Conjugation process development
A conjugation that works once at two milligrams is not a process. Making it reproducible means knowing which inputs move the output, controlling those, and being able to demonstrate the rest do not matter. The work is unglamorous and it is where most of the batch failures are prevented.
Know your starting material
Antibody lots differ in free thiol content, and free thiol is a direct input to the reduction stoichiometry. Measure it on every lot rather than on the first one. An antibody arriving with half an equivalent of free thiol will overshoot the DAR target if the reductant equivalents were set on a fully oxidized assumption.
Buffer composition carries over. Residual histidine, trace metals from a stainless line, or a surfactant from a prior step all change reaction rate. Diafiltering into a defined conjugation buffer before starting removes a whole class of variability, and it is worth the yield.
Concentration affects the reaction more than people expect. Conjugation is bimolecular, and running at five milligrams per milliliter rather than twenty changes the rate and the apparent equivalents needed.
Parameters that move DAR, in order
Reductant equivalents dominate. This is the primary control and it wants a narrow, justified range.
Hold time between reduction and conjugation comes second, because reoxidation is happening throughout it. This is the parameter most often left uncontrolled at bench scale and most likely to break on scale-up, where transfers take longer. Fix it as a defined window with EDTA present, not as "proceed when ready".
Payload equivalents matter less than the thiols available, once the payload is in reasonable excess. Excess drives the reaction to completion against the sites that exist, and the sites that exist were set by the reduction.
Temperature and pH set the rate rather than the endpoint for maleimide chemistry, within the normal working range. They matter more for the slower alternatives.
Quenching, clearance and the hold nobody documents
Quench the unreacted maleimide deliberately, usually with excess N-acetylcysteine or cysteine. Leaving it to react with whatever is available produces adducts at unintended sites during the hold before purification.
Free payload and payload-related species have to be cleared to a specification. Tangential flow filtration handles most of it, activated carbon is effective for hydrophobic small molecules, and a polishing chromatography step removes the rest along with aggregate. Measure the clearance rather than assume it, and measure it after the final step.
Write down every hold time in the process, including the ones between purification and formulation. Conjugates degrade during holds, particularly before the succinimide has been hydrolyzed, and an undocumented overnight hold is a reproducibility problem waiting for the first campaign that runs late.
In-process controls worth having
Free thiol after reduction, by Ellman. Five minutes, and it tells you whether to proceed.
DAR by UV during or immediately after conjugation. Fast, no distribution, and enough to catch a reaction that did not go.
DAR by HIC and aggregate by SEC on the purified material, before formulation. These are the decision points.
Free payload on the final product, and again on stability.
Scaling
Mixing is the thing that changes. At bench scale a reaction is effectively instantaneously mixed, and at scale it is not, so a concentrated payload stream added quickly creates a local excess that over-conjugates part of the batch. Controlled addition rate and defined mixing become real process parameters rather than details.
Payload is usually delivered in an organic cosolvent, most often DMSO or DMA. The cosolvent fraction has to stay below the point where the antibody unfolds, typically around ten percent, and adding a concentrated organic stream too fast produces a local spike well above the bulk value. Dilute the payload stock and add slowly.
Run one intermediate scale before the production campaign. The problems that appear between two milligrams and two grams, a filtration that takes all day, a hold that cannot be met, an aggregate peak that was never visible at low load, are considerably cheaper to find there.