CuAAC and Strain-Promoted Click Chemistry for Peptide Conjugates

Azides and terminal alkynes are close to unreactive toward everything in a peptide except each other, which is why the pair became the default bioorthogonal handle. The uncatalysed reaction, though, is not usable for conjugation. The thermal Huisgen 1,3-dipolar cycloaddition carries an activation barrier of approximately 25 kcal mol⁻¹ for methyl azide and propyne, which results in exceedingly low reaction rates for unactivated reactants even at elevated temperature; and because the HOMO–LUMO energy differences of azides and alkynes are of similar magnitude, both dipole-HOMO- and dipole-LUMO-controlled pathways operate and a mixture of regioisomeric 1,2,3-triazole products is usually obtained when the alkyne is unsymmetrically substituted ^[1]. Copper(I) removes both defects at once. It transforms organic azides and terminal alkynes exclusively into the corresponding 1,4-disubstituted 1,2,3-triazoles, and the rate of the copper-catalysed reaction is increased by a factor of 10⁷ relative to the thermal process, making it conveniently fast at and below room temperature ^[1].

What you get is a bond, not a release mechanism. The triazole is an aromatic, highly stable product ^[14], so a click ligation decides how two halves are joined and says nothing about whether the payload ever comes off again. Triggers and payload release belong to a separate design axis, treated in the companion guide on conjugate linker chemistry.

The catalysis is stepwise, and that sets a floor on copper concentration

Copper does not lower the barrier of a concerted cycloaddition; it replaces the mechanism. The key carbon–nitrogen bond-forming event takes place between the nucleophilic, vinylidene-like β-carbon of the copper(I) acetylide and the electrophilic terminal nitrogen of the coordinated organic azide, giving a strained copper metallacycle. That step has a calculated barrier of 18.7 kcal mol⁻¹, considerably lower than the barrier for the uncatalysed reaction of approximately 26.0 kcal mol⁻¹, which accounts for the observed rate acceleration ^[1].

More than one copper atom is involved, and this is the part with direct bench consequences. Real-time monitoring of a representative cycloaddition by heat-flow reaction calorimetry showed that monomeric copper acetylide complexes are not reactive toward organic azides unless an exogenous copper catalyst is added, and crossover experiments with an isotopically enriched exogenous copper source illustrated the stepwise nature of the bond-forming events and the equivalence of the two copper atoms within the cycloaddition steps ^[3]. Kinetics measured as a function of ligand-to-metal ratio and catalyst concentration were separately found to be consistent with an active Cu₂L formulation ^[6]. Because the active species is multinuclear, halving the copper concentration does not simply halve the rate. Finn and co-workers accordingly recommend copper concentrations generally between 50 and 100 µM for bioconjugation, the lower limit being necessary to achieve a sufficient concentration of the proper catalytic complex, which incorporates more than one metal centre, and note that more than 100 µM copper is usually not needed to achieve high rates ^[5].

Copper source, reductant, and the order of addition

The usual route to Cu(I) in aqueous buffer is copper(II) sulfate plus sodium ascorbate generated in situ, and the ascorbate has two jobs: reducing the metal and consuming dissolved oxygen that would otherwise reoxidise it. In the optimisation study, reactions with 100 µM Cu and 500 µM of the water-soluble ligand THPTA run in air stopped before completion when initiated with 1 mM or less of reducing agent, while the next concentration tested, 2.5 mM, proved sufficient and further increases did not enhance the rate — consistent with the need to remove oxygen from the aqueous solution, approximately 0.27 mM at room temperature plus whatever diffuses in during the reaction ^[5].

Two procedural details from the same work are rules rather than preferences. Ascorbate should not be added to copper-containing solutions in the absence of the ligand; the described routine is to mix CuSO₄ with the ligand first, add that mixture to the azide and alkyne, and only then initiate by adding ascorbate ^[5]. And the catalyst has an upper substrate limit: the Cu–THPTA complex in water is inhibited by excess alkyne, so that protocol is useful for alkyne concentrations below approximately 5 mM, and a different ligand is suggested for more concentrated solutions ^[5].

The accelerating ligand is also a sacrificial reductant

Polytriazolylamines are themselves made by copper-catalysed ligation of azides and alkynes, and the C₃-symmetric member TBTA was shown to be a powerful stabilising ligand for copper(I), protecting it from oxidation and disproportionation while enhancing its catalytic activity ^[4]. THPTA is the water-soluble relative used for aqueous bioconjugation. Not all accelerating ligands behave alike, though: a survey of tris(heterocyclemethyl)amines identified two classes, those giving high rates in coordinating solvents but inhibiting the process when used in excess relative to copper, and those that give fast catalysis in water and are not inhibitory in excess ^[6]. That distinction is why an excess-ligand protocol works in water and can fail in a coordinating organic solvent.

The excess is not there for rate. It is there to keep the copper from oxidising your peptide. Using N-benzoylhistidine as a model compound, the substrate was stable in the presence of CuSO₄ or ascorbate alone at pH 7, but the combination of the two induced oxidation of approximately 16% of it in 90 min, increasing to 65% after 20 h. THPTA protected the histidine moiety in a manner proportional to the ligand concentration: at a ligand/Cu ratio of 2:1 no oxidation was observed after 90 min and only approximately 15% was lost after 20 h, while at 5:1 less than 5% was oxidised after 20 h. The ligand was itself consumed over the same period, with approximately 0.7–1 mM lost at either two or five equivalents relative to copper. The authors therefore suggest that the ligand protects against histidine oxidation as a sacrificial reductant, intercepting reactive oxygen species in the coordination sphere of the metal as they are generated ^[5]. Those percentages were measured on a benzoylated free amino acid, not a peptide; they license an expectation about His-containing sequences, not a predicted loss figure for your construct.

Excess ligand is affordable because at a ligand/Cu ratio of 5:1 the overall reaction rate was only reduced by half, and five equivalents of tris(triazolyl)methylamine ligands are recommended in most cases and especially where substrate oxidation is a danger ^[5]. The published protocol states the logic compactly: five equivalents of a copper-binding ligand are used with respect to metal because the convenient ascorbate conditions produce reactive oxygen species, and the ligand both accelerates the reaction and serves as a sacrificial reductant that protects biomolecules from oxidation ^[7]. The same dual role has been shown on cells, where THPTA served both to accelerate the reaction and to protect the cells from damage by oxidative agents produced by the copper-catalysed reduction of oxygen by ascorbate ^[8].

Ascorbate itself generates a second, non-oxidative hazard. Dehydroascorbate and other ascorbate byproducts can react with lysine amine and arginine guanidine groups, leading to covalent modification and potential aggregation of proteins, which is the reason a carbonyl-trapping additive is used ^[5]. Aminoguanidine is only a modest inhibitor of the catalysis: at 100 µM copper the rate was unaffected by 1 mM aminoguanidine but noticeably lowered at 5 mM and above, whereas at 0.5 mM copper it had very little inhibitory effect even up to 20 mM ^[5]. On a lysine- or arginine-rich peptide, inspect the mass spectrum after a long ascorbate-driven reaction for adducts, not only for the expected triazole mass.

The copper itself is not innocent in cell work. Across four human cell lines, the effects and related toxicity of copper on mammalian cells were found to be dependent on the ligand environment; copper complexes can be highly toxic, can induce changes in cellular metabolism, and can be rapidly taken up by cells ^[9]. A catalyst chosen for rate on the bench is therefore not automatically the catalyst for live cells.

Choosing the alkyne partner

Alkyne identity matters less than most protocols imply, with specific exceptions. Comparing classes of terminal alkynes under typical bioconjugative and preparative conditions gave only modest differences in reactivity; propargyl compounds represent an excellent combination of azide reactivity, ease of installation, and cost, while electronically activated propiolamides are slightly more reactive at the expense of increased propensity for Michael addition, and certain alkynes including tertiary propargyl carbamates are not suitable for bioconjugation because of copper-induced fragmentation ^[10]. The propiolamide caveat is also an orthogonality caveat: an alkyne prone to Michael addition is an alkyne that competes with any thiol chemistry running on the same peptide.

Strain-promoted cycloaddition removes the metal and adds a reagent problem

The copper-free alternative replaces catalysis with ring strain. A strain-promoted [3+2] cycloaddition between cyclooctynes and azides proceeds under physiological conditions without the need for a catalyst, and was demonstrated by selective modification of biomolecules in vitro and on living cells with no apparent toxicity ^[11]. A cyclooctyne bearing electron-withdrawing fluorine substituents combines ring strain and electronics to promote the cycloaddition; that Cu-free reaction is reported to possess comparable kinetics to the Cu-catalysed reaction and to proceed within minutes on live cells with no apparent toxicity ^[12]. That comparison is a statement about their labelling system, not a universal equivalence.

The spread of published cyclooctyne rate constants is wide but must be read with the solvent attached. In a tabulation of practically stable cyclooctynes reacting with benzyl azide or a similar aliphatic azide, second-order rate constants (×10⁻³ M⁻¹ s⁻¹) run from 2.4 for plain cyclooctyne in CD₃CN, through 76 for DIFO in CD₃CN, 120 for DIBO in CD₃CN:D₂O (3:1), 140 for BCN and 310 for DIBAC in methanol, to 960 for BARAC in CD₃CN ^[14]. Rows measured in different solvents are not directly comparable, and the same review notes that BARAC, despite a rate constant of nearly 1 M⁻¹ s⁻¹, is inherently unstable and rapidly decomposes ^[14]. Where a single laboratory measured one reagent against a fixed partner the numbers are cleaner: cyclopropane ring fusion in BCN gave a near 100-fold rate enhancement over plain cyclooctyne (k ≈ 2 × 10⁻³ M⁻¹ s⁻¹), measuring 0.14 M⁻¹ s⁻¹ for the endo isomer and 0.11 for the exo with benzyl azide, rising to 0.29 and 0.19 respectively in the more polar CD₃CN/D₂O (1:2) ^[13].

Faster is bulkier, and that is the trade. Benzoannulation has been shown to be beneficial for cyclooctyne reactivity but inevitably also leads to enhanced steric interactions and lipophilicity, which is typically not beneficial when SPAAC in water is envisaged; attempts to fix the solubility have cost reactivity, with DIMAC displaying excellent water solubility but severely compromised reactivity, and the more hydrophilic DIBO variants TMDIBO and S-DIBO showing much improved water solubility while their rate constants remain rather low ^[14]. A systematic study of biarylazacyclooctynone analogues separates the two contributions: electronic perturbation of the aryl rings had a modest effect on reaction rate, whereas steric hindrance in the transition state can significantly retard the reaction, and the results suggest a correlation between decreased alkyne bond angle and increased cyclooctyne reactivity ^[15].

Two further properties of SPAAC bear on peptide work directly. First, regiochemistry is lost: formation of a mixture of regioisomeric triazoles is typical for all the cyclooctynes except the C₂-symmetric versions ^[14]. A SPAAC conjugate is therefore usually a pair of isomers with one mass, and a clean deconvoluted mass spectrum is not evidence of a single species. Second, cyclooctynes are not inert to cysteine. Evaluating DIBO, DIBAC and BCN for azide-independent protein labelling in vitro, thiol-yne addition with reduced peptidylcysteines was found to be responsible for most of the azide-independent polypeptide labelling, with the product identity confirmed by LC-MS and NMR; alkylating the cysteine thiols with iodoacetamide prevented the undesired reaction, and iodoacetamide is compatible with SPAAC ^[16]. Set against that, the parent BCN alcohols did not undergo structural change upon stirring in the presence of 5 mM glutathione for 48 h in CD₃CN/D₂O (1:2) ^[13]. The two results are not in conflict — one tracks consumption of the cyclooctyne by a small-molecule thiol, the other tracks adduct formation on a polypeptide — but only the second answers the question you actually have about a cysteine-containing peptide.

Azide and alkyne amino acids

The handles go in as building blocks during chain assembly. In the founding solid-phase work, primary, secondary and tertiary alkyl azides, aryl azides and an azido sugar were all used successfully, producing diversely 1,4-substituted triazoles in peptide backbones or side chains, and novel Fmoc-protected amino azides derived from Fmoc-amino alcohols were prepared by the Mitsunobu reaction ^[2]. Side-chain azides can also be generated on the resin rather than bought in: for an i to i+4 side-chain-to-side-chain cyclisation, the linear precursor carrying ε-azido norleucine and propargylglycine at those positions was assembled either by subjecting the resin-bound peptide to selective on-resin diazo transformation of a lysine into Nle(ε-N₃) or by incorporating Fmoc-Nle(ε-N₃)-OH during the stepwise build-up ^[17]. Propargylglycine is the standard alkyne residue and needs no special handling beyond ordinary Fmoc coupling.

The triazole bridge is not a conformationally neutral staple. In that study the resulting cyclopeptide was compared with the corresponding lactam-bridged analogue: circular dichroism and NMR in water/hexafluoroacetone (50:50, v/v) revealed a high prevalence of turn-helical structures involving in particular the cyclic regions of both molecules, but the two differed in where the turn-helical segment sat, which in the lactam involved non-cyclised residues while in the triazole compound it overlapped with the residues involved in the cyclic structure ^[17]. Swapping a lactam bridge for a triazole of nominally similar span is a structural change that has to be measured, not assumed.

On resin or in solution

Both work, and they fail differently. On-resin copper(I) catalysis was mild and efficient, giving greater than 95% conversion and purity in most cases, and the reaction conditions were fully compatible with solid-phase peptide synthesis on polar supports ^[2]. The arguments for staying on resin are that the side chains are still protected, so nucleophilic residues cannot coordinate the metal, and that copper, ligand and ascorbate wash away before cleavage rather than needing a purification step on the finished peptide. Against that, macrocyclisation of long protected sequences is not automatic: cyclisation efficiency in complex peptides remains a significant challenge, and a robust on-resin protocol for side-chain to side-chain macrocyclisation by CuAAC was developed to address it and applied to a series of 21-residue helical peptides ^[18]. Go to solution when the linear precursor is worth isolating and characterising first — in the i to i+4 case above, solution-phase copper(I)-catalysed cycloaddition converted the purified linear nonapeptide into the triazole-containing cyclopeptide ^[17].

Where the orthogonality actually breaks

Azides and alkynes survive Fmoc chemistry, but copper does not ignore the rest of the molecule. Free thiols such as glutathione at more than two equivalents with respect to copper are strong inhibitors of the CuAAC reaction as described, so a peptide carrying a reduced cysteine can inhibit its own ligation once the free thiol exceeds that ratio relative to copper ^[5]. Sequences that chelate copper behave the same way by sequestering the metal; hexahistidine-tagged proteins were found to inhibit the reaction, and the suggested adjustments are to increase the metal–ligand complex to a maximum of 0.5 mM, or two to three equivalents with respect to the His₆ sequence, to employ an accelerating ligand with greater affinity for copper ions, or to add other metal ions such as Ni²⁺ or Zn²⁺ to occupy the metal-binding site ^[5]. The copper-free route trades this set of interferences for the thiol-yne one ^[16] rather than eliminating them: neither chemistry is orthogonal to a free cysteine without an explicit protection or capping step.

The sequencing rule that follows is unglamorous. Run the azide–alkyne ligation while cysteine is still protected or already alkylated, keep histidine-rich sequences and free thiols in mind when the reaction stalls at otherwise adequate copper ^[5], and verify the product by intact mass plus fragmentation rather than by mass alone — the mass shift confirms that a triazole formed somewhere, and on a peptide bearing more than one potential handle it does not tell you where.