Cyclic and stapled peptides
A linear peptide in solution samples an enormous number of conformations, and only a small fraction resemble the bound state. Constraining the backbone reduces that entropic penalty, and the same constraint usually blocks proteolysis. Those two effects are the reason cyclization is the first thing to try on a peptide that binds weakly and disappears quickly.
The chemistries, and when each fits
Disulfide cyclization is the simplest, oxidizing two cysteines to close a loop. It is cheap, it is how nature does it, and it is reduced in the cytosol and in serum over time. Good for a first pass and for extracellular targets, poor for anything that has to survive a reducing environment.
Lactam bridges form an amide between a lysine side chain and an aspartate or glutamate. The bond is stable to reduction and to most proteases, and the chemistry is standard on resin using orthogonal protecting groups. This is the usual upgrade from a disulfide when the disulfide worked but did not survive.
Head-to-tail cyclization joins the N and C termini, removing both exopeptidase entry points at once. It is the strongest single intervention against plasma turnover, since aminopeptidases and carboxypeptidases account for much of the degradation, and it works only if neither terminus is part of the binding interface.
Hydrocarbon stapling installs non-natural olefin-bearing residues and closes them by ring-closing metathesis. The staple is placed on the face of the helix opposite the binding surface, at i and i+4 for one turn or i and i+7 for two. This is the method for stabilizing a helix specifically, and it is the one most associated with cell permeability claims.
Bicyclic formats react a scaffold such as a tribromomethylbenzene with three cysteines to create two fused loops. The rigidity is much higher than a single macrocycle and the format is well suited to phage display, which is how most bicyclic binders are found.
What the constraint actually buys
Affinity gains are real and modest. Ten to a hundred fold is a normal result when the constraint genuinely preorganizes the bound conformation. Gains much larger than that usually mean the linear peptide was a poor starting point rather than that the staple was brilliant.
Protease resistance is the most reliable benefit. Head-to-tail cyclization and terminal modification can move a plasma half-life from minutes to hours, and this improvement is more dependable than the affinity one.
Permeability is the claim to treat carefully. Some macrocycles cross membranes well, and the property tracks with the ability to shield backbone amide hydrogens by intramolecular hydrogen bonding, which lets the molecule present a lipophilic face in membrane and a polar one in water. Cyclization alone does not confer this. Many stapled peptides that were reported as cell active turned out to be entering through membrane disruption or to be stuck in endosomes, and the field spent several years correcting that. Demonstrate cytosolic access directly with a chloroalkane penetration or split-luciferase assay rather than inferring it from an activity readout.
Where it disappoints
If the bound conformation is unknown, the placement is a guess. A staple in the wrong position rigidifies the wrong conformer and loses affinity outright. Structural or mutational information about the binding face is close to a prerequisite, and without it the honest approach is a small scan of positions rather than one designed construct.
The chemistry raises cost and lowers yield. Non-natural residues, an on-resin metathesis step with a ruthenium catalyst to remove afterward, and a cyclization that competes with oligomerization all add steps. Macrocyclization is usually run at high dilution for that reason, which limits throughput.
Solubility often falls. Hydrocarbon staples add substantial lipophilicity, and a stapled analog can be markedly harder to formulate than the linear parent.
A sensible order of work
Establish the linear peptide binds and identify the binding face before constraining anything. Scan constraint positions rather than designing one. Measure binding, plasma stability and solubility on every analog, since the three move in different directions and the best affinity is frequently the worst to handle. Confirm the disulfide or lactam formed where intended by mapping rather than by intact mass, because every regioisomer weighs the same.