Six cysteines can be paired fifteen ways; eight can be paired one hundred and five. Only one of those arrangements is the connectivity you drew, and oxidative folding of the fully reduced chain will find it only if the sequence carries enough of a folding preference to get there. An orthogonal protecting-group scheme removes the question by never letting more than one pair of thiols be free at a time. Everything difficult about building a peptide with two, three or four bridges follows from a single requirement: each protecting group has to survive, intact and still on the sulfur it started on, every condition used to remove the others.
Orthogonality is a property of the set
The term entered synthetic practice through Barany and Merrifield's orthogonal protecting-group strategy, which opened the way for solid-phase peptide synthesis and for the assembly of oligosaccharides and oligonucleotides, and has since spread into general use as a term about chemical reactivity and selectivity [1]. Applied to cysteine it is narrower and more testable than the general usage suggests. More than sixty individual protecting groups have been reported for the cysteine thiol [2], and not one of them is orthogonal on its own. A group is orthogonal only with respect to a named list of partners and the specific reagents used to strip them; swap one reagent and the set can stop being orthogonal without any group in it changing.
That is not a semantic point, because the test is empirical and the answer comes back graded rather than binary. In a direct measurement, five 10-mer peptides were built, each carrying a different protecting group on one cysteine and an Acm pair on the two cysteines destined to be bridged. On-resin N-chlorosuccinimide (3 equivalents in DMF, 3.5 min, repeated once) removed the Acm pair and closed that bridge, and the HPLC purity of the cyclised product then reports how well the remaining protecting group tolerated the treatment: 76.1% for Trt, 68.4% for Msbh, 67.1% for Dpm, 48.1% for Mmt and 25.4% for Thp [3]. Trt survives well enough to call that pair orthogonal for working purposes; Thp does not. Nothing in the structures alone tells you where the cut-off falls.
A workable scheme therefore has to state four things: the reagent that removes each group, the stability of every other group to that reagent, the residue-level liabilities of that reagent given the actual sequence, and whether the step runs on the support or in solution. Leave one of them unspecified and what you have is a drawing.
The triads that get used
Three independently addressable pairs need three groups with three separable removal conditions, and a small number of combinations recur. Trityl, acetamidomethyl and 4-methoxybenzyl is one such combination. In one implementation that triad was used to build five conotoxins carrying three disulfides each, with overall yields of 20–30% and connectivities reported to be obtainable independently of the sequence [4]. The removal logic in that scheme runs: the Trt pair comes off during acidolytic cleavage with the rest of the side-chain protection and is closed in solution with 4,4′-dipyridyl disulfide; the Acm pair is then opened and oxidised in a single operation with excess iodine; and the Mob pair needs a separate strong-acid treatment before a final iodine oxidation. For the peptide reg3b those steps gave 71.4%, 90.2% and 46.8% respectively [4].
The choice of a non-iodine reagent for the first closure is forced rather than stylistic. Iodine removes trityl along with Acm [3], so in a Trt/Acm/Mob set the trityl pair cannot still be masked when the iodine step runs; it has to be unmasked already and closed by something that leaves Acm and Mob alone, which is what 4,4′-dipyridyl disulfide is doing there.
Trityl, Acm and S-tert-butyl is the other common triad, and it extends. Trt, Acm, tBu and 4-methylbenzyl together were used for the regioselective formation of an α-conotoxin dimer peptide carrying four disulfide bridges [5]. Substituting the acid lability of one member is how a step gets moved onto the support: 4-methoxytrityl comes off in far weaker acid than trityl, and an Mmt/Trt/Acm set allows the first bridge to be closed on resin before cleavage [6].
One naming trap changes the entire deprotection plan and is worth stating plainly. S-tert-butyl (tBu) is a thioether and needs acid or an oxidant to remove; S-tert-butylthio (StBu) is a mixed disulfide and comes off by reduction. StBu is difficult to remove, which is why trimethoxyphenylthio (S-Tmp) was introduced as a replacement — S-Tmp and the dimethoxy analogue S-Dmp are both cleared in 5 minutes with mild reducing agents [7]. A scheme written for one and executed with the other simply leaves the group on.
Iodine, thallium, silver — and what each one damages
Most of the classical Acm reagents do deprotection and oxidation in one step. That is convenient, and it is also why their side reactions land on the product rather than on an intermediate you would have discarded.
Iodine is long established, and it is not a selective reagent. Its cross-reactivity with trityl was covered above [3]. Its residue-level liabilities can dominate the outcome: in an attempt to close the N-terminal loop of a snake α-neurotoxin from the corresponding bis-Acm linear peptide, the iodine method failed to give the expected product and delivered the mono-iodinated tyrosine analogue instead [8].
Thallium(III) trifluoroacetate is a mild oxidant with soft-acid character that cleaves various S-protecting groups of cysteine in TFA with spontaneous formation of cystine. Histidine and tyrosine were reported to remain intact under it, with unmasked tryptophan and methionine the exceptions; the reaction was demonstrated by direct conversion of three model S-protected cysteine peptides into oxytocin, urotensin II and human calcitonin gene-related peptide [9]. The complementarity to iodine is the useful part — iodine attacks tyrosine, thallium leaves tyrosine alone but wants Trp and Met masked.
Thallium and mercury share a separate failure that has nothing to do with oxidation. In the same neurotoxin loop work, both the Tl(III) and the Hg(II) acetate/potassium ferricyanide routes gave cyclised peptide whose observed mass exceeded the calculated value by a multiple of the Acm moiety: the Acm group had migrated rather than left, shifting from sulfur to the oxygen of serine or threonine. The assignment was supported by demonstrating Ser(O-Acm) formation when Boc-Cys(Acm) was treated with Tl(TFA)3 in the presence of an equimolar amount of serine, and glycerol was reported as an effective scavenger in both the thallium and the mercuric/ferricyanide procedures [8]. A peptide rich in serine and threonine is the wrong substrate for a single-step Acm deprotection–oxidation whichever metal you reach for.
Silver was tested specifically against the oxidation-sensitive residues. Treating an S-Acm cysteine with silver trifluoromethanesulfonate and then with DMSO/aqueous HCl was reported to give quantitative conversion, and under those conditions no significant side reactions were observed with the oxidation-sensitive residues Met, Tyr and Trp; oxytocin and the Trp-containing peptide urotensin II were prepared this way. Combining air oxidation with the AgOTf–DMSO/HCl system made regioselective two-disulfide formation feasible, and that was applied to tachyplesin I and endothelin I, each of which has two disulfide bonds and a tryptophan [10]. The cost is that it is two operations rather than one, with the peptide sitting in aqueous acid in between.
N-chlorosuccinimide is the more recent entry, and the reports on its residue tolerance have to be read with their conditions attached. Brønsted-acid activation — 2% TFA — gave rapid, clean disulfide product in about 15 minutes, with roughly 2.2 equivalents of NCS in 2% TFA–DCM needed for Acm- and tBu-protected cysteines, and no significant oxidised products from Met or Trp were found in those trials [11]. Under the different conditions of the on-resin DMF protocol described above, the conclusion was that, like iodine, Met- and Trp-containing peptides show sensitivity towards NCS treatments [3]. Both can hold; neither licenses a general statement about NCS and tryptophan.
Palladium sidesteps the oxidant problem by changing the mechanism. Oxidation of Cys(Acm) to the corresponding disulfide was achieved within minutes in a one-pot operation in aqueous medium using palladium with diethyldithiocarbamate, demonstrated on the peptide oxytocin and on the protein thioredoxin-1 [12]. The protein result is a protein result; what it establishes for a synthetic peptide is that the chemistry runs in water, not that the yield transfers.
Which bridge to close first
Two different constraints set the order, and they are easy to conflate.
The first is chemical and absolute. If any protecting group is even partly labile under the conditions used to remove another, the labile one goes first. In the Trt/Acm/Mob scheme, S-Acm was found to be unstable in TFA at 40 °C with removal increasing over time, which is what forces the Acm bridge to be closed before the Mob deprotection [4]. That is not a preference; reversing it destroys the regiochemistry.
The same work exposes a second-order cost of closing bridges early. Deprotecting the Mob pair on an intermediate that already carried two disulfides reached a maximum of 59% after 18 h at 45 °C in TFA/TIS/water (95:2.5:2.5), the slow rate being attributed to the structural rigidity the existing bonds impose [4]. Each closure makes the next deprotection harder to drive.
The second constraint is topological, and it is weaker than folklore suggests. The order question has been examined directly across three disulfides: complete synthetic flexibility with respect to the order of oxidation was found in targets that lack the knot motif, while an essential order of disulfide bond formation was revealed in the knotted miniprotein EETI-II [13]. The reading that survives that result is narrow. The flexibility and the constraint alike were reported for one one-pot oxidation approach on those particular targets, not established as a property of knotted and unknotted topologies in general. The common practice of closing the most conformationally restricted bridge first is consistent with the knotted case and is not established by it for unknotted scaffolds — on those, treat the order as a variable to optimise against yield rather than a rule to obey.
One-pot approaches change what order costs rather than removing the question. A strategy combining a small molecule, ultraviolet light and palladium for chemo- and regioselective activation of cysteine was reported to give ultrafast, high-yielding one-pot formation of two and three disulfide bonds in peptides and proteins [14]. What that removes is the step-by-step overhead: a conventional orthogonal route requires purification before and after each deprotection, which lengthens the process and costs yield [12].
On the resin or in solution
Closing a bridge on the support is done under pseudodilute conditions, and the argument for it is largely operational: running the whole process in solution generally demands high dilution, which brings its own liabilities in peptide solubility and in laborious product isolation [11]. In the acid-activated NCS work, syntheses were run at 0.05 mmol on Rink-amide AM resin with NCS charged at 14 mM in 2% TFA–DCM for 15 minutes per cross-linking step, and semipreparative HPLC purification of the crude reactions gave moderate isolated yields of 40–50%; two regioselective disulfide bonds were installed in an α-conotoxin and in gomesin by iodine/persulfate followed by 2% TFA-activated NCS, about 30 minutes over the two steps [11].
Mixed routes are common and often the right answer. In one α-conotoxin SI synthesis the Cys2–Cys7 bridge was closed on resin with NCS from an Acm pair at 81.7% HPLC purity, and the Cys3–Cys13 bridge was then closed in solution, the final cyclised product coming in at 83.1% [3]. In the three-disulfide conotoxin work all folding steps were run in solution at 2 mg/mL [4]. In the PMP-D2 route the Mmt group was removed and the first disulfide formed on the resin, with the two remaining bridges formed in solution [6].
Assume nothing about which bridges formed
A regioselective scheme is a plan for the connectivity, not a measurement of it, and the gap between the two is documented. During the synthesis of α-conotoxin GI using a PhS(O)Ph/CH3SiCl3 oxidation, rearrangement of disulfide bonds was observed. The protecting scheme — which of the Acm and tBu groups sat where — and the reaction time both influenced the ratio of native to mispaired compound, while the temperature of the reaction mixture had no significant effect, and in all cases the non-native derivative was produced in high amount. The isomers were identified by combining enzymatic digestion with mass spectrometry, and the conclusion for that target was that air oxidation followed by Tl(tfa)3 gave the appropriate compound while the PhS(O)Ph/CH3SiCl3 system gave the non-native isomer [15].
Two things follow for any multicyclic synthesis. Disulfide isomers of one sequence are isobaric, so an intact mass matching the target excludes a retained protecting group and excludes a missing or extra bridge, and says nothing at all about which cysteine is paired with which. A single retention time is not an assignment either, since intended and mispaired isomers may or may not resolve on the gradient you happen to be running.
The assignment has to come from an experiment that localises the linkage. Enzymatic digestion under conditions that suppress exchange, followed by mass spectrometry, is the general route [15]; where the target has published reference data, agreement of two-dimensional NMR chemical shifts with literature values for the correctly folded form has been used as the connectivity check, as it was for two of the three-disulfide conotoxins [4]. Designing that experiment, and the controls that bound its own scrambling rate, is the subject of the disulfide mapping guide.
When the scheme is not worth building
PMP-D2, a 35-residue peptide with three disulfides, is the cleanest published comparison of the two philosophies on one target. Simultaneous air oxidation of all six cysteines gave 13% overall. Two regioselective routes on the same peptide — Trt/Acm/Mob, with the first bridge by air oxidation, the second by iodine oxidation of the Cys(Acm) pair and the third after TFMSA/TFA removal of Mob, and an Mmt/Trt/Acm route with the first bridge closed on resin — gave 2.5% and 2%, and the Mob deprotection step was poorly reproducible on a large scale [6]. That is one target and it evidently folds well, so the comparison does not transfer to a sequence that does not; what it establishes is the question to ask before committing to a scheme, which is whether the fully reduced peptide has been given a fair chance to fold on its own.
Protecting groups are also not the only way to direct pairing. A minimal CXC motif has been used to direct both intramolecular and intermolecular cysteine pairing in complex redox media, producing dimeric peptides and bicyclic and tricyclic peptides from fully reduced peptides containing one to six cysteine residues, without protecting groups, post-translational modification or extensive manipulation of the primary sequence [16]. Where the sequence tolerates the extra cysteine, that is one fewer deprotection step whose side reactions you have to characterise.