Bicyclic Peptides by Cysteine Alkylation onto Tris-Halomethyl Scaffolds

A linear peptide carrying three free cysteines and a small aromatic core carrying three benzylic bromides need only be mixed. Each thiolate displaces one halide by nucleophilic substitution, the peptide ends up tethered at three points to a single core, and two fused macrocycles close in the same reaction. Heinis and co-workers established this with 1,3,5-tris(bromomethyl)benzene (TBMB) on peptides fused to the phage gene-3 protein, noting that the reaction occurs in aqueous solvents at room temperature (20–25 °C) and that the threefold rotational symmetry of TBMB ensures the formation of a unique structural and spatial isomer ^[1]. The mass signature is small and specific: the expected shift on conjugation is 114 Da, corresponding to addition of the mesitylene core ^[1].

That is a different kind of operation from building two rings one after the other. Bicyclic peptides in general are two macrocyclic rings joined through a chemical linker, and relative to monocyclic peptides of comparable molecular mass they are more constrained in conformation, which is the reason they can bind targets with higher affinity and show greater resistance to proteolytic degradation ^[2]. The question is how you get the second ring.

Why a trivalent scaffold is not sequential ring closure

Close two rings by oxidation and the connectivity is not settled by the chemistry alone. Chen and colleagues built this ambiguity into a library on purpose: phage-encoded peptides of the format X_mCX_nCX_oCX_p were oxidised to connect two pairs of cysteines, and because the fourth cysteine can appear in any of the randomised positions the format gives 3 × (m + n + o + p) different peptide topologies ^[3]. Their X-ray analysis showed the disulfide bridges playing an important structural role, not merely holding the ends together ^[3]. Worth reading that paper in the direction it actually runs: the authors presented disulfide oxidation as the simpler and more robust procedure, and phage-peptide alkylation as the technically complex one they were trying to avoid ^[3].

A tris-halomethyl scaffold trades that flexibility for determinacy. The branch point lives in the reagent, so with three cysteines and a C₃-symmetric core there is only one way for the three arms to be assigned, and no orthogonal thiol protection, no staged deprotection and no controlled oxidation order are involved. What you give up is the ability to generate topological isomers from one sequence, which is exactly what the disulfide format was designed to exploit ^[3].

Scaffold geometry and loop length are the design variables

The core is not an inert spacer. Chen and colleagues made two new cyclisation reagents, each carrying three thiol-reactive groups, that efficiently and selectively cyclised three-cysteine linear peptides; when the mesitylene linker of the kallikrein-inhibiting bicycle PK15 was replaced by these new linkers, inhibitory activity dropped by a factor of more than 1000, and solution-state NMR revealed different NOE constraints across the three bicyclic peptides ^[5]. Their reading is that these relatively small linkers, sitting at the centre of the structure, significantly influence peptide conformation ^[5]. Substitution pattern matters as well as identity: 1,2,3-tris(bromomethyl)benzene gives bicyclic peptides with conformations that differ from the conventional bicyclisation products formed with 1,3,5-TBMB, while still bicyclising cysteines efficiently in aqueous buffer ^[6].

Loop length is the other lever, and it is set entirely by where you place the cysteines. The original repertoire displayed peptides with an N-terminal alanine followed by three cysteines each spaced apart by six random amino acids, giving two six-residue loops on the mesitylene core and an estimated 4.4 × 10⁹ variants ^[1]. Loop geometry decides what surface the molecule can present: the crystal structure of a bicyclic inhibitor of urokinase-type plasminogen activator (K_i = 53 nM) showed an extended peptide with both loops engaging the target across an interaction surface of 701 Ų, with multiple hydrogen bonds and complementary charge interactions ^[7]. Pushing scaffold diversity further requires leaving the single-core format: cyclising peptides with two chemical bridges rather than one gives access to thousands of macrocyclic scaffolds, an approach built on more cysteines and two bi-reactive linkers rather than one trivalent core ^[8].

Conditions, stoichiometry and the concentration regime

Start from a fully reduced peptide with the reductant removed. In the original protocol the peptide–D1-D2 fusion was reduced with tris(2-carboxyethyl)phosphine, the phosphine was removed by size-exclusion chromatography, and only then was TBMB added; the reaction buffer was 20 mM NH₄HCO₃, 5 mM EDTA, pH 8 with 20% (v/v) acetonitrile, run at 30 °C for 1 h rather than at the 20–25 °C at which the chemistry itself is described as proceeding, and 10 µM TBMB was sufficient for quantitative reaction, giving predominantly one product at the expected mass ^[1]. The same scaffold conditions were used on phage, and 10 µM TBMB in 20 mM NH₄HCO₃, 5 mM EDTA, pH 8 for 1 h is the version repeated in the macrocyclisation literature as the reference point for scaffold thioether formation on a display system ^[9]. The peptide has to be fully reduced when the scaffold arrives and stay that way until the third bond forms, which is why the published order of operations is reduction with TCEP, removal of the phosphine, and only then addition of TBMB ^[1].

Reaction time is a variable, not a formality. Working with a three-cysteine test peptide under the published ammonium carbonate conditions at pH 8 and 30 °C, one group saw the linear peptide almost quantitatively (>95%) converted to the bicycle in under 5 min — but as the reaction continued past 10 min, the bicyclic product itself was considerably further modified by the 1.5 equivalents of TBMB present ^[10]. Excess scaffold does not simply sit there once the intended three bonds have formed.

The competing pathway is polymer. Most peptide cyclisations need to be run in dilute solution so that the intramolecular reaction outcompetes intermolecular oligomerisation ^[9]. For a trivalent scaffold this applies twice over, because after the first alkylation the remaining two electrophiles can be captured either by the remaining two cysteines of the peptide already anchored to that core — which is the intended path and enjoys a high effective local concentration — or by cysteines on a second peptide molecule, which starts a chain. On phage there is a second, measurable penalty for raising scaffold concentration: conjugation caused a fivefold loss of phage infectivity at 10 µM TBMB but much greater losses at higher concentrations, which the authors suggested may be due to cross-linking of the phage coat proteins through lysine residues ^[1].

Do not assume near-quantitative conversion transfers to every display format or every linker. A two-fold symmetric linchpin that bridges two cysteine side chains and an N-terminal serine-derived aldehyde converted phage-displayed libraries to bicycles in 45 ± 15% yield — a different chemistry from tris-alkylation, but a reminder that conversion on a displayed library is a number to measure rather than assume ^[11].

Residue tolerance and the free-cysteine rule

Chemoselectivity here is preferential, not absolute, and the original characterisation says so precisely. Reacting TBMB with a fusion containing only two cysteines gave a product whose mass was consistent with reaction of both cysteines and the α-amino group at the peptide N terminus; a construct with one cysteine and a lysine gave a mass consistent with reaction of the cysteine, the N-terminal α-amino group and the ε-amino group of the lysine. The stated conclusion is that the functional groups of TBMB preferentially react with the thiol groups of cysteines but can also react intramolecularly with primary amines in the absence of cysteine ^[1]. In other words, a missing or blocked third cysteine does not give you a monocycle plus a dangling benzyl bromide; it gives you a scaffold whose unreacted arm can instead be captured intramolecularly by a primary amine.

The corollary is that no other free or reducible cysteine may be present anywhere in the construct. This constraint was not negotiable even in the founding work: the authors could not selectively conjugate the three peptide cysteines while sparing the three disulfide bridges of the D1 and D2 domains of the phage protein, which is why they moved to a disulfide-free gene-3 protein ^[1]. Synthetic bicyclisation formats state the same rule as a sequence specification — CX_nCX_nC, where X is any proteinogenic amino acid except cysteine ^[10].

Within that constraint the residue tolerance is broad. Bicyclisation with 1,2,3-TBMB gave clean conversion (>90%) for 96 of 115 diverse peptides tested ^[6]. In a related scaffold chemistry — tetravalent scaffolds carrying two benzyl bromides and two alkynes, reacting with cysteines and azides to give tricycles — 18 canonical amino acids were successfully implemented, and there appeared to be no limitation on the number and types of amino acids present ^[12]. In that tetravalent-scaffold system, scaffold flexibility governed product purity: the flexible scaffolds mostly promoted formation of single isomeric tricycles while the rigid ones did not yield clean products ^[12].

Characterisation: the mass is necessary, not sufficient

A +114 Da shift on a three-cysteine peptide is consistent with the intended bicycle ^[1]. It is consistent with other things too, and the useful habit is to ask what the measurement leaves open. It does not tell you which nucleophiles reacted: an adduct in which one arm has captured a primary amine instead of the third cysteine still adds one core and still loses three HBr, so the arithmetic gives the same nominal mass. Nor does raw m/z separate the monomeric bicycle from a cyclic dimer of two peptides and two cores — the doubly charged dimer ion sits at the same m/z as the singly charged monomer. Deconvoluted masses and an analytical HPLC trace of the crude, not a single peak in a raw spectrum, are the readout that distinguishes them.

The founding paper shows what an orthogonal check looks like. Digesting cleaved PK15 with leucyl aminopeptidase removed the N-terminal alanine, and that result is what confirmed the N-terminal amine had not reacted with TBMB ^[1]. Mass alone was silent on the point; a residue-level digestion answered it.

Connectivity ambiguity is a general hazard of multiply bridged peptides. In the disulfide case, a peptide with two disulfide bonds has three linkage isomers — side-by-side, overlapped, and looped-within-a-loop — and distinguishing them required selective disulfide cleavage chemistry followed by MS³ CID rather than a single mass or a single MS/MS spectrum ^[4]. Tris-alkylation onto a C₃-symmetric core sidesteps the arm-assignment version of this problem by symmetry ^[1], but the escape is specific to that symmetry: with a non-symmetric core the conformational outcome changes ^[6], and with four cysteines cyclised on phage by a bi-reactive scaffold, three isomers with different conformations arise per linear sequence ^[9].

For structure rather than connectivity, the honest limits are on record. Two-dimensional ¹H NMR of PK15 gave a sequence-specific assignment of TOCSY and NOESY chemical shifts, but there was no evidence of interactions between the loops, no NOEs across the loops, and only one NOE between protons separated by more than two residues; the absence of long-range and medium-range NOEs precluded a restraint-driven minimisation and hinders the definition of a single defined structure with confidence ^[1]. NMR of a bicycle of this kind can confirm that the loops behave as independent, non-interacting segments around the core without delivering one rigid conformer. A co-crystal with the target is the measurement that pins the bound geometry down ^[7]. Step-by-step protocols for library generation, affinity selection and ligand characterisation, with troubleshooting for the frequent failure modes, are published in protocol form ^[13].

Thioether versus disulfide stability

The stability argument was an expectation before it was a measurement: compared with disulfide-bonded cyclic peptides, the TBMB cross-links were expected to be inert to exchange and stable in reducing environments ^[1]. What the thioether avoids is easiest to see in the deliberately reversible counterpart. TPSMB carries the same planar trivalent benzene core but attaches through pyridyl-activated disulfides rather than thioethers, and the resulting bicycles are reducible on purpose: at 2 mM glutathione the bicycle was completely reduced within 60 min in vitro, whereas at 20 µM glutathione only about 15% was reduced in 1 h, and a 10-fold excess of TCEP reduced it quantitatively in under 15 min ^[10]. A benzylic thioether bicycle offers no comparable handle, which is the point.

On proteolysis, the evidence is consistent but should be reported at the resolution it was collected. PK15 was cleaved in one of the loops only after prolonged incubation with plasma kallikrein, and even then remained active ^[1]. Across 21 diverse bicyclic macrocycles of different ring sizes, sequences and cross-linker geometries, general stability was observed in mouse serum and in a protease mixture ^[11]. For the double-bridged format built from two chemical bridges, an interlaced bridge topology was associated with a plasma half-life greater than three days ^[8] — a property of that architecture, measured in that study, not a general figure for TBMB bicycles.

Choosing between the routes

If you want two rings and you can place three cysteines where you want them, scaffold alkylation is the shortest path: one reagent, one aqueous step, a defined branch point, and cross-links that contain no disulfide for a reductant to cleave. If you want topological diversity from a single sequence, or you are working in a system where a fourth cysteine is unavoidable, the oxidative disulfide format gives you isomers rather than a single product and was developed precisely as the more accessible procedure ^[3]. For head-to-tail and side-chain-to-side-chain closures that do not involve a scaffold at all, and for the dilution and chemoselectivity considerations shared by all of them, see the macrocyclisation guide; for controlling which cysteine pairs with which under oxidation, see the regioselective disulfide guide.