NMR structure and dynamics of the agonist dynorphin peptide bound to the human kappa opioid receptor

Significance

The human kappa opioid receptor (KOR) is implicated in addiction, pain, reward, mood, cognition, and perception. Activation of KOR by the neuropeptide dynorphin is critical in mediating analgesia and tolerance. Our solution NMR study of dynorphin (1–13) provided quantitative data on a KOR-bound conformation. Analysis of the peptide structure and dynamics revealed a central helical turn bounded on both sides by flexibly disordered peptide segments. Future drug development will benefit from knowledge of the dynorphin structure bound to its human receptor.

Abstract

The structure of the dynorphin (1–13) peptide (dynorphin) bound to the human kappa opioid receptor (KOR) has been determined by liquid-state NMR spectroscopy. ¹H and ¹⁵N chemical shift variations indicated that free and bound peptide is in fast exchange in solutions containing 1 mM dynorphin and 0.01 mM KOR. Radioligand binding indicated an intermediate-affinity interaction, with a K_d of ∼200 nM. Transferred nuclear Overhauser enhancement spectroscopy was used to determine the structure of bound dynorphin. The N-terminal opioid signature, YGGF, was observed to be flexibly disordered, the central part of the peptide from L5 to R9 to form a helical turn, and the C-terminal segment from P10 to K13 to be flexibly disordered in this intermediate-affinity bound state. Combining molecular modeling with NMR provided an initial framework for understanding multistep activation of a G protein-coupled receptor by its cognate peptide ligand.

G protein-coupled receptors (GPCRs) are the largest superfamily of membrane proteins in the human genome and play a critical role in human physiology by initiating signal transduction in response to extracellular stimuli (1, 2). Since 2007, 89 GPCR crystal structures have been reported, including receptors in inactive and active states, as well as the beta-2 adrenergic receptor (β₂-AR) bound to heterotrimeric G proteins (3). NMR spectroscopy has revealed that the intrinsic conformational heterogeneity of GPCRs is influenced by ligand pharmacology, membrane composition, and effector interactions (4–6). These structural biology studies have provided atomic-resolution insights of systems defined by dynamic structural rearrangements that are correlated with diverse cellular and physiological outcomes.

The classic opioid receptors (δ/κ/μ) are GPCRs activated in response to binding enkephalin-like peptide agonists and are the primary targets of widely prescribed pain medications (7). The kappa opioid receptor (KOR) and its cognate peptide dynorphin are implicated in neuronal pathways associated with addiction, pain, reward, mood, cognition, and perception (8, 9). Nonselective KOR antagonists such as naltrexone have been prescribed for alcohol dependence with limited efficacy in humans, and next-generation KOR antagonists continue to be developed to treat drug addiction and other disorders. Although much is known regarding the antagonist-bound, inactive state of GPCRs, including the crystal structure of JDTic-bound KOR, the interaction of these receptors with neuropeptide agonists remains largely unknown (10). Peptide agonist-bound structures have thus far been limited to a conformationally stabilized neurotensin receptor, likely corresponding to a low-energy peptide-receptor state (11–13).

Dynorphin was discovered by Goldstein and Chavkin as the endogenous activating neuropeptide for KOR, with a “low-resolution” structural model of interaction proposed to PNAS in 1981 (14, 15). Dynorphins are derived from the precursor prodynorphin, with dynorphin A(1–17), dynorphin B(1–13), and alpha neoendorphin sharing a highly conserved N-terminal sequence and charge distribution (16). Dynorphin A(1–13) was shown to act as an agonist on opioid kappa receptors in vivo (17). Physiological activation of KOR is mainly associated with unwanted effects such as dysphoria, anhedonia, and hallucinations, and a current hypothesis in the field is that KOR functionally selective ligands may produce analgesia without dysphoria (18, 19). Functional selectivity has emerged as the leading model to understand the ability of a ligand to activate a subset of signaling cascades, providing a framework for developing next-generation drugs with rationally designed pharmacological profiles (20).

The seminal work of Schwyzer in the 1970s and 1980s led to a model of KOR activation by dynorphin that proceeds via a multistep binding mechanism (14, 21, 22). Thereby, low- to intermediate-affinity binding states of dynorphin correspond to binding to cell-surface membranes or to extracellular loops of the GPCR. A “message–address” paradigm has been formulated based on structure–activity relations observed with dynorphin analogs (21–26). Accordingly, the N-terminal YGGF “message” sequence, which is common to all opioid peptides, was found to be responsible for receptor activation. A C-terminal “address” sequence was further found to contribute via electrostatically driven interactions to KOR subtype specificity. In the context of this paradigm, the present study yields intriguing data on the N-terminal segment of the KOR-bound opioid peptide dynorphin. The methods used, NMR in solution and molecular dynamics simulations, enabled us to define structural ensembles of KOR-bound dynorphin and to characterize internal peptide motions in the presently prepared low-affinity receptor-bound state.

Results

We prepared KOR with the approach used to solve the crystal structure in complex with JDTic by X-ray crystallography (10). The membrane fraction of Spodoptera frugiperda cells transiently expressing KOR was isolated and radioligand binding assays performed to assess receptor function. Dynorphin binding affinity was determined as ∼200 nM at pH 7.4 and 6.1, consistent with an intermediate-affinity interaction (Fig. S1A). Following reconstitution of the receptor in detergent micelles and purification to >95% homogeneity, similar binding was measured and the preparation was stable at 7 °C for over a week (Fig. S1B). The following observations, made in the absence of G protein, thus characterize an intermediate state of KOR along its activation pathway, similar to what was recently described by ¹³C-NMR for β₂-AR (27).

Fig. S1.

Radioligand binding and protein purification. (A) Radioactive diprenorphine was displaced by dynorphin (1–13), giving K_i of 210 nM. Similar displacement profiles were observed at pH 6.1 and 7.4. Ligand binding was performed according to the Psychoactive Drug Screening Program radioligand protocol (see https://pdspdb.unc.edu/pdspWeb/) with a radioactive ligand concentration of 1 nM in binding buffer of 50 mM Tris⋅Cl, 0.1 mM EDTA, and 10 mM MgCl₂. (B) SDS/PAGE of the purification of His-tagged KOR (NuPAGE precast gels, Bis-Tris 4–12%, MES running buffer); lane 3: KOR after IMAC elution, lane 2: after PD-10 desalting, lane 1: after TEV cleavage and reverse IMAC.

The dynorphin (1–13) peptide, YGGFLRRIRPKLK, was ¹⁵N-labeled at residues G2, G3, F4, L5, R6, R7, I8, R9, and L12. [¹⁵N, ¹H]-heteronuclear single quantum correlation (HSQC) NMR spectra of ¹⁵N-(GFLIR)-dynorphin at pH 6.1 and pH 7.4 indicated the peptide adopts a random coil conformation in aqueous solution. Owing to little dependence of binding on pH, measurements were made at pH 6.1 to reduce proton exchange with solvent. Resonance assignments were obtained with 2D [¹H, ¹H]-total correlation spectroscopy (TOCSY) and nuclear Overhauser effect spectroscopy (NOESY), 2D [¹⁵N, ¹H]-HSQC, [¹³C, ¹H]-HSQC, and 3D CBCANH experiments, using standard pulse sequences and a classical ¹H assignment strategy for small peptides (28). ¹H, ¹³C, and ¹⁵N resonance assignments were deposited to the Biological Magnetic Resonance Data Bank (BMRB; accession no. 25597) (29).

Detergent-reconstituted KOR was added to an aqueous solution of dynorphin at a ratio of 1:100, with a 1 mM final concentration of peptide. KOR-specific binding of dynorphin was reversed by the addition of the high-affinity ligand JDTic at a molar ratio of 1:1 with respect to dynorphin. We thus isolated by difference analysis the nonspecific binding of dynorphin to the mixed detergent-sterol micelle of dodecyl maltoside/cholesteryl hemisuccinate (DDM/CHS). NMR-derived values such as chemical shift, NOE intensity, and relaxation rate constants (R₁, R₂, and ¹H-¹⁵N NOEs) were obtained with and without JDTic to characterize a KOR-bound conformation of dynorphin.

Chemical Shift Perturbations.

KOR-binding produced distinct patterns of dynorphin ¹H, ¹⁵N, and ¹³C chemical shift changes detected in 1D proton and 2D heteronuclear correlation NMR experiments, as shown in Fig. 1. The observation of chemical shift changes shows that the exchange rate between KOR-bound and KOR-unbound states is fast on the millisecond time scale, as the frequency difference between bound and unbound state is on the order of 10³ Hz (Fig. 1C; see Supporting Information for further considerations on exchange rates). The sequence-specific changes were largest from L5 to R9, with smaller changes observed from G2 to F4 and K11 to K13. ¹⁵N-¹H correlation experiments acquired without proton decoupling in the in-phase/anti-phase method (HSQC-IPAP) were obtained to determine ¹J_NH (30). No significant variations on the ¹⁵N-¹H couplings was observed in HSQC-IPAP experiments, consistent with negligible residual dipolar couplings, and therefore anisotropic interactions do not contribute to the observed phenomena. The shifts observed on ¹H and ¹⁵N are too large to be accounted for by secondary structure alone and likely reflect a ring current effect due to the receptor proximity. We measured the ¹³C chemical shifts of C′, C^α, and C^β of ¹³C-labeled arginine residues (R6, R7, and R9), because these values report on secondary structure (31). Based on carbon chemical shifts, a helical conformation was expected for R6 and R7, in contrast to R9. In addition to JDTic competition, the inhibition of dynorphin-specific binding was also observed using the highly potent KOR agonist ICI-199,441 (32). The observations and analysis performed here therefore describe a property of KOR–dynorphin-specific interaction reversible by agonists and antagonists.

Fig. 1.

Influence of KOR and JDTic on NMR spectra of dynorphin. (A) (top, black) 1D ¹H-NMR expansion of the NH region of dynorphin in aqueous solution at pH 6.1 acquired at 280 K. (middle, red) Addition of DDM/CHS reconstituted KOR induces significant line broadening for all resonances. The least broadened resonances were K11, L12, and K13. (bottom, blue) The addition of JDTic largely reversed the observed broadening and chemical shift changes, particularly for residues F4 to R9. (B) Overlay of two [¹⁵N, ¹H]-HSQC-IPAP spectra: (red) KOR + dynorphin and (blue) KOR + dynorphin + JDTic. L5 chemical shift perturbations are mainly in the ¹⁵N dimension whereas I8 changes were observed in both ¹H and ¹⁵N dimensions. The complete spectrum is available in Fig. S4. (C) Chemical shift perturbations of dynorphin as a result of KOR binding, calculated as the difference of chemical shifts of dynorphin with KOR and KOR + JDTic. Gray bars (left axis): ¹H chemical shifts; white bars (right axis): ¹⁵N chemical shifts. The [KOR]/[dynorphin] ratio was 1/100. Almost no chemical shift variations were observed for the C-terminal tripeptide segment.

Nuclear Overhauser Effects and Structure Determination.

Where ligand exchange between a bound and free state is considered fast (k_off > R₁), transferred NOEs (trNOEs) originating from the receptor-bound state may be observed on the free peptide spectrum. The measured ¹H R₁ values for amide protons in the free state depended on the residue and ranged from 1.2 to 1.8 s⁻¹ (Fig. S2). As expected from simulations (Fig. S3), our measurements confirm the dominant contribution of unbound dynorphin on R_1. When k_off > σ_ij, the cross-relaxation rate σ_ij is the weighted average between the bound and free state contributions (33). The σ_ij range from 100 to 600 Hz in the bound state and from 0.5 to 3 Hz in the free state and are thus lower than k_off for dynorphin–KOR exchange in our experimental conditions. [¹H, ¹H]-NOESY experiments recorded in the presence and absence of JDTic revealed distinct cross-peak distributions (Fig. 2A). NOE build-up curves were obtained from NOESY experiments at four mixing times: 50, 100, 200, and 500 ms. Fig. 2B shows a characteristic long-range NOE between the G3α and R6 amide protons. For many resonances, a significant qualitative difference was observed for the build-up rate of NOE intensities with and without JDTic.

Fig. S2.

¹⁵N relaxation rates of dynorphin bound to KOR without (red) and with (blue) JDTic, measured at 800 MHz on ¹⁵N-(GFLI)–labeled dynorphin. (A) R₁ relaxation rates (s⁻¹, accuracy 1%). (B) R₂ relaxation rates (s⁻¹, accuracy 5%). (C) ¹⁵N-¹H NOE (accuracy 5%). Note that R₁ and hetNOEs are hardly affected by the small bound fraction, whereas R₂ is largely increased. These relaxation properties are congruent with simulations shown in Fig. S3.

Fig. S3.

¹⁵N relaxation times at 600 MHz as a function of the rotational correlation times τ_c (x axis, logarithmic scale) related to the KOR-bound (KOR), micelle-bound (DDM), and free dynorphin peptide (Dyn) states. R₁ and R₂ are given in seconds⁻¹ and derived from the equations described in ref. 59; hetNOE has no dimension. The experimental values of R₂ determined in the presence of 1 mol % of KOR indicate that the rotational correlation time for the most immobilized residues bound to KOR is above 10⁻⁷ s, much larger than that of the free or micelle-bound peptide.

Fig. 2.

Structure determination of a KOR-bound conformation of dynorphin. (A) Overlay of two 2D [¹H, ¹H]-NOESY spectra: (red) KOR + dynorphin and (blue) KOR + dynorphin + JDTic; 800 MHz, 100-ms mixing time, 280 K. (B) NOE build-up curves for dynorphin cross-peak G3 H^α–R6 H^N in the presence of (red) KOR and (blue) KOR + JDTic. (C) Summary of the most significant NOEs defining the secondary structure. These NOEs are the difference between the observed NOEs in the absence and in the presence of JDTic and are thus representative of KOR specific binding. Note the absence of medium-range NOEs in the N and C termini. (D) Conformational ensemble of 10 structures of dynorphin bound to KOR after restrained MD protocols, using the constraints given in Table 1.

The observed NOE build-up differences were thus converted into NMR-derived structural distance restraints using a biexponential fitting routine. Fifty-six significant restraints were used for structure determination (Table 1). Fig. 2C qualitatively shows the most significant restraints. Medium- and long-range restraints are clearly concentrated in the central part of the peptide, from F4 to R9. The distance restraints and the R6–R7 dihedral angle restraints were used to determine the structure of dynorphin bound to KOR using a restrained molecular dynamics (MD) protocol as described in Materials and Methods and Supporting Information. The structure of KOR-bound dynorphin is shown in Fig. 2D as an ensemble of 10 best-fit structures derived from the structure determination protocol. In agreement with the restraint table, we found that the central part (from L5 to R9) of the peptide forms a well-defined α-helical turn whereas the N- and C-terminal portions, Y1–F4 and P10–K13, are flexibly disordered.

Table 1.

Statistics for the KOR-bound dynorphin structure determination

No. of restraints for calculations
Sequential	42
Medium-range (n to n + 2,3,4)	14
Dihedral angles	4
Restraint violations
NOE restraints, Å	0.07
Coordinate precision: rmsd of backbone atoms, Å
Residues 1–4	1.76 ± 0.53
Residues 5–8	0.46 ± 0.16
Residues 9–13	5.47 ± 1.92
Residues 1–13	3.57 ± 1.11

Rmsds were computed on the bundle of 10 best conformers used to represent the KOR-bound structure of dynorphin (1–13) (Fig. 2D). More detailed statistics and structure validation parameters can be found in Table S2.

¹⁵N Relaxation Times.

Measurement of ¹⁵N relaxation rates are routinely applied in protein NMR to characterize internal dynamics. We determined ¹⁵N T₁, T₂, and ¹H-¹⁵N heteronuclear NOEs (hetNOEs) on a peptide ¹⁵N-labeled on GFLI residues to assess the peptide dynamics (Fig. S2). Significant variation of the ¹⁵N relaxation parameters was observed exclusively for T₂ relaxation times, and thus the related relaxation rate constant R₂. The exclusive variation of R₂ correlates with fast ¹⁵N-dynorphin exchange between free and bound states, with a KOR-bound correlation time (τ_c) of 10⁻⁷ to 10⁻⁶ s, whereas nonspecifically bound peptide has a shorter τ_c of 10⁻⁹ to 10⁻⁸ s (Fig. S3). ¹⁵N T₁ and ¹⁵N-¹H hetNOEs do not contain significant information on the KOR-bound state (Materials and Methods and Figs. S2 and S3). Following these observations, a second set of ¹⁵N T₂ relaxation measurements were made on ¹⁵N-(GFLI)-,¹⁵N,¹³C-(R)-dynorphin, as shown in Fig. 3A.

Fig. 3.

Characterization of internal dynorphin dynamics. (A) ¹⁵N R₂ relaxation rate constants (s⁻¹) measured at 600 MHz at 280 K on ¹⁵N-(GFLI)-,¹⁵N,¹³C-(R)-dynorphin in the presence of (red) KOR and (blue) KOR + JDTic. (B) (gray) Order parameters profile S²/S²_max of NH bond vectors derived from R₂ as described in Supporting Information. The order parameters describe the amplitude of the NH bond fluctuations in the KOR bound state, normalized to L5. (white) Best-fit S² profiles calculated with the ensemble of conformers identified by docking and MD simulations.

The difference in R₂ due to KOR binding was recognized as directly proportional to the order parameter, S², which describes the amplitude of internal motion of the NH vector (Fig. 3B). Changes to R₂ were greater for residues L5–R9 than for G2–F4, and L12 was least affected. In the KOR-bound state, the NH bonds of L5 to R9 are clearly more immobilized than the N and C termini. The ¹H-NMR spectrum also indicates the resonances of K11, L12, and K13 were not as broadened by chemical exchange, confirming that these last three residues remain mobile. The internal dynamics of dynorphin are therefore consistent with measured chemical shift perturbations, NOE patterns, and structure in terms of sequence-specific changes. Collectively, these NMR measurements of ¹⁵N-dynorphin reflect significant residual mobility in the KOR-bound conformation studied here.

Quantitative analysis of intraresidual NOEs was performed for the side chains of Y1, F4, and P10 where interatomic distances are known. From the initial build-up rates of trNOEs we derived order parameters and found S² for F4 was approximately twice that of Y1, and fourfold greater than that of P10. The mobility profile obtained from ¹⁵N measurements was thus extended, revealing the phenolic ring of Y1 remains mobile in a KOR-bound state. Interestingly, the F4 aromatic ring is more immobilized than the F4 NH vector (Fig. 3B). Similarly, the R9 NH vector has a high S² value, whereas the R9 ¹³C chemical shift indicated the helix is interrupted at the R9 dihedral angles. From the rmsd of the structure determination (Table 1), the helix is defined from L5 to I8, and extends slightly toward F4 (but not its NH) and toward R9 NH (but not its C^α or C′ atoms).

Molecular Modeling of KOR–Dynorphin.

Starting from the opioid receptor structures 4DJH (KOR), 4N6H (DOR), and 4DKL (MOR), we generated an ensemble of six structures representative of the opioid receptor binding pocket. Flexible docking of dynorphin was performed with a rigid helical turn and MD simulations performed with explicit water (Fig. S5). Five major dynorphin conformations were identified, revealing significant structural diversity in both N and C termini (Table S1; the corresponding pdb files of the peptide receptor complexes are available in Datasets S1–S5). Averaging of NH orientations over five MD simulations, starting from these five poses, in equal proportion, was required to reproduce experimental data.

Fig. S5.

Decomposition of the dynorphin rmsd during a 50-ns molecular dynamic simulation. Contributions from the N-terminal (black), helical core (blue), and the C-terminal (red) segments are shown. Over the same trajectory, the receptor’s rmsd is ∼2 Å (i.e., similar to the helical core value, which is the least mobile part of the peptide over the trajectory). The C-terminus mobility is apparent after several nanoseconds, whereas the N terminus is more static over 50 ns; a longer time scale such as those involved in ¹⁵N R₂ relaxation may be required to observe mobility. A much longer MD simulation (1 µs) of the KOR-2 conformation (Fig. 4) performed with NMR restraints in the central helix indicated that conformations KOR-1 to KOR-5 did not interchange within this microsecond time scale.

Table S1.

List of persistent contacts between dynorphin and KOR in the major conformations (KOR-1 to KOR-5) obtained from docking and MD simulations

Dynorphin	KOR-1	KOR-2	KOR-3	KOR-4	KOR-5
Interaction energy, kcal/mol	−22.68	−21.11	−20.21	−14.96	−9.12
Y1	Y1(HA)-D138 (O)	Y1(HA)-D138 (OD1)	Y1(H1)-D138 (OD1)	Y1(HH)-D105 (OD2)	Y1(HH)-D105 (OD2)
	Y1(OH)-M142 (HB3)	Y1(HB3)-N141(OD1)	Y1(HB2)-N141 (OD1)	Y1(HB3)-N141 (OD1)	Y1(O)-N141 (HD21)
	Y1(HH)-V230 (O)	Y1(HD1)-W287(HE3)	Y1(OH)-F231 (HA)	Y1(OH)-N322 (HD22)	Y1(OH)-N322 (HD22)
	Y1(HB2)-Y320(OH)		Y1(O)-I316 (HA)
G2	G2(H)-D138 (O)	G2(H)-D138 (OD2)	G2(H)-F231(HE2)	G2(O)-I316 (HA)
	G2(HA2)-Y139(HE1)
G3	G3(H)-D138 (OD2)	G3(H)-D138 (OD1)		G3(HA2)-I290(HG12)
F4			F4(HA)-Y312(OH)	F4(HZ)-S211(HA)
L5
R6		R6(HH21)-E209 (OE2)	R6(HH11)-E203 (OE1)	R6(O)-S211 (HB2)
			R6(HH12)-E209 (OE2)
			R6(O)-S211 (HB3)
R7	R7(HH12)-D223 (OD2)	R7(HH21)-D138 (OD2)	R7(O)-S211	R7(HH22)-E297 (OE2)	R7(HH11)-E297 (OE2)
	R7(HH22)-M226 (O)		(HB2)
			R7(O)-L212 (H)
I8
R9	R9(HH22)-E203 (OE1)				R9(HH22)-E209 (OE1)

Characters in parentheses refer to the atoms for which contacts have been observed during at least 80% of the length of the MD trajectory. The numbering refers to the full length KOR sequence. The Protein Data Bank files corresponding to the 3D structures of these five KOR–dynorphin complexes are available in Datasets S1–S5.

A comparison of opioid receptor crystal structures provided the basis for interpreting the results from MD. We observed that the position of a phenol-like functional group was largely conserved, in terms of position in the orthosteric site, for antagonist-bound structures of KOR-JDTic, DOR-DIPP, and MOR-funaltrexamine (Fig. 4A). Following MD simulations the “KOR-1” structural model revealed Y1 in a position near the phenol-piperidine fused-ring system of JDTic, resembling a previously proposed pose (Fig. 4B) (9). The “KOR-2” pose of dynorphin placed the Y1 side chain in the sodium allosteric site (Fig. 4C). D^3.32 made polar contacts with Y1, G2, and G3 in both KOR-1 and KOR-2 models, whereas R7 made a polar contact in KOR-2 but not KOR-1. W^6.48 and N^7.45 made contacts with Y1 in the KOR-2 model exclusively. Table 2 and Table S1 summarize the findings from modeling and MD simulations, with further details given in Materials and Methods and Supporting Information.

Fig. 4.

Ligand poses from modeling the dynorphin–KOR complex. (A) Antagonist binding poses from structures of KOR-JDTic (4DJH), DOR-DIPP (4RWA), and MOR-FNA (4DKL). JDTic is shown in green, DIPP in purple, and β-funaltrexamine in salmon. (B) Structure of dynorphin in complex with “KOR-1” from MD simulations with Y1 in a position near the fused phenol-piperidine ring system of JDTic. (C) Zoom on dynorphin Y1 in the “KOR-2” complex. Y1 is positioned toward the sodium allosteric binding site. Dynorphin–KOR contacts are given in Table 2 and Table S1. (D) Visual representation of order parameters derived from NMR relaxation measurements. The width of the cone indicates the flexibility of G2 (orange), R6 (green), and L12 (red) dynorphin residues in a KOR-bound state.

Table 2.

Most important contacts between dynorphin and KOR

Dynorphin	KOR-1	KOR-2
Interaction energy	−22.68 kcal⋅mol−1	−21.11 kcal⋅mol−1
Tyr1	Asp138 (3.32)	Asp138 (3.32)
	Met142 (3.36)	Asn141 (3.35)
	Val230 (5.42)	Asn322 (7.45)
		Trp287 (6.48)
Gly2	Asp138 (3.32)	Asp138 (3.32)
	Tyr139 (3.33)
Gly3	Asp138 (3.32)	Asp138 (3.32)
Arg6		Glu209
Arg7	Asp223 (5.35)	Asp138 (3.32)
	Met226 (5.38)

The receptor contacts are shown with Ballesteros and Weinstein numberings for helical residues in parentheses (63). A more detailed list of contacts for the five major conformations of dynorphin is provided in Table S1. The interaction energies were computed using contacts of dynorphin 1–8 residues.

Discussion

Since the 1970s, pharmacologists and biochemists have attempted to determine opioid peptide conformations capable of explaining their activity and to develop drugs mimicking these conformations (34). In aqueous solutions, peptides often exist as a dynamic ensemble of random coil conformers, with specific folds stabilized by organic solvents or micelles (35–37). Studies of liposome-bound peptides likewise indicate that nonpolar or membrane-like environments stabilize peptide structure (23, 24, 38–41). Schwyzer introduced the “membrane compartment concept” that postulated the membrane-bound state as part of the binding mechanism, thereby reducing the available peptide conformations toward an activating conformation in complex with receptor (21, 22, 25). For many years, the direct analysis of the peptide–receptor complex was not possible due to the lack of suitable receptor preparations. Recently, several neurotensin–receptor complexes have been studied both by X-ray crystallography and solid-state NMR (11–13), and the structure of the Leukotriene B4 (a proinflammatory lipid mediator), in complex with the human BLT2 receptor, was determined by liquid-state NMR (41).

Recent advances in producing the opioid receptors (and other GPCRs) in milligram quantities via transient insect cell expression and stabilization in a bicelle-like architecture of mixed detergent-sterol micelles has opened new avenues for opioid receptor structural biology (42). This progress culminated in 2012, with reports of inactive state structures determined by X-ray crystallography of the four opioid receptors in complex with antagonists or inverse agonists (10, 43–46). Currently, the only reported 3D structure of the human KOR is in complex with JDTic, a highly potent KOR antagonist.

NMR using (¹⁵N-¹³C)–labeled ligand represents a powerful complementary alternative to crystallographic approaches in obtaining structural information of receptor activation by peptide agonists. Owing to the moderate affinity of dynorphin with KOR reconstituted in detergent micelles and fast association rate, the ligand dissociation rate is also fast on the NMR chemical shift time scale (47). This context made possible the observation of trNOEs and a straightforward interpretation of ¹⁵N relaxation rates (33). The receptor–peptide interaction was therefore ideal to determine a KOR-bound conformation of dynorphin via the trNOE method, as well as to characterize internal peptide dynamics in a bound state. Such an approach would be prohibited in a comparable study of the high-affinity state, which is characterized by low nanomolar K_d and longer off-rate. In this case, changes to dynorphin structure and dynamics are expected as a result of G protein binding to KOR. Nevertheless, NMR observation of the high-affinity state of dynorphin can be pursued by deuteration of the receptor, peptide, and cognate inhibitory G proteins and preparation of a 1:1:1 complex at millimolar concentrations.

The computational approaches were insufficient to determine reasonable models of dynorphin–KOR binding, because multiple conformers of similar energy were observed. The NMR observations of structure and dynamics were required to limit the starting poses of dynorphin for MD. We identified the pose KOR-1 as typical of an inactive state based on the proximity of Y1 to the phenol-like functional group of JDTic (Fig. 4B and Fig. S6B). In contrast, we speculate the conformational change of the peptide found in KOR-2 correlates with an activated state (Fig. 4 B and C and Fig. S6D). The position of Y1 in the KOR-2 model corresponds with an established role in activation, with the dynorphin (2–13) peptide previously reported to bind weakly and not activate KOR (48). In this “active” conformation, the N terminus of dynorphin forms polar interactions with N^3.35 and D^3.32 side chains, whereas the Y1 phenol ring is involved in a π-stacking interaction with W^6.48 and an H-bond with N^7.45. As part of the allosteric sodium site these residues stabilize an inactive state of the receptor, with changes to dynorphin structure providing agonist–receptor contacts among highly conserved residues in transmembrane helices 6 and 7 (43). Interestingly, this conformational change may be associated with an increased penetration of water into the receptor cavity, which has been linked to the activation mechanism upon agonist binding (49). Mutations of the sodium site in DOR shift nalfurafine, an antagonist, to an arrestin-biased ligand, establishing this sodium site as a key mediator of receptor function (43). The KOR-2 conformation may hint at the mechanism of dynorphin functional selectivity, which has been reported as a G protein-biased agonist in pharmacological in vitro studies (50, 51). The proposed model may be tested in future studies by mutagenesis of KOR residues that surround dynorphin in the KOR-2 conformation.

Fig. S6.

Structural modeling of the dynorphin–KOR complex. KOR is represented in gray cartoon and dynorphin in orange sticks. Blue and red coloring on KOR show side chains with positive and negative charges, respectively. (A) Location of JDTic in the binding site. (B) KOR-1: a stable pose of dynorphin with the side chain of Tyr1 occupying a location close to the aromatic part of JDTic. (C) KOR-3: another pose with Tyr1 taking the place of JDTic. (D) KOR-2: a stable pose of dynorphin with the side chain of Tyr1 extended toward the sodium-binding site. The Protein Data Bank files corresponding to the 3D structures of the five KOR–dynorphin complexes are available in Datasets S1–S5.

The direct quantification of neuropeptide dynamics in an intermediate-affinity receptor-bound state yields useful insights for the design of KOR-targeting antagonists as well as the biology of peptide-activated receptors. More surprising than the well-defined α-helical turn between L5 and R9 was the significant motion observed for the N and C termini. Such motion was expected for the C-terminal part of dynorphin: Indeed, within the message–address paradigm the highly positively charged C-terminal “address” is expected to produce favorable but nonspecific electrostatic interactions with the negatively charged extracellular loop 2. In contrast, this mobility was entirely unexpected for the first four residues YGGF known to be crucial for the activation of opioid receptors and form the so-called “message” part of the peptide.

The biology of peptide agonists may also be reflected in flexibly disordered N and C termini in a bound, but not activating, state. It is feasible that a combination of attractive and repulsive forces have been selected for through evolution so that peptide agonists do not remain bound for excessively long periods, allowing enkephalinases to degrade potent bioactive neuropeptides (52). Whereas GPCRs exist in an array of states with variable ligand affinity, the observations of dynorphin in complex with KOR indicate that there are multiple bound states of the peptide that correspond with various ensembles of activated receptor. We postulate that the KOR-bound conformation reported here, retaining a significant degree of freedom, reflects the mechanism of receptor binding and activation. It would involve, in an initial stage, the association of helix L5 to R9 into the binding pocket and, in a second stage, the conversion of the receptor into the active conformation concomitant with structural immobilization of the N-terminal “message” part of the peptide.

Materials and Methods

Peptide Synthesis.

Peptide synthesis was performed using standard solid-phase synthesis as described in Supporting Information.

KOR Expression, Purification, and Reconstitution in Detergent Micelles.

KOR samples were prepared as previously published and described in Supporting Information (10).

NMR Experiments.

The NMR data were measured at 280 K on a Bruker Avance III 800 MHz for the ¹⁵N-(GFLI)–labeled dynorphin and on a Bruker Avance III 600 MHz for the ¹⁵N(GFLI)- and ¹⁵N-¹³C(R)–labeled dynorphin. The peptide was dissolved to 1 mM in a buffer containing 40 mM deuterated MES (Mes_d), pH 6.1, 150 mM KCl, 100 µM 2,2 dimethyl-2-silapentane-5-sulfonic acid (DSS), and 10% D₂O for frequency lock. To a 1 mM peptide solution was added KOR reconstituted in detergent micelles to a final concentration of 10 µM, a 1:100 ratio of receptor to ligand. The concentration of DDM was 8 mM and CHS 1.6 mM, respectively, as measured by ¹H-NMR. Standard 1D ¹H, 1D ¹³C, 2D [¹H, ¹H]-TOCSY, and NOESY experiments were acquired using excitation sculpting for water suppression (53–55). [¹⁵N, ¹H]-HSQC, [¹³C, ¹H]-HSQC, [¹⁵N, ¹H]-IPAP-HSQC, and CBCANH pulse programs were used to perform the ¹H, ¹⁵N and ¹³C-Arg assignments of dynorphin in aqueous solvent, with KOR, and with KOR and JDTic (30, 56–58). The assignments were obtained using the standard reported strategy for ¹H, with ¹⁵N and ¹³C assignments transferred to heteronuclear correlation experiments based on ¹H assignments (28). The ¹H, ¹⁵N, and ¹³C assignments of the free peptide have been deposited in the BMRB (accession no. 25597).

NOESY experiments were acquired at four mixing times: 50, 100, 200, and 500 ms to generate build-up curves. ¹⁵N relaxation rates, R₁ (inversion recovery), R₂ (CPMG), and ¹H-¹⁵N hetNOEs were measured with established experiments (59). After data acquisition in the presence of KOR, JDTic was added to a final concentration of 1 mM and the complete set of NMR experiments was performed again to report on the nonspecific binding of dynorphin.

Structure Determination.

Direct comparison of NOE spectra were normalized in the following manner. NOE volumes in spectra with and without JDTic were integrated for all mixing times. The integrals were rescaled to take into account the interactions between groups of equivalent spins. The build-up curves were fitted to a biexponential analytic function, which permitted estimation of the cross-relaxation rates. The NOEs in the bound state were calculated assuming a weighted average of the free and bound states, with weights equal to populations in both states. Finally, the NOEs were calibrated with respect to the H^N-H^α peaks from the backbone of the peptide. The initial set of NOEs contained 105 peaks, from which 22 indirect NOEs were eliminated, characterized by sigmoidal build-up curves, as well as 20 NOEs with uncertain integrals due to peak overlap. The intraresidual H^N-H^α NOEs were used only for calibration. Hence, in total, 56 NOEs and 4 dihedral angle restraints (for R6 and R7 dihedral angles) were used for the structure determination (Table 1). Further details of the structure determination protocol are given in Table S2. The ensemble of 10 best structures have been deposited in the Protein Data Bank (ID code 2N2F).

Table S2.

Input for the structure calculations and validation of the bundle of 10 energy-minimized conformers used to represent the NMR structure of dynorphin (1–13)

NOE distance constraints
Intraresidual	2
Sequential	42
Medium-range ($2\le \left|i-j\right|\le 4$)	14
Dihedral angle constraints	4
Residual NOE violations
No. >0.2 Å	7 ± 2
Maximum, Å	0.35 ± 0.04
Residual dihedral angle violations
No. >2°	0
Maximum,°	0
Amber energies, kcal/mol
Total	−110,523 ± 233
Van der Waals	25,209 ± 85
Electrostatic	−142,889 ± 260
Rmsd from ideal geometry
Bond lengths, Å	0.0152 ± 0.0006
Bond angles, °	2.08 ± 0.74
Rmsd to the mean coordinates, Å
Backbone (1–4)	1.76 ± 0.53
Backbone (5–8)	0.46 ± 0.16
Backbone (9–13)	5.47 ± 1.92
Backbone (1–13)	3.57 ± 1.11
Heavy atoms (1–4)	2.66 ± 0.81
Heavy atoms (5–8)	2.14 ± 0.46
Heavy atoms (9–13)	6.26 ± 1.65
Heavy atoms (1–13)	4.38 ± 0.99
Ramachandran plot statistics, %
Most favored regions	41.25
Additional allowed regions	51.25
Generously allowed regions	7.50
Disallowed regions	0
Protein Data Bank validation suite	OK

¹⁵N Relaxation Rates.

Relaxation rates were analyzed using the standard equations described by Farrow (59). In conditions of fast exchange between three states, namely receptor-bound, nonspecific binding, and the free peptide in solution, relaxation parameters were measured as a weighted average of their respective values in each state (60). Owing to the R₁ and hetNOE dependence on rotational diffusion correlation times (Fig. S3), it is academic to establish R₁ and hetNOE should not be significantly affected by the small fraction of bound receptor, whereas the R₂ contribution arising from the bound fraction is proportional to the NH order parameters in the bound state (Supporting Information).

Molecular Modeling of KOR–Dynorphin Complexes.

Details of the molecular modeling protocols are given in Supporting Information and briefly summarized here. We started from the 3D structure of KOR-JDTic published in 2012 (9). Because JDTic induces certain conformational changes by disrupting the salt bridge involving Gln115, Asp138, and Tyr320, we mutated the known structure of MOR into KOR (45). Missing side chains were added and optimized using the SCWRL4 software (61). Out of thousands of possible poses, 10 poses per structure were retained based on a combination of (i) docking score, (ii) interaction of the positively charged C terminus with the negatively charged extracellular loops, and (iii) the competitive binding with JDTic. Each of these poses was then submitted to a 50-ns MD simulation in explicit water. The equilibrated parts of the trajectories (the last 20 ns) were used for subsequent analyses.

The energies of intermolecular interaction between dynorphin and KOR were determined using the MMPBSA method (62). The flexibility of the C terminus on the nanosecond scale was clearly demonstrated in MD simulations (Fig. S4). In contrast, the N terminus is fairly rigid in each simulation, with the existence of distinct starting conformations consistent with reorientations on a slower time scale. The interconversion between these conformations was still not observed after a 1-µs MD simulation performed on KOR-2, with NMR restraints required fixing the central helix. The MD runs were performed over 50 ns in a periodic box with explicit solvent, including four water molecules present in crystal structures of KOR, DOR, and MOR, and with ions neutralizing the charges of the system. The equilibrated parts of the trajectories have been subject to detailed analysis.

Fig. S4.

(A) Overlay of two [¹⁵N, ¹H]-HSQC-IPAP spectra, performed on ¹⁵N-GFLI–labeled peptide, 800 MHz, 280 K, pH 6.1: (red) KOR + dynorphin, (blue) KOR + dynorphin + JDTic. Asterisk indicates impurity. (B) Overlay of two [¹⁵N, ¹H]-HSQC spectra performed on ¹⁵N-GFLI, ¹⁵N-¹³C-R–labeled peptide, 600 MHz, 280 K, pH 6.1: (red) KOR + dynorphin, (blue) KOR + dynorphin + JDTic. G2 cross peak is attenuated by NH exchange with water and shows up at lower contour level in all spectra.

Peptide Synthesis

Synthesized peptides were produced through a solid-phase strategy and purified by semipreparative HPLC on an Aquapore column (C8, 10 × 220 mm, 20 µm; Brownlee Labs). The purity of the final product was assessed by analytical reverse-phase liquid chromatography with a linear gradient from 5 to 80% over 56 min. Solvent A consisted of 0.1% TFA in H₂O milli-Q and solvent B consisted of 0.1% TFA in acetonitrile. Molecular weights were confirmed by MALDI-TOF Voyager DE STR (Applied Biosystems). Fmoc-protected amino acids were purchased from Novabiochem. The starting Fmoc-Lys(Boc)-HMP resin was synthesized by standard method in our laboratory. First couplings were controlled by ninhydrin.

¹⁵N-(GFLI)-Dynorphin (1–13) Peptide.

The coupling was achieved with the COMU-DIEA [(1-(cyano-2-ethoxy-2-oxoethylideneaminooxy)-dimethylamino-morpholinomethylene)] methanaminium hexafluorophosphate– diisopropylethylamine] method. Fmoc¹⁵N-Gly-OH, Fmoc¹⁵N-Phe-OH, Fmoc¹⁵N-Leu-OH, and Fmoc¹⁵N-Ile-OH were synthesized by standard method in our laboratory. COMU was purchased from Iris Biotech GMBH.

¹⁵N-(GFLI)-¹⁵N,¹³C-R-Dynorphin (1–13) Peptide.

The coupling was achieved with the HOAT-DIC [1-hydroxy-7-azabenzotriazole–N,N′-diisopropylcarbodiimide] method. Fmoc¹⁵N-Gly-OH, Fmoc¹⁵N-Phe-OH, Fmoc¹⁵N-Leu-OH, and Fmoc¹⁵N-Ile-OH were synthesized by standard method in our laboratory. Fmoc-Arg(Pbf)-OH (U-¹³C₆, U-¹⁵N₄)was purchased from AnaSpec. HOAT was purchased from PerseptiveBioSystems.

Expression of KOR in Sf9 Cells

The wild-type (OPRK, Uniprot accession no. P41145) human kappa opioid receptor gene was subcloned into a modified pFastBac1 vector with a truncated N terminus (ΔN⁴²). An N-terminal expression cassette included hemagglutinin signal sequence followed by FLAG epitope, 10x-His, and TeV protease recognition site. The mutation I135L was introduced to increase expression and stability. The N-terminal sequence was identical to that used to solve the 2012 structure in complex with JDtic. Recombinant baculoviruses were generated with the Bac-to-Bac system (Invitrogen) and used to infect Sf9 insect cells at a density of 2 × 10⁶ cells per mL at a multiplicity of infection of 5 as previously described (10). Expression and trafficking was assessed by fluorescent detection of the FLAG epitope. Infected cells were grown at 27 °C for 48 h before harvesting, with resulting cell pellet stored at −80 °C.

KOR Purification and Reconstitution in DDM/CHS Micelles

KOR was purified for NMR in a manner similar to preparations used for X-ray crystallography, briefly outlined here (10). Lysis was performed by a combination of thawing the frozen cell pellet, hypotonic shock, and gentle shearing forces accomplished via dounce homogenization in the presence of EDTA-free complete protease inhibitor mixture tablets (Roche), followed by ultracentrifugation at 200,000 × g for 35 min. The resulting pellet containing the membrane fraction was homogenized in the presence of 1 M NaCl followed by ultracentrifugation (twice) to complete membrane isolation. Preceding solubilization, the membrane fraction was resuspended and incubated in a solution containing 250 μM naltrexone, 2 mg⋅mL⁻¹ iodoacetamide, 800 mM NaCl, and 50 mM Hepes (pH 7.5) and incubated for 1 h at 4 °C.

Solubilization of membranes was accomplished by a 1:2 dilution with a membrane solubilization buffer containing 50 mM Hepes (pH 7.5), 300 mM NaCl, 40 mM (wt/vol) n-dodecyl-β-d-maltopyranoside (DDM; Anatrace), 8 mM cholesteryl hemisuccinate (CHS; Sigma), and 250 μM naltrexone for 3 h at 4 °C. Solubilized membranes were isolated by ultracentrifugation at 180,000 × g for 45 min. Purification of His-tagged proteins was initiated by incubation in 20 mM imidazole (pH 7.4) and 800 mM NaCl with 2 mL TALON IMAC resin (Clontech) for 6–18 h (“overnight”). Unbound proteins were removed by centrifugation (700 × g) at 4 °C followed by batch-washing of beads in 25 column volumes (CV) of wash buffer I [1 mM DDM, 0.2 mM CHS, 10 mM Hepes (pH 7.4), and 150 mM KCl]. Unbound proteins were effectively removed after an additional 3 × 5 CV wash buffer II [1 mM DDM, 0.2 mM CHS, 10 mM Hepes (pH 7.4), 20 mM imidazole, and 150 mM KCl]. The receptor was eluted in 5 CV of elution buffer [25 mM Hepes (pH 7.4), 150 mM KCl, 1 mM DDM, 0.2 mM CHS, and 200 mM imidazole] with ¹⁵N-dynorphin (1–13), or dynorphin, added to a final concentration of 25 μM to the elution fractions. Imidazole was removed and a deuterated MES buffer exchanged by gravity-flow size exclusion using the PD-10 miniTrap G-25 column (GE Healthcare) with a buffer containing 1 mM Mes_d (pH 6.1), 150 mM KCl, 1 mM DDM, and 0.2 mM CHS. ¹⁵N-dynorpin was added to the desalted protein fraction to a final concentration of 25 μM. The N-terminal expression cassette was removed by treatment with His-tagged TeV protease (50 µL, 5 mg/mL) at 4 °C for 1–3 h and incubation with TALON IMAC resin for 6–18 h. Protein purity was judged as greater than 95% by SDS/PAGE (Fig. S1) and protein quality judged as monodisperse by analytical size-exclusion chromatography. The purified KOR sample was concentrated to a final concentration of ∼30 µM in 100 µL of H₂O, 10% D₂O, 40 mM Mes_d (pH 6.1), 150 mM KCl, 100 µM DSS, 8 mM DDM, and 1.6 mM CHS. The receptor concentration was estimated by UV absorbance at 280 nm using a theoretical molar extinction coefficient of 48,400 M⁻¹⋅cm⁻¹. The KOR solution was then added to ¹⁵N-dynorpin, 1 mM in the same buffer, to get the desired KOR:dynorphin molar ratio of 1:100 and appropriate line broadening effect.

Radioactive Ligand Binding Experiments

Saturation binding was performed with washed Sf9 membranes using [³H]diprenorphine in the presence and absence of 10 µM JDTic. The binding assays were done in 96-well plates with a final volume of 125 µL per well: 25 µL radioligand (0.16–20 nM), 25 µL binding buffer (for total binding) or 25 µL JDTic (for nonspecific binding), and 75 µL washed Sf9 membranes expressing the KOR construct, IMPT1280. The binding buffer consists of 50 mM Tris⋅HCl, 10 mM MgCl₂, and 0.1 mM EDTA, pH 7.4 (or 5.0 or 6.0), at room temperature. Approximately 0.5 µg of total membrane protein was added to each well and the reaction incubated for 1 h in the dark at room temperature. The reaction was stopped by vacuum filtration onto cold 0.3% polyethyleneimine-soaked 96-well glass fiber filter mats using the 96-well Filtermate harvester (Perkin-Elmer). The filter was then washed three times with cold standard wash buffer (50 mM Tris⋅HCl, pH 7.4, at 4 °C) and a wax scintillation mixture melted on the filter and radioactivity counted in a MicroBeta2 counter (Perkin-Elmer). Total binding and nonspecific binding results were analyzed to determine the K_d and K_i values. Competition binding assays were performed under similar conditions; however, a constant dose of 1 nM [³H]diprenorphine was used with the competing ligand dynorphin (1–13) ranging from 0–10 µM. The counts were pooled and fitted to a three-parameter logistic function for competition binding to determine K_i.

Observation of Fast Exchange Rate

In the fast exchange limit, the observed chemical shift reflects the weighted average of bound and free states. In our experimental conditions, the bound fraction of dynorphin is ∼1%, as [dynorphin] = 1 mM and [KOR] = 10 μM. The relationship δ_av = x_bδ_b + x_fδ_f may be rewritten as δ_b − δ_f = (δ_av – δ_f)/x_b, indicating that the ∼10 Hz of observed shifts are due to ∼1,000-Hz (∼1 ppm) shifts between the bound and free state. Because the observed shifts are on the order of 10³ Hz and proportional to the bound fraction, we concluded that the exchange is fast on the millisecond time scale. Strictly speaking, the infinitely fast exchange hypothesis may not apply here, and the system may be somewhere between fast and intermediate exchange with respect to the millisecond time scale, which was sufficient for the observation and analysis of trNOEs.

NMR Observation of Fast Off-Rate with a 200 nM K_d

Applying the most simple model of bimolecular interaction gives K_d = k_off/k_on. The experimental association rates for protein–protein pairs cover a wide range of k_on, from ∼10³ to 10¹⁰ M^–1⋅s^–1 (64). Therefore, it is difficult to determine a priori that a certain K_d will be compatible with a fast off-rate and with the observation of trNOEs. In our case, the NMR measurements revealed that the interaction we observe was due to a fast exchange ligand binding condition being satisfied, suggesting that the on-rate should be of the fast side. The theory of diffusion controlled on rates of association is well described by Fersht (64). For two molecules of the same radius in water at 25 °C, the encounter frequency is equal to 7 × 10⁹ M⁻¹⋅s⁻¹. When one partner is larger than the other one (e.g., a peptide–receptor interaction) it is faster and scales with (r_A + r_B)²/r_Ar_B,. The actual rate can be lower due to nonproductive binding and activation energy (due, for instance, to a structural rearrangement of the receptor). It can be higher due to favorable electrostatic interaction energies, which can bring the encounter frequency up to 10¹¹ M⁻¹⋅s⁻¹. Thus, for a K_d of 200 nM and a k_on of 10¹⁰ M⁻¹⋅s⁻¹, the k_off is equal to 2,000 Hz, which is compatible with our observed fast exchange rate on the millisecond time scale. On rates may be faster if the dominant binding mechanism is not a 3D random diffusion search, but rather a two-step binding mechanism. For instance, a first collision of dynorphin with the micelles followed by a 2D diffusion on the surface to the receptor leads to an increased k_on by the so called “reduction of dimensionality effect” (65). Furthermore, we follow the interaction between positively charged (+5) dynorphin peptide and negatively (−6) charged extracellular surface of KOR (mostly within ECL2). This may lead to a collision factor close to one, the peptide being driven toward to entry site by the electrostatic potential, as well as increased collision rate. In a similar situation, despite a K_d of 62 nM, trNOEs have been observed in systems where electrostatics significantly increased k_on (47). For a similar reason, at low ionic strength, the barnase–barstar association rate was found to be >5 × 10⁹ M⁻¹⋅s⁻¹ (64), and the association rate of AD2 to TAZ2 has been estimated to be 1.7 × 10¹⁰ M^–1⋅s^–1 (66).

Details of Structure Determination Protocols

An extended structure of the peptide was created as the starting point for a series of MD simulations coupled with the simulated annealing algorithm, in the presence of NMR distance restraints (Table 1). In total, 1,000 putative structures were generated, with each MD run covering 320 ps. The best structures retained were those with the lowest distance violations. To compare directly two sets of spectra, acquired with and without JDTic, the NOE volumes in each series of spectra were divided by a scale factor, defined as the product of number of scans, receiver gain, and the volume of a reference peak. As reference, we chose the amide proton of K13 in dynorphin, because the signal of this highly mobile C-terminal residue is not affected by the peptide’s binding to KOR. The NOE volumes have been integrated for all mixing times τ_m in both series of spectra. In case of interaction between two groups of equivalent spins, the volumes have been rescaled by the factor $2\text{nm}/(n+m)$, where n and m are the numbers of spins in each group. The build-up curves for each NOE have been fitted to the function given by the expression $I\left({\tau }_m\right)=a\exp (-{\rho }_S{\tau }_m)\cdot \left(1-\exp (-2{\sigma }_S{\tau }_m)\right)$, where a is the signal intensity, ρ_S the relaxation rate in state S, and σ_S the cross-relaxation rate in state S (S = averaged or free). “Free” refers to the relaxation rates observed in the presence of JDTic that reflect the averaged relaxation between the free peptide and the multiple possible species resulting from nonspecific interactions. The averaged, observable cross-relaxation rate σ_av for dynorphin in the presence of KOR in DDM/CHS micelles is a weighted average of the rates in the free and specifically bound states: ${\sigma }_{av}=p_f{\sigma }_f+p_b{\sigma }_b$, where p_f and p_b are fractions of the peptide in the free and bound states, respectively. In our experiments the fractions of free and bound amounted to ∼99% and 1%, respectively. Hence, σ_b could be calculated from the above equation. However, these values are still affected by an unknown amplification factor, resulting from experimental settings, such as tuning quality, analog-to-digital converter resolution, and processing parameters. For this reason the obtained cross-relaxation rates have been calibrated using the intraresidual H^N-H^α NOEs, for which the distances are known (2.9 Å). The NOEs have been divided into groups (strong, medium, weak, and very weak) with appropriate upper limits for distances (2.7, 3.3, 5, and 6 Å, respectively). The lower distance limit was held at 1.8 Å in all cases.

Computations of the peptide structure involved NMR restraints-driven MD simulations with simulated annealing. After the initial structure minimization, the system was heated to 1,000 K within 20 ps, followed by an MD run of 300 ps, divided into three stages. During the first 100 ps the temperature was maintained at 1,000 K. In the second stage the system was slowly cooled to 300 K, and during the last 100 ps the temperature was decreased to 100 K. In parallel with the temperature changes, the force constants were also evolving. At the beginning they were reduced to 1% of their nominal values. During the first 20 ps they increased to 5%, during the following 20 ps they increased up to 20%, then during the next 40 ps they were increased to 100% and kept at their maximal value to the end of the procedure. The structures were minimized again at the end of calculations. The quality of the final structures was assessed from the violations of NOE distances and dihedral angles and deposited in the Protein Data Bank (ID code 2N2F).

R₂ Analysis

For dynorphin, R₂ is strongly enhanced by the receptor-bound fraction due to the dominant contribution of J(0). Given R_2KOR/JDTic to represent the contribution to R₂ relaxation of all of the nonspecific interactions and R_2KOR the contribution to R₂ of all states, one may write R_2KOR − R_2KOR/JDTic = x_b·R_2bound ∼ x_b·(0.5 d² + 1/6 c²)·4J(0). In this expression, J(0) dominates R₂ relaxation at high correlation time τ_c, such as observed in the KOR-bound state, d and c are the dipolar and chemical shift anisotropy constants, respectively, and x_b is the peptide-bound fraction, equal to 1% (59).

Furthermore, under the Lipari–Szabo formalism and assuming an internal motion characterized by the correlation time τ_i and amplitude S², $\mathrm{J}\left(0\right)={\mathrm{S}}^2\cdot {\mathrm{\tau }}_{\mathrm{c}}+\left(1-{\mathrm{S}}^2\right)\cdot \left[{\mathrm{\tau }}_{\mathrm{i}}/(1+{\mathrm{\omega }}^2{\mathrm{\tau }}_{\mathrm{i}}^2\right]\sim {\mathrm{S}}^2\cdot {\mathrm{\tau }}_{\mathrm{c}}$ (again due to the large value of τ_c compared with τ_i).

Thus, the difference of R₂ relaxation rate constants in the absence and in the presence of JDTic,

$${\mathrm{R}}_{\text{2KOR}}-{\mathrm{R}}_{\text{2KOR}/\text{JDTic}}\sim (0.5{\mathrm{d}}^2+1/6\hspace{0.17em}{\mathrm{c}}^2)\hspace{0.17em}{\mathrm{x}}_{\mathrm{b}}\cdot 4·{\tau }_{\mathrm{c}}\cdot {\mathrm{S}}^2,$$

is simply proportional to S², and thus describes directly the residual amplitude of motion of each NH vector in the KOR-bound state.

Details of the Molecular Modeling Protocols

The initial structures used as starting points for MD simulations were obtained from flexible docking. Each complex was placed at the center of a periodic box. To hydrate the system, the volume occupied by the complex was extended by 10 Å on each side, and 12,268 water molecules were added to the box. The final system had the dimensions of 73 × 100 × 72 Å and contained 41,727 atoms. The MD simulations were performed with the Amber 12 software (67), using the FF03 force field for the protein and the peptide and the TIP3P water model for the solvent. The equilibration of the entire system was achieved in several steps. Initially, the energy of the system was minimized by 1,000 cycles of the steepest descent (SD) algorithm, with the solute held fixed, by constraining its Cartesian coordinates using a harmonic potential with the force constant k equal to 100 kcal(mol)⁻¹⋅(Ų)⁻¹. In the second step, the energy was minimized by 500 cycles of SD and 1,500 cycles of the conjugate gradient algorithm, with weakly restrained solute (k = 10 kcal⋅mol⁻¹⋅Å⁻²). Next, a short 20-ps MD run was performed on weakly restrained solute with temperature varying linearly from 0 to 300 K. The integration step used in this run was 1 fs. Throughout the calculations a cutoff of 12 Å was used for electrostatic interactions. The MD simulation continued for 100 ps at constant temperature at 300 K with no restraints, with the integration step of 2 fs. Finally, a 50-ns run with constant pressure of 1 bar was launched, with atomic coordinates saved every 10 ps. The Langevin dynamics was used to control the temperature, with γ = 1.0 ps⁻¹, and the pressure was controlled by the Berendsenbarostat with the pressure relaxation time τ_p = 2 ps. Bonds involving hydrogen were constrained with the SHAKE algorithm. Calculations were performed using GeForce GTX TITAN Black GPU cards, which worked at the speed of nearly 1 ns⋅h⁻¹. Trajectories obtained from MD simulations were analyzed with the AmberTools 13 programs as well as with in-house software.

Pairwise decomposition of the interaction energy, along with a detailed analysis of persistent intermolecular contacts, determined the conserved contact residues between the two molecules. The analysis of individual trajectories shows that the N terminal of dynorphin and its helical core have low rmsd, whereas the C terminal is flexible with a high rmsd value (Fig. S5), consistent with the experimental findings.

An initial MD simulation of KOR was performed in a periodic box with explicit solvent. The simulation covered 100 ns, of which ∼25 ns were necessary to equilibrate the system. The frames in the remainder of the trajectory were used in the clustering analysis. With the cluster radius set to 2 Å, six main-chain conformers were found. We selected 16 residues inside the pocket of the binding site whose side chains protrude to the interior of the pocket, to be flexible during docking. As for dynorphin, we have fixed the single bonds in the backbone of residues 5–8 to preserve the helical turn in the core of the peptide and kept all of the other bonds rotatable, both in the backbone and in the side-chains. The docking was performed on all six major KOR conformers, with 1,320 peptide poses generated for each conformer. The resulting structures showed significant diversity while having similar scoring function values; therefore, we have filtered the results with respect to experimental evidence available: The aromatic residues (Y1, F4) should be found within the pocket, whereas the C-terminal part (K11) should face the outside of the site to be able to interact with the KOR’s extracellular loops. These restraints reduced the number of poses down to several dozen per conformer. In the end, we have selected some 10 poses per receptor conformer for validation via MD simulations.

The local order parameters S² for selected NH vectors in a given trajectory have been calculated by measuring the degree of dispersion of these vectors, quantified by the ratio of the volume of the space they occupy to the volume of the sphere corresponding to the isotropic reorientations. The volume of a cone formed by fluctuating NH vectors is given by

$$V\left(\theta \right)={\displaystyle \underset{0}{\overset{r}{\int }}{\displaystyle \underset{0}{\overset{2\pi }{\int }}{\displaystyle \underset{0}{\overset{\theta }{\int }}{r}^2\sin \theta \hspace{0.17em}d\theta \hspace{0.17em}d\varphi \hspace{0.17em}dr}=\frac{2}{3}\pi r^3(1-\cos \theta )}},$$

where r is the NH distance and θ is the cone semiangle. The isotropic reorientations are described by θ = π, with V_iso = (4/3)πr³. The mobility M of a fluctuating NH vector can be quantified by its normalized volume V/V_iso, given by (1 − cosθ)/2. Consequently, the order parameter S² can be defined as

$$S^2=1-M=\frac{1+\cos \theta }{2}$$

and varies between 1 and 0 for angle θ varying between 0 and π. For each fluctuating NH vector, the temporal dispersion was determined from the MD trajectory and the angle θ of the corresponding cone has been calculated as the width of the NH vector distribution. Then it was converted to the order parameter according to the above equation. For comparison with NMR derived order parameters, S² was averaged over the last 20 ns of MD trajectories from the five best conformers (Table S1). A mixture of these conformations was required to reproduce correctly the order parameter profile, particularly for the N-terminal residues G2 to L5.