.

Sample preparation and image acquisition of Cryo-TEM. Liposomes containing Pd@RCC3
were prepared using egg-yolk phosphatidylcholine (EYPC) and 1,2-dioleoyl-sn-glycero-3phospho-L-serine (DOPS) in a mole ratio of 4:1 using the reverse-phase method. EYPC (79.2 μL, 10 mg mL -1 )/DOPS (20.8 μL, 10 mg mL -1 ) in chloroform and 0.03 fmCLR Pd@RCC3 dissolved in chloroform were added to a round-bottom flask. Chloroform, diethyl ether, and HEPES buffer were then added in the same flask in a volume ratio of 2:1:1. The flask was kept degassed under dry argon. Subsequently, the mixture was sonicated at 0-4 °C until a homogeneous water-in-oil mixture was obtained. The organic solvents were then removed under a reduced pressure using a rotary evaporator (178 rpm, 45 °C, in air). The liposomes obtained were extruded through a hand-held extruder with 0.2 μm track-etched membrane for 21 times to obtain monodisperse, unilamellar vesicles. The cryo-TEM images were recorded at a nominal S5 magnification of 59,000× giving a pixel size of 1.43 Å at the specimen. The images were acquired under high dosage of 50 e -Å -2 .
Sample preparation of POC-coated porous anodic aluminum oxide (AAO) substrate. POCs were coated on AAO substrates by spin-coating. POC solutions (100 μL, 1 wt% POC in dichloromethane (< 2 wt% methanol was added to increase solubility of CC3, CC19, and FT-RCC3)) were added to fully cover the AAO substrate before coating (1000 rpm, 1min). CC5 has very limited solubility (ca. 0.0002 wt% was used), hence multiple coating was conducted. The POC-coated substrates were kept under vacuum overnight to remove residual organic solvents before testing.   Figure 9│AFM images of blank liposome and liposome with CC3. 2D AFM images of blank supported lipid bilayer (SLB) (a) and CC3-incorporated (fmCLR 0.03) SLB (b) on mica sheet observed using tapping mode in buffer (10 mM HEPES). The same images were observed in 3D showing blank SLB (c) being smoother compared to CC3-incorporated (fmCLR 0.03) SLB (d) on mica sheet. Solid AFM was performed in tapping mode for blank SLB (e) and CC3-incorporated (fmCLR 0.03) SLB (f) on mica sheet. Some weak protrusions were observed in CC3-incorporated SLB and they are marked with white arrows. Larger CC3 nanoaggregates deposited on the SLB (circled in white) were also observed. Note: Solid SLB appears regularly patterned due to the smoothness of blank lipid bilayer and system noise. Note: For the cation transport tests, liposomes were suspended in buffer solution at pH 7. At 50 s, a base pulse (additional of NaOH) induced a change in extravesicular environment. This induces the efflux of protons from the liposome, deprotonating the pyranine and results in an increase in the fluorescence intensity ratio. We did not observe significant increase in fluorescence intensity ratio in both blank liposome and liposome containing POCs. Gramicidin A dissolved in dimethyl sulfoxide (265 nM) was added to each sample at 100 s inducing large increase in fluorescence intensity ratio, which indicates significant cation transport. At the end of the experiment, detergent was added to completely destroy the liposomes and release all pyranine, which is shown as a sudden spike in fluorescence intensity ratio. Ratiometric results have been confirmed by Dr. Mihail Barboiu's group (figures f-i). Samples of liposomes with CC3 and FT-RCC3 were prepared accordingly: FT-RCC3 (0.5 mg, 0.00041 mmol) was evenly premixed with EYPC (10 mg, 0.013 mmol) in chloroform before evaporation; CC3 (0.26 mg, 0.00023 mmol) was evenly premixed with EYPC (10 mg, 0.013 mmol) in chloroform before evaporation. Liposomes were prepared in buffer A (100 mM NaCl, 10 μM pyranine, 10 mM phosphate buffer at pH 6.4) and then exposed to buffer B (either 100 mM NaCl or CsCl, 10 mM phosphate buffer at pH 6.4) during ratiometric measurements. The calculation of single channel permeability is based on equations [1], [3], and [4] in the main text assuming 17 or 75 POC molecules per nanoaggregate. Error bars represent standard deviation of three independent replicates. The calculation of single channel permeability is based on equations [2], [3], and [4] in the main text assuming 17 or 75 POC molecules per nanoaggregate. Error bars represent standard deviation of three independent replicates. Only CC5 showed some ion permeability, which can be attributed to its large pore size. No ion rejection was observed in small CC5 nanoaggregate before the channels were totally blocked by lipid tails.

Supplementary Table 5. Unit area permeability of aquaporin, carbon nanotube and CC3
The permeability and packing density of Aqp1 are obtained from Murata et al. 1 and Walz et al 2 .
The permeability and packing density of CNT are obtained from Tunuguntla et al. 3 and Holt et al. 4 . The unit area permeability is the water permeability per micrometre-squared area, obtained from the product of single-channel permeability and packing density. The CC3 data were calculated based on a cross-sectional area of 22.4 nm 2 of the nanoaggregate.