There is a length past which building a peptide one residue at a time stops working. The line is not sharp, and it moves with the sequence, the support and the operator, but it is real. Native chemical ligation is the chemistry developed to get past it: a way of joining two fully unprotected peptide segments in water, at a defined point, to give an ordinary amide bond with nothing left at the junction to mark where the join happened.
This guide covers the reaction itself, the thioester that drives it, the cysteine it requires, and the strategies developed to work around that requirement. It also covers what the method costs, because none of it is free.
Why stepwise synthesis runs out of road
Solid-phase peptide synthesis extends a chain anchored to an insoluble support, one residue per cycle, with excess reagent removed by filtration rather than by isolating an intermediate. The strength of that architecture is that no purification is needed between steps. The weakness is the same fact stated differently: because nothing is purified between steps, every failed coupling stays in the pot and is carried forward into the final crude.
The arithmetic is worth doing once, carefully. A 100-residue target requires 99 couplings. If every coupling ran at 99.5% completion, the fraction of chains that never missed a residue would be 0.995 raised to the 99th power, or about 0.61 — roughly 61% of chains are the full-length target and roughly 39% are not. Now split the same target into two 50-residue segments. Each needs 49 couplings, so each arrives at about 0.995 to the 49th power, or 78%. Multiply those two crude purities and you get about 61% again. The stepwise arithmetic on its own gains you essentially nothing, because you have done nearly the same number of couplings either way. This calculation counts couplings only, ignores deprotection steps and the ligation step itself, and assumes a constant efficiency that real syntheses do not have; it is an illustration, not a yield prediction.
What convergent assembly actually buys is separability. Each 50-residue segment can be purified to homogeneity before it is used, so the roughly 22% of failure products in each segment never reach the junction. And the species you must separate after ligation are different in kind: unreacted starting segments differ from the ligated product by roughly half its mass, whereas a deletion peptide in a 100-residue stepwise crude differs from the target by one residue in a hundred.
The second limit is physical rather than statistical. Resin-bound chains associate with one another as they grow, and the association is sequence-dependent. Measurements of swollen peptidyl-resin volume across 87 sequence-unrelated peptides found that nonrandom incomplete acylation is sequence-dependent and is caused by aggregation of the peptide chains, manifested by a decreased swelling capacity; the authors derived a per-residue aggregation parameter and used it to predict potentially difficult sequences [1]. A shielded N-terminus does not couple, whatever the reagent excess. This is why chain length and sequence character interact: a longer chain has more opportunity to associate, and the couplings that fail tend to fail in runs rather than at random.
The reaction: a thioester, a cysteine, and an acyl shift
Native chemical ligation joins two unprotected segments. One carries a thioester at its C-terminus; the other begins with a free cysteine. As originally reported, the chemoselective reaction of the two unprotected segments gives an initial thioester-linked species, and spontaneous rearrangement of that transient intermediate yields a full-length product with a native peptide bond at the ligation site [2]. Two steps, one flask, no protecting groups on any side chain.
The mechanistic reading usually given for those two steps is stated here without citation, as standard textbook description rather than a finding. The cysteine thiol attacks the C-terminal thioester in a reversible thiol–thioester exchange, tethering the two segments through sulfur. That arrangement places the free amine of the same cysteine four bonds from the thioester carbonyl, and an intramolecular S-to-N acyl shift through the resulting five-membered ring — carbonyl carbon, sulfur, C-beta, C-alpha and nitrogen — delivers the amide. The exchange step is reversible and the acyl shift is not, which is what pulls the equilibrium forward. Chemoselectivity follows from the same geometry: internal cysteines can exchange with the thioester, but only an N-terminal cysteine has its amine positioned to close the ring, so exchange at internal sites simply reverses.
Getting a thioester in the first place
The thioester is the harder half of the problem. It does not survive the repeated basic treatments of Fmoc chemistry, which is the dominant route to synthetic peptides today, so the generic workaround is a masked precursor that survives assembly and is converted to a thioester afterward.
One widely used version is the peptide hydrazide. A published protocol describes the solid-phase synthesis of peptide hydrazides by Fmoc chemistry, followed by native chemical ligation of those hydrazides through in situ activation with sodium nitrite and thiolysis; its authors note that the route does not involve the use of hazardous hydrogen fluoride, and that it may help with targets that are not easily produced recombinantly or that carry acid-sensitive modifications [3]. The same protocol is candid about cost: it puts the whole synthesis of a 140-residue target at milligram scale at approximately 30 working days [3]. That figure is a useful anchor for anyone comparing a synthetic route with a biological one.
Why the reaction needs a catalyst
Peptide alkyl thioesters are easy to prepare and correspondingly unreactive, so ligations are run with an added thiol that exchanges into the thioester in situ and generates a better leaving group. A systematic study of thiol additives reported that despite the use of the older catalysts — a thiophenol and benzyl mercaptan mixture, or the alkanethiol MESNA — ligation reactions typically take 24 to 48 hours, and that substituted thiophenols with a pKa above 6 best combined the ability to exchange rapidly and completely with thioalkyl esters with good leaving-group behavior toward the cysteine thiol [4]. The catalyst identified on that basis, 4-mercaptophenylacetic acid, gave an order of magnitude faster reaction in model studies and in the synthesis of one small protein [4]. That is a rate result on specific systems rather than a universal constant, and the authors presented it as such.
What the junction residue costs
Two constraints govern where a ligation site can be placed. The obvious one is the cysteine on the N-terminal side. The less obvious one is the identity of the residue donating the thioester.
A combinatorial study surveyed all twenty naturally occurring amino acids in the X-Cys position by MALDI mass spectrometry of ligations between LYRAX C-terminal thioester peptides and the peptide CRANK; on that model pair the authors reported that all twenty are suitable for ligation, with valine, isoleucine and proline representing less favorable choices because of slow ligation rates [5]. That result is more permissive than it is often remembered as being — the constraint is kinetic, not absolute — but it does mean a junction placed after a beta-branched residue is a deliberate decision about reaction time. The same work assembled two 124-residue proteins by a three-step, four-piece ligation [5], which is the more general point: the number of junctions, not only their chemistry, sets the length of the campaign.
Desulfurization: converting the cysteine to alanine
Cysteine is uncommon in most sequences, so requiring one at every junction is a severe restriction. The most direct answer is to put a cysteine where an alanine belongs, ligate, and then remove the sulfur. That was reported as a general method: a specific alanine in the parent sequence is replaced with cysteine to enable the ligation, and the unprotected ligation product is then selectively desulfurized with hydrogen and metal reagents to convert that cysteine to alanine, demonstrated on targets including a 56-residue domain and a 110-residue variant [6].
Metal-free variants followed. One report describes a new thiol additive, methyl thioglycolate, used to accomplish one-pot native chemical ligation and metal-free desulfurization, assembling ubiquitin from two or three peptide segments; the authors used circular dichroism and racemic protein crystallography to confirm correct folding of the product [7]. Note what that paper does and does not claim — it demonstrates that the one-pot sequence delivers material of sufficient purity for crystallographic study, not a mechanistic account of the desulfurization step.
Protecting the cysteines you want to keep
Desulfurization is not selective for the junction cysteine. Any free cysteine in the sequence, including one that belongs in the final structure, is a substrate. The standard answer is to protect the native cysteines with a group that survives both ligation and desulfurization, then remove it afterward. Acetamidomethyl is a widely used choice [8], and its removal has historically been the awkward step: one report describes conventional Acm removal as requiring harsh reaction conditions and additional purification steps before and after, which extends the synthetic process and reduces overall yield, and offers palladium(II) complexes in aqueous medium as a rapid alternative [8]. The same work assembled three peptide fragments in one pot with Acm protecting the N-terminal cysteine of the middle fragment, and carried out ligation, desulfurization and Acm deprotection in a single one-pot operation on a protein containing two native cysteines [8].
Auxiliaries and selenium
Two other routes avoid the cysteine requirement without a cysteine-to-alanine conversion.
A removable thiol-bearing auxiliary can be attached to the N-terminal amine of the second segment, playing the cysteine's role and then being cleaved away. The honest summary of that field comes from a study that set out to improve it: ligation auxiliaries overcome the requirement for cysteine, but the reported auxiliaries remained limited to glycine-containing ligation sites, and the acidic conditions applied to cleave the typical N-benzyl-type linkages promote side reactions [9]. That study found auxiliary-mediated peptide couplings to be fastest when the ligation proceeds through five-membered rather than six-membered rings, reported that substituents alpha to the amine should be avoided while additional beta-substituents on a mercaptoethyl scaffold accelerated the ligation, and described cleavage with an aqueous solution of tris(2-carboxyethyl)phosphine and morpholine at pH 8.5 [9]. On mechanism the authors hedged, and the hedge should be kept: NMR analysis of a carbon-13-labeled auxiliary showed that cleavage most likely proceeds through a radical-triggered oxidative fragmentation [9].
The selenium route substitutes selenocysteine for cysteine and then removes the selenium. One protocol converts selenocysteine to serine in a single, high-yielding step, which when coupled with ligation at selenocysteine amounts to a programmed ligation at a serine junction; the authors report compatibility with a wide range of functionality including unprotected side chains and appended glycans, and demonstrate the method on a structured cysteine-free protein [10]. A tutorial review places these developments in context, describing thiol- and selenol-derived amino acids, desulfurization and deselenization methods, improved access to peptide thioesters, and combination with recombinantly expressed fragments as the extensions that have expanded the reach of ligation chemistry, including new records for the size of proteins that can be accessed [11].
Epimerization at the thioester carbon
The carbon alpha to a thioester carbonyl is acidified, which raises the question of whether the residue donating the thioester racemizes. The literature here is more specific than a blanket yes or no, and the distinctions matter.
A study of peptide thioester condensation — silver-ion activation followed by coupling to an amino component, which is a different reaction from aqueous native chemical ligation — reported that during preparation of peptide thioesters by the Boc solid-phase method, no substantial epimerization of the C-terminal amino acid was detected, but that epimerization was observed during a thioester–thiol exchange reaction and during segment condensation in dimethyl sulfoxide in the presence of a base [12]. Critically for ligation practice, the same authors report that thioester–thiol exchange reactions in aqueous solution gave no epimerization [12]. In their condensation work a less polar solvent suppressed the problem substantially: the epimer content of an 89-residue coupling product formed from 45- and 44-residue segments fell from 27% to 6% [12].
The reasonable reading is that the risk is real for the activated, basic, organic-solvent regime those authors studied, and that the aqueous thiol-exchange conditions of native chemical ligation sit outside the regime in which they observed it. That is not the same as demonstrating that no ligation protocol can epimerize, and it should not be read that way.
What is not settled
The prediction of difficult sequences is genuinely contested. The older framework derives a per-amino-acid aggregation parameter from measured swelling behavior and uses it to flag potentially difficult sequences in advance [1]. A recent modeling study built on existing peptide synthesis datasets, supplemented with further experimental data, reports that composition-dependent aggregation was a stronger predictor than sequence-based patterns, and states that the authors identified and experimentally validated that finding [13]. Both are describing the same phenomenon with different variables, and the disagreement is about which representation carries the signal. Anyone choosing where to place a ligation junction is choosing under that uncertainty.
Several other things are simply not known in general form. There is no reliable rule stating in advance how long a given segment can be assembled stepwise before its particular sequence defeats the support. Rate results for ligation catalysts and for junction residues are reported on specific model systems, and the extent to which they transfer to an arbitrary sequence has not been established. And the yields quoted in the literature are overwhelmingly those of successful syntheses; the distribution of attempts that were abandoned is not in the record.
What ends up in the material
What follows is the expected analytical consequence of the chemistry above. It is stated without citation because it is a straightforward reading of the reaction scheme rather than a reported finding. A ligation-assembled peptide is likely to carry a different impurity profile from a stepwise one: unreacted N-terminal-cysteine segment and unreacted thioester segment, both far from the product in mass; the hydrolyzed thioester segment, which is the free acid and can no longer ligate; the thiol catalyst and its adducts; and, where desulfurization was used, any material in which the sulfur was not removed, which sits 32 mass units above the target. Because each segment is purified before ligation, the classical deletion peptides of stepwise synthesis appear at segment mass rather than at full length. None of this substitutes for an actual chromatogram and mass spectrum on the specific lot in hand.