🔬 Fc Engineering for Half-Life & Effector Function

Engineering the CH2-CH3 Interface — Structure-Guided Mutations That Amplify FcRn Recycling

The neonatal Fc receptor (FcRn) is the master regulator of IgG serum persistence: it rescues pinocytosed antibody from lysosomal degradation by binding IgG at the acidic pH of the early endosome (pH~6.0) and releasing it back into circulation at the neutral pH of the blood (pH~7.4). Two structurally adjacent, independently discovered mutation sets — YTE and LS — increase this pH6 affinity at the CH2-CH3 interdomain interface, extending antibody half-life from the typical 21 days of wild-type IgG1 to 60-100+ days in the clinic.

• M252Y/S254T/T256E: YTE mutations (Dall'Acqua et al. 2006)
• M428L/N434S: LS mutations (Zalevsky et al. 2010)
• ~21 days: WT IgG1 half-life (typical human serum t1/2)
• 60–100+ days: YTE/LS clinical half-life (nirsevimab, tixagevimab)

Structural rationale for pH-selective affinity engineering

The FcRn/Fc interaction site: FcRn is a non-covalent heterodimer of a MHC class I-like heavy chain and beta-2-microglobulin. It contacts the IgG Fc at the CH2-CH3 domain interface, with the key histidine residues on FcRn (His166 and others) requiring protonation — and therefore acidic pH — to engage the Fc's His310, His435, and adjacent residues efficiently. At neutral pH, these histidines are deprotonated and the interaction weakens sharply, releasing the antibody.

Why engineer this interface rather than simply increase overall affinity: increasing Fc/FcRn affinity uniformly at both pH values would be counterproductive — high affinity at neutral pH would cause the antibody to remain receptor-bound at the cell surface and in blood, effectively sequestering it rather than releasing it back into free circulation, and could also saturate FcRn capacity, reducing endogenous IgG and albumin recycling (FcRn also recycles serum albumin via a distinct site). The engineering goal is therefore pH-selective: increase pH6 affinity substantially while leaving pH7.4 affinity essentially unchanged (already very weak, Kd typically >10-30 uM for wild-type).

YTE mutations (M252Y/S254T/T256E): identified by alanine-scanning mutagenesis and phage display selection at Medimmune, these three substitutions sit near the CH2 domain FcRn contact surface and introduce additional hydrogen-bonding and electrostatic contacts specifically favorable at pH6 protonation states, without introducing new pH7.4 contacts. First deployed clinically in motavizumab-YTE, later in nirsevimab (RSV prophylaxis, approved 2023).

LS mutations (M428L/N434S): identified independently by Xencor via a similar pH6-selective screen, situated at an overlapping but distinct region of the CH3 domain interface. Often combined with other Fc modifications since the LS site is spatially separable from FcγR-binding and complement-binding surfaces, allowing independent tuning of half-life and effector function within the same molecule. Used in tixagevimab/cilgavimab (Evusheld) among others.

Combining YTE/LS with effector-function mutations: because the FcRn-binding site (CH2-CH3 interface, involving residues ~250-256 and ~428-436) is spatially distinct from the FcγR-binding site (lower hinge-proximal CH2, residues ~234-239 and ~330-332) and the complement C1q-binding site, half-life-extending mutations can in principle be combined with effector-silencing (e.g., LALA-PG) or effector-enhancing (e.g., afucosylation) modifications on the same Fc scaffold without direct mutational conflict — though real designs still require experimental confirmation that the combination doesn't introduce unexpected allosteric coupling.

Measuring pH-Selective Affinity — Surface Plasmon Resonance at pH 6.0 versus pH 7.4

A YTE or LS mutation is only useful if it delivers the intended pH-selective affinity profile — a large affinity gain at pH6 (the endosomal recycling condition) combined with negligible affinity at pH7.4 (the condition required for release back into circulation). Surface plasmon resonance run in parallel at both pH values is the standard biophysical assay for confirming this, and is typically the first hard go/no-go filter applied to any candidate Fc variant before committing to in vivo PK studies.

• 970 nM: WT Fc Kd, pH 6.0 (Biacore, immobilized FcRn)
• 75 nM: YTE Fc Kd, pH 6.0 (~13-fold affinity gain)
• >30 µM: WT/YTE Kd, pH 7.4 (both essentially non-binding)
• >100x: Required pH7.4 fold-weaker (vs pH6, for clean release)

SPR assay design and interpretation of the pH-selectivity ratio

Assay format: recombinant human FcRn (heterodimer of FCGRT ectodomain and beta-2-microglobulin) is immobilized on a Biacore CM5 sensor chip via amine coupling, or captured via an anti-His antibody if expressed with a His-tag. The Fc variant of interest (either as isolated Fc fragment or full IgG) is flowed over the chip as analyte at a range of concentrations (typically 5-7 point, 2-3 fold dilution series), in two entirely separate running-buffer conditions: MES-based buffer at pH6.0 and HEPES-based buffer at pH7.4, matched for ionic strength.

Kinetic fitting: 1:1 Langmuir binding model fit to the association/dissociation sensorgrams at each pH independently, extracting ka, kd, and Kd=kd/ka. At pH6, wild-type IgG1 Fc typically shows Kd in the 700nM-1.7uM range depending on the specific FcRn allotype and assay format; YTE and LS variants typically show 10-15 fold tighter Kd in the 50-100nM range. At pH7.4, both wild-type and engineered variants typically show Kd too weak to fit reliably (>10-30uM, often reported simply as "no measurable binding" at physiological analyte concentrations).

Why the pH7.4 measurement matters as much as the pH6 measurement: a hypothetical Fc mutation that increased pH6 affinity 50-fold but also increased pH7.4 affinity 20-fold would likely perform worse in vivo than wild-type, because retained receptor engagement at the cell surface (neutral pH) promotes non-productive receptor occupancy and can accelerate rather than slow antibody clearance via alternative degradative pathways. The FcRn recycling advantage specifically requires the affinity gain to be confined to the acidic-pH state.

pH-selectivity ratio: defined as Kd(pH7.4)/Kd(pH6.0). Wild-type IgG1 typically shows a ratio in the range of 30-50x (already substantially weaker at neutral pH). A well-engineered half-life-extending mutation preserves or improves this ratio while shrinking the absolute pH6 Kd — the YTE variant characterized here shows a pH-selectivity ratio exceeding 400x (>30,000nM / 75nM), consistent with clean, mechanistically sound engineering rather than a generic affinity-increasing mutation.

A common engineering failure mode is a mutation that increases pH6 affinity but "bleeds through" into increased pH7.4 affinity as well — in early Fc engineering campaigns at several companies, candidate mutations near the FcRn interface were deprioritized specifically because SPR at pH7.4 revealed measurable, non-negligible binding, predicting (correctly, in follow-up PK studies) impaired rather than extended half-life. The two-pH SPR assay is therefore treated as a mandatory paired measurement, never run at pH6 alone.

Tuning Effector Function Independently of Half-Life — Glycan and Charge Engineering at the FcγR Interface

While YTE/LS engineering targets the FcRn-binding CH2-CH3 interface to extend half-life, a separate and spatially distinct region of the Fc — the lower hinge-proximal CH2 domain, together with the N297-linked glycan — governs engagement of activating Fc-gamma receptors, principally FcγRIIIa (CD16a) on NK cells and macrophages, which drives antibody-dependent cellular cytotoxicity (ADCC). This stage tunes that engagement independently, either up (enhanced killing, e.g. for oncology antibodies) or down (silenced effector function, e.g. for pure receptor-blocking antibodies where killing of the target cell is undesired).

• ~85 ng/mL: ADCC EC50, WT fucosylated (NK-cell reporter assay)
• 12 ng/mL: ADCC EC50, afucosylated (~7-fold potency gain)
• 10–50x: FcγRIIIa affinity gain (afucosylation or S239D/I332E)
• obligatory: N297 glycan role (N-linked, required for FcγR/C1q binding)

Two independent levers for effector function: glycoengineering and protein mutation

The N297 glycan and its role: every IgG Fc carries a conserved N-linked glycan at Asn297 in the CH2 domain, buried between the two CH2 domains of the Fc homodimer. This glycan is not merely decorative — it is structurally required to hold the two CH2 domains in the "open" conformation necessary for FcγR and C1q engagement. Enzymatic deglycosylation (e.g., PNGase F treatment) essentially abolishes ADCC, ADCP, and CDC activity, though it does not affect FcRn binding (a spatially distinct site).

Lever 1 — Afucosylation: the glycan's core structure normally carries a fucose residue attached to the innermost GlcNAc. Removing this fucose (afucosylated or "low-fucose" antibody, produced by expression in glycoengineered CHO lines lacking FUT8, the alpha-1,6-fucosyltransferase gene, or in Lec13 CHO mutants) relieves a steric clash between the fucose and FcγRIIIa's own N-glycan, increasing FcγRIIIa binding affinity 10-50 fold and translating to a 7-fold or greater improvement in cell-based ADCC EC50 (85 ng/mL fucosylated to 12 ng/mL afucosylated in this case study). This is the mechanism behind commercial afucosylated antibodies such as obinutuzumab and mogamulizumab.

Lever 2 — Protein engineering (S239D/I332E, "DE" mutations): amino-acid substitutions in the lower hinge/CH2 region directly increase the protein-protein contact surface with FcγRIIIa. S239D/I332E (often combined with A330L as the "DEL" triple mutant) can increase FcγRIIIa affinity by up to 100-fold in some published series and is glycosylation-independent, making it usable even in expression systems that cannot easily be glycoengineered.

Effector silencing (the opposite direction): for applications where Fc effector function is undesired — e.g., an antibody meant only to block a receptor without killing the cell it is bound to — the LALA-PG mutation set (L234A/L235A/P329G) or the N297A/N297Q glycosylation-knockout mutation abolish FcγR and C1q engagement almost completely, while leaving the FcRn-binding site (and therefore half-life, and any YTE/LS enhancement) completely intact, since the two functional sites are structurally independent.

Assay: FcγRIIIa engagement is quantified both by SPR (direct Fc/receptor affinity, accounting for the two common CD16a allotypes, V158 and F158, which differ in baseline affinity) and functionally by a reporter-cell ADCC bioassay (engineered Jurkat effector cells expressing FcγRIIIa and an NFAT-luciferase reporter, co-cultured with antibody-opsonized target cells) — the reporter assay EC50 is the more clinically predictive readout and the one reported in the metrics panel.

Tg32 Humanized FcRn Mice — Translating Binding Affinity Gains into an Actual Half-Life Number

A favorable pH-selective FcRn binding profile by SPR is necessary but not sufficient to predict clinical half-life extension — the relationship between binding affinity and in vivo recycling efficiency is nonlinear and depends on the full endosomal trafficking pathway, not just the isolated binding event. Tg32 mice, in which the endogenous mouse Fcgrt gene is knocked out and replaced with a human FCGRT transgene, provide an in vivo system where human FcRn (not mouse FcRn, which has different Fc affinity and would give a misleading result) governs antibody recycling, making Tg32 PK the standard translational bridge between SPR data and human dosing.

• mFcgrt-/-, hFCGRT+: Tg32 genotype (Roopenian lab-derived model)
• IV, single dose: Dosing route (typically 5-10 mg/kg)
• 21 days: Sampling window (serial serum, ELISA/LC-MS quant)
• ~7–9 days: WT IgG1 Tg32 t1/2 (mouse-scale, not directly = human days)

Why mouse FcRn cannot be used, and how the Tg32 PK study is designed

The species-mismatch problem: mouse FcRn actually binds human IgG Fc with higher affinity than human FcRn does (a long-recognized cross-species quirk), which means wild-type mouse PK studies systematically over-predict human half-life and, worse, fail to discriminate between Fc variants that would behave very differently in a human FcRn context. A YTE mutation optimized for improved binding to human FcRn may show a blunted or even absent PK improvement in a standard wild-type mouse, because mouse FcRn is already saturating the benefit.

Tg32 model construction: Roopenian and colleagues generated a mouse line with the endogenous Fcgrt gene (encoding the mouse FcRn heavy chain) deleted and replaced with a human FCGRT transgene under an appropriate promoter, expressed on the mouse's own beta-2-microglobulin background (which is permissive for human FCGRT heavy-chain pairing). Hemizygous Tg32 animals express human FcRn at levels and tissue distribution reasonably approximating human physiology, making Fc/FcRn interactions in this model governed by the same binding parameters measured by human-FcRn SPR in Stage 2.

Study design: animals (typically n=5-8 per group) receive a single IV bolus dose (5-10 mg/kg) of each Fc variant under comparison — wild-type IgG1, YTE, LS, and any combination variants. Serial retro-orbital or tail-vein blood sampling at time points spanning 30 minutes to 21 days post-dose; serum antibody concentration quantified by generic human-IgG ELISA (Fc-independent capture/detection, so as not to bias for or against the engineered epitope) or by LC-MS/MS for higher precision at low concentrations.

PK modeling: concentration-time data fit to a two-compartment model with first-order elimination, yielding distribution half-life (alpha phase, typically hours, reflecting initial tissue distribution) and terminal elimination half-life (beta phase, typically days, reflecting the FcRn-recycling-dominated slow clearance phase this whole engineering campaign is designed to extend). It is the terminal beta-phase half-life that is reported and compared across variants.

Results in this case study: wild-type IgG1 showed Tg32 terminal half-life of approximately 7-9 days (consistent with published benchmarks for this model, notably shorter in absolute mouse-day terms than the ~21-day human half-life of wild-type IgG1, since Tg32 mice still have faster overall physiological turnover than humans — the model is used for relative, not absolute, half-life comparison). The YTE variant showed an approximately 2.5-3 fold extension relative to wild-type in the same study, consistent with its SPR-measured pH6 affinity gain, and this relative fold-extension is what translates most reliably to the clinical setting.

From Mouse Fold-Change to Human Dosing Interval — PBPK Scaling and Approved-Drug Benchmarks

The final stage of Fc half-life engineering connects preclinical binding and PK data to an actual clinical outcome: how infrequently can this antibody be dosed in patients? Allometric scaling and physiologically-based pharmacokinetic (PBPK) modeling translate the relative half-life extension observed in Tg32 mice into a predicted human terminal half-life, which is then benchmarked against real approved YTE- and LS-engineered antibodies to validate (or correct) the translational model before it informs first-in-human dose selection.

• ~85 days: Predicted human t1/2 (YTE) (PBPK-scaled from Tg32 fold-change)
• ~71 days: Nirsevimab (YTE) clinical t1/2 (approved RSV mAb, benchmark)
• ~90 days: Tixagevimab/LS clinical t1/2 (Evusheld component, benchmark)
• q3–6 months: Dosing interval enabled (vs. q2-4 weeks for WT IgG1)

PBPK scaling methodology and validation against approved antibodies

Scaling approach: rather than naive allometric (body-weight-based) scaling, which works reasonably for small molecules but poorly for FcRn-mediated antibody disposition, half-life engineering programs typically use a fold-change transfer approach: the ratio of engineered-variant half-life to wild-type half-life observed in Tg32 mice (a ratio, not an absolute value, since both numerator and denominator are measured in the same humanized-FcRn biological context) is assumed to transfer reasonably well to humans, and is applied multiplicatively to the well-established wild-type human IgG1 half-life of approximately 21 days.

Worked calculation for this case study: Tg32 mouse data showed YTE half-life approximately 2.7-fold longer than wild-type IgG1 in the same study (Stage 4). Applying this fold-change to the human wild-type reference (21 days) gives a predicted human YTE half-life of approximately 21 x 2.7 ≈ 57-85 days depending on the specific fold-change confidence interval used — consistent with, and validated by, the actual clinical performance of approved YTE antibodies.

Benchmark 1 — Nirsevimab (Beyfortus, YTE-engineered anti-RSV F antibody, approved 2023): observed clinical terminal half-life of approximately 71 days in infant PK studies, enabling a single intramuscular dose to provide protection through an entire RSV season (approximately 5 months) — directly validating that YTE's Tg32-predicted half-life extension translates quantitatively to human infants, a population with immature but functional FcRn-mediated recycling.

Benchmark 2 — Tixagevimab (a component of Evusheld, LS-engineered anti-SARS-CoV-2 antibody): observed clinical terminal half-life of approximately 87-90 days, supporting a dosing interval of every 6 months for pre-exposure prophylaxis in immunocompromised patients — again consistent with LS's preclinical fold-change prediction.

Remaining sources of translational uncertainty: PBPK scaling from Tg32 mice to humans is more reliable for the FcRn-recycling-driven terminal elimination phase than for target-mediated drug disposition (TMDD) effects, which depend on the specific antigen target's expression level, turnover, and whether the antibody is itself being consumed by target engagement — a soluble, low-abundance target produces PK dominated by FcRn recycling (well-predicted by this method), while a highly-expressed, rapidly-internalizing cell-surface target can produce PK dominated by target-mediated clearance (poorly predicted by Fc-engineering-focused Tg32 studies alone, requiring additional target-specific PK/PD modeling).

The consistent 3-4 fold half-life extension achieved by both YTE and LS mutations across dozens of independent antibody programs — from Tg32 mouse fold-change, through PBPK-scaled human prediction, to actual approved-drug clinical PK — represents one of the most reliably translatable pieces of antibody engineering in the field, precisely because it exploits a single, well-characterized, target-independent mechanism (FcRn pH-dependent recycling) rather than a target-specific pharmacological effect. This is why YTE and LS are now considered near-default modular add-ons for any antibody program prioritizing infrequent dosing, from RSV prophylaxis to long-acting HIV and oncology antibodies.