Synergistic regulation of nonbinary molecular switches by protonation and light

Significance

The work reported here could allow for better control over the storage and processing of information at the molecular level, as well as an increased understanding of the determinants that regulate the function of molecular switches and logic devices. We report switchable molecular shuttles that are regulated synergistically by both protonation and light. This allows hierarchical control over structure in both a kinetic and thermodynamic sense. The shuttles were constructed using azobenzene-4,4′-dicarboxyate dianions 2 ²⁻ or 4,4′-stilbenedicarboxyate dianions 3 ²⁻ as the threading elements and a tetracationic macrocycle, cyclo[2](2,6-di(1H-imidazol-1-yl)pyridine)[2](1,4-di-methylenebenzene) (referred as the Texas-sized molecular box; 1 ⁴⁺), as the encircling wheel. This present dual input strategy provides a complement to more traditional orthogonal stimulus-based approaches to molecular switching.

Abstract

We report a molecular switching ensemble whose states may be regulated in synergistic fashion by both protonation and photoirradiation. This allows hierarchical control in both a kinetic and thermodynamic sense. These pseudorotaxane-based molecular devices exploit the so-called Texas-sized molecular box (cyclo[2]-(2,6-di(1H-imidazol-1-yl)pyridine)[2](1,4-dimethylenebenzene); 1⁴⁺, studied as its tetrakis-PF₆ ⁻ salt) as the wheel component. Anions of azobenzene-4,4′-dicarboxylic acid (2H⁺•2) or 4,4′-stilbenedicarboxylic acid (2H⁺•3) serve as the threading rod elements. The various forms of 2 and 3 (neutral, monoprotonated, and diprotonated) interact differently with 1⁴⁺, as do the photoinduced cis or trans forms of these classic photoactive guests. The net result is a multimodal molecular switch that can be regulated in synergistic fashion through protonation/deprotonation and photoirradiation. The degree of guest protonation is the dominating control factor, with light acting as a secondary regulatory stimulus. The present dual input strategy provides a complement to more traditional orthogonal stimulus-based approaches to molecular switching and allows for the creation of nonbinary stimulus-responsive functional materials.

Multifactor regulation of biomolecular machines is essential to their ability to carry out various biological functions (1 –11). Construction of artificial molecular devices with multifactor regulation features may allow us to understand and simulate biological systems more effectively (12 –31). However, creating and controlling such synthetic constructs remains challenging (16, 32 –37). Most known systems involving multifactor regulation, including most so-called molecular switches and logic devices (38 –43), have been predicated on an orthogonal strategy wherein the different control factors that determine the distribution of accessible states do not affect one another (20, 44 –56). However, in principle, a greater level of control can be achieved by using two separate regulatory inputs that operate in synergistic fashion. Ideally, this could lead to hierarchical control where different states are specifically accessed by means of appropriately selected nonorthogonal inputs. However, to our knowledge, only a limited number of reports detailing controlled hierarchical systems have appeared (57). Furthermore, the balance between specific effects (e.g., kinetics vs. thermodynamics) under conditions of stimulus regulation is still far from fully understood (54). There is thus a need for new systems that can provide further insights into the underlying design determinants. Here we report a set of pseudorotaxane molecular shuttles that act as multimodal chemical switches subject to hierarchical control.

Results and Discussion

The devices reported here exploit the so-called Texas-sized molecular box (cyclo[2]-(2,6-di(1H-imidazol-1-yl)pyridine)[2](1,4-dimethylenebenzene) (1 ⁴⁺; studied as its tetrakis-PF₆ ⁻ salt) as the wheel component (58). The anionic species of azobenzene-4,4′-dicarboxylic acid (2H⁺•2) or 4,4′-stilbenedicarboxylic acid (2H⁺•3) serve as the threading elements. Both guests are subject to photoinduced interconversion between their limiting cis and trans configurations, and both can exist as their respective neutral diacid, monoanionic singly protonated, and dianionic doubly deprotonated forms. These species interact differently with 1 ⁴⁺ to form either threaded pseudorotaxane structures, externally bound complexes, or nonbound host–guest mixtures (Fig. 1). Thus, the combination of receptor 1 ⁴⁺ and the various forms of 2 and 3 (neutral, monoprotonated, and diprotonated) constitutes a nonbinary molecular switch that can be regulated in synergistic fashion through protonation/deprotonation and photoirradiation. Both inputs can be used to regulate the distribution of the linear anionic guest inside or outside the cavity of 1 ⁴⁺ and thus to tune the equilibrium balance of states and the speed with which it is obtained (Fig. 1A ). As a general rule, guest protonation is the dominating control factor, with light acting as a secondary regulatory stimulus.

Fig. 1.

(A) Schematic representation of the synergistic protonation- and light-based regulation of the pseudorotaxane switch considered in this study. (B) Also shown are the chemical structures of receptor 1 ⁴⁺ and guests 2H⁺•2 2H⁺•3, (H⁺•2)⁻, (H⁺•3)⁻, 2 ²⁻, and 3 ²⁻. For series 2, X = N whereas, for series 3, X = CH. Note that a solid green ball is used to show the protonated carboxylic acid forms of the guest, whereas a green ball outlined with a dashed line is meant to denote a site that is either protonated or deprotonated, and an empty cleft represents the fully deprotonated carboxylate form.

As noted above, understanding the factors that regulate the switching process might allow for better insights into these switching events. Therefore, various guest forms (acid, monoanion, and dianion obtained via protonation/deprotonation) were subject to photoirradiation, and then compared with similar photoisomerization processes in the presence of 1 ⁴⁺.

In our initial study, the monoanions (H⁺•2)⁻ and (H⁺•3)⁻ and the corresponding dianions 2 ²⁻ and 3 ²⁻ (Fig. 1B ) were subject to photoirradiation in the absence of 1 ⁴⁺. A 100-W LED 365 nm light source (radiant flux = 5.1 W in all tests unless otherwise noted) was used to illuminate 1 mM solutions of 2H⁺•2 or 2H⁺•3. Conversion was monitored by ¹H NMR spectroscopy. Note that all NMR spectroscopic studies were performed in dimethyl sulfoxide (DMSO)-d ₆. The monoanions and dianions were obtained by adding one or two molar equivalents of TMA⁺ •OH⁻ •5H₂O, respectively, to solutions of 2H⁺•2 or 2H⁺•3.

The photoisomerization behavior of 2H⁺•2, 2H⁺•3, (H⁺•2)⁻, (H⁺•3)⁻, 2 ²⁻, or 3 ²⁻ were found to differ. The reactions were thus analyzed in terms of opposing first-order isomerization processes (59). The elementary reactions were subject to a global kinetic analysis ( SI Appendix, Figs. S2, S4, S6, S8, S10, and S12). This allowed a relationship between y' (the percent of cis-isomer after photoirradiation with 365 nm light) and k ₁ (apparent forward rate constant) to be established,

y′=−1.0×103c2′e−1.0×10−3c2′k1t+1.0×103c2′, [1]

where c′ ₂ is the concentration of the cis form of 2H⁺•2, 2H⁺•3, (H⁺•2)⁻, (H⁺•3)⁻, 2 ²⁻, or 3 ²⁻. The apparent rate constants (k) were calculated for each species in question (Table 1). It was found that the k ₁ values for 2H⁺ •2, (H⁺•2)⁻, and 2 ²⁻, 0.34(1), 1.7(7), 3.2(5) (× 10⁻² s⁻¹), respectively, were lower than the corresponding values for 2H⁺ •3, (H⁺•3)⁻, and 3 ²⁻ (k ₁ = 1.1(5), 4.9(3), 4.8(1) (× 10⁻² s⁻¹), respectively). These values thus lend support to the notion that deprotonation of 2H⁺•2 or 2H⁺•3 serves to increase the rate of photoisomerization (Fig. 2).

Table 1.

Conversion efficiencies and apparent rate constants of the photoisomerization processes of 2H⁺•2, 2H⁺•3, (H⁺•2)⁻, (H⁺•3)⁻, 2 ²⁻, or 3 ²⁻

Guest	ym * (%)	ys * (%)	k 1 † (10−2 s−1)	k -1 † (10−2 s−1)	R 2 ‡
2H+•2	29	29	0.34(1)	0.81(6)	0.993
(H+•2)−	92	95	1.7(7)	0.10(5)	0.990
2 2−	93	96	3.2(5)	0.11(9)	0.997
2H+•3	38	39	1.1(5)	1.7(5)	0.997
(H+•3)−	77	77	4.9(3)	1.4(9)	0.996
3 2−	78	79	4.8(1)	1.2(7)	0.998

*Conversion efficiencies for the trans → cis photoisomerization process.

^†The k ₁ and k _{− 1} represent the apparent forward and reverse rate constants, respectively.

^‡ R ² corresponds to the adjusted R-square value.

Fig. 2.

Relationship between the degree of isomerization and illumination time at 365 nm for 2H⁺•2, 2H⁺•3, (H⁺•2)⁻, (H⁺•3)⁻, 2 ²⁻, and 3 ²⁻.

The asymptotic photoconversion yields y(%) corresponding to the trans → cis isomerization of each azobenzene or stillbene species were monitored. (Note that y _s(%) refers to the simulated value derived from Eq. 1 , whereas y _m refers to the experimental values measured via ¹H NMR spectroscopy.) The simulated value (y _s) was found to agree well with the measured value (y _m) (Table 1), although, as a rule, the y _m values proved slightly smaller than the corresponding y _s values. This discrepancy may reflect limitations in the kinetic model. In the case of 2H⁺ •2, (H⁺•2)⁻, and 2 ²⁻, the y(%) values were found to be 29%, 95%, and 96%, respectively, while y(%) values of 39%, 77%, and 79% were found for 2H⁺ •3, (H⁺•3)⁻ and 3 ²⁻, respectively (Table 1). That the y(%) values increase as a function of deprotonation leads us to propose that there is a competition between photoinitiated proton transfer and photoisomerization and that this competition, reflected in the k ₁ and y(%) values, becomes more intense as the degree of protonation increases (60).

The above results led us to consider that it might be possible to construct nonbinary molecular switches based on 2 ²⁻ and 3 ²⁻ that could be regulated by both the degree of protonation and exposure to light. To test this hypothesis, macrocycle 1⁴⁺ , a receptor known to form interpenetrated structures with linear aromatic anions (61, 62), was tested in conjunction with 2H⁺•2 and 2H⁺•3 and their anionic forms. It was expected that the choice of substrate and the degree protonation (2H⁺•2, 2H⁺•3, (H⁺•2), (H⁺•3)⁻, 2 ²⁻, or 3 ²⁻), the cis vs. trans configuration, and the degree of threading would all define the state of the system and that the latter could be controlled via the synergistic application of light and chemical-based control over the degree of protonation.

Prior to commencing studies of the putative switching effects, ¹H NMR spectroscopic analyses were carried out in an effort to assess the interaction between 2H⁺•2, 2H⁺•3, (H⁺•2)⁻, (H⁺•3)⁻, 2 ²⁻, or 3 ²⁻ and 1 ⁴⁺. Mixing a 1 mM solution of 2H⁺•2 or 2H⁺•3 with one molar equivalent of 1 ⁴⁺ in solution gave rise to no obvious chemical shift change ( SI Appendix, Figs. S13 and S16). This finding is interpreted in terms of there being little or no complexation between either 2H⁺•2 or 2H⁺•3 and 1 ⁴⁺ in DMSO-d ₆.

In contrast, distinct chemical shift changes for proton signals of (H⁺•2)⁻, (H⁺•3)⁻, 2 ²⁻, or 3 ²⁻ and 1 ⁴⁺ were observed after mixing any of these four anionic guests and 1 ⁴⁺ in a 1:1 ratio ( SI Appendix, Figs. S14, S15, S17, and S18). These findings were taken as evidence of host–guest complexation. ¹H NMR spectroscopic titrations and Job plot analyses were then carried out in an effort to probe the putative binding interactions.

In a first study, the changes in the ¹H NMR spectrum were monitored as increasing quantities of dianion 2 ²⁻ were added to 1 ⁴⁺ in DMSO-d ₆. The resulting Job plot revealed that the value for y _Jobplot [defined as the product of chemical shift change for the signal corresponding to H(1) on 1 ⁴⁺ and the host concentration] reached a maximum when the mole fraction was 0.5 ([H]/([H]+[G]) ( SI Appendix, Fig. S23, where [H] and [G] represent [host] (1 ⁴⁺) and [guest] (i.e., 2 ²⁻), respectively). Although subject to caveats (63), such a finding is consistent with a 1:1 ([H]: [G]) interaction between 1 ⁴⁺ and 2 ²⁻ dominating in solution. An association constant of lgK _a1 = 5.3(1), corresponding to this inferred 1:1 binding interaction, could then be calculated using standard curve fitting methods ( SI Appendix, Fig. S25).

A one-dimensional nuclear Overhauser enhancement (1D-NOE) spectroscopic analysis revealed cross-peaks between H ₂ (b) on 2 ²⁻ and H(2,7) on 1 ⁴⁺. This is as expected for a structure wherein 2 ²⁻ is inserted into the cavity of 1 ⁴⁺ to produce a pseudorotaxane-type structure as shown in Fig. 1 ( SI Appendix, Fig. S26).

The interactions between host 1 ⁴⁺ and the other anionic guests considered in this study, namely, (H⁺•2)⁻, (H⁺•3)⁻, and 3 ²⁻, were analyzed in a similar way (i.e., by means of Job plots, ¹H NMR spectral titrations, and 1D-NOE spectroscopy). In all cases, the results proved consistent with these anions binding 1 ⁴⁺ to form pseudorotaxane structures. However, for anions (H⁺•2)⁻ and (H⁺•3)⁻, y _Jobplot reached a maximum at 0.6 and 0.67 shown in the Job plots, values that correspond to formation of 2:3 or 1:2 (host/guest), respectively, complexes with 1 ⁴⁺ ( SI Appendix, Figs. S19 and S27). Further support for the proposed interactions between (H⁺•2)⁻ and (H⁺•3)⁻ with 1 ⁴⁺ came from theoretical calculations ( SI Appendix, Tables S7 and S8). Such non-1:1 stoichiometries have been seen previously in the case of 1 ⁴⁺ and other anionic guests (61, 62, 64). The binding stoichiometries and association constants (as lgK _a) inferred from these studies are summarized in Table 2.

Table 2.

Summary of the interactions between (H⁺•2)⁻, (H⁺•3)⁻, 2 ²⁻, and 3 ²⁻ with 1 ⁴⁺ as Inferred from ¹H NMR spectroscopic analyses carried out in DMSO-d ₆

Guest	[H]:[G]	lg K a
2 2−	1:1*	5.3(1)
(H+•2)−	2:3 †	4.0(1), 4.7(1), 2.5(1)
3 2−	1:1*	4.2(1)
(H+•3)−	1:2‡	2.7(1), 3.4(1)

Equations governing the relevant equilibria are given as footnotes to the [H]:[G] stoichiometry entries.

*[H]+[G]⇄Ka1[HG].

^† [H]+[G]⇄Ka1[HG],[HG]+[G]⇄Ka2[HG2], and [HG]+[HG2]⇄Ka3[H2G3].

^‡ [H]+[G]⇄Ka1[HG] and [HG]+[G]⇄Ka2[HG2].

Further evidence for the formation of complexes between (H⁺•2)⁻, (H⁺•3)⁻, 2 ²⁻, or 3 ²⁻ and 1 ⁴⁺ came from electrospray ionization high-resolution mass spectrometric (ESI-HRMS) analyses. For instance, ESI-HRMS analyses of 1:1 mixtures of 1 ⁴⁺ and 2 ²⁻ revealed a peak corresponding to [1 ⁴⁺ + 2 ²⁻] ²⁺ (m/z 449.1728; calculated m/z 448.1721) ( SI Appendix, Fig. S35). Corresponding results were obtained for mixtures of 1 ⁴⁺ and the other test anions considered in this study ( SI Appendix, Table S7). On this basis, we infer that the complexes formed between 1 ⁴⁺ and (H⁺•2)⁻, (H⁺•3)⁻, 2 ²⁻, or 3 ²⁻ in question are quite stable, even in the vapor phase.

Insights into the interactions between 1 ⁴⁺ and the azobenzene-4,4′-dicarboxylate (2 ²⁻) and 4,4′-stilbenedicarboxylate (3 ²⁻) dianions came from single-crystal X-ray diffraction analyses of single crystals of [1 ⁴⁺•2 ²⁻ ₂•2DMF•16.5H₂O] and [1 ⁴⁺•3 ²⁻ ₂•2DMF•16H₂O]. The single crystals used for these studies were obtained via the slow evaporation of mixtures containing 1 ⁴⁺•4PF₆ ⁻ (2.0 mM) and either 2 ²⁻ or 3 ²⁻ (10.0 mM) in the mixture containing N,N-dimethylformamide (DMF) and H₂O (1/1, vol/vol) over the course of 5 d ( SI Appendix, Table S10). In both cases, 1 ⁴⁺ adopts a similar “chair-like” conformation. The carboxylate anions 2 ²⁻ or 3 ²⁻ exist in planar trans configuration and are inserted into the cavity of 1 ⁴⁺ to produce a pseudorotaxane-type structure as shown in Fig. 3. In both structures, the planes of the dianions are vertical to the bridging benzene subunits in 1 ⁴⁺ and parallel to the 2,6-di(1H-imidazol-1-yl) pyridine moieties of 1 ⁴⁺. The short distances (less than 3.6 Å) between the carbon atoms on benzene rings of 2 ²⁻ or 3 ²⁻ and the nitrogen atoms on the pyridine rings of 1 ⁴⁺ are taken as evidence for the presence of stabilizing intermolecular π–π donor–acceptor interactions ( SI Appendix, Figs. S38 and S39).

Fig. 3.

Pseudorotaxane structures of (A) [1 ⁴⁺•2 ²⁻] (A ₁ and A ₂) and (B) [1 ⁴⁺•3 ²⁻] (B ₁ and B ₂) as seen in single crystals of [1 ⁴⁺•2 ²⁻ ₂•2DMF•16.5H₂O] and [1 ⁴⁺•3 ²⁻ ₂•2DMF•16H₂O]. Some and/or all solvent molecules have been omitted for clarity; A, 1 and B, 1 are side views and A, 2 and B, 2 are top views of the individual pseudorotaxane subunits.

Previous studies revealed that, when the width of a putative anionic substrate is larger than the cavity diameter of 1 ⁴⁺ (ca. 5.8 Å), pseudorotaxane structures are not obtained; rather, various so-called outside bound complexes are formed (65). It was thus expected that photoirradiation of 2 ²⁻ or 3 ²⁻, which would serve to stabilize the more compact, cis forms, would translate into a dethreading of the pseudorotaxanes and possible stabilization of more loosely associated host–guest species. The ¹H NMR spectroscopic experiments were thus carried out to probe the photoisomerization behavior of dianions 2 ²⁻ and 3 ²⁻ in the presence of 1 ⁴⁺. Irradiation of a 1:1 mixture of 2 ²⁻ and 1 ⁴⁺ with 365 nm light was found to generate 90% of the cis form of 2 ²⁻ (“cis-2 ²⁻”) at steady state. Furthermore, the yield of cis-2 ²⁻ (y _m.M(%)) relative to the corresponding trans form is observed to decrease only slightly over the course of 3 h (90 to 82%) in the presence of macrocycle 1 ⁴⁺ as determined via ¹H NMR spectroscopy ( SI Appendix, Fig. S40 and Table S11). A 1D-NOE spectroscopic analysis of this 1:1 mixture of 1 ⁴⁺ and cis-2 ²⁻ revealed correlations between protons H(2 to 7) on 1 ⁴⁺ and H ₂ (a') on cis-2 ²⁻ ( SI Appendix, Fig. S50). However, only very small chemical shift changes were observed for the signals of 1 ⁴⁺ (e.g., the change in the chemical shift of H(6) on 1 ⁴⁺ is ca. 0.15 ppm; SI Appendix, Fig. S48). We thus suggest that 1 ⁴⁺ adopts a “boat-like” conformation and acts as a molecular tweezer that binds cis-2 ²⁻ to form a sandwich-like structure (Fig. 4 A , 1–3). Support for this suggestion comes from theoretical calculations ( SI Appendix, Table S13). Using the molecular mechanics (MM+) force field included in the HyperChem 7.5 program (66), the geometries of the putative complexes formed from 1 ⁴⁺ and cis-2 ²⁻ were optimized. In the resulting optimized model, the benzene ring on cis-2 ²⁻ is essentially parallel to the 2,6-di(1H-imidazol-1-yl) pyridine subunit and orthogonal to the linking benzene rings present in 1 ⁴⁺. In addition, the chemical environment of H(6) on the bridged benzene rings of the tweezers binding mode (Fig. 4 A, 1–3 ) is similar to that in the pseudorotaxane structure [1 ⁴⁺•2 ²⁻] (Fig. 3 A, 1 and 2 ), such that H(6) experiences little in the way of an upfield chemical shift (0.15 ppm) upon photoisomerization of 2 ²⁻.

Fig. 4.

Optimized geometries of complexes formed from 1 ⁴⁺ and cis- 2 ²⁻ (A) and 3 ²⁻ (B). (A, 1 and B, 1) Front, (A, 2 and B, 2) top, and (A, 3 and B, 3) side views of the optimized structures showing the framework of 1 ⁴⁺ and cis-2 ²⁻ and cis-3 ²⁻ in space-filling form.

When 3 ²⁻ was subject to photoirradiation with 365 nm light in the presence of 1 ⁴⁺, 87% of the cis form was produced in the photostationary state as inferred from ¹H NMR spectroscopic studies. Notably, the cis-3 ²⁻ formed in this way exhibited high stability, with the percentage of the cis form dropping from 87 to 84% over the course of 2 d following photoirradiation ( SI Appendix, Fig. S43 and Table S12). In contrast to what was seen for cis-2 ²⁻, a 1D-NOE spectroscopic analysis of a mixture of 1 ⁴⁺ and cis-3 ²⁻ revealed no distinct cross-peaks between the protons of 1 ⁴⁺ and those of cis-3 ²⁻ ( SI Appendix, Fig. S57). This finding is interpreted in terms of cis-3 ²⁻ either not binding to 1 ⁴⁺ or being bound outside its cavity. We favor the latter interpretation, since an upfield chemical shift change of 0.22 ppm is seen for the H(6) signal of 1 ⁴⁺ ( SI Appendix, Fig. S55). Geometry optimizations were thus carried out. On the basis of these calculations, we conclude that, in the presence of cis-3 ²⁻, host 1 ⁴⁺ adopts a “basket-like” conformation to which cis-3 ²⁻ is bound as the “basket handle” (Fig. 4 B, 1–3 ). Furthermore, the bridging phenyl subunits on 1 ⁴⁺ are subject to shielding from the benzene rings present in 3 ²⁻, leading to an upfield chemical shift (0.22 ppm) in the signal corresponding to H(6).

As noted above, the photoresponsive behavior of 2H⁺•2, 2H⁺•3, (H⁺•2)⁻, (H⁺•3)⁻, 2 ²⁻, and 3 ²⁻ differed, with the greatest trans → cis conversion being observed in the anionic forms. This led us to explore whether changes in degree of protonation could be used to influence the extent to which the cis vs. trans ratio of the photoresponsive guests would be influenced by host 1 ⁴⁺. In a similar vein, we were keen to determine whether protonation changes could be used to regulate the nature of the complexes, if any, formed between 1 ⁴⁺ and (H⁺•2)⁻, (H⁺•3)⁻, 2 ²⁻, or 3 ²⁻ upon photoirradiation. In other words, could light and protonation be used as synergistic control elements to turn on or turn off the binding events that define the chemical switches comprised of 1 ⁴⁺ and 2H⁺•2 or 2H⁺•3 and their deprotonated, anionic forms?

To address the above question, the photoisomerization behavior of 2H⁺•2, 2H⁺•3, 2 ²⁻, and 3 ²⁻ were investigated in the presence of 1 ⁴⁺. Initial time-dependent ¹H NMR studies were carried out on each 1:1 host–guest mixture in DMSO-d ₆ after photoirradiation with 365 nm light ( SI Appendix, Figs. S44, S46, S48, S51, S53, and S55). The apparent forward and reverse rate constants (k _m1 and k _m-1) were calculated for each combination of guest (i.e., 2H⁺•2, 2H⁺•3, (H⁺•2)⁻, (H⁺•3)⁻, 2 ²⁻, or 3 ²⁻) and host 1 ⁴⁺ (Table 3). It was found that the photoisomerization rate constants (k _m1) and the steady-state photoconversion yields (y _s.m) (note that the subscript “m” denotes the presence of macrocycle 1 ⁴⁺; thus, k _m1 and y _s.m represent the corresponding apparent rate constants and simulated photoconversion yields, respectively) increase with deprotonation, as is true in the absence of 1 ⁴⁺ ( SI Appendix, Figs. S45, S47, S49, S52, S54, and S56). This leads us to suggest that the extent of protonation is the main determinant as to whether a complex is formed with host 1 ⁴⁺; however, it was expected that light could be used to regulate the nature of the complex obtained (pseudorotaxane vs. an outside binding mode).

Table 3.

Conversion efficiencies and apparent rate constants of the photoisomerization processes of 2H⁺•2, 2H⁺•3, (H⁺•2)⁻, (H⁺•3)⁻, 2 ²⁻, and 3 ²⁻ in the presence of 1 ⁴⁺

Guest	y m.m * (%)	y s . m * (%)	k m1 † (10−2 s−1)	k m-1 † (10−2 s−1)	R 2 ‡
1 4+ + 2H+•2	22	27	0.37(1)	1.0(9)	0.999
1 4+ + (H+•2)−	70	72	1.2(1)	0.48(1)	0.998
1 4+ + 2 2−	90	95	2.6(9)	0.12(4)	0.991
1 4+ + 2H+•3	30	34	0.73(8)	1.4(9)	0.999
1 4+ + (H+•3)−	73	74	4.1(7)	1.4(7)	0.999
1 4+ + 3 2−	87	91	4.9(6)	0.47(6)	0.991

*Conversion efficiencies for the photoisomerization process.

^†The k _{m 1} and k _m−1 represent apparent forward and reverse rate constants, respectively. The values of  km1 and  km−1 were calculated by curve fitting as detailed in SI Appendix .

^‡ R ² corresponds to the adjusted R-square value. Note that “m” refers to measurements carried out in the presence of 1 ⁴⁺.

Consistent with the above expectation, the kinetic and thermodynamic parameters associated with guest photoisomerization were found to change in the presence of 1 ⁴⁺ in the case of 2H⁺•2, 2H⁺•3, (H⁺•2)⁻, (H⁺•3)⁻, 2 ²⁻, and 3 ²⁻ (Fig. 5). Notably, the asymptotic photoconversion yields are similar within the error range (10%) in the presence and absence of the 1 ⁴⁺ as shown in Fig. 5. In particular, light-induced trans → cis conversion yields (y_{s .m}(%)) of 95% and 72% were seen for (H⁺•2)⁻ in the absence and presence, respectively, of 1 ⁴⁺. In the case of 3 ²⁻, the corresponding y _s.m(%) values are 79% and 91% in the absence and presence of 1 ⁴⁺ (Tables 1 and 3). We thus suggest that the pseudorotaxane-type structure stabilizes the trans form of the anion, while binding to 1 ⁴⁺ serves to stabilize the cis form of the anionic guest. The final conversion yield (y _s.m) depends, therefore, not just on the inherent photoisomerization features of the guest but also on whether a given species benefits from stabilization upon binding to 1 ⁴⁺. Such a finding is consistent with our underlying design principle, namely, that the ensemble consisting of 1 ⁴⁺ and 2H⁺•2 or 2H⁺•3 and their anionic forms act as multimodal chemical switches subject to synergistic external stimulus control.

Fig. 5.

(A and B) Bar graphs showing the differences in photoconversion efficiencies seen in the absence (yellow columns) and presence (green columns) of receptor 1 ⁴⁺ as a function of substrate and protonation state [2H⁺•2, (H⁺•2)⁻, and 2 ²⁻ (A) or 2H⁺•3, (H⁺•3)−, and 3 ²⁻ (B)]. (C and D) Bar graphs showing the corresponding differences in the apparent forward rate constants (note: the absence (blue columns) and presence (purple columns) of receptor 1 ⁴⁺ and substrate as 2H⁺•2 (H⁺•2)⁻ and 2 ²⁻ (C) or 2H⁺•3 (H⁺•3)⁻ and 3 ²⁻ (D)].

We also explored whether the presence or absence of 1 ⁴⁺ has an effect on the photoisomerization kinetics of 2H⁺ •2 and 2H⁺ •3 and their anionic forms. As can be seen from an inspection of the values in Table 1, in most cases, the differences in the apparent rate constants (k ₁ and k _m1) corresponding to trans → cis photoisomerization in the absence and presence of 1 ⁴⁺ are not dramatic (typically less than a factor of 2). Exceptions to the generalized trend are seen in the case (H⁺•2)⁻, where the off rate is enhanced by factors of roughly 6 and 3, where a possible slowing (by slightly less than a factor of 3) in seen the presence of 1 ⁴⁺ (Fig. 5 C and D and Table 3). Although further analysis is needed, on the basis of the aggregated data, we conclude that the effect of 1 ⁴⁺ on the photostationary cis–trans ratio is largely thermodynamic in nature and that the generally modest differences in the on and off rates simply reflect such effects (Fig. 6).

Fig. 6.

Relationship between isomerization conversion of guests 2H⁺•2, 2H⁺•3, (H⁺•2)⁻, (H⁺•3)⁻, 2 ²⁻, and 3 ²⁻ and illumination time at 365 nm in the presence of 1 ⁴⁺.

Next, the photoresponsive nature of pseudorotaxane [1 ⁴⁺•2 ²⁻] was tested. Upon irradiation with ultraviolet (UV) light (365 nm), a decrease in the intensity of the peak around 334 nm corresponding to the π–π* transition and an increase in absorption intensity at ca. 437 nm corresponding to the n–π* transition is observed, indicating the trans → cis photoisomerization of 2 ²⁻ (67 –69). As discussed above, this conversion induces dethreading of the dianion from 1 ⁴⁺. To test the reversibility of the system, conversion back to the predominantly trans photostationary state was induced by irradiating with a 100-W LED 450 nm light source (radiant flux = 5.1 W in all tests unless otherwise noted). Under these conditions, threading and dethreading caused by 365 nm and then 450 nm photoinduced isomerization could be affected for at least 12 cycles without obvious photofatigue (Fig. 7 A and B and SI Appendix, Figs. S58 and S60). Rethreading could also be induced by heating at 80 °C (Fig. 7 C and D and SI Appendix, Figs. S61 and S62). It is worth noting that thermally promoted (80 °C) photoisomerization shows a dampening in the isomerization cycles when compared to that promoted by 450 nm photoirradiation. This difference is ascribed to decomposition of the receptor at higher temperatures (70) ( SI Appendix, Fig. S63). Nevertheless, switching through multiple cycles proved possible. These complementary dethreading approaches, in conjunction with an ability to control the degree of protonation, serve to underscore our belief that the nonbinary switching systems made up from 1 ⁴⁺ and the anionic forms of 2H⁺•2 or 2H⁺•3 are robust and readily subject to synergistic control.

Fig. 7.

(A and C) UV-visible (UV-vis) absorption spectra (black lines; 1 mM in DMSO at 298 K) of 1 ⁴⁺ + 2 ²⁻ recorded after subjecting to UV 365 nm irradiation (blue lines) to achieve the cis-photostationary state and the subsequent irradiation with 450 nm (A) or heat (80 °C; C) to produce the corresponding trans-photostationary state (pink lines). (B and D) Plots showing the percentage of cis-2 ²⁻ inferred from UV-vis absorption spectral analyses as a mixture of 1 ⁴⁺ and 2 ²⁻ is subject to cycles of 365 nm−induced cis-2 ²⁻ and 450-nm (B) or thermally promoted (80 °C; D) trans-2 ²⁻ photoisomerization in DMSO at 298 K.

In order to expand the application of our synergistic regulation strategy, the molecular switching processes involving anion 2 ²⁻ were explored under nonequilibrium dynamic conditions (71, 72) (Figs. 8 and 9). In dark, dianion 2 ²⁻ (2 mM in DMSO-d ₆/D₂O (9/1, vol/vol), created via mixing azobenzene-4,4′-dicarboxylic acid (2H⁺•2) (1.2 µmol, 2 mM) and Et₃N (6 µmol, 10 mM) at pH 9.5) was treated with a “chemical fuel” CCl₃COOH (24 µmol, 40 mM; pH of the overall system = 4.3) at 303 K (Fig. 8). The time-dependent proton shift change trends from 2H⁺•2 to 2 ²⁻ were similar (charactered by t_1/2 = 10.4 s vs. 9.5 s, t₁ = 27 s vs. 30 s, where t_1/2 is the time required for 50% of the proton chemical signal for 2H⁺•2 to convert to that of 2; t₁ is the time required to fully deprotonate 2H⁺•2 as judged by a stable chemical shift to give a final system pH of ≥8.6) in the absence or presence of 1 ⁴⁺ ( SI Appendix, Figs. S65 and S67). These results lead us to suggest that these dark processes reflect a pH-controlled system with dynamic character.

Fig. 8.

Schematic representation of switching systems driven by a “chemical fuel” pulse (i.e., CCl₃COOH, 24 µmol, 40 mM) operating in the absence (A) or presence (B) of 1 ⁴⁺ without light.

Fig. 9.

(A) Schematic representation of molecular shuttling effects driven by a “chemical fuel” pulse (i.e., CCl₃COOH) in the presence or absence of 1 ⁴⁺ under conditions of 365 nm photoirradiation. (B) Relationship between the extent of isomerization of guest 2 ²⁻ and illumination time at 365 nm in the presence or absence of 1 ⁴⁺.

Chemical fuel–driven nonequilibrium dynamic molecular switching of 2 ²⁻ was further probed under conditions of 365 nm photoisomerization. Here, addition of 1 ⁴⁺ resulted in an acceleration in the rate of photoisomerization of 2 ²⁻ with t_1/2.365nm now ≤52 s, that is, faster than without 1 ⁴⁺ (t_1/2.365nm ≥140 s) (note that t_1/2.365nm is the time required for 50% of the trans form of 2 ²⁻ and related species to convert to the corresponding cis isomers) (Fig. 9). Furthermore, the presence of 1 ⁴⁺ leads to an increase in the y _m(%) values (measured via ¹H NMR spectroscopy) from 50 to 80% ( SI Appendix, Tables S20 and S21). These effects are ascribed to a stepwise increase in the effective pH value that favors trans to cis photoisomerization with additive 1 ⁴⁺ serving to strengthen these effects. As such, we conclude that the present synergistic approach may be used to regulate dynamic systems that are not subject to inherent equilibrium control.

In the present system, the various inputs translate into differences in the absorbance features (Fig. 7 A and C ). In addition, an upfield chemical shift of 0.05 ppm or 0.22 ppm is seen for the H(6) signal of 1 ⁴⁺ after photoisomerization ( SI Appendix, Figs. S48 and S55). Finally, 1D-NOE spectroscopic analysis revealed correlations between H(a') on cis-2 ²⁻ (or H(b') on cis-3 ²⁻) and H(1 or 2 to 7) on 1 ⁴⁺ after photoisomerization ( SI Appendix, Figs. S50 and S57). We thus suggest that the multimodal chemical switches produced from 1 ⁴⁺ are not subject only to hierarchical control via the use of nonorthogonal inputs but that the resulting effects may be readily monitored by means of several different spectroscopic outputs.

In conclusion, we report a pair of molecular switching ensembles regulated by protonation state and photoirradiation. In this system, the so-called Texas-sized molecular box (1 ⁴⁺; studied as its tetrakis-PF₆ ⁻ salt) is used as a wheel-like host, and either the azobenzene-4,4′-dicarboxyate (2 ²⁻) or 4,4′-stilbenedicarboxyate (3 ²⁻) (di)anions are used as axle moieties to produce pseudorotaxane in the absence of photoirradiation. Exposure to UV light was found to cause trans → cis photoisomerization and dethreading. Further exposure to visible light or heat, which resulted in cis → trans conversion, restored the pseudorotaxane bound forms. The original binding events were also found to be dependent on the degree of protonation with the neutral forms, 2H⁺•2 and 2H⁺•3, showing little propensity to bind to 1 ⁴⁺. The present use of synergistic stimuli allows a high level of control over chemical switching events. These switching events can be followed using standard spectroscopic outputs. As such, the approach detailed here represents a potentially generalizable strategy that could see use in applications requiring stimuli-responsive functional materials, thereby setting the stage for the future development of logic-controlled nanomachines.

Materials and Methods

Azobenzene-4,4′-dicarboxylic acid (2H⁺•2) and 4,4′-stilbenedicarboxylic acid (2H⁺•3) were purchased commercially (Energy) and used without further purification. Deuterated solvents were purchased from Cambridge Isotope Laboratory. NMR spectra were recorded on Bruker Advanced instrument (600 MHz). The ¹H NMR chemical shifts are reported relative to the residual solvent (¹H: DMSO-d ₆ at 2.49 ppm) (73).

Photoisomerization studies were carried out under 365- or 450-nm light (CEL-LED100-365 or CEL-LED100-450, Beijing China Education Au-Light Co., Ltd.). Nonlinear data fitting was carried out using the Origin Pro-8.5 software. ESI-HRMS data were acquired by infusion into a Fourier transform ion cyclotron resonance mass spectrometer.

The structures shown in SI Appendix, Tables S7 and S13 were optimized in vacuum via molecular mechanics (MM+) using the force field in the HyperChem 7.5 program. Further details of calculations are described in SI Appendix .

The data crystals were cut from clusters of crystals and had the approximate dimensions given in the .cif documents available from the Cambridge Crystallographic Data Centre (CCDC) by quoting CCDC numbers 2074970 and 2074971. The structures were solved and refined by full-matrix least squares on F² with anisotropic displacement parameters for the non-H atoms using SHELXL-18 (74). Neutral atom scattering factors and values used to calculate the linear absorption coefficient are from the International Tables for X-ray Crystallography (1992) (75). All ellipsoid figures were generated using SHELXTL/PC (76).

The monoanions (H⁺•n)⁻ and dianions n ²⁻ (n = 2 or 3) were obtained by adding separately one or two molar equivalents of tetramethylammonium hydroxide pentahydrate (TMA⁺ •OH⁻ •5H₂O) to a solution of 2H⁺•2 or 2H⁺•3. The anions obtained in this way were used to study the binding interactions with macrocycles 1 ⁴⁺.

The single crystals used for these studies were obtained via the slow evaporation of mixtures containing 1 ⁴⁺•4PF₆ ⁻ (2.0 mM) and 2 ²⁻ or 3 ²⁻ (10.0 mM) in DMF/H₂O (1/1, vol/vol) over the course of 5 d.

The “chemical fuel” pulse-driven molecular shuttle operations were achieved by adding CCl₃COOH at 303 K (70, 71) after first testing the system in its absence. Further details of these procedures are described in SI Appendix .

Data Availability

Crystal structures have been deposited in CCDC (2074970, 2074971). All other study data are included in the article and/or SI Appendix.