The Cation−π Interaction in Chemistry and Biology

Abstract

The cation−π interaction is an important noncovalent binding force that impacts all areas of chemistry and biology. Extensive computational and gas phase experimental studies have established the potential strength and the essential nature of the interaction. Previous reviews have emphasized studies of model systems and a variety of biological examples. This work includes discussion of those areas but emphasizes other areas that are perhaps less well appreciated. These include the novel cation−π binding ability of alkali metals in water; the application of the cation−π interaction to organic synthesis and chemical biology; cooperative behaviors of multiple cation−π interactions, including adhesive proteins from mussels and similar organisms and the formation and modulation of biomolecular condensates (phase separation); and cation−π interactions involved in recognizing DNA/RNA.

1. Introduction

Over the past 40 years the cation−π interaction has come to be recognized as a major noncovalent binding force, with important examples across all of chemistry and molecular/structural biology. Thousands of studies of cation−π interactions have been published, including several valuable reviews. We published a relatively comprehensive (for its time) review in Chemical Reviews in 1997.¹ In 2013 we published a more personal perspective in Accounts of Chemical Research,² and a very comprehensive and especially informative review was published in this journal by Mahadevi and Sastry.³ Other reviews on aspects of the cation−π interaction have appeared.^4,5

In the decade since the Sastry review, the pace of research involving cation−π interactions has quickened, with many thousands of papers describing various examples and insights. The present manuscript cannot exhaustively cover all that work. We will begin with a brief overview of the essential aspects of the cation−π interaction, providing a key framework for discussion. From the beginning, we and others have highlighted many examples of cation−π interactions in biological systems, and we will describe select examples here. However, cation−π interactions are seen in a wide variety of chemical systems, some of which have received less attention, and a major focus here will be to highlight areas that perhaps were not as strongly emphasized in previous reviews. In particular, we will describe the surprising nature of cation−π interactions involving alkali metal ions in fully aqueous solutions. Also, we will describe systems in which multiple cation−π interactions work cooperatively to produce remarkable effects. Our emphasis will be on naturally occurring systems, with only a few select examples of studies involving model systems.

We note again that it is not possible to cover all work that presents important insights into or examples of cation−π interactions. We apologize in advance to scientists whose work we were not able to include.

2. The Fundamentals—Gas Phase Studies

It is generally agreed that the starting gun for studies of the cation−π interaction was fired by Kebarle, with a 1981 study that showed that K⁺ binds to benzene in the gas phase with a −ΔH° of 19 kcal/mol.⁶ Importantly, in the same study it was shown that K⁺ binds to water with a −ΔH° of 18 kcal/mol. This stunning result established that a cation−π interaction can be quite strong, and it presaged the subsequent findings that it could be important in aqueous media, and hence in biological systems.

Gas phase studies provide the best data for revealing the basic nature of the cation−π interaction, without complications from solvation effects (which must be considered, of course, and will be below). Scores of measurements of various cations binding diverse π systems have been tabulated.⁷ In addition, we and others established early on that relatively simple levels of theory faithfully reproduce the experimental data. Some variation is seen, both in different experimental approaches and in different computational methods. For the present purposes, though, the key question is not whether the K⁺···benzene binding energy is 18.7 or 19.2 kcal/mol. The point is that it is strong and on the order of 19 kcal/mol. In addition, it is usually the trends in binding energies that are of most interest to experimentalists. In such comparisons, it is best if all measurements are made under identical conditions, and for this reason all cation−π interaction energies shown here (Figure 1) were computed at a level of theory that has consistently been shown to reproduce results from experiment and from higher levels of theory.^8,9 To be clear, essentially all the results in Figure 1 have been reported previously using, in some cases, much higher levels of theory and/or experimental measurement. However, none of these studies lead to meaningfully different conclusions about the cation−π interaction, and, again, we feel there is value in having all numbers generated in the same way.

Figure 1.

A. Cation−π binding energies for representative systems. B. Binding energies for Na⁺ to various aromatic systems. All values (kcal/mol) were obtained by computation using M06-2X DFT with the 6-31+G* basis set.

Figure 1 shows representative cation−π binding energies. This relatively small selection of data reveals many essential aspects of the cation−π interaction. First, ethylene binds cations quite well. The cation−π interaction has nothing to do with aromaticity, or, to a large degree, any kind of orbital interactions, as the HOMO and LUMO of ethylene are energetically out of range for any reasonable mixings. Second, as we move down the periodic table—as the cation gets larger—the cation−π interaction gets weaker. This is the hallmark of an electrostatic interaction. Third, any kind of cation binds—note ammonium and tetramethylammonium (TMA).^10,11

Interestingly, K⁺ and ammonium have the same binding energies to benzene. They are also very similar in ionic radius and aqueous solvation energy, so much so that ammonium ions can often pass through biological ion channels that are otherwise quite selective for K⁺.

These findings establish the key role that electrostatic interactions play in the cation−π interaction. To first order, simple π systems like ethylene or benzene have a buildup of negative electrostatic potential on their face, and cations are naturally attracted to it. The simplest way to think about the origin of the electrostatic potential is to consider that sp² C is more electronegative than H, creating 4 (6) bond dipoles that converge of the center of the π system of ethylene (benzene). The overall charge distribution in ethylene and benzene is that of a quadrupole, and one can think of the electrostatic interaction as being between a cation and a quadrupole moment. In contrast, sp³ C is not meaningfully more electronegative than H, and so alkanes such as cyclohexane do not bind cations strongly. Alkanes are quite polarizable, though. Cyclohexane is more polarizable than benzene, but cyclohexane does not bind cations well. As such, polarizability is not the defining feature of the cation−π interaction in simple systems, although it can be a factor in larger systems. As we move away from very simple molecules like ethylene and benzene, the origin of the buildup of electrostatic potential is less easily traced to polarized C–H bonds, but it is still there. Very early on we established the value of calculated electrostatic potential surfaces in predicting/rationalizing cation−π interactions.¹²

The fourth takeaway from Figure 1 is that trends in cation−π binding energies for substituted benzenes do not follow the intuition that most chemists have developed through studies of electrophilic aromatic substitution.^8,13 It makes sense that fluorobenzene should be a weaker binder than benzene, as the highly electronegative F should withdraw electron density from over the ring. However, organic chemists think of phenol as “electron rich”, yet it is not meaningfully better than benzene in the cation−π interaction. Aniline does produce stronger cation−π interactions, and consistent with that, indole shows a very strong cation−π interaction.

An interesting consequence of the major role of electrostatics in the cation−π interaction is that the distance dependence of the cation−π interaction is not very steep. A fully electrostatic interaction would scale with distance simply as 1/r, and, indeed, we found that the distance dependence of a cation−π interaction is not strong, varying as 1/rⁿ, with n < 2.^9,14 There are two important implications of this finding. First, while in simple molecules the optimal geometry has the cation over the center of the ring and at van der Waals contact, deviating from that geometry, either by moving from the center of the ring or moving further away from the ring, is not overly costly. For example, a Na⁺ ion that is 1.6 Å further from the benzene centroid than is optimal still shows an interaction energy of 10 kcal/mol. As such, in protein structures or drug-receptor interactions, the cation does not have to be positioned perfectly to achieve an energetically meaningful interaction. A second implication is that regions of electrostatic build up can be fairly far from the cation and yet still contribute. We noted above that substituent effects do not follow expectations based on polarization of the π system. Wheeler and Houk have noted that such π polarization is not important in rationalizing substituent effects, but a predominantly electrostatic model is appropriate.⁸ Interestingly, for most substituents, a direct electrostatic interaction between the cation and the substituent itself is an important contributor to the binding. Again, the long-range nature of electrostatic interactions facilitates this effect.

With larger π systems, some interesting effects are seen. As shown in Figure 1B, going from benzene to naphthalene, anthracene, and triphenylene gives a steady, but small, increase in cation−π binding energies. Sastry has performed extensive, higher level computational studies of cation−π interactions to larger π systems, with the aim of understanding the binding of ions to graphene (see below for a discussion of graphene).¹⁵ With Li⁺ as the probe cation, a general trend was seen in which bigger ring systems produce stronger interaction energies. The variation is not large. The interaction energy rises from 41.7 to 50.5 kcal/mol on going from a system with one ring (benzene) to the linear polyacene with 10 rings. A plot of interaction energy vs polarizability is linear, but the intercept at zero polarizability is ∼40 kcal/mol, indicating that the contribution to the overall interaction energy from polarizability is not large.

In these larger systems there are no strong biases as to whether a terminal ring or an inner ring is the preferred binding site. In earlier work we noted the usefulness of electrostatic potential surfaces for predicting relative cation−π binding abilities.¹² This can be effective for simple, closely related systems, but comparisons across more diverse structures are not straightforward. For example, electrostatic potential surfaces for benzene vs naphthalene (Figure 2) do not clearly indicate that naphthalene should bind ions more tightly. The maximum negative electrostatic potential in benzene is larger than in naphthalene. Again, the shallow distance dependence of the cation−π interaction is an issue here. While an ion centered over one ring of naphthalene may experience a weaker electrostatic attraction than an ion centered over benzene, the second ring of naphthalene can contribute significantly to the overall binding,

Figure 2.

Electrostatic potential surfaces for benzene and naphthalene. The level of theory is as in Figure 1. The range for the electrostatic potential surfaces is ±100 kJ/mol.

The strongest Na⁺ interaction energy in Figure 1 is for indole, and of course, benzene, phenol, and indole are the aromatic portions of the side chains of phenylalanine (Phe), tyrosine (Tyr), and tryptophan (Trp). It has consistently been seen that Trp plays an especially important role in biological cation−π interactions. The data from Figure 1 suggest, however, that Phe and Tyr should be comparable, but that is not at all the case. Tyr is substantially over-represented at cation−π binding sites compared to Phe. Two aspects of Tyr could explain this. First, the cation−π interaction has a significant directional component, and the OH of Tyr, through hydrogen bonding, could position the Tyr side chain optimally. Perhaps more importantly, it makes sense that when the OH of phenol is hydrogen bonded, the aromatic ring can make significantly stronger cation−π interactions. We showed this some time ago in a simple model system,¹⁶ and Sastry has done a systematic study to show that, indeed, when the OH of phenol is engaged in a hydrogen bond, the cation−π interaction becomes stronger.¹⁷

We have emphasized that simple electrostatics provide a good understanding of the cation−π interaction, and that relatively modest computational models are adequate. That is not to say that other effects, such as induced dipoles and polarizability, are not important, but rather that they do not explain the basic trends. Computationally, one could imagine that as systems get larger, polarizability might become more important; an ion like TMA is presumably more polarizable than Na⁺. We were surprised, then, that the simple model of Figure 1 performed well for TMA, for which the experimental binding energy is 9 kcal/mol.¹⁰

Experimentally, using substituents to systematically modify the electrostatic potential of a ring, and hence the cation−π interaction, has proven to be quite powerful. While simple enough in model systems,¹⁸ to perform such studies in biological receptors typically requires sophisticated methodologies for incorporating noncanonical (unnatural) amino acids site-specifically into proteins. Nevertheless, a large number of such studies has been reported.^4,19,20 An especially effective approach has been creation of a so-called “fluorination plot”, in which an aromatic of the protein is progressively fluorinated, and the binding of the cation is progressively diminished. Remarkably, despite the complexity of the biological binding site, a linear free energy relationship is often seen.^20,21

Taken together, gas phase studies of the cation−π interaction provide a clear picture of this important noncovalent interaction. Electrostatic effects are paramount, and relatively simple models can provide excellent guidance for the experimentalists. In particular, rational ways to modulate the cation−π interaction through substitution are evident and have been frequently exploited by experimentalists.

3. Cation–π Interactions in Water

As noted above, Kebarle established in 1981 that, in the gas phase, K⁺ binds more tightly to benzene than to water, presaging thousands of observations that cation−π interactions can be energetically consequential in aqueous solution, and hence in biological systems. An immediate question was whether cation−π interactions could indeed contribute to binding small molecules in water at natural or designed binding sites. We discuss several aspects of this question in this section, but we begin with some general issues regarding aqueous solvation.

3.1. Basic Considerations–the Desolvation Penalty

Of course, simple ion pairs are fully dissociated in water–water is an excellent solvent for ions. In a protein context, salt bridges that are on the surface of a protein, and hence substantially water solvated, are not tightly bound and contribute little or nothing to protein stability. More buried salt bridges can be energetically significant.

Consider the fundamental differences between forming a salt bridge—R–CO₂^–···H₃N⁺–R—vs forming a comparable cation−π interaction—C₆H₆···H₃N⁺–R—in water. Again, if the salt bridge dissociates, both partners experience very favorable aqueous solvation that, energetically, outweighs any benefit from forming the ion pair. Upon dissociating a cation−π interaction, however, the cation experiences very favorable solvation, but the benzene does not. Benzene is hydrophobic, and so to first order does not interact favorably with water. One could argue that placing an ion over the face of benzene could be considered to have favorable solvation consequences, in that the hydrophobic entity has less surface area exposed to water. In fact, the C₆H₆···H₃N⁺–R unit can be expected to have quite favorable aqueous solvation.

A more subtle feature concerns long-range solvation. An ion dissolved in water experiences a great deal of stabilization due to long-range interactions with water molecules that are well separated from the ion, due to reorienting water dipoles that, on average, interact favorably with the ion (Born solvation). When two ions come together to form a tight ion pair, such long-range solvation would be diminished, since, from a distance, a water molecule sees both positive and negative charges. Stated differently, when two ions come together, they, in a sense, neutralize each other from the perspective of water molecules that might be 10–15 Å away. However, when an ion binds benzene, the charge is still fully present–a cation−π complex can enjoy long-range water solvation the same as an ion can. All these arguments suggest that simple cation−π interactions could be meaningful even in the presence of full, aqueous solvation. In addition, in many cases it may be that when a molecular ion enters a biological (or artificial) binding site, the full solvation energy of the ion may not be lost. The desolvation penalty will still be substantial, but it may not be equal to the full hydration energy of the dissociated ion.

Just to get a sense of the potential magnitude of the desolvation penalty in water, calculations show that NH₄⁺ is better solvated than CH₄ by ∼96 kcal/mol; (CH₃)₄N⁺ (tetramethylammonium, TMA) is better solvated than (CH₃)₄C by ∼70 kcal/mol.²² Again, these are upper limits to the desolvation penalty, but they are large numbers.

Concerning a cation−π interaction vs a salt bridge, early calculations clearly indicated that a cation−π interaction (C₆H₆···NH₄⁺) would be viable in water, but a comparable salt bridge (CH₃CO₂^–··· NH₄⁺) would not.²³ More recent DFT studies with a standard PCM water model conclude that biologically relevant cations (ammonium, TMA, guanidinium) bind to prototype aromatics (indole, phenol) with ∼4–7 kcal/mol of binding energy.²⁴ We will see throughout this review the special role of Trp in cation−π interactions.²⁵ A recent experimental study of cation−π interactions in water confirmed the stronger cation−π interaction with Trp vs Tyr. As described in Section 5, when two mica surfaces are coated with polymers and brought together in water, a surface force apparatus can measure the attraction between the two surfaces. In the present case, one surface was coated with a polyaromatic, the other a polycation such as polylysine. In all cases, the polyindole-coated surface (mimicking Trp) showed stronger adhesion to the polycation than the polyphenol-coated surface (mimicking Tyr).²⁴

3.2. Binding of Alkali Metal Ions in Water: The Unique Role of K⁺

While the binding of Li⁺, Na⁺, etc. to benzene in the gas phase is understandable, and model systems have revealed such interactions in organic solvents and the solid state, the notion that cation−π interactions involving alkali metals in water may seem far-fetched. However, a variety of recent studies have clearly established cation−π interactions with alkali metals in water. These studies highlight the unique nature of the cation−π interaction in water, with interesting implications for materials science and other areas.

Again, the cation−π interaction follows a classical electrostatic trend in the gas phase: Li⁺ > Na⁺ > K⁺ > Rb⁺ (Figure 1A). In 1993 we published a computational study of prototype cation−π interactions in water.²⁶ The study was prompted by emerging information on biological ion channels that are highly selective for K⁺. In particular, the channels have several very highly conserved aromatic amino acids in a region that was thought to be near the selectivity filter, and we and others wondered whether a cation−π interaction could contribute to ion selectivity. The computations produced a surprising outcome. For M⁺···C₆H₆ complexes in water, K⁺ binds better than Na⁺ or Rb⁺, while Li⁺ is preferred over K⁺, although the bias is quite small compared to that in the gas phase. For C₆H₆···M⁺···C₆H₆ complexes, K⁺ is preferred over all three ions in water.

In studies like this, one must always consider a competition between two forces. One is the solvation of the ion, or rather the cost of desolvating the ion so it can bind. The other is the intrinsic binding energy of the ion. Clearly Li⁺ has the strongest cation−π interaction but also the largest water solvation (hydration) energy. Apparently, K⁺ strikes a particular balance, in which the cost of desolvation is not too high, yet the cation−π interaction is still strong. To be clear, there are many examples of trends like this, and the results are reminiscent of, and no doubt related to, the Hofmeister series. We believed these calculations from 30 years ago were fundamentally sound, and we will see below much experimental and modern theoretical support for them. Concerning prototype biological K⁺ channels, however, MacKinnon’s groundbreaking 1998 structure of the KcsA bacterial channel²⁷ showed that the cation−π interaction does not contribute to the remarkable ion selectivity of these channels. Recently, however, it has been proposed that an aromatic amino acid near the outer mouth of some K⁺ channels does facilitate outward K⁺ currents.²⁸ In addition, more recent studies of different ion channels and transporters do implicate a role for cation−π interactions in channel selectivity (see below).

Computationally, more recent and more sophisticated studies show results similar to our early work. For example, extensive, high-level calculations—both DFT and MP2(full)—by Reddy and co-workers probed the effect of successively adding water molecules to a M⁺···C₆H₆ system.²⁹ Initially (1 to 3 waters), the gas phase trends hold up. However, the (H₂O)₇··· M⁺···C₆H₆ system shows a new trend, with K⁺ binding tighter than Li⁺ or Na⁺. The reason for the switch is clear. With 7 water molecules, the Li⁺ and Na⁺ complexes show a fully hydrated ion docked on the face of the ring, with the ion making no direct contact with the benzene. However, in the K⁺ complex the ion is partially desolvated, and the K⁺ makes direct contact with the benzene ring. This leads to a stronger binding interaction. Clearly, the balance between desolvation and binding is at play here. With K⁺, the cation−π interaction is strong enough to induce partial desolvation of the ion, allowing for a direct cation−π interaction. With Na⁺ and Li⁺ the solvation is too strong.

While being intellectually interesting, these studies have important implications in many areas, but perhaps none more so than in studies of graphene and other carbonaceous materials. There have been extensive studies of possible cation−π interactions to graphene, graphene oxide, single walled carbon nanotubes and similar structures. Alkali metal ions bind to these carbon structures, often with interesting consequences. As implied by the results of Figure 1 and the extensive studies of Sastry, larger aromatic systems bind ions more tightly than benzene and other simple aromatics. Mu et al. performed DFT calculations on a C₈₄H₂₄ model of graphene and found gas phase binding energies of 54, 40, and 31 kcal/mol for Li⁺, Na⁺, and K⁺, respectively.³⁰ On adding waters, a similar trend to that mentioned above is seen. Once enough water is added, K⁺ binds most tightly, and, again, the reason is that the K⁺ is partially desolvated and can make direct contact with the graphene surface. A later study on the graphene model showed that a Ca²⁺ ion can become partially desolvated and thus make direct contact with the graphene surface, but Mg²⁺ cannot.³¹ An implication of this work is that when graphene and similar materials are exposed to aqueous salt solutions, we can anticipate selective aggregation of certain ions at the surface, and this could affect the materials and electronic properties of the graphene.

Studies of graphene oxide membranes (GOMs) also show interesting ion-specific effects. Bathing a GOM in a salt solution leads to swelling, which affects the materials and electronic properties of the GOM. The interlayer spacing depends on the identity of the ion. Interestingly, K⁺ produces the narrowest interlayer spacing, compared to Na⁺, Ca²⁺, Li⁺, Mg²⁺ or even pure water. Remarkably, when the GOM is exposed to a solution with two salts, one being KCl and the other being either NaCl, CaCl₂, LiCl, or MgCl₂, the interlayer spacing is always the same as that seen with KCl alone.³² Clearly, intercalated K⁺ is preventing the other ions from entering the interlayer space. Again, the notion of an altered hydration structure for K⁺ within the GOM is invoked to explain these results. In a similar manner, in the studies described in Section 3.1 that measure the force needed to separate a polylysine surface from a polyaromatic surface, the adhesion is greatly diminished when K⁺ ions are added to the solution, but Li⁺ and Na⁺ ions have little or no effect.²⁴

A similar pattern is seen for interactions of ions within narrow carbon nanotubes (CNTs).³³ Again, K⁺ is partially desolvated and makes direct contact with the walls of the CNT, while Na⁺ does not (nor does Cl^–). Several review articles highlight the role of cation−π interactions in modulating the properties of carbonaceous materials.³⁴

A remarkable consequence of the special interaction of K⁺ with carbonaceous materials involves the interaction of single-layer graphene (SLG) with neurons.³⁵ When SLG is brought into contact with cultured neurons, neuronal excitability is affected, such that there is an increase in cell firing. Additionally, neurons exposed to SLG upregulate K⁺ currents. The authors propose a key role for cation−π interactions, which cause K⁺ ions to accumulate at the graphene surface. This local, altered ratio of ions at the neuron causes a change in the neuronal excitability from what is seen in the absence of the SLG.

Experimental support for the reversal of the order of cation affinities also comes from novel studies of hydrogel swelling.^36,37 A hydrogel comprised of a random copolymer of monomers that contain a quaternary ammonium ion or a phenyl ring was prepared. The expectation was that the ion and aryl would experience extensive cation−π interactions. The hydrogel was then exposed to aqueous salt solutions, and the molar ratio of metal ions absorbed was measured as an indication of binding affinity of the cation. The order seen was Cs⁺ > K⁺ > Rb⁺ > Na⁺ > Li⁺. Again, the gas phase binding order was completely upended in aqueous solution.

We began this section with a reference to biological K⁺ channels and the possibility that a cation−π interaction could be involved in ion selectivity. Alas, Nature did not choose this path for canonical K⁺ channels. However, the story does not stop there. Matile and co-workers prepared synthetic ion channels comprised of p-septiphenyl units and showed they produced an ion selectivity sequence of K⁺ > Rb⁺ > Cs⁺ > Na⁺ > Li⁺.³⁸ Xin et al. prepared synthetic ion channels from tetrarylmethanesulfonates linked to para-substituted bis(diaminomethyl)phenyl groups, producing structures with 4-fold symmetry (natural K⁺ channels are tetramers). The channels show high cation conductance, with selectivity rations for K⁺/Li⁺ and K⁺/Na⁺ of 360 and 31, respectively.³⁹ In both model systems, cation−π interactions were invoked to explain ion selectivity.

Perhaps most excitingly, K⁺ selectivity induced by cation−π interactions has been seen in naturally occurring channel proteins. The KCR channel rhodopsins are K⁺-selective, light-gated ion channels that provide valuable optogenetic tools. A recent crystal structure has revealed the origin of the selectivity. Based on the structure and molecular dynamics (MD) simulations, the authors describe a “selectivity triad” comprised of a tryptophan, a tyrosine, and an asparagine.⁴⁰ The Trp makes a cation−π interaction, while the OH of the Tyr and the carbonyl of the Asn also bind the K⁺. The K⁺ is almost fully dehydrated (two coordinating water molecules remain) when it occupies the selectivity triad. A Na⁺ ion is not accommodated by the selectivity triad, at least in part because of the diminished strength of the aqueous cation−π interaction for Na⁺. Another study evaluated a series of homologous KCRs, most of which are K⁺ selective, but a few homologues showed Na⁺ selectivity. A key Trp residue that was present in the K⁺-selective channels and absent in the Na⁺-selective channels was considered to rationalize the different selectivities, via a cation−π interaction.⁴¹

Another relevant study involved microbial rhodopsins, which are light-driven ion transporters. KR2 is a Na⁺-transporting rhodopsin that shows good selectivity against K⁺ and Cs⁺. Two mutations were found to markedly increase K⁺ conductance. One replaced an Asn by a Phe at the entrance to the channel. Another (an Asn-to-Leu mutation) led to a structural rearrangement that flipped a Phe residue into position so it can make a cation−π interaction with passing ions.⁴²

One more comment about cation−π interactions of simple ions in water is in order. Divalent ions such as Mg²⁺ and Ca²⁺ have much higher hydration energies than the monovalent ions we have been considering: Na⁺ is 98 kcal/mol and Mg²⁺ is 476 kcal/mol. While Mg²⁺ certainly binds strongly to benzene in the gas phase (Figure 1), it is inconceivable that a divalent ion would give up its full hydration shell in order to bind a benzene in water.²⁹ However, when surrounded by a single aqueous solvation shell, with the oxygens of the waters pointing toward the ion, the surface is made up of hydrogens, which certainly carry significant positive charge and can make cation−π interactions. Indeed, Mg²⁺(H₂O)₆ shows a cation−π interaction in the gas phase comparable to that of K⁺ (Figure 1).

3.3. Binding of Small Organic Molecules in Water

Section 3.2 focused on alkali metals as prototypes for the cation−π interaction. The novel behaviors of these ions in water have implications across materials science and other areas. But of course, the overwhelming majority of cation−π interactions in water are not to alkali metals, but to cationic organic molecules, frequently protonated amines and quaternary ammonium ions. Examples abound both in natural biological receptors and in model systems.

Model systems have always played an important role in evaluating cation−π interactions. Arguably, cyclophane studies from our laboratories in the 1980s were the first demonstration of cation−π interactions in water. Since then, a wide range of artificial systems have been developed, often with a goal of determining the energetic magnitude of a cation−π interaction. Many take the form of molecular “torsion balances” or “seesaw balances” and “lariats”.^43,44 Among other things, these approaches have probed substituent effects, solvent effects, and counterion effects, while also providing a measure of the energetic strength of prototype cation−π interactions. There is considerable variation across the work, with some systems showing interaction energies of 2–3 kcal/mol, while other systems give smaller values. Context matters, as does the method of measurement and the choice of reference state. Again, previous reviews have amply documented these studies with model systems. Here we will focus on more recent examples that we feel are illustrative of general principals or involve important biological systems.

Stepping back for a moment, it is amazing that an ion will leave water and dock into an undeniably hydrophobic binding site, as long as that binding site is not fully aliphatic but rather contains π systems. Also, it should not be surprising, and has been understood from the very beginning,⁴⁵ that hydrophobic molecules—often neutral isosteres of cationic ligands—will bind at “cation−π” sites. Many studies have probed this particular issue.

Almost 40 years ago we compared the binding of N-methylquinolinium vs the isosteric 4-methylquinoline to a cyclophane artificial receptor (Figure 3A).^22,46 In this case, the cation was more tightly bound by ∼2.5 kcal/mol, even though the hydration energy of the cation was greater than that of the neutral by ∼46 kcal/mol.

Figure 3.

Examples of isostructural comparisons used to evaluate the role of hydrophobicity vs electrostatic attraction in cation−π binding sites. A. Structures compared in the cyclophane system developed by Dougherty. B. Analogues used on the model calix[4]pyrrole system developed by Hunter. C. Trimethyllysine and tert-butyl norleucine, which were compared by Waters and Houk in scores of natural proteins that recognize methylated lysines on chromatin proteins.

Much more recent studies by Hunter and co-workers with a different model system—a synthetic calix[4]pyrrole receptor that would be anticipated to make cation−π interactions to encapsulated guests—produced a different result. In two separate studies they asked whether neutral isosteres of cationic guests could bind at the “cation−π” sites. One study contrasted trimethylammonium guests to tert-butyl analogues,⁴⁷ and another study⁴⁸ contrasted N-methylpyridinium cations to neutral toluene or pyridine N-oxide guests (Figure 3B). In both studies the neutral guests tended to bind more tightly than the cationic by 2–3 kcal/mol.

Similar comparisons have been made with biological receptors. Diederich studied the binding of inhibitors to the protease enzyme Factor Xa and showed that quaternary ammonium ions (RNMe₃⁺) bound well to the S4 pocket. Neutral analogues (RCMe₃) showed a 100-fold drop in affinity, indicating a 2.8 kcal/mol preference for the cation vs the neutral.⁴⁴ Note that the Factor Xa binding site contains an “aromatic box” (see below) comprised of three aromatic side chains. It could be argued that the energetics imply each cation−π interaction to an individual side chain is worth ∼0.9 kcal/mol. Other studies have found similar numbers for a cation−π interaction to an individual ring.

Recently, Waters, Houk and co-workers have thoroughly probed the same issue in native biological recognition.⁴⁹ One of the most important cation−π interactions in biology involves so-called chromatin reader proteins. An important posttranslational modification of chromatin proteins is the methylation of lysine, typically to make trimethyllysine (KMe₃⁺), although other degrees of methylation are also seen. Of course, KMe₃⁺ is a quat, and extensive studies have shown that reader proteins recognize KMe₃⁺ through cation−π interactions, and the binding is involved in regulating transcription. Again, an “aromatic box” (see below) is seen in these proteins, even in systems that share no sequence/structural similarity. Extensive studies by Waters showed that cation−π interactions were essential to binding KMe₃⁺.⁵⁰

In the recent study,⁴⁹ roughly 200 KMe₃⁺ reader proteins were evaluated for their ability to bind the tert-butyl analogue of KMe₃⁺—tBuNle—vs KMe₃⁺ itself (Figure 3C). There was a range of selectivities, with some reader proteins showing a >120-fold preference for KMe₃⁺, while others showed a much smaller but nearly 7-fold preference for tBuNle. Interestingly, the two classes show different thermodynamics of binding. Those that are driven by the cation−π interaction show enthalpically favorable binding, while sites that favor tBuNle show a significantly reduced enthalpic driving force and a favorable entropy of binding. This is exactly what one would expect if KMe₃⁺ binding is driven by an attractive noncovalent binding interaction—the cation−π interaction—while the binding of tBuNle is driven by hydrophobic forces. In a very impressive test of the analysis, the authors progressively altered the electrostatic binding potential of key aromatic residues in the binding site and showed that at a given binding site, the affinity for KMe₃⁺ was progressively diminished as the electrostatics on the aromatic ring became less favorable, while the binding of tBuNle was unaffected. This beautiful study convincingly demonstrates that a natural cation−π binding site is, of course, hydrophobic, but that in addition to hydrophobic forces, at many binding sites there is an enthalpically favorable binding to cations–a cation−π interaction.

Overall, studies of cation−π interactions in water reveal a potent binding force that is manifest in many different contexts. In addition, some surprising results arise from the balance between the cation−π interaction and aqueous solvation. While in any particular system results may vary, most workers have come to the view of Diederich that “the cation−π interaction is one of the strongest driving forces in biological complexation processes”.⁴⁴

4. Cation–π Interactions in Organic Synthesis/Chemical Biology

For many years after the development of the cation−π interaction, with countless examples from biology being reported, exploitation of the cation−π interaction for organic synthesis and catalysis was limited. In the past decade the situation has changed dramatically, and there are now many examples. A common theme is that a cation−π interaction favors a specific conformation or bimolecular association that then influences reaction stereochemistry. There is now a quite large body of work, and here we rely heavily on several excellent recent reviews.⁵¹⁻⁵³ Prototype reactions are shown in Figure 4.

Figure 4.

Representative synthetic reactions in which cation−π interactions influence reactivity and/or selectivity. Cation−π interactions are denoted by red dashed lines; hydrogen bonds by black dashed lines. Ovals represent arenes. A. A Schmidt reaction featuring a cation−π interaction between a diazonium and an arene.⁵⁴ B. Prototype iminium chemistry that benefits from an intramolecular cation−π interaction.⁵⁵ C. A nucleophilic addition to a pyridinium, one face of which is blocked by an intramolecular cation−π interaction.⁵⁶ D. An example of the remarkably effective thiourea catalysts developed by Jacobsen.⁵⁵ The large arene on the left can be varied systematically, revealing trends that support the role of cation−π interactions. Also shown is a schematic of catalysis of episulfonium ring opening. The thiourea complexes an anion through hydrogen bonds; a carbonyl hydrogen bonds to the nucleophilic indole; and cation−π interactions between the sulfonium and the arene (oval) orient the substrate, allowing control of stereochemistry.

Jacobsen emphasizes the role of cation−π interactions in small-molecule catalysis and distinguishes several classes of reaction.⁵¹ The first involves a cation−π interaction contained within the reaction substrate. Lewis acid-promoted reactions are common here, forming a cation that can interact favorably with an arene on the substrate. Schmidt-type reactions involving a diazonium group are also common (Figure 4A). A second class involves “native cationic catalysts”, including reactive species such as thiazolium salts, guanidinium ions, oxoammonium groups, and heterocycles involved in acyl transfer reactions. These cationic components of the catalyst make a cation−π interaction with an arene of the catalyst, imparting a specific shape that leads to control of reaction stereochemistry. Iminium ions formed in organocatalysts developed by MacMillan and others can experience similar interactions (Figure 4B).

The third class involves a cation−π interaction between the catalyst and the substrate, and a number of fascinating examples exist. Whether it is an alkali metal sandwiching between two arenes in a diarylcyclopropane, the cation−π interaction of a imminium ion to one face of an arene influencing reaction stereochemistry (Figure 3C), or the binding of tetrabutylammonium to an arene also with stereochemical implications, bimolecular cation−π interactions are seen to influence a range of reactions. Acyl transfer reactions are prominent, and S_N2 reactions of the Mentschutken class that form pyridinium-type products are accelerated by cation−π interactions. The formal reverse reaction—S_N2-type dealkylation of sulfonium compounds—is also accelerated by cation−π interactions in a manner mimicking biological alkylations involving S-adenosylmethionine (SAM). A number of examples invoking cation−π interactions involve Jacobsen’s remarkably versatile organocatalysts derived from arylpyrrolidino thioureas and other thioureas. Catalyzed reactions include polycyclizations, episulfonium ring openings, alkylation of isoquinolines, hydroamination of olefins, chlorination of silyl ketene acetals, and selenoetherification of olefins. It is clear that the link between cation−π interactions and small-molecule catalysis is a strong one.

The very extensive review by Yamada⁵² includes hundreds of synthetic chemistry examples that involve cation−π interactions, and again reactions are grouped according to the role the cation−π interaction plays. Reactions that involve a cationic intermediate or transition state include: nucleophilic additions to pyridiniums and related structures; conjugate additions; Friedel–Crafts reactions; S_N1/S_N2 reactions; a range of cycloadditions; nucleophilic acylation catalysis; and sigmatropic rearrangements. In another class, a cation, typically an onium ion, assists a reaction by complexing to a reactant or intermediate, blocking a face of the structure and thereby influencing reaction stereochemistry. Examples include: nucleophilic additions and substitutions; cyclization of alkynes; photochemical reactions, both in solution and in the solid state; template-directed syntheses; and dynamic combinatorial synthesis. A remarkable collection of reactions that are metal-cation assisted are seen, including Li⁺, Na⁺, K⁺, and Cs⁺. Also, many examples of supramolecular catalysis are presented, in which reaction occurs in supramolecular receptors and self-assembled cages.

We have noted the special role that Trp plays in biological cation−π interactions, and very recently a remarkable piece of synthetic chemistry/chemical biology has been reported by Chang, Toste, and co-workers.⁵⁷ The goal was to develop reagents that react specifically with Trp residues in a biological context, much as there are reagents that are specific for Cys and Lys residues. Combining a strategy that mimics an indole-based alkaloid biosynthetic path with theoretical modeling, they identified oxaziridine reagents that show remarkable reactivity (Figure 5). The reaction shown, termed Trp-CLiC, is extremely selective for Trp over the other canonical amino acids. Through the R groups shown, Trp-CLiC allows the delivery of a range of probes and reactive structures to biological macromolecules. Remarkably, the chemistry shows a strong preference for modifying Trp residues that are involved in cation−π interactions. This includes a “privileged” cation−π interaction motif: WxxxK. These new chemical tools were used to probe the important role that Trp-mediated cation−π interactions play in phase separation (see Section 5.2 below).

Figure 5.

Basic chemistry of Trp-CLiC chemistry.

Very recently another example of Trp-specific chemistry has appeared, in this case focused on the reader proteins discussed above for trimethyllysine modifications on chromatin.⁵⁸ Such proteins generally contain an aromatic box, and Trp residues are common. Wu and co-workers reasoned that a sulfonium ion mimic of trimethyllysine would also bind to such an aromatic box. Exposure to UV light leads to single electron transfer and, ultimately, covalent modification of the Trp residue (Figure 6).

Figure 6.

Modification of trimethyllysine binding sites by a sulfonium analogue.

It is clear that cation−π interactions can be rationally employed to control reactivity and, especially, reaction stereochemistry. We can anticipate a steady stream of new examples across a broad range of reaction types.

5. Multivalency/Cooperativity in Cation–π Interactions

While many have argued that the cation−π interaction is one of the strongest noncovalent interactions, it is still not a covalent bond; it is a weak interaction. However, as with other weak or moderately strong interactions, combining multiple cation−π interactions can produce very large effects. Of course, multivalency is a common theme in biology, and examples have been seen with cation−π interactions. Here we describe several systems in which multiple cation−π interactions produce large effects.

5.1. Mussel Adhesion Proteins and Related Systems

Mussels, barnacles, and other marine organisms have developed a remarkable strategy to adhere to various surfaces when immersed in seawater. When a crevice is encountered, the organism creates a sort of vacuum chamber by forcing out the air and water using suction. The byssus of the organism then injects into the crevice a group of mussel foot proteins that are highly adhesive and allow the organism to remain attached even under strong tidal forces. These foot proteins are very rich in cationic and aromatic amino acids, and it is clear that cation−π interactions play a critical role in their function.^59,60 Early work emphasized the noncanonical amino acid DOPA, as it is enriched in many adhesive proteins. It was proposed that the adjacent hydroxyl groups on the aromatic ring were directly involved in adhesion, or perhaps they chelated metals that facilitated adhesion. However, more recent work has downplayed the importance of DOPA. For example, DOPA is found minimally or not at all in highly adhesive proteins of barnacles and other organisms. Conversely, DOPA is quite prevalent in nonadhesive coating proteins of some marine organisms. In addition, DOPA is oxidatively unstable, and when oxidized it loses its adhesive ability.

More recent studies have firmly established the major role that cation−π interactions play in these adhesion proteins. Many studies involve coating an inert surface such as mica with polymers/peptides that are enriched in cationic and aromatic moieties. Two complementary surfaces are then brought together in water, and a surface force apparatus measures the force required to pull them apart as an indicator of adhesive strength. For example, a naturally occurring DOPA-deficient adhesion protein was shown to be just as effective as DOPA-containing proteins.⁶¹ In another study, a protein that contained DOPA but also other aromatic residues was still quite adhesive under conditions in which the DOPA was expected to be oxidized.⁶⁰ A recent systematic study considered a series of synthetic peptides in which one surface was coated with an aromatic moiety—Phe, Tyr, DOPA, or, as a control, Leu—and the other surface was coated with polylysine.⁶² The adhesive ability of the Leu protein was minimal. The results with the other aromatics were surprising. DOPA produced the weakest adhesive, with Tyr a bit better. However, the Phe-containing peptide was much more adhesive, ruling out a special role for DOPA and strongly supporting a cation−π model. Solid state NMR studies also supported the formation of cation−π interactions. Interestingly, in many of these studies adding K⁺ ions to the solution significantly weakened adhesion, but Na⁺ and Li⁺ did not, consistent with the work described in Section 3.2.

Other workers have developed hydrogels and related materials, the properties of which are substantially impacted by cation−π interactions. For example, in aqueous environments many solid surfaces are negatively charged, including rocks, glasses and metals. However, developing materials that will bind to such surfaces through conventional electrostatic forces is challenging, especially in saline water, in which electrostatic interactions are attenuated by the high ionic strength. Gong and co-workers found that in copolymers of cationic and aromatic groups, cation−π interactions between adjacent monomers enhance the electrostatic interaction with the surface, allowing for strong, reversible adhesion to negatively charged surfaces in seawater.³⁶ The authors also propose that a precomplexation of cationic and aromatic monomers in the radical polymerization leads to an enhancement of adjacent cation-aromatic monomers in the polymer, which facilitates adhesion.⁶³ Yan and co-workers found that incorporating cation−π interactions into double-network elastomers enhanced the compressive strength and toughness, leading to high impact resistance.⁶⁴ A recent clever design took advantage of the high acidity of the gastric environment.⁶⁵ Polymers with protonatable amines and aromatic moieties formed cation−π interactions in an acidic environment, allowing the preparation of self-healing hydrogels for gastric perforation repair.

Integrins are dimeric cell adhesion proteins that mediate cell–cell, cell–matrix, and cell–pathogen interactions, including adhesion of leucocytes. A number of studies have established that cation−π interactions play an important role in this process.⁴ For example, Pan et al. showed that a cation−π interaction is critical to ligand binding and signaling in integrin α4β7.⁶⁶

5.2. Phase Separation

Over the past ∼15 years it has become evident that cells contain a number of membraneless compartments, variously referred to as granules, puncta, droplets, P granules, stress granules, nucleoli, etc., but now known as biomolecular condensates.⁶⁷⁻⁷¹ These structures contain a range of biological macromolecules that associate and carry out crucial cellular functions. Malfunction of biomolecular condensates is associated with a number of diseases. It is clear that cation−π interactions play a key role in forming biomolecular condensates and in modulating their properties. The phrase liquid–liquid phase separation (LLPS) is associated with this phenomenon, and while the macromolecules are clearly associated strongly, the lack of a membrane allows exchange with the cellular medium and the rapid influx/efflux of small molecules and ions. In particular, cancer therapeutics and other small molecules have been shown to selectively partition into these condensates, opening up interesting new therapeutic strategies.⁷² The physical chemistry associated with phase separation is a topic of intense current investigation. However, it is clear that, in many systems, cation−π interactions play an essential role.

A prototype and well-studied system involves the RNA-binding protein FUS (fused in sarcoma), which is found in ribonucleoprotein granules and other biomolecular condensates.⁷³ As in other systems, phase separation in FUS is driven by an intrinsically disordered low complexity domain of the protein. It has been convincingly demonstrated that cooperative cation−π interactions between tyrosines in the low complexity domain and arginines in the structured C-terminal domain of FUS play an important role in phase separation. FUS has 27 Tyr in the low complexity domain and 37 Arg, mostly in the C-terminal region. Arginines are very much over-represented in biomolecular condensates in general,⁶⁹ and an interesting feature of Arg is that it frequently undergoes posttranslational modification. In particular, Arg residues can be methylated to varying degrees on the guanidinium group of the side chain. In the FUS system it is clear that the methylation state of the Arg residues in the C-terminal domain influences biomolecular condensate formation. For example, hypomethylation encourages formation of stable β-sheets rather than phase separation, and it is associated with FUS-associated frontotemporal lobar degeneration (FTLD). Similar biomolecular condensates are thought to play a key role in the pathology of amyotrophic lateral sclerosis (ALS).⁷⁴

Another well-characterized system involves Ddx4, a DEAD-box helicase that is a major component of nuage (germline granules that contain DNA).^70,75 This DNA-binding protein contains a low complexity domain. Again, cation−π interactions play an important role in phase separation, and Arg methylation influences the process.

Characterization of biomolecular condensates can be challenging, but a consensus is emerging that cation−π interactions frequently play an important role. As noted in a recent review: “cation−π interactions are demonstrably important to phase separation in many commonly studied proteins... (and) are also important for tuning the material and transport properties of condensates”.⁶⁸ We can anticipate much further work in this fascinating area.

6. Select Studies of Cation–π Interactions in Biology

6.1. Well-Established Cation−π Interactions Involving Proteins

A major goal of this work has been to describe examples of cation−π interactions in systems that perhaps have not received as much attention in previous reviews. In contrast, cation−π interactions in proteins are well established and vast in number. This includes stabilizing protein secondary structure, ligand binding, and protein–protein interactions. Here we highlight select systems, some of which have been discussed in previous reviews, that provide especially compelling examples and/or have special biological significance.

6.1.1. Interactions within Proteins

Within a protein, cation−π interactions can form between Lys/Arg and Phe/Tyr/Trp. Histidine can also be considered, either as the cation if the side chain is protonated, or as the π if not. It is not always straightforward to ascertain the protonation state of a His side chain from a protein crystal structure. As such, broad analyses of the role of cation−π interactions in proteins generally do not include His, and that will be the case here.

Comprehensive studies of cation−π interactions contributing to protein secondary structure via Lys/Arg···Phe/Tyr/Trp interactions have appeared.^16,76 Such interactions are common, such that there are over 1,000,000 such interactions in the Protein Data Bank (based on a finding of 1 cation−π interaction for every 77 residues in the PDB).¹⁶ In addition, cation−π interactions of this sort are more common in proteins from thermophiles, suggesting that cation−π interactions generally enhance protein stability.⁷⁷

A spectacular example is the chain of six consecutive cation−π interactions in the erythropoietin receptor shown in Figure 7. Very similar motifs are seen in human growth hormone receptor and human prolactin receptor.⁷⁸

Figure 7.

Extended cation−π system found in the crystal structure of the erythropoietin receptor (PDB 3hhr).

Surveys have also indicated that cation−π interactions are important contributors to protein–protein interfaces.⁷⁹ An interesting recent example was seen in voltage-gated ion channels, which are comprised of multiple subunits. Structural studies revealed a “cation−π pocket” that spans the interface between the channel proper and an important chaperone protein.⁸⁰

6.1.2. Binding Endogenous Substrates

A wide range of key bimolecular recognition events in biology exploit cation−π interactions. The ubiquitous G protein-coupled receptors use cation−π interactions to bind signaling molecules such as ACh, biogenic amines and related structures.^4,81 Also, many transporters for small molecules employ cation−π interactions to bind their substrate. An example is a transporter for choline/ethanolamine, the dysfunction of which is implicated in several diseases, which uses cation−π interactions in ligand recognition.⁸² Also, a recent structure of a member of the betaine/carnitine/choline transporter family clearly shows cation−π interactions in binding.⁸³

A number of systems that bind biological “onium” ions make extensive use of cation−π interactions. Enzymes that use S-adenosylmethionine (SAM, a sulfonium ion) to methylate biomolecules invariably use cation−π interactions to bind SAM and facilitate catalysis.^14,84 The mRNA “cap” at the 5′ end of essentially all RNA transcripts is the cationic 7-methylguanosine. As pointed out by Quiocho, a “cation−π sandwich” between two aromatics is a universal feature of proteins that bind the mRNA cap. Figure 8 shows two such proteins—VP39 and eIF4E—which, despite “total structural dissimilarity”, show the same binding motif.⁸⁵ The essential role of cation−π interactions in recognizing methylated Lys residues in chromatin—an important component of transcriptional regulation—was discussed above.⁸⁶

Figure 8.

Two proteins binding the 7-Me-G mRNA cap: left, VP39 (PDB 1JSZ); right, eIF4E (PDB 1EJ1).

Extensive studies by Roberts and co-workers have convincingly documented cation−π interactions to phosphocholine headgroups in membranes and shown they are influential in many systems involving membrane proteins and transmembrane processes. In particular, the aromatic box (see below) has been described as a “specific phosphatidylcholine membrane targeting motif”.⁸⁷ As a specific example, phosphatidylinositol-specific phospholipase C, a virulence factor secreted by Gram-positive bacteria, has two cation−π binding sites for the choline headgroup of phosphatidylcholine (PC) that are critical for vesicle binding. Important cation−π interactions to PC have been seen in a number of other phospholipases.^4,88 More generally, Trp residues tend to accumulate at membrane interfaces.⁸⁹

Terpene biosynthesis involves extensive cationic rearrangements/cyclizations, and aromatic amino acids are prominent at enzyme active sites, controlling regiochemistry and stereochemistry.⁹⁰ Examples abound, including one study that included modulation of the electrostatic potential of aromatic side chains using the unnatural amino acid methodology noted above.⁹¹

6.1.3. Small-Molecule Binding: Drug–Receptor Interactions

Throughout pharmaceutical research, cation−π interactions have been employed to optimize drug–receptor interactions.^44,92,93 One system that has been extensively studied is the binding of a range of agonists, antagonists, and pharmaceuticals to members of the pentameric (Cys loop) class of neurotransmitter-gated ion channels,²⁰ including the following:

• Nicotine, varenicline (Chantix), cytisine (Tabex), and epibatidine binding to the nicotinic acetylcholine receptor (nAChR)

• Binding to the 5-HT₃ receptor by the natural agonist serotonin, and the pharmaceuticals ondansetron (Zofran), granisetron (Kytril)

• Binding of GABA to various GABA_A receptors and the RDL receptor

• Binding of glycine, β-alanine, and taurine to the glycine receptor

The role of cation−π interactions in the binding of inhibitors to the blood coagulation protein Factor Xa was discussed above.⁹⁴ Voltage-gated sodium channels are blocked by a number of structures, and systematic studies have shown that cation−π interactions are involved in blockade by tetrodotoxin,⁹⁵ lidocaine,⁹⁶ and Ca²⁺.⁹⁷

An interesting reversal of the usual pattern for small molecules binding to proteins is seen in TREK-1, a member of the two-pore domain ion channel family. Here, the drug that is binding is neutral and a cation−π interaction is made to a Lys of the protein.⁹⁸

6.2. The Aromatic Box

A common feature of biological recognition sites that employ cation−π interactions is that there is often more than one aromatic amino acid that shapes the binding site, frequently forming an “aromatic box” that surrounds the cation.^9,44 The prototype example is arguably the ACh binding site for the nAChR, as first revealed in the crystal structure of the highly homologous ACh binding protein (AChBP)⁹⁹ (Figure 9) and subsequently confirmed in many structures of full receptors. Many of the examples cited above contain multiple aromatics in the binding site. The box structure is associated with binding of ammonium ions, while flat substrates like 7-Me-G form more of a sandwich interaction. Not all cation−π binding sites involve multiple aromatics—a classic example is the ACh esterase, in which a single Trp positions the ACh for hydrolysis in a typical serine “protease” site.¹⁰⁰

Figure 9.

Prototype aromatic box of three Tyr and two Trp as revealed in the structure of AChBP (PDB 1i9b). The structure is a homopentamer, with four residues from one subunit (designated A) and one from an adjacent subunit (designated B).

It is not clear why multiple aromatic amino acids are common at biological cation−π binding sites. Perhaps a cation−π interaction to a single ring is just not enough. Perhaps a highly hydrophobic binding site is desired and can be shaped better by flat side chains like Phe, Tyr, and Trp rather than bulky hydrophobic residues like Val, Leu, and Ile.

It is challenging to sort out what individual rings contribute to binding in an aromatic box or similar binding site. We noted above the shallow distance dependence of the cation−π interaction. This means that when a cation inserts into a binding site defined by three, four, or even five aromatics, meaningful cation−π binding interactions can be made to all the rings even if van der Waals contact with each ring is not possible. Computational studies on prototype natural cation−π binding sites reveal that, indeed, many rings can contribute to the ion binding.⁹

6.3. Recognition of DNA/RNA

From the beginning it has been tempting to consider that cation−π interactions could be important in binding to the bases of nucleic acids. Sorting out that issue is complicated by a number of factors, including the overall negative charge of DNA/RNA, the alteration of the electrostatic potentials of the bases by the hydrogen bonding they typically engage in, and the likely energetic cost, in the case of DNA, of disrupting the base stacking. These potential complications perhaps slowed the progress in identifying clear-cut examples of cation−π interactions involving nucleic acids. However, it is now clear that such interactions can be important, and many laboratories have invoked cation−π interactions involving DNA/RNA. For proteins that bind to DNA/RNA, we can anticipate a special role for Arg, as the shape of the guanidinium group seems especially well suited to interacting with DNA/RNA bases while still also making important hydrogen bonds. We will see that this is the case.

One interaction that has received considerable attention is the “stair motif”, described by Rooman and co-workers.¹⁰¹ This is a tripartite interaction involving two successive bases along the B-DNA stack and a protein guanidinium side chain. Figure 10 shows a prototype structure. The Arg guanidinium makes strong hydrogen bonds with the Hoogstein face of G8 in the major groove of the DNA strand. This kind of hydrogen bonding interaction is quite common.¹⁰² Directly above the guanidinium is the base of G7, and it is clear that a favorable electrostatic interaction is expected. Guanine, as shown in Figure 10, has a very large dipole moment (>7D), the negative end of which is directed toward the Arg side chain. Recalling the shallow distance dependence of the cation−π interaction, the positive charge of the Arg side chain will certainly interact favorably with the negative end of the G dipole. Biot et al. performed high level calculations on this triad.¹⁰³ They concluded that, while the hydrogen bonding interaction is the largest contributor to the stability of the unit, the electrostatic interaction—a cation−π interaction—between Arg and G7 is a substantial contributor. The G-G stacking interaction is negligible.

Figure 10.

“Stair motif” for a cation−π interaction involving DNA. On the left is an image from the DNA binding domain of Tc3 transposase (PDB 1tc3), showing the Hoogstein hydrogen bonds from an Arg side chain to a G (the conventional DNA hydrogen bonds would come out toward the viewer) and the electrostatic interaction with the next G in the strand. On the right is the electrostatic potential surface for guanine in the same orientation, showing the large buildup of negative electrostatic potential that would interact favorably with the positive charge of the Arg.

The stair motif is fairly common. A survey of 52 high resolution crystal structures of protein/double-stranded DNA complexes found 77 stair motifs; 36 of the 52 structures had at least one.¹⁰¹ The stair motif has also shown selectivity for methylated DNA, and it is important in the function of the methyl-CpG binding domain (MBD) proteins, which are critical in epigenetic regulation.¹⁰⁴

The stair motif involves binding to double-stranded DNA. Also common is binding to a DNA base that has flipped out of the double helix or binding to RNA that is less structured.¹⁰⁵ For example, an interaction similar to the stair motif is seen in proteins that recognize O⁶-alkylguanine lesions.¹⁰⁶ The modified G is flipped out of the double helix, and the recognition protein provides an Arg that can bind to the alkylated G, flagging it for nucleotide excision repair (Figure 11). These enzymes are quite selective for O⁶-alkylguanine vs G. The authors conclude that the cation−π interaction is key to this selectivity, based on the contrasting electrostatic potential surfaces of the natural vs the modified base.

Figure 11.

A cation−π interaction involving a DNA base that has flipped out of the double helix and an Arg residue of a recognition protein. In this particular structure, the base is 2-aminopurine (2-AP). DNA backbone is in blue; protein backbone is in silver. PDB: 4HDU.¹⁰⁶

An interesting and important example involves recognition of RNA of the HIV Rev-responsive element (RRE) by the Rev protein, an essential aspect of viral replication.¹⁰⁷ In the RRE a U is flipped out, enabling binding of the Rev protein via an Arg that stacks on the RNA base. Another interesting recognition of a flipped-out U is seen in the spliceosome, which splices precursor mRNA (pre-mRNA).¹⁰⁸ A U at the 3′ end is sandwiched between a His and an Arg, with the latter certainly forming a cation−π interaction, while the exact nature of the interaction with His would depend on the protonation state. Actually, there is a string of four U’s at the 3′ end, and all four engage in a similar sandwich motif. Another splicing reaction, this one with tRNA as a substrate, also involves a cation−π interaction.¹⁰⁹ Structural studies reveal a “cation−π sandwich” in which two Arg residues bind a flipped-out A, positioning the substrate for catalysis. Binding to single-stranded DNA in telomeres has also been proposed to involve cation−π interactions.¹¹⁰

Finally, several guanidine RNA riboswitches show the guanidinium ion stacking on a base.¹¹¹ In addition, cations from solvent have been shown to make cation−π interactions with the faces of bases in unstacked conformations.¹¹² Tl⁺ has been observed to bind directly, while for Mg²⁺ and Ca²⁺ it is the first hydration shells that make the cation−π interaction, as anticipated in Section 3.2 above.

7. Conclusions

The present work has sought to generate an understanding of the broad range of fields impacted by the cation−π interaction, from physical organic chemistry to synthetic chemistry; from chemical biology to biochemistry to structural biology; from materials science to graphene and related structures. It is clear that the cation−π interaction is a universal and important binding force.

Looking forward, we can anticipate a steady stream of new examples and new contexts for the cation−π interaction. The interplay between solvation and the basic binding force will continue to be a focus for developing a fuller understanding of the effect. One particularly interesting topic is the role of multiple cation−π interactions in organizing biomolecular condensates. Since these systems are generally associated with intrinsically disordered regions of proteins, full characterization can be challenging, but it will be fascinating to learn the full energetic consequences of multiple cation−π interactions in assembling these structures.

Biography

Dennis A. Dougherty received a BS/MS from Bucknell University and a Ph.D. from Princeton University working with Kurt Mislow, and did postdoctoral work at Yale University with Jerome Berson. He then joined the faculty of Caltech, where he is now the George Grant Hoag Professor of Chemistry.

The author declares no competing financial interest.

Special Issue

Published as part of Chemical Reviewsspecial issue “Weak Interactions in Chemistry and Biology”.