A disulfide bond is a covalent link between the sulfur atoms of two cysteine residues. Counting how many a peptide carries is easy. Saying which cysteines are joined to which is a different problem, and the one most documentation quietly skips. A certificate reporting the correct molecular mass and a single sharp chromatographic peak has told you almost nothing about connectivity. This guide covers how disulfides are formed on purpose, how they rearrange when you are not looking, and which methods map the pairing rather than infer it from the sequence.
Connectivity is not composition
If a peptide contains 2n cysteines and all of them are paired, the number of possible pairing patterns is the double factorial (2n-1)!!. Work it through: four cysteines give 3!! = 3 x 1 = 3 possible patterns; six give 5!! = 5 x 3 x 1 = 15; eight give 7!! = 7 x 5 x 3 x 1 = 105; ten give 9!! = 9 x 7 x 5 x 3 x 1 = 945. Every one of those isomers has the same molecular formula, the same monoisotopic mass and the same amino acid composition. No measurement of intact mass, and no amino acid analysis, can distinguish among them.
Connectivity is conventionally written by cysteine index in sequence order — Cys I to Cys III, Cys II to Cys IV, abbreviated 1-3, 2-4. Residue numbering is less ambiguous but only if the scheme is stated. Either notation is a documentation convention, not a measurement; it records a claim whose evidence lives elsewhere.
Oxidative folding: what is actually happening in the tube
Disulfide formation and rearrangement both proceed by thiol-disulfide exchange, and the nucleophile in that reaction is the thiolate anion, not the neutral thiol. Reactivity therefore tracks ionization. In a study of 16 model peptides carrying single cysteines with thiol pKa values from 7.4 to 9.1, exchange rates against cystine and 2-hydroxyethyl disulfide followed Bronsted correlations similar to those seen with small-molecule models, but against the charged reagents oxidized glutathione and cystamine the apparent Bronsted coefficients were 0 and 0.8 — a divergence the authors attributed largely to electrostatic interactions between charged groups on the peptides and the disulfide reagents [1]. The practical reading is narrow: the local charge environment around a cysteine, not just its pKa, can dominate how fast it reacts, so two peptides with similar cysteine spacing need not fold on similar timescales.
Folding does not proceed cleanly from reduced to native. In an analysis of human epidermal growth factor, a 53-residue peptide with three disulfides, reduced and denatured material was shown to refold through successive populations of one-disulfide, two-disulfide, and fully oxidized three-disulfide "scrambled" isomers before reaching the native structure; at least 27 intermediate fractions were identified, and among the six disulfide bonds present in two isolated scrambled species, only one was native [2]. Non-native bonds are not simply failure modes. In that system they dominated the well-populated intermediates — the authors report that native disulfides make up only 25 percent of those found in the six isolated well-populated species — though the same authors caution that a well-populated intermediate does not necessarily lie on the productive folding pathway.
The reaction therefore has two stages that respond to conditions differently. Oxidative folding of a 26-residue, three-disulfide peptide was described as proceeding through (I) formation of disulfides and (II) rearrangement of mismatched ones, with both temperature and salt strongly affecting the equilibrium of the second stage; in that work a low temperature of 5 degrees C and a high salt concentration such as 2 M ammonium sulfate were needed to obtain the correctly folded product, whereas folding at 21 degrees C in 0.1 M ammonium acetate predominantly gave two mismatched products with or without added glutathione redox reagents [3]. Those numbers belong to that peptide and that buffer. The transferable point is that the shuffling stage, not the oxidation stage, is usually where yield is won or lost.
Conditions that set the outcome
Several handles are standard and largely conventional. Folding buffers are held mildly alkaline, because the thiolate fraction rises with pH and exchange needs thiolate. Peptide concentration is kept low to favor intramolecular pairing over intermolecular dimers. A redox pair — usually reduced and oxidized glutathione — keeps exchange reversible so mismatched bonds can be undone rather than trapped. None of these rests on a single primary result; they are accumulated convention, and the optimum for a given sequence is found empirically.
Simple air oxidation works, and it also produces isomers. Solid-phase synthesis of a three-disulfide peptide, omega-conotoxin GVIA, followed by air oxidation yielded, alongside the target, an isomeric peptide that could be separated from it only by repeated HPLC; digestion and peptide mapping identified its pairings as 1-2, 3-5, 4-6 [4]. Two things follow. A single peak on a first-pass gradient is weak evidence of a single isomer; and when regioisomers do resolve, they can resolve poorly enough that a routine purity method averages them together.
Regioselective strategies
The alternative to letting thermodynamics choose is to build the bonds in order. The approach is standard practice: cysteines destined for different bonds carry different S-protecting groups, removed and oxidized in sequence so each pairing forms while the others stay masked. The specific group combinations are conventional choices refined over decades of synthesis rather than conclusions from a decisive experiment.
The chemistry of each step is documented. In one system, an S-acetamidomethyl cysteine was reported to be converted quantitatively to cysteine by deprotection with silver trifluoromethanesulfonate followed by treatment with dimethyl sulfoxide in aqueous HCl, with no significant side reactions observed at the oxidation-sensitive residues methionine, tyrosine and tryptophan; combining air oxidation with that silver-then-DMSO sequence gave regioselective formation of two disulfides in peptides containing tryptophan [5]. A more recent approach used 2,2'-dipyridyl diselenide in trifluoroacetic acid to convert cysteines protected as either the 4-methoxybenzyl or acetamidomethyl derivative into S-pyridylselenenyl adducts, which the authors showed could be reduced chemoselectively by a five-fold excess of ascorbate at pH 4.5 — a condition they expected would spare already-installed disulfides; when pairs of acetamidomethyl-protected cysteines were treated, the second disulfide formed spontaneously, and the authors offered a mechanism for that as their most likely explanation rather than a demonstrated one [6].
A regioselective route documents intent. It does not by itself document outcome, because deprotection can be incomplete and oxidation can be intermolecular. The connectivity of the isolated product still has to be measured.
Scrambling under storage
Scrambling is thiol-disulfide exchange running in the wrong direction, and it needs a thiolate to start. A single free thiol anywhere in the vial — an unpaired cysteine, a partly reduced impurity, residual reducing agent, a thiol leached from a stopper — is catalytic rather than stoichiometric, because the exchange regenerates a free thiol on the other side. That is why trace free-thiol content matters more than its low absolute level suggests, and why free-thiol determination (Ellman's assay, or maleimide labeling followed by chromatography) is run as a routine check. Both the catalytic argument and the practice of routine free-thiol monitoring are convention, not traceable to one experiment.
The dry state is more stable than solution but is not inert. Lyophilized bovine somatotropin photolyzed with near-ultraviolet light between 305 and 410 nm was found to degrade in a way that exclusively targeted its disulfide bonds; the reaction was initiated by photoionization of tryptophan and one-electron reduction of the disulfide, was strictly dependent on molecular oxygen but did not involve singlet oxygen, and yielded products the authors assigned as alpha-disulfoxide, thiosulfinate and/or thiosulfonate [7]. That is one protein in the solid state under a defined light source, and the authors noted its behavior contrasted with other proteins studied in solution. It is not a general law, but it is a concrete reason that "protect from light" on a cysteine-containing peptide is more than boilerplate.
Disulfides can also gain a sulfur. Trisulfide bonds have been characterized as a product-related species in recombinant protein products, with LC-MS analysis finding measurable basal levels of trisulfide in one such product's formulation and the authors describing the impact of these species on structure and stability as still under investigation [8]. The mass consequence is a trap worth doing on paper. Inserting one sulfur shifts monoisotopic mass by 31.97207 Da; adding two oxygens, as in double oxidation of a methionine or tryptophan, shifts it by 2 x 15.994915 = 31.98983 Da. The two differ by 0.01776 Da, which a low-resolution instrument cannot separate. Single methionine oxidation, at 15.99491 Da, is a different shift entirely and is not the confusable one. Distinguishing a trisulfide from double oxidation requires either high resolving power or localization of the change to a specific peptide.
What is not known is the part practitioners most want. No published relationship converts storage temperature and elapsed time into a scrambling rate for an arbitrary sequence; rates depend on sequence, pH, buffer, free-thiol burden, trace metals, oxygen and light in ways that have not been generalized. Measure connectivity on a stored sample rather than predict it.
The analysis creates artifacts of its own
This is the failure mode that ruins otherwise careful work. A review of mass spectrometric disulfide characterization identifies the central sample-preparation problem as obtaining a solution in which non-native disulfides have not been formed by high pH, high temperature or free sulfhydryl groups, and states that such non-native bonds can lead to erroneous annotation [9]. The digestion step is the usual culprit, because the majority of enzymes used in protein characterization are most efficient at alkaline pH [10] — exactly where thiolate is abundant.
Two ways out exist, each with a cost. Digesting at acidic pH suppresses exchange but, as one group noted while developing an alternative, increases missed and non-specific cleavages, complicating the resulting map; their method instead kept the digestion at basic pH and added cystamine as an oxidizing agent with a low concentration of iodoacetamide, which significantly reduced the scrambled peptides arising from light chain-heavy chain inter-disulfide disruption, relative to conventional methods, in the two human IgG1 antibodies they used as model proteins [10]. The principle is that free thiols generated during preparation must be suppressed or capped before they can attack an existing bond.
Scrambling has also been reported after the sample leaves solution. Photo-induced radical disulfide scrambling in the gas phase during MALDI analysis was described for tryptic peptides from insulin and confirmed during a separate assignment, the authors proposing a possible mechanism and noting it as, to their knowledge, the first such report; even so, they assigned their target's disulfide pattern unambiguously by combining multiple enzymatic digestions with MALDI [11]. Treat a single ionization method that produces an unexpected linkage as a hypothesis, not a result.
Methods that map connectivity rather than assume it
Non-reduced peptide mapping is the workhorse. The peptide is digested without reduction, and any fragment held by a disulfide appears at the summed mass of two chains; comparing that map with a reduced-and-alkylated map identifies which fragments were linked. The method has a structural requirement that is routinely overlooked: there must be a cleavage site between every pair of cysteines you want to distinguish. Cysteines that are adjacent, or separated by too few residues to place a cut, cannot be resolved by digestion alone. Where the enzyme cannot cut, the assignment is an inference from the sequence, not an observation.
Partial reduction with chemical cleavage addresses exactly that case. In one established scheme the denatured protein is limitedly reduced by tris(2-carboxyethyl)phosphine in pH 3.0 citrate buffer to give a mixture of partially reduced isomers, the nascent sulfhydryls are cyanylated immediately with 1-cyano-4-dimethylaminopyridinium tetrafluoroborate under the same conditions, the isomers are separated by reversed-phase HPLC, and cleavage on the N-terminal side of each cyanylated cysteine in aqueous ammonia gives truncated peptides still held by the remaining disulfides, which are then fully reduced and mass mapped; the authors report that a beta-elimination side reaction accompanying cleavage produces overlapping peptides giving complementary confirmation, and describe the strategy as minimizing scrambling and as feasible for proteins containing adjacent cysteines [12]. The acidic conditions throughout are the point — they hold the thiols protonated and unreactive.
Nuclear magnetic resonance offers an orthogonal route, with a caveat its practitioners state plainly: NMR spectra contain no direct evidence of a disulfide bond, so connectivity has conventionally been inferred from three-dimensional structure calculations, which can leave the assignment ambiguous. One method addresses this by applying Bayesian inference to the local topology around cysteine residues; applied to peptides with complex disulfide topologies, it returned a connectivity for mu-conotoxin KIIIA of 1-15, 2-9, 4-16 that differed from previously published results, with the authors presenting additional evidence in support of the revised assignment [13]. That case is worth remembering for a reason beyond method choice: a published connectivity, repeated across the literature, can be wrong.
A mass calculation that tells you how many, not which
Reduction and alkylation give a clean count of cysteines. Reducing one disulfide adds two hydrogens; capping each resulting thiol with iodoacetamide installs a carbamidomethyl group, whose monoisotopic mass is that of C2H3NO: (2 x 12.000000) + (3 x 1.007825) + 14.003074 + 15.994915 = 57.021464 Da, that is 57.02146 Da.
So for a peptide with six cysteines in three disulfides, going from fully oxidized to fully reduced and carbamidomethylated should raise the monoisotopic mass by (6 x 1.00783) + (6 x 57.02146) = 6.047 + 342.129 = 348.176 Da. Observing that shift confirms six cysteines were engaged. It says nothing about which were paired with which — the reduction destroyed precisely the information you are after. That is why connectivity always costs an extra experiment.
What is measured and what is convention
Reported connectivity is a measurement only when a method that localizes the linkage has been run on the actual lot. Purity by reversed-phase HPLC is a statement about a specific gradient; regioisomers may co-elute, and at least one documented case required repeated chromatography to separate them [4]. Intact mass confirms the number of disulfides, not their arrangement. Free-thiol content answers a separate question.
The corresponding uncertainty is worth stating outright. Much of disulfide bench practice — the alkaline folding buffer, the dilute conditions, the glutathione redox pair, the protecting-group combinations, the habit of storing dry and cold and dark — is convention supported by accumulated experience rather than by controlled comparison for the peptide in front of you. What rests on primary measurement is the mechanism of thiol-disulfide exchange, the structure of scrambled folding intermediates in specific model systems, the chemistries of directed disulfide formation, and the analytical methods above. Where a document tells you which cysteines are joined, ask which method produced the answer. If it cannot say, the connectivity is a prediction from the sequence and should be read as one.