A peptide with six cysteines can be oxidised into fifteen different fully bridged isomers, every one of them the same mass and the same composition, and the oxidant has no preference among them. Two routes exist. Let the reduced chain find its own pairing under oxidising conditions, betting that the sequence encodes one dominant fold; or protect the cysteines in orthogonal pairs and close each bridge deliberately, paying in steps and in yield for the certainty. Choosing between them is a question about the peptide, not about taste, and it is expensive to get wrong in either direction: a six-step directed route on a sequence that folds cleanly wastes weeks, and a folding buffer on a sequence with no thermodynamic funnel hands you an isomer mixture you then have to separate and assign.
The two routes, and the property that decides between them
Decades of air oxidation of cysteine-rich peptides have largely confirmed Anfinsen's principle of self-assembly: a continuously growing number of cysteine-rich peptides, from diverse sources and with differing cysteine patterns, were found to retain the critical sequence-encoded structural information for correct oxidative folding into the native structures as dominant isomers [1]. The same review is equally explicit about the exception. There are numerous cases where folding and disulfide formation are not thermodynamically coupled, and where applying a defined succession of regioselective cysteine pairings still represents the method of choice [1]. That is the fork, and which side of it you are on is a property of the individual sequence rather than of the peptide class.
The chemistry on either side of the fork falls into three classes: oxidation starting from unprotected thiols, oxidation starting from protected thiols, and directed methods for forming unsymmetrical disulfides, any of which can be carried out in solution or on a polymeric support [2]. Overall yields and purities depend on the protecting-group combination chosen, the precise reaction conditions and the targeted structure, and no procedure can be guaranteed to give outstanding results for all cases [2]. Budget for screening rather than for transferring someone else's protocol intact.
When letting it fold is the right call
Disulfide-coupled refolding of five ω-conotoxins — peptides 23 to 26 residues long, each stabilised by three disulfides — was promoted by the disulfide form of glutathione in the presence of reduced glutathione. All five reached the native conformation, as judged by reversed-phase HPLC behaviour, with efficiencies of 16% for ω-MVIIC, 28% for ω-MVIID, and 50% for ω-MVIIA, ω-GVIA and ω-SVIA [3]. Although the primary structures show only limited sequence similarity, the mature forms of at least some of these peptides contain sufficient information to direct correct folding and disulfide formation [3].
For a peptide in that class the arithmetic favours folding: a single operation delivering half the material as the wanted isomer beats a multi-step orthogonal route whose stepwise yields multiply against you. The premise, though, is that the native pattern sits at the thermodynamic minimum your buffer can reach.
Three Ser→Cys mutants of human epidermal growth factor, a 53-residue chain carrying three disulfides, tested that premise directly by placing a competing cysteine at position 2, 4 or 9 alongside the native Cys6, each free to compete for Cys20. At equilibrium the native Cys6–Cys20 was favoured over Cys2–Cys20 and Cys4–Cys20 by 4.5-fold and 9-fold respectively; in the S9C mutant, a non-native Cys9–Cys20 was thermodynamically more stable than the native Cys6–Cys20 by a free-energy difference of 1.12 kcal/mol [4]. Moving a competing thiol by a few residues inverted which bridge the equilibrium prefers. Where the cysteine spacing in your sequence is unusual, the assumption behind random oxidation is worth testing on milligrams before it is scaled.
Concentration decides intramolecular against intermolecular
Oxidation cannot tell two thiols on one chain from two thiols on two chains. Only relative rates separate the two outcomes, and they scale differently with bulk peptide concentration: the intramolecular closure does not depend on it, the intermolecular one does.
Rattusin, an α-defensin-related peptide, is the limit case. Synthesised by solid-phase peptide synthesis and refolded by air oxidation in vitro, it gives a C2-symmetric disulfide-linked dimeric structure, and dimerisation of the refolded material resulted from five intermolecular disulfide bond exchanges formed by ten cysteines within two protomer chains [5]. Air oxidation did not fail there; it produced a defined, symmetric, entirely intermolecular product. Nothing in the oxidation step itself protects you against that outcome.
The standard countermeasure is high dilution, and its cleanest expression is oxidising the chain while it is still anchored. Polymer-supported oxidations take advantage of the pseudo-dilution phenomenon, which favours intramolecular processes; across the model peptides Ac-Cys-Pro-DVal-Cys-NH2 and Ac-Pen-Pro-DVal-Cys-NH2, which both readily assume a type II β-turn stabilised by a 14-membered disulfide-containing ring, and the conformationally mobile 20-membered ring of oxytocin, with Acm, Fm or Trt thiol blocking under both Boc and Fmoc Nα-protection, chains were released from the support without substantial damage to the disulfide bridges at overall yields as high as 60–90% [6].
The complementary trick keeps the peptide in solution and immobilises the oxidant instead. Ellman's reagent attached to a cross-linked ethoxylate acrylate resin — CLEAR-OX — was used to produce several model cyclic disulfides with improved purities and yields compared with solution oxidations, including a gram-scale oxidation of a peptide containing a hindered penicillamine unit; the problems it is built against are the formation of oligomeric by-products, other side reactions, and occasional solubility problems in aqueous oxidising media [7].
Dilution is not monotonically good, and this is the part usually left out of the protocol. In the solution synthesis of μ-conotoxin GIIIB, 22 residues with six cysteines, the hexa-SH peptide oxidised at a peptide concentration of 1 × 10−5 M gave three major products in a ratio of 1:4:3, and the native product was the minor component; analysis of the folding process at that concentration showed the disulfide between Cys10 and Cys15 forming first, leading to the predominant generation of the two mismatched products. The optimal conditions for the native isomer were obtained by increasing the peptide concentration in the oxidation reaction mixture or by using redox reagents, both of which functioned as promoters of the thiol-disulfide exchange reaction of the mismatched Cys10–Cys15 bond [8]. Dilution suppresses the intermolecular reaction you do not want and, in the same stroke, suppresses the concentration-dependent thiol–disulfide exchange that repairs an intramolecular mistake. That gap is exactly what a redox shuffling agent exists to fill.
Oxidants, and what each one costs
Air is the default for the random route and carries the whole self-assembly literature behind it [1], with the caveat the rattusin case makes concrete [5]. Its weakness is that almost nothing about it is under your control except pH, concentration and time.
Dimethyl sulfoxide is the reagent for substrates that will not tolerate an aqueous alkaline buffer at all. For a 37-residue peptide carrying a single disulfide between residues two and seven, very insoluble and aggregation-prone in vitro and particularly so at basic pH, formation of the intramolecular disulfide using DMSO was the most effective oxidation method, reducing the reaction time from 24 to 5 hours; using DMSO or 1,1,1,3,3,3-hexafluoro-2-propanol increased the solubility of the hydrophobic sequence, avoided the use of aqueous basic solutions and eliminated the need for continuous stirring during oxidation, which in turn allowed more consistent HPLC retention times [9].
A glutathione redox pair does two jobs at once: it supplies oxidising equivalents, and it maintains a thiol pool that lets a wrong pairing reopen. Both jobs respond to how much you add. Folding of all the cyclotides examined in one study — cyclic-backbone plant molecules stabilised by three disulfide bonds that form a cystine knot — was heavily influenced by the concentration of redox reagents, with the folding rate and final yield of the native isomer greatly enhanced by high concentrations of oxidised glutathione; addition of hydrophobic solvents to the buffer also enhanced the folding rates and appeared to alter the folding pathway [10]. The same work records what a slow reaction costs elsewhere: significant deamidation and isoaspartate formation were seen when oxidation conditions were conducive to slow folding [10]. A folding window measured in days is a window in which the backbone is degrading.
Iodine is the fastest of the common oxidants and the least selective, and in practice it belongs to the directed route, where it does double duty as an Acm-removal reagent. Its side reactions are dealt with below, alongside the rest of the deprotection chemistry.
Folding buffers: pH, chaotrope, cosolvent
Chaotropes are the most commonly misapplied handle on this reaction. Intuition says a denaturant should stop the chain trapping in a misfolded state. In at least one well-characterised case the measurement runs the other way: the folding efficiency of ω-MVIIA was greatly reduced by the addition of 8 M urea, which the authors read as indicating that formation of the correct disulfides is determined largely by noncovalent interactions, as opposed to steric constraints arising from the spacing between cysteine residues [3]. A chaotrope strong enough to dissolve an aggregating intermediate is also strong enough to erase the structure that selects the pairing. Reach for it when solubility is the binding constraint, and expect to pay for it in selectivity.
Organic cosolvents are the gentler version of the same lever. α-Conotoxin ImI is a 12-residue peptide that requires bridges between Cys2–Cys8 and Cys3–Cys12; direct oxidative folding resulted in high yields of correctly folded material where a one-pot oxidation of t-butyl/4-methylbenzyl-protected cysteine pairs did not, and the addition of organic cosolvents such as methanol, ethanol or isopropanol into the folding mixture significantly increased the accumulation of the native peptide, an effect also observed for α-PnIA [11]. Keep the authors' own hedge with the result: cosolvent-assisted direct oxidation might be of general use for other bicyclic α-conotoxins, but efficiency should be assessed on a case-by-case basis [11]. Note also which way that comparison ran — on this substrate the direct route beat the one-pot oxidation of protected cysteine pairs, the two protocols the authors explored as alternatives to the traditional two-step orthogonal route [11].
Directed closure: the deprotection order is the design
In a regioselective synthesis the connectivity comes from the order in which thiol protecting groups are removed, not from the fold. A three-way scheme using Mob, Trt and Acm protection enabled the efficient synthesis of peptides with the desired disulfide bond connectivities independent of their sequences, and delivered five conotoxins carrying three disulfides each at yields of 20–30% [12]. Sequence independence is the whole selling point, and the reason that yield is acceptable.
The scheme is only as orthogonal as its reagents. Treating an S-Acm peptide with iodine removes the protecting group and forms the disulfide in one step, but iodine also removes Trt, which constrains what else may be on the peptide at that point; N-chlorosuccinimide removes Acm with concomitant disulfide bridge formation without affecting the Trt group, can be used in the presence of other commonly used cysteine thiol protections, and has been applied on resin in the synthesis of multi-disulfide peptides including α-conotoxin SI [13].
Then there is what the reagent does to the rest of the sequence. In the preparation of the amino-terminal loop of a snake α-neurotoxin, [Cys3,Cys23,Ser17](1–24) amide, the iodine method failed to yield the expected peptide and gave instead the mono-iodinated tyrosine analogue. Thallium(III) trifluoroacetate and mercuric acetate/potassium ferricyanide both gave cyclised peptide whose observed molecular weight exceeded the calculated value by a multiple of the Acm moiety, traced to an S→O(Ser,Thr) Acm shift, and in a separate Ser/Thr-rich cyclised decapeptide from the same study the Tl(TFA)3 process additionally showed dimerisation; glycerol proved efficient as a scavenger in both the Tl(TFA)3 and the mercuric/ferricyanide methods [14].
Tryptophan is the other residue that forces a redesign. In the chemical synthesis of human insulin-like peptide-6, which shares the distinctive disulfide bond configuration of human insulin, the presence of an oxidation-sensitive tryptophan prompted two new orthogonal methodologies: identification of an additive to suppress the oxidation of tryptophan during iodine-mediated S-Acm deprotection, and an iodine-free, sulfoxide-directed disulfide bond formation [15]. Audit the sequence against the oxidant before fixing the protecting-group scheme — tyrosine and tryptophan against iodine [14, 15], serine and threonine against reagents that liberate the Acm cation [14].
Monitoring: RP-HPLC and the free-thiol count
Reversed-phase HPLC carries most of the load on both routes. The folding efficiencies of the five ω-conotoxins were assigned from RP-HPLC behaviour [3], and the three GIIIB products were separated well enough for the disulfide structure of each to be determined [8]. What a chromatogram licenses is a count of species and their proportions, plus the disappearance of reduced starting material. What it does not license is an assignment: co-elution with a reference is consistent with the reference connectivity and is not evidence of it, and isomers that differ in retention need not differ enough to resolve at analytical scale. Which cysteines are joined is a separate measurement, and belongs to the disulfide mapping workflow rather than to this one.
The orthogonal completeness check is a free-thiol assay. DTNB reacts with a thiol to release 2-nitro-5-thiobenzoate, whose anionic form has a molar extinction coefficient of 14,150 M−1 cm−1 at 412 nm; tris(2-carboxyethyl)phosphine reduces DTNB rapidly and stoichiometrically to generate two equivalents of that chromophore [16]. Read the result against a fully reduced control of the same peptide rather than against a molar concentration calculated from weighed mass, which drags counter-ion and water content into the answer.
The reduced control is where this assay usually goes wrong. Dithiothreitol readily oxidises above pH 7.5, whereas TCEP is very stable in acidic solutions and also highly stable in basic solutions, and below pH 8 TCEP is significantly more effective than DTT at reducing DTNB [16]. A control reduced with an aged DTT stock at the alkaline pH typical of a folding buffer can therefore be incompletely reduced, which depresses the apparent thiol content of the reference and risks making an incomplete oxidation look finished.
Intact mass is the third readout and the cheapest. Each disulfide formed removes two hydrogens, so the measured mass falls by 2 Da per bridge relative to the fully reduced chain, and a covalently linked dimer appears near twice the monomer mass. Run the intact-mass measurement before interpreting the chromatogram, or a symmetric dimer carrying the right number of bridges per chain will read as a clean single product.
What a finished oxidation entitles you to say
A single sharp peak at the expected mass with no detectable free thiol establishes that the cysteines are oxidised, that the material is monomeric under the conditions of that separation, and that one species dominates. It establishes nothing about which cysteines are joined to which.
If you took the random route, the evidence for your pairing is that one isomer dominated in your buffer. The epidermal growth factor mutant series is the standing reminder that dominance at equilibrium can belong to a non-native bridge [4], so that evidence is about the conditions you chose, not about the molecule.
If you took the directed route, the evidence is the deprotection order, and it holds exactly as far as the orthogonality held. Iodine stripping Trt alongside Acm [13] produces an isomer of the intended mass and will not announce itself in a mass check; the Acm cation migrating onto a serine or threonine hydroxyl will, as a mass too high by a multiple of the Acm moiety [14]. Both are ways for a scheme to be right on paper and wrong in the flask.
In either case the pairing remains an assumption until it is mapped. Schedule the mapping experiment against the same batch as the oxidation, because it is the only step in this workflow that produces the claim everything else is presumed to support.