Transforming Conotoxins Into Cyclotides: Backbone Cyclization of P-Superfamily Conotoxins Muharrem Akcan,1* Richard J. Clark,1† Norelle L. Daly,1‡ Anne C. Conibear,1 Andrew de Faoite,2 Mari D. Heghinian,3 Talwar Sahil,4 David J. Adams,2 Frank Marı,3 David J. Craik1 1

Institute for Molecular Bioscience, The University of Queensland, Brisbane QLD 4072, Australia

2

Health Innovations Research Institute, RMIT University, Bundoora VIC 3083, Australia

3

Department of Chemistry and Biochemistry, Florida Atlantic University, FL 33431, USA

4

Queensland Brain Institute, The University of Queensland, Brisbane, QLD 4072, Australia

Received 9 March 2015; revised 17 June 2015; accepted 4 July 2015 Published online 14 July 2015 in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/bip.22699

ABSTRACT: Peptide backbone cyclization is a widely used approach to improve the activity and stability of small peptides but until recently it had not been applied to peptides with multiple disulfide bonds. Conotoxins are disulfide-rich conopeptides derived from the venoms of cone snails that have applications in drug design and development. However, because of their peptidic nature, they can suffer from poor bioavailability and poor stability in vivo. In

disulfide-bonded peptides with cystine knot motifs. Cyclic gm9a was more potent at high voltage-activated (HVA) calcium channels than its acyclic counterpart, highlighting the value of this approach in developing active and stable conotoxins containing cyclic cystine knot motifs. C 2015 Wiley Periodicals, Inc. Biopolymers (Pept Sci) V

104: 682–692, 2015. Keywords: conotoxins; cyclic peptide; cyclization; cystine knot; drug design

this study two P-superfamily conotoxins, gm9a and bru9a, were backbone cyclized by joining the N- and Ctermini with short peptide linkers using intramolecular native chemical ligation chemistry. The cyclized derivatives had conformations similar to the native peptides

This article was originally published online as an accepted preprint. The “Published Online” date corresponds to the preprint version. You can request a copy of any preprints from the past two calendar years by emailing the Biopolymers editorial office at [email protected].

showing that backbone cyclization can be applied to three *Present affiliation: Dumlupınar University, Faculty of Arts and Sciences, Department of Biochemistry, 43100, K€ utahya, Turkey † Present affiliation: School of Biomedical Sciences, The University of Queensland, Brisbane, QLD 4072, Australia ‡ Present affiliation: Centre for Biodiscovery and Molecular Development of Therapeutics, James Cook University, Cairns, QLD 4878, Australia Correspondence to: David J. Craik, Institute for Molecular Bioscience, The University of Queensland, Brisbane 4072; e-mail: [email protected] Contract grant sponsor: Australian Research Council Contract grant number: DP1093115 Contract grant sponsor: National Health and Medical Research Council Contract grant number: APP1076136 Contract grant sponsor: NHMRC Contract grant number: APP1026501 Contract grant sponsor: ARC Future Fellow, ARC Australian Professorial Fellow C 2015 Wiley Periodicals, Inc. V

682

INTRODUCTION

T

he use of peptides in drug design and development has become increasingly prominent in recent years.1–3 One class of peptides that has attracted particular interest are conotoxins, which are disulfiderich peptides of 10–40 amino acids isolated from the venoms of marine snails of the genus Conus.4,5 More than 200 conotoxins have been characterized and their selective targeting of mammalian ion channels and transporters, as well as membrane receptors, make them attractive drug candidates for neurological diseases.6,7 For example, x-conotoxin

PeptideScience Volume 104 / Number 6

Transforming Conotoxins Into Cyclotides

MVIIA from Conus magus is approved in the US and Europe with the generic name ziconotide for the treatment of chronic pain.8 However, despite this success, a general susceptibility to proteolysis, poor bioavailability and short in vivo half-lives of peptides potentially limits their therapeutic applications. To overcome these challenges, several approaches, including truncation of the N- and C- termini,9–11 replacement of cysteine residues with selenocysteine,12,13 stereochemical modification of residues,14–16 addition of lipids to side chains,17–19 N- and C- terminal capping,20 nonpeptidic spacers,21 stabilization with dicarba bridges,22–25 and backbone cyclization of the peptide framework, have been developed.26–36 The idea of using backbone cyclization of peptides to enhance their stability was inspired by cyclic peptides already present in nature.37 For example, the antimicrobial peptides bacteriocin AS-48 and rhesus theta defensin (RTD-1) from bacteria and monkeys, respectively, have higher stabilities than their linear counterparts.38,39 Cyclotides are the largest known family of cyclic peptides found in nature.40 These plantderived peptides have a unique head-to-tail cyclized peptide backbone with a knotted arrangement of three conserved disulfide bonds.37,41,42 The Rubiaceae, Violaceae, and Cucurbitaceae plant families were the first known sources of these peptides43 but novel cyclotide sequences from the Fabaceae have also been reported recently.44,45 More than 300 cyclotide sequences have been described so far and it is predicted that there may be tens of thousands of different cyclotides.46–48 In addition to the cyclic backbone, cyclotides have three disulfide bonds that form a cystine knot, a motif comprising an embedded ring formed by two disulfide bonds (CysI-CysIV and CysIICysV) and their connecting backbone segments that is threaded by a third disulfide bond (CysIII-CysVI). The cystine knot and cyclic backbone together form the cyclic cystine knot (CCK) motif, which is highly resistant to thermal, enzymatic or chemical degradation.40,49 The CCK motif is exclusively found in plant cyclotides, but the cystine knot is also present in peptides from fungi, cone snail, spider and scorpion venom peptides, where it is referred to as the inhibitor cystine knot (ICK). Conotoxins with an ICK motif are generally from the O1- and P-superfamily conotoxins such as MVIIA and gm9a, respectively. Cyclization of these conotoxins with an ICK motif will produce a CCK framework. Cyclization has so far been applied only to conotoxins from the a-, v- and x-families, including a-MII,33 v-MrIA,35,50 aVc1.1, a-ImI,51 a-RgIA,34 a-AuIB,36,52 and x-MVIIA53 and in most cases produced favorable results. For example, cyclization of MII increased its serum stability by about 20%, and cyclic Vc1.1 was not only more stable, but also more potent at one of its molecular targets, the GABAB receptor, than native (acyclic) Biopolymers (Peptide Science)

683

FIGURE 1 Cystine knot motif of cyclotides and P-superfamily conotoxins. A: Three-dimensional structures of conotoxin gm9a, the cystine knot motif and cyclotide kalata B1. One disulfide bond is shown in green ball and stick format; it threads the embedded ring formed by the other two disulfide bonds, shown in orange ball and stick format. B: Amino acid sequences of acyclic and cyclic P-superfamily conotoxins gm9a and bru9a, and cyclotides kalata B1 and kalata B15. Two disulfide bonds are shown as orange lines and the third is shown as a green line. “*” at the end of gm9a sequence indicates an amidated C-terminus and “O” in the bru9a and cyclic bru9a-GLP sequences indicates a hydroxyproline residue.

Vc1.1 and showed oral activity in a rat model of neuropathic pain.26 P-superfamily conotoxins have a cystine pattern (C-C-CCXaaC-C) identical to cyclotides and among all conotoxins would be the closest match to cyclotides if they were cyclized (Figure 1). The conotoxin tx9a from a mollusk-hunting cone snail, C. textile, was the first member of this superfamily to be described. It has two post-translationally modified ccarboxyglutamate (Gla) residues at positions 8 and 13, and causes spasmodic contractions when injected into mice.54 Another conotoxin belonging to the P-superfamily is gm9a, isolated from C. gloriamaris, has a Ser residue at the N-terminus instead of a Gly, and Ser and Ala residues instead of the Gla found in Tx9a.55 The ICK motif of gm9a, containing the three disulfide bonds, CysI-CysIV, CysII-CysV and CysIII-CysVI, is illustrated in Figure 1. Another member of the P-superfamily, bru9a, isolated from C. brunneus, comprises 24 residues, including a post-translationally modified hydroxyproline residue. In this study, we investigated the backbone cyclization of gm9a and bru9a. As shown in Figure 1, their sequence and structural similarity with cyclotides make them excellent candidates for backbone cyclization. Three residues, Gly-Leu-Pro were incorporated as a linker to span the distance between the

684

Akcan et al.

N- and C-termini of each peptide. It was postulated that these residues might facilitate cyclization because of their presence in the cyclization loop of cyclotides (Figure 1).

tral width of 12 ppm. Structures were calculated using a torsion angle simulated annealing protocol within the program CYANA.58 Structures were analysed using PROMOTIF and PROCHECK.59,60

Serum Stability Assays

MATERIALS AND METHODS Solid Phase Peptide Synthesis Manual solid-phase peptide synthesis (SPPS) was used to synthesize the cyclic peptides with standard protecting groups [e.g. Asn(Xan), Asp (OcHex), Arg(TOS), Cys(MeBzl), Lys(ClZ), Ser(Bzl), Thr(Bzl) and Tyr(BrZ)]. The peptides were assembled onto phenylacetamidomethylglycine (PAM-Gly) resin using an in situ neutralization/HBTU [2-(1Hbenzotriazol-1-yl)-1,1,3,3-tetramethyluronium hexafluoro phosphate] protocol for Boc (tert-butoxycarbonyl) chemistry via a thioester linker.56,57 Firstly, the resin was soaked in DMF overnight to swell before the synthesis. Before coupling the thioester linker, the Boc group on the preloaded resin was removed by TFA. After washing the resin with DMF, S-trityl-b-mercaptopropionoic acid in HBTU/DIPEA was double coupled for 30 min to generate the thioester linker. The trityl group on the linker was removed with 2% triisopropylsilane/2% H2O/96% TFA and the resin was washed with DMF. A Gly-Leu-Pro sequence was used to join the N- and C-termini of the peptides. Peptide gm9a was also cyclized without any linker residues. Cleavage of the peptides from the resin was achieved using hydrogen fluoride (HF) with p-cresol and p-thiocresol as scavengers [9:0.8:0.2 (vol/vol) HF:p-cresol:p-thiocresol] at –5 to 08C for 1.5 h. Crude peptides were purified by reversed phase-high performance liquid chromatography (RP-HPLC) on a Phenomenex C18 column using a gradient of 0–80% solvent B (solvent A: H2O/0.05% TFA; solvent B: 90% CH3CN/10% H2O/0.045% TFA) in 80 min, with the eluant monitored at 215 and 280 nm. Electro spray ionization mass spectrometry (ESI-MS) confirmed the molecular mass of the fractions collected, and those displaying the correct molecular mass of the thioester peptide were pooled and lyophilized for oxidation.

Cyclization and Oxidation An intramolecular native chemical ligation reaction was used for cyclization of the three peptides. The peptides gm9a and bru9a were cyclized and oxidized with stirring in an aqueous buffer solution consisting of 0.1M NH4HCO3/DMSO (50%/50%) (v/v) at pH 8 at room temperature overnight. gm9a without linker residues was cyclized and oxidized in a buffer solution of 0.1M Tris-HCl/0.1 mM EDTA (pH 8.7) at room temperature with stirring overnight, as reported previously for the linear peptide.55 RP-HPLC was used to purify the cyclic peptides, and their purity and molecular weights were confirmed by analytical RP-HPLC and ESI-MS, respectively.

NMR Spectroscopy NMR spectra were recorded on Bruker Avance 600 MHz or 900 MHz NMR spectrometers. The peptides were dissolved in 90% H2O and 10 % D2O and the one-dimensional (1H), total correlation spectroscopy (TOCSY) and nuclear Overhauser effect spectroscopy (NOESY) spectra were recorded at 298 K. For TOCSY spectra an 80 ms mixing time, and for NOESY spectra a 250 ms mixing time were used. Twodimensional spectra were generally collected over 4096 data points in the f2 dimension, and 512 increments in the f1 dimension over a spec-

Serum stability assays were carried out in 100% human male serum (Sigma) using a peptide concentration of 20 lM. The serum was centrifuged at 14,000g for 10 min to remove the lipid component, then the supernatant was incubated at 378C for 15 min before the assay. Each peptide was incubated in serum at 378C and 40 lL triplicate aliquots were removed at 0, 1, 2, 3, 6, 10, 16 and 24 h. Each serum aliquot was quenched with 40 lL of 6M urea and incubated for 10 min at 48C. Then, each serum aliquot was quenched with 40 lL of 20% trichloroacetic acid and incubated for another 10 min at 48C to precipitate serum proteins. The samples were centrifuged at 14,000g for 10 min, and 100 lL of the supernatant were analyzed using analytical RP-HPLC using a gradient of 5–80% solvent B (0.3 mL/min flow rate). The control samples contained equivalent amount of peptides in phosphate-buffered saline (Sigma, pH 7.4) subjected to the same treatment procedure. The percentage recovery of peptides was detected by integration at k 5 215 nm.

Dissociated Dorsal Root Ganglion (DRG) Neurons and Patch Clamp Recording DRG neurons were enzymatically dissociated from ganglia of 4- to 16day-old Wistar rats as described previously61–63 and used for experiments within 24 h. Rats were killed by cervical dislocation as approved by the RMIT University Animal Ethics Committee. The spinal column was hemi-segmented and the spinal cord was removed. Ganglia were removed and rinsed in cold Hank’s Balanced Salt Solution (HBSS) (Life Technologies, Invitrogen, Australia). They were minced and incubated in 1 mg/mL collagenase (type 2; 405 U/mg) (Worthington Biochemical Corp., Lakewood, NJ) in HBSS at 378C for 30 min. After incubation, ganglia were rinsed three times with warm (378C) Dulbecco’s Modified Eagle Media (Invitrogen) supplemented with 10% fetal calf serum and 1% penicillin/streptomycin, then they were gently triturated with a fire-polished Pasteur pipette. Cells were plated on glass coverslips and incubated at 378C in 95% O2/5% CO2. Cells were transferred into a small volume (200 mL) recording chamber constantly perfused with a solution containing (in mM):150 tetraethylammonium chloride (TEA-Cl), 2 CaCl2, 10 D-glucose, 10 HEPES, pH 7.4 (with NaOH). Borosilicate glass electrodes were filled with an internal solution containing (in mM): 140 CsCl, 1 MgCl2, 5 MgATP, 0.1 Li-GTP, 5 BAPTA-Cs4, 10 HEPES, pH 7.3 (with CsOH), and had resistances of 1.5–2.5 MX. Whole-cell patch clamp recordings of depolarization-activated calcium currents were performed with a Multiclamp 700B amplifier controlled by Clampex9.2/DigiData1332 acquisition system (Molecular Devices), at room temperature (23– 258C). Isolated DRG neurons were voltage-clamped at a holding potential of –70 mV and membrane currents were filtered at 2 kHz and sampled at 5 kHz. Leak and capacitative currents were subtracted using a –P/4 pulse protocol. Peptides were diluted to the final concentration (1 lM) immediately before the experiment and were applied via perfusion in the bath solution. Data are presented as mean 6 SEM (n  3). Statistical analyses were performed using one-way ANOVA; differences were considered significant if P < 0.05.

Biopolymers (Peptide Science)

Transforming Conotoxins Into Cyclotides

685

FIGURE 2 Comparison of the secondary aH chemical shifts of gm9a and bru9a. A: Comparision of the secondary aH chemical shifts of native, cyclic gm9a-GLP and cyclic gm9a without linker residues shows that the cyclic version is correctly folded like native peptide. B: Comparision of the secondary aH chemical shifts proves that cyclic bru9a-GLP is correctly folded like the native peptide.

Glycine Receptor Assays Stably expressing a1 GlyR-HEK293 cells were cultured in medium supplemented with 0.5 mg/mL G-418 and 10 lM strychnine in a T75 flask. After 2–3 days incubation, cells were trypsinized, pelleted by centrifugation and then resuspended in extracellular/external recording solution at a final density of 1 3 106 – 5 3 107 cells/mL. The extracellular solution contained (in mM) 140 NaCl, 4 KCl, 2 CaCl2, 1 MgCl2, 10 HEPES/NaOH, and 5 glucose (pH 7.4). Intracellular/ internal solution consisted (in mM) of 50 KCl, 10 NaCl, 60 KF, 2 MgCl2, 10 HEPES/ KOH (pH 7.2), and 20 EGTA. Seal enhancer contained (in mM) 80 NaCl, 3 KCl, 35 CaCl2, 10 MgCl2, 10 HEPES/ NaOH (pH 7.4). The cell suspension was transferred to the cell hotel (Cell reservoir on Patchliner station, Nanion Technologies GmbH, Munich, Germany) immediately before experiments. Intracellular solution, external solution, seal enhancer and compound (drug) solutions was placed on the allocated workstation. The 16-channel chip was placed on the chip wagon. Briefly, the protocol comprises the following steps—the robotic pipette first applies internal solution into the chip, and then external solution, the chip was docked, and the machine then formed electrical contact between the internal and external solutions. Cell suspension was applied to the external side of the chip. Once a cell was caught in each channel (i.e. sucked onto the aperture), seal enhancer solution was applied into the chip. Finally, Patch Con-

Biopolymers (Peptide Science)

trol HT performed several seal formation and optimization steps before it established the whole-cell clamp. Standard patch-clamp techniques were used to record from the whole cell configuration using an EPC-10 amplifier (HEKA Elektronik). All experiments were performed at room temperature. Conotoxin gm9a concentrations (0.3, 1, 3, 10, 30, and 100 lM) were prepared with EC20 glycine concentration (20 lM) from the stocks. Results are expressed as mean 6 SEM of five independent experiments. Glycine EC20 current was considered as baseline current (0%). Response to conotoxin is plotted as % potentiation of EC20 current.

Functional Characterization in the Drosophila Melanogaster GFS Linear and cyclic gm9a were tested on the Drosophila GFS using the paired electrophysiology/nanoinjection bioassays as described previously.64,65 P[Gaw- B]OK307 (Stock #6488; Bloomington Stock Center, Bloomington, IN; referred to as A307 henceforth) fly stocks were kept at either 228C or 258C in vials containing standard media. All peptides were resuspended in 0.7% saline for injection and tested at 2 lg/mg for activity (n 5 10). Control flies (n 5 10) were injected with a 0.7% saline solution. The GF-TTM and GF-DLM pathways were evaluated for changes in the following frequency (FF), which is the total number

686

Akcan et al.

FIGURE 3 Three dimensional structures of cyclic gm9a-GLP and bru9a-GLP: A superimposition of the 15 lowest energy structures of cgm9a (A) and cbru9a (C) are shown on the left with the corresponding ribbon representations (B) and (D) on the right. Structures were determined using NMR spectroscopy. The six cysteine residues are numbered with Roman numerals, the disulfide bonds are shown in orange ball and stick format and the b-strands shown as arrows.

of responses recorded for each pathway when stimulated with 10 trains of 10 stimuli given at 100 Hz with a 1-s interval between the trains. This evaluation was performed before and 1, 5, 10 and 15 min after treatment. Statistical analysis was performed with GraphPad Prism 5.04 software (GraphPad Software Inc, La Jolla, CA). All groups underwent a two-way repeated measure ANOVA followed by a Bonferroni test.

RESULTS The designed cyclic peptides were synthesized using solid phase methods based on Boc chemistry.57 To preserve the peptide scaffold, a three-residue linker, GLP, was used to span the distance between the N- and C-termini of the native peptides. It was anticipated that the GLP linker might facilitate cyclization because of its presence in the cyclization loop of cyclotides. In addition, attempts were made to join the N- and C- termini of gm9a without any linker residues. An intramolecular native chemical ligation strategy56,66–68 was used to join the two ends of the peptides, resulting in backbone cyclization.69 The linear form of the P-superfamily conotoxin gm9a (i.e. without linker residues) was oxidized in a

buffer solution containing 0.1M Tris-HCl/0.1 mM EDTA, pH 8.7 at room temperature overnight.55 Additionally, both gm9a and bru9a with GLP linker residues were cyclized and oxidized in a buffer solution consisting of 0.1M NH4HCO3/DMSO (50/ 50) (v/v) at pH 8 at room temperature overnight. RP-HPLC and ESI-MS analysis revealed two isomers for cgm9a-GLP (45% correctly folded peptide) and seven isomers for cbru9aGLP (25% correctly folded peptide) with the same peptide mass but different retention times were obtained. Each isomer was lyophilized separately for NMR analysis to determine which isomers were correctly folded. Peptides were dissolved in 90% H2O and 10% D2O, and one dimensional TOCSY and NOESY spectra were recorded at 298 K. The NMR spectra were assigned using well-established techniques.70 Secondary aH chemical shifts of native gm9a, cyclic gm9a, and cyclic gm9a with GLP linker residues are shown in Figure 2A. Cyclization of gm9a without linker residues (cgm9a) resulted in a misfolded peptide as its secondary aH chemical shifts do not match to native gm9a. In contrast, correctly folded cyclic gm9a (cgm9a-GLP) was obtained by using three linker residues between the N- and C-termini of Biopolymers (Peptide Science)

Transforming Conotoxins Into Cyclotides Table I NMR Structural Statistics for cgm9a-GLP and cbru9a-GLP

Target function average Distance constraints Total number of NOEs Sequential (ji-jj 5 1) Medium range (ji-jj < 5) Long range (ji-jj  5) RMSD (15 structures), residues 1–27 Average backbone RMSD to mean Average heavy atom RMSD to mean Ramachandran statistics Residues in most favored regions Residues in additionally allowed regions Residues in generously allowed regions Residues in disallowed regions

cgm9a-GLP

cbru9a-GLP

0.03 6 0.00992

0.0313 6 0.00612

574 436 53 85

500 390 39 71

0.16 6 0.02 A˚

0.46 6 0.18 A˚

0.64 6 0.13 A˚

0.89 6 0.19 A˚

52.0

73.7

48.0

26.3

0.0

0.0

0.0

0.0

the peptide, based on the similarity of secondary shifts between the native and cyclic peptide. The fingerprint region in the NOESY spectrum of each peptide shows a complete cycle of aH–NH sequential connectivities with the exception of the Pro30 residue, which lacks an amide proton. However, consistent with a trans geometry of the Pro residue in loop 6 of cyclotides, NOEs were observed between Ha of Leu29 and the d protons of Pro30 of cgm9a-GLP. As illustrated in Figure 2B, cyclization and folding of bru9a was achieved. The amide peaks of cyclic bru9a are well dispersed, similar to the synthetic linear version and the secondary aH chemical shifts of the cyclic version overlay well with the native version indicating that the peptide is correctly folded. The fingerprint region in the NOESY spectrum of the bru9a peptides shows a complete cycle of aH–NH sequential connectivities, with the expected exception of the hydroxyproline (Hyp12) and Pro27 residues since they lack amide protons. However, NOEs were observed from the d protons of the proline residues and their preceding residue, which completed the backbone cycle of the peptide. The three-dimensional NMR structures of cyclic gm9a-GLP and cyclic bru9a-GLP (PDB ID: 2MSQ) were calculated with the program Cyana 3.058 and the 15 lowest energy structures were overlayed using line representations as shown in Figure 3.71 The structural statistics given in Table I indicate that the structures are well-defined. The major element of secondary Biopolymers (Peptide Science)

687

structure in both molecules is a small b-sheet or b-hairpin. Additionally, cyclic gm9a-GLP contains a helical turn in loop 2. Both structures comprise a CCK topology. Conotoxins bru9a and gm9a, and their respective cyclic analogues, were tested for their ability to inhibit high voltageactivated (HVA) calcium currents in rat DRG neurons (Figure 4). All peptides were tested at 1 mM and added to isolated DRG neurons for 400 s each, after a steady-state control current amplitude was achieved in response to depolarizing voltage steps to 0 mV from a holding potential of –70 mV. Neither linear nor cyclic bru9a (1 mM) had any inhibitory effect on HVA calcium currents in DRG neurons (n 5 4; P 5 0.94), whereas linear and cyclic gm9a exhibited small but not significant inhibition of calcium channel currents. In the presence of cgm9a (1 mM), depolarization-activated calcium current amplitude was reduced by 25.8 6 10.2% (n 5 3) and by 16.0 6 7.2% (n 5 3) in the presence of linear gm9a. However, the inhibition observed was not statistically significant (P  0.62) and the incomplete recovery of current amplitude upon washout of the peptide may reflect in part current rundown. It has been proposed that gm9a or tx9a might target glycine receptors, because their effect on mice is consistent with the spasmodic phenotype of mouse strains deficient in glycine receptors.54,55 Therefore gm9a and cgm9a-GLP were screened for their modulatory activity on a1 recombinant homomeric glycine receptors. Coapplication gm9a (linear and cyclized) with EC20 glycine (20 lM) did not potentiate or inhibit glycine-induced currents (Figures 4F and 4G). Additionally, higher concentration of peptide (

Transforming conotoxins into cyclotides: Backbone cyclization of P-superfamily conotoxins.

Peptide backbone cyclization is a widely used approach to improve the activity and stability of small peptides but until recently it had not been appl...
636KB Sizes 0 Downloads 9 Views