Introduction

In the microscopic world, molecules are in stochastic motion driven by thermal energy. Despite this randomness, biological molecular machines such as ribosomes are capable of executing complex functions by precisely controlling relative motion of their components. Inspired by these natural systems, chemists have developed artificial molecular machines based on mechanically interlocked architectures like (pseudo)rotaxanes and catenanes1,2. These architectures enable controlled motion between subcomponents, allowing synthetic systems to perform work at the molecular scale.

Pseudorotaxane dethreading, where the threaded ring moves relative to the axle (Fig. 1), plays a pivotal role in the operation of artificial molecular machines3,4,5,6,7,8,9,10,11. Previous studies have identified key structural factors that influence (pseudo)rotaxane (de)threading12,13,14,15,16,17,18,19,20,21,22,23,24,25,26, including size complementarity between stoppers and macrocycles27,28,29,30,31,32,33,34, structural features of the axle (e.g. length and flexibility)35,36, and electrostatic interactions between components37,38,39,40. While qualitative control over dethreading motion has been demonstrated, achieving predictable and programmable modulation of dethreading kinetics remains challenging. Furthermore, practical applications of pseudorotaxane dethreading—particularly in contexts requiring precise kinetic control, such as drug delivery—are limited24,41. This is primarily due to the lack of a generalizable framework for systematic and programmable tuning of dethreading rates.

Fig. 1: Design of a modular pseudorotaxane platform with programmable dethreading kinetics.
figure 1

An unthreadable stopper (red), crown ether macrocycle (blue), and adjustable stopper (orange) assemble into a pseudorotaxane. The modular design enables tuning of dethreading rates through stopper and macrocycle variations.

Full size image

Here, we report a modular pseudorotaxane platform that enables programmable control over dethreading kinetics through rational component engineering (Fig. 1). The system, comprising 24-crown-8-based macrocycles and benzylic amine stoppers, is assembled via a metal-free active template strategy developed by Leigh and co-workers42,43,44. Systematic variation of these components enables precise tuning of dethreading rates. Through X-ray crystallography and DFT calculations, we elucidate the dethreading pathway and reveal structure-kinetic relationships. To demonstrate the translational potential of this system, we engineered camptothecin-conjugated pseudorotaxanes. Their release kinetics and corresponding cytotoxicity profiles were programmed through component variation. Finally, we derived general guidelines for the rational selection of pseudorotaxane components to tailor dethreading behavior. This work establishes a versatile platform for kinetic regulation in interlocked systems and advances the development of drug delivery systems and molecular machines.

Results

Design and synthesis of pseudorotaxanes

The investigated pseudorotaxanes (Fig. 1) incorporate the following key components: 1) Unthreadable stopper: a trityl group was used as the unthreadable stopper due to its bulkiness, preventing the crown ether from slipping off on that side. 2) Size adjustable stopper: benzylic amines were used as adjustable stoppers. Their steric profiles can be finely tuned by varying alkyl substituents on the oxygen atom. Additionally, these stoppers are compatible with a metal-free active template strategy, enabling rapid and modular pseudorotaxane synthesis43,44. 3) Crown ether macrocycle: the macrocycles include 24-crown-8 ether, monobenzo-24-crown-8 ether, and dibenzo-24-crown-8 ether. Crown ethers were selected due to the lack of systematic studies on their effect in rotaxane dethreading. Addressing this gap could guide the design of crown ether-based artificial molecular machines and broaden the applicability of our platform45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62.

Based on this modular design, a library of 11 pseudorotaxanes (designated as ROT1–11) was synthesized (Fig. 2). The resulting interlocked architectures, incorporating various combinations of adjustable stoppers and crown ether macrocycles, were confirmed by high-resolution mass spectrometry, NMR spectroscopy, and X-ray crystallography (see Supporting Information Section 10).

Fig. 2: Component investigation to tune dethreading rate.
figure 2

a Structures of the investigated macrocycle components. b Structures of the investigated pseudorotaxanes.

Full size image

Dethreading Kinetics

With pseudorotaxanes ROT1–11 in hand, dethreading kinetics were measured with 1H-NMR at specific temperatures in deuterated dimethyl sulfoxide (DMSO-d6). The activation Gibbs free energies (ΔG‡) were determined using the Eyring equation. The results are summarized in Table 1.

Table 1 Rate constants, half-lives and activation Gibbs free energies for the dethreading of the pseudorotaxanes ROT1–11 in DMSO-d6
Full size table

The macrocycle significantly influences the dethreading rate. As the macrocycle is systematically varied from 24-crown-8 to benzo-24-crown-8 and further to dibenzo-24-crown-8, higher temperatures are required to achieve comparable half-lives. For instance, ROT1 has a half-life of 0.75 h at 294 K (Table 1, entry 1). Compound ROT4, which features the same stopper but incorporates benzo-24-crown-8 as the macrocycle, displays a half-life of 1.07 h at 333 K (Table 1, entry 4). In contrast, ROT8, bearing dibenzo-24-crown-8 ether as the macrocycle, exhibits a half-life of 1.44 h at 373 K (Table 1, entry 8).

When the macrocycle is fixed, the stopper fine-tunes the dethreading rate. For pseudorotaxanes with the same macrocycle, the half-lives follow the order: pseudorotaxane with methyl group <pseudorotaxane with propargyl group <pseudorotaxane with allyl group. For example, within the 24-crown-8 series, the half-life increases with the steric bulk of the stopper, from 0.75 h for ROT1 to 1.9 h for ROT2 and then 27 h for ROT3 (Table 1, entries 1–3).

Substituents with similar electronic property on the macrocycle benzene rings exert a negligible influence on the dethreading rate. ROT6, bearing a bulky tert-butylcarbamoyl group (Table 1, entry 6), exhibits a half-life comparable to that of ROT5, which contains a methoxycarbonyl group (Table 1, entry 5). A similar trend is observed for ROT9 and ROT10 (Table 1, entries 9 and 10). These results suggest that meta-substituents on the aromatic ring of the macrocycle are well tolerated and do not significantly affect the dethreading rate, thereby enabling predictable kinetics and offering flexibility for further derivatization.

Noncovalent interactions analysis

To better understand the structural features and noncovalent interactions between the axle and the macrocycle, we analyzed the single-crystal X-ray structure of ROT9. In the solid state, the dibenzo-24-crown-8 macrocycle is positioned over the amide group and adopts an elliptical shape that snugly accommodates the central part of the axle. The amide hydrogen atom forms a hydrogen bond with an oxygen atom in the macrocycle (Fig. 3a). Close contacts were observed between the carbonyl oxygen and the C–H groups of the macrocycle (Fig. 3a), as well as between the macrocycle oxygens and both the benzylic –CH2– group and the aromatic C–H groups (Fig. 3b). The macrocycle adopts a boat conformation, with its two aryl rings oriented toward the opposite ends of the axle. The benzene ring of the macrocycle and that of the stopper have a centroid-to-centroid distance of 4.31 Å and an interplanar angle of 28.1° (Fig. 3c). Within the adjustable stopper, the smaller methyl group is oriented toward the ester group on the macrocycle, whereas the bulkier propargyl group is oriented away to minimize steric repulsion (Fig. 3a).

Fig. 3: Analysis of ROT9 by X-ray diffraction and 1H-NMR.
figure 3

a Side view of ROT9 showing intercomponent interactions (in green). Distances: N1(–H)···O2, 3.09 Å; C8(–Hh)···O14, 3.61 Å; C9(–Hg)···O14, 3.67 Å. Bond angles: N1–H···O2, 167.0°; C8–Hh···O14, 150.5°; C9–Hg···O14, 146.8°. b Side view of ROT9 showing intercomponent interactions (in green). Distances: C37(–H5)···O1, 3.41 Å; C37(–H5)···O8, 3.28 Å; C41(–H6)···O4, 3.26 Å; C41(–H6)···O4A, 3.68 Å; C41(–H6)···O6, 3.26 Å. Bond angles: C37–H5···O1, 149.5°; C37–H5···O8, 148.1°; C41–H6···O4, 142.5°; C41–H6···O4A, 164.1°; C41–H6···O6, 141.7°. c Side view of ROT9 showing weak interactions between the macrocycle’s aryl ring and the adjustable stopper’s benzene ring. centroid-to-centroid distance, 4.31 Å, dihedral angle, 28.1°. Carbon, gray; hydrogen, ivory; chlorine, green; nitrogen, light purple; oxygen, red. Solvent molecules and other hydrogen atoms omitted for clarity. Thermal ellipsoids (atomic displacement parameters) are depicted at the 50% probability level. d Partial 1H NMR spectra (400 MHz, CDCl3, 296 K) of macrocycle M4 (top, 20 mM), pseudorotaxane ROT9 (middle, 20 mM) and the corresponding unthreaded axle ROT9-Axle (bottom, 30 mM).

Full size image

To investigate the interaction between the axle and the macrocycle in solution, we compared the 1H-NMR spectra of macrocycle M4, pseudorotaxane ROT9, and the corresponding axle (Fig. 3d and Supporting Information). Notably, the homotopic geminal protons of the crown ether in ROT9 were desymmetrized by the interaction with the axle. Together with characteristic chemical shift changes observed for multiple protons on both the axle and macrocycle, these findings support the existence of intercomponent interactions in solution. Further analysis of the chemical shift variations for the N–H and H1–H6 protons revealed a progressive decrease in chemical shift changes from N–H to H6 and then to H1. This decreasing trend indicates that the macrocycle resides closer to the amide part rather than the terminal part of the axle. To further elucidate the position of the macrocycle on the axle, ROESY spectroscopy was employed (see Supporting Information, page S110). Observed ROESY signals between H5–H7 and Hf –Hh confirmed the macrocycle’s proximity to the amide group. Conversely, the absence of ROESY signals between the macrocycle and the terminal groups of the adjustable stopper, as well as the trityl group of the unthreadable stopper, indicates their spatial separation.

Computational studies

Although the dethreading process appears deceptively simple with the cartoon depictions of a ring slipping off a rod, extensive computational studies reveal that the actual mechanism involves intricate conformational changes and complex structural preferences at different stages of the dethreading processes.

Before discussing the reaction pathway, we emphasize the necessity of careful conformational search (Fig. 4). Both the axle and the 24-membered macrocycle exhibit considerable conformational flexibility. The macrocycle can adapt its shape to accommodate the substructures passing through it (see discussed later in Fig. 5). The six aryl groups—four on the axle and two on the macrocycle—can engage in either steric repulsion or attractive dispersion interactions in different conformers. Notably, the terminal triaryl group on the axle remains in contact with the macrocycle’s aromatic rings even when the axle has mostly dethreaded. Failing to identify the most stable conformer, or assuming that the optimal conformer in the previous stage of the pathway persists in subsequent steps, would significantly overestimate the activation Gibbs free energy.

Fig. 4: Conformational search of the dethreading transition state of ROT8.
figure 4

a Coverage for a conformational search targeting the dethreading transition state of ROT8. The image overlaps 10% of the investigated conformers (randomly picked). The thicker bonds belong to the macrocycle, and the thinner ones, the axle. Hydrogen atoms are hidden for clarity. The conformers are colored according to their serial number. b A typical example of significant energy difference between conformers for the key dethreading transition state of ROT8. For the conformer candidates, solvated Gibbs free energies were calculated at ωB97X-D/def2-SV(P) level.

Full size image
Fig. 5: Solvated Gibbs free energy profile for the dethreading process of ROT9 at ωB97X-D/def2-TZVP//def2-SV(P) level.
figure 5

For each structure, the top illustration shows the side-on view of the complex, and the lower illustration shows the head-on view of the macrocycle itself with the axle hidden for clarity. The macrocycle is colored based on the distance of each atom to the best-fit plane of the 24-membered-ring in a rainbow color scale, with red signifies the atom is far away from the plane towards the tail end of the axle. Arrows depicts the direction of key geometric changes. The 4 plots on the bottom show the electron density (darker red-green-blue colormap with red/blue corresponding to 0–0.3 a.u.), and reduced density gradient map (brighter rainbow colormap with the blue to red color scale corresponding to 0.15–0.65 a.u.), which shows the nonbonding interactions, in the best-fit plane of the 24-membered-ring66,67. SXRD: single crystal X-ray diffraction.

Full size image

To address this, extensive conformational searches were conducted for each key transition state and intermediate. As a representative case, Fig. 4a illustrates the coverage of conformers examined to approach the global minimum structure for TS8-ROT8. The importance of this effort was demonstrated in Fig. 4b, where two conformers TS8-ROT8-722 and TS8-ROT8-1532—differing only in the dihedral angle around the amide bond—exhibited a Gibbs free energy difference of 7.8 kcal/mol. This difference far exceeds the typical error associated with computational level selection, as well as the experimental differences in effective ΔG‡ between the compounds of interest (ROT8, ROT9 and ROT11), making the effort indispensable. Intuitively, one might expect TS8-ROT8-722 to have a higher energy than TS8-ROT8-1532, due to the axle distortion caused by steric repulsion between the triaryl end of the axle and the aryl ring of the macrocycle. However, TS8-ROT8-722 is actually significantly lower in Gibbs free energy than its less contorted counterpart, TS8-ROT8-1532. This counterintuitive outcome can be attributed to dispersion interactions between the aromatic rings, which help maintain a favorable alignment between the axle dimethyl group and the macrocycle, thereby guiding the axle through the macrocycle along a lower-resistance dethreading trajectory.

The optimal dethreading pathway of ROT9 after extensive conformational search was located at ωB97X-D/def2-TZVP//def2-SV(P) level63,64 with SMD solvation in DMSO65 (Fig. 5). For clarity, we will refer to the triaryl end of the axle as the “tail”, and the opposite end as the “tip”. The process begins with determining the most stable conformation of the substrate complex in solution, which serves as the reference point for the Gibbs free energy. The conformer extracted from the single crystal structure (Sub, major component in the disordered SXRD structure) was found to be 2.6 kcal/mol higher in Gibbs free energy in solution than the conformer IM1, differing primarily in the C-O4-C-C(-O5) dihedral angle. Such a discrepancy is not unexpected: for relatively flexible molecules, the optimal conformation in solid state often differs from that in solution due to differences in the optimal intermolecular interactions required for periodic packing versus solvation. The disorder in the crystal structure also occurs near this region (Figs. 3a–3c).

Starting from IM1 in the solvated state, a series of dihedral angle changes (TS2 … TS3) in the bottom part of the macrocycle leads to a metastable conformer, IM3, in which the axle moves towards the dethreading direction while the tail end tilts towards aryl ring B of the macrocycle. This induces the flipping of ring B from the tail direction to the tip direction (TS4), bringing it to the same side of the macrocycle as ring A. This conformational rearrangement resembles a “syn-anti flip” and is crucial for the benzene linker in the axle to dethread from the macrocycle. In the “anti” conformation, π-π interaction stabilizes the close contact between the benzene linker in the axle and ring A in the macrocycle (see those in TS3, Fig. 5). The “syn-anti flip” replaces the π-π interaction with two sets of edge-on C-H…π interactions involving both ring A and ring B (as seen in TS5, Fig. 5), allowing for the benzene linker in the axle to dethread from the macrocycle (TS7). In addition, to accommodate the flat intersection profile of the exiting benzene ring, the “round/square” shaped macrocycle (see TS5) also needs to morph into an “oval” shape (see TS7). This transformation is enabled by the O3-C-C-O4 dihedral angle change in TS5. The conformational shift is clearly visualized in the cross-sections of the electron density and RDG66,67 maps.

For the departure of the stopper at the tip, four possible dethreading profiles have been postulated34: (1) simultaneous passage of both methyl groups and alkoxy group through the macrocycle cavity; (2) simultaneous passage of the two methyl groups, followed by sequential slippage of the alkoxy group; (3) slippage of a single methyl group first, followed by the simultaneous transit of the remaining methyl and alkoxy groups; or (4) a stepwise process involving the consecutive passage of each group individually. Among all identified conformers for the transition state and the corresponding intrinsic reaction coordinate pathways, only scenarios (1) and (2) exist, with scenario (2) overwhelmingly dominant for ROT9 and ROT11. In the few cases resembling scenario (1), the motion exhibits significant asynchrony and is more accurately regarded as a special case of scenario (2), where the two methyl groups pass through a barriered transition, followed by a nearly barrierless slippage of the alkoxy group.

Similar to what has been discussed in Fig. 4b, the transition state TS7-ROT9 features a rotation of the amide bond that induces a bending of the axle into a hook-like shape and causes ring A of the macrocycle to flip towards the axle’s tail. This conformational change enables the system to adopt the preferred geometry (TS8) with a roughly round-shaped macrocycle for the exit of the gem-dimethyl group. TS8 is the rate-determining transition state in the overall dethreading process, with a calculated activation Gibbs free energy of 28.8 kcal/mol. The electron density and RDG maps clearly show that the gem-methyl groups go through the ring in an asynchronous manner, with the upper methyl group passing first. Following this step, the exit of the alkyne group requires only a low barrier of 1.0 kcal/mol, resulting in the formation of the detached intermediate IM9. After solvation and relaxation, the combined Gibbs free energy of the optimal product structures (Prod1 and Prod2) is 4.8 kcal/mol lower than that of the most stable complex (IM1), making the dethreading process both kinetically and thermodynamically irreversible.

The key gem-dimethyl dethreading process was also studied for ROT8 and ROT11. Compared to the 28.8 kcal/mol barrier observed for ROT9, the corresponding transition state TS8 for ROT8 (TS8-ROT8-722, Fig. 4b) requires a slightly lower activation Gibbs free energy of 27.5 kcal/mol. In contrast, a similar transition state for ROT11 (TS8-ROT11, Fig. 6) presents a higher activation Gibbs free energy of 29.7 kcal/mol. These relative differences align closely with the experimental trends observed (Table 1). The lower activation Gibbs free energy required for ROT8 could be attributed to the fact that the -OMe group on the axle of ROT8 is short in comparison to the -O-propargyl in ROT9 or the -O-allyl group in ROT11, such that it does not offer additional interaction with the macrocycle. The final dethreading behavior of ROT8 is significantly different from ROT9 or ROT11. The dethreading of the dimethyl group and the -OR group is concerted for ROT8 (i.e., the equivalent process involving TS9 is barrierless for ROT8), while the same process is stepwise for ROT9 and ROT11. This further confirms the absence of further interaction between the -OMe group and the macrocycle. For a comparison between TS8-ROT9 and TS8-ROT11, wavefunction analysis indicates that the slightly increased barrier of ca. 1 kcal/mol for the latter could be the secondary effect of the interaction between the -O-allyl group and the closely situated hydrogen atoms on the macrocycle (shown in Fig. 6), which causes both a rotation along the axle and a slight upward tilt of the tip end of the axle, increasing the steric repulsion between the two dimethyl groups and the macrocycle. This can be seen in both the geometry (top figures, Fig. 6) and the RDG cross section.

Fig. 6: Comparison between the key transition state (TS8) for the dethreading pathway of ROT9 (left) and ROT11 (right).
figure 6

The top illustration shows the structures around the macrocycles (side-on view). Atoms other than the skeletal atoms and the interaction-relevant hydrogen atoms were hidden for visual clarity. The reference plains (shown in transparent yellow) are the best-fit planes of the macrocycle. The bottom plots show the electron density (darker red-green-blue colormap with red/blue corresponding to 0–0.3 a.u.) and RDG map (brighter rainbow colormap with the blue to red color scale corresponding to 0.15–0.65 a.u.) which shows the weak interactions, in the best-fit plane of the 24-membered-ring66,67.

Full size image

Drug release

Building on our understanding of the system’s dethreading kinetics, we explored its potential for modulating the release rate of functional molecules. As a proof-of-concept, camptothecin (CPT), a potent antineoplastic agent that exerts its effect by inhibiting DNA topoisomerase I, was selected as a model payload. In the designed CPT pseudorotaxanes, the payload CPT is connected to adjustable stoppers via a self-immolative linker68,69. Upon dethreading, this linker undergoes spontaneous cleavage, releasing CPT (Fig. 7a).

Fig. 7: Controlled release of CPT via adjustable molecular dethreading.
figure 7

a Reaction scheme. b Stacked UPLC-MS selected ion monitoring (SIM) traces of CPT-Axle3 self-immolation reaction (DMSO:PBS buffer = 1:2). Blue traces: SIM of m/Z = 636.3 for CPT-Axle3 (M-AllylO-). Yellow traces: SIM of m/Z = 349.1 for CPT (M + H+). Red traces: SIM of m/Z = 367.2 for compound 12 (M+Na+). From lighter colored traces to darker traces: reaction mixture at 0 min, 15 min and 30 min, respectively. c Rate constants, half-lives and activation Gibbs free energies for the dethreading of the CPT pseudorotaxanes in DMSO-d6 at 37 °C. ΔG‡ calculated using the Eyring equation at 310 K. See Supporting Information Section 6 for uncertainties of t1/2 and ΔG‡. d IC50 ± standard error of the mean (μM) for the cytotoxicity of CPT, CPT-Axle1–3 and CPT-ROT1–3 in A549 cells, averaged from three independent experiments. e–g Comparison of dose-response curves of A549 cells treated with (e) CPT, CPT-Axle1, and CPT-ROT1, (f) CPT, CPT-Axle2, and CPT-ROT2, and (g) CPT, CPT-Axle3, and CPT-ROT3. Each condition was tested in triplicate per experiment.

Full size image

Tuning the dethreading rate of CPT pseudorotaxanes can be achieved by varying the -OR groups on the adjustable stoppers or the fused aromatic ring on the crown ether. The dethreading kinetics were measured at physiological temperature 37 °C. CPT-ROT1, with -OMe group on the stopper, displayed the fastest dethreading rate with a half-life of 3.2 h (Fig. 7c, entry 1). When the -OMe is changed to -O-propargyl (CPT-ROT2) or -O-allyl group (CPT-ROT3), the dethreading half-life extends to 7.7 h and 59 h respectively (Fig. 7c, entries 2 and 3). Changing the macrocycle from 24-crown-8 ether to monobenzo-24-crown-8 ether significantly increases the dethreading half-life. For example, CPT-ROT4 has a half-life of 35.6 h, about 10 times longer than its counterpart CPT-ROT1 (Fig. 7c, entry 4). Further modification of CPT-ROT4 by replacing the methyl group with propargyl group extends the half-life to 91 h (Fig. 7c, entry 5). By modular engineering of the components, we obtained a series of CPT pseudorotaxanes with dethreading half-lives ranging from a few hours to a couple of days.

To ensure efficient payload release following pseudorotaxane disassembly, a glycine-succinate tether was incorporated into the axle (Fig. 7a). This motif has been reported to undergo rapid intramolecular transacylation under physiological pH70. The kinetics of self-immolation of CPT-Axle3 in PBS buffer (pH 7.4), which releases CPT and generates transacylation byproduct pyrrolidine-2,5-dione (12), were monitored by LC-MS over a 30-minute period (Fig. 7b). Compound 12 was identified by NMR spectroscopy (see Supporting Information, Figure S36).

The cytotoxicity of representative CPT pseudorotaxanes and their corresponding CPT axles were evaluated against the A549 lung cancer cell line. As shown in Fig. 7e–g, the cytotoxicity profiles of CPT-Axle1–3 and free CPT suggest rapid release of the active drug from the self-immolative linker under biological conditions. In contrast, the interlocked CPT molecules exhibited reduced cytotoxicity compared to their axle counterparts, indicating that the mechanical bond effectively cages the drug and slows its release. Among CPT-ROT1–3, cytotoxicity followed the trend CPT-ROT1 > CPT-ROT2 > CPT-ROT3, consistent with their respective dethreading rates (Fig. 7d). Notably, the difference in IC50 values between each CPT-ROT and its corresponding CPT-Axle increased as the dethreading rate decreased. Although CPT-Axle1 and CPT-Axle3 exhibited comparable cytotoxicity, CPT-ROT1 showed an IC50 value approximately three times that of CPT-Axle1, whereas the IC50 of CPT-ROT3 is 17 times that of CPT-Axle3. This difference aligns with the significantly slower dethreading rate of CPT-ROT3, which was reduced to approximately 1/18 of that of CPT-ROT1. By tuning the dethreading kinetics of CPT pseudorotaxanes, we achieved predictable modulation of their cytotoxicity in A549 lung cancer cells.

Discussion

To guide further applications and studies of crown ether-based (pseudo)rotaxane systems, we summarize the factors influencing dethreading as follows: (i) Benzo-fused substitution on the macrocycle has the most significant impact on the dethreading rate. In the dibenzo-24-crown-8 ether systems, fused 1,2-benzo substituents on the macrocycle have a significant role in both limiting the conformational change of the macrocycle, as well as interacting with the axle in different stages of the dethreading process. Since dethreading requires a series of conformational changes in the macrocycle to accommodate the passing groups along the axle, the rigidity caused by the benzo substitution reduces the macrocycle’s flexibility and hinders these conformational changes, which slow down the dethreading process15. (ii) Adjusting the stopper size typically fine-tunes the dethreading kinetics. Variations in the alkoxy substituent on the adjustable stopper led to systematic modulation of the dethreading kinetics. Contrary to the intuitive expectation that a larger or more rigid group directly interacts with the macrocycle to affect the dethreading rate, our findings indicate that the rate modulation arises from conformational changes induced by the stopper substituents. The rate-determining step involves the slip-off of the dimethyl groups. The slip-off of the alkoxy groups is either low barrier (for -O-propargyl and -O-allyl groups) or barrierless (for the -OMe group). (iii) Substituents with similar electronic properties on the benzene ring(s) of the macrocycles have a negligible influence on the dethreading process. In summary, we have established a modular pseudorotaxane platform that enables programmable control over molecular dethreading kinetics. Combined crystallographic and computational analyses provide mechanistic insight into the dethreading pathway and structure-kinetic relationship. This understanding was translated into a tunable drug delivery platform, where camptothecin release kinetics and corresponding anticancer activity directly correlated with the engineered dethreading rates. These findings lay the groundwork for mechanically interlocked architectures as programmable release systems. This work integrates synthetic design, kinetic study, supramolecular characterization, computational modeling, and biological application to provide a blueprint for developing advanced molecular machines with biomedical functionality.

Methods

Synthesis of pseudorotaxanes

The synthesis and characterization of the pseudorotaxanes and building blocks are provided in the Supplementary Information. Representative synthesis of pseudorotaxane ROT1: To a stirring solution of S2 (0.13 mmol, 23 mg, 2.0 eq.) and M1 (0.065 mmol, 23 mg, 1.0 eq.) in toluene (600 µL) was added S1 (0.13 mmol, 42 mg, 2.0 eq.). The mixture was stirred at rt overnight. The reaction mixture was purified with preparative TLC (SiO2, Hexane/EtOAc 4:1 then Hexane/EtOAc 1:1) to afford ROT1 (30.9 mg, 52%) as a colorless solid.

Dethreading experiment of pseudorotaxanes

Kinetic measurements were performed to determine the dethreading rate constants in DMSO-d6. The rate constants k for the dethreading reactions of pseudorotaxanes at specific temperatures were obtained by tracking the decrease in selected proton resonances using 1H-NMR spectroscopy. The half-lives were calculated based on this data. The activation free energy ΔG‡ for each process was calculated using the Eyring equation.

Computational levels

The conformational search candidates for the minima and first order saddle points were first evaluated at GFN2-xTB level, and then ranked at ωB97X-D/def2-SV(P) level. The structures in the optimal pathways were further evaluated at ωB97X-D/def2-TZVP//def2-SV(P) level. SMD solvation was included both during the optimization process and single point energy calculations. The critical points were checked to have near-zero gradients, and zero (for the minima) or one (for the transition states) imaginary frequencies evaluated at the level of optimization. The conformational searches are either fully relaxed to locate the global minima (e.g., IM1) or encouraged by an energy penalty to explore different regions of a collective variable, until it is completely dethreaded. The collective variable was defined by the dot products of two vectors: the first one is a unit vector originated from the centroid of the 24 atoms on the macrocycle, with a direction aligned with the normal vector for the best-fit plane of the 24 atoms pointing towards the tail of the axle; the second vector also originated from the centroid of the 24 atoms on the macrocycle, but ends at one of the several key atoms on the axle (including the CAr3 carbon atom, the nitrogen atom, the two benzylic atoms, and the carbon atom of the CH2 or CH3 on the OR substituent) depending on the dethreading process.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.