Cucurbit[8]uril Binds Nonterminal Dipeptide Sites with High Affinity and Induces a Type II β-Turn

Abstract

In an effort to target polypeptides at nonterminal sites, we screened the binding of the synthetic receptor cucurbit[8]uril (Q8) to a small library of tetrapeptides, each containing a nonterminal dipeptide binding site. The resulting leads were characterized in detail using a combination of isothermal titration calorimetry, ¹H NMR spectroscopy, electrospray ionization time-of-flight mass spectrometry (ESI-TOF-MS), and X-ray crystallography. The equilibrium dissociation constant values determined for the binding of Q8 to nonterminal dipeptide sites Lys-Phe (KF) and Phe-Lys (FK) were 60 and 86 nm, respectively. These are to the best of our knowledge the highest affinities reported to date for any synthetic receptor targeting a nonterminal site on an unmodified peptide. A 0.79 Å resolution crystal structure was obtained for the complex of Q8 with the peptide Gly-Gly-Leu-Tyr-Gly-Gly-Gly (GGLYGGG) and reveals structural details of the pair-inclusion motif. The molecular basis for recognition is established to be the inclusion of the side chains of Leu and Tyr residues, as well as an extensive network of hydrogen bonds between the peptide backbone, the carbonyl oxygens of Q8, and proximal water molecules. In addition, the crystal structure reveals that Q8 induces a type II β-turn. The sequence-selectivity, high affinity, reversibility, and detailed structural characterization of this system should facilitate the development of applications involving ligand-induced polypeptide folding.

Graphical Abstract

INTRODUCTION

The targeted molecular recognition of polypeptides with synthetic ligands has driven many areas of biomedical science. Effective ligands are typically obtained by identifying small molecule leads from large library screens and optimizing for affinity, selectivity, and pharmacokinetics.¹ Such small molecule ligands fit within a cavity on the surface of a protein and therefore require their binding sites to be predominantly folded. By contrast, recognition of intrinsically disordered polypeptides using synthetic ligands is exceedingly rare.^2,3 Disordered regions of polypeptides are highly solvent exposed, and due to the lack of folding, their structural information is encoded simply into the sequence of amino acid residues. Therefore, targeting disordered regions of polypeptides has the potential to be predictive on the basis of the amino acid sequence. Toward this end, cucurbit[n]uril (Qn) synthetic receptors have shown great promise.

Qn receptors have received much attention over the past few decades for their unique supramolecular properties and myriad applications.^4–7 The n glycoluril units of Qn receptors linked by 2n methylene bridges form highly stable, aqueous-soluble structures that contain a weakly hydrated cavity accessible via two C=O lined entrances (Figure 1). These properties drive the binding of small molecule ligands via inclusion of nonpolar groups within the Qn cavity and electrostatic interactions between Qn C=O oxygens and ligand groups with a positive electrostatic potential.

Figure 1.

(a) Structural formula of cucurbit[8]uril (Q8). (b) Schematic illustration of a pair-inclusion complex of Q8 with a peptide containing nonterminal Leu-Tyr.

There have been numerous studies of the binding properties of polypeptides with Qn receptors.^8–17 Cucurbit[7]uril (Q7) and cucurbit[8]uril (Q8) recognize the amino acid residues Phe (F), Tyr (Y), and Trp (W) when located at the N-terminal position. Recognition is mediated by inclusion of the aromatic side chain within the Qn cavity and via ion-dipole interactions between the N-terminal ammonium group and C=O groups on the Qn.⁴ The smaller cavity size of Q7 can accommodate the side chain of a single aromatic residue.^10,11 The selective binding of Q7 at N-terminal phenylalanine translates to selective binding in complex mixtures of the proteins human insulin and human growth hormone, which each have an N-terminal Phe.^12,15,18 The larger cavity size of Q8 can accommodate two guests simultaneously. The reversible, noncovalent dimerization by Q8 of two FGG peptides (G is Gly) translates effectively to the noncovalent dimerization of proteins tagged with this sequence.^19–23 The ability of Q8 to bind two different but complementary guests, such as 2,6-dihydroxynaphthalene and methyl viologen,²⁴ has been applied to the noncovalent coupling of N-terminal tryptophan with methyl viologen or other electron-deficient guest,^25–30 as well as the coupling of N-terminal aromatic residues with N-terminal pentafluorophenylalanine.³¹ These homoternary and heteroternary complexes mediated by Q8 have enabled numerous applications, such as the reversible assembly of protein dimers,^19,20,23 oligomers,²² polymers,^32–35 aggregates,^36,37 and hydrogels.^38–47

While investigating the effects of sequence context on Q8•peptide binding, we discovered that Q8 can bind as a 1:1 complex with a peptide via the inclusion of a pair of neighboring side chains, e.g., Tyr-Leu (YL) (Figure 1). Using a semiempirical model with NMR-derived restraints, we predicted that the backbone would need to fold in order to accommodate pair-inclusion.¹³ Subsequently, we showed that this motif enables the recognition of N-terminal Met and entirely nonaromatic sites.¹⁴ Scherman and co-workers confirmed that binding occurs as a 1:1 complex, uncovered additional high-affinity sequences, and studied the chirality that is induced in the peptide upon Q8 binding.^16,17 The relative simplicity of this motif versus ternary complexes, the ability to target N-terminal methionine, and the greater potential for sequence control offered by two residues versus one have encouraged us to further explore the scope and limitations of the pair-inclusion motif.

An area of polypeptide recognition that is generally underexplored for Qn receptors is the binding at nonterminal positions—that is, at any residue other than the first or last. It is interesting that the highest affinity complexes have been observed at the N-terminal position.^11,13,16 Recognition of terminal sites can be advantageous due to the propensity of termini to be solvent accessible and disordered,⁴⁸ but there are also limitations. For example, there are only two terminal residues in any polypeptide chain, and one or both may be necessary for protein function or may be inaccessible to binding. Moreover, numerous post-translational modifications occur at protein termini (e.g., N-acetylation, N-myristoylation, and cleavage of N-Met by methionine aminopeptidase) that can remove or alter the N-terminal residue. Therefore, we are motivated to identify target sites for Q8 that do not contain the N-terminal residue.

Prior work has shown a few examples of cucurbit[n]urils binding to nonterminal sites on peptides and proteins.^{8,12–14,16,17,49,50} Q7 binds to nonterminal F and Y with equilibrium dissociation constant (K_d) values in the millimolar to high-micromolar range.^11,12,51,52 By contrast, Q8-induced peptide dimerization at nonterminal sites occurs at low micromolar concentrations in certain contexts,⁴⁹ but not in others.⁹ The pair-inclusion motif, however, is particularly promising for targeting nonterminal sites. Studies have reported a few measurable binding constants for nonterminal sites, such as YL, MY, and FL, but a systematic investigation of nonterminal binding in the pair-inclusion motif has not been reported until now.^13,14,16

RESULTS AND DISCUSSION

Design and Screen of a Peptide Library.

To investigate the binding of Q8 to nonterminal dipeptide sites, we designed a library of 64 peptides of sequence G–X₁X₂–G, in which X₁X₂ is the dipeptide target site and is flanked by G residues. The variable (X) residues were chosen to be aromatic (W, F, Y), hydrophobic (L, M, P), or cationic (K, R) based on prior work involving Q8 binding in the pair-inclusion motif (P is Pro; R is Arg).^13,14,16,17 The tetrapeptides were synthesized on Synphase Lanterns with Rink amide resin using Fmoc solid-phase chemistry, as described previously,^13,14,53 and yielded peptides containing N-terminal amine and C-terminal primary amide groups (see Supporting Information for experimental details).

The peptides were screened for binding to Q8 using a fluorescent indicator displacement assay with the indicator tetramethylbenzobis(imidazolium) (MBBI), as described previously.^13,14 The conditions used in the assay (20 μM Q8, 20 μM MBBI, 200 μM peptide, 10 mM sodium phosphate, pH 7.0) were optimized to give a large dynamic response using the reported K_d values for Q8 in complex with MBBI (1.2 μM),²⁶ AYL (3.1 μM),¹³ and AMY (6.1 μM)¹⁴ to estimate the lower end of the range of dynamic response, thus allowing us to observe stronger binders than these peptides. This buffer condition was chosen to match the buffer used in our prior studies in order to maximize the comparative analysis. In prior work, we observed that certain peptides competitively displace MBBI from Q8, resulting in an increase in the observed fluorescence emission intensity.^13,14 For peptides containing W residues, however, we have observed a decrease in fluorescence intensity upon binding, which we attribute to the simultaneous inclusion within the Q8 cavity of MBBI and the indole side chain of W.²⁶ In the current study (Figure 2), we observed an increase in fluorescence intensity ranging from 3 to 116% upon addition of most peptides to Q8•MBBI, but for seven peptides, all containing a W residue, we observed a decrease in fluorescence intensity ranging from 1 to 33%. One peptide induced no change in the Q8•MBBI fluorescence. All values are listed in Table S2.

Figure 2.

Fluorescent indicator displacement assay heat map. The changes in fluorescence of an equimolar mixture of Q8 and MBBI (20 μM) were measured upon the addition of 200 μM peptide of sequence GX₁X₂G in 10 mM sodium phosphate, pH 7.0. Color saturation indicates the intensity of change in fluorescence, in which dark red corresponds to a large increase in fluorescence and dark blue corresponds to a large decrease in fluorescence.

As anticipated from prior work with N-terminal dipeptide sites,^13,14 the peptides GYLG (95%), GLYG (101%), GFLG (106%), GLFG (99%), GYKG (85%), GKYG (97%) showed large increases in fluorescence. The peptides GFRG (107%), GRFG (94%), GLKG (101%), and GKLG (91%) also showed large increases that were not predicted from prior work. The largest change in fluorescence intensity was observed for peptide GFKG (116%), and its sequence isomer, GKFG, also showed a large increase (102%). Surprisingly, with the exception of GFMG (97%), the methionine-containing sequences all showed relatively lower changes in fluorescence (−8% to 65%).

In-Depth Study of Q8•Peptide Complexes.

The finding that some peptides gave unexpectedly high changes in fluorescence intensity inspired us to conduct an in-depth analysis of the thermodynamic and structural determinants of binding for a representative set of fourteen purified peptides of sequence GGX₁X₂GGY. The target X₁X₂ binding site is flanked by two Gly residues on each side, and a C-terminal Tyr residue was included for convenient quantitation by UV spectroscopy. The dipeptide sequences YL and LY were tested as benchmarks due to their characterization in prior work.^13,16,17 FK was included because it showed the largest increase in the fluorescence response of all 64 peptides in the library screen. The sequence isomer KF was included and also showed a large increase in the fluorescence response. FR and RF were included due to their large increases in fluorescence response and to the lack of prior investigation of peptides containing the basic Arg residue (with the exception of MRGGY).¹⁴ In our prior work on the pair-inclusion motif, we specifically excluded Trp due to its interference with the fluorescence assay, typically showing decreases in fluorescence upon addition of peptide to Q8•MBBI. In the current study, however, we found that the peptides GWKG and GKWG showed significant increases in fluorescence intensity, suggesting that they displace MBBI upon binding to Q8, perhaps by binding in the pair-inclusion motif. Therefore, we were excited to learn more about these Trp-containing peptides. Heptapeptides containing the dipeptide sites YY, LL, FF, KK, RR, and WW were included as sequence controls for the other eight peptides. The purified heptapeptides were obtained commercially and confirmed by ¹H NMR spectroscopy (Figures S20–S41) and electrospray ionization time-of-flight mass spectroscopy (ESI-TOF-MS) (Figures S42–74).

Isothermal titration calorimetry (ITC) was used to determine the thermodynamics and stoichiometry of Q8•peptide binding at 300 K in 10 mM sodium phosphate, pH 7.0 (Table 1 and Figures S1–S19). The data show that Q8 binds to GGYLGGY and GGLYGGY with K_d values of 0.55 (±0.05) and 0.23 (±0.02) μM, respectively, which is significantly stronger than the previously reported AAYLAA (5.6 μM) and AYL (3.1 μM) but more similar to ALY (0.77 μM).^13,16 We found that Q8 binds to the corresponding sequence controls GGLLGGY and GGYYGGY with K_d values of 18 (±1) μM and >1 mM, respectively. Therefore, submicromolar binding in this system appears to require both Y and L in the target site. We address the influence of flanking G residues further below. Q8 binds to heptapeptides GGFKGGY and GGKFGGY with K_d values of 0.086 (±0.004) and 0.060 (±0.001) μM, respectively, with 100-fold weaker binding for GGFFGGY (11 ± 1 μM) and GGKKGGY (9.3 ± 0.8 μM). Q8 binds to GGFRGGY and GGRFGGY with K_d values of 0.17 (±0.03) μM and 3.5 (±0.3) μM, respectively, pointing to a marked 21-fold selectivity for FR vs RF. We also observed significantly lower affinity for the sequence controls GGRRGGY (>1 mM) and GGFFGGY (11 ± 1 μM). Q8 binds to GGWKGGY and GGKWGGY with K_d values of 1.5 (±0.1) μM and 0.46 (±0.04) μM, respectively, with significantly lower affinity for the sequence controls GGWWGGY (>1 mM) and GGKKGGY (9.3 ± 0.8 μM). In all cases where a K_d value was determined, the Q8:peptide stoichiometry of complexation was 1:1, and binding was enthalpically driven, a hallmark of the nonclassical hydrophobic effect (ΔH ≤ 0, |ΔH| > |ΔS|).⁵⁴ In all cases, a combination of two different residues in the target dipeptide sequence (e.g., WK and KW) yielded a significantly higher affinity than only one type of residue (e.g., WW or KK). In all controls except GGWWGGY and GGYYGGY (which gave no signal), the Q8:peptide binding stoichiometry was confirmed to be 1:1 by ESI-TOF-MS. For X₁X₂ = YL, LY, and LL, Q8:peptide stoichiometry was observed to be 1:1 by ITC and by ESI-TOF-MS, and binding was observed to be enthalpically driven and entropically unfavorable. These results are consistent with binding in the pair-inclusion motif and form the basis for the comparative analysis of the remaining peptides.

Table 1.

Thermodynamic Data for Q8•Heptapeptide Complexes

peptide	Kd (M)	ΔH (kcal mol-1)	-TΔS (kcal mol-1)
GGFKGGY	(8.6 ± 0.4) × 10−8	−13.4 ± 0.1	3.7 ± 0.1
GGKFGGY	(6.0 ± 0.1) × 10−8	−14.2 ± 0.1	4.3 ± 0.1
GGFFGGY	(1.1 ± 0.1) × 10−5	−5.8 ± 0.6	−1.1 ± 0.6
GGKKGGY	(9.3 ± 0.8) × 10−6	−8.7 ± 0.2	1.8 ± 0.2
GGFRGGY	(1.7 ± 0.3) × 10−7	−14.3 ± 0.3	5.0 ± 0.4
GGRFGGY	(3.5 ± 0.3) × 10−6	−11.9 ± 0.3	4.4 ± 0.3
GGRRGGY	>10−3a	NDb	ND
GGWKGGY	(1.5 ± 0.1) × 10−6	−12.0 ± 0.1	4.0 ± 0.1
GGKWGGY	(4.6 ± 0.4) × 10−7	−13.0 ± 0.2	4.3 ± 0.2
GGWWGGY	>10−3a	ND	ND
GGYLGGY	(5.5 ± 0.5) × 10−7	−14.9 ± 0.2	6.3 ± 0.1
GGLYGGY	(2.3 ± 0.2) × 10−7	−14.9 ± 0.5	5.8 ± 0.5
GGYYGGY	>10−3a	ND	ND
GGLLGGY	(1.8 ± 0.1) × 10−5	−13.6 ± 0.7	7.1 ± 0.7

Mean values measured from three ITC experiments at 300 K in 10 mM sodium phosphate buffer, pH 7.0. Errors indicate 1σ from the mean.

^aBinding was too weak to be measured by ITC.

^bNot determined.

Collectively, these data demonstrate that Q8 can bind peptides at nonterminal sites with K_d values in the submicromolar range in neutral aqueous buffer and that binding is sequence-selective. The dipeptide sequence isomers FK and KF are the strongest binders to Q8, and there is little effect of the sequence order on the binding affinity. Q8 binds KF 3-fold more tightly than FR, 8-fold more tightly than KW, and 4-fold more tightly than LY. Q8 binds FR 20-fold more tightly than its sequence isomer RF. Q8 binds KW 3-fold more tightly than its sequence isomer WK. Q8 binds YL only 2-fold more tightly than its sequence isomer LY. In all cases, Q8 binds to the control peptides containing two identical target residues with an affinity significantly lower than that of the corresponding peptides with two different target residues. We were particularly excited by the high affinity of Q8 to the peptides containing the nonterminal sites FK and KF. To the best of our knowledge, these are the highest affinity nonterminal dipeptide sites for any synthetic receptor reported thus far.

Investigation of the Neighboring Sequence Context.

Given the potential utility of the high affinity complexation of Q8 with nonterminal FK and KF, we were particularly interested to learn whether the neighboring residues can have a significant influence on binding. We have reported that substituting a neighboring Gly with the more sterically demanding Ala or Val can diminish binding affinity.¹⁴ Therefore, we studied a series of heptapeptides of the sequence GAX₁X₂AGY and GVX₁X₂VGY, with X₁X₂ = FK and KF.

ITC data show that flanking the target dipeptide site with Ala or Val residues can decrease the binding affinity of Q8 (Table 2). Flanking the target site FK with Ala residues on both sides (i.e., GAFKAGY) yielded a K_d value of 28 (±7) μM, which is a 330-fold decrease in affinity compared with that of GGFKGGY. By contrast, flanking the KF dipeptide target site with Ala residues on both sides (i.e., GAKFAGY) yielded a K_d value of 880 nM, which is a 15-fold decrease in affinity compared with that of GGKFGGY. Flanking FK and KF with Val residues (i.e., GVFKVGY and GVKFVGY) resulted in insoluble peptides. The larger impact of flanking Ala residues on Q8 binding to the target FK vs KF binding sites was unexpected. To probe this phenomenon in more detail, we characterized the binding of Q8 to nonterminal FK with only a single flanking Ala residue, i.e., GAFKGGY and GGFKAGY, to ask whether one or the other Ala has a greater influence on binding. Q8 binds to GAFKGGY with an affinity of 10 (±0.4) μM and to GGFKAGY with an affinity of 4.4 (±0.4) μM. Therefore, one Ala residue reduces the binding affinity by 120-fold when on the N-terminal side of the target site and 51-fold when on the C-terminal side of the target site.

Table 2.

Influence of Flanking Sequence on Binding at KF and FK Sites^a

peptide	Kd (M)	ΔH (kcal mol−1)	−TΔS (kcal mol−1)
GGKFGGY	(6.0 ± 0.1) × 10−8	−14.2 ± 0.1	4.3 ± 0.1
GAKFAGY	(8.8 ± 0.3) × 10−7	−12.0 ± 0.2	3.7 ± 0.2
GVKFVGY	b
GGFKGGY	(8.6 ± 0.4) × 10−8	−13.4 ± 0.1	3.7 ± 0.1
GAFKAGY	(2.8 ± 0.7) × 10−5	−11.3 ± 2.0	5.1 ± 2.2
GVFKVGY	b
GAFKGGY	(1.0 ± 0.1) × 10−5	−10.8 ± 0.3	4.0 ± 0.2
GGFKAGY	(4.4 ± 0.4) × 10−6	−9.0 ± 0.5	1.6 ± 0.5

^aMean values measured from three ITC experiments at 300 K in 10 mM sodium phosphate buffer, pH 7.0. Errors represent 1σ from the mean.

^bPeptide was insoluble.

Structural Characterization by ¹H NMR Spectroscopy.

The binding of several peptides with Q8 was characterized by ¹H NMR spectroscopy (Figures S20–S41). GGWWGGY, GGLYGGY, GGLLGGY, GVFKVGY, and GVKFVGY were insoluble alone or in complex with Q8 at concentrations suitable for NMR spectroscopy. For each peptide, we included an overlay of the one-dimensional spectra of pure peptide, 2:1 peptide:Q8, and 1:1 peptide:Q8. When necessary for peak assignment, ¹H–¹H correlation spectroscopy (COSY) was included. The NMR data were analyzed for changes in the peptide spectra induced by the addition of Q8, including perturbation in chemical shift (Δδ), where a shift of signals to lower frequencies indicates shielding from the spectrometer magnetic field of functional groups on a guest molecule when bound inside Q8, and a perturbation of signals to higher frequencies (i.e., deshielding) indicates noncovalent interactions between functional groups on a guest and C=O groups on Q8. We also analyzed the data for differences in the kinetics of chemical exchange on the NMR time scale for the peptides in the presence of Q8. Peptides with K_d values less than 10 μM were observed to have slow exchange on the NMR time scale, whereas peptides with K_d values greater than or equal to 10 μM were observed to have fast or intermediate exchange on the NMR time scale.⁵⁵

Upon the addition of Q8 to each peptide, we observed shielding of the signals corresponding to the side chain protons of both residues in the target dipeptide binding site, indicating their inclusion within the cavity of Q8. For example, in the spectra for GGKFGGY, we observed a Δδ for the meta protons on the Phe phenyl ring of ~0.8 ppm, while the signals for the ε-methylene protons on Lys had a Δδ of ~1.1 ppm. In cases where one of the bound residues consists of an aromatic residue (e.g., F, Y, W) and the other an alkyl chain (e.g., K, R, L), the chemical shift dispersion of the signals of the nonaromatic target residue is enhanced by magnetic field anisotropy of the proximal bound aromatic ring.⁵⁶ In fact, the resonances for diastereotopic protons (e.g., β CH₂ of F) can be differentiated in this binding motif, as we have observed in the ¹H NMR spectrum of the GGKFGGY•Q8 complex: whereas the average Δδ of the γ-methylene protons is −1.62 ppm upon binding to Q8, one of the geminal protons is more shielded than the other by −0.67 ppm, which is consistent with the close proximity of the Lys ϵ-CH₂ and the Phe phenyl ring.⁵⁷ This effect can also be seen in the NMR spectra of GGYLGGY (Figure 3): the γ-CH of Leu shows the greatest chemical shift perturbation upon Q8 binding at −2.6 ppm. The greater chemical shift perturbation of γ-CH of Leu in this complex compared to the previously reported, nonaromatic ML binding site (−0.6 ppm, Leu γ-CH)¹⁴ points to a magnetic shielding effect greater than what the cavity of Q8 would provide and suggests an interaction of the methine proton with the electron-rich surface of the Tyr aromatic ring in a CH···π interaction. The Leu δ-CH₃ resonances experience a smaller chemical shift perturbation (ca. −1 ppm) upon the addition of Q8, which may be due to a larger inter-side-chain distance between the face of the Tyr aromatic ring and Leu δ-CH₃ protons versus Leu γ CH protons. It is important to note that we did not observe a chemical shift perturbation of signals corresponding to the C-terminal Tyr side chain protons for any of the peptides in the presence of Q8, indicating that the C-terminal Tyr does not interact appreciably with Q8 under these conditions.

Figure 3.

Overlay of ¹H NMR spectra (500 MHz, D₂O, 25 °C) showing a partial titration of Q8 into a 1 mM solution of H-GGLYGGY-NH₂. Relevant peaks were labeled and corresponded to the protons indicated in the structures above. Asterisks denote signals corresponding to the bound peptide.

With regard to the control peptides containing dipeptide sites KK, FF, and RR, the signals corresponding to the side chains of the residues at the target dipeptide binding site are broadened upon addition of Q8 but do not have appreciably perturbed chemical shifts; this result is consistent with weak binding at the dipeptide site. We also observed that for sequences involving K or R, the NMR signals are better resolved when K or R is located at the third position in the sequence (i.e., GGKXGGY or GGRXGGY) versus the relatively broad peaks of the sequence isomers that place K or R at the fourth position (i.e., GGXKGGY or GGXRGGY). A good example of this effect can be seen with GGKWGGY versus GGWKGGY, in which the side chain resonances of W and K become progressively broader as the titration progresses, as contrasted with those of sequence isomer GGKWGGY, which remain well-resolved throughout the titration.

Regarding the residues immediately flanking the dipeptide binding site (i.e., at the second and fifth positions in the sequence), Q8 binds to the target dipeptide site, but mutating G to A or V does not alter the binding mode. In this mode, we would expect any substituent at the α-position of the second and/or fifth residues to be proximal to the electron-dense carbonyl groups of Q8. Indeed, the signals corresponding to the Ala β-CH₃ protons perturb to higher chemical shifts upon Q8 binding, suggesting a steric interaction that may mitigate binding affinity.

Structural Characterization by X-ray Crystallography.

To investigate the structural details of the pair-inclusion binding motif, we obtained a 0.79 Å resolution crystal structure of the binary complex of Q8 with the peptide H-GGLYGGG-OH. Co-crystals were grown from a solution containing equimolar peptide and Q8 at 3 mM dissolved in D₂O containing 10 mM sodium acetate buffer, pH 4.75 (Table S1, CCDC 2312293). The complex crystallized in the P2₁2₁2₁ space group, with one Q8•peptide complex and ten associated water molecules in the asymmetric unit. The experiment gives some insight into the way in which the complex is solvated. Ordered water molecules surround the complex and congregate around the hydrogen bond donors and acceptors. The long-range packing of the complexes in the crystal reveals channels that run parallel to the crystallographic a axis with channel diameter ranging from 4.1–6.5 Å (cf. 2.4 Å for the mean diameter of H₂O used as the probe^58,59).

In the Q8-peptide complex, the Q8 molecule has been deformed from a structure approximating D_8h symmetry in the uncomplexed state⁶⁰ to an ellipse with eccentricity of 0.47. This deformation (or “puckering”) is typical of cucurbit[n]urils bound to appropriately sized guests (Figure 4).^4,5,61 The side chains of the L and Y residues are included within the cavity of Q8, which corroborates the NMR data reported in this study, as well as the semiempirical structure we reported previously, and further substantiates our proposed model of the pair-inclusion motif.¹³ In order to accommodate the inclusion of the side chains of both L and Y residues, the plane of the peptide bond connecting L and Y adopts a conformation that is close to parallel to the plane of oxygen atoms comprising the Q8 portal. The γ-CH of L3 is proximal to the center of the aromatic ring of Y4, which corroborates the NMR chemical shift perturbation (vide supra) and is consistent with a CH···π interaction.⁶² Remarkably, Q8 binding induces a type II β-turn at the dipeptide binding site in which the carbonyl oxygen of G2 forms an intramolecular hydrogen bond^†,63,64 with the amide NH of G5.

Figure 4.

Crystal structure of the Q8·H-GGLYGGG-OH complex. (a) Rendering that highlights the deformation of Q8 (depicted as sticks with a transparent solvent-accessible surface). The L and Y side chains are shown as space-filling models. (b) Rendering that highlights the Q8-induced β-turn in the peptide and the corresponding inter-residue hydrogen bond, the Leu NH···Q8 hydrogen bond, and the L3-Y4 CH···π interaction. Dashed lines in magenta show noncovalent electrostatic interactions. Nitrogens are blue. Oxygens are red. Hydrogens are white. Q8 carbons are gray. Peptide carbons are green.

There is a network of intermolecular hydrogen bonds between the peptide and six neighboring carbonyl oxygens on one portal of Q8 (Figure 5). The first Q8 carbonyl oxygen forms a CH···O=C hydrogen bond with the α-CH of L3. The second Q8 carbonyl oxygen forms a bifurcated hydrogen bond with the amide NH and α-CH of Y4, which also forms a bifurcated hydrogen bond with the third Q8 carbonyl oxygen. The β-CH₂ of Y4 forms two CH···O=C hydrogen bonds with the fourth and fifth carbonyl oxygens of Q8. In addition, the seventh carbonyl oxygen forms a hydrogen bond with the amide NH of L3. This network of electrostatic interactions defines the orientation of most of the peptide backbone with respect to the Q8. Extensive electrostatic interactions between the peptide backbone and the Q8 portal were also reported in Q8•peptide complexes in other binding motifs,⁹ showing the remarkable structural complementarity of these supramolecular systems.

Figure 5.

Crystal structure of the Q8·H-GGLYGGG-OH complex. Rendering highlights the hydrogen bonding network between the peptide and Q8. The hydrogen bonds are shown as dashed lines. Host hydrogen atoms and all water molecules have been hidden for clarity. Nitrogens are blue. Oxygens are red. Hydrogens are white. Q8 carbons are gray. Peptide carbons are green.

We observed several ordered water molecules in the crystal structure. The most notable are the two water molecules located between the Tyr phenolic OH and proximal carbonyl oxygens on the portal of Q8 opposite the one interacting with the peptide backbone (Figure 6). Compared to many of the other water molecules, these two waters are surrounded by the host–guest complex and are limited in their hydrogen bond positioning, further supporting the placement of the hydrogen atoms in a bridging manner. The other water molecules are congregated near the C-terminus.

Figure 6.

Crystal structure of the Q8·H-GGLYGGG-OH complex. Rendering highlights the ordered water molecules that bridge the Tyr phenolic OH to the Q8 portal carbonyls. Dashed lines in magenta show hydrogen bonding interactions between the host and the guest, both depicted in stick form. Host hydrogen atoms were hidden for clarity. Nitrogens are blue. Oxygens are red. Hydrogens are white. Q8 carbons are gray. Peptide carbons are green.

The C-terminus of the peptide, while well-solvated, has relatively large anisotropic displacements in the oxygen atoms of the carboxylic acid. This group is proximal to the electropositive equator of a nearby Q8 molecule, as seen in the Q8•WGG and Q8•FGG•FGG crystal structures.⁹ The N-terminal ammonium group forms a hydrogen bond with a Q8 carbonyl oxygen of a neighboring complex (Figure 7). These intercomplex electrostatic interactions are believed to aid in cocrystallization of the complex.

Figure 7.

Crystal structure of the Q8·H-GGLYGGG-OH complex. Rendering highlights the coordination of one complex to another via dipolar interactions. The peptide C-terminal carboxylate and the Gly6 carbonyl oxygen each coordinate to the outer equatorial surface of an adjacent Q8 molecule. The N-terminal ammonium group of one complex forms ion-dipole interactions with Q8 carbonyl oxygens of an adjacent complex. Hydrogen atoms and water molecules have been hidden for clarity. Molecules are rendered in stick form, where the host is in gray, and the peptides are colored with carbons in green, nitrogens in blue, and oxygens in red.

CONCLUSIONS

This study investigates sequence-activity relationships for the binding of Q8 to nonterminal dipeptide sites. We found that Q8 binds most tightly to nonterminal FK and KF with K_d values of 86 and 60 nM, respectively, the highest affinity reported for a synthetic receptor with an unmodified peptide at a nonterminal site. The K_d values for Q8 with nonterminal dipeptide sites FR and KW were also found to be submicromolar, along with the benchmark YL and LY sites. The 0.79 Å resolution cocrystal structure of the Q8•H-GGLYGGG-OH complex establishes the molecular basis of the pair-inclusion motif for the first time, including the details of the interactions that drive molecular recognition and peptide folding in the complex. The structure reveals the unexpected result that Q8 induces a type II β-turn. To the best of our knowledge, this is the first reported evidence for sequence-selective β-turn induction by a synthetic ligand. The sequence-selectivity, high affinity, reversibility, and structural characterization of this system should facilitate the development of applications involving the recognition of nonterminal sites as well as ligand-induced polypeptide folding.

Supplementary Material

ASSOCIATED CONTENT

Supporting Information

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacs.3c14045.

Experimental details, crystallographic data, ITC data, ¹H NMR spectroscopy data, mass spectrometry data, and fluorescence spectroscopy data (PDF)