Editing the CHO cell N-glycan pathway — knocking out FUT8 core fucosylation to unlock enhanced FcγRIIIa binding and antibody-dependent cellular cytotoxicity
Every therapeutic antibody produced in Chinese hamster ovary (CHO) cells carries an N-linked glycan attached to Asn297 in the Fc CH2 domain. This single sugar tree, invisible to most potency assays, controls how strongly the antibody engages Fc receptors on immune effector cells — and in wild-type production, nearly all of it arrives decorated with a core fucose that dampens that engagement.
N-glycosylation begins in the endoplasmic reticulum, where a pre-assembled 14-sugar oligosaccharide (Glc3Man9GlcNAc2) is transferred en bloc onto the asparagine of the Asn-X-Ser/Thr sequon by oligosaccharyltransferase. As the antibody transits the Golgi stack, this precursor is progressively trimmed and rebuilt:
• cis-Golgi: glucosidases and mannosidase I trim the structure down to Man5GlcNAc2 • medial-Golgi: GlcNAc transferase I (GnTI) adds a GlcNAc branch; mannosidase II removes two more mannose residues; GlcNAc transferase II (GnTII) adds the second branch • medial-Golgi (critical step): FUT8 (α1,6-fucosyltransferase) transfers fucose from GDP-fucose onto the innermost GlcNAc of the chitobiose core — this must occur before GnTII acts, making the enzyme order a key control point • trans-Golgi: galactosyltransferases and, more rarely, sialyltransferases cap the antennae, producing the final G0F, G1F, or G2F glycoform mixture typically seen in CHO-derived IgG1
The result in unmodified CHO production is a biantennary complex glycan, core-fucosylated in roughly 90–98% of Fc domains, with only a small residual population of naturally afucosylated species.
The Fc region engages FcγRIIIa (CD16) on natural killer cells and macrophages to trigger antibody-dependent cellular cytotoxicity (ADCC). Structural studies (Ferrara et al. 2011; Shields et al. 2002) show that the core fucose on the Fc glycan sits directly at the Fc–FcγRIIIa interface, sterically and energetically interfering with a favorable carbohydrate-carbohydrate contact between the antibody glycan and an N-glycan on FcγRIIIa itself.
Removing that single fucose residue does not change the antibody's amino-acid sequence, its antigen-binding specificity, or its CDR loops — it purely modifies a post-translational sugar. Yet this modification increases FcγRIIIa binding affinity by roughly 10 to 50-fold and can boost ADCC potency severalfold in vitro, because the unobstructed glycan-glycan contact stabilizes a tighter, more productive receptor engagement.
Every gram of a standard IgG1 biologic manufactured in ordinary CHO cells is already ~90–98% core-fucosylated by default — afucosylation is not a mutation to be avoided, it is an engineering target that must be deliberately introduced by editing the host cell glycosylation machinery.
Reducing core fucosylation can be achieved at three different points in the pathway: deleting the enzyme that adds fucose, starving the enzyme of its GDP-fucose donor substrate, or rerouting glycan branching so FUT8 never gets access to its substrate. Each approach trades off potency, robustness, development timeline, and manufacturing complexity differently.
Selecting a defucosylation strategy means weighing durability against speed:
FUT8 gene knockout permanently removes the enzyme from the genome. Because the edit is heritable, afucosylation is stable across cell banking, scale-up, and multi-year manufacturing campaigns — no risk of "leaky" fucosylation returning. This is the approach behind commercial afucosylated platforms such as BioWa/Kyowa Kirin's Potelligent CHOK1SV and its successor PotelliCHO, both used to manufacture mogamulizumab and other approved low-fucose antibodies.
GDP-fucose pathway inhibitors (for example, 2-fluorofucose / 2FF, a cell-permeable fucose analog) block de novo GDP-fucose synthesis without touching the genome. They are fast to deploy — simply supplement production media — but require continuous dosing, add a raw-material cost and regulatory characterization burden, and can leave residual fucosylation if uptake or salvage pathway activity varies between runs.
GnTIII / RMD overexpression takes a different route: it does not remove fucose capacity but instead adds a "bisecting" GlcNAc to the glycan core. This bisecting sugar sterically block FUT8's access to the same core GlcNAc it would otherwise fucosylate, indirectly suppressing fucosylation while also altering antennary branching and galactosylation patterns.
| Product | Indication | Trial Design | Key Result |
|---|---|---|---|
| FUT8 gene knockout (CRISPR) | Enzyme deletion, biallelic | Cas9 double-strand break in FUT8 exon; NHEJ frameshift eliminates the enzyme entirely | Permanent, heritable, >95% afucosylation |
| GDP-fucose analog (2-FF) | Substrate pool depletion | Fluorinated fucose analog blocks de novo GDP-fucose synthesis pathway | No genome edit; fast media-based deployment |
| GnTIII / RMD overexpression | Competing branch enzyme | Bisecting GlcNAc sterically blocks FUT8 access to the glycan core | Tunable expression level, alters branching too |
Turning a strategy into a manufacturing cell line requires editing every relevant allele in the host, then rebuilding a clonal population from a single edited cell — verifying, at each step, that the knockout is complete and that the resulting clone still grows and produces antibody at commercially useful titers.
A guide RNA directs Cas9 to a early exon of FUT8, common to both alleles in the diploid CHO genome. The resulting double-strand break is repaired predominantly by non-homologous end joining (NHEJ), which frequently introduces small insertions or deletions (indels). An indel that shifts the reading frame produces a truncated, non-functional transcript — silencing that allele's fucosyltransferase activity.
Because NHEJ is stochastic, a single transfection produces a mixed population: some cells with no edit, some heterozygous (one allele disrupted), and some biallelic knockouts. Only biallelic knockout eliminates FUT8 activity completely, so the editing step is always followed by single-cell cloning and genotypic screening (Sanger sequencing or next-generation amplicon sequencing across the cut site) to identify true null clones.
Limiting-dilution or single-cell printing generates hundreds to thousands of clonal colonies from the edited pool. Each candidate clone must clear several bars before advancing:
• Genotype confirmation: biallelic frameshift at the FUT8 locus, no detectable wild-type allele • Phenotype confirmation: glycan mapping of secreted antibody shows the expected afucosylated profile • Growth and productivity: viable cell density, doubling time, and specific productivity (pg/cell/day) must remain compatible with fed-batch production economics
Higher selection stringency — tighter cutoffs on titer, growth rate, and genotypic purity — yields a cleaner, more reliable production clone but shrinks the pool of clones that pass every gate, which is why aggressive editing and stringent selection can trade off against final manufacturing titer. Balancing knockout completeness against clone productivity is the central optimization of this stage.
Kyowa Kirin's Potelligent CHOK1SV platform was one of the first to demonstrate that a FUT8 double-knockout host line could still support commercial-scale fed-batch titers — decoupling the long-held assumption that glycoengineering necessarily sacrifices productivity.
Confirming that a cell line edit translated into a real change in the secreted product requires releasing the N-glycans from purified antibody and separating them analytically. The resulting glycan map is the definitive readout of engineering success: a quantitative shift of peak area away from fucosylated glycoforms and toward their afucosylated counterparts.
Glycan analysis starts with purified antibody (post-Protein A capture). PNGase F, an amidase, cleaves the entire N-glycan intact from the Asn297 side chain, converting the asparagine to aspartate and releasing a free oligosaccharide into solution. The released glycans are then labeled at their reducing end with a fluorophore such as 2-aminobenzamide (2-AB), enabling sensitive fluorescence detection downstream.
An orthogonal, faster route skips release entirely: the intact antibody (or a glycopeptide generated by protease digestion, typically trypsin) is analyzed directly by liquid chromatography coupled to high-resolution mass spectrometry (LC-MS), reading glycoform mass differences straight off the intact protein or glycopeptide envelope. The mass difference between a fucosylated and afucosylated glycoform (loss of one deoxyhexose, −146 Da) is easily resolved by modern Orbitrap or Q-TOF instruments.
HILIC (hydrophilic interaction liquid chromatography) separates labeled glycans primarily by hydrophilicity and glycan size, producing a characteristic multi-peak trace. In wild-type CHO product, the chromatogram is dominated by fucosylated peaks — G0F (agalactosylated, core-fucosylated), G1F (mono-galactosylated), and G2F (di-galactosylated) — collectively representing the vast majority of total peak area.
In a successfully engineered FUT8-knockout clone, those same peaks collapse and new peaks emerge at the retention times corresponding to the non-fucosylated equivalents — G0, G1, and G2 — which shift slightly earlier in the HILIC separation because they lack the additional fucose monosaccharide. Integrating peak areas and expressing the afucosylated fraction as a percentage of total glycan is the standard potency-relevant release specification tracked from process development through commercial batch release.
The glycan map predicts function, but only a cell-based bioassay proves it. Co-culturing the engineered antibody, antigen-positive target cells, and FcγRIIIa-expressing NK effector cells directly measures whether the afucosylation engineering translated into the intended increase in immune-mediated killing.
A standard ADCC bioassay combines three components: antigen-positive target cells, the test antibody at a concentration series, and effector cells expressing FcγRIIIa — either primary NK cells or an engineered reporter cell line (e.g., Jurkat-NFAT-luciferase cells stably expressing FcγRIIIa-V158).
When the Fab arms of the antibody bind target-cell surface antigen, its Fc domain becomes available to engage FcγRIIIa on the effector cell. Afucosylated Fc glycans form a tighter, unobstructed carbohydrate-carbohydrate contact with the FcγRIIIa glycan, stabilizing the Fc-receptor complex. This directly increases the avidity of the immune synapse, lowers the antibody concentration required for half-maximal receptor engagement (EC50), and — for primary NK cell assays — increases downstream signaling, degranulation, and release of cytotoxic granules (perforin, granzyme B) that lyse the target cell.
The functional consequence of afucosylation is measured as a leftward shift and an upward shift of the ADCC dose-response curve: lower EC50 (more potent at low antibody concentration) and often higher maximal killing (Emax) compared to the fucosylated parent molecule.
This translates clinically into two related benefits: enhanced-potency antibodies that can achieve equivalent or superior efficacy at a lower administered dose, and stronger effector engagement in patients whose own FcγRIIIa is a lower-affinity allelic variant (F158 rather than V158) — a genetic factor known to blunt ADCC for standard fucosylated therapeutics. Approved and clinical-stage afucosylated antibodies, including mogamulizumab (anti-CCR4) and obinutuzumab (anti-CD20, GlycArt/Roche glycoengineering platform), were developed specifically to exploit this mechanism for hematologic and oncologic indications.
Obinutuzumab, engineered for reduced core fucosylation via the GlycoMAb platform, demonstrated superior B-cell depletion and clinical efficacy compared to the fucosylated CD20 antibody rituximab in head-to-head trials — one of the clearest clinical demonstrations that glycan engineering alone, without any change to antigen specificity, can meaningfully improve a therapeutic antibody.