Coupling Reagents in Peptide Synthesis: Carbodiimides, Uronium and Phosphonium Salts

Every amide bond in a synthetic peptide is made by turning a carboxylic acid into something an amine will attack. That conversion is the whole job of a coupling reagent, and the differences between the families — carbodiimides, uronium and aminium salts, phosphonium salts — come down to what intermediate they generate, how long it survives, and what else it can react with while it waits. Those differences show up as distinct species in the crude mixture, and some are chemically invisible to the assay reported on a certificate.

What activation has to accomplish

A carboxylate is not electrophilic enough for an amine to attack at a useful rate. Activation converts it into an acyl donor — an active ester, a symmetrical anhydride, or an acylisourea — that an amine displaces readily. Every activation method faces the same three-way competition once that species exists: the intended amine can attack it, something else can attack it, or the activated species can rearrange or eliminate into a dead end. The families differ in how they weight that competition, not in whether it happens.

A second constraint runs underneath all of it. Activation raises the acidity of the alpha proton on the residue being coupled, which opens the door to loss of stereochemistry. Most of the development work in this field has been about breaking that trade-off between rate and configurational stability rather than accepting it.

Carbodiimides and the O-acylisourea

Carbodiimides are the oldest activators still in routine use. Dicyclohexylcarbodiimide (DCC), diisopropylcarbodiimide (DIC) and the water-soluble EDC all work the same way: the carboxylic acid adds across the C=N bond to give an O-acylisourea, a highly reactive acyl donor. If the amine reaches it, the amide forms. If it does not, the acyl group can migrate from oxygen to the adjacent nitrogen, giving an N-acylurea that is unreactive and is a dead end for that amino acid.

The provenance of that N-acylurea was worked out directly. A study reacting alkoxycarbonylamino acid symmetrical anhydrides with diisopropylcarbodiimide found the reaction slow and indirect, and presumed a route in which the anhydride is converted to oxazolone plus free alkoxycarbonylamino acid, with that free acid then reacting with carbodiimide to give the O-acylisourea, which in turn rearranges to N-acylurea; the authors take these experiments to suggest that the reaction between anhydride and carbodiimide is not the source of N-acylurea in peptide synthesis ^[1].

Whether the same O-acylisourea is also the species responsible for racemization is less settled than a typical mechanism scheme suggests. Working with DCC, one group found that the N-acylurea by-product preserved its chiral integrity while peptides formed simultaneously in the same reaction were racemized to a large extent, and stated that this observation is inconsistent with the generally accepted opinion that a racemization-prone O-acylisourea is a common intermediate for both products ^[2]. That is a live disagreement about mechanism, worth holding onto when someone tells you which reagent gives the cleanest stereochemistry.

Additives: HOBt, HOAt, Oxyma

A carbodiimide is almost never used alone. An additive intercepts the O-acylisourea and converts it into a less reactive but more selective active ester — the benzotriazolyl ester from HOBt, the 7-azabenzotriazolyl ester from HOAt, or the oxime ester from Oxyma. The additive suppresses both the N-acylurea rearrangement and the loss of stereochemistry, at the cost of a somewhat slower acyl transfer.

Oxyma, ethyl 2-cyano-2-(hydroxyimino)acetate, was introduced explicitly as a replacement for the benzotriazole additives after those were reported to exhibit explosive properties. Tested as an additive in the carbodiimide approach, Oxyma displayed a remarkable capacity to inhibit racemization together with coupling efficiency superior to HOBt and at least comparable to HOAt, surpassing HOAt in the more demanding peptide models used; stability assays showed no risk of capping the resin under standard coupling conditions, and differential scanning calorimetry and accelerating rate calorimetry gave decomposition profiles consistent with the reported explosivity of the benzotriazole additives while suggesting a lower risk of explosion for Oxyma ^[4]. DIC/Oxyma is now the workhorse pairing in a great deal of Fmoc synthesis for exactly these reasons.

Where the stereochemistry goes

The route usually drawn for loss of configuration at the activated residue is the 5(4H)-oxazolone: the urethane or acyl carbonyl closes onto the activated carboxyl, the resulting ring has an acidic alpha proton, and base removes and returns it without regard to face. A study of copper(II) chloride as a racemization suppressor in carbodiimide couplings describes the 5(4H)-oxazolone as a species that may be formed from an activated carboxyl component during the coupling, and reports that copper(II) chloride has a strong ability to suppress racemization of that oxazolone ^[3]. The hedge in that phrasing is the authors' own.

How much epimerization a given protocol produces depends less on the reagent family than on the base that accompanies it and on how long the activated species sits before the amine arrives. Angell and co-workers, summarizing earlier work from their laboratories, report that several common in situ activating agents — HATU, HBTU and BOP, each in the presence of N-methylmorpholine or diisopropylethylamine — gave levels of cysteine racemization they describe as unacceptable, quantified as 5 to 33 percent. The protocols they go on to recommend include a pentafluorophenyl ester, DIC/HOBt, and HBTU/HOBt with the hindered base 2,4,6-trimethylpyridine at preactivation times of 3.5 to 7.0 minutes, or the same combination in dichloromethane/DMF with no preactivation at all; they also note that several highly hindered bases may replace diisopropylethylamine or N-methylmorpholine to minimize cysteine racemization even with in situ coupling protocols ^[5]. The lesson is that base identity and preactivation window are variables of the same order as reagent choice.

Uronium and aminium salts: HBTU, HATU, COMU

These reagents deliver a preformed active ester in one step: the amino acid carboxylate attacks the reagent's central carbon, displacing the benzotriazolate or oxime, which then closes on the acyl group to give the active ester. HATU and HBTU differ only in that leaving group — 7-azabenzotriazol-1-yl for HATU, benzotriazol-1-yl for HBTU. The azabenzotriazole is generally the more reactive of the two, which is why HATU is reached for on hindered junctions and in cyclizations. A naming caveat: catalogs draw these as uronium salts, with the activating group attached through oxygen, while the peptide literature also describes the N-attached aminium form; the two names refer to the same bottles.

COMU is the same architecture with the benzotriazole replaced by Oxyma and one of the two dimethylamino groups on the central carbon replaced by a morpholino one. Its developers report that the morpholino group markedly influences solubility, stability and reactivity; that COMU performs well in the presence of only one equivalent of base, which they attribute to the hydrogen-bond-acceptor effect; that its by-products are water-soluble and easily removed; and that it shows a less hazardous safety profile than benzotriazole-based reagents such as HATU and HBTU, which they describe as exhibiting unpredictable autocatalytic decompositions ^[7].

COMU's weakness is hydrolytic stability in the solvent it is normally dissolved in. Measured after 24 hours, COMU remaining was 14 percent in DMF against 88 percent in gamma-valerolactone and 89 percent in acetonitrile; the same work assembled the demanding Aib-ACP and JR decapeptides with COMU dissolved in gamma-valerolactone or acetonitrile while the Fmoc amino acids stayed in DMF ^[8]. A reagent that has largely hydrolyzed in a synthesizer reservoir is no longer a coupling reagent — a mundane route to an impurity profile that does not match the protocol on paper.

Guanidinylation: the failure mode that belongs to this class

HBTU and HATU carry the same bis(dimethylamino)methylene carbon, and that fragment can be transferred to a free amine instead of activating a carboxyl. The result is a permanently capped, guanidinylated chain that cannot elongate further.

The best-characterized demonstration is a cyclization that failed. HBTU was used to close a strained ten-membered lactam by side-chain-to-side-chain cyclization between adjacent diaminobutyric acid and D-glutamic acid residues in an enkephalinamide analog. The cyclization did not occur; instead, characterization by mass spectrometry, amino acid analysis, sequencing and NMR identified the major products as the tetramethylguanidinium derivative at the diaminobutyric acid side-chain amine, together with a dimeric linear tetramethylguanidinium-containing peptide, both resulting from transfer of the tetramethyluronium moiety from HBTU to that side-chain amine. The authors concluded that formation of these tetramethylguanidinium side products limits HBTU's usefulness for lactam formation ^[6].

The mass consequence is worth computing, because the charge bookkeeping decides the answer. Capping a side-chain amine this way converts R-NH2 into the tetramethylguanidine R-N=C(N(CH3)2)2, so the free base gains C5H10N2 relative to the unmodified neutral peptide: 98.08 monoisotopic, nominally +98 Da. The guanidinium it forms carries charge without an added proton, so it sits 99.09 above the parent's neutral mass but only 98.08 above the parent's own [M+H]+. The species is detectable, but which offset a spectrum shows depends on how each species ionizes.

Two conditions favor it: reagent in excess of the carboxylic acid component, and a free amine more accessible than the acid is reactive. Both describe a slow cyclization or a sluggish coupling into an aggregated sequence — the situations in which someone reaches for HATU in the first place.

Phosphonium salts: PyBOP and its relatives

Phosphonium reagents — BOP, PyBOP, PyAOP — generate the same benzotriazolyl or azabenzotriazolyl active esters through a phosphonium center rather than an amidinium carbon. Because the transferable fragment in the uronium case is that amidinium carbon and the phosphonium reagents do not carry one, guanidinylation of a free amine is not a described failure mode of this class. That structural argument is the usual reason phosphonium salts are chosen for head-to-tail cyclizations.

They are not inert toward the substrate, though. In work on amidation of glycopeptide-antibiotic scaffolds, reaction with an excess of PyBOP and triethylamine without any added amine or ammonia gave a mixture containing the corresponding N-unsubstituted carboxamides; with bulky amines such as decylamine and adamantylamine the unsubstituted carboxamide appeared as an admixture, whereas the same reactions with highly reactive amines were usually not accompanied by side-product formation ^[11]. The generalizable point is that an excess of activating reagent facing a slow nucleophile finds something else to do.

The Oxyma logic has been carried into this family too. The phosphonium salts PyOxP and PyOxB were reported to exhibit higher capacity to suppress racemization in various peptide models and enhanced solubility in DMF and dichloromethane than benzotriazole-based reagents; PyOxP combined stability with efficiency in assembling penta- and decapeptides containing consecutive Aib residues, and in cyclization models rendered a higher percentage of cyclic material than other potent phosphonium salts tested ^[9].

Direct head-to-head comparisons are rare, and the ones that exist are narrow. Screening PyBOP, HATU and HOBt for coupling cyclic beta-sugar amino acids into alpha/beta-chimera peptides, with time-resolved proton NMR used to compare formation and stability of the Fmoc-protected active esters, found PyBOP the best choice for both beta-sugar amino acid prototypes examined ^[10]. That is a result about two unusual monomers, not a general ranking.

How reagent choice shows in the crude

The clearest comparative evidence concerns difficult sequences rather than easy ones. Investigating aggregating and otherwise hard-to-obtain model peptides, one group compared Oxyma/DIC against HATU/sym-collidine as the second reagent in a double-coupling protocol; comparative analytical data on the unpurified products showed that Oxyma/DIC largely improved the content of the target molecule in the final crude material ^[12]. Note the scope carefully: this is a second-coupling comparison on aggregating models, not a claim that DIC/Oxyma beats HATU everywhere.

Beyond that, the reagent leaves a signature in a few predictable ways.

• Urea by-products. DIC gives diisopropylurea, soluble in DMF and washed out; DCC gives dicyclohexylurea, which precipitates — the practical reason DIC displaced DCC in solid-phase work.
• N-acylurea. It consumes the activated amino acid without coupling it. It adds no peak to the peptide trace; it removes acylating capacity, and the consequence appears downstream as a deletion sequence missing that residue.
• Guanidinylation. A capped chain is a truncation product, and unlike a deletion it stopped growing at a known point.
• Epimerization. A diastereomer has the identical molecular formula. No mass shift, no new molecular ion; detection depends entirely on whether the chromatographic method resolves it.

Only the last of those is reagent-diagnostic in reverse, and it is the one a routine LC-UV plus MS certificate is least equipped to see.

Reagent hazards and handling

Two hazards belong in any discussion of these reagents. The benzotriazole additives HOBt and HOAt have reported explosive properties, and calorimetry gave decomposition profiles consistent with those reports while suggesting a lower risk for Oxyma ^[4]. Separately, the DIC/Oxyma combination is not hazard-free: after a report that Oxyma and DIC can react to generate hydrogen cyanide at 20 °C, a computational study derived two pathways for decomposition of the Oxyma/DIC adduct and, on the basis of computed observed activation energies, identified attack of the sp2 nitrogen at the oxime carbon as the most kinetically favorable route, a conclusion the authors report is supported by closer agreement between predicted and experimental carbon-13 NMR data ^[13]. That work is theoretical, intended to inform strategies for suppressing HCN generation; it does not measure how much is produced in any particular synthesis.

What is not settled

Several things in this area are genuinely open. Whether the O-acylisourea is the common intermediate for both peptide formation and racemization in carbodiimide couplings has been directly challenged on the evidence that the N-acylurea by-product retains chirality while the peptide does not ^[2]. Which reagent racemizes least is model-dependent: the published comparisons use different test peptides, different bases, and different preactivation times, and a ranking established on Gly-Cys-Phe does not transfer to a hindered Aib junction. Guanidinylation is well demonstrated in a strained cyclization ^[6]; how much guanidinylated material an ordinary linear coupling carries under normal reagent stoichiometry is not something the literature reviewed here quantifies.

Finally, a practical limit. Nothing on a routine certificate of analysis identifies which coupling reagent was used. Most reagent-derived by-products — ureas, released additives, spent phosphine oxides — are removed by washing before cleavage and never reach the analytical trace. What survives is the shape of the failure: deletions where activation died, truncations where a chain was capped, and diastereomers where the alpha center was lost. Those are the marks reagent choice actually leaves, and reading them is the subject of our guide to solid-phase peptide synthesis impurities. Terms used above are defined in the research glossary.

All material discussed here is for laboratory research use only.