Maintaining transmembrane electrochemical gradients via tightly controlled ion homeostasis is crucial for regulating cellular activities, a process that largely depends on channel proteins [1,2]. Through regulating the movement of Cl⁻ anions, multiple chloride channels help to determine cell membrane potential, modify cell cycle, adhesion and cell motility, maintain intracellular pH and cell volume, etc. Malfunctioning chloride channels may result in respiratory, neuromuscular and renal dysfunction and diseases such as cystic fibrosis. This has inspired the search for artificial transmembrane anion transporters to emulate functions of natural systems, shed light into transport mechanisms or serve as possible therapeutics [3–20].

Although investigation of transmembrane anion transporter has intensified in recent years, anion transport has been predominantly mediated by diverse anionophores capable of H-bonding to anions [3–16,21–37]. Yet, a great variety of innovative strategies that do not rely on intermolecular H-bonds for facilitated cross-membrane anion transport have also emerged, creatively applying halogen bonds [38–41], chalcogen bonds [42–44], pnictogen bonds [45,46], anion−π interactions [47–49], charge repulsion [50], bidentate phosphonium boranes [51] and cage structures [52,53] as well as side chain-side chain interactions [54–56] or intramolecular folding of aromatic foldamers [57] that create hollow sizable cavities.

Another inspiring strategy takes advantage of limitedly studied dynamic metal-anion interactions of metalloporphyrins to induce anion transport across the membrane [58–61]. Metalloporphyrins feature four pyrroline subunits interconnected via methane bridges in a circular fashion, with a metal ion sitting in the ring center that leaves up to two labile axial coordination sites available for anion binding. As early as 1996, Cuppoletti et al. reported increased anion permeability in lung epithelial cells caused by a Mn(Ⅲ)-bound metalloporphyrin chloride. Mao and his co-workers extended the study further, establishing that this metalloporphyrin induced autophagy and immunogenic cell death [59]. Structurewise, these anion transporters are based on monomeric [58,59] organometallic complexes. Here, we describe another alternative approach, employing a dimeric Co²⁺-porphyrin confined within an adaptive macrocyclic tubular cavity to achieve very substantial enhancements in anion transport activity by up to 35 folds over the monomeric Co²⁺-porphyrin complex.

Recently we reported a naphthalene diimide-based giant macrocycle 1, possessing a giant hexagon-shaped tubular cavity of up to 15.3 and 14.6 Å when fully stretched (Fig. 1a) [60]. This giant cavity demonstrates an adaptive feature in response to the guest molecule pagoda[5]arene [61], exhibiting a smaller cavity in the guest's absence that becomes much larger in its presence (Fig. 1b). Inspired by these early results, we hypothesized that the adaptive giant cavity in 1 might be capable of binding electron-rich porphyrins (G1) or metalloporphyrins (Gns, n = 2-5).

Indeed, through our current continued investigation of 1-mediated host-guest chemistry, we demonstrate here that 1 can bind two Gn molecules in its giant cavity, forming a sandwiched 1:2 host-guest complex 1·(Gn)₂ (Fig. 1c). And to our best knowledge, this is the first example of cavity-containing hosts that can bind two porphyrin molecules in its adaptive cavity. Further, we show that such a dimeric Co²⁺-porphyrin (G3)₂ geometrically confined in 1 readily promotes far more efficient anion transports than its free monomeric counterpart G3 by up to 35 folds through a dynamic metal-anion bond (Fig. 1d).

Commercially available G1-G5 were selected to investigate their binding properties with host 1. ¹H NMR experiments were first carried out to evaluate the host-guest complexation in CDCl₃. As shown in Figs. 2a-c, mixing 1 and G3 at 1:2 molar ratio generated one set of proton signals distinctively different from those of free host and free guest, suggesting not only the formation of a new complex but also the association and dissociation between 1 and G3 to be a fast-exchange process. Specifically, the H₁ signal in 1 shifts up field largely by 0.46 ppm, which indicated that 1 experienced a shielded magnetic environment provided by G3. Together with upfield changes in the chemical shift for protons H₂ and H₃ and those from 3.5–4.0 ppm (Fig. S1 in Supporting information) in 1, we can conclude that the bound G3 molecules largely reside in the cavity of 1, likely forming a sandwich supramolecular structure 1·(G3)₂. Similarly, 1 is also found to be capable of binding guest molecules of G1, G2, G4 and G5 at 1:2 molar ratio, possibly producing complexes 1·(Gn)₂ (Figs. S2-S5 in Supporting information). Additionally, UV-vis and fluorescence experiments were also conducted to investigate the complexation process between 1 and Gn. As shown in Fig. S34 and S35 (Supporting information), the ultraviolet absorption and fluorescence emission intensity of Gn changed dramatically after the addition of 0.5 equiv. of 1, which further confirmed the formation of complexes 1·(Gn)₂.

To quantitatively measure the association constants (K_a) between 1 and Gns (n = 1-5), ¹H NMR titrations were conducted by monitoring changes in chemical shift of H₁ in 1. The mole-ratio plot confirms all binding stoichiometries between 1 and Gns to be 1:2 (Figs. S6-S25 in Supporting information). Accordingly, K_a values were calculated for the 1:2 complexes 1·(Gn)₂ using the BindFit software. As summarized in Table 1, while we are not exactly sure why G3 demonstrates a considerable allosteric binding cooperativity value of 1.37 with K_a1 = (4.78 ± 0.15) × 10² L/mol and K_a2 = (1.63 ± 0.23) × 10² L/mol, other non-G3 guests however show negligible cooperativity, with the first Gn bound more tightly than the second Gn by 23-334 folds. Moreover, UV-vis spectroscopy Job plot experiments were also performed to determine the stoichiometry of the complex, and the results indicate the formation of a 1:2 complex between 1 and Gn (Fig. S32 in Supporting information), which are consistent with those obtained from NMR molar ratio analysis.

Prompted by the high likelihood that the formed sandwich structures 1·(Gn)₂ might facilitate cross-membrane anion transport via dynamic metal-ligand interaction, we conducted the pH-sensitive HPTS assay, having the intravesicular region filled with 0.1 mmol/L HPTS and 67 mmol/L Na₂SO₄ at pH 7.0 and the extravesicular region with 65 mmol/L Na₂SO₄ and 2.5 mmol/L NaX (X⁻ = Cl⁻, Br⁻, I⁻, NO₃⁻ and ClO₄⁻) at pH 6.0. If 1·(Gn)₂ does not transport or transports cations not as well as the anions, the transporter-mediated influx of anions will have to be accompanied by the influx of H⁺ or efflux of OH⁻, resulting in an increase in intravesicular pH and the fluorescence intensity of the HPTS dye (Fig. 3a). Using this assay and as illustrated in Fig. 3b, among all the tested complexes 1·(Gn)₂ at 2.5 µmol/L, only 1·(G3)₂ displays a fractional ion transport activity value (R_Cl⁻) of 92%, which is much larger than the combined value of 37% (20% from 1 at 2.5 µmol/L and 17% from G3 at 5.0 µmol/L). Coincidentally yet intriguingly, 1·(G3)₂ is the only complex whereas 1 exhibits a positive cooperativity in binding two molecules of G3 (Table 1). After considering the transport activities of 1 and G3 as conservatively estimated in Table S1 (Supporting information), the chloride transport activity of 1·(G3)₂ at 2.5 µmol/L is at least 92%, which is 5.4 times that of G3 at 5.0 µmol/L.

Applying the same HPTS assay in Fig. 3a after replacing 65 mmol/L Na₂SO₄ + 2.5 mmol/L NaX with 67 mmol/L Na₂SO₄ at pH 8.0, we found that 1·(G3)₂ at 2.5 µmol/L elicits no detectable ion transport activity (Fig. S27 in Supporting information). Substituting Na₂SO₄ with K₂SO₄ in both intra- and extravesicular regions leads to the same observation. These findings confirm the inability of 1·(G3)₂ to transport both Na⁺ and K⁺ ions.

Furthermore, enhanced ion transport activities arising from the confined dimeric G3 are also seen when anions such as Br⁻, NO₃⁻ and ClO₄⁻ are examined (Figs. 3c-e). Backed by the same method for conservatively estimating transport activity (Table S2 in Supporting information), the anion-transporting capacity of 1·(G3)₂ increases by 30 folds for Br⁻, 35 folds for NO₃⁻ and 3.4 folds for ClO₄⁻ with respect to those of G3 (Fig. 3c vs. 3d).

Using the well-established Hill analysis, the EC₅₀ value at which 50% ion transport activity is reached was determined to be 0.69, 0.57, 1.30 and 0.74 µmol/L for Cl⁻, Br⁻, NO₃⁻ and ClO₄⁻, respectively (Figs. 3f-h and Fig. S28 in Supporting information).

To evaluate whether the introduction of 1·(G3)₂ maintains membrane integrity, we conducted membrane leakage assays. As shown in Fig. S33 (Supporting information), it was found that membrane-lytic melittin results in efflux of 58% and 99% CF dye from LUVs at 5 and 25 nmol/L, respectively, but incorporation of 1·(G3)₂ at a concentration as high as 2.5 µmol/L produces an only negligible CF efflux of 3%. These comparative studies clearly support well-maintained membrane integrity in the presence of 1·(G3)₂ at high concentrations.

Consistent with the above anion transport studies, ¹H NMR titrations verify the ability of 1·(G3)₂ to bind various anions. Specifically, addition of 10 equiv. Br⁻ ion to the solution of 1·(G3)₂ in CDCl₃ results in significant downfield changes in the chemical shift of H_a and H₁, suggesting the effective binding of Br⁻ anion by 1·(G3)₂. The fact that Br⁻ exerts considerable impact on H_a from G3 (Fig. 2b vs. 2f) but no influence on H₁ from 1 (Fig. 2c vs. 2g) confirms G3 from 1·(G3)₂ to be responsible for the 1·(G3)₂-mediated binding of Br⁻ anions. The additional fact that Br⁻ makes no difference in chemical shift but does make the shape and height different for both H_a and H_b from G3 and 1·(G3)₂ (Fig. 2e vs. 2f) suggests Br⁻ to interact mainly with G3 from 1·(G3)₂. Further investigations show that 1·(G3)₂ can also bind Cl⁻, NO₃⁻ and ClO₄⁻ (Figs. S29-S31 in Supporting information).

To elucidate the structural basis for the anion transport activity of complex 1·(G3)₂, we sought to characterize its solid-state structure. Despite repeated attempts, we were unable to obtain single crystals of the complex. Instead, crystallization invariably resulted in the separation of pure host 1 or guest G3 (CCDC No. 2492337, Fig. 4a). With a stoichiometry 1:2 molar ratio inferred from the solution binding profile, we applied M06-2X/6-31G(d) to optimize complexes of 1·(G3)₂ with and without Cl⁻ or Br⁻ (Figs. 4b-d). The computed structures reveal the formation of a Co²⁺···Co²⁺ bond of 2.63 Å inside the macrocycle-confirmed dimeric metalloporphyrin (Fig. 4b). This results in a tightly packed complex 1·(G3)₂, seemingly leaving no entrance space for anions. Nevertheless, the presence of Cl⁻ and Br⁻ appears to break the Co²⁺···Co²⁺ bond, increasing the Co²⁺···Co²⁺ distances to 4.75 Å and 4.97 Å (Fig. 4c), respectively. Together with the overlapped structures shown in Fig. 4d, these computational findings suggest a highly dynamic nature of the cage structure and its associated complexes as well as dynamic metal-anion interactions, enabling the dynamic formation of the Co²⁺···Co²⁺ and Co²⁺-anion bonds and their subsequent breaking for the anion to bind and release when crossing the membrane as sketched in Fig. 1d.

To summarize, we have successfully demonstrated a previously unexplored strategy toward construction of an artificial anion-transporting system. This was achieved by using a giant macrocycle to bind and constrain two Co²⁺-porphyrin complexes in its large adaptive tubular cavity while leaving sufficiently large, confined space for anions to bind and release through dynamic metal-anion interactions. This strategy effectively converts an uncompetitive transporter Co²⁺-porphyrin into a highly efficient anion-transporting, sandwiched host-guest complex, displaying good EC₅₀ values of 0.57–1.30 µmol/L for Cl⁻, Br⁻, ClO₄⁻ and NO₃⁻ and activity enhancements by up to 35 folds over monomeric Co²⁺-porphyrin. This geometric confinement-based approach might instigate others or find new inspiring uses in designing diverse types of artificial transmembrane transporter systems, possibly benefiting multidisciplinary applications in chemistry, materials sciences, biology and medicines.

Declaration of competing interest

The authors declare that they have no known competing financialinterestsor personal relationships that could have appeared to influence the work reported in this paper.

CRediT authorship contribution statement

Fei Zeng: Writing – original draft, Investigation. Xinrui Pan: Investigation. Zihong Yang: Investigation, Data curation. Jie Shen: Writing – review & editing. Wenju Chang: Writing – review & editing. Huaqiang Zeng: Writing – review & editing, Supervision.

Acknowledgments

This work was supported by the National Natural Science Foundation of China (Nos. 22501049, 21602055, 22271049 and 22371048), Natural Science Foundation of Hunan Province (No. 2023JJ50414), Natural Science Foundation of Fujian Province (No. 2025J08021) and a start-up grant from Fuzhou University.

Supplementary materials

Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.cclet.2025.112355.