Solvent Effects and Driving Forces in Pillararene Inclusion Complexes

May 8, 2015 - Pillararenes, a recently discovered class of aromatic macrocycles, form inclusion complexes with a large number of guest molecules, but ...
1 downloads 9 Views 4MB Size
Article pubs.acs.org/JPCB

Solvent Effects and Driving Forces in Pillararene Inclusion Complexes Christian Schönbeck,†,‡,§ Hui Li,‡ Bao-Hang Han,*,‡ and Bo W. Laursen*,† †

Nano-Science Center & Department of Chemistry, University of Copenhagen, Universitetsparken 5, 2100 Copenhagen, Denmark National Center for Nanoscience and Technology, Beijing 100190, China § Sino-Danish Center for Education and Research, Niels Jensens Vej 2, DK-8000 Aarhus C, Denmark ‡

S Supporting Information *

ABSTRACT: Pillararenes, a recently discovered class of aromatic macrocycles, form inclusion complexes with a large number of guest molecules, but not much is known about the driving forces of complexation, including the role of the solvent. We have measured the binding thermodynamics for a small number of model complexes in several solvents and used computational chemistry to rationalize the obtained results and identify the driving forces of complexation. Favorable electrostatic interactions between the host and guest are obtained when the charge distribution in the guest matches the negative electrostatic potential in the cavity of the pillararene. Polar guests, however, also interact strongly with polar solvents, thereby shifting the complexation equilibrium away from the complex. The shape of the solvent molecules is another important factor as some solvents are sterically hindered from entering the pillararene cavity. By changing solvent from acetonitrile to o-xylene the binding constant in one case increased more than 4 orders of magnitude. Even electrostatically similar solvents such as o-xylene and p-xylene have very different impacts on the binding constants due to their different abilities to fit into the cavity. The study illustrates the importance of taking into account the interactions between the solvent and the complexing species in the investigation and design of molecular host:guest systems.



cationic guests14−16 and guests with electron-withdrawing groups17 this suggests that electrostatic attraction between the PA and the guest is an important driving force.9,12 The direct interaction between host and guest is not the only important interaction that contributes to the stability of host:guest complexes. Although sometimes overlooked, the interaction between the solvent molecules and the host/guest molecules may be just as important as the direct host:guest interaction. There are not many studies of the effect of solvents on the stability of PA complexes, but there seems to be a trend that the complexation constants decrease with increased polarity of the solvent.4,7,17−19 This may be explained by the favorable interaction of polar solvents with the polar guest molecule and with the negatively charged cavity of the PA. Likewise, nonpolar solvents are expected to interact less strongly with the PA and the guest, thereby shifting the complexation equilibrium toward the complex. The affinity of solvent molecules for the PA cavity depends not only on the polarity of the solvent but also on the molecular size of the solvent molecules, as some solvent molecules may be too bulky to fit into the PA cavity. This is expected to promote binding of the guest molecule as no energy is required to first expel the solvent molecules from the cavity. Differences in the molecular size of the solvent have been used to explain variations in the self-inclusion of an alkyl

INTRODUCTION First synthesized in 2008,1 pillararenes (PAs) are a relatively new class of macrocyclic molecules. The basic structure consists of phenyl groups linked by methylene bridges to form a highly symmetrical tube-like structure (Figure 1). The most common

Figure 1. Structure of per-methoxylated pillararene, DMP5A.

PAs are the 5-membered pillar[5]arenes (P5A), but larger analogues have also been synthesized,2 and all together these form the basis for a large number of modified PAs.3 Like other macrocycles, PAs are capable of forming inclusion complexes with suitably sized guest molecules, and a large number of inclusion complexes between various PAs and guest molecules have been reported.4−11 Nevertheless, not much is known about the driving forces for the formation of PA inclusion complexes. A specific feature of PAs is the presence of a significant negative electrostatic potential in the cavity, a consequence of the π-electron-rich building blocks.12,13 Together with the observation that PAs has a preference for © XXXX American Chemical Society

Received: March 15, 2015 Revised: May 8, 2015

A

DOI: 10.1021/acs.jpcb.5b02515 J. Phys. Chem. B XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry B

gas phase and solvents, and frequency calculations were also performed in both gas phase and solvents to obtain the enthalpies of the guests. For the larger species, the free DMP5As and the complexes, structural optimization and frequency calculations were feasible in the gas phase but were computationally demanding in solvent and were only in a few test cases performed in the solvent. Instead, another less computationally demanding strategy was chosen to obtain the enthalpies of the larger species in the solvents. The gas phase optimized structures were immersed in the solvent and the electronic energies were calculated. The zero-point energies and the thermal corrections obtained from the gas phase frequency calculations were added to the electronic energies of the solvated species to yield to enthalpies of the solvated species. Omitting the structure optimization and the frequency calculations in the solvents is a source of error but several tests revealed that the resulting errors in the complexation enthalpies were less than 1 kJ/mol, which must be considered minor relative to other possible sources of error. Complexation enthalpies were calculated in vacuum and in the various implicit solvents by subtracting the enthalpies of the free species from the enthalpies of the complex. In this approach the solvent was treated as a pure continuum. Another approach took into account the specific interactions between one solvent molecule and the PA and thereby treated the solvent as a competitor:

chain in a modified PA. Inclusion was observed when chloroform was used as solvent but not when the less bulky dichloromethane was used.20 In the present study, we explore the effects of various organic solvents on the formation of pillararene inclusion complexes with the aim of gaining insight into the driving forces of complexation. To assess the effect of direct electrostatic attraction between the guest and the PA, guest molecules with different electron-withdrawing groups at the end of the alkyl chain were used. The interaction between the guests and the solvent was probed by varying the solvent polarity. Interactions between the PA and the solvent depend not only on the polarity but also on the ability of the solvent to fit into the cavity, and the importance of the solvent−host interaction could therefore be estimated by varying the molecular size of the solvents. As model compounds we used the per-methoxylated P5A (DMP5A, Figure 1) as host molecule and 1,4-dibromobutane (BrC4Br), dicyanoethane (CNC2CN) and 1,3-dicyanopropane (CNC3CN) as guest molecules. BrC4Br and CNC2CN have previously been shown to bind to P5As.5,17 For experimental determination of the binding constants, isothermal titration calorimetry (ITC) was employed. This widely used technique is capable of providing highly precise values of binding constants as well as direct determination of the binding enthalpy. The obtained experimental thermodynamic parameters may, together with quantum mechanical calculations and molecular dynamics simulations, provide a detailed picture of the different energetic contributions to the formation of PA inclusion complexes.

DMP5A: Solvent + Guest ⇄ DMP5A: Guest + Solvent (1)

The second approach required calculating the enthalpy of binding of one solvent molecule to the PA in the relevant solvent, e.g., the binding of one o-xylene molecule to DMP5A was calculated in implicit o-xylene. Second, the binding enthalpy of the solvent molecule was subtracted from the previously calculated binding enthalpy of the guest to yield the final complexation enthalpy corresponding to the equilibrium in eq 1. The calculated binding enthalpies of the various host:guest complexes are found in Table S-1. Molecular Dynamics Simulations. Molecular dynamics simulations of DMP5A in 40 Å solvent boxes were run for 10 ns (ns) with the NAMD25 code, and the equations of motion were integrated with 2 fs timesteps. The pressure and temperature was maintained at 1.013 bar and 298 K by the use of a barostat and a thermostat. Periodic boundary conditions were used, and the cutoff for the van der Waal’s interactions was set to 12 Å with a switching function starting at 10 Å. Charges and force field parameters were all from the Charmm General Force Field v. 2b8.26 Specifically, the atom types and charges for DMP5A were taken from the pethylphenol and methoxybenzene residues in the corresponding topology file. Initial coordinates for DMP5A was obtained from its crystal structure.1 Pure solvent boxes were generated by Packmol,27 then equilibrated for 1 ns before solvating the DMP5A molecule, which was performed by means of the “solvate” function in VMD.28



EXPERIMENTAL SECTION Chemicals. Solvents were purchased from Sinopharm Chemical Reagent Co., Ltd. Guest molecules 1,4-dibromobutane (BrC4Br), dicyanoethane (CNC2CN), 1,3-dicyanopropane (CNC3CN), and methyldibromoglutaronitrile (MDBG) were purchased from J&K Scientific Ltd. and used as received. Per-methoxylated pillar[5]arene (DMP5A) was prepared according to reported procedures.21 The isolated material contained strongly bound solvent, which was identified and quantified by H NMR (spectrum is given in the Supporting Information). ITC. Calorimetric titrations were conducted on a VP-ITC titration calorimeter from GE Health Care, Life Sciences, USA. Prior to titration the solutions were filtered through a 0.22 μm pore size syringe filter for organic solvents. Solutions of the guest molecules were titrated into a filtrated solution of the DMP5A host molecule, typically in 10 μL aliquots. Concentrations of the guest molecules ranged from 0.8 to 1500 mM and host concentrations ranged from 0.09 to 2 mM. The large variations in concentrations reflect the large variations in binding constants. Low binding constants requires high concentrations in order to obtain precise binding parameters from the titrations. Control titrations in which the guest molecule was titrated into pure solvent were conducted and subtracted from the binding experiments. The resulting enthalpograms were analyzed in the Origin 7 data analysis application that was supplied with the calorimeter. Quantum Chemical Calculations. All quantum chemical calculations were performed in Gaussian 0922 with the camB3LYP functional and the 6-31g(d,p) basis set. The IEFPCM solvation model23,24 was used for the calculations in solvent. The structures of all guest molecules were optimized in both



RESULTS AND DISCUSSION Matching Electronic Distribution in Host and Guest Molecules. As described in the Introduction, electrostatic attraction between host and guest molecules is considered to be an important driving force for the formation of PA inclusion complexes. To investigate this hypothesis, quantum mechanical B

DOI: 10.1021/acs.jpcb.5b02515 J. Phys. Chem. B XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry B calculations of the electrostatic potential around the host and guest molecules were performed. The results are visualized in Figures 2 and 3 as electrostatic potential mapped onto

Figure 2. Calculated electrostatic potential in DMP5A mapped onto an electronic isodensity surface with a density of 0.001. A significant negative electrostatic potential is found in the cavity of DMP5A while a positive electrostatic potential is found at the rims.

isoelectronic surfaces. For the host molecule, DMP5A, the πelectrons on the phenyl rings create a strongly negative electrostatic potential inside the cavity of DMP5A. Conversely, a positive electrostatic potential is found around the methyl groups that decorate the openings of DMP5A. For the guest molecules, the presence of electron-withdrawing groups at the ends of the alkyl chains leads to a positive electrostatic potential around the central hydrocarbon moieties and a negative electrostatic potential around the electron-withdrawing groups at the ends of the guest molecules. This is of course more pronounced for the dicyano guests than for BrC4Br. For comparison, the quite uniform and more or less neutral electrostatic potential around butane is also shown in Figure 3. When comparing the electrostatic potential around the host and guest molecules, it is striking how the positive charge around the central parts of the alkyl chains matches the negative charge in the cavity of DMP5A, and the negative charge at the ends of the guest molecules matches the positive charge at the rims of DMP5A. It seems very likely that the complementary charge distribution in host and guest molecules is a major driving force for the formation of PA inclusion complexes. This interpretation is corroborated by the observations that alkanes, which appear electrically neutral (see Figure 3), have very low affinities for PAs.9,29 Experimental Determination of Binding Thermodynamics. The titration of guest molecules into DMP5A produced enthalpograms that could be fitted by The One Set of Sites binding model (Figure 4), which assumes one set of identical and independent binding sites on each host molecule.30 For each titration, this yielded a binding constant, K, a binding enthalpy, ΔH, and a binding stoichiometry, N. By use of the relations in eq 2, the binding constant was converted into Gibbs free energy of complexation, ΔG°, which was split into contributions from enthalpy, ΔH, and entropy, ΔS°. −RT ln(K ) = ΔG° = ΔH − T ΔS°

Figure 3. Calculated electrostatic potential mapped onto electronic isodensity surfaces with a density of 0.001. From top to bottom: CNC2CN, CNC3CN, BrC4Br, and butane. Electron-withdrawing groups at the end of the alkyl chains results in a positive electrostatic potential around the central parts of the alkyl chains and a negative electrostatic potential around the end groups. All guests contain four carbons except CNC3CN, which is five carbons long.

For each complex in each solvent, several titrations with different concentrations of host and/or guest were conducted to validate the binding parameters. The results from the titrations conducted in the most suitable concentration range are presented in Table 1. The binding stoichiometries, N, were in all cases significantly lower than unity. In the strongly binding systems, that is, the experiments in o-xylene and toluene, the enthalpograms exhibited a sigmoidal shape (Figure 4A) and N could be determined with high precision. These experiments consistently yielded values of N close to 0.75. There is no reason to believe that the investigated hosts and guests form complexes with a stoichiometry different from 1, and the low values of N are assigned to an overestimation of the concentration of DMP5A. This was confirmed by inspection of the H NMR spectrum (see Supporting Information) which showed the presence of close to one equivalent of strongly bound dichloromethane as well as alkane solvent residues. The bound solvent corresponds to a

(2) C

DOI: 10.1021/acs.jpcb.5b02515 J. Phys. Chem. B XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry B

Figure 4. Enthalpograms in the case of a strong complex (A) and a weak complex (B). The sigmoidal shape obtained for the strong complex allows for a precise determination of the stoichiometry, N, in contrast to the weak system where N cannot be determined and must be kept fixed during the iterative fitting procedure. (A) Titration of 0.09 mM DMP5A with 0.8 mM CNC2CN in o-xylene. (B) Titration of 2.0 mM DMP5A with 1.5 M CNC3CN in acetonitrile.

Table 1. Experimental Binding Parameters from the ITC Experiments CNC2CN

BrC4Br

CNC3CN

solvent

K (M−1)

ΔG° (kJ/mol)

ΔH (kJ/mol)

ΔS° (J/(mol·K))

N

acetone DMSO acetonitrile toluene tolueneb o-xylene p-xylene acetone DMSO acetonitrile toluene o-xylene p-xylene acetone DMSO acetonitrile toluene o-xylene p-xylene

× × × × × × × × × × × × × × × × × × ×

−17.6 −14.2 −9.3 −31.0 −31.0 −35.8 −28.2 −13.9 −13.5 −5.8 −19.6 −24.1 −17.1 −13.6 −11.6 −4.2 −24.8 −29.2 −22.3

−25.7 −28.4 −19.7 −32.3 −31.7 −43.6 −12.2 −34.9 −39.6 −20.4 −35.1 −45.9 −4.9 −24.1 −21.1 −15.9 −27.5 −39.1 −8.5

−27.3 −47.3 −35.1 −4.3 −2.5 −26.5 53.7 −70.6 −87.8 −48.9 −51.9 −73.4 40.8 −35.3 −31.8 −39.5 −9.2 −33.0 46.3

0.92 0.83 0.82a 0.73 0.87 0.76 0.49 0.84 0.84 0.82a 0.76 0.75 0.44 0.68 0.89 0.75c 0.74 0.76 0.14

1.2 3.1 4.2 2.7 2.7 1.8 8.7 2.7 2.3 1.0 2.7 1.6 9.7 2.4 1.1 5.3 2.2 1.3 8.0

3

10 102 101 105 105 106 104 102 102 101 103 104 102 102 102 100 104 105 103

The value of N was fixed at 0.82, rather than 1, during the fitting procedure to accommodate errors in the concentration of the DMP5A sample. Experiments performed on an extra dried sampled (vacuum for 2 h at 190 °C, see Supporting Information for details). cThe value of N was fixed at 0.75 during the fitting procedure to accommodate errors in the concentration of the DMP5A sample. These experiments were conducted with a second batch of DMP5A. a b

fitting parameters. This problem was circumvented by keeping N fixed at 0.75 during the fitting procedure instead of treating it as a floating fitting parameter. Very low values of N, sometimes as low as 0.14, were observed for the titrations in p-xylene. This was attributed to a low solubility of DMP5A in this solvent such that the concentration in the filtered samples was significantly decreased. If this is the cause for the low values of N, then the obtained values of K and ΔH are correctly describing the system (see Supporting Information), and this was confirmed by the consistent values of these parameters, obtained from

mass fraction of approximately 10−15%. When the concentration of DMP5A is corrected for this known solvent content the N values are close to the usual range for 1:1 complexes (0.9−1.1). Even though the concentration errors result in artificially low values of the stoichiometry, N, the values of K and ΔH are not affected by the solvent residues, as shown by measurements on a control sample dried at 190 °C in vacuum for 2 h (see Supporting Information for details). In the weakly binding systems the enthalpograms were less curved (Figure 4B) and the values of N and ΔH could not be precisely determined due to a strong correlation between these two D

DOI: 10.1021/acs.jpcb.5b02515 J. Phys. Chem. B XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry B

in the guests, as illustrated by the ESP maps in Figures 2 and 3. The withdrawal of negative charge from the center of the guests toward the electron-withdrawing groups at the ends of the guest increases in the order BrC4Br < CNC3CN < CNC2CN and it is therefore no surprise that the binding constants increase in the same order. Everything else being equal, the increased electrostatic attraction between host and guest molecules should result in more favorable complexation enthalpies (ΔH) but this is not the case. Instead, the enthalpic contributions to the binding increased in the order CNC3CN < CNC2CN < BrC4Br. The assumed relation between ΔH and the electrostatic attraction between host and guest is too simplistic since there are many other contributions to the complexation enthalpy, including the interaction between guest and solvent which will be discussed below. Quantum mechanical (QM) calculations of some of the host:guest complexes were conducted to investigate the different contributions to the complexation enthalpy, most of which will be discussed below in relation to the solvent effects. The calculated enthalpies are shown in Figure 6 and reproduces

various titrations with different concentrations of host and guest (see Figures S-5−S-11). Effect of Guest Structure. The thermodynamic parameters for the complexations are plotted in Figure 5 as a function of the solvent. The influence of solvent will be discussed below−in the present section the impact of the guest molecular structure is discussed. In most of the solvents, but most pronounced in the less polar solvents, the guests bound to DMP5A in the following order: BrC4Br < CNC3CN < CNC2CN. The explanation seems straightforward considering the electronic distributions

Figure 6. Calculated enthalpies for the binding of three guest molecules to DMP5A in o-xylene, toluene, acetonitrile and DMSO. As explained in further detail in the section on the solvent effects, the solvent is modeled as a continuum with one explicit solvent molecule binding to the cavity of the free DMP5A.

the dominant experimental trends (compare with Figure 5B), although the calculated enthalpic stabilization of the BrC4Br complex is strongly overestimated. The calculated binding enthalpy of the hypothetical guest, Butane, is much less favorable than for the other guests, due to the lack of matching charge distributions in host and guest (Figure 3). The entropic contributions (Figure 5C) are relatively constant when measured in 4 out of the 6 solvents and TΔS° is ≈-10 kJ/mol for the complexes of the two dicyano guests and somewhat lower (≈-20 kJ/mol) for the complexes with BrC4Br. The entropic destabilization of BrC4Br relative to the two other guest molecules, which leads to the weaker binding of BrC4Br, could stem from the greater flexibility of the longer alkyl chain and the concomitant larger reduction in conformational entropy upon complexation. To further test the hypothesis that the electrostatic match between host and guest is a fundamental driving force for the formation of PA inclusion complexes, a fourth guest molecule, methyldibromoglutaronitrile (MDBG), was investigated. The structure is similar to those of the other guest molecules and

Figure 5. Thermodynamic parameters for the complexation of three different guests with DMP5A in different organic solvents. E

DOI: 10.1021/acs.jpcb.5b02515 J. Phys. Chem. B XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry B

The possible origins of the solvent effects are numerous but can be grouped into three contributions: (1) guest−solvent interactions, (2) host−solvent interactions, and (3) interactions between solvent and the formed complex. Favorable interactions between host/guest and the solvent (contributions 1 and 2) shifts the equilibrium toward the noncomplexed state, while the opposite is true if the solvent stabilizes the complex (contribution 3). The interaction of solvent with the free and complexed species may be calculated in various ways. The simplest approach, which forms the basis for many computational solvation models,23 is to treat the solvent as a dielectric continuum. The central physical property of this continuum is the dielectric constant, ε, which is a measure of the polarity of the medium. Increasing the dielectric constant polarizes the molecules and leads to increased electrostatic interactions between solute and solvent. Furthermore, a high dielectric constant more effectively shields electrostatic charges. A plot of the binding constant as a function of the dielectric constant (Figure 5A) reveals a weak tendency of the binding constant to decrease with increasing dielectric constant of the solvent. It is obvious, however, that other factors are more important than the dielectric constant. This is especially clear for the solvents toluene, o-xylene and p-xylene which have very similar dielectric constants but have quite different impacts on the binding thermodynamics of the guests. In the following, the observed solvent effects will first be discussed in the framework of the continuum model, and afterward in terms of specific interactions between solvent molecules and the PA. Solvent as a Dielectric Continuum. Theoretical complexation enthalpies of the complexes between DMP5A and three guests (CNC2CN, BrC4Br, and butane), were calculated in the gas phase and in different implicit solvents. Theoretical estimation of the solvent effects (ΔΔHsol gas) were calculated as the difference between the gas phase and the implicit solvent calculations:

consists of two cyano groups and two bromines attached to the ends of an alkyl chain (Figure 7). As for the other nitrile guests

Figure 7. Energy-minimized structure of DMP5A:MDBG complex shown along with the chemical structure of MDBG. The central positively charged methylene groups on MDBG cannot reach into the electronegative center of DMP5A due to the bulky end group. Thus, the electrostatic match between host and guest is not optimal.

the nitrogens on the cyano groups are in close proximity of the methyl hydrogens of the PA and may engage in stabilizing CH−N interactions. However, in contrast to the other guest molecules the positively charged central alkyl chain cannot be properly included in the central part of the PA due to the bulky end-group on MDBG (see Figure 7). This prevents a good electrostatic match between MDBG and the PA, and a low binding constant is therefore expected if this is the dominant driving force. Indeed, calorimetric titrations in o-xylene reveals a binding constant of 92 M−1 (ΔG° = −11.2 kJ/mol) which is much lower than for the other guests. The complexation enthalpy (−32 kJ/mol) is less favorable than for the other guests but only by 7−14 kJ/mol, which is not enough to be the sole cause of the much smaller binding constant. QM calculations in implicit o-xylene also shows that the binding enthalpy of MDBG (−51.6 kJ/mol) is only 3.2 kJ/mol less favorable than for the binding of CNC2CN (−54.8 kJ/mol). Overall, the experimental binding constants strongly support the hypothesis that matching charge distributions in host and guest is an important factor. Increasing the positive charge on the part of the guest molecule that is located in the electronegative center of the PA increases the binding constant. However, this driving force is expected to be enthalpic in origin but neither the experimental nor the calculated complexation enthalpies conform to this picture. The binding enthalpy of BrC4Br is larger (more negative) than for CNC2CN despite the more electropositive center on CNC2CN. Similarly, the binding enthalpy of MDBG is only slightly smaller than for the other guests despite the seemingly poor electrostatic match with the host. Effect of Solvent. The influence of the solvent on the complexation thermodynamics is very similar for each of the three guest molecules. A clear trend is observed for the binding constants (Figure 5A) where the solvent promotes complexation in the order o-xylene > toluene > p-xylene > DMSO ≈ acetone > acetonitrile. Extremely strong solvent effects are observed, as illustrated by the CNC2CN:DMP5A complex where K increases from 42 to 1.8 × 106 M−1 upon changing solvent from acetonitrile to o-xylene. The trends observed for K (and ΔG°) (Figure 5A) is to a large extent brought about by differences in complexation enthalpy (Figure 5B), which follows the same pattern as ΔG°. Toluene, and especially pxylene, seems to entropically stabilize the complexes, as compared to the other solvents.

sol ΔΔHgas = ΔH(sol) − ΔH(gas)

where ΔH(sol) and ΔH(gas) are the complexation enthalpies in the solvent and the gas phase, respectively. The solvent effect can also be calculated as the difference between the solvation enthalpy of the complex and the solvation enthalpies of the free species: sol sol sol sol ΔΔHgas = ΔHgas (PA :G) − {ΔHgas (PA) + ΔHgas (G)}

Thus, if the solvent-mediated stabilization of the complex, ΔHsol gas(PA:G), is larger than the total stabilization of the free sol species, ΔHsol gas(PA) + ΔHgas(G), the solvent will favor the complex, as compared to the gas phase. It turns out that ΔΔHsol gas is positive in all solvents (see Figures S-12−S-14), meaning that the formation of complexes is enthalpically destabilized by the presence of solvent. Both the free species and the complexes are stabilized by the solvent but the solvent interacts stronger with the free species than with the complex, thus favoring dissociation of the complex. This effect is more pronounced in the polar solvents (DMSO, acetonitrile and acetone) than in the nonpolar solvents (toluene and the xylenes) and agrees with the lower binding constants measured in the polar solvents. To compare the polar to the nonpolar solvents, it is more illustrative to look at the difference in complexation enthalpies between solvents than between gas phase and solvent. The solvent effects are almost invariant within the group of polar solvent and within the group of F

DOI: 10.1021/acs.jpcb.5b02515 J. Phys. Chem. B XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry B

capable of including some solvents in the cavity, while others are excluded for steric reasons. The two smallest solvent molecules, acetonitrile and acetone, are capable of being fully included, and during most of the simulation time the PA cavity contained a single one of these molecules. This is consistent with the report that DMP5A form weak inclusion complexes with acetonitrile (K = 30 M−1) and acetone (K = 5 M−1) in CDCl3.31 On the other hand, DMSO is a little too bulky to be completely included, and during the simulation the DMSO molecules were rarely located in the middle of DMP5A, unlike acetone and acetonitrile. The incomplete inclusion of DMSO led to a slightly cone-shaped structure of DMP5A in which some of the aromatic rings were slightly tilted. Only rarely were two DMSO molecules partially included, so for most of the simulation time the cavity was partially empty. The poor interaction of DMSO with the PA may explain the increased binding in DMSO compared to acetonitrile. Inclusion of the aromatic solvent molecules in the DMP5A crucially depended on the methyl groups. The aromatic rings were too large to be included but a methyl group attached to the aromatic ring filled up the cavity pretty well. This is in agreement with the experimentally observed binding mode of p-toluenesulfonate with a modified P5A.32 Thus, toluene and pxylene were partially included and interacted with the cavity. Conversely, when two methyl groups were in the orthoposition, as in o-xylene, the solvent molecule was too bulky and was excluded from the cavity. In the simulation of DMP5A in oxylene a shallow inclusion of the aromatic ring was observed but for long intervals the solvent was completely excluded. This might explain the extraordinary high binding constants observed in o-xylene. p-Xylene contains two methyl groups that are capable of being included in the PA while toluene only contains one. The effective concentration of competitive solvent molecules may thus be regarded as approximately twice as high in p-xylene. This may be the reason for the decreased binding of guest molecules in p-xylene. QM calculations of PA:solvent complexes were made in an attempt to quantify the explicit binding of solvent molecules to the PA, and thereby evaluate the ability of solvent molecules to act as competitors for the binding to PA. The calculations were performed in implicit solvent and it was assumed that only one solvent molecule would bind to the cavity of DMP5A, in line with what is suggested by the MD simulations and crystal structures.31 Complexation enthalpies were calculated for four different solvent molecules, o-xylene, toluene, acetonitrile, and DMSO, and were in all cases negative, but they did only partially conform to the experimental results and the conclusions from the MD simulations. As expected, o-xylene can only be vaguely included in the PA cavity and thus yields a binding enthalpy (−9.0 kJ/mol) that is somewhat smaller than for toluene (−19.1 kJ/mol). The difference between these two solvents agrees well with the observed differences in binding enthalpies when conducting titrations in these two solvents. The binding of CNC2CN, CNC3CN, and BrC4Br to DMP5A were all more exothermic in o-xylene, by 11.3, 11.6, and 10.9 kJ/mol, respectively. It seems evident that the stabilization of PA inclusion complexes in o-xylene is due to the weak inclusion of solvent molecules in the PA cavity, as the energetic cost of removing o-xylene from the cavity to make room for the guest molecule is very low. It was expected that the same mechanism could explain the stabilization of inclusion complexes when going from acetonitrile to DMSO. After all, the MD simulations

nonpolar solvents (see Figures S-12−S-14), and the difference between the polar and nonpolar solvents is exemplified by the differences between DMSO and p-xylene. The calculated enthalpies of transfer from p-xylene to DMSO are plotted in Figure 8 along with the calculated difference in complexation

Figure 8. Different contributions to the solvent effect, defined as the destabilizing effect of a polar solvent (DMSO) relative to a nonpolar solvent (p-xylene). The size of the red bar is significantly diminished throughout the series of complexes, showing that the interaction of the free guest with the solvent is the main reason for the variation in solvent effects among the three complexes.

enthalpies between these two solvents. All species are stabilized in DMSO relative to p-xylene but the stabilization of the free species is larger than for the complexes, leading to less favorable complexation enthalpies in DMSO. The solvent effect, expressed as the destabilizing effect of the polar solvents relative to the nonpolar solvents, decreases in the order CNC2CN > BrC4Br > butane. A closer look at Figure 8 reveals that the decrease in solvent effect is primarily brought about by increased interaction of the guest molecule with the solvent. CNC2CN interacts much stronger with the polar solvent than does BrC4Br, and Butane only has a very weak preference for the polar solvent. As shown in the previous section, the withdrawal of negative charge to the ends of the guest molecules ensures a good electrostatic match with the PA. However, it also promotes interaction with the solvent and may thus be an overall destabilizing factor in polar solvents. Solvent Molecules as Competitive Guests. The above calculations, which treated the solvent as a continuum, suggested that formation of complexes is favored in nonpolar solvents, as compared to polar solvents. The experimental observations also showed a trend of increased binding in nonpolar solvents, but with strong variations among the nonpolar solvents that are not predicted by the continuum model. For example, the complexation constants are higher in o-xylene than in toluene and p-xylene, despite the slightly higher polarity of o-xylene. Similarly, DMSO favors complexation compared to acetonitrile, despite being a more polar solvent. These observations must be explained by the specific interactions of solvent molecules with the PA, such as the ability to fill up the cavity,13 and these were explored by molecular dynamics (MD) simulations of DMP5A in boxes of explicit solvent molecules. These revealed that DMP5A is G

DOI: 10.1021/acs.jpcb.5b02515 J. Phys. Chem. B XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry B showed weak inclusion of DMSO in the PA, while acetonitrile is known to form inclusion complexes with PAs.17,31 However, the QM calculations showed that the binding enthalpy of DMSO (−21.5 kJ/mol) was slightly more favorable than the binding enthalpy of acetonitrile (−18.6 kJ/mol). Despite this result from the calculations, we still consider the poor inclusion of DMSO into the PA cavity to be the main reason for the higher binding constants in DMSO, as compared to acetonitrile. The QM calculation only provides the enthalpy of the lowest energy conformation and it is possible that this conformation is very unlikely in the case of DMSO, which for steric reasons is difficult to fit into the PA cavity. Discussion of Solvent Effects. As stated in the introduction, the stability of a given host:guest complex depends on the strength of interactions between (i) host− guest (ii) host−solvent, and (iii) guest−solvent. The above results show that treating the solvent as a dielectric continuum is a grave simplification, and the strong variations in binding thermodynamics in different solvents cannot be accounted for by the dielectric constant of the solvent alone. Still, the solvent polarity is an important parameter and the largest binding constants were observed in the nonpolar solvents. It turned out that the general destabilization of the complexes in polar solvents to a large extent could be attributed to the increased guest−solvent interaction. Since PA guests must be polar to match the charge distribution inside the PA cavity, similar strong solvent effects are expected for all other PA inclusion complexes. While the guest−solvent interaction may be reasonably predicted by a continuum model, it is necessary to treat the host-solvent interaction explicitly. Some solvents are too bulky to be included in the host and will thereby favor binding of the guest in the empty host. The three aromatic solvents have similar dielectric constants, and the implicit solvent model therefore resulted in very similar binding parameters in these solvents, contrary to what was observed experimentally. Only by considering the specific interactions of the solvent with the host molecule the experimental observations could be rationalized. Indeed, the change of solvent from p-xylene to toluene to o-xylene produced changes in ΔG° and ΔH that were almost independent of the guest (see Table 1). This is a strong indication that the solvent effects for these three solvents are induced by differences in host-solvent interactions rather than guest-solvent interaction. Roughly speaking, the guestsolvent interaction is determined by the dielectric constant of the solvent, while the host-solvent interaction is dominated by the specific interactions between the host and the solvent molecules, in particular the ability of the solvent molecule to fit into the cavity of the PA. Figure 9 and Figures S-15 and S-16 summarize the ability of the computational approaches to reproduce the experimentally obtained binding enthalpies. Reasonable agreement with experiment is obtained when the implicit solvation model is supplemented by one explicit solvent molecule that binds to the uncomplexed host. In relation to the overall discussion of the driving forces for formation of PA inclusion complexes it is noticeable that a large number of reported binding constants have been measured in CDCl3 or CHCl3. Chloroform is a relatively nonpolar solvent (εr = 4.8) and at the same time there are indications that it is a bit too bulky to be properly included in the cavity of P5A,20,33 which is confirmed by modeling (Figure S-17). Thus, chloroform is expected to be a complexation-promoting solvent

Figure 9. Calculated and experimental enthalpies for the binding of CNC2CN to DMP5A. A mixed solvent model in which the binding of one solvent molecule (a competitor) to the free PA is treated explicitly and the rest of the solvent as a continuum results in better agreement with experimental data than the purely implicit solvation in which all solvent is treated as a continuum.

for the same reasons as o-xylene is a complexation-promoting solvent, although the reported binding constants for the BrC4Br:DMP5A complex in CDCl3 (K = 1.6 × 103 M−1)5 suggest that o-xylene (K = 1.6 × 104) has a stronger effect. Nevertheless, one may speculate that the formation of many of the reported inclusion complexes with P5As are driven by the solvent rather than strong attractive interactions between host and guest. The inability of the solvent to fit into the PA is likely to be an important driving force. Another consequence of the present work is that it highlights the importance of taking the competitive inclusion of solvent molecules into account when comparing experimental data to computationally predicted complexation free energies, obtained by the use of implicit solvent models. As an illustration of this issue, a large bias was observed for the calculated binding energies for a series of guests binding to cucurbit[7]uril. It was speculated that the inability of the implicit solvent model to accurately calculate the energy of the cavity bound water was responsible for this bias,34 a speculation which was soon after supported by another study of the energetics of cavity bound water.35 Indeed, a recent review characterizes the release of cavity bound high-energy water as “the essential enthalpic driving force for complexation”.36 Water is in many regards a peculiar solvent and it is meaningless to compare the presently reported solvent effects to those induced by water, but it nevertheless shows the importance of treating the solvent explicitly rather than as a dielectric continuum. This is highlighted by the experiments in toluene, o-xylene and pxylene which, despite having similar dielectric constants, exhibit very different solvent effects due to minor structural differences.



CONCLUSION A strong negative electrostatic potential is found in the cavity of pillararenes. In order to form stable complexes, the guest molecules must present a matching charge distribution, and this is accomplished by the introduction of electron-withdrawing groups to the parts of the guest molecules that protrude from the pillararene cavity. Increasing the polarity of the guest increases the electrostatic interaction with the pillararene but it also increases the interaction of the uncomplexed guest with H

DOI: 10.1021/acs.jpcb.5b02515 J. Phys. Chem. B XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry B

(7) Li, C.; Chen, S.; Li, J.; Han, K.; Xu, M.; Hu, B.; Yu, Y.; Jia, X. Novel Neutral Guest Recognition and Interpenetrated Complex Formation from Pillar[5]arenes. Chem. Commun. 2011, 47, 11294− 11296. (8) Li, C.; Ma, J.; Zhao, L.; Zhang, Y.; Yu, Y.; Shu, X.; Li, J.; Jia, X. Molecular selective binding of basic amino acids by a water-soluble pillar[5]arene. Chem. Commun. 2013, 49, 1924−1926. (9) Hu, X. S.; Deng, H. M.; Li, J.; Jia, X. S.; Li, C. J. Selective Binding of Unsaturated Aliphatic Hydrocarbons by a Pillar[5]arene. Chin. Chem. Lett. 2013, 24, 707−709. (10) Yu, G.; Zhou, X.; Zhang, Z.; Han, C.; Mao, Z.; Gao, C.; Huang, F. Pillar[6]arene/Paraquat Molecular Recognition in Water: High Binding Strength, pH-Responsiveness, and Application in Controllable Self-Assembly, Controlled Release, and Treatment of Paraquat Poisoning. J. Am. Chem. Soc. 2012, 134, 19489−19497. (11) Zhang, H.; Zhao, Y. Pillararene-Based Assemblies: Design Principle, Preparation and Applications. Chem.Eur. J. 2013, 19, 16862−16879. (12) Ogoshi, T.; Yamagishi, T. Pillar[5]- and Pillar[6]arene-Based Supramolecular Assemblies Built by Using Their Cavity-Size-Dependent Host-Guest Interactions. Chem. Commun. 2014, 50, 4776−4787. (13) Park, C.; Kim, K. T. Pillar[n]arenes and Other Cavitands: Aspects of Complex Thermodynamics. Chin. J. Chem. 2015, 33, 311− 318. (14) Han, C.; Yu, G.; Zheng, B.; Huang, F. Complexation between Pillar[5]arenes and a Secondary Ammonium Salt. Org. Lett. 2012, 14, 1712−1715. (15) Dong, S.; Yuan, J.; Huang, F. A Pillar[5]arene/Imidazolium [2]rotaxane: Solvent- and Thermo-Driven Molecular Motions and Supramolecular Gel Formation. Chem. Sci. 2014, 5, 247−252. (16) Ogoshi, T.; Kayama, H.; Yamafuji, D.; Aoki, T.; Yamagishi, T. a. Supramolecular Polymers with Alternating Pillar[5]arene and Pillar[6]arene Units from a Highly Selective Multiple Host-Guest Complexation System and Monofunctionalized Pillar[6]arene. Chem. Sci. 2012, 3, 3221−3226. (17) Shu, X.; Chen, S.; Li, J.; Chen, Z.; Weng, L.; Jia, X.; Li, C. Highly Effective Binding of Neutral Dinitriles by Simple Pillar[5]arenes. Chem. Commun. 2012, 48, 2967−2969. (18) Li, C.; Zhao, L.; Li, J.; Ding, X.; Chen, S.; Zhang, Q.; Yu, Y.; Jia, X. Self-Assembly of [2]pseudorotaxanes Based on Pillar[5]arene and Bis(imidazolium) Cations. Chem. Commun. 2010, 46, 9016−9018. (19) Zhang, Z.; Han, C.; Yu, G.; Huang, F. A Solvent-Driven Molecular Spring. Chem. Sci. 2012, 3, 3026−3031. (20) Ogoshi, T.; Akutsu, T.; Yamafuji, D.; Aoki, T.; Yamagishi, T. a. Solvent- and Achiral-Guest-Triggered Chiral Inversion in a Planar Chiral pseudo[1]Catenane. Angew. Chem., Int. Ed. 2013, 52, 8111− 8115. (21) Wang, K.; Tan, L. L.; Chen, D. X.; Song, N.; Xi, G.; Zhang, S. X.-A.; Li, C.; Yang, Y. W. One-pot Synthesis of Pillar[n]arenes Catalyzed by a Minimum Amount of TfOH and a Solution-Phase Mechanistic Study. Org. Biomol. Chem. 2012, 10, 9405−9409. (22) Gaussian 09, Revision B.01, Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.; Robb, M. A.; Cheeseman, J. R.; Scalmani, G.; Barone, V.; Mennucci, B.; Petersson, G. A.; et al. Gaussian Inc.: Wallingford CT, 2010. (23) Tomasi, J.; Mennucci, B.; Cammi, R. Quantum Mechanical Continuum Solvation Models. Chem. Rev. 2005, 105, 2999−3094. (24) Scalmani, G.; Frisch, M. Continuous Surface Charge Polarizable Continuum Models of Solvation. I. General Formalism. J. Chem. Phys. 2010, 132, 114110. (25) Phillips, J. C.; Braun, R.; Wang, W.; Gumbart, J.; Tajkhorshid, E.; Villa, E.; Chipot, C.; Skeel, R. D.; Kalé, L.; Schulten, K. Scalable Molecular Dynamics with NAMD. J. Comput. Chem. 2005, 26, 1781− 1802. (26) Vanommeslaeghe, K.; Hatcher, E.; Acharya, C.; Kundu, S.; Zhong, S.; Shim, J.; Darian, E.; Guvench, O.; Lopes, P.; Vorobyov, I.; et al. CHARMM General Force Field: A Force Field for Drug-Like Molecules Compatible with the CHARMM All-Atom Additive Biological Force Fields. J. Comput. Chem. 2010, 31, 671−690.

polar solvents. Consequently, the complexes are strongly destabilized in polar solvents such as acetone, acetonitrile and DMSO. However, the specific interactions between host and solvent molecules may be an even more important factor than the polarity of the solvent. When the solvent is too bulky to be included in the pillar-arene, which is the case for o-xylene, the binding constants are greatly enhanced. To illustrate the magnitude of the solvent effects, the binding constant of the DMP5A:CNC2CN complex increases from 42 M−1 in acetonitrile to 1.8 × 106 M−1 in o-xylene. The solvent-induced variation in the binding constants, the binding enthalpies and the binding entropies, follows a similar pattern for each of the three guest molecules. This suggests that the solvent effects are general, and similar effects are expected for other complexes.



ASSOCIATED CONTENT

S Supporting Information *

H NMR of DMP5A and quantification of residual solvents, assessment of the impact of sample impurities on the obtained thermodynamic parameters, calculated complexation enthalpies and solvation enthalpies, ITC enthalpograms for the titrations in p-xylene, plots of solvation enthalpies and solvent effects, energy-minimized structures of complexes between DMP5A and dichloromethane and chloroform. The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.jpcb.5b02515.



AUTHOR INFORMATION

Corresponding Authors

*(B.W.L.) E-mail: [email protected]. *(B.-H.H.) E-mail: [email protected]. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS Financial support from the National Natural Science Foundation of China (Grant No. 61261130092), the SinoDanish Center for Education and Research, and the DanishChinese Center for Molecular Nanoelectronics funded by the Danish National Research Foundation is acknowledged.



REFERENCES

(1) Ogoshi, T.; Kanai, S.; Fujinami, S.; Yamagishi, T. a.; Nakamoto, Y. para-Bridged Symmetrical Pillar[5]arenes: Their Lewis Acid Catalyzed Synthesis and Host-Guest Property. J. Am. Chem. Soc. 2008, 130, 5022−5023. (2) Ogoshi, T.; Ueshima, N.; Sakakibara, F.; Yamagishi, T. a.; Haino, T. Conversion from Pillar[5]arene to Pillar[6−15]arenes by Ring Expansion and Encapsulation of C60 by Pillar[n]arenes with Nanosize Cavities. Org. Lett. 2014, 16, 2896−2899. (3) Cao, D.; Meier, H. Pillar[n]arenes - a Novel, Highly Promising Class of Macrocyclic Host Molecules. Asian J. Org. Chem. 2014, 3, 244−262. (4) Li, C.; Xu, Q.; Li, J.; Feina, Y.; Jia, X. Complex Interactions of Pillar[5]arene with Paraquats and Bis(pyridinium) Derivatives. Org. Biomol. Chem. 2010, 8, 1568−1576. (5) Shu, X.; Fan, J.; Li, J.; Wang, X.; Chen, W.; Jia, X.; Li, C. Complexation of Neutral 1,4-Dihalobutanes with Simple Pillar[5]arenes That is Dominated by Dispersion Forces. Org. Biomol. Chem. 2012, 10, 3393−3397. (6) Strutt, N. L.; Forgan, R. S.; Spruell, J. M.; Botros, Y. Y.; Stoddart, J. F. Monofunctionalized Pillar[5]arene as a Host for Alkanediamines. J. Am. Chem. Soc. 2011, 133, 5668−5671. I

DOI: 10.1021/acs.jpcb.5b02515 J. Phys. Chem. B XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry B (27) Martínez, L.; Andrade, R.; Birgin, E. G.; Martínez, J. M. PACKMOL: A Package for Building Initial Configurations for Molecular Dynamics Simulations. J. Comput. Chem. 2009, 30, 2157− 2164. (28) Humphrey, W.; Dalke, A.; Schulten, K. VMD: Visual Molecular Dynamics. J. Mol. Graphics 1996, 14, 33−38. (29) Ogoshi, T.; Demachi, K.; Kitajima, K.; Yamagishi, T. a. Selective Complexation of n-Alkanes with Pillar[5]arene Dimers in Organic Media. Chem. Commun. 2011, 47, 10290−10292. (30) MicroCal ITC Data Analysis in Origin. Tutorial Guide. 2004. (31) Tan, L. L.; Zhang, Y.; Li, B.; Wang, K.; Zhang, S. X.-A.; Tao, Y.; Yang, Y. W. Selective Recognition of ″Solvent″ Molecules in Solution and the Solid State by 1,4-Dimethoxypillar[5]arene Driven by Attractive Forces. New J. Chem. 2014, 38, 845−851. (32) Gómez, B.; Francisco, V.; Fernández-Nieto, F.; Garcia-Rio, L.; Martín-Pastor, M.; Paleo, M. R.; Sardina, F. J. Host-Guest Chemistry of a Water-Soluble Pillar[5]arene: Evidence for an Ionic-Exchange Recognition Process and Different Complexation Modes. Chem.Eur. J. 2014, 20, 12123−12132. (33) Ogoshi, T.; Ueshima, N.; Akutsu, T.; Yamafuji, D.; Furuta, T.; Sakakibara, F.; Yamagishi, T. a. The Template Effect of Solvents on High Yield Synthesis, Co-Cyclization of Pillar[6]arenes and Interconversion Between Pillar[5]- and Pillar[6]arenes. Chem. Commun. 2014, 50, 5774−5777. (34) Muddana, H. S.; Gilson, M. K. Calculation of Host-Guest Binding Affinities Using a Quantum-Mechanical Energy Model. J. Chem. Theory Comput. 2012, 8, 2023−2033. (35) Biedermann, F.; Uzunova, V. D.; Scherman, O. A.; Nau, W. M.; De Simone, A. Release of High-Energy Water as an Essential Driving Force for the High-Affinity Binding of Cucurbit[n]urils. J. Am. Chem. Soc. 2012, 134, 15318−15323. (36) Biedermann, F.; Nau, W. M.; Schneider, H. J. r. The Hydrophobic Effect Revisited - Studies with Supramolecular Complexes Imply High-Energy Water as a Noncovalent Driving Force. Angew. Chem., Int. Ed. 2014, 53, 11158−11171.

J

DOI: 10.1021/acs.jpcb.5b02515 J. Phys. Chem. B XXXX, XXX, XXX−XXX