Site-specific conjugation is usually summarised in a single sentence: attach the payload at a defined position, obtain a homogeneous DAR, avoid the heterogeneity of lysine chemistry or reduced interchain cysteines. The summary is accurate, but it skips the part we find most interesting, which is how that defined position is created in the first place.

Two clinical antibody-drug conjugates make the point nicely. JSKN003 (Alphamab Oncology) and IBI343 (Innovent Biologics) both place the payload at Asn297, the single conserved N-glycosylation site in the antibody Fc. They reach it by two quite different enzymatic routes, and the difference says a lot about how much chemistry sits behind the phrase “site-specific”.

Why Asn297 is such an attractive attachment point

Every IgG1 carries a biantennary N-glycan at Asn297, two copies per antibody, one on each heavy chain. The position is far from the complementarity-determining regions, so modifying it does not interfere with antigen binding. More importantly, the glycan is a chemical handle that no amino acid side chain can offer: a branched carbohydrate structure with terminal residues that enzymes can recognise and modify with complete selectivity.

It also requires no protein engineering. There is no engineered cysteine to introduce, no unnatural amino acid, no change to the antibody sequence at all. For a programme that already has a validated antibody, that is a significant practical advantage.

The complication is that native Fc glycans are heterogeneous. A typical antibody batch contains a distribution of G0F, G1F and G2F glycoforms, with variable fucosylation and variable terminal galactose. Any conjugation chemistry that builds on the native glycan inherits that distribution. The two platforms handle this problem in opposite ways.

Route one: extend the native glycan

JSKN003 keeps the antibody’s own glycan tree. An engineered galactosyltransferase transfers an azide-modified galactose from a UDP-sugar donor onto the exposed GlcNAc termini of the biantennary structure. The glycan remains essentially intact; the enzyme simply installs an azide at the end of each branch.

This is the minimal intervention. One enzymatic step, no removal, no rebuilding, and the native glycosidic architecture is preserved. The trade-off is that the number of azides installed depends on how many GlcNAc termini are actually exposed, which in turn depends on the glycoform profile of the starting material. The chemistry follows the batch rather than overriding it.

Route two: strip the glycan and build it back

The platform behind IBI343 goes considerably further, and the enzymology is elegant.

EndoS2 is an endoglycosidase that hydrolyses the Fc glycan between the two core GlcNAc residues, leaving a single GlcNAc stub on the protein. A single active-site point mutation converts this hydrolase into a glycosynthase. The mutant can no longer hydrolyse the glycosidic bond it forms, but it retains the ability to catalyse the transfer step, provided it is supplied with an activated donor.

That donor is a disaccharide oxazoline: a strained bicyclic sugar that mimics the oxocarbenium-ion-like transition state of the natural enzymatic reaction. Because the oxazoline is already activated, no nucleotide sugar is required, and the mutant enzyme transfers it onto the GlcNAc stub efficiently.

The practical consequence is a one-pot operation. The heterogeneous native glycan is trimmed away, and a single, defined, azide-tagged disaccharide is installed in its place. One antibody goes in, one glycoform comes out. Heterogeneity is eliminated rather than tolerated.

The same click step finishes both

Once the azide is installed, both platforms converge on the same reaction: strain-promoted azide-alkyne cycloaddition. A DBCO-functionalised linker-payload reacts directly with the azide, with no copper catalyst and no competing reactivity toward the antibody’s cysteines, lysines or histidines.

From a linker chemist’s perspective, the requirements are specific but manageable. The linker-payload must carry a strained alkyne, it must remain stable in the aqueous, near-neutral buffer where the conjugation is run, and it must stay soluble despite DBCO being a distinctly lipophilic group. Both programmes reach DAR 4 with homogeneous product.

The bond matters as much as the site

There is a second advantage that is easy to overlook. In a conventional cysteine-maleimide conjugate, the thiosuccinimide formed by Michael addition can undergo the reverse reaction in circulation, transferring the payload to serum albumin before the ADC ever reaches the tumour. We have written about strategies for preventing that premature release and about the phosphonamidate chemistry used in TUB-040 to avoid the maleimide junction altogether.

Glycan conjugation solves the same problem from a different direction. The antibody-to-linker connection is a native glycosidic bond formed enzymatically, and the triazole produced by the click step is about as chemically inert as anything in organic chemistry. There is no retro-Michael pathway because there is no Michael adduct.

Both platforms are now in late-stage development

This is no longer preclinical chemistry. IBI343 (arcotatug tavatecan), targeting CLDN18.2, met its progression-free survival primary endpoint in an international Phase 3 study in previously treated gastric and gastroesophageal junction adenocarcinoma, and its NDA was accepted for priority review by China’s NMPA in June 2026, making it the first CLDN18.2 ADC to reach regulatory review. JSKN003, a biparatopic HER2-targeting ADC, has received FDA Fast Track and Breakthrough Therapy designations in HER2-expressing platinum-resistant ovarian cancer.

Two late-stage programmes built on enzymatic glycan conjugation is a reasonable signal that the approach has moved from interesting to practical.

What this means for linker-payload design

Choosing glycan conjugation shapes the chemistry upstream of it:

• The linker-payload must carry DBCO or another strained alkyne, which adds a synthetic step and a purification challenge, since DBCO reagents are both hydrophobic and reactive.
• DAR 4 is the natural ceiling for one handle per glycan branch point. Higher loading requires branched architectures of the kind we discussed in our article on dual-payload ADCs.
• Hydrophilicity management does not go away. DBCO is lipophilic, and a PEG or other hydrophilic spacer between the click handle and the payload is usually necessary to keep the construct soluble and well behaved.
• The whole assembly has to survive enzymatic conjugation conditions, which rules out some acid- or base-labile motifs.

How SigutLabs can help

Glycan conjugation shifts the chemical difficulty from the protein to the linker-payload, which is exactly where we work. At SigutLabs we design and synthesise custom ADC linkers and linker-payloads, including azide- and DBCO-ready constructs, hydrophilic spacer architectures and payload derivatives prepared for copper-free click conjugation. If your programme is moving toward site-specific chemistry and you need the linker side built to match, get in touch.

References:

• Innovent Biologics Phase 3 result and NMPA priority review for IBI343 (arcotatug tavatecan), June 2026. Fierce Biotech report
• Phase 3 study of IBI343 in CLDN18.2-positive gastric cancer, NCT06238843