Evaluation of Efficient Non-reducing Enzymatic and Chemical Ligation Strategies for Complex Disulfide-Rich Peptides

Abstract

Double-knotted peptides identified in venoms and synthetic bivalent peptide constructs targeting ion channels are emerging tools for the study of ion channel pharmacology and physiology. These highly complex and disulfide-rich peptides contain two individual cystine knots, each comprising six cysteines and three disulfide bonds. Until now, native double-knotted peptides, such as Hi1a and DkTx, have only been isolated from venom or produced recombinantly, whereas engineered double-knotted peptides have successfully been produced through enzymatic ligation using sortase A to form a seamless amide bond at the ligation site between two knotted toxins, and by alkyne/azide click chemistry, joining two peptide knots via a triazole linkage. To further pursue these double-knotted peptides as pharmacological tools or probes for therapeutically relevant ion channels, we sought to identify a robust methodology resulting in a high yield product that lends itself to rapid production and facile mutational studies. In this study, we evaluated the ligation efficiency of enzymatic (sortase A5°, butelase 1, wild-type OaAEP 1, C247A-OaAEP 1, and peptiligase) and mild chemical approaches (α-ketoacid-hydroxylamine, KAHA) for forming a native amide bond linking the toxins while maintaining the native disulfide connectivity of each pre-folded peptide. We used two Na_V1.7 inhibitors: PaurTx3, a spider-derived gating modifier peptide, and KIIIA, a small cone snail-derived pore blocker peptide, which have previously been shown to increase affinity and inhibitory potency on hNa_V1.7 when ligated together. Correctly folded peptides were successfully ligated in varying yields, without disulfide bond shuffling or reduction, with sortase A5° being the most efficient, resulting in 60% ligation conversion within 15 min. In addition, electrophysiology studies demonstrated that for these two peptides, the amino acid composition of the linker did not affect the activity of the double-knotted peptides. This study demonstrates the powerful application of enzymes in efficiently ligating complex disulfide-rich peptides, paving the way for facile production of double-knotted peptides.

Graphical Abstract

INTRODUCTION

Double-knotted peptides are increasingly being identified in spider venoms.^1–3 Two double-knotted peptides, DkTx and Hi1a, were recently found to target the heat- and acid-sensitive tetrameric transient receptor potential ion channel TRPV1 and the acid-sensitive ion channel ASIC1a, respectively,^1,2,4,5 whereas CpTx1 has been reported to damage membranes in a receptor-independent manner.³ DkTx and Hi1a comprise 79 and 75 amino acid residues, respectively, and contain two individual inhibitor cystine knot (ICK) domains, with three disulfide bonds each, linked together via a short amino acid chain.^1,2 CpTx1, on the other hand, is a much larger 134 amino acid residue peptide containing two domains of four disulfide bonds each.³ In addition to these three native double-knotted venom peptides, identified using traditional assay-guided fractionation, a recent study using data-mining of the UniProtKB database identified more than 60,000 secreted cysteine-rich repeat peptides (SCREPs) for a range of putative targets including bovine pancreatic trypsin inhibitor/Kunitz domains, with 836 non-redundant SCREPs targeting ion channels.⁶ This highlights the prevalence of these multi-domain peptides in nature, their potential as pharmacological tools or drug leads for a range of therapeutic targets, and the importance of the development of robust synthetic pathways to further explore the pharmacology of this class of peptides.

For DkTx and Hi1a, although both domains are independently active to varying degrees, neither of the individual domains can reach the potency of their respective full-length peptide.^1,2 This suggests that both domains must simultaneously bind to the receptors to promote full activity. Despite the similarity in the channel architecture between TRPV1 and Na_Vs,^4,7–9 so far there are no reports of naturally occurring multi-domain peptides that target voltage-gated sodium channels, albeit ~10% of SCREPs are predicted to target Na_Vs.⁶ The first attempt to produce an artificial double-knotted peptide targeting Na_Vs was by Murray et al., who produced a homodimer of the gating modifier toxin (GMT) GpTx-1 by means of linking the two domains with a polyethylene glycol (PEG) linker using alkyne/azide click chemistry.¹⁰ Dimerization caused a 10-fold increase in potency and increased the peptide residence time on Na_V1.7 compared to a single GpTx-1 molecule.¹⁰ In addition, a recent study from our group showed successful homo- and hetero-dimerization of GMTs CcoTx-1, CcoTx-2, gHwTx-IV, and D1la targeting Na_V1.7 using the evolved sortase A5° (SrtA5°) and a four-residue glycine peptide linker.¹¹ These peptides did not display improved affinity for Na_V1.7, but the results suggested the presence of cooperative binding.¹¹ These findings have emphasized the potential of homo- and hetero-bivalent neurotoxins targeting Na_Vs. In addition to studies focusing on homo- and hetero-ligation dimerization of two ICK GMTs targeting the voltage-sensing domains of Na_V1.7, the pharmacological activity of hetero-bivalent peptides comprising the GMTs mHwTx-IV and PaurTx3 ligated to the Na_V channel pore blocker μ-conotoxin KIIIA has also been explored.^12,13

Although there are multiple ways to join two peptide fragments together using a variety of linkers and ligation strategies,^{10,11,14–19} our aim was to utilize a ligation method leaving a native amide bond to mimic bivalent peptides found in nature. The challenge in producing these bivalent peptides synthetically arises from our desire to fold each of the domains prior to ligation to ensure correct disulfide bond connectivity. This means that the chosen ligation method must be non-reducing to avoid disulfide bond shuffling during the ligation reaction. Popular chemical ligation strategies, including a variety of native chemical ligation (NCL) methods, which are elegant, rapid, and high yielding reactions, usually require reducing agents like tris(2-carboxyethyl)phosphine (TCEP) and thiol additives like 4-mercaptophenylacetic acid or sodium 2-mercaptoethanesulfonate^20–23 and are thus not suitable for this study. On the other hand, enzymatic ligation utilizing enzymes including sortase A (SrtA) (Figure 1A),^{11,12,24–26} asparaginyl endopeptidases (OaAEP 1) (Figure 1B), butelase 1 (Figure 1C),^27–34 and peptiligase (Figure 1D)^35–39 or chemical ligation using α-ketoacid-hydroxylamine amide (KAHA) (Figure 1E)^40–44 approaches has been shown to be able to form native amide bonds in the absence of reducing agents.

Figure 1.

Scheme showing enzymatic and chemical ligation approaches examined in this study. The recognition sequences of the enzymes are shown as single letter amino acid residue codes. The N-terminal peptide is shown in blue, and the C-terminal peptide is shown in orange. (A) SrtA, (B) OaAEP 1, (C) butelase 1, (D) peptiligase, and (E) KAHA ligation.

In this study, we sought to evaluate these ligation strategies using PaurTx3, a spider-derived GMT from Phrixotrichus auratus,^45,46 and KIIIA, a μ-conotoxin Na_V pore blocker isolated from the venom of Conus kinoshitai.⁴⁷ These peptides were originally selected in a previous study due to their moderate and approximately equipotent activities at hNa_V1.7, one of the best-validated pain targets, (PaurTx3 IC₅₀ = 290 nM, KIIIA IC₅₀ = 363 nM), allowing us to detect any potential improvements in activity upon ligation.¹² Following evaluation of the two recently reported cryo-EM structures of KIIIA in complex with Na_V1.2⁴⁸ and the structure of the gating modifier toxin ProTxII in complex with Na_V1.7,⁴⁹ measuring the distances between the two toxin binding sites, we applied an SrtA5°-mediated approach using polyglycine linkers to evaluate various linker lengths in order to optimize the Na_V1.7 activity of ligated PaurTx3 to KIIIA.¹² The synthetic bivalent peptide with an optimized 9-glycine linker length showed improved potency at Na_V1.7 compared to either PaurTx3 or KIIIA individually with an IC₅₀ = 58 nM.¹² This work suggested that novel and improved Na_V1.7 inhibitors can be designed by combining a pore blocker toxin and a gating modifier toxin to confer desired pharmacological properties from both the voltage sensing domain and the pore domain. However, SrtA5°-mediated ligation reactions resulted in hydrolysis products that proved to be challenging to separate from the starting material as they elute close to PaurTx3 by HPLC. In addition, the five-residue-long SrtA motif ultimately proved to be susceptible to degradation in human serum.¹² Our aim with this study was therefore to evaluate alternative approaches to serve our purpose of creating bivalent disulfide-rich peptides via native amide bonds while maintaining their structure, potency, and stability under harsh proteolytic conditions encountered in vivo.

Herein, we present a comparative study evaluating the ability of both enzymatic (SrtA5°, OaAEP 1, butelase 1, and peptiligase) and chemical (KAHA) ligation approaches for creating bivalent peptides from disulfide-rich monomers of KIIIA and PaurTx3. Using NMR, we confirmed that the different ligation strategies did not compromise the overall fold in the individual domains, we evaluated the impact of different linker sequences on bivalent peptide activity using electrophysiology, and finally we evaluated the ligated peptide’s stability in serum.

RESULTS

In our previous study, we found that the linker length for connecting PaurTx3 and KIIIA to reach maximum potency at NaV1.7 was nine amino acid residues.¹² We therefore maintained this linker length for all ligation strategies tested here in order to evaluate the efficiency of different ligation strategies in addition to assessing if different linker compositions affected activity, adding or subtracting Gly residues to accommodate the different ligation motives to achieve the desired linker length. Peptide sequences, enzyme recognition sequences, linkers, and theoretical and observed masses for peptides used in this study are summarized in Table 1 and Table S1. All results of SrtA5°-mediated ligation in this study were extracted from Tran et al.¹² and are used for comparative purposes.

Table 1.

Summary of Peptide Sequences Synthesized for This Study^a

Strategy	Peptide	Sequence
	PaurTx3	DCLGFLWKCNPSNDKCCRPNLVCSRKDKWCKYQI
	KIIIA	CCNCSSKWCRDHSRCC*
SrtA5°	PaurTx3[LPAT|GG]	DCLGFLWKCNPSNDKCCRPNLVCSRKDKWCKYQILPATGG
	[G5]KIIIA	GGGGGCCNCSSKWCRDHSRCC *
	P[LPATG5]K	DCLGFLWKCNPSNDKCCRPNLVCSRKDKWCKYQILPATGGGGGCCNCSSKWCRDHSRCC *
wt-OaAEP 1	PaurTx3[N|GL]	DCLGFLWKCNPSNDKCCRPNLVCSRKDKWCKYQINGL
	[G8]KIIIA	GGGGGGGGCCNCSSKWCRDHSRCC*
	Pr[NG8]K	DCLGFLWKCNPSNDKCCRPNLVCSRKDKWCKYQINGGGGGGGGCCNCSSKWCRDHSRCC*
	PaurTx3[N|GL]	DCLGFLWKCNPSNDKCCRPNLVCSRKDKWCKYQINGL
	[GVG6]KIIIA	GVGGGGGGCCNCSSKWCRDHSRCC*
	P[NGVG6]K	DCLGFLWKCNPSNDKCCRPNLVCSRKDKWCKYQINGVGGGGGGCCNCSSKWCRDHSRCC*
C247A-OaAEP 1	PaurTx3[N|GL]	DCLGFLWKCNPSNDKCCRPNLVCSRKDKWCKYQINGL
	[GVG6]KIIIA	GVGGGGGGCCNCSSKWCRDHSRCC*
	PrNGVG6lK	DCLGFLWKCNPSNDKCCRPNLVCSRKDKWCKYQINGVGGGGGGCCNCSSKWCRDHSRCC*
Butelase 1	PaurTx3[N|HV]	DCLGFLWKCNPSNDKCCRPNLVCSRKDKWCKYQINHV
	[GVG6]KIIIA	GVGGGGGGCCNCSSKWCRDHSRCC*
	P[NGVG6]K	DCLGFLWKCNPSNDKCCRPNLVCSRKDKWCKYQINGVGGGGGGCCNCSSKWCRDHSRCC*
KAHA	PaurTx3-ka#	DCLGFLWKCNPSNDKCCRPNLVCSRKDKWCKYQL-COOH
	[OprG8]KIIIA	OprGGGGGGGGCCNCSSKWCRDHSRCC*
	P[hSG8]K	DCLGFLWKCNPSNDKCCRPNLVCSRKDKWCKYQLhSGGGGGGGGCCNCSSKWCRDHSRCC*
Omniligase 1	PaurTx3[AVRL|-OCamL]	DCLGFLWKCNPSNDKCCRPNLVCSRKDKWCKYQAVRL-OCamL
	[FIG3]KIIIA	FIGGGCCNCSSKWCRDHSRCC*
Thymoligase 1	PaurTx3[ITTK|-OCamL]	DCLGFLWKCNPSNDKCCRPNLVCSRKDKWCKYQITTK-OCamL
	[DLG3]KIIIA	DLGGGCCNCSSKWCRDHSRCC*

^a|: enzyme cleavage site, red: C-terminal recognition motifs/mutations, blue: N-terminal polyglycine/recognition motif.

^*C: terminal amide

^#ka: Leu-a-ketocarboxylic acid, Opr: oxaproline

^hS: homoserine, and OCam: carboxyamidomethyl ester.

OaAEP 1 Ligation.

All peptide precursors used in this study were synthesized using fluorenylmethyloxycarbonyl (Fmoc) chemistry and cleaved individually. KIIIA was assembled with an eight-residue polyglycine at the N-terminus ([G₈]KIIIA), while PaurTx3 was assembled with the OaAEP 1 recognition site, NGL, at its C-terminus (PaurTx3[NGL]) (Table 1). Under thermodynamic oxidizing conditions, PaurTx3[NGL] formed only one major product during oxidation (28% yield) (Figure S1A), while [G₈]KIIIA formed two isomers as reported previously (Figure S2A).¹² While μ-conotoxins typically display a disulfide connectivity of C1–C9, C2–C15, and C4–C16,⁵⁰ full 3D structure characterization and direct mass spectrometric collision-induced dissociation fragmentation of KIIIA carried out by Khoo et al. confirmed that the major isomer observed following thermodynamic oxidation of KIIIA display a C1–C15, C2–C9, and C4–C16 disulfide bond connectivity.⁵¹ As reported previously,¹² comparison of the 1D ¹H NMR amide region of [G₈]KIIIA with published KIIIA (PDB ID: 2lxg)⁵¹ (Figure S3A) and secondary Hα chemical shifts of both peptides suggests that the major isomer following folding of [G₈]KIIIA exhibits a similar overall structure compared to the major KIIIA isomer with C1–C15, C2–C9, and C4–C16 connectivity reported by Khoo et al. (Figure S3B).⁵¹ The major oxidation isomer of [G₈]KIIIA (yield: 38%) was subsequently used for ligation. Ligation conditions and reaction lengths were optimized for all strategies (Figure S4) and are reported in Table 2. Folded PaurTx3[NGL] and [G₈]KIIIA were ligated in phosphate buffer using OaAEP1 (Table 2), with reaction progress monitored at 0, 2, 4, 6, 8, and 24 h (Figure S4A). The desired product was identified using ESI-MS and purified by reversed-phase high-performance liquid chromatography (RP-HPLC) at final yield of 12% (Figures 2A and 3A). Since it has been reported that a GV recognition site in the N-terminal region of the C-terminal fragment improves OaAEP 1 catalysis and that ligation with a GV recognition site is irreversible,⁵² we also synthesized [GVG₆]KIIIA. Under thermodynamic oxidizing conditions, [GVG₆]KIIIA, like [G₈]KIIIA, also formed two isomers. The predominant [GVG₆]KIIIA isomer (40% yield) (Figure S2B), evaluated by 1D and 2D ¹H NMR (as described above) to confirm correct folding (Figure S3), was subsequently used for ligation. The ligation was conducted using both the wild-type and the C247A-OaAEP 1 enzyme. When treated with the wild-type enzyme, [GVG₆]KIIIA (yield 16%) (Figure 2B) showed little improvement in ligation yield compared to [G₈]KIIIA (yield: 12%) (Figure 2A). However, as expected, the reaction in which [GVG₆]KIIIA was incubated with the C247A-OaAEP 1 mutant produced a higher yield (35% yield) (Figure 2C and Figure S4B). The 1D ¹H NMR spectra of the ligated peptides superimposed well with the individual peptide domain spectra, thus suggesting that the disulfide bonds were intact during ligation (Figure S5A). Furthermore, no disulfide bond shuffling was observed during ligation, as evidenced by a single ligation peak in the HPLC chromatograms (Figure 2 and Figure S4). In addition, Hα secondary shift analysis derived from 2D NMR TOCSY and NOESY spectra of the bivalent peptides overlapped well with previously published PaurTx3 (PDB ID: 5we3)⁵³ and KIIIA (PDB ID: 2lxg),⁵¹ confirming that the enzymatic ligation does not shuffle already formed disulfide bonds (Figure S5A,B and Figure 3D).

Table 2.

Summary of Ligation Conditions and Yields of each Oxidation and Ligation Approach

strategy	KIIIA precursor oxidation yielda	PaurTx3 precursor oxidation yielda	conditionc	reaction time	ligation yieldb
SrtA5°	40%	50%	100 μM PaurTx3[LPATGG]:300 μM [G8]KIIIA:30 μM enzyme 50 mM Tris:150 mM NaCl:10 mM CaCl2, pH 7.4, 37 °C	15 min	60%
wt-OaAEP 1	38%	28%	250 μM PaurTx3[NGL]:750 μM [G8]KIIIA:3 μM enzyme 0.1 M Na2HPO4:0.1 M NaCl, pH 6.5	4 h	12%
C247A-OaAEP 1	40%	18%	20 μM PaurTx3[NGL]:60 μM [GVG6]KIIIA:75 nM enzyme 0.1 M Na2HPO4:0.1 M NaCl, pH 6.5	6 h	35%
butelase 1	40%	29%	20 μM PaurTx3[NHV]:60 μM [GVG6]KIIIA:100 nM enzyme 0.1 M Na2HPO4:0.1 M NaCl, pH 6.5	6 h	38%
KAHA	25%	28%	22 mM PaurTx3-ka:15 mM [OprG8]-KIIIA	6 h	36%
peptiligased	40%	n/d	90% DMSO:10% H2O, 0.1 M oxalic acid, 60 °C 10 mM PaurTx3[linker]-OCamL:15 mM [linker]-KIIIA 10% DMSO in 1 M potassium phosphate, pH 8–8.5 (with/without 2	n/a	n/d

^aOxidation yield was calculated based on the precursor weight and isolated oxidized peptide weight.

^bLigation yield was calculated by dividing the observed weight by the theoretical weight calculated from PaurTx3 analogs.

^cReactions carried out at room temperature unless specified.

^dLigation was not carried due to insolubility of peptide fragments.

Figure 2.

Enzymatic and chemical ligation approaches of PaurTx3 and KIIIA. RP-HPLC chromatograms of PaurTx3 and KIIIA starting material (top) and ligated products (bottom) following completion of the ligation reaction. (A) Ligation of PaurTx3[NGL] and [G₈]KIIIA using wt-OaAEP 1. (B) Ligation of PaurTx3[NGL] and [GVG₆]KIIIA using wt-OaAEP 1. (C) Ligation of PaurTx3[NGL] and [GVG₆]KIIIA using C247A-OaAEP 1. (D) Ligation of PaurTx3[NHV] and [GVG₆]KIIIA using butelase 1. (E) KAHA ligation between PaurTx3-ka and [OprG₈]KIIIA. (F) Ligation of PaurTx3[LPATGG] and [G₅]KIIIA using SrtA5° (SrtA5° data adapted from Tran et al. and included for comparison purposes¹²).

Figure 3.

Enzymatic and chemical ligation products following PaurTx3 and KIIIA ligation. Analytical RP-HPLC traces and observed masses (ESI-MS) of purified bivalent peptides. (A) Product from wt-OaAEP 1 ligation using PaurTx3[NGL] and [G₈]KIIIA. (B) Product from wt-OaAEP 1 (identical product from C247A-OaAEP 1, not shown) and butelase 1 ligation using PaurTx3[NGL] and PaurTx3[NHV], respectively, and [GVG₆]KIIIA. (C) Product from KAHA ligation using PaurTx3-ka and [OprG₈]KIIIA. (D) Secondary Hα chemical shifts for P[NG₈]K, P[NGVG₆]K, and P[^hSG₈]K, overlaid with published KIIIA (from Khoo et al.),⁵¹ and native PaurTx3 (from Agwa et al.).⁵³ The linker region was omitted for clarity.

Butelase 1 Ligation.

Butelase 1 is an asparaginyl endopeptidase like OaAEP 1 with a similar ligation mechanism and recognition motif, and therefore, [GVG₆]KIIIA could again be used as the C-terminal fragment for butelase 1 ligation.^31,33,34 PaurTx3 was assembled with the NHV butelase 1 recognition site at its C-terminus (PaurTx3[NHV]) (Table 1). Folding of PaurTx3[NHV] formed one primary oxidation product (29% yield) (Figures S1B and S3). The two fragments were dissolved in butelase 1 ligation buffer (Table 2), and ligation reaction was monitored by LC/MS at room temperature for 6 h (Figures 2D and 3B and Figure S4C). The desired product was formed at 38% yield and was examined by 1D and 2D NMR (Figure 3D and Figure S5C).

Peptiligase.

The substrate-tailored variants of peptiligase, omniligase 1, and thymoligase 1 were used for ligation of bivalent peptides. KIIIA was synthesized with three Gly residues followed by either FI or DL at the N-terminus ([FIG₃]KIIIA or [DLG₃]KIIIA) for omniligase and thymoligase, respectively. PaurTx3 was assembled with glycolic acid to yield the ester linkers ARVL-OCamL or ITTK-OCamL at its C-terminus (PaurTx3[ARVL-OCamL] or PaurTx3[ITTK-OCamL]) for omniligase and thymoligase, respectively (Table 1). Peptides were cleaved from the resin, isolated by HPLC, and oxidized individually. Under thermodynamic oxidizing conditions previously described, the KIIIA analogs [FIG₃]KIIIA or [DLG₃]KIIIA formed one major isomer each (Figure S2C,D). Unfortunately, while PaurTx3[ARVL-OCamL] and PaurTx3[ITTK-OCamL] displayed one main HPLC peak following thermodynamic oxidation using two different folding conditions (Figure S1C,D), these peptides appeared misfolded when examined using 1D ¹H NMR (Figure S3). Furthermore, both the linear and the misfolded PaurTx3-OCamL esters (5–10 mM) precipitated in ligation buffer (0.1–1 M phosphate buffer pH 8.3, with or without 2 M urea, 1 M Gn·HCl, or DMSO). Thus, ligation of PaurTx3 and KIIIA using omniligase 1 or thymoligase 1 could not be attempted. Peptiligase ligation reaction require an optimal pH between 8.0 and 8.5 as at lower pH (<7.5), a significant amount of the N-terminal amine is protonated and thus unreactive.⁵⁴ It is likely that both linear and oxidized PaurTx3 analogs precipitated in potassium phosphate ligation buffer as the theoretical pI of these PaurTx3 analogs is 8.8 and thus incompatible with these conditions.

KAHA Ligation.

The C-terminal KIIIA-fragment used for KAHA-ligation was synthesized with eight Gly residues followed by an oxaproline (Opr) at the N-terminus ([OprG₈]-KIIIA), while the N-terminal PaurTx3-fragment was assembled with a Leu-α-ketocarboxylic acid (ka) at the C-terminal in place of Ile34 (PaurTx3-ka) (Table 1). Peptides were cleaved from the resin and oxidized individually. Under thermodynamic oxidizing conditions, both PaurTx3-ka and [OprG₈]-KIIIA formed one primary product, with yields of 28 and 25%, respectively (Figures S1E, S2E, and S3). Folded peptides were ligated for 6 h in 90% DMSO/10% H₂O and 0.1 M oxalic acid (Table 2), and the desired product was purified with a yield of 36% (Figures 2E and 3C and Figure S4D). Following ligation, treatment with NH₄HCO₃ resulted in a rightward shift in RP-HPLC retention time, indicating that O → N shift (Figure S4E) was completed and P[^hSG₈]K, containing a homoserine residue in its linker, was formed via a native amide bond (Figure 3C). The ligated peptide P[^hSG₈]K was examined by 1D and 2D NMR and compared to native PaurTx3 and KIIIA to ensure that the integrity of the disulfide bonds was intact and the overall fold was maintained (Figure S5D and Figure 3D).

In addition to generating the individual oxidized domains of PaurTx3-ka and [OprG₈]-KIIIA and ligating them, we also produced the linear peptide P[^hSG₈]K using KAHA ligation prior to oxidation of the two individual domains (Figure S6A). Following successful ligation and purification of the linear peptide, we attempted to fold the full-length linear peptide. However, the desired folding of the peptide was not achieved using our standard PaurTx3 oxidation condition (0.1 M Tris/0.81 mM GSH/0.81 mM GSSG) nor the KIIIA oxidation condition (0.1 M NH₄HCO₃/100 equiv GSH/10 equiv GSSG, pH 8) (Figure S6B), highlighting the importance of our modular folding strategy.

Automated Patch-Clamp Electrophysiology.

To assess the impact of linker types on the activity of the bivalent peptides, all successfully ligated bivalent peptide constructs were tested for activity at human Na_V1.7 (hNa_V1.7) expressed in HEK293 cells using automated whole cell patch-clamp electrophysiology (Figure 4A–C and Figure S7).

Figure 4.

Pharmacology at hNa_V1.7 and stability of bivalent peptides. (A) Representative current trace before and after the addition of 100 nM P[NG₈]K. (B) Representative current trace before and after the addition of 100 nM P[NG₈]K. (C) Representative current trace before and after the addition of 100 nM P[^hSG₈]K. (D) Concentration–response curves of bivalent peptides on hNa_V1.7 expressed in HEK293 cells assessed using automated whole cell patch-clamp electrophysiology (n = 3). (E) Serum stability of P[NGVG₆]K (ligated with wt-OaAEP 1, C247A-OaAEP 1, or butelase 1), P[^hSG₈]K (KAHA), and P[LPATG₅]K (ligated with SrtA5°) in human serum over 24 h. The linear linker sequence KYQILPATG₅ (SrtA5° linker sequence) was used as the positive control. P[LPATG₅]K and linear linker sequence data are adapted from Tran et al.¹² for comparison purposes.

All bivalent peptides produced in this study performed better than PaurTx3 and KIIIA (Figure 4D) individually, with minor differences in the linker sequences resulting in negligible effects on activity at Na_V1.7 (P[NG₈]K IC₅₀ = 74 ± 0.9 nM, P[NGVG₆]K IC₅₀ = 55 ± 0.8 nM, P[^hSG₈]K IC₅₀ = 59 ± 0.8 nM, P[LPATG₅]K IC₅₀ = 57 ± 0.9 nM) (Figure 4D).

Serum Stability.

Proteolytic stability of P[NGVG₆]K, P[^hSG₈]K, and P[LPATG₅]K was determined by incubation in human serum and compared to the linear linker used in SrtA5° ligation (Figure 4E).¹² Interestingly, incorporation of different linkers, arising from the different enzymes and ligation strategies employed in this study, gave rise to differences in serum stability. After 24 h of incubation, P[^hSG₈]K, chemically ligated using KAHA-ligation, proved to be the most proteolytically stable analog with 73.3 ± 3.3% of the peptide remaining, followed by P[LPATG₅]K, ligated by SrtA5° with 66.9 ± 3.3% remaining,¹² and P[NGVG₆]K, ligated with either wt-OaAEP 1, C247A-OaAEP 1, or butelase 1, with 39.4 ± 1.0% peptide remaining (Figure 4E and Figure S8).

DISCUSSION

Peptide toxins that target Na_Vs are hailed as promising candidates for future analgesic treatments. However, despite extensive discovery and development efforts focused on potent and selective venom-derived gating-modifier toxins targeting hNa_V1.7, this potential is yet to be realized.^55–60 Recently discovered double-knotted spider toxins^1–3 and enabling synthetic methods^{10,11,14–19} have paved a path for the development of an exciting new class of pharmacological tools and potential drug leads, involving conjugation of two potent Na_V inhibitors with different modes of action to create synthetic bivalent constructs. However, the route to synthetically accessing this new class of bivalent peptides is not straightforward due to complications introduced by the presence of complex disulfide bond frameworks in each individual peptide domain, prohibiting the use of most standard chemical ligation techniques. We therefore evaluated a series of enzymatic and chemical strategies to delineate the most facile and high yielding method of producing these peptides.

By employing automated Fmoc solid-phase peptide synthesis protocols, individual peptide precursors, including the enzyme-specific recognition sequences incorporated at the N- or C-termini of the peptides designed for the different ligations evaluated in this study, were successfully assembled in high yield and purity. All peptide precursors, apart from PaurTx3- [ARVL-OCamL] and PaurTx3[ITTK-OCamL], were successfully oxidized, and structures of the individual peptides incorporating the different ligation recognition motifs were evaluated by 1D and 2D NMR and shown to correlate well with their native counterparts, confirming that the correct overall secondary structure of the peptides was maintained. Although none of the ligation approaches evaluated achieved complete conversion, four approaches, including SrtA5°, OaAEPs, butelase 1, and KAHA-ligation, proved their ability to ligate complex disulfide-rich peptides under mild conditions without any detectable disulfide bond reduction or shuffling.

OaAEP 1 ligation resulted in a clean reaction with the desired product and unreacted starting materials being easy to recover. Due to the removal of the hydrophobic Gly-Leu recognition motif in the C-terminal of P[GVG₆]K following the ligation, the ligated product was easily separated from the more hydrophobic starting material PaurTx3[NGL] by RP-HPLC. Additionally, the small amount of added enzyme resulted in easy separation of the desired product during purification. Despite this, wt-OaAEP 1 reaction was the least efficient approach evaluated in this study. After several attempts to ligate PaurTx3[NGL] and [G₈]KIIIA, a maximum yield of only 12% was achieved after 4 h. OaAEP 1 has been reported to be an effective catalyst when cyclizing cyclotide kB1, but it was also observed that the kinetic cyclization rate of OaAEP 1 on other peptides was inferior to that of kB1.^27,61 The interaction between the peptide and the binding pocket of OaAEP 1 has been shown to play a vital role in catalytic efficiency, and peptides with an N-terminus containing GL or GV present better kinetics than other residues, including VF or VK.^52,62,63 The C-terminal amino acids following N in the recognition motif have also been shown to be critical, and NGL, NGI, and NGV have been shown to produce more efficient cleavage sites.^52,62,63 Thus, OaAEP 1 ligation was also attempted with PaurTx3[NGL] and [GVG₆]KIIIA. This combination allowed ligation of the two fragments to proceed using less peptide starting material and enzyme, and the yield also increased slightly to 16%. When the ligation was performed using C247A-OaAEP 1 with the same peptide fragments, an improved yield of 35% was observed. This result agrees with the findings of Rehm et al.⁵² and Yang et al.²⁸ that C247A-OaAEP 1 displays improved kinetics compared to wt-OaAEP 1, resulting in the more efficient ligation of peptides and proteins. Taken together, these results show that OaAEP 1 is a highly site-specific enzyme, but both N- and C-terminal Asn residues of peptide fragments interfere with OaAEP 1 activity, thus affecting the ligation rate of the peptide. Conversely, OaAEP 1 ligation leaves only an Asn residue footprint in the sequence following ligation. Several studies have demonstrated that the addition of small molecules can further improve the yields of ligated products through quenching the released nucleophiles when using OaAEP 1.⁶⁴ This can be attempted in future strategies when ligating bivalent peptides with OaAEP 1.

Despite their sequence homology, butelase 1 and OaAEP 1 produced different ligation yields in this study. Butelase 1 ligation of PaurTx3[NHV] and [GVG₆]KIIIA took about 6 h to reach the maximum yield (38%) and showed a similar kinetic rate to C247A-OaAEP 1, although butelase 1 has a reported ligation rate of 90 times wt-OaAEP 1.^27,28 In our hands, the yield from butelase 1 ligation was only double that of the yield from using wt-OaAEP 1 when applied to the same fragments, and the reaction was slower, 6 h for butelase 1 compared to 4 h for wt-OaAEP 1.

The only chemical approach evaluated in this study, the KAHA ligation, was found to be a promising approach for the chemical conjugation of two peptides each comprising a complex disulfide bond network. With both peptides highly concentrated (15–25 mM), the ligation reaction produced a 36% yield after 6 h. Therefore, ligation using the KAHA method was comparable with respect to yields resulting from butelase 1 and C247A-OaAEP 1 ligations. The P[^hSG₈]K bivalent peptide produced using KAHA ligation was also the most stable in serum compared to those produced using the enzymatic ligations.

To demonstrate the importance of our modular ligation strategy, we produced linear P[^hSerG₈]K using KAHA ligation (Figure S6A) and attempted to fold the peptide using our standard PaurTx3 oxidation condition or KIIIA oxidation (Figure S6B), but the desired folding of the peptide was not achieved. While it was straightforward to oxidize each precursor domain in good yield, it was challenging for each domain to obtain its native fold following ligation of the linear peptides. This result emphasized the importance of pre-folding disulfide-rich peptide fragments into their correct structure before ligating them with their partners.

Overall, the SrtA5° approach proved to be highly efficient for ligating disulfide-rich peptide fragments. Despite ligated peptides containing the residual SrtA5° recognition sequence being vulnerable to hydrolysis by the enzyme, we were able to isolate conjugated peptide with a yield of 60% within 15 min. In addition, the synthesis of both SrtA5° precursor fragments produced more materials than other strategies (Table 2).

Evaluation of the activity of the ligated peptides on hNa_V1.7 using automated whole cell patch-clamp electrophysiology demonstrated that the amino acid sequence in the linkers in this case is less important than having the correct linker length, which has been previously proven to be critical for optimal activity of ligated PaurTx3 and KIIIA at hNa_V1.7.¹² All successful ligation approaches delivered bivalent peptides with similar potency, suggesting that the linker sequence neither significantly contributes to channel interactions nor to the orientation of the two ICK domains. On the other hand, evaluation of serum stability of the bivalent peptides included in this study proved to be affected by the composition of amino acids in the linker sequence. The most stable peptide was found to be the chemically KAHA-ligated P[^hSG₈]K, with 73.3 ± 3.3% of the peptide remaining after 24 h. This is most likely due to the presence of an unnatural residue (homoserine) in the linker sequence of P[^hSG₈]K, leading to increased resistance to proteases present in human serum compared to the other peptides. While the SrtA5° [LPATG₅] linker in P[LPATG₅]K also displayed promising stability in serum with 66.9% ± 1.8% remaining after 24 h, the [GVG₆] linker used for ligation with OaAEP 1 (wt and C247A) and butelase 1 of P[GVG₆]K resulted in only 39.4% ± 1.0% of ligated peptide remaining after 24 h.

Although both OaAEP 1 and butelase 1 have previously been used to cyclize disulfide-rich peptides,^{27,29–34,52,61–63} to our knowledge, this is the first study to evaluate their ability to ligate two previously oxidized disulfide-rich peptides. In addition, we highlight the versatility of the KAHA-ligation strategy, being able to chemically ligate complex disulfide-rich peptides without reducing or scrambling the disulfide bonds. Although peptiligases, including thymoligase 1 and omniligase 1, have been proven to be promising enzymes for protein and peptide labeling and ligation in previous studies,^{35–39,52,54,62,65,66} due to the oxidation of PaurTx3[ARVL-OCamL] and PaurTx3[ITTK-OCamL] resulting in misfolded peptide and the pI of the peptides being too close to the required pH for enzyme activity, the peptide fragments precipitated and we could not investigate the efficiency of thymoligase 1 and omniligase 1 in ligating two folded disulfide-rich peptides.

All the ligation strategies used in this work are summarized in Table 3 and are suitable for peptide ligation and cyclization. However, they all come with their own specific advantages and disadvantages. Sortase A-mediated ligation is the most efficient when a more extended enzyme recognition motif in the ligation product is not an issue. The enzyme is straightforward to produce in high yield, resulting in good accessibility to any laboratory. For peptide cyclization, OaAEP 1 and butelase 1 are two of the most efficient enzymes to date, although application and use of these enzymes have been limited due to the poor accessibility, resulting from challenging recombinant expression and low yield during extraction from native materials.^33,34 For traceless ligation and cyclization of peptides, peptiligases are considered the best option. The ligation reaction is fast with high ligation yield, and the enzymes are easy to produce. However, an active ester is required as the starting material, which can be challenging to synthesize. Chemical ligation using the KAHA methodology was not the most efficient ligation strategy in terms of conversion of starting materials, but the method requires a short motif and leaves only a homoserine as a footprint at the ligation site, making this ligation suitable for shorter linkers or for mutational studies of the linker itself. After considering all aspects of the ligation strategies investigated here, SrtA5° is, in our hands, the most efficient strategy for ligation of complex disulfide-rich peptides overall. Despite leaving a large footprint of LPATG at the ligation site, in terms of the straightforward production of enzyme, efficiency of ligation, final yield, and ease of purification of the desired product, SrtA5° is the best, provided that the linker is of sufficient length for the enzyme to access the recognition motifs. We envisage that easy synthetic access will open the door to future development of this hitherto underexplored class of molecules.

Table 3.

Advantages and Disadvantages of Ligation Strategies Applied in this Study

ligation strategies	advantages	disadvantages
sortase A	ultrafast ligation, good accessibility to the enzyme, high folding/ligation yield	long enzyme motif, motif is intolerant to modification, susceptible to hydrolysis
OaAEP 1/butelase 1	short enzyme motif, traceless ligation, ease of purification	medium ligation, yield poor, accessibility to enzyme challenging
peptiligase	traceless amount of used enzyme	synthesis of ester peptides is challenging; not suitable for long, hydrophobic peptides; ligation requires concentrated starting materials
KAHA	short recognition motif, homoserine footprint increases protease stability	limited access to the building blocks, ligation requires concentrated starting materials

METHODS

Peptide Synthesis.

An automated microwave Liberty Prime synthesizer (CEM, USA) was used to assemble peptide precursors on Fmoc-polystyrene resin (RAPP Polymere, 0.72 mmol/g, 0.1 mmol scales) for peptide C-terminal carboxamides and either ketocarboxylic Leu on rink amide resin (Bode Lab, 0.28 mmol/g, 0.1 mmol scales) for PaurTx3-ka or pre-loaded Wang (LL) resin (CEM, 0.31 mmol/g, 0.1 mmol scales) for C-terminal peptide carboxylates. Couplings were performed in dimethylformamide (DMF) using 5 equivalents (equiv) of Fmoc-protected amino acid/5 equiv ethyl cyanohydroxyiminoacetate (Oxyma Pure)/10 equiv diisopropylcarbodiimide (DIC), relative to resin substitution, for 1 min at 105 °C. Pyrrolidine (25% v/v in DMF) was used to remove the Fmoc protecting group (40 s at 100 °C). Cleavage and simultaneous removal of sidechain protecting groups were accomplished by treatment with 95% trifluoracetic acid (TFA)/2.5% triisopropylsilane (TIPS)/2.5% H₂O for 2 h at room temperature. Following filtration of cleavage solution, ice-cold diethyl ether (Et₂O) was added to precipitate the crude peptide. The crude peptides were centrifuged (5000 rpm for 5 min × 3) in Et₂O then lyophilized in 0.05% TFA/50% acetonitrile (ACN)/H₂O.

Synthesis with Fmoc-Leu-OCam-Ester.³⁸

Fmoc-glycolic acid (2 eq) was manually coupled at room temperature on Leu pre-loaded Rink amide resin (RAPP Polymere, 0.63 mmol/g) using DIC (5 equiv) and Oxyma pure (5 equiv) in DMF for 30 min. The Fmoc group was deprotected by 25% piperidine in DMF for 2 min × 2. The four amino acids after glycolic acid were coupled manually using 4 equiv of amino acid/4 equiv of DIC/4 equiv of Oxyma pure (20 min × 2). The treated resin was then transferred to Liberty Prime to complete the sequences. Couplings were performed in DMF using 5 equiv of Fmoc-protected amino acid/5 equiv of Oxyma pure/10 equiv of DIC for 2 min at 90 °C, and Fmoc was deprotected by 25% pyrrolidine in DMF 1 min at 90 °C.

Oxidative Folding.

Purified linear PaurTx3 analogs were dissolved in a minimum volume of aqueous ACN and then diluted to a peptide concentration of 0.1 mg/mL with 0.1 M Tris buffer at pH 8 containing 0.81 mM GSH/0.81 mM GSSG and stirred for 24 h at room temperature.⁵³ The reaction was stopped by adding 1% TFA/H₂O to pH ~2, and the correctly folded peptide was isolated by HPLC. Purity and correct mass were confirmed by RP-HPLC and ESI-MS, respectively, and 1D ¹H NMR experiments were used to confirm the correct fold.

[OprG₈]KIIIA was folded in 20% DMSO/H₂O for 24 h.Unless otherwise specified, linear KIIIA analogs were folded with a peptide concentration of 0.3 mg/mL in a 0.1 M NH₄HCO₃ buffer at pH 8 containing 100 equiv of GSH/10 equiv of GSSG and stirred for 24 h at room temperature. Following addition of 1% TFA/H₂O to lower the pH to ~2, the peptide was isolated by preparative HPLC and the correct fold was confirmed by ESI-MS and NMR.

Reversed-Phase High-Performance Liquid Chromatography.

Preparative and analytical HPLC were carried out on a Shimadzu LC-20AT system equipped with an SPD-20A Prominence UV/VIS detector, an SIL-20AHT auto-injector, and an FRC-10A fraction collector. An Eclipse XDB–C18 column (Agilent; 7 μm, 21.2 cm × 250 mm, 80 Å, flow rate 8 mL/min) with a gradient from 0 to 50% solvent B in 50 min was used to purify the peptides. A Hypersil GOLD–C18 column (Thermo Fisher; 3 μm, 2.1 × 100 mm, 175 Å, flow rate 0.3 mL/min) with a gradient 0–60% B over 30 min was used to monitor oxidation and ligation and to analyze the purity of the peptides. Absorbance was monitored at 214 and 280 nm. (Solvent A: 0.05% TFA in H₂O; solvent B: 90% ACN/0.05% TFA in H₂O).

Sortase Ligation.

Peptide ligation was performed using a procedure adapted from Agwa et al.^11,67 Briefly, PaurTx3[LPATGG] and [G₈]KIIIA were dissolved in SrtA ligation buffer containing 50 mM Tris, 150 mM NaCl, and 10 mM CaCl₂ at pH 8 at concentrations of 100 and 300 μM, respectively. SrtA5° was added to this solution to a concentration of 30 μM. The mixture was incubated at 37 °C for 15 min and then acidified with 1% aqueous TFA to stop the reaction and purified by RP-HPLC.

OaAEP 1 Ligation.

For OaAEP 1 ligation, 250 μM PaurTx3[NGL] and 750 μM [G₈]KIIIA were dissolved in ligation buffer containing 50 mM Tris, 150 mM NaCl, and 10 mM CaCl₂ at pH 6. wt-OaAEP 1 was added to this solution to a concentration of 3 μM. The mixture was left to react at room temperature for 4 h and then acidified with 1% aqueous TFA to stop the reaction. PaurTx3[NGL] (20 μM) and 60 μM [GVG₆]KIIIA were dissolved in ligation buffer containing 50 mM Tris, 150 mM NaCl, and 10 mM CaCl₂ at pH 6, and wt-OaAEP 1 (or C274A-OaAEP 1) was added to this solution to a concentration of 75 nM. The mixture was left to react at RT for 6 h and then acidified with 1% aqueous TFA to stop the reaction, and the ligated peptide was purified by RP-HPLC.

Butelase 1 Ligation.

PaurTx3[NHV] and [GVG₆]KIIIA were dissolved in ligation buffer containing 50 mM Tris, 150 mM NaCl, and 10 mM CaCl₂ at pH 6 at concentrations of 20 and 60 μM, respectively. Butelase 1 was added to this solution to a concentration of 100 nM. The mixture was left to react for 6 h and then acidified with 1% aqueous TFA to stop the reaction, and the ligated peptide was purified by RP-HPLC.

Omniligase 1 and Thymoligase 1.

The protocol was previously described by Schmidt and Nuijens.⁵⁴ Briefly, peptide fragments were dissolved in 50% ACN/H₂O or DMSO before being added to enzymatic ligation buffer (1 M potassium phosphate, pH 8.2–8.4, with or without 2 M urea or 1 M Gn·HCl) to make final concentrations of 10 mM Pep-OCamL:15 mM H₂N-Pep:8 μM omniligase 1 (or thymoligase 1).

KAHA Ligation.

Oxidized PaurTx3-ka (22 mM) and oxidized [OprG₈]KIIIA (15 mM) were dissolved in 0.1 M oxalic acid in DMSO:H₂O (9:1) and incubated at 60 °C for 6 h to yield the peptide ester. Similarly for the linear peptide ligation, 22 mM linear PaurTx3-ka and 15 mM linear [OprG₈]KIIIA were dissolved in 0.1 M oxalic acid in DMSO:H₂O (9:1) with 50 equiv of TCEP and incubated at 60 °C for 7 h to yield the peptide ester. The reactions were monitored by LC/MS, then diluted with buffer A, and purified by RP-HPLC. Lyophilized peptide ester was dissolved in 0.1 M NH₄HCO₃ at pH 9.5 for 2 h. The O → N shift was monitored by HPLC, then acidified by 1% TFA, and purified.

NMR.

Peptides were dissolved in 500 μL of MilliQ water (MilliPore, USA) and 50 μL of D₂O (Cambridge Isotopes). A Bruker 600 MHz Avance III spectrometer equipped with a cryoprobe (Bruker, Billerica, MA, USA) was used to acquire 1D ¹H NMR spectra, and 2D ¹H-¹H TOCSY (80 ms mixing time) and NOESY (200 ms mixing time) were collected,^68,69 as described by Agwa et al.¹¹ Spectra were processed using TopSpin 3.5 (Bruker), and CCPNMR Analysis 2.5.2 (CCPN, University of Cambridge, Cambridge, UK) was used to assign resonances.^70,71 The chemical shift of water at 4.76 ppm was used as reference.

Serum Stability.

Serum stability assays were performed as previously described.^12,72 Briefly, male human AB serum (Sigma) was thawed to room temperature and centrifuged at 17,000g for 15 min to separate and remove the lipid component. The supernatant was incubated at 37 °C for 15 min, and then each peptide was incubated in either 100% serum or PBS at 37 °C at 20 μM. Samples (40 μL) were removed at time points in triplicate (0, 2, 4, 6, 12, and 24 h). Each sample was quenched with 40 μL of 6 M urea and incubated at 4 °C for 10 min. Serum proteins were precipitated with 40 μL of 20% trichloroacetic acid (TCA) followed by another 10 min incubation at 4 °C. Samples were centrifuged at 17,000g for 10 min to remove precipitated proteins. The supernatant was analyzed by analytical HPLC with a linear 0– 50% solvent B gradient over 50 min. Peptides were quantified by integration at 215 nm and identified by mass spectrometry.

HEK 293 Cell Culture.

HEK 293 cells heterologously expressing human Na_V1.7 and β1 subunit (SB Drug Discovery, Glasgow, UK) were used for all electrophysiology experiments. Cells were cultured in minimum essential medium Eagle supplemented with 10% (v/v) heat-inactivated fetal bovine serum (HI-FBS), 1% (v/v) GlutaMAX, 4 μg/mL blasticidin, and 0.6 mg/mL geneticin at 37 °C and 5% CO₂. Cells were passaged every 3–4 days at 70–80% confluence using TrypLE Express (Thermo Fisher Scientific).

QPatch Electrophysiology.

Electrophysiology assays were performed as previously described using a QPatch-16 automated electrophysiology platform (Sophion Bioscience, Ballerup, Denmark).⁷³ HEK293 cells expressing hNaV1.7/β1 were dissociated and suspended in Ham’s F12 containing 25 mM HEPES, 100 U/mL penicillin–streptomycin, and 0.04 mg/mL soybean trypsin inhibitor. The suspension was stirred for 30–60 min prior to use. The extracellular solution contained (in mM) 70 NaCl, 70 choline chloride, 4 KCl, 2 CaCl₂, 1 MgCl₂, 10 HEPES, and 10 glucose and was adjusted to pH 7.4 and osmolarity to 305 mOsm. The intracellular solution contained (in mM) 140 CsF, 1 EGTA, 5 CsOH, 10 HEPES, and 10 NaCl and was adjusted to pH 7.3 using CsOH and osmolarity to 320 mOsm.

As described previously,^12,74 peptides were diluted in extracellular solution containing 0.1% bovine serum albumin. Each peptide concentration was incubated for 5 min, and the peak current was compared to the buffer control. Concentration–response curves were acquired using a holding potential of −90 mV and a 50 ms pulse to −20 mV every 20 s (0.05 Hz). Peak current post-peptide addition (I) was normalized to the buffer control (I₀). IC₅₀ values were determined by plotting the difference in peak current (I/I₀) and log peptide concentration. Calculated IC₅₀ values were compared across subtypes, and statistical differences were determined by ordinary one-way ANOVA. I–V curves were obtained with a holding potential of −90 mV followed by a series of 500 ms step pulses that ranged from −110 to +55 mV in 5 mV increments (repetition interval: 5 s). Conductance– voltage curves were obtained by calculating the conductance (G) at each voltage (V) using the equation G = I/(V – V_rev), where V_rev is the reversal potential and was fitted with a Boltzmann equation. Voltage dependence of steady-state fast inactivation was assessed using a 10 ms pulse of −20 mV immediately after the 500 ms step (described above) to assess the availability of non-inactivated channels.

Data Analysis.

GraphPad Prism, v8.0.2 was used for all data analysis and plotting. Concentration–response curves were fitted with a four-parameter Hill equation with variable Hill coefficient. Data are presented as mean ± SEM.