TECHNOLOGY EXPLAINER
Antibody Drugs and ADCs
— How do you attach a 1 kDa molecule to a 150 kDa protein?
An antibody–drug conjugate (ADC) is built from three parts: an antibody, a linker and a payload. By mass the drug portion is only about 5%. Yet what defines an ADC's character is an organic chemistry design question: where that 5% is attached, how many copies, and through what kind of bond.
- Antibody drugs and ADCs (the short version)
- The structure of an antibody — IgG, Fab and Fc
- How antibodies have been made — from hybridomas to phage display
- Production in CHO cells and purification with Protein A
- Glycans — why "the same antibody" does not come out the same
- The three elements of an ADC — antibody, linker, payload
- How the drug is attached — lysine, cysteine, site-specific
- Our calculation: DAR heterogeneity grows with the number of combinations
- A materials engineer's view (1): DAR is a distribution, not a mixing ratio
- Linker chemistry — cleavable and non-cleavable
- A materials engineer's view (2): a linker is a material that must not break, yet must break
- Approved ADCs (as confirmed in material published by regulators)
- Glossary / References / Claim-to-source audit
Sourced = stated in published material (link given)
Our calculation = a figure this article derived, with the assumptions spelled out
Not yet confirmed = a plan or a target with no confirmed track record
Structural summaries and readings from a materials or process-design standpoint are marked separately as Commentary.
This article explains chemistry, materials and manufacturing technology. It does not assess the efficacy or safety of any treatment, and it is not medical advice. Approval status and product specifications are quoted only where they appear in material published by the regulators (the FDA and Japan's PMDA). The article does not deal in any way with how individual drugs are used.
1. Antibody drugs and ADCs (the short version)
- What an antibody drug is: of all the antibodies the body makes, a single chosen kind is produced in quantity (a monoclonal antibody) and used as a medicine. An antibody has the property of binding only to one particular partner
- What an ADC is: Tsuchikama and An describe the general structure of an ADC as comprising a humanised or human monoclonal antibody, a cleavable or non-cleavable chemical linker, and a cytotoxic payloadSourced. In short, a potent drug tied to an antibody that brings binding specificity
- Why it is hard: the antibody is a huge protein with a molecular weight of about 150,000; the payload is a small molecule of around 1,000. The two must be joined so that they stay together in the body but come apart at the intended site
An ADC is not "an antibody with a drug stuck on"; it is a material whose character is set by the chemistry of the bond. With the same antibody and the same drug, attaching to lysine or to cysteine can give anything from a single uniform species to a mixture of dozens. Tsuchikama and An record that lysine conjugation gave an average DAR of 3.5 to 4 with a distribution from 0 to 7, whereas site-specific engineered cysteines gave DAR 2 with more than 90% homogeneitySourced.
2. The structure of an antibody — IgG, Fab and Fc
Like other isotypes, an IgG immunoglobulin molecule consists of four polypeptide chains: two identical 50 kDa γ heavy (H) chains and two identical 25 kDa κ or λ light (L) chains, linked by interchain disulfide bonds.Sourced
Each heavy chain consists of an N-terminal variable domain (VH) and three constant domains (CH1, CH2, CH3), with a "hinge region" between CH1 and CH2. The light chain consists of VL and CL, and the light chain associates with the VH and CH1 domains to form a Fab arm (Fab = fragment antigen binding)Sourced
The role of the Fc region matters too. Vidarsson and colleagues state that the Fc region also contains the binding epitope for the neonatal Fc receptor (FcRn), which is responsible for the extended half-life, placental transport and bidirectional transport of IgG to mucosal surfacesSourced. This mechanism is why antibodies stay in the blood for so long. For an ADC, it means a carrier that can transport a drug for a long time (our commentary).
3. How antibodies have been made — from hybridomas to phage display
The wish to be able to produce monoclonal antibodies of a predetermined specificity came true in 1975, when Georges J.F. Köhler and César Milstein described the hybridoma technique for producing monoclonal antibodies. They immortalised antibody-producing cells by fusing them with tumour cells. With this method, monoclonal antibodies of a predetermined specificity can be produced in unlimited quantities.Sourced
The announcement ranks the hybridoma technique as one of the most important methodological advances in biomedicine in the 1970sSourced.
In 1985, George Smith developed an elegant method known as phage display, in which a bacteriophage — a virus that infects bacteria — can be used to evolve new proteins. Gregory Winter used phage display for the directed evolution of antibodies, with the aim of producing new pharmaceuticals. The first one based on this method, adalimumab, was approved in 2002.Sourced
The prescribing information for approved ADCs shows that the antibody part comes from a range of origins: chimeric (cAC10 in ADCETRIS), humanised (KADCYLA, ENHERTU, POLIVY and others), and fully human (AGS-22C3 in PADCEV, and TIVDAK) — each stated explicitly on the labelSourced.
4. Production in CHO cells and purification with Protein A
Antibody manufacture has become an industry-wide shared "platform". In a 2009 review, Kelley summarised it as followsSourced.
Suspension mammalian cell culture bioreactors run in fed-batch mode deliver high product titres in 10 to 14 days. After harvest by centrifugation and depth filtration, Protein A chromatography captures the product, and two further chromatographic polishing steps complete the purification. Two membrane steps assure the viral safety of the product and concentrate and formulate the drug substance.Sourced
| Item | What the paper says |
|---|---|
| Cell line | Mammalian cells are used for all commercially produced therapeutic monoclonal antibodies. Most come from a handful of lines — CHO, NS0 and Sp2/0 — and CHO is the dominant choice because of its long track record since the approval of tissue plasminogen activator in 1987 |
| Culture | Fed-batch, 7 to 14 days, with nutrients added periodically |
| Titre | Initially well below 1 g/L. Fed-batch cultures give 1 to 5 g/L, and some companies report 10 to 13 g/L in extended cultures |
| Bioreactor | 5,000 L to 25,000 L |
| Capture | Protein A chromatography, which includes a low-pH elution step that also serves as a viral inactivation step |
| Polishing | Usually two further steps, most commonly anion exchange and cation exchange |
| Viral safety | A virus-removal filtration step is added |
| Formulation | A final ultrafiltration step formulates and concentrates |
| Overall yield | Purification yield from the culture fluid is 70 to 80% |
| Batch size | If 5 g/L can be purified, a 10 kL to 25 kL bioreactor gives a batch of 15 to 100 kg |
All Sourced (Kelley 2009 [Reference 5]). Note that this describes the situation as of 2009.
5. Glycans — why "the same antibody" does not come out the same
An antibody is a protein, but it is not only protein. The Fc region carries a glycan (a sugar chain).
Each heavy chain carries an N-linked glycan attached at position 297. The glycan is present as a core structure common to all human and rodent IgG, and is also found with fucose, bisecting N-acetylglucosamine (GlcNAc), one or two galactoses and one or two sialic acids addedSourced
Differences in this glycan substantially change what the antibody does. A 2002 paper by Shields and colleagues (Journal of Biological Chemistry) reports the following for human IgG1 lacking fucose on the sugar attached at Asn297Sourced.
- The lack of fucose on IgG1 had no effect on binding to human FcγRI, C1q or the neonatal Fc receptorSourced
- By contrast, binding of fucose-deficient IgG1 to human FcγRIIIA was improved by up to 50-foldSourced
- Antibody-dependent cellular cytotoxicity assays showed enhanced cytotoxicity, particularly at lower antibody concentrationsSourced
The BLENREP prescribing information describes its antibody component as an afucosylated humanised immunoglobulin G1 monoclonal antibodySourced. An antibody with deliberately controlled glycosylation has actually become a product.
What makes glycans interesting to a materials engineer is that they are not written directly into the blueprint (the gene sequence). A glycan is a modification the cell adds after translation, and it can change with culture conditions. So with the same cell line and the same sequence, a change in process conditions can change the properties of the product (our commentary).
Whether a single fucose is present can change binding to a particular receptor by up to 50-foldSourced — in other words, one sugar in a molecule of about 150,000 is making the difference. In materials terms it resembles a trace additive or a surface end group dominating the properties (our commentary). That is exactly why the glycan profile is controlled as a characterisation item for the drug substance. Indeed, the PMDA review report lists the N-linked glycan profile among the characterisation items for the antibodySourced.
6. The three elements of an ADC — antibody, linker, payload
Tsuchikama and An set out the general sequence of ADC action as followsSourced.
The ADC binds to its target cell-surface antigen receptor (step 1), forming an ADC–antigen complex that leads to endocytosis of the complex (step 2). The internalised complex undergoes lysosomal processing (step 3), and the cytotoxic payload is released inside the cell (step 4).
This is what sets how potent the payload has to be. The same paper explains how little actually arrives.
If the efficiency of each step of the ADC mechanism (biodistribution, antigen binding, internalisation, payload release, intracellular stability of the payload and binding of the payload to its target) is assumed to be 50%, only 1.56% of the administered drug molecules can enter the target cells.Sourced
It adds that actual uptake is estimated to be far lower than this assumption (less than 0.01% of the injected dose per gram of tumour), and that the payload's cytotoxic activity therefore needs to be potent enough to eliminate target cells effectively, ideally in the picomolar rangeSourced.
7. How the drug is attached — lysine, cysteine, site-specific
This is where chemical engineers will find the most common ground: where on the antibody the drug is attached, and by what reaction. The method makes a large difference to how uniform the product is.
(1) Lysine amide coupling
Amide coupling, which uses a linker bearing an activated carboxylic ester to join the payload to solvent-exposed lysine residues on the antibody, is a major conjugation method for ADCs.Sourced
However, a typical antibody has about 80 lysine residues, of which about 10 are chemically accessible. This conjugation method therefore often gives multiple ADC species with different DARs and conjugation sites.Sourced
For a maytansinoid ADC, it records an average DAR of 3.5 to 4 with a distribution from 0 to 7Sourced.
The paper also points out that lysine residues important for the antigen–antibody interaction can be modified, which can reduce binding affinitySourced. Not being able to choose the attachment site translates directly into variation in quality.
(2) Cysteine coupling
Human IgG1, the most commonly used isotype in modern ADCs, has four interchain disulfide bonds and twelve intrachain disulfide bonds. The four interchain disulfides, which are generally not critical to the structural stability of IgG1, can be selectively reduced under mild conditions to give 2, 4, 6 or 8 free thiols.Sourced
Because the number of attachment sites is limited and the thiol group is exceptionally reactive, cysteine-based conjugation is considered superior to lysine-based conjugation in terms of DAR control and heterogeneitySourced. The paper also notes, however, that with the simple cysteine conjugation described above, the DAR distribution can range from 0 to 8Sourced.
(3) Site-specific conjugation
| Method | How it works | Resulting DAR |
|---|---|---|
| Engineered cysteines (THIOMAB) | Two new cysteine residues are introduced, one per heavy chain, for selective conjugation | DAR 2 (more than 90% homogeneity) |
| Disulfide re-bridging | Scaffolds such as dibromomaleimide, dibromopyridazinedione and 1,3-bis(p-toluenesulfonyl)propane take up the two cysteines from a reduced interchain disulfide and re-bridge the antibody | Mainly DAR 4 (for dibromomaleimide) |
| Non-natural amino acids | A non-natural amino acid carrying a reactive handle is introduced by genetic engineering and conjugated by oxime ligation, or by copper-catalysed or strain-promoted (copper-free) azide–alkyne cycloaddition | Fixed exactly by the number of residues introduced |
| Microbial transglutaminase | An enzyme from Streptomyces mobaraensis transfers a linker bearing a primary amine to the side-chain amide of heavy-chain Q295 in a deglycosylated antibody. Native antibodies can be modified directly, with no genetic engineering | DAR 2 (one site per heavy chain; 1.8 to 2 in the figure, high homogeneity). Introducing an N297Q mutation first adds two more sites, giving DAR 4 |
| Sortase | Attaches an oligoglycine-modified linker to an LPETG tag on the antibody | Set by the number of tags |
| Glycan remodelling | An endoglycosidase and a galactosyltransferase install an azide-bearing sugar (GalNAz) in the glycan, which is then joined to a strained cyclooctyne linker (GlycoConnect) | DAR 2 (high homogeneity) |
All Sourced (Tsuchikama and An 2018 [Reference 4]). For the non-natural amino acid approach the paper also lists a limitation: the genetic engineering steps need special techniques and biological reagents, and the introduced non-natural residues may trigger an unwanted immune responseSourced.
8. Our calculation: DAR heterogeneity grows with the number of combinations
"An average DAR of 4" does not settle the molecule: where the four drugs sit makes it a different one. Here we count the possibilities as combinationsOur calculation.
- Assumption: about 10 chemically accessible lysine residues (as stated in the paper)Sourced
- Assumption: up to 8 free thiols from reducing the four interchain disulfides (as stated in the paper)Sourced
- Assumption: all attachment sites are equally reactive, and only the choice of positions is counted (in reality reactivity differs)
Calculation
- Lysine conjugation at DAR 4: choose 4 from 10 = 210 ways
- Cysteine conjugation at DAR 4: choose 4 from 8 = 70 ways
- Cysteine conjugation at DAR 8: choose 8 from 8 = 1 way (every site is filled, so there is no positional freedom left)
- Site-specific (THIOMAB, DAR 2): the drug goes on the two introduced sites, so 1 way
Assumptions and limits: this is a count of positional choices, not the actual ratio of products formed. In practice the distribution is set by the reactivity of each site, steric hindrance and the reaction conditions. The figure of "about 10" lysines is also an approximate number given in the paper.
The PMDA review report states the following for ENHERTU (trastuzumab deruxtecan)Sourced.
- Molecular weight of the whole ADC: about 157,000
- Deruxtecan is attached to an average of 8 Cys residues of the recombinant monoclonal antibody
- Molecular formula and molecular weight of the deruxtecan portion: C52H57FN9O13, molecular weight 1,035.06
- Drug attachment sites: light-chain C214, heavy-chain C223, heavy-chain C229, heavy-chain C232
From these figuresOur calculation:
- Total mass of drug plus linker: 1,035.06 × 8 = about 8,280
- Share of the whole ADC: 8,280 ÷ 157,000 = about 5.3%
- Antibody portion: 157,000 − 8,280 = about 148,700. About 95% of the mass is antibody
- Number of attachment sites: the four sites (light-chain C214; heavy-chain C223, C229 and C232) are each doubled because there are two light chains and two heavy chains, giving 8 sites in total. An average DAR of 8 therefore means almost every attachable site is occupiedOur calculation
Assumptions and limits: the molecular weight of 157,000 is the value stated in the report and is an approximate figure including the glycans. The 8,280 and 5.3% are calculated by this article. "Almost every site is occupied" is this article's reading, based on the number of attachable sites matching the average DAR; it does not show that the actual DAR distribution is a single species.
9. A materials engineer's view (1): DAR is a distribution, not a mixing ratio
Engineers who work with formulated materials are used to designing by averages, such as "4 phr of additive". With an ADC, however, the same average DAR can hide different contents. Tsuchikama and An state this plainly:
in general, a broad DAR distribution can lead to reduced efficacy, so the distribution needs to be tightly controlled. A high DAR can raise not only potency but also the risk of aggregation, the clearance rate and the risk of premature release of the toxic payload in circulationSourced.
All three of the concerns named here are phenomena familiar to materials people (our commentary).
- Aggregation: load many highly hydrophobic molecules onto a surface and the particles stick together. Indeed, the paper states that combining a hydrophobic linker with a hydrophobic payload often promotes aggregation of ADC molecules, and cites a case where non-covalent dimerisation was observed in an antibody carrying doxorubicin through a multi-loaded hydrophobic dipeptide linkerSourced
- Clearance: change the surface state and the rate at which it is removed from the body changes
- Premature release: if the bond is weak, the drug comes off before it arrives
As a countermeasure, the paper says this risk can be reduced by using hydrophilic, sufficiently stable linkers, and concludes that for each ADC it is critical to identify the optimal DAR value, with a controlled distribution, that maximises the balance of efficacy, tolerability and cytotoxicitySourced.
In other words, what has to be designed is not a single DAR target but the shape of the distribution. For engineers used to particle-size or molecular-weight distributions, that way of thinking should feel quite natural (our commentary). Indeed, the average DARs stated on the labels of approved products range widely, from 2.3 to about 8Sourced. The "optimal DAR" differs from product to product.
10. Linker chemistry — cleavable and non-cleavable
A linker has to meet a seemingly contradictory requirement: it must not break in the blood, yet it must break inside the cell. Tsuchikama and An divide linkers broadly into cleavable and non-cleavable, according to how the payload is releasedSourced.
(1) Hydrazone linkers — the earliest design, and its lesson
Hydrazone, an acid-labile group, is used as a cleavable linker that releases the free drug by hydrolysis when the ADC is trafficked into acidic endosomes (pH 5.0 to 6.0) and lysosomes (pH about 4.8).Sourced
The paper then describes the limit of this design: in practice, ADCs with hydrazone linkers undergo slow hydrolysis even under physiological conditions (pH 7.4, 37 °C), leading to slow release of the toxic payload.Sourced
Of the early ADCs built on this structure, the paper summarises that all suffered from toxicity and poor tolerability, apparently owing to the instability of the hydrazone linker in circulationSourced. On Mylotarg it records that it was approved in 2000 but withdrawn from the market in 2010 because of a lack of clinical benefit over standard chemotherapy and a high rate of fatal toxicitySourced.
(2) Cathepsin B-responsive linkers — today's mainstream
Cathepsin B is a lysosomal protease overexpressed in a variety of cancer cells. [...] It preferentially recognises particular sequences such as phenylalanine–lysine (Phe-Lys) and valine–citrulline (Val-Cit), and cleaves the peptide bond on the C-terminal side of the sequence.Sourced
In particular, Val-Cit and Val-Ala linkers combined with p-aminobenzyloxycarbonyl (Val-Cit-PABC, Val-Ala-PABC) are the most successful cleavable linkers for ADCs.Sourced
The PABC moiety is explained as acting as a spacer between the Val-Cit moiety and the payload, allowing cathepsin B to exert its full protease activity even on linkers carrying bulky payload molecules such as doxorubicinSourced.
(3) Disulfide linkers — exploiting a concentration difference
This strategy relies on the concentration of reducing molecules such as glutathione being higher in the cytoplasm (1 to 10 mmol/L) than in the extracellular environment (about 5 µmol/L in blood).Sourced
To increase stability in circulation further, a methyl group is often introduced next to the disulfide bond.Sourced
(4) Non-cleavable linkers
Non-cleavable linkers consist of stable bonds that resist proteolytic degradation, ensuring greater stability than cleavable linkers. They rely on complete degradation of the antibody component of the ADC by cytoplasmic and lysosomal proteases, which releases payload molecules still attached to an amino acid residue derived from the degraded antibody.Sourced
With this design, the payload structure must be carefully chosen and designed so that it exerts equal or greater antitumour activity even in that modified formSourced. The success story cited is trastuzumab emtansine (T-DM1, Kadcyla), a conjugate of an anti-HER2 antibody and maytansineSourced.
11. A materials engineer's view (2): a linker is a material that must not break, yet must break
Lay out the requirements for a linker again and you have a textbook stimulus-responsive materials problem (our commentary).
- In environment A (blood: pH 7.4, 37 °C, glutathione about 5 µmol/L), it must not breakSourced
- In environment B (lysosome at pH about 4.8; cytoplasmic glutathione 1 to 10 mmol/L), it must break quicklySourced
In practice there are only three stimuli to work with.
- pH: the gap between 7.4 and 4.8 — 2.6 pH units, or about 400 times in hydrogen-ion concentrationOur calculation
- Redox potential: the difference in glutathione concentration. Between 5 µmol/L and 1 to 10 mmol/L it is 200 to 2,000 timesOur calculation
- Enzymes: cathepsin B, a molecule that recognises specific sequences
What the failure of hydrazone shows is that pH alone did not give enough selectivity (our commentary). Hydrolysis proceeds slowly even at pH 7.4Sourced — a stimulus under which the reaction rate changes continuously cannot deliver an on/off switch.
Peptide linkers cleaved by cathepsin B, by contrast, switch to a more discrete condition: whether an enzyme that recognises a particular sequence is present or not. When designing a stimulus-responsive material, a discrete condition (the presence or absence of a particular molecule) makes a cleaner on/off switch than a continuous variable (temperature, pH) — and the history of ADC linkers can be read as a worked example of that general rule (our commentary).
The explanation of what PABC (p-aminobenzyloxycarbonyl) does is instructive for materials engineers: it acts as a spacer between the Val-Cit moiety and the payload, [...] allowing cathepsin B to exert its full protease activity even on linkers carrying bulky payload moleculesSourced.
In other words, a problem arose in which the cleavage site itself was designed correctly, but a bulky neighbouring molecule got in the way and the reaction did not proceed, and it was solved by inserting a molecule in between (a spacer).
Avoiding steric hindrance at a reactive functional group by adjusting the spacer chain length — that is exactly the thinking used every day in designing coupling agents, silane coupling treatments and crosslinkers (our commentary). PABC is also designed to release the payload tracelesslySourced, so that the spacer itself does not remain on the released product.
The paper states that cleavable linkers are in most cases designed to release the original payload molecule after bond cleavage, and explains the advantage: such traceless drug-release mechanisms allow researchers to estimate the cytotoxic activity of the conjugated payload from the known pharmacological parameters of the free payloadSourced. Leave no residue, for the sake of design predictability — a principle that carries over to materials design too (our commentary).
12. Approved ADCs (as confirmed in material published by regulators)
The table below lists only products for which the FDA's Drugs@FDA database gives a first approval date and whose composition could be confirmed in the prescribing information. Indications and clinical results are not covered. Only the product name, approval date and chemical make-up are given.
| Product (nonproprietary name) | First approval (FDA) | Payload | Linker | Average DAR |
|---|---|---|---|---|
| MYLOTARG (gemtuzumab ozogamicin) | 1 September 2017 (Initial U.S. Approval on the label: 2000) | N-acetyl-γ-calicheamicin | Acid-cleavable (hydrazone-based) | 2 to 3 (0 to 6) |
| ADCETRIS (brentuximab vedotin) | 19 August 2011 | MMAE (microtubule inhibitor) | Protease-cleavable | About 4 |
| KADCYLA (ado-trastuzumab emtansine) | 22 February 2013 | DM1 (maytansine derivative) | Non-cleavable thioether (MCC) | 3.5 |
| BESPONSA (inotuzumab ozogamicin) | 17 August 2017 | N-acetyl-γ-calicheamicin | Acid-cleavable (AcBut plus dimethylhydrazide) | About 6 (2 to 8) |
| POLIVY (polatuzumab vedotin-piiq) | 10 June 2019 | MMAE | Protease-cleavable mc-vc-PAB | 3.5 |
| PADCEV (enfortumab vedotin-ejfv) | 18 December 2019 | MMAE | Protease-cleavable vc (SGD-1006) | About 3.8 |
| ENHERTU (fam-trastuzumab deruxtecan-nxki) | 20 December 2019 | DXd (topoisomerase I inhibitor, an exatecan derivative) | Cleavable, tetrapeptide-based | About 8 |
| TRODELVY (sacituzumab govitecan-hziy) | 22 April 2020 | SN-38 (topoisomerase inhibitor) | Hydrolysable (CL2A) | 7 to 8 |
| ZYNLONTA (loncastuximab tesirine-lpyl) | 23 April 2021 | SG3199 (PBD dimer, an alkylating agent) | Protease-cleavable valine–alanine | 2.3 |
| TIVDAK (tisotumab vedotin-tftv) | 20 September 2021 | MMAE | Protease-cleavable vc | About 4 |
| ELAHERE (mirvetuximab soravtansine-gynx) | 14 November 2022 | DM4 (maytansine derivative) | Disulfide (sulfo-SPDB) | 3.4 |
| DATROWAY (datopotamab deruxtecan-dlnk) | 17 January 2025 | DXd | Cleavable, tetrapeptide-based | About 4 |
| EMRELIS (telisotuzumab vedotin-tllv) | 14 May 2025 | MMAE | Protease-cleavable vc | About 3 |
| BLENREP (belantamab mafodotin-blmf) | 23 October 2025 (BLA 761440) (Initial U.S. Approval on the label: 2020) | mcMMAF | Protease-resistant maleimidocaproyl | About 4 |
All Sourced (approval dates from FDA Drugs@FDA [Reference 7]; composition from each product's prescribing information [Reference 8]). For MYLOTARG and BLENREP, the "Initial U.S. Approval" on the label differs from the first approval date of the current BLA (Biologics License Application); the table gives both. In Japan, ENHERTU was approved on 25 March 2020Sourced (material published by the PMDA and the Ministry of Health, Labour and Welfare [References 6 and 9]).
- The antibody is almost always made in CHO cells. The exceptions are MYLOTARG, stated as made in NS0 cells, and TRODELVY, in mouse myeloma cellsSourced
- Payloads fall into three broad families: microtubule inhibitors (MMAE, MMAF, DM1, DM4), topoisomerase inhibitors (DXd, SN-38) and agents acting on DNA (calicheamicin, the PBD dimer)Sourced
- Cleavable linkers are in the majority. The non-cleavable ones are KADCYLA and BLENREP, which use a maytansinoid and MMAF respectivelySourced
- DAR runs from 2.3 to about 8. There is no single "right" value (our commentary)
- Most approvals came in 2019 or later: 10 of the 14 products in the tableOur calculation
- IgG consists of two 50 kDa heavy chains and two 25 kDa light chains. Fab binds the antigen; Fc binds receptorsSourced
- The standard is fed-batch CHO culture for production and Protein A for purification. The low-pH elution doubles as viral inactivationSourced
- A single fucose on the glycan can change binding to a particular receptor by up to 50-foldSourced
- About 95% of an ADC's mass is antibody. The drug and linker are only about 5%Our calculation
- DAR is controlled as a distribution, not an average. At DAR 4 on lysine, there are 210 ways to choose positions aloneOur calculation
- A linker is a stimulus-responsive material through and through. pH, redox potential and enzymes — in practice, there are three stimuli to work withSourced
13. Glossary
- Monoclonal antibody
- An antibody derived from a single cell clone that binds only one particular target.
- IgG
- The most abundant immunoglobulin in the blood, made of two heavy chains and two light chains.
- Fab
- Fragment antigen binding. The arm formed when a light chain associates with VH and CH1.
- Fc
- The stem of the antibody, where FcγR, C1q and FcRn bind.
- FcRn
- The neonatal Fc receptor, responsible among other things for the long half-life of antibodies.
- Hybridoma
- A cell immortalised by fusing an antibody-producing cell with a tumour cell. Established in 1975.
- Phage display
- A method of evolving proteins using bacteriophages. Developed in 1985.
- CHO cells
- Chinese hamster ovary cells, the dominant cell line for antibody production.
- Fed-batch culture
- A mode of culture in which nutrients are added during the run. Antibody production runs for 7 to 14 days.
- Protein A
- A protein that binds the Fc of antibodies, used as the affinity ligand in the capture step.
- N-linked glycan
- A glycan attached to asparagine. In IgG it sits at position 297.
- Afucosylation
- Removing fucose from the glycan, which changes binding to particular receptors.
- ADC
- Antibody–drug conjugate, made of three elements: antibody, linker and payload.
- DAR
- Drug-to-antibody ratio. The number of drug molecules per antibody, expressed as an average and a distribution.
- Payload
- The cytotoxic small molecule an ADC carries, such as MMAE, DM1 or DXd.
- Cleavable linker
- A linker that breaks in response to pH, reduction, enzymes or similar triggers to release the payload.
- Non-cleavable linker
- A linker that does not break; the payload is freed when the antibody is degraded.
- Cathepsin B
- A lysosomal protease that recognises and cuts sequences such as Val-Cit.
- PABC
- p-Aminobenzyloxycarbonyl. A spacer between the cleavage site and the payload.
- THIOMAB
- A technology that introduces new cysteines into the heavy chains for site-specific conjugation.
14. References (primary sources)
- The Nobel Assembly at the Karolinska Institute "The Nobel Prize in Physiology or Medicine 1984 — Press release" — nobelprize.org
- The Royal Swedish Academy of Sciences "The Nobel Prize in Chemistry 2018 — Press release", 3 October 2018 — nobelprize.org
- Vidarsson, G., Dekkers, G., Rispens, T. "IgG subclasses and allotypes: from structure to effector functions", Frontiers in Immunology 5, 520 (2014). doi:10.3389/fimmu.2014.00520 — pmc.ncbi.nlm.nih.gov
- Tsuchikama, K., An, Z. "Antibody-drug conjugates: recent advances in conjugation and linker chemistries", Protein & Cell 9(1), 33–46 (2018). doi:10.1007/s13238-016-0323-0 — pmc.ncbi.nlm.nih.gov
- Kelley, B. "Industrialization of mAb production technology: the bioprocessing industry at a crossroads", mAbs 1(5), 443–452 (2009). doi:10.4161/mabs.1.5.9448 — pmc.ncbi.nlm.nih.gov
- Pharmaceuticals and Medical Devices Agency (PMDA) "Enhertu for Intravenous Drip Infusion 100 mg: report on the deliberation results and review report", 2020 (PDF, in Japanese) — pmda.go.jp
- FDA "Drugs@FDA: FDA-Approved Drugs" (approval dates for each BLA) — accessdata.fda.gov
- DailyMed (U.S. National Library of Medicine) FDA-approved prescribing information (MYLOTARG, ADCETRIS, KADCYLA, BESPONSA, POLIVY, PADCEV, ENHERTU, TRODELVY, ZYNLONTA, TIVDAK, ELAHERE, DATROWAY, EMRELIS, BLENREP) — dailymed.nlm.nih.gov
- Ministry of Health, Labour and Welfare (MHLW), Pharmaceutical Safety and Environmental Health Bureau "Points to note on the use of trastuzumab deruxtecan (genetical recombination) products", notification PSEHB/PED No. 0325-1, 25 March 2020 (PDF, in Japanese) — pmda.go.jp
- Shields, R.L. et al. "Lack of fucose on human IgG1 N-linked oligosaccharide improves binding to human FcγRIII and antibody-dependent cellular toxicity", Journal of Biological Chemistry 277(30), 26733–26740 (2002). doi:10.1074/jbc.M202069200 — pubmed.ncbi.nlm.nih.gov
15. Claim-to-source audit
| Claim in the text | Basis | Label |
|---|---|---|
| That Köhler and Milstein described the hybridoma technique in 1975; that they immortalised antibody-producing cells by fusing them with tumour cells; that monoclonal antibodies of a predetermined specificity can be produced in unlimited quantities; and that the hybridoma technique was called one of the most important methodological advances in biomedicine in the 1970s | Nobel Foundation press release, 1984 Prize in Physiology or Medicine[Reference 1] https://www.nobelprize.org/prizes/medicine/1984/press-release/ | Sourced |
| That George Smith developed phage display in 1985; that Gregory Winter used it for the directed evolution of antibodies; and that adalimumab, the first medicine based on this method, was approved in 2002 | Nobel Foundation press release, 2018 Prize in Chemistry[Reference 2] https://www.nobelprize.org/prizes/chemistry/2018/press-release/ | Sourced |
| That IgG consists of two identical 50 kDa heavy chains and two identical 25 kDa light chains linked by interchain disulfides. That the heavy chain consists of VH, CH1, CH2 and CH3 with a hinge region between CH1 and CH2; that the light chain consists of VL and CL; and that the light chain associates with VH and CH1 to form the Fab arm. That the Fc region contains the FcRn binding epitope responsible for the extended half-life, placental transport and bidirectional transport to mucosal surfaces. That each heavy chain carries an N-linked glycan at position 297, which besides the core structure may carry fucose, bisecting GlcNAc, galactose and sialic acid | Vidarsson et al. (2014)[Reference 3] https://pmc.ncbi.nlm.nih.gov/articles/PMC4202688/ | Sourced |
| The general structure of an ADC (humanised or human monoclonal antibody, cleavable or non-cleavable chemical linker, cytotoxic payload) and the four steps of its action. The 1.56% figure when each step is assumed to be 50% efficient; that actual uptake is estimated at under 0.01% of the dose per gram of tumour; and that payloads ideally need activity in the picomolar range. That human IgG1 has 4 interchain and 12 intrachain disulfides, and that the 4 interchain ones can be reduced under mild conditions to give 2, 4, 6 or 8 free thiols. That a typical antibody has about 80 lysines, of which about 10 are chemically accessible; that lysine amide coupling gave a maytansinoid ADC an average DAR of 3.5 to 4 with a distribution of 0 to 7; and that lysines important for antigen–antibody interaction can be modified, reducing binding affinity. That the DAR distribution for simple cysteine conjugation can range from 0 to 8. That THIOMAB gives DAR 2 with more than 90% homogeneity. That re-bridging with dibromomaleimide and similar scaffolds gives mainly DAR 4. The non-natural amino acid method and its limitations. Transfer by microbial transglutaminase to Q295 of a deglycosylated antibody (DAR 2; 1.8 to 2 in the figure; DAR 4 with an N297Q mutation) and its advantage of needing no genetic engineering; sortase conjugation to an LPETG tag; azide installation with an endoglycosidase, a galactosyltransferase and GalNAz, and conjugation to a strained cyclooctyne (GlycoConnect, DAR 2). That a broad DAR distribution can reduce efficacy; that a high DAR can raise the risks of aggregation, clearance and premature release; that hydrophilic, stable linkers can reduce the risk; and that identifying the optimal DAR value with a controlled distribution is critical. That combining a hydrophobic linker with a hydrophobic payload promotes aggregation, and that non-covalent dimerisation was observed in an antibody carrying doxorubicin via a multi-loaded hydrophobic dipeptide linker. That linkers are broadly divided into cleavable and non-cleavable by payload-release mechanism; that cleavable linkers are in most cases designed to release the original payload tracelessly; and that this allows activity to be estimated from the known pharmacological parameters of the free payload. That hydrazone is acid-labile and hydrolysed in endosomes (pH 5.0 to 6.0) and lysosomes (pH about 4.8); that it slowly hydrolyses even under physiological conditions (pH 7.4, 37 °C), slowly releasing the toxic payload; and that early ADCs suffered from toxicity and poor tolerability, apparently owing to hydrazone instability in circulation. That Mylotarg was approved in 2000 and withdrawn in 2010 because of a lack of clinical benefit over standard chemotherapy and a high rate of fatal toxicity. That cathepsin B is a lysosomal protease that preferentially recognises Phe-Lys and Val-Cit and cleaves on the C-terminal side; that Val-Cit-PABC and Val-Ala-PABC are the most successful cleavable linkers; and that PABC acts as a spacer letting cathepsin B exert its full protease activity even with bulky payloads. That disulfide linkers rely on the difference between cytoplasmic glutathione (1 to 10 mmol/L) and blood (about 5 µmol/L), and that a methyl group next to the disulfide increases stability in circulation. That non-cleavable linkers resist proteolysis and, through complete degradation of the antibody, release the payload with an amino acid residue still attached; that the payload structure must be designed to be active in that form; and that the success story is trastuzumab emtansine (T-DM1, Kadcyla) | Tsuchikama and An (2018)[Reference 4] https://pmc.ncbi.nlm.nih.gov/articles/PMC5777969/ | Sourced |
| That mammalian cells are used for all commercially produced therapeutic monoclonal antibodies, most coming from CHO, NS0 and Sp2/0, with CHO dominant because of its track record since the approval of tissue plasminogen activator in 1987. That fed-batch culture runs 7 to 14 days; that titres were initially well below 1 g/L; and that 1 to 5 g/L in fed-batch and 10 to 13 g/L in extended cultures have been reported. That bioreactors are 5,000 to 25,000 L. That harvest uses continuous disc-stack centrifugation and depth and membrane filtration. That Protein A chromatography performs capture and its low-pH elution doubles as a viral inactivation step. That the two polishing steps are anion and cation exchange, with virus-removal filtration and a final ultrafiltration. That overall yield is 70 to 80%. That if 5 g/L can be purified, a 10 to 25 kL bioreactor gives a batch of 15 to 100 kg | Kelley (2009)[Reference 5] https://pmc.ncbi.nlm.nih.gov/articles/PMC2759494/ | Sourced |
| That trastuzumab deruxtecan is an antibody–drug conjugate with a molecular weight of about 157,000; that deruxtecan is attached to an average of 8 Cys residues of the recombinant monoclonal antibody; that the deruxtecan portion has the formula C52H57FN9O13 and molecular weight 1,035.06; that the antibody is a humanised monoclonal antibody produced in Chinese hamster ovary cells, with two heavy chains of 450 amino acid residues and two light chains of 214; that the drug attachment sites are light-chain C214 and heavy-chain C223, C229 and C232; and that the glycosylation site is heavy-chain N300. That the drug-substance characterisation items include the N-linked glycan profile and the drug-to-antibody ratio, and that the drug-to-antibody ratio is set as a specification test for drug substance and drug product. That the review report is dated 17 February 2020 | PMDA report on the deliberation results and review report (Enhertu for Intravenous Drip Infusion 100 mg)[Reference 6] https://www.pmda.go.jp/drugs/2020/P20200420002/430574000_30200AMX00425_A100_1.pdf | Sourced |
| The FDA first approval dates for each ADC in the table in Section 12 (MYLOTARG 1 September 2017; ADCETRIS 19 August 2011; KADCYLA 22 February 2013; BESPONSA 17 August 2017; POLIVY 10 June 2019; PADCEV 18 December 2019; ENHERTU 20 December 2019; TRODELVY 22 April 2020; ZYNLONTA 23 April 2021; TIVDAK 20 September 2021; ELAHERE 14 November 2022; DATROWAY 17 January 2025; EMRELIS 14 May 2025; BLENREP (BLA 761440) 23 October 2025) | FDA Drugs@FDA overview for each BLA[Reference 7] https://www.accessdata.fda.gov/scripts/cder/daf/index.cfm | Sourced |
| The payload, linker and average DAR of each ADC in the table in Section 12. The origin of the antibodies (chimeric cAC10 in ADCETRIS; fully human AGS-22C3 in PADCEV; a human antibody in TIVDAK; humanised antibodies in KADCYLA, ENHERTU, POLIVY and others). The producing cells (NS0 cells for MYLOTARG, mouse myeloma cells for TRODELVY, CHO cells for the rest). That the BLENREP antibody is an afucosylated humanised IgG1. That the MYLOTARG and BLENREP labels give "Initial U.S. Approval" as 2000 and 2020 respectively. That the BESPONSA linker is acid-cleavable, formed by condensing AcBut with 3-methyl-3-mercaptobutane hydrazide (dimethylhydrazide); that the ELAHERE linker is sulfo-SPDB; and that the BLENREP linker is a protease-resistant maleimidocaproyl | FDA-approved prescribing information on DailyMed[Reference 8] https://dailymed.nlm.nih.gov/dailymed/ | Sourced |
| That trastuzumab deruxtecan products were approved in Japan on 25 March 2020 | MHLW Pharmaceutical Safety and Environmental Health Bureau notification (PSEHB/PED No. 0325-1, 25 March 2020)[Reference 9] https://www.pmda.go.jp/files/000234462.pdf | Sourced |
| That for human IgG1 lacking fucose on the Asn297-linked sugar, binding to FcγRI, C1q and the neonatal Fc receptor was unaffected; that binding to human FcγRIIIA improved by up to 50-fold; and that antibody-dependent cellular cytotoxicity was enhanced, particularly at lower antibody concentrations | Shields et al. (2002), abstract[Reference 10] https://pubmed.ncbi.nlm.nih.gov/11986321/ | Sourced |
| Putting the positional choices at DAR 4 at 210 for lysine conjugation and 70 for cysteine conjugation, and at 1 for DAR 8 and for site-specific conjugation. Putting the total mass of the deruxtecan portion at about 8,280, its share of the whole ADC at about 5.3% and the antibody portion at about 148,700. Counting 8 attachable sites on ENHERTU, so that an average DAR of 8 fills nearly all of them. 15,000 L × 5 g/L = 75 kg, and about 56 kg at 75% yield. Putting the difference between pH 7.4 and 4.8 at about 400 times in hydrogen-ion concentration, and the glutathione concentration difference at 200 to 2,000 times. That 10 of the 14 products in the table were approved in 2019 or later | Our calculation. Equal reactivity of attachment sites, about 10 accessible lysines, and 15,000 L, 5 g/L and 75% yield are all assumptions set by this article. The combinatorial counts do not show actual product ratios, and nothing here addresses therapeutic meaning or safety | Our calculation |
| The manufacturing cost of each ADC, process conditions for making linkers and payloads, and yields by company | Not stated because no published primary source could be confirmed within the scope of this article | Commentary |
| The indications, clinical results and safety assessment of each product, and comparisons between products | This article explains chemistry, materials and manufacturing technology and does not assess therapeutic efficacy or safety. For approved products it quotes only composition and approval dates stated in material published by the regulators. The account of Mylotarg's withdrawal in Section 10 is quoted as recorded by Tsuchikama and An (2018) | Commentary |
| Which conjugation methods and linkers will become mainstream in future | At the time of writing (September 2026) this article could not confirm any primary source indicating future trends | Not yet confirmed |
| Framing glycans as a modification not written into the blueprint that changes with process conditions; likening the effect of one fucose to a trace additive or surface end group; treating DAR as "distribution design" of the same kind as particle-size or molecular-weight distributions; reading linkers as stimulus-responsive materials with three usable stimuli (pH, redox potential, enzymes); reading the failure of hydrazone as "a continuous stimulus cannot make an on/off switch"; likening PABC to spacer design in coupling agents and crosslinkers; the reading that the optimal DAR differs by product; and the summary of trends visible in the table | Our summary and commentary based on published content. Not views expressed by the authors of the papers or by the regulators | Commentary |
| That Figs. 1 to 6 are explanatory drawings rather than real 3D structures, equipment or chemical structural formulas, and that the hero image is an AI-generated image | Our note | Commentary |
Last updated 23 September 2026. Sources are limited to primary material (Nobel Foundation announcements, FDA-approved prescribing information and Drugs@FDA, a PMDA review report, an MHLW notification and peer-reviewed papers). Because the article includes readings from a chemical and process-design standpoint, those are marked as Commentary and kept separate from sourced fact. Manufacturing costs, process conditions for linkers and payloads, and yields by company are not stated here because they could not be confirmed in published primary sources. The production data from Kelley (2009) describe the situation as of 2009. This article explains chemistry, materials and manufacturing technology; it does not assess the efficacy or safety of any treatment, and it is not medical advice. All figures are explanatory concept graphics. Figs. 1 to 6 are vector drawings and the hero image is an AI-generated image; none of them shows a real molecule, piece of equipment or product.