Probing pattern and dynamics of disulfide bridges using synthesis and NMR of an ion channel blocker peptide toxin with multiple diselenide bonds

A biologically active peptide toxin containing four diselenide bonds was synthesized. The diselenide network and its dynamics were disclosed using a combined NMR and MD approach.


Introduction
The disulde pattern of polypeptides containing multiple disulde bridges is of considerable interest because the presence of disulde bonds is essential for maintaining the tertiary structure responsible for the observed biological activity. Therefore, a key issue in the structural characterization of these peptides is an unambiguous determination of the pattern of disulde pairing. Since small, exible disulde-rich peptides are typically difficult to crystallize, therefore nuclear magnetic resonance (NMR) spectroscopy and mass spectrometry (MS) are the most commonly used techniques for elucidating their structures. 1 MS determination of a disulde bonding pattern is a challenging task primarily relying on manual interpretation. 2,3 It requires either enzymatic digestion of the protein where disul-de scrambling may take place, or chemical modication if the target protein is cysteine rich, or else contains an unexpected folding pattern with unknown disulde bonds.
NMR provides powerful tools for determining the threedimensional structures of these small proteins and peptides, but, due to the unfavorable NMR properties of sulfur isotopes, disulde bonds oen present a 'blind spot' in NMR structural investigations. 1 Experimental observations supported by DFT quantum-chemical calculations proved that the redox state of cysteine residues can be safely disclosed on the basis of Cys 13 C b chemical shis. 4 These data, however, do not provide information about the disulde bridge connectivity. Heretofore the most commonly used approach for determining S-S connectivity relied on the detection of interresidue dipolar (NOE) interactions between the b-methylene protons of the covalently linked cysteine residues. 5 Detection of these weak inter-disul-de NOE contacts is oen hampered by spin-diffusion effects due to the strong intraresidue cross-relaxation between the geminal beta protons of Cys and/or in some cases by an enhanced conformational exibility around the disulde bonds, leading to an ambiguous assignment of the disulde network. However, such uncertainties could be resolved by stereospecic deuterium labeling of the cysteine beta protons at the cost of laborious isotope labeling. Besides, in the case of closely packed disulde bonds, the NOE contacts can be also ambiguous and lead to incorrect pairing of Cys residues. Alternatively, as has been recently suggested, replacement of the NMR-inactive 32 S nucleus by 77 Se possessing more favorable magnetic properties 6,7by mutation of cysteine into selenocysteine (Sec) residues 8-13may allow the detection of diselenide pairing directly through the three-bond 3 J( 1 H, 77 Se) couplings across the diselenide bond. 5 Importantly, it was also conrmed that the isosteric replacement of the disulde bond with diselenide has no signicant effect, in general, on the three-dimensional structure and the associated biological activity of peptides. 14 In addition, using 77 Se NMR both the conformational and dynamic properties of diselenide (disulde) bonds can be explored.
The Sec residues can be introduced into peptides or proteins either by chemical synthesis 11,12,[14][15][16] or by recombinant expression. 7,9,10,17 Using chemical synthesis, substitution of only one or two disulde bridges has been reported in most cases whereas multiple disulde bonds can be replaced with diselenides by applying recombinant expression. Recently, an increasing interest also has emerged for the synthesis of selenium-containing polymers due to their potential applicability as biomaterials. 18,19 Biological background Peptide toxins form an important class of disulde bond containing peptides and proteins. One particular group of peptide toxins isolated from scorpions and other venomous animals targets Kv1.3, the voltage-gated ion channel of human T lymphocytes. 20 The Kv1.3 channel plays a crucial role in antigeninduced activation and proliferation of human T cells in general, and in those of effector memory T cells in particular. 21 These latter cells are responsible for tissue damage in certain autoimmune diseases such as multiple sclerosis, type I diabetes, and rheumatoid arthritis. 22 High affinity and selectivity Kv1.3 inhibitors persistently inhibit the proliferation of effector memory T cells thereby ameliorating the symptoms of the disease in animal models. This indicates the great therapeutic value of the Kv1.3 inhibitors. [22][23][24] Anuroctoxin (AnTx), a 35-amino-acid scorpion toxin having four disulde bridges, is a high affinity blocker of Kv1.3 channel, but also blocks Kv1.2. 25 Improving selectivity of the toxin for Kv1.3 over Kv1.2 while keeping the high affinity for Kv1.3 opens the door to the toxin's therapeutic use. In line with these, we have constructed a double substituted analog of AnTx, [N17A/F32T]-AnTx, which is characterized by comparable Kv1.3 affinity to the wild-type peptide but shows a 2500-fold increase in the selectivity for Kv1.3 over Kv1.2 and other ion channels tested (Kv1.1 and KCa3.1). 26 Thus, [N17A/F32T]-AnTx may serve as an excellent template for a novel therapeutically applicable peptide, and, interestingly, understanding the structural features of this peptide may lead to a better understanding of the nature of the selectivity of scorpion toxins between Kv1.3 and Kv1.2.
In the present work we report the chemical synthesis of a Secanalog of [N17A/F32T]-AnTx during whichfor the rst time to our knowledgethe replacement of all four disulde bonds with isosteric diselenide bonds has been achieved. Using a combined NOE-and 77 Se NMR-based approach unequivocal assignment of the diselenide (disulde)-bond connectivities was also accomplished. Moreover, the signicant broadening of some of the 77 Se resonances revealed the presence of slow conformational exchange processes which have been subsequently interpreted by molecular dynamics (MD) simulations.

Synthesis of the Sec-analog of [N17A/F32T]-AnTx mutant
For the preparation of the anuroctoxin derived peptides, both the Boc and Fmoc chemistry were used previously with success. 26 Since the stability of several Boc compatible Se protecting groups is signicantly lower than the corresponding sulfur protecting groups (e.g., BnSe À is a much better leaving group than the corresponding sulfur derivative 12,27 ), for the solid-phase synthesis of selenocysteine analog of [N17A/F32T]-AnTx mutant, the Fmoc/ t Bu strategy was applied. Correspondingly, the related N a -Fmoc derivative of the 4-methoxy-benzyl protected selenocysteine 8 was synthesized and used as replacement for all eight cysteines in the original sequence. For the linear Sec(Mob) protected [N17A/F32T]-AnTx analog, a standard synthetic protocol was applied with DCC/HOBt (1 : 1) as coupling agent in a 3-fold molar excess. The removal of the Mob group from the selenocysteine residues can be carried out, according to previous studies, under various conditions, e.g., acidic treatment, I 2 , or applying a sub-stoichiometric amount of 2,2 0 -dithiobis(5-nitropyridine). With regard to the complexity of the amino acid sequence of the peptide, we decided to achieve the cleavage from the resin together with the protecting groups commonly used in Fmoc/ t Bu synthesis and removal of the Mob group from the selenocysteine side-chains in a two-step protocol. Acidolytic cleavage of the Sec(Mob) peptide from the resin was performed under various conditions (TFA/H 2 O/TIS cocktail in 93/5/2, v/v, at low temperature for 2 h; TFA/H 2 O/TIS cocktail in 93/5/2, v/v at room temperature for 2.5 h) and in both cases the crude product obtained consisted mainly of the compound with all eight Mob groups intact besides compounds which had partially removed one, two, or even three Mob groups. In order to avoid incomplete removal of the side-chain protecting groups, and considering the fact that it follows Mob group de-protection, we decided to carry on with this complex reaction mixture. For Mob de-protection/cyclization we tried procedures well established in cysteine chemistry, 27,28 like the DMSO/TFA (v/v) procedure (10% DMSO/TFA and 4% DMSO/TFA) and the I 2 -mediated method (added quantitatively or in excess), but in all cases we obtained a very complex mixture of compounds from which the desired diselenide-bridged peptide could not be isolated. Cleavage of the p-methoxybenzyl group and intramolecular oxidation to the diselenide peptide was successfully achieved with the more gentle 2,2 0 -dithiobis(5-nitropyridine) (DTNP) in TFA. Incubation of the protected Sec-containing peptide in TFA with DTNP resulted in effective protecting group removal and its substitution by the 2-(5-nitropyridyl) (Npys) group derived from DTNP fragmentation, independent of the number of DTNP equivalents and peptide concentration used. Aer test reactions, a nal 4 mM peptide concentration in TFA was used with 1 equiv. DTNP incubation for 1 h. The reaction was quenched via precipitation into cold diethyl ether and, aer isolation of the precipitate, the crude material was treated with one equivalent of L-cysteine in pH 8 NH 4 OAc buffer solution in order to induce collapse into the desired multiple diselenide-peptide. HPLC analysis of the resulting solution showed an almost instantaneous transformation of the hemi-reduced intermediate into the multiple diselenide compound (Scheme 1).
The HPLC trace of the folded, puried tetra diselenide toxin analog is shown on Fig. S1. † The spectra of mass spectrometric analyses of the multiple selenocysteine peptide recorded with quadrupole ESI and Q-TOF spectrometers are depicted in Fig. S2-S5. † The characteristic isotopic selenium abundance (Fig. S5 †) is more complicated than the sulfur containing ones, but a careful study led to the justication of the correct molecular ion (Fig. S4 †). Subsequent NMR studies and biological measurements of the isolated main product proved that all four diselenide bridges are completely analogous with the disulde connectivities from the native [N17A/F32T]-AnTx peptide.
The Sec-analog of [N17A/F32T]-AnTx retains its biological activity Prior to obtaining detailed structural information one key question remained to be answered, namely, whether the Secanalog of [N17A/F32T]-AnTx still blocks the Kv1.3 channel with high affinity. To this end, the Sec-analog was tested on Kv1.3 expressed in human peripheral blood lymphocytes. Fig. 1 shows that the peptide in 3 nM concentration blocks approx. 75% of the peak current (Fig. 1A), which is comparable to the inhibition of Kv1.3 with the double mutated peptide [N17A/F32T]-AnTx at identical concentration (89% inhibition in the latter case, 26 Fig. 1C). -AnTx) which argues that the peptide:channel stoichiometry of inhibition is 1 : 1, as predicted by currently accepted models of toxin binding to the channels. 29 In summary, the key pharmacological property of [N17A/ F32T]-AnTx, i.e., the high affinity block of the Kv1.3 current, is retained in Sec-[N17A/F32T]-AnTx so the peptide is functionally active.

NMR investigation of the Sec-analog of [N17A/F32T]-AnTx
Almost complete assignments of 1 H, 13   on two-dimensional (2D) homonuclear correlation spectra including 1 H-1 H COSY, TOCSY and NOESY. Whereas the wellresolved cross peaks of 1 H-13 C and 1 H-15 N HSQC spectra allowed the assignment of 13 C and 15 N resonances (see Table S1 in the ESI †), it should be mentioned that some of the H b -C b cross peaks, such as of Sec4, Sec10 and Sec24 residues, are signicantly broadened and have, therefore, lower intensities in the 1 H-13 C HSQC correlation map (Fig. 2); this is indicative of slow conformational dynamics occurring in the vicinity of those residues. Note that similar exchange mediated line-broadening was observed in the 1 H-15 N HSQC spectrum for T5, Q8, H9, U10, T11 and K30 residues (data not shown). It was also found that the Sec-derivative possesses almost identical 1 H, 13 C, and 15 N chemical shis as the S-containing peptide (selected regions of 1 H-13 C HSQC spectra of the two peptide analogs are shown in Fig. S6 †), suggesting high similarity of their three-dimensional structures.
The oxidized state of all eight Sec residues incorporated into the peptide could be unequivocally proved on the basis of the characteristic H b /C b and Se chemical shis of the Sec residues (Table S1 †). The MS data also supported the formation of four intramolecular diselenide bonds.
To establish the location of bridged residues, 1 H-1 H 2D NOESY spectra were recorded with 150 ms and 300 ms mixing. However, even using longer mixing time (300 ms), NOE cross peaks between the b-methylene protons of the covalently linked Sec residues could be hardly detected. This may be due partly to the strong intraresidue cross-relaxation between the geminal b-protons of Sec, partly to the line broadening caused by slow conformational dynamics around the Se-Se bridge, and also to the severe signal overlap in the region of interest. Specically, inspection of the NOESY spectrum shown in Fig. 3 reveals only one Se-Se bond unequivocally, which can be assigned to the Sec4-Sec24 bridge. Unambiguous assignment of the other inter-Sec(H b ) cross peaks is hampered by the strong signal overlap.
To take advantage of our 'Se NMR-spies' incorporated into the peptide sequence, multiple bond 1 H-77 Se correlation spectra were recorded using standard HMQC or HMBC experiments. Sites around disulde/diselenide bridges in peptides/proteins, however, are oen regions of conformational exchange. The exchange process, in general, can be detrimental to the quality of NMR spectra, resulting in broad and very weak, or even unobservable signals, and which may be detrimental to structure elucidation and, in particular, to the NMR-based characterization and assignment of disulde/diselenide bonds. To our disappointment, even aer several attempts using a different experimental setup (see ESI † for details and selected examples), only a few and exclusively intraresidue 1 H-77 Se correlations could be observed, probably due to conformational exchange line broadening.
To improve the sensitivity of 1 H-77 Se multiple bond correlation measurements, a modied, 77 Se-decoupled variant of the CPMG-HSQMBC experiment [30][31][32] has been developed in our group. This experiment considerably reduces the adverse line broadening effect of conformational exchange and also efficiently suppresses the undesired co-evolution of proton-proton couplings (such as the large geminal coupling between the H b protons of Sec) during the long scalar coupling evolution period matched to the intraand interresidue multiple bond 1 H-77 Se couplings. To minimize heating of the probe electronics and the sample during the long duration of CPMG pulse train, low-power composite 180 pulses were applied on both the proton and Se nuclei. Note that, similarly, CPMG pulse trains have been successfully employed in HSQC experiments to reduce the undesired effect of exchange processes and, as a consequence, for better observation of exchange broadened signals in nucleic acids and proteins. [33][34][35] The details of the experiment proposed here (scheme in Fig. S7 † and Bruker code of the pulse sequence are given in ESI †) and comparison of its performance to the classical HMQC/HMBC methods are demonstrated in the ESI (Fig. S8-S10 †). Recording several 1 H-77 Se CPMG-HSQMBC spectra with the sequence of   Fig. S7 † at different temperatures and using evolution times matched to different long-range couplings, we found that the largest number of connectivities could be detected at 318 K with the coupling evolution time set to 30-35 ms. The resulting spectrum shown in Fig. 4 allowed unambiguous assignment of all eight Se-resonances (as given on the projection at le), though some of the correlations are weak due to exchange broadening as noticed. Nevertheless, the cross peaks enlarged and shown in insets at lower intensity threshold could be assigned to provide direct evidence of two additional Se-Se pairings, between Sec14-Sec31 and Sec19-Sec34 residues, respectively.
Since the chemical shis (C b , Se) of Sec-s already conrmed that all eight Sec residues are involved in diselenide bonds, thus the remaining, unassigned Se-Se connectivity could be unambiguously assigned to the Sec10 and Sec29 residues.

Molecular dynamics (MD) simulation
MD calculations were found to be useful tools to analyze disulde bridge connectivities in small disulde-rich proteins. 36,37 In order to nd an explanation for the observed NMR behaviour of peptide sites around diselenide bridges, we carried out a 100 ns molecular dynamics (MD) simulation on the [N17A/F32T]-AnTx peptide. Since replacement of disulde bridges with diselenide bonds has been demonstrated to be bioisoteric due to high similarity between the conformational properties of sulfur and selenium, 11 the simulations were carried out on the disulde bridge containing peptide. The c 3 angle has taken up conformations around the canonical AE90 and has mostly persisted in the conformation presented in the starting structure shown in Fig. 5a. The dynamic behavior of the different S-S bridges in the trajectory could be classied into three groups according to the conformation along the side chain c 2 dihedral angles of the cysteine residues ( Fig. 5b and c

Conclusions
In summary, a Sec-analog of the biologically active [N17A/F32T]-AnTx peptide toxin containing four diselenide bonds has been produced with chemical synthesis. To the best of our knowledge, this is the rst time all disulde bonds in a protein were replaced by chemical synthesis with isosteric diselenides. We found that the Sec-derivative displayed similar voltage-gated potassium channel (Kv1.3) blocking potency and almost identical 1 H, 13 C and 15 N chemical shis as the S-containing peptide, suggesting high similarity of their three-dimensional structures including their Se-Se/S-S bridge pattern, respectively.
Using a combined NMR approach relying on characteristic chemical shis, 1 H-1 H NOEs and 1 H-77 Se multiple bond connectivities, the diselenide network of the Sec-analog was unequivocally disclosed. To improve the sensitivity of 1 H-77 Se correlation NMR experiment for peptides with multiple diselenide bonds and in the presence of exchange broadening, a modied CPMG-HSQMBC pulse sequence was designed and applied. The CPMG-cycles employed efficiently reduce the loss of spin coherence due to both chemical exchange and undesired evolution of proton-proton couplings.
Finally, NMR experiments complemented with MD simulations allowed us to explore and characterize the conformational dynamics of diselenide/disulde bonds.

Experimental
Electrophysiology Cells. Kv1.3 currents were measured on human peripheral lymphocytes obtained from healthy volunteers. Mononuclear cells were isolated using the Ficoll-Hypaque density gradient separation technique and cultured in 24-well culture plates at 37 C in a 5% CO 2 incubator in RPMI 1640 medium supplemented with 10% fetal calf serum (Sigma-Aldrich), 100 mg mL À1 penicillin, 100 mg mL À1 streptomycin, and 2 mM L-glutamine (density, 5 Â 10 5 cells per mL) for 2 to 5 days. To increase K + channel expression 5, 7.5, or 10 mg mL À1 phytohemagglutinin A (Sigma-Aldrich) was added to the medium.
Electrophysiology Kv1.3 currents were elicited by 15 ms long depolarization impulses to +50 mV from a holding potential of À100 mV every 15 s. To acquire and analyze the measured data a pClamp10 soware package was used. Current traces were lowpass-ltered by the analog four-pole Bessel lters of the ampliers. The effect of the toxin in a given concentration was determined as remaining current fraction (RF ¼ I/I 0 , where I 0 is the peak current in the absence of the toxin and I is the peak current at equilibrium block at a given toxin concentration). Points on the dose-response curves represent the mean of 5 independent measurements where the error bars represent the S.E.M. Data points were tted with a two-parameter Hill equation  10, 12, 15, 17, 20, and 30 Hz, typically in overnight experiments. 2D 1 H-1 H NOESY spectra using a Watergate sequence for water suppression were recorded with 150 and 300 ms mixing, respectively. TOCSY mixing was set to 80 ms. The 1 H, 13 C, 15 N, and 77 Se 90 hard pulses were of 11.5, 13.0, 29.0, and 15.0 ms duration, respectively.

Computational methods
All MD simulations were performed using AMBER version 12. 38 The AMBER ff99SB force eld 39 for the peptides and the TIP3P model 40 for water were used. The cut-off used for non-bonded interactions was 8Å. The particle-mesh Ewald 41 procedure was used to describe long-range electrostatic interactions with a maximal grid spacing of 1Å. Periodic boundary conditions were applied with a truncated octahedron geometry. The SHAKE 42 algorithm was used to keep the bond lengths of hydrogen atoms rigid allowing a timestep of 2 fs to be used. Since there is no selenium atom type implemented in the chosen force eld, we used sulfur in place of selenium. Comparison of the properties of sulfur and selenium shows that the structure and dynamics of the selenium-containing peptide should be very similar to the sulfur-containing one. This assumption was conrmed by comparison of the 1 H, 13 C, and 15 N chemical shis, which were strongly correlated.
Protocol. First minimization in 10 000 steps was performed to remove bad contacts switching from steepest descent to a conjugate gradient algorithm aer 100 steps. Aer minimization at a constant pressure MD was carried out to equilibrate the system for 0.5 ns, during which the density stabilized at around 0.9965 g cm À3 , and the temperature settled at ca. 300 K. Subsequently, a classical constant total energy MD producing micro canonical NVE ensemble was carried out for 100 ns.
All simulations were run in a single Nvidia GTX680 GPU using the GPU 43,44 implementation of the pmemd provided in the AMBER12 simulation package. Using the described protocols a 100 ns-long calculation of the peptide took about 25 h to complete.
Analysis. Translational centre of mass motions were removed every 1000 steps for NVE simulations. 50 000 coordinate snapshots were saved for analysis in all cases. The trajectories were analyzed with ptraj 45 and visualized in VMD. 46 The stability of the trajectories were analyzed using the mdout_analyser.py.