How to choose ADC DAR: why more drug is not always better

Did DAR 8 beat DAR 4 in animals? The 2004 measurement says no

This is the study that set the direction for the field, so it is worth seeing the numbers.

Hamblett and colleagues made three versions of an anti-CD30 antibody carrying MMAE, with 2, 4 and 8 drugs. The linker was a valine-citrulline peptide, and the attachment points were the 8 cysteines released by opening the interchain disulfides. In cultured cells, toxicity rose with DAR as expected.

In animals the result changed. At the same antibody dose, DAR 4 had antitumor activity equal to DAR 8, even though DAR 8 carried twice as much MMAE. Clearance from plasma explains why: DAR 8 disappeared 3 times faster than DAR 4 and 5 times faster than DAR 2. The therapeutic window also widened as DAR went down.

DAR looks like a knob for the amount of drug, but it is also a knob for how long the molecule lives in the body.

Why are DAR 7 to 8 ADCs approved anyway?

Because what changed is the payload and the linker, not DAR.

Trastuzumab deruxtecan and sacituzumab govitecan were approved at DAR 7 to 8. By the 2004 conclusion, those should be disadvantaged designs. Newer payloads, such as DNA topoisomerase I inhibitors, can pass through cell membranes. Drug released inside the target cell reaches the cells around it as well, so it also works on cells with low antigen expression. Linkers are more stable in blood, so less drug comes off before the conjugate arrives. A linker made hydrophilic can also hold down the rise in hydrophobicity at high DAR.

So the 2004 conclusion was never "do not raise DAR".

DAR cannot be decided alone; it only means something together with the payload's properties and the linker's design, and that is the correct way to read the 2004 result.

Lysine conjugation: why does it make so many different molecules?

Because an antibody has dozens of lysines that can react, and the drug can land on any of them.

Lysine is the oldest of the three chemistries. An IgG1 has 80 to 90 lysine residues, and about 40 of them are exposed to solvent and can react. With an active ester, any of them reacts, so the average DAR is tuned by the reaction equivalents. Trastuzumab emtansine is a lysine product: DM1 attached to lysines at an average DAR of 3.5.

The problem is the distribution. Two DAR 3 molecules are different molecules if the three drugs sit on different residues. Choosing 3 residues out of 40 gives 9,880 combinations. A real product is a population of these positional isomers.

You can measure the average. Controlling "how many molecules carry the drug at which position" as a product attribute is not realistic.

For lysine products, control extends to the average DAR and the width of the distribution; position is left outside.

Cysteine conjugation: why does DAR come only in even numbers?

Because the method opens disulfide bonds in pairs.

Using the interchain disulfides is the mainstream approach today. An IgG1 has 4 interchain disulfides, which is 8 cysteines. Gentle reduction with TCEP or a similar agent opens the 4 bonds and releases 8 thiols, and a drug-linker with a maleimide group then attaches to them. Reducing agents in use include Thermo Fisher Scientific's Bond-Breaker and immobilized TCEP, and reagents from Merck, FUJIFILM Wako and Tokyo Chemical Industry (TCI).

Each disulfide that opens gives 2 thiols, so DAR takes only the values 0, 2, 4, 6 and 8. The distribution is not continuous as it is with lysine; it appears as five peaks. Brentuximab vedotin, polatuzumab vedotin and enfortumab vedotin are all designed around an average DAR of 4.

The price is that you have cut the bonds that were holding the antibody together. A molecule that has lost its interchain disulfides keeps its shape through non-covalent interactions alone.

The amount of reducing agent sets the average DAR directly. More TCEP opens more disulfides and creates more attachment points. Open too many, and fragments in which the heavy and light chains have come apart increase. What to ask the reagent supplier is not "what equivalent do you recommend" but "at that equivalent, what percentage of chain-separated fragments appears". Even if the average DAR is right, a lot with more fragments is a different product.

The distribution tightens, but the premise remains that the antibody's backbone has been loosened once before the drug is attached.

Site-specific conjugation: what does a single attachment point cost?

It costs you upstream, in how the antibody is made.

The third approach specifies the attachment site by sequence. If a cysteine is introduced by mutation at a chosen position, only that position reacts, and DAR comes out uniform at 2. Another route is to modify the glycan with an enzyme to build a scaffold for a click reaction; Synaffix (Lonza) GlycoConnect belongs to this family. Zedira's microbial transglutaminase enzymes, GenScript's sortase and tag technology, and Ambrx's incorporation of non-natural amino acids aim at the same goal. Click reagents are supplied by Vector Laboratories, Jena Bioscience and BroadPharm.

Because the product is uniform, the average and the distribution can be described by the same number. Unconjugated and over-conjugated species are in principle reduced, which also lightens the purification burden.

The cost comes back upstream. The antibody itself has to be remade, and introducing a mutation can move expression level and stability. If the method uses an enzyme, that enzyme becomes a new raw material, and there is more to remove and test for. A 2026 review paper compares the three families by process integration rather than chemical merit; we read it in a separate piece (in Japanese).

A uniform DAR makes analysis and specifications easier, and you pay for that ease in the antibody manufacturing process.

What keeps the conjugation reaction under control?

Three things: the organic solvent, the way the reaction is stopped, and the way you measure.

The first is the organic solvent. Most drug-linkers do not dissolve in water, so they are dissolved in DMA or DMSO before being added. Anhydrous grades are supplied by Merck, Honeywell, FUJIFILM Wako and TCI. Solvent fractions of 5 to 20% are used. The higher the fraction, the more easily the antibody denatures and the more aggregates (in Japanese) form.

The second is how you stop the reaction. If you move on with unreacted drug-linker still present, it shows up later as free drug. A quencher is added to stop the reaction, and the leftover is removed by ultrafiltration or hydrophobic interaction chromatography (HIC). Preparative HIC media include Thermo Fisher Scientific's POROS Benzyl, Tosoh Bioscience's TSKgel HIC-ADC Butyl and Cytiva's Capto Phenyl ImpRes.

The third is how you measure. The DAR distribution appears as differences in hydrophobicity, so analytical HIC is the main method. Columns used include Tosoh Bioscience's TSKgel Butyl-NPR, Waters' Protein-Pak Hi Res HIC, YMC's BioPro HIC BF, Thermo Fisher Scientific's MAbPac HIC-Butyl and Agilent's AdvanceBio HIC. Adding native mass spectrometry confirms the mass of each DAR species.

Three controls decide whether the reaction stays under control: the solvent fraction, the quench and the HIC trace.

In what order should you settle DAR?

Work through these decisions in this order when you design the DAR of an ADC.

1. Confirm the payload's properties. Can it pass through membranes, how hydrophobic is it, and do you expect it to reach neighboring cells? Until this is settled, the DAR target cannot be set. For a classical payload that does not cross membranes, raising DAR behaves close to the 2004 conclusion.
2. Look at the linker's hydrophilicity. A design that holds down the rise in hydrophobicity limits the worsening of clearance even at high DAR.
3. Choose the conjugation chemistry. If you want the position to be uniform, choose site-specific. If you want the distribution tightened to even numbers, choose cysteine. If you want to use an existing antibody as it is, choose lysine. This is the point where you find out whether the upstream process must be remade.
4. Fix the reaction conditions: reducing agent equivalents, organic solvent fraction, temperature and time. Watch the average DAR and, at the same time, how much fragment and aggregate form.
5. Decide what the specification will say. Will it state only the average DAR, or also the width of the distribution, and will it set an upper limit for unconjugated antibody? What nonclinical safety assessment (in Japanese) and pharmacokinetics (in Japanese) ask about is not the average but which side of the distribution holds what.