Enzymatic conjugation
Enzymes place a modification at a sequence they recognize, which buys positional control without needing a free thiol or an unnatural amino acid. What you pay for it is an engineered tag, an enzyme to remove afterward, and a conversion that is rarely quantitative.
Sortase A
Sortase recognizes an LPXTG motif near the C terminus, cuts between threonine and glycine to form an acyl-enzyme intermediate, and resolves it with an incoming oligoglycine nucleophile. The payload carries the glycines.
That reaction is reversible, because the released glycine-containing fragment competes as a nucleophile. That caps conversion, often well short of complete. Two fixes work: a large excess of the glycine nucleophile, which is wasteful if the payload is expensive, or a depsipeptide substrate whose leaving group cannot re-enter, which drives the reaction forward and is the better answer when the chemistry allows.
Evolved variants with faster kinetics and calcium independence have made the method considerably more practical than the original enzyme.
Positional freedom is the constraint. A sortase tag must sit at a terminus, so a payload cannot go in the middle of a domain, and a C-terminal tag on a heavy chain sits close to the FcRn interface.
Microbial transglutaminase
The enzyme forms an isopeptide bond between a glutamine side chain and a primary amine. On an IgG, the accessible glutamine is Q295, which is masked by the N297 glycan, so deglycosylating with PNGase F exposes it and gives clean DAR 2. Introducing an N297Q mutation gives DAR 4.
Alternatively an LLQGA or similar tag can be placed anywhere accessible, which restores positional freedom.
The reaction runs in mild aqueous buffer with no metals and no reduction step, which is the main practical attraction: no interchain disulfides are disturbed, so there is no chain dissociation and no reoxidation window to control.
Specificity is the thing to check. Transglutaminase will use lysines as the amine donor as well, so an excess of a small-molecule amine nucleophile is needed to outcompete intramolecular crosslinking, and the product needs checking for high molecular weight species.
Formylglycine-generating enzyme
FGE oxidizes the cysteine in a CxPxR consensus to a formylglycine, an aldehyde. Nothing else on a protein carries an aldehyde, so the handle is genuinely orthogonal.
Conversion by co-expression is often incomplete, leaving a mixture of converted and unconverted protein that is difficult to separate, and that is the main limitation. Co-expressing FGE at higher levels or treating in vitro with purified enzyme improves it.
The aldehyde is then addressed by hydrazino-Pictet-Spengler or trapped-Knoevenagel chemistry, both of which give stable C-C linked products rather than the hydrolytically reversible hydrazones and oximes that simpler aldehyde chemistry produces.
Choosing between them
If you want no reduction step and a straightforward DAR 2, transglutaminase after deglycosylation is the shortest path, and losing the glycan is acceptable where effector function is not wanted.
If you need the payload at a defined C terminus and can tolerate sub-quantitative conversion, sortase with a depsipeptide substrate works well.
If you need true orthogonality against a protein with other reactive groups, formylglycine is the cleanest handle available.
Across all three, budget for the enzyme removal and for the unconverted species. Separating conjugated from unconjugated protein is usually harder than the conjugation itself, and it is where the yield actually goes. Plan the polishing step at the same time as the chemistry rather than afterward.