Impact of antibody Fc engineering on translational pharmacology, and safety: insights from industry case studies

ABSTRACT

Therapeutic monoclonal antibodies (mAbs) are often designed to not only bind targets via their antigen-binding domains (Fabs) but to also engage with cell surface receptors, FcγRs and FcRn, through their Fc regions, which may result in a variety of functional outcomes, including antibody- dependent cellular cytotoxicity (ADCC), antibody-dependent cellular phagocytosis (ADCP), complement-dependent cytotoxicity (CDC) and alteration of circulating half-lives. Engineering the Fc regions to achieve desirable pharmacology and pharmacokinetics is a widely adopted strategy in drug development. Fc regions can be modified through amino acid substitutions and glycoengineering, resulting in enhanced or reduced effector functions, preferential binding to FcR subtypes, or pH-dependent binding to FcRns. These alterations in binding and effector activities of mAbs may potentially also be accompanied by undesirable effects or safety concerns. Critical assessment of pharmacology and safety in the nonclinical setting is essential before exposing humans to the engineered mAb. For Fc-modified mAbs, the choice of in vitro and in vivo nonclinical pharmacology and safety models need to account for species differences in FcR expression and function, potentially divergent effects of Fc modifications in humans versus nonclinical species, impact of target and cognate ligand expression patterns, and potential impact of emergent anti-drug antibodies directed against the mAb. Using a variety of industry case studies, we highlight key aspects of nonclinical pharmacology and toxicology testing strategies, factors that influence choice of nonclinical models, translatability of findings, input from health authorities and suggest best practice approaches for nonclinical testing of Fc modified mAbs.

1. Introduction

The crystallizable fragment (Fc) region of therapeutic monoclonal antibodies (mAbs) of the IgG isotype is often engineered to optimize their pharmacological and pharmacokinetic (PK) properties. Accordingly, safety characteristics have become an increased focus of therapeutic mAb development. Enhancing specific effector functions can be achieved by modifications that increase engagement with activating FcγRs or the complement component 1q (C1q). Fc-engineered IgG mAbs can have enhanced Antibody-Dependent Cell-mediated Cytotoxicity (ADCC), Antibody-Dependent Cell-mediated Phagocytosis (ADCP), Complement-Dependent Cytotoxicity (CDC), or Complement-Dependent Cellular Cytotoxicity (CDCC) of target cells. In addition, receptor clustering-based agonism can be achieved through amino acid substitutions that potentiate Fc–Fc hexamerization and FcγRIIB interactions. Alternatively, Fc domains have also been (partially) silenced by reducing FcγR and C1q binding. Fc domains have been multimerized to block Fc gamma receptors (FcγRs) and neonatal Fc receptors (FcRn). Furthermore, Fc domains have also been modified to enhance or decrease binding to FcRn, to fine-tune IgG half-life. Due to this complexity in engineering strategies, there is an increased need for informed non-clinical pharmacology and safety assessment of such modified mAbs due to their altered receptor binding affinity in the most used pharmacology and toxicology species, namely cynomolgus monkeys and rodents. This is important to minimize the risk of unexpected or unwanted events in human clinical trials.

Here, we introduce key considerations and present 15 industry case studies that illustrate nonclinical pharmacology and toxicology testing approaches, in vitro and in vivo model selection, and risk assessment considerations for the most common IgG modifications that result in altered effector function and half-life, namely FcR and C1q binding enhancement and ablation. The clinical translatability and predictivity of non-clinical findings and learnings from health authority (HA) interactions are shared where available. As such, this review is intended to provide practical information and informal recommendations for pharmacological and toxicological testing of Fc-modified drugs, while highlighting remaining gaps and suggesting future directions for experts in this expanding field.

2. FcR expression and function in humans and nonclinical species

2.1. FcγR expression and function in humans

Functionally, human FcγRs can be divided into activating and inhibitory receptors, which transmit signals through intracellular immunoreceptor tyrosine-based activation motifs (ITAMs) or immunoreceptor tyrosine-based inhibitory motifs (ITIMs), respectively, to affect cellular function after cross-binding to target-bound IgG antibodies.¹ In humans, hFcγRI (CD64), hFcγRIIA (CD32a), hFcγRIIC (CD32c), and hFcγRIIIA (CD16a) are activating receptors, and FcγRIIB (CD32b) is the sole inhibitory receptor (Table 1). Most FcγRs have a low affinity for the IgG1 Fc (Table 2), but form strong, functional interactions when multiple antibodies cluster into immune complexes. This clustering enables cross-linking of FcγRs, which is essential for triggering downstream immune signaling and function. The exception is the high-affinity receptor FcγRI, which can bind monomeric and non-aggregated IgG. The FcγRIIIB (CD16b) is unique in that it lacks transmembrane and cytoplasmic parts and, thus, direct signaling activity but may remain tethered to the cell membrane via a glycophosphatidylinositol (GPI) anchor. However, its high-level expression on numerate neutrophils, sequence similarity to FcγRIIIA, and absence from the mouse and most non-human primates (NHP), demand thoughtful consideration of its potential functions.⁵

Table 1.

Key features of human Fc gamma receptors.

	FcγRI (CD64)	FcγRIIA (CD32A)	FcγRIIB (CD32B)	FcγRIIC (CD32C)	FcγRIIIA (CD16A)	FcγRIIIB (CD16B)
Receptor
Key Features	Activating. Likely saturated by endogenous IgG; may have role in antigen processing, cytokine release, and anti-tumor immunity, but minor role in overall effector function of mAbs	Activating. Central role in phagocytosis of Ab-opsonized antigens. Two major allelic variants in humans: R131/H131. R131 has decreased IgG binding especially IgG2 whilst H131 has enhanced binding to multiple subclasses including IgG2. A notable receptor on platelets and neutrophils.	Single inhibitory receptor in humans (and other species). Highly expressed by phagocytes, sole FcγR on B cells. Expressed by liver sinusoidal endothelial cells for recognition and uptake of ICs. All major human variants of FcγRIIB have identical extracellular domains with R131, making them similar to the low affinity FcγRIIA variant. Two splice variants of FcγRIIB exist that differ in the expression of a cytoplasmic region that affects internalization.	Formed by a genetic crossover between the activating cytoplasmic domain of FcγRIIA and ECD of FcγRIIB. Expressed on subsets of human B cells and NK cells. Can trigger ADCC in NK cells.	Activating. Drives ADCC by NK cells but also expressed on other cells. Humans have a V158 SNP resulting in increased IgG1, IgG3, and to a lesser extent IgG4 binding compared to the F158 variant. V158 is less frequent and promotes enhanced ADCC and anti-tumor effects of mAb therapy.	Activating. Uniquely expressed on human granulocytes. ECD closely homologous to FcγRIIIA. No signaling domain but associated syk signaling. Complement receptors and FcγRIIA may provide accessory signaling. FcγRIIA may synergize with FcγRIIIB to increase the flux of Ca2+ into PMNs and promote release of cytotoxic granules and phagocytosis. FcγRIIIb may be a decoy that competes with activating receptors for ICs. Increased affinity of IgG for FcγRIIIB results in decreased neutrophil ADCC
Expression	Macrophage, monocyte, dendritic cell, mast cell, neutrophil**	Macrophage, neutrophil, platelet, mast cell, eosinophil	B cell, macrophage, dendritic cell, basophil, monocyte*, neutrophil*	B cell*, NK cell*	NK cell, monocyte*, macrophage*, CD4+ T cell*, basophil*, mast cell*	Neutrophil, basophil*, eosinophil**

*Expressed on a subset. **induced expression. ECD, Extracellular domain; SNP, single nucleotide polymorphism.

Humans have four IgG subclasses that differ in their binding affinity for the various FcγRs and ability to mediate effector functions (Table 2). IgG1 and IgG3 bind strongly to most FcγRs and fix the C1q component of complement, whereas IgG4 and IgG2 show greatly reduced FcγR binding (IgG4 does not bind C1q) and correspondingly lower levels of effector function. Therapeutic antibodies are commonly derived from IgG1, 2, or 4 depending on the effector function desired, but not typically IgG3 due to manufacturing challenges and shorter circulating plasma half-life caused by altered binding to FcRn, although the latter phenotype is allotype-dependent.²

Table 2.

Binding affinities of IgG subclasses to FcRn, FcγRs, and complement C1q.

Pharmacological Response	Subclass	IgG1	IgG2	IgG3	IgG4
Pharmacokinetics*	Serum half-life (days)	21	20	7	21
	FcRn binding affinity (nM) at pH 5.8 and 37°C	998	768	898#	973
Immune Activating**	FcγRI binding affinity (nM)	15	NB	16	29
	FcγRIIA binding affinity (nM)	192	2222	1124	5882
	FcγRIIC binding affinity (nM)	286	10000	1099	4762
	FcγRIIIA binding affinity (nM)	854 (F158) 500 (V158)	33333 (F158) 14286 (V158)	130 (F158) 102 (V158)	5000 (F158) 4000 (V158)
	FcγRIIIB binding affinity (nM)	4762	NB	1000	NB
Immune Inhibitory	FcγRIIB binding affinity (nM)	286	10000	1099	4762
	C1q binding affinity	++	+	+++	-

*Abdiche et al.³ **Bruhns P et al. ⁶. NB: non-binding. ##allotype-dependent: IgG3-R435 (most prevalent in Caucasian populations) binds weaker at pH 6.0 and has some residual FcRn binding at pH 7.4, this is different for IgG3-H435⁴ Determination of affinity values of Fc receptor binding strength to IgG strongly depends on the experimental method. Therefore, these values are a guide.

FcγRs are expressed on all subsets of human immune cells and platelets (Table 1). However, mouse immune cells express different receptors to humans and have no FcγR on their platelets. NHP FcγRs match human sequence and expression profile more closely, but their neutrophils do not express an FcγRIIIB equivalent.⁷ These key differences regarding expression profiles together with distinct-binding properties of engineered IgG to FcγRs across species are the source of most challenges in the pre-clinical study of Fc-engineered IgG, as discussed hereinafter. It is also important to note that some cells express only one receptor, making them liable to act in a polarized manner (e.g. human B cells and the inhibitory FcγRIIB; most human natural killer (NK) cells express only the activating FcγRIIIA). Other cells express multiple receptors, making dynamic competition for activating vs inhibitory receptor signaling based on size and isotype of immune complex a feature. For example, human macrophages and dendritic cells express activating and inhibitory FcγRs in various combinations. FcγR expression is also dynamic as their expression can be increased by IFN-γ, TNF, IL-4, IL-10, or IL-13, for example, IFN-γ potently and differentially induces expression of the activating FcγRIIa versus the inhibitory FcγRIIb on immature dendritic cells (DCs)^8–10 and some are also internalized upon interaction with ICs. For a comprehensive overviews of FcγR expression and function, refer to Hogarth & Pietersz;¹¹ Bournazos et al.¹² and DiLillo & Ravetch.¹³

The key effector functions of IgG are shown in Figure 1. These are comprehensively reviewed elsewhere.^14,15 To summarize key points, IgG/IC engagement with:

• FcγRIIIA on NK cells promotes ADCC-mediated target cell killing.

• FcγRIIA and FcγRIIIA on macrophages and FcγRIIIB on neutrophils can promote ADCP and subsequent killing of antibody-bound targets

• FcγR engagement (e.g., FcγRIIB) can also enhance the activation of target effector cells by promoting receptor cross-linking/agonism by the Fab arms of the IgG.¹⁶ “Super cross-linking” of IgG by FcγRIIB on one cell can generate signals in target cells, e.g., apoptosis or activation. Cross-linking of FcγRIIB by IgG can inhibit B cell activation (Karnell et al., 2014). FcγRIIB can be engaged for ‘sweeping’ or internalization of small Immune Complexes by phagocytes and liver sinusoidal endothelial cells.¹⁷

• FcγRIIA on platelets or neutrophils can result in secretion of soluble mediators, clotting or netosis (Anania et al., 2019).

• FcγRIIIB is thought to play a role in clearance of circulating immune complexes without producing neutrophil activation. Under certain conditions, however, hFcγRIIIB can cooperate with other signaling molecules, such as hFcγRIIA, and contribute to IC-mediated neutrophil activation (Wang and Jonsson, 2019). The high-level expression of hFcγRIIIB on human neutrophils, sequence similarity to hFcγRIIIa, and absence from mouse and most NHP demand thoughtful consideration of its potential functions.

Figure 1.

Key Fc-mediated effector functions of IgG. (a) Fc binding to activating FcγRs on NK cells and macrophages to promote their activation and mediation of antibody-dependent cell cytotoxicity (ADCC) and antibody-dependent cell phagocytosis (ADCP), respectively. Fc binding to C1q to mediate complement activation and direct complement-mediated cell cytotoxicity (CDC), and via C3b generation, opsonizing of target cells for complement-mediated cell cytotoxicity (CDCC) via the C3b receptor (C3bR) on macrophages and neutrophils. (b) Binding of the Fc within IgG-antigen immune complexes to the inhibitory FcγRIIB on B cells, NK cells and macrophages to inhibit their activation via the B cell receptor (BCR) or activating FcyRs, respectively. (c) Binding of the Fc to FcγRIIB on B cells which acts as a scaffold to enhance cross-linking of Fab-bound receptors on T cells and other cells leading to target-cell activation or apoptosis depending on the receptor type, (d) Binding of the Fc within IgG-antigen immune complexes to FcγRIIB on antigen-presenting cells (APCs) to promote the internalization, clearance, and presentation to T cells of Fab-bound antigen.

The Complement System “complements” humoral and cellular immunity for functions including: direct defense against pathogens, promoting inflammation, and removal of immune complexes through classical, lectin, or alternative pathways (reviewed in Monach).¹⁸ IgGs can bind C1q to activate the classical complement pathway and drive CDC,¹⁹ with differing activity across IgG subclasses (IgG1 and IgG3 > IgG4).^20,21 Complement activation also leads to the generation of C3b and the anaphylatoxins C3a and C5a that opsonize target cells for phagocytosis by C3bR, C3aR and C5aR-bearing macrophages and neutrophils.

2.2. FcγR expression and function in nonclinical animal species

It is critical to consider the translational effects of any differences between animal and human expression and function. FcγR biology in the commonly used species for these assessments, NHP and rodent, including differences from human FcγRs and their interaction with human IgGs, has been previously reviewed in detail.^22–26

2.2.1. Non-human primate

Cynomolgus monkey (cyno) and rhesus FcγRI, FcγRIIIA and FcγRIIA all show qualitatively similar expression, function, and binding patterns with human, cyno and rhesus IgG1–IgG4.^22,26 For example, FcγRIIA is expressed on cynomolgus monkey platelets, so human platelet activation risk via FcγRIIA engagement can generally be assessed in monkeys. However, there are some differences, as summarized in Supplementary Table 1. Notably, FcγRIIC is not present in macaques, and macaque granulocytes do not express FcγRIIIB, but express higher levels of FcγRII than human cells. While human IgG1 and IgG3 show greater binding to both cynomolgus monkey and human FcγRs than human IgG2 and IgG4, human IgG2 shows increased binding to cynomolgus monkey FcγRIIB as compared to the human equivalent.²⁶ Human IgG1 and IgG3 mediate similar high ADCC activity (via NK cell FcγRIIIA) with human or cyno peripheral blood mononuclear cells (PBMCs), but human IgG4 is slightly more active on cyno PBMCs compared to human PBMCs.²⁶ While unmodified human therapeutic mAbs have been found to have a broadly similar effector function profile in humans and cynomolgus monkeys, Fc modification could lead to species differences in effector properties, as in the case of afucosylated human IgG1 exhibiting enhanced ADCC with cyno FcγRIIIA.²⁷

2.2.2. Rodents

Rodent FcγRs show important differences in FcγR expression and function compared to humans and NHPs (summarized in Figure 4), complicating the use of mouse models. The key difference is that mice do not express FcγRIIIB, so their granulocytes (neutrophils) lack it. Mouse NK cells express mFcγRIII, (more equivalent to huFcγRIIA) whilst human NK cells express hFcγRIIIA (equivalent to mouse FcγRIV).²⁸ Mouse FcγRIV on macrophages is the key mediator of ADCC (mainly from FcγRIIIA on human NK cells). Indeed, monocytes or macrophages are the main immune cell responsible for ADCC/ADCP in the mouse. Critically, mouse platelets do not express FcγR (FcγRIIA in humans).

Human IgGs bind to mouse FcγRs with remarkably similar binding strengths as to human ortholog receptors, with relative affinities IgG3 > IgG1 > IgG4 > IgG2 and FcγRI >> FcγRIV > FcγRIII > FcγRIIB²⁹. Mice may underestimate the activity of human IgG1 for ADCC activity. Human IgG1 is the most potent human isotype at inducing NK cell-mediated ADCC of target cells with mouse effector cells (via mouse FcγRIII), but it is not as potent as mouse IgG2a. Hence, in mice, afucosylated human IgG1 exhibits enhanced ADCC with mouse FcγRIII and FcγRIV vs hFcγRIIIA.²⁷ Human IgG1 is capable of ADCP in mice, as long as they possess macrophages³⁰. Human IgG2 mAbs activate a different set of effector cells in mice and humans, and human IgG4 potently activates mouse macrophages (whereas it is more inert in humans).²⁸ Mouse can exhibit an IgG-driven anaphylactoid response (mediated by platelet activating factor (PAF) and neutrophils)³¹ that is not or rarely observed in humans. Since there is no FcγR expression on mouse platelets,³² any potential risk of platelet activation cannot be assessed in standard mice, unless they are transgenic for this purpose (see Human FcγR Transgenic and Immune System Humanized Mice, section 4.2; Case Studies 8 and 9).

As regards the comparative biology of complement, all species have the same general components,³³ but may differ in their activity. Mice have been observed to elicit weak CDC, the reasons for which are not well understood, although protective CDC-mediated effects have been observed with Fc-modified hexameric IgGs.³⁴

2.3. FcRn expression and function

FcRn is expressed in immune and nonimmune tissues and is critically important for recycling and transcellular transport of IgG).^35,36 FcRn is a heterodimer of an MHC-class-I-like heavy chain and β2-microglobulin light chain; however, the receptor does not contain or associate with domains capable of signal transduction or peptide presentation.³⁷ The critical importance of FcRn for IgG half-life was shown in FcRn-deficient mice, where mIgG half-life is reduced >6-fold, from 9 to 1.4 days, an effect overcome by the introduction of a human FcRn transgene (discussed below in Human FcRn Transgenic Mice, section 4.2). FcRn is predominantly present within endosomes with little (~10%) plasma membrane expression.³⁸ As a result, the interaction between Fc regions of IgG and FcRn occurs at mildly acidic pH (as within late endosomes), mediated via critical Fc histidines. In the acidic environment of the endosomes, tight binding between FcRn and IgG results in recycling of IgG back to the circulation where IgG is released at physiological pH (Figure 2). Vascular endothelial and bone marrow-derived cells, such as monocytes, are key contributors to IgG half-life.^39,40 FcRn has also been shown to actively transfer IgGs across the placenta during pregnancy, allowing transport of maternal IgG from mother to unborn, a process which gave rise to the neonatal Fc receptor name. Finally, the liver hepatocytes, Kupffer, and sinusoidal epithelial cells have all been shown to express FcRn. For an in-depth review of FcRn biology and function, refer to Pyzik et al.³⁶

Figure 2.

Recycling of IgG by FcRn general schematic of FcRn-mediated recycling of IgG antibodies. Circulating plasma proteins (e.g., antibodies, albumin, cytokines) can undergo fluid phase endocytosis into endosomes. Upon entry into, maturation, and acidification of endosomes, IgG antibodies (red) may bind to FcRn in a pH-dependent manner. Once formed, the FcRn-IgG complex may be recycled to the plasma membrane, where exposure to the neutral pH in circulation facilitates dissociation. Antibodies and serum proteins that are unbound to FcRn or simply cannot bind to FcRn are targeted for lysosomal degradation, where catabolizing enzymes will break down proteins into peptides and amino acids.

Human IgG binds more strongly throughout the endosomal pH gradient to human and mouse FcRn than mouse IgG, with an important impact on pharmacology and PK assessments in both wild-type and human FcRn transgenic mice. Wild-type human IgG is rescued in conventional mice; however, when Fc-engineering is added to achieve extended plasma half-life, pH-dependent binding to mouse FcRn is compromised, and as such results in shorter half-life, in contrast to human FcRn transgenic mice.^41,42 In addition, as human FcRn does not bind mouse IgGs, the levels of mouse IgG are very low in human FcRn transgenic mice.^43,44 This fact should thus be considered when evaluating IgG mAb formats in such transgenic mouse models, as natural competition for FcRn binding will not be present unless they are loaded with pooled human IgG, such as IVIG, prior to the half-life study. A detailed analysis of various species and cross-species binding was conducted by.³ Human IgG1 binds cynomolgus monkey FcRn with a 2-fold higher affinity than human FcRn and binds both mouse and rat FcRn with a 10-fold higher affinity than human FcRn.

Another key feature of FcRn is that it also binds and recycles albumin via an independent and non-overlapping binding site for that of IgG,^45,46 and as such, extends the plasma half-life of not only IgG but also albumin.^45,47 However, the stoichiometry differs, as while one FcRn binds to one albumin, two FcRn molecules can bind to one Fc and in a so-called ‘upside down’ manner, orienting the Fabs proximal to the membrane. ⁴⁸ The latter has consequences for the cellular handling properties of Fc-containing molecules.^49,50

2.4. Unwanted effects of Fc-effector function in humans

Exaggerated FcγR and complement activation can have potential safety risks for Fc-bearing therapeutics (reviewed by Brennan & Kiessling),⁵¹ as shown in Figure 3. Unwanted activation of FcγR-bearing cells driving ADCC, ADCP and C1q/CDC-mediated killing of target cells can result in inappropriate immune system activation and cytokine release (including tumor lysis syndrome and other sequelae), as observed with OKT3, rituximab, obinutuzumab and alemtuzumab.⁵² Unwanted cross-linking and activation of FcγR-bearing immune cells can also lead to activation-related apoptosis, margination, proliferation, inhibition, or internalization, depending on the biology of the target. Cross-linking of multivalent soluble molecules in circulation may lead to soluble immune complex formation, triggering inflammatory reactions. The binding to FcγRIIA on platelets may promote their activation and potentially thrombotic events, as observed with an anti-CD40L mAb in animals and humans.⁵³ Complement activation can generate C3a and C5a (anaphylatoxins) to induce the undesirable activation of mast cells, basophils and endothelial cells bearing C3aR/C5aR.⁵⁴

Figure 3.

Safety-relevant aspects of enhanced Fc-effector function created in https://biorender.com (A) IgG structure indicating locations of critical immune receptor / ligand interactions. (B) Complement related killing and immune cell activation. The classical complement pathway kills cells directly after engagement on clustered IgG (CDC), effected by membrane attack complex (MAC) or by opsonisation of target cells with complement ‘b’ fragments (or opsonins), which are recognised by phagocytic effector cells (ADCP). Complement ‘a’ fragments (or anaphylotoxins) are highly potent effectors of immune cell infiltration and activation of immune and endothelial cells with potential for further cytokine release. (C) IgG binding to FcRn in endothelial, macrophages and other cells prolongs IgG half-life and hence exposure to its pharmacological activity potentially increasing the risk of pharmacology-associated toxicity. (D). Receptor clustering driven directly by Ab format engineering or facilitated in trans by CD32b on a neighbouring cell can lead to receptor signalling and hence cell activation or inhibition, depending on receptor biology and cellular context. (E) Cellular killing mechanisms effected through FcγRIIIa on NK cells (ADCC) or primarily FcγRIIa on phagocytic cells (ADCP). NK cells secrete local effectors such as perforin, IFNγ, granzyme B to kill cells directly. Phagocytic cells engage and kill target cells, via IgG and / or opsonins to directly ‘eat’ target cells or strip off sections of membrane ‘trogocytosis’. (F) IgG aggregates or target antigen related immune-complexes can activate FcγR resulting in ‘off-target’ / ‘off biology’ related immune mediator / cytokine release toxicology.

3. Enhancing or ablating antibody Fc function by protein and glyco-engineering

Many of the effector functions of IgG Fc outlined in section 2.2 contribute to the therapeutic efficacy of therapeutic mAbs and related Fc-containing scaffolds. For Fc-based therapeutics where a strong effector function is required for efficacy, the IgG1 isotype is typically used. The effector function and half-life can be further enhanced or reduced by antibody engineering depending on the clinical application. However, when targeting inflammatory diseases, it may be undesirable to have Fc-mediated activation of immune cells, complement activation and cytokine release. Unless cell depletion is a desired pharmacologic effect, mAbs that bind to cellular receptors, such as, to activate NK or T cells for cancer therapy or to inhibit the function of cells involved in inflammatory immune responses, must be designed to avoid Fc-mediated effects. Non-activating IgGs are typically generated using the more inert IgG4 or IgG2 wild-types or proactively through Fc silencing of IgG1/4, as described below.

The principal focus areas of Fc engineering efforts are Fc-effector enhancement, Fc-effector ablation, half-life modulation (most typically extension), and direct FcγR and FcRn blockade. Enhancement of both FcγR and FcRn engagement and silencing of Fc-effector function of therapeutic mAbs by protein and glyco-engineering have been comprehensively reviewed elsewhere.^13,55 The use of Fc fragments and Fc multimers to block FcRn and FcγR engagement, resulting in enhanced clearance and blockage of pathogenic FcγR-mediated effectors of autoantibodies, has also been reviewed.^56,57 Hence, these are only briefly summarized below and are further explored in the industry case studies for each focus area in section 6. Fc-engineered mAbs that are approved or in late-stage clinical trials are summarized in Table 3.

Table 3.

Summary of case studies.

Case No.	Fc modification	MoA	Indication	Key features
1	Fc-enhanced (afucosylated); 2 different molecules	Deplete target- expressing cells	Cancer and Autoimmunity	Non-clinical safety assessment; rat and cyno tox. studies; pro-inflammatory AEs (decreased neutrophils) observed with 1 of the mAbs.
2	Fc-enhanced (afucosylated)	Deplete B cell tumors	Cancer	Used huFγR Tg mice to assess PD & safety; AEs (thrombosis) observed related to higher FcγRIIA tg mouse platelets than on human platelets.
3	Fc-enhanced (triple mutant)	Deplete tumor cells	Cancer	Non-clinical safety assessment; unexpected toxicity & immunogenicity due to promiscuous FcγR binding.
4	Fc-enhanced (afucosylated)	Deplete eosiniphils	Asthma	Non-clinical safety assessment; cyno tox. & reprotox studies; prolonged but reversible target cell depletion.
5	Fc-enhanced (afucosylated)	Deplete B & TFH cells	Autoimmunity	Non-clinical safety assessment; cyno tox. studies; prolonged but reversible target cell depletion.
6	Fc-silenced (LALA, LALA-PG, STR)	Bind model antigens on lymphocytes	Model system only	Projection and evaluation of human PK parameters using cyno and FcRn transgenic mice. Functional characterization (binding to huFcγRs & C1q, in silico & vitro immunogenicity assessment, cytokine release).
7	Fc-silenced (DAPA)	Binds model antigen on target cells	Model system only	Developability assessment and functional characterization (binding to FcγR, FcRn, C1q); PK assessment in FcRn Tg mice.
8	Fc-silenced (N297A, D265A, LALA, IgG4- S228P, novel mutants)	Target on immune cells (not disclosed)	Autoimmunity	Functional characterization & safety assessment (target cross-linking/agonism risk) in vitro and in vivo in huFγR Tg mice.
9	Fc-silenced (N297A; aglycosylated)	Inhibits T & B cell responses	Autoimmunity	Functional characterization (lack of ADCC, CDC. agonism) & safety assessment (platelet activation risk) of an Fc-silenced mAb in vitro and in vivo in NHPs and in huFcγR Tg mice.
10	Fc-silenced (LALA)	Target on immune cells (not disclosed)	Autoimmunity	Platelet activation assessment in vitro comparing human and cyno
11	Fc-silenced (DAPA, LALA, others)	Multiple mAbs binding to immune cells	Autoimmunity	HA feedback and responses relating to multiple silenced mAbs targeting both membrane and soluble targets.
12	HLE (enhanced FcRn binding)(N434H)	Prolonged half-life	Autoimmunity	Half-life extension in NHPs not observed in human patients.
13	HLE, sweeping (enhanced FcRn & FcγRIIB binding) (mutations not disclosed)	Prolonged half-life, removal of soluble antigen	Autoimmunity	Used cyno surrogate for PK/PD and safety assessment.
14	Fc multimer	Blocks FcγRs	Autoimmunity	In vitro immunosafety assessment with human cells (FcγR binding; assays for immune cell and platelet activation, cytokine release)
15	Fc fragments, HLE	Blocks FcRn	Autoimmunity	Non-clinical (cyno) and clinical PK/PD & safety assessment.

3.1. Fc modifications for enhanced binding and function

Most common engineering efforts aim to increase binding to activating FcγRIIIA and FcγRIIA to promote enhanced ADCC- and ADCP-mediated target cell killing. Other efforts aim to selectively increase binding to FcγRIIB to enhance receptor clustering agonistic mAbs, drive inhibitory B cell signaling, or contribute to target antigen ‘sweeping’.

3.1.1. Glycoengineering for enhancement of Fc effector function

IgG antibodies have a conserved N-glycosylation site at Asn297 of the Fc CH2 domain that bears complex- and heterogeneous-type N-glycans, which fine-tune IgG effector function, complement activation, and half-life. Fucosylation, galactosylation, sialylation, bisection and mannosylation all generate glycoforms that interact in a specific manner with different cellular antibody receptors and are linked to a distinct functional profile (reviewed by Garcia-Alija et al.).⁵⁸ N-glycosylation contributes to Fc structure, critically enabling engagement with FcγRs^59,60 and C1q. The absence of the CH2 glycan leads to near total loss of binding to C1q and the low-affinity FcγRs, but FcγRI binding is substantially retained.⁶¹ More than 90% of human serum IgG glycans contain fucose linked to the first core GlcNAc in α1–6 position, thus called core fucose.⁶² The absence (afucosylation) of the core fucose on IgG1 results in an increased affinity for FcγRIIIA and enhanced ADCC activity.^63–66 Afucosylation is by far the most frequently used strategy applied to increase the therapeutic effect of anti-cancer mAbs and is achieved by the use of production cell lines deleted for the FUT8 gene, or cell lines with naturally low levels of fucosylation.

Modifying other sugar constituents can also enhance Fc-effector function, but to a lesser extent than of fucose. Galactosylation of the IgG N-glycan is reported to increase FcγRIIA/IIIA and C1q binding and Fc hexamerization potential of IgG ^67,59,68,69 and adding GlcNAc (N-acetylglucosamine) to the mannose fork (called bisecting GlcNAc) increases ADCC.^59,70 High levels of terminal sialic acid are thought to increase IgG PK.⁷¹ A high proportion of the mannosylated structures can also increase the ability to interact with FcγRIIIA and decrease binding to C1q.^68,72 However, the impact of these modifications on mAb clearance by mannose receptors on liver macrophages and endothelial cells needs to be considered,⁷³ as does the impact of enhanced binding to FcγR on mAb clearance in general.⁷⁴ Several glycoengineered mAbs are being tested in clinical trials and some approved, including afucose, hypofucose and aglycosyl are described in our case studies, as summarized in Table 4. Case studies 1–3 and 5–7 describe afucosylated Fc-enhanced mAbs.

Table 4.

Fold difference between allometrically scaled PK parameters and human popPK parameters in Tg mouse models and monkey.

PK Parameter	Scaled from Tg32†	Scaled from Tg32-SCID†	Scaled from Tg276†	Scaled from monkey†
CL	0.5	0.7	0.9	0.6
V total	0.8	0.9	1.0	0.6
V1	0.7	0.7	0.8	0.9
V2	1.0	1.1	1.2	0.3
CLD	6.7	5.5	1.6	0.3

^†Projected fold-difference in human PK parameters based on allometry using the respective species. Fold-difference calculated as scaled mouse or monkey parameter value/human PK parameter value.

3.1.2. Amino acid substitutions to enhance Fc-effector function

Amino acid substitutions can also be made in the IgG Fc to alter the activatory:inhibitory ratio and so promote enhanced FcγR-mediated effector function (e.g., ADCC, ADCP). Approaches taken include alanine⁶¹ and higher content scanning (US2006173170A1). These studies showed that the binding site for human FcγRs and C1q was found in the lower hinge and proximal CH2 regions of hIgG1, whilst the FcRn site was at the ‘elbow’ region. Computational structure-based modeling combined with high throughput screening are also used to identify novel variants with enhanced FcγR binding. A number of amino acid substitutions in Fc regions resulting in enhanced activation potential have been identified and described in Table 5. A different approach to gaining enhanced effector function of therapeutic mAbs is to incorporate E345K or E430G mutations that promote hexamer formation following target binding of the mAb and so strongly promote complement activation, and CDC- and ADCC-mediated killing of target cells.⁹⁰ A more recent example is an Fc-engineered variant with three amino acid substitutions (Q311R/M428E/N434W) that not only enhance CDC activity but also extend the plasma half-life, an approach that could also be combined with afucosylation for enhanced ADCC without compromising half-life.⁴¹ Multiple mAbs that contain Fc mutants with enhanced Fc effector function are being evaluated in clinical trials (see Table 4). Case study 3 describes a mAb with Fc mutations for enhanced ADCC activity.

Table 5.

Fc substitutions increasing FcγR binding.

Fc Modification	Abbreviation	Binding Affinity (vs Wild-Type)	Effector Function	Refs.
Afucosylation	Potelligent	↑↑↑FcγRIIIAF158	↑↑ADCC	Yamane-Ohnuki et al.75
S298A/E333A/K334A	AAA	↓↓FcγRIIAH131 ↑↑FcγRIIIAF158/↑FcγRIIIAV158 ↓↓FcγRIIB	↑ADCC	Shields et al.61
S239D/I332E	DE	↑↑↑FcγRIIIAF158/↑↑↑↑FcγRIIIAV158 ↑↑↑FcγRIIB	↑↑ADCC; ↑ADCP	Lazar et al.76
S239D/A330L/I332E	DLE	↑↑↑FcγRIIIAF158/↑↑↑↑FcγRIIIAV158 ↑↑↑FcγRIIB	↑↑↑ADCC; ↑ADCP	Lazar et al.76
S239D/I332E/E345K	DEK	↑↑↑FcγRIIIAF158/↑↑↑↑FcγRIIIAV158 ↑↑↑FcγRIIB	↑↑↑ADCC; ↑ADCP; ↑CDC	Gehlert et al.77
G236A	G236A	↑↑FcγRIIAH131/↑↑FcγRIIAR131 ↑FcγRIIB	↑ADCP	Brinkhaus et al.78
Q311R/M428L	QRML		↑FcRn; ↑ADCC; ↑CDC	Ko et al.79
G236A/S239D/I332E	GASDIE/ADE	↑↑FcγRIIAH131/↑↑↑FcγRIIAR131 ↑↑↑FcγRIIIAF158/↑↑↑FcγRIIIAV158 ↑↑↑FcγRIIB	↑ADCC; ↑ADCP	Richards et al.80
G236A/A330L/I332E	GAALIE	↑↑FcγRIIAH131/↑↑FcγRIIAR131 ↑↑FcγRIIIAF158/↑↑FcγRIIIAV158 ↓↓FcγRIIB	Not stated.	Ravetch and Bournazos81
G236A/S239D/A330L/I332E	GASDALIE	↑↑↑FcγRIIAH131/↑↑↑FcγRIIAR131 ↑↑↑FcγRIIIAF158/↑FcγRIIIAV158 ↑FcγRIIB	No stated	82
F243L/R292P/Y300L/V305I/P396L	LPLIL	↑↑FcγRIIAH131/↑FcγRIIAR131 ↑↑↑FcγRIIIAF158/↑↑↑FcγRIIIAV158 ↑FcγRIIB	↑↑ADCC	Stavenhagen et al.83
L235V/F234L/R292P/Y300L/P396L	VLPLL	↑FcγRIIAH131/↓↓FcγRIIAR131 ↑↑FcγRIIIAF158/↑FcγRIIIAV158 ↓↓FcγRIIB	↑ADCC	Nordstrom et al.84
1 Heavy chain: L234Y/L235Q/G236W/S239M/H268D/D270E/S298A1 Heavy chain: D270E/K326D/A330M/K334E	Asym-mAb1	↑↑FcγRIIAH131/↑↑FcγRIIAR131 ↑↑↑↑FcγRIIIAF158/↑↑↑↑FcγRIIIAV158	↑↑ADCC	Mimoto et al.85
S267E/H268F/S324T	EFT	↑↑C1q	↑↑CDC	Moore et al.86
K326W/E333S	WS	↑C1q	↑↑CDC	Idusogie et al.87
E345R/E430G/S440Y	RGY	IgG hexamer formation	↑↑↑↑CDC	Diebolder et al.88
IgG1/IgG3 cross subclass		↑C1q	↑↑CDC	Natsume et al.89

3.1.3. Strategies to enhance FcγRIIB binding

Enhanced binding to FcγRIIB was first developed to inhibit autoreactive B-cells. Still, such a strategy has found the most use for enhancing target-based clustering agonism, generally driven by a proximal B-cell. This was first reported for agonistic anti-DR5 mAbs that require cross-linking by FcγRIIB for tumor killing and then for mAbs targeting the TNF superfamily molecules CD40, 4-1BB and O×40 in pre-clinical models.¹⁶ Some Fc mutations relevant for FcγRIIB binding and/or activation are summarized in Table 6, the most commonly used mutation being ‘SELF’ and ‘V11’. Multiple mAbs that harbor Fc variants with enhanced FcγRIIB binding are in clinical trials (see Table 4).

Table 6.

Fc substitutions increasing FcγRIIB binding.

Fc Modification	Abbreviation	Binding Affinity (vs Wild-Type)	Refs.
S267E/L328F	SE/LF	↑↑↑↑FcγRIIB ↑FcγRI no binding to FcγRIIIA-V158.	Chu et al.91
P238D		↑↑FcγRIIB ↓↓FcγRI no binding to FcγRIIA-H131 or FcγRIIIA-V158 ↓↓↓FcγRIIA-R131.	Mimoto et al.85
E233D/G237D/P238D/H268D/P271G/A330R	V12	↑↑↑↑FcγRIIB ↓↓FcγRI, ↓↓FcγRIIA-H131, ↑↑ FcγRIIA-R131, no binding to FcγRIIIA-V158 or FcγRIIIA-F158	Iwayanagi et al.92;Mimoto et al.85

3.2. Fc silencing

Fc silencing of mAbs is primarily used in autoimmune disease when the therapeutic goal is to block the function of immune cells without the pro-inflammatory effects associated with effector cell or complement activation, particularly when targeting cell membrane antigens. In infectious diseases, Fc silencing can also help reduce the risk of antibody-dependent enhancement (ADE) of disease or infection.¹² Fc silencing is critically used for T cell-engaging antibodies such as checkpoint inhibitors, costimulatory agonist, and CD3 engagers where the ADCC-mediated killing of the effector cells should be avoided. Therapeutics targeting multiple receptors might also be Fc-silenced to minimize receptor clustering/activation and immunogenicity in general due to Fc-driven uptake by myeloid antigen presenting cells (APCs).

3.2.1. Glycoengineering and amino acid substitutions for Fc silencing

Use of the single amino acid substitutions N297A, N297Q, and N297G is a very simple method for achieving high levels of silencing as it eliminates glycan attachment and therefore yields an aglycosylated antibody. This greatly decreases the binding to all low-affinity FcγRs and C1q, resulting in greatly decreased ADCC and CDC activity. However, substantial FcγRI binding is retained, and the Fc loses some thermal stability. Despite this, a number of aglycosyl IgG1 are approved or in late-stage trials (reviewed by.⁵ Importantly, immune complexes display higher avidity-mediated binding than monomeric formats and should ideally be used to assess FcγR silencing. Aglycosylated hIgG1, hIgG2, and hIgG3 mAbs presented as large immune complexes retain wild-type levels of binding to FcγRIIA-131 R, while previous studies using monomeric, aglycosylated hIgG1/3 showed no detectable binding.²⁴ Case study 8 describes a mAb with an N297A mutation that is aglycosylated and lacking detectable Fc effector function.

After identifying FcγR and C1q binding sites, mutagenesis was used to generate mAb variants with reduced effector functions. The anti-CD3 mAb OKT3 with Fc disabled hIgG1 L234A/L235A showed no detectable binding to the low-affinity FcγRs and C1q, leading to a significant reduction in ADCC and CDC.⁹³ A combination of substitutions and deletion (E233P/L234V/L235A and deletion of residue G236) in a human IgG1 background resulted in a mAb that showed no binding to any human FcγR and became the backbone for Xencor’s XmAb® bispecific platform.⁹⁴ Crystal structure analyses revealed extensive contacts between FcγR and CH2 domain of IgG1, particularly in the Fc C/E loop.⁹⁵ S298G/T299A mutations were designed which abolished/significantly reduced binding to C1q and most FcγRs except for FcγRIIA-R131 and FcγRIIB.⁹⁵ The triple mutant L234A/L235A/P329G had no detectable binding to C1q or FcγRs, resulting in abrogated ADCC when introduced into a hIgG1 anti-EGFR mAb, an obinutuzumab variant, faricimab (Vabysmo), and glofitamab-gxbm (Columvi).

Engelberts et al.⁹⁶ identified the L234F/L235E/D265A FEA mutant, which completely ablates binding to all FcγRs and C1q. This FEA mutation is currently applied in the approved DuoBody- CD3×CD20 epcoritamab. Another simple approach for Fc silencing is to use the hIgG4 Fc domain to minimize the effector function (reviewed in Liu et al.⁵⁵), as it has low affinity for all FcγR, except FcγRI. These normally contain a core hinge S228P mutation to stop half molecule formation during production and the in vivo Fab arm exchange process.⁹⁷ IgG4 is now commonly additionally disabled with mutations, such as ‘FALA’ (F234A L235A). Specific amino acids from human IgG4 have also been used in IgG1 and IgG2 to greatly reduce/remove FcγR and C1q binding.^98–100 Fc amino acid substitutions that reduce Fc effector function are summarized in Table 7. Case studies 8–13 describe mAbs with Fc mutations for Fc silencing.

Table 7.

Fc amino acid substitutions ablating FcγR binding.

Fc Modification	Abbreviation	Binding Affinity (vs Wild-Type)	Effector Function (vs Wild-Type)	Refs.
Aglycosylation (N297A/Q/G)	NA	No binding to FcγRIIAH131/FcγRIIAR131 FcγRIIIAF158/FcγRIIIAV158 FcγRIIB C1q. ↓↓↓FcγRI.	↓ADCC; ↓↓CDC	Walker et al.101
L235A/G237A/E318A	AAA	↓↓FcγRI ↓↓FcγRIIAH131/↓↓FcγRIIAR13 ↓↓FcγRIIIAF158/ ↓↓FcγRIIIAV158 ↓↓FcγRIIB.	↓↓ADCC	Hezareh et al.102;Hutchins et al. 103
L234A/L235A	LALA	No binding to FcγRIIAH131 /FcγRIIAR131 FcγRIIIAF158/FcγRIIIAV158 FcγRIIB C1q. ↓↓↓↓FcγRI.	↓↓ADCC; ↓↓CDC	Xu et al.93
S228P/L235E	IgG4-PE	No binding to FcγRIIIAV158 C1q. ↓↓↓↓FcγRI ↓↓↓FcγRIIAH131 /↓↓↓FcγRIIAR131 ↓↓FcγRIIB.	No ADCC	Schlothauer et al.104
G236R/L328R	RR	No binding to FcγRIIAH131, FcγRIIIAV158 FcγRIIB. ↓↓↓↓FcγRI.	Not determined	Chu et al.91
S298G/T299A	GA	No binding to FcγRIIIAF158/FcγRIIIAV158 C1q. ↓↓↓FcγRI ↓FcγRIIAH131/ ↑↑FcγRIIAR131 ↑FcγRIIB.	Not determined	Sazinsky et al.95
L234F/L235E/P331S	FES	No binding to FcγRI ↓↓ FcγRIIAH131/FcγRIIAR131 ↓↓↓FcγRIIIAV158 ↓↓↓C1q.	Not determined	Oganesyan et al.105
H268Q/V309L/A330S/P221S	IgG2m4	No binding to FcγRI FcγRIIIAF158/FcγRIIIAV1588 C1q. ↓↓FcγRIIB	Not determined	An et al.98
E233P/L234V/L235A/G236del/S267K	XmAb® bispecific	No binding to FcγRI FcγRIIIAV158 FcγRIIB	Not determined	Moore et al. 94
L234A/L235A/P329G	LALA-PG	No binding to FcγRI FcγRIIAH131/FcγRIIAR131 FcγRIIIAF158/FcγRIIIAV158 FcγRIIB, C1q	No ADCC	Schlothauer et al.104
V234A/G237A/P238S/H268A/V309L/A330S/P331S	IgG2c4d	No binding to FcγRI FcγRIIA FcγRIIIA C1q	No ADCC, ADCP, CDC	Strohl and Vafa 99
L234F/L235E/D265A	FEA	No binding to FcγRI FcγRIIA FcγRIIIA FcγRIIB C1q	Not determined	Engelberts et al.96
L234S L235T G236R	STR	No binding to FcγRI FcγRIIA FcγRIIIA FcγRIIB C1q	Not determined, no activation of FcgR+ cells, no cytokine release	Wilkinson et al.106
Q311R/M428E/N434W	REW	IgG oligomerization in an on-target manner	↑↑↑↑CDC	Foss et al. 41

3.3. Half-life extension

For some therapeutics, half-life extension (HLE) can be desirable with several benefits beyond enhancing mAb exposure including increased duration of their pharmacological activity for the same equivalent molar dose, decreased dosing frequency, patient-friendly dosing regimens and reduced cost of goods. However, HLE is not always an appropriate strategy as certain classes of mAb therapeutics, such as agonist mAbs, may drive adverse events through long-term target engagement. Additionally, it may not be useful with targets with significant target-mediated elimination or rapid turnover (see case study 12). Caution should be taken as such engineering may modulate FcγR and rheumatoid factor binding. The primary strategy to prolong mAb half-life to improve FcRn binding at pH 6, but not pH 7.4, as reviewed by Saxena and Wu.¹⁰⁷ Table 8 includes descriptions of Fc motifs with impact on hFcRn binding, PK, and antibody effector functions. mAbs that have been Fc-engineered for HLE are described in case studies 12–13 and 15.

Table 8.

Fc amino acid substitutions enhancing FcRn binding and half-extension.

Mutations	Abbreviation	Effect on hFcRn Binding	PK Observations	Effector function impact	Ref
M428L/N434S	LS or XtendTM	11.3-fold increase at pH 6	Bevacizumab: Decrease in CL by 5.3- fold in hFcRn Tg276 mice; 4.2-fold increase in t1/2 in Tg mice; decrease in CL by 3.1- fold in monkey; increase in t1/2 by 3.2- fold Cetuximab: decrease in CL by 5.2-fold and increase in t1/2 by 4.8- fold in Tg276 mice; decrease in CL by 3.2- fold and increase in t1/2 by 3.1-fold in monkey Ravulizumab: half-life in patients ~50 days, 4- fold longer than eculizumab	Small decrease in binding for FcγRIIA-R131, FcγRIIA-H131, FcγRIIB, FcγRIIIA-V158, FcγRIIIA- F158, and FcγRIIIB. Shown to increase RF binding somewhat	Lee et al.108; Mackness et al.109; Maeda et al.110; Peffault de Latour et al.111; Sahelijo et al.112; Sheridan et al.113; Zalevsky et al. 114
M252Y/S254T/T256E	YTE	6- to 11-fold increase at pH 6	Decrease in CL by 5- to 8-fold in cynomolgus monkey;	>100-fold decrease in ADCC activity, due to reduced FcγRIIIA and C1q binding – reversed with additional mutations of S239D/A330L/I332E.	Dall’Acqua et al.115; Grevys et al.116; Mackness et al.109; Zalevsky et al.114
			2- to 5-fold increase in human serum t1/2, up to 100 days	Strongly reduces RF binding
T250Q/M428L	QL	29- to 40-fold increase at pH 6	Decrease in CL by 2.3- to 2.8-fold in rhesus monkey, increase in terminal t1/2 by 1.8- to 2.5-fold	No negative impacts on ADCC/CDC activity	Datta-Mannan et al.117; Hinton et al.118; Hinton et al.119
N434A		Stronger binding to hFcRn. 3.5-fold increase to hFcRn at pH 6 Increase in terminal t1/2 by 1.7-fold in Tg276 homozygous mice	No effect on FcγRI, FcγRIIA, FcγRIIB, or FcγRIIIA binding		Petkova et al.120; Valente et al.121; Shields et al.61; Burvenich et al.122
		Increase in terminal half-life by 2.2-fold in Tg276 heterozygous mice
		Increase in terminal half-life by 1.6-fold in Tg32 heterozygous mice
T256D/T307Q	DQ	10.3-fold increase at pH 6	Decrease in CL by 2.3- and 1.5-fold in Tg32	Less pronounced effects than YTE (qualitative analysis)	Mackness et al.109
			mice and cynomolgus monkey, increase in terminal t1/2 by 2.1- fold in Tg32 mice and 2.1-fold in monkey
T256D/T307W	DW	14.1-fold increase at pH 6	Decrease in CL by 2.1- and 2.2-fold in Tg32 mice and cynomolgus monkey, increase in terminal t1/2 by 1.7- fold in Tg32 mice and 2.1-fold in monkey	Less pronounced effects than YTE (qualitative analysis)	Mackness et al.109
M252Y/T256D	YD	25.4-fold increase at pH 6	Decrease in CL by 1.6- and 2.4-fold in Tg32 mice and cynomolgus monkey, increase in terminal t1/2 by 1.5- fold in Tg32 mice and 2.4-fold in monkey	Similar effects to YTE (qualitative analysis)	Mackness et al.109
H433K/N434Y	HN or NhanceTM	6-fold increase at pH 6	Increase in terminal t1/2 by 4-fold in Swiss Webster mice; ARGX- 111 had t1/2 in cancer patients of 64–116 h	Decreased binding for FcgRI, FcgRIIA-R131, FcgRIIA-H131, FcgRIIIA-V158, FcgRIIIA-F158, and FcgRIIIB	Grevys et al.116
T307A/E380A/N434A	AAA	11.8-fold increase at pH 6	Increase half-life in Tg276 hemizygous mice by 2.5-fold	Single mutation of T307A increased binding to FcgRIIA and FcRIIB; single mutation of E380A had no impact on FcgR binding; single mutation N434A had no impact on FcgR binding; triplemutant data not available	Petkova et al.120; Shields et al.61
M252Y/N268E/N434Y	YEY†	172-fold increase at pH 6, 700-fold increase at pH 7.0	Decrease in CL by 1.1- fold in Tg32 mice in presence of hIL-6R	Not tested	Igawa et al.17; Igawa et al.123
M252Y/V308P/N434Y	YPY†	459-fold increase at pH 6, 2500-fold increase at pH 7.0	Increase in CL by 4.3- fold in Tg32 mice in presence of hIL-6R	Not tested	Igawa et al.17; Igawa et al.123
M428L/N434A	LA†	10-fold increase at pH 6.0	Decrease in CL by 3.1- fold in cynomolgus monkey Crovalimab: half-life increased to 30 days as compared to Eculizumab, 11-16 days, but similar to Ravulizumab, ~32 days.	Not tested	Fukuzawa et al. (2017); Roth et al. (2020); Stern and Connell (2019)
H285D/T307Q/A378V	DQV	13-fold increase at pH 6.0	Increase in terminal half-life by 8.4-fold and decrease in CL by 3.0- fold in Tg276 mice	Increase in FcgRIIIA binding by 3.9-fold, increase in ADCC and CDC activity by 1.5-fold and 2.7- fold, similar C1q activity.	Booth et al. (2018)
V308P	VP	43 to 390-fold increase at pH 6.0	Increase in terminal half-life by 2.5-fold and decrease in CL by 2.1- fold in cynomolgus Monkey		Datta-Mannan et al.167
L309D/Q311H/N434S	DHS	5-fold increase at pH 5.8 No detectable binding at pH 7.4	Increase in terminal half-life by 5.9-fold and decrease in CL by 6- fold in Tg276 mice Increase in terminal half-life by 4.1-fold in Scarlett mice (hFcRnKI hβ2mKI hFcγRKI hIgG1, κKI)	FcγR binding affinities and ADCC in vitro potency similar to wild- type IgG CDC potency approximately 2- fold greater than WT IgG and 10- fold more potent than YTE mutation	Lee et al.164
Q311R/M428L	PFc29	4.4-fold increase at pH 6.0	Increase in terminal half-life and decrease in CL by 2-fold in hFcRn Tg276 mice Increase in terminal half-life by 2-fold and decrease in CL by 2.2- fold in cynomolgus Monkey	2.6-fold increase in ADCC activity (EC50) compared to WT antibody 3.6-fold increase in CDC activity (EC50) compared to WT antibody	Ko et al.79

Note. † Antibody Fc mutations used for sweeping antibodies.

Variants have also been designed to modulate binding at extracellular pH, so-called “sweeping antibodies”, to clear away soluble targets from the circulation.^17,123 A sweeping antibody incorporates two antibody engineering technologies: 1) conditional Ag binding to bind to a target in plasma but not the endosome (typically pH- or Ca2 ± related), and 2) Fc engineering to increase the cellular uptake of the mAb-target complex into the endosome (typically FcγRIIB and/or FcRn related and/or PI engineering).^123,124 Hence, ‘sweepers’ have enhanced cellular association and uptake at neutral pH and reduced Ag binding at acidic pH, resulting in a net in-flow of antigen for endosomal destruction. Sweeping mAbs are highly effective at clearing soluble targets coupled to an extended half-life and shown value with soluble targets with high serum concentrations (e.g., complement factors) and targets with high turnover rates (e.g., cytokines, chemokines). Case Study 13 describes a sweeping antibody.

3.4. FcγR and FcRn blockade

Autoantibody-mediated diseases may be treated with intravenous immunoglobulin (IVIG), which is thought to act in part via blockade and modulation of Fcγs, FcRn, and complement via Fc aggregates in the preparation.¹²⁵ However, IVIG has limited supply, high cost, and patient inconvenience due to multi-hour/day infusions. These drawbacks have driven the development of recombinant, high avidity Fc multimers to recapitulate these effects (reviewed by Fitzpatrick et al.⁵⁶ Zuercher et al.),¹²⁶ Stradomers^TM are generally human IgG1 hinge-Fc with a C-terminal addition of a human IgG2 hinge region.^127,128 The human ‘stradomer’, GL-2045 (PF-06755347), avidly binds human FcγRI, FcγRIIA, FcγRIIB and FcγRIIIA, as well as rat, mouse and cynomolgus FcγRs, and is protective in CIA and immune thrombocytopenia (ITP) models. However, Phase 1 trial (NCT03275740) was terminated for business reasons. Fc hexamers are human IgG1 hinge-Fc with a C-terminal fusion to an 18 amino acid ‘tailpiece’ sequence from human IgM with an optional leucine 309 to a cysteine mutation (L309C). Fc-hexamers show greater avidity for FcγRs and C1q than IVIG and block and/or cause internalization of FcγRs, resulting in ADCP and ADCC blockade and protect mice from autoimmunity.^129–132 IgG1- Fc hexamers caused profound cytokine release and platelet and complement activation in human whole blood assays, whereas the IgG4 Fc-hexamer did not.¹³² Ortiz et al.¹³³ showed that trimeric Fc multimers avidly bind FcγRs, but do not induce FcγR activation. Fc-trimer competitively inhibits several IC-mediated FcγR functions and protects mice from ITP, in the absence of proinflammatory effects. Human Phase 1 trials with Fc-trimer were terminated for commercial reasons. Fc-multimers are increasingly being engineered to have enhanced C1q binding and/or altered FcγR binding, presumably with the aim of overcoming some of these translational and safety challenges, some of which are described in case study 14.

Blocking FcRn to increase the degradation of IgG has been successfully explored for the treatment of autoantibody-mediated diseases.^57,134–136 This can be achieved with engineered Fc fragments,¹³⁷ anti-FcRn antibodies,^138–140 non-antibody-based short peptides, affibodies (high affinity FcRn-binding proteins) and small molecules.^141–143 A potential drawback of these antibody-degrading technologies is the lack of selectivity for pathogenic, over protective, IgG. Two such strategies have so far been granted marketing approvals, the engineered Fc fragment efgartigimod (VYVGART; described in case study 15) and the full-length IgG4 rozanolixizumab (RYSTIGGO; with S228P). An alternative is the development of antigen-specific approaches, such as antigen-Fc fusions (Seldeg). These can bind pathogenic IgG only, leading to rapid receptor-mediated internalization and degradation of IgG. An example is fusion of myelin oligodendrocyte glycoprotein (MOG) to selectively degrade anti-MOG autoantibodies.^144,145

4. Pharmacology and safety testing of Fc-engineered mAbs

For Fc-modified mAbs, a relevant species for pharmacology and safety testing is one in which the antigen-binding arms of the mAb have a similar pharmacological activity to that intended in humans, i.e., binds to mouse target equivalently and with relatable biology but also that the Fc effector function is comparable to human. In vitro studies with Fc-modified mAbs should characterize the FcγR and FcRn-binding profiles and effector functions, as well as identify any safety-related effects, relative to wild-type IgG, with humans and animal cells (cyno and mouse usually), prior to conducting in vivo pharmacology and toxicology studies.

4.1. In vitro studies to characterize FcR binding and effector function of Fc-engineered mAbs

In vitro assays commonly used to evaluate Fc-mediated IgG effector functions include binding assays for FcγR, FcRn, and C1q, and activity assays, such as ADCC, ADCP and CDC (see case studies 8–12). For molecules that might have enhanced risks such as Fc-enhanced IgGs and Fc blockers at greater risk of unintended FcγR activation, assessment of effector function and immune activation such as cytokine release due to unwanted activation of FcγR-bearing cells, including platelets, neutrophils, macrophages and NK cells, is an important part of nonclinical safety strategy. This is also true for novel Fc-silenced molecules, demonstrating a lack of FcγR binding and effector function, ideally in the presence and absence of competing irrelevant IgG, such as IVIG. Half-life engineered molecules should be assessed for FcRn binding at pH 6 and 7.4 but ideally by studying how the mAbs engage the receptor throughout a pH gradient by, for instance, the use of an analytical FcRn retention/elution column assay. However, such in vitro biochemical assays do not provide insight into how IgG mAbs behave in a cellular context. As a solution to this, cellular assays can be used to map cellular uptake, FcRn-mediated rescue from degradation, and accumulation/degradation. One example is a human endothelial cell-based recycling assay (HERA) that can be used to study how Fc-engineered strategies for altered FcRn engagement affect cellular handing, the results of which can guide selection of variants to be tested in vivo.^42,146 In addition, this assay can also be used to study how distinct features of the variable regions of mAbs or Fc-fusion partners affect cellular uptake and FcRn transport properties, predictive for half-life in vivo.^50,147 Of note, if a mAb binds to cell surface receptors indirectly through engaging soluble proteins, the assessment of potential effector function may be needed.¹⁴⁸

The binding of Fc-engineered molecules to purified FcγRs, FcRn, and C1q can be done using surface plasmon resonance (SPR), and binding to FcγR- and FcRn-expressing cells can be assessed by flow cytometry. However, the low-affinity nature of these monomeric interactions makes definitive or accurate conclusions challenging. Therefore, binding studies should use both monomeric and immune-complexed drugs since use of monomeric IgG will underpredict binding to the low-affinity FcγRs. Reporter gene assays utilizing FcγR-bearing cells can also be used to assess the activation of the cells following FcγR binding. Cell-based effector function assays are used to demonstrate enhanced (or reduced) potency and to assess unwanted effects of Fc-engineered mAbs, including blood cell depletion, effector cell activation, whole blood, and PBMC-based cytokine release assays, ADCC, ADCP and CDC assays, platelet activation and aggregation assays, and neutrophil and basophil activation assays. The assays used will depend on the primary mechanism of action of the drug (e.g., ADCC, ADCP, CDC, FcγRIIB agonism, and silencing) and the FcγR-binding profile. To increase the clinical translatability of these assays, the target cells for ADCC, ADCP and CDC assays should have a disease-relevant target and FcγR expression and ADCC/CDC sensitivity (e.g., low expression of complement regulatory proteins, CD55 and CD59). The effector cells (e.g., in ADCC and ADCP assays) should be FcγRIIIA-expressing NK cells and FcγRIIA-expressing macrophages respectively, that reflect those in patients. Complement sources used in the CDC assays should be compatible with relevant complement defense factors. The Fc-engineered drug is compared with the parental drug and with the wild-type. Assays should be performed in the presence of physiological levels of human IgG, which would compete for FcR binding¹⁴⁹ in vivo. The goal is to understand the balance between intended mAb activity and any unwanted/unknown toxicity. The impact of preexisting anti-drug antibodies (ADA) and, hence, drug aggregation can be assessed by pre-incubation of healthy volunteer or patient-relevant serum samples.

After assessment in human in vitro assays, a similar range of assays should be performed whenever possible using recombinant proteins and/or cells from the animal toxicology species that will be used, often cynomolgus macaque and rodent. The aim is to confirm whether the expected human functional profile is recapitulated in the animal species and identify any key activities in humans that cannot be recapitulated in animals.

Use of assays that include human-derived organoids allows the evaluation of effector functions in a more physiologically relevant environment. Organoid systems are three-dimensional structures derived from stem cells, which resemble structural and functional properties of human tissues.¹⁵⁰ They are developed by inducing stem cell differentiation under specific culture conditions, replicating the natural microenvironment of specific organs, and enabling cellular self-assembly into tissue architectures.¹⁵¹ The translational benefit of organoids lies in their enhanced predictive accuracy over traditional two-dimensional cell cultures and animal models, providing physiologically relevant platforms for assessing PK/pharmacodynamics (PD) relationships.¹⁵² For example, kidney organoids have been used to evaluate nephrotoxicity and clearance of antibody-based therapies;¹⁵³ intestinal organoids have facilitated investigations into mucosal antibody absorption and barrier interactions¹⁵⁴; and patient-derived tumor organoids have aided in predicting therapeutic efficacy and resistance mechanisms of antibody therapies in personalized oncology.¹⁵⁵ Although there is promise for this methodology, scaling organoid-based assays for high-throughput screening remains technically challenging, and variability in organoid cultures may lead to reproducibility and standardization challenges,¹⁵¹ and translating findings from organoids directly into clinical outcomes requires cautious interpretation and additional validation ¹⁵⁰. The most notable challenge is incorporation of immune cell components to evaluate effector function pharmacology. Recent advances have enabled the use of patient-derived pancreatic cancer organoids to evaluate combination therapies of PD-L1 inhibitors and trastuzumab.¹⁵⁶

4.2. In vivo studies in standard NHP and Rodent models to assess pharmacology and safety of Fc-modified mAbs

As discussed above, for wild-type and afucosylated human IgG1 that function primary through ADCC via FcγRIIIA/IIA, the cynomolgus monkey and mouse are generally considered predictive for target-mediated killing and associated safety assessment.^26,27 Hence, cyno and sometimes mouse (where target cross-reactivity exists) have been used as the relevant pharmacology and toxicology species for many mAbs with wild-type, enhanced and null effector functions (see case studies). For Fc-enhanced mAbs, these studies aim to assess the ADCC-mediated target cell killing (potentially in blood and tissues), the specificity, depth, duration and recovery of these effects, and whether there is any killing of non-target cells, or any unwanted immune activation (e.g., cytokine release, platelet activation, complement activation) or sequelae associated with prolonged target cell depletion (e.g., reduced immune function, increased infection). The goal of these studies is to determine the relationship between drug exposure and target cell depletion.

4.2.1. Alternative animal models

When in vitro and/or in vivo studies demonstrate that the primary effector functions, exposure, and potential related safety risks of Fc-modified mAbs cannot be fully recapitulated in cynomolgus monkey or mice, then alternative models should be considered.

4.2.2. Fc surrogates

When the effector function of an Fc-engineered mAb is not recapitulated in cynomolgus monkeys or mice (and/or the mAb does not bind to the cyno target via its Fab arms), then the use of a rodent surrogate molecule might be considered. For example, for an Fc-enhanced IgG1 that mediates its effect via enhanced FcγRIIIA-mediated ADCC activity or enhanced FcγRIIB-mediated T cell agonism, an Fc-engineered mouse IgG2a isotype with the same level of mouse FcγR binding/effector function activity in mice as the human could conceptually be used. Alternatively, mice transgenic (knock-in) for human FcγRs can be used. Surrogate Fab arms specific to the mouse target might also be required. The use of a surrogate mAb might provide important mechanism-related safety data that is important for hazard identification. Case study 13, for instance, describes use of a surrogate mAb.

4.2.3. Human FcγR transgenic and immune system-humanized mice

Disadvantages with surrogate mAbs (e.g., species differences in FcγR expression, uncertain human translatability) can be mitigated by testing the clinical mAb candidates in human FcγR transgenic mice. These mouse models (huFcγR) aim to replicate the expression patterns, levels, and signaling capabilities of human FcγRs, and have been used to test the effector function and efficacy of Fc-engineered molecules.^157,158 HuFcγR Tg mice are increasingly available from academic, commercial, and pharma sources. Researchers have used huFcγR models to test engineered antibodies such as S239D/I332E (SDIE) with enhanced binding to all human FcγRs, G236A/S239D/A330L/I332E (GASDALIE) with enhanced binding to FcγRIIA and FcγRIIIA, S267E (SE) and S267E/L328F (SELF) with enhanced binding to inhibitory FcγRIIB and an N297A aglycosylated silent mutant. HuFcγR mice have also been used to compare the effector functions of Fc-engineered mAbs in several models of antibody-driven inflammatory disease (e.g., ITP, Type III/immune-complex-mediated hypersensitivity reaction), infection (influenza, anthrax, SARS-COV2), and cancer.^159,160 HuFcγR mice recapitulate ADCP, ADCC, platelet depletion, leukocyte depletion, tumor clearance via FcγRIIA and FcγRIIIA, and FcγRIIB clustering-based agonism, including the enhancement of these effects by Fc-engineered and afucosylated mAbs.^159,160 Crossbreeding of huFcγR mice with Rag2 knockout mice (RhuFR) results in immunodeficient mice lacking functional B and T cells, which can be engrafted with human tumors.¹⁵⁷ A further variant mouse strain (RhuFR1-) lacks huFcγRI to preserve the long half-life of human IgG by preventing the target-mediated drug disposition (TMDD) of human IgG via FcγRI due to the lack of endogenous IgG competition. Increasingly, mice are being created with multiple key factors influencing the PK of human IgG, such as hFcRn, human serum albumin, and human IgG. FcγR humanized mice have also been used for the safety assessment of an Fc-engineered anti-CD40 agonist mAb with enhanced FcγRIIB binding. To express the human target, huFcγR mice were crossed with a huCD40 transgenic mouse line. The resultant mice were engrafted with a CD40-expressing tumor. In this model, the engineered mAb demonstrated increased anti-tumor activity compared to the parental mAb.^161,162 The model also recapitulated platelet (thrombocytopenia) and hepatic (transaminitis) toxicity observed with anti-CD40 mAbs in patients. The Fc variant of the antibody had dose-related increased toxicity (intravascular thrombi, hepatocyte necrosis) compared to the wild-type IgG2 mAb. The huFcγR model was used to optimize dosing and identify a delivery regimen with tolerable toxicity and optimal anti-tumor activity. The adverse effects of anti-DR5 mAb, requiring inhibitory FcγRIIB binding for anti-tumor and associated hepatotoxic effects in humans, were also recapitulated in the transgenic FcγR model.¹⁶³ Notably, the dose-related toxicities observed in patients and in the transgenic model were not replicated in WT mice and NHPs.

The PK of human mAbs in mice differs from that in humans, making the direct comparison of dose-related effects problematic. To overcome this issue, a new strain has recently been developed that expresses all huFcγRs, hFcRn, and human IgG to provide competition for FcγR binding¹⁶⁴ (see next section). However, the levels of IgG will still be lower than what is found in humans, as the mice are kept under pathogen-free housing, and boosting by immunization is then needed.

Limitations of huFcγR mice and variants include subtle differences in the expression of human FcγRs on mouse cells compared to human cells (see case study 4) and the need to create mouse strains with complex genetic configurations to express human target antigens such as CD20, HER2, EGFR, CD40, CTLA-4, and PD-1. PAF-driven anaphylaxis (via immune complexes and FcγRs) is possible in mice, but rarely observed in humans. There is also limited experience in safety characterization and background data in these mice compared to standard toxicology species. It is important to note that repeated dosing will likely result in an anti-human IgG ADA response, which might compromise the toxicology assessment and limit the dosing duration. Hence, the drug product may need to be murinized if longer dosing is required. Expressing a human IgG1 transgene in huFcγRs mice will induce tolerance to human IgG1 and reduce anti-constant region ADA production, but adds to the complexity of the model and does not solve the issue for other IgG subclasses. To fully recapitulate human FcγR function, other Fc receptors (e.g., FcRn (see above), IgA and IgE receptors, Type II FcγRs) might need to be expressed in these mice, increasing the effort to generate and maintain the models. The advantage of using huFcγR mice is that simultaneous assessment of pharmacology-related toxicity, nonspecific toxicity, and local tolerance of the clinical drug product is possible. However, due to the limitations mentioned above, careful characterization of each model is required to facilitate the interpretation of any findings.

A different approach is to use the immune system of humanized mice. These models are based on the engraftment of highly immunodeficient strains, such as the NOG or the NSG, with either human hematopoietic stem cells or circulating PBMCs. The use of these models to test the activity of engineered human IgGs has been considered problematic for several reasons, including the lack of a functional complement system, TMDD mediated by a mouse FcγRII with high affinity toward human IgGs, non-physiological antibody PK caused by the interaction between the human mAb and the mouse FcRn, and the lack of human platelets in the blood. On the other hand, the value of having functional human immune system cells with the complete FcγR system compensates for at least some of these shortcomings, especially if the goal is to compare different Fc-engineered molecules. Different strategies can be used to address some of the issues associated with using these humanized models. Pre-treatment with IVIG provides binding competition on FcRn and FcγRs, deleting the gene coding for mFcγRII to avoid excessive TMDD¹⁶⁵ or introduction of a functional Hc1 gene to restore complement function.¹⁶⁶ The development of new models based on the currently available immunodeficient mice will likely address some of the major deficits in the near future. However, before conducting antibody testing in immune system humanized mice, it is crucial to characterize them thoroughly and understand their limitations. Case studies 2, 8, and 9 describe studies in huFcγR transgenic mice used to assess the safety of Fc-engineered molecules.

4.2.4. Human FcRn transgenic mice

For the most part, hIgG1 and hIgG2 mutants with increased affinity for hFcRn exhibit increased serum half-life in non-human primates.^118,119,167 However, there are examples where the monkey PK data has overestimated human exposure, likely due to human IgG having a 3-fold higher affinity for monkey FcRn than human FcRn. Rodent models (the preferred species for screening of multiple Fc-engineered candidates during lead optimization) are even less informative for evaluating half-life-engineered mAbs because of the known differences between rodent and human IgG-FcRn interactions. Human IgG has a 10-fold higher affinity for mouse FcRn than mouse. Consequently, half-life engineered IgGs do not show increased half-life in wild-type mice, but are shorter than that of the wild-type counterpart^{44,168,41,42,120} To address this issue, mice lacking endogenous mouse FcRn and transgenic for human FcRn (hFcRn Tg mice) are available for evaluating the PK of human IgG mAbs, in which these can be used to evaluate Fc-engineering extension strategies.^41,42 Different variants of hFcRn transgenic mice are available from The Jackson Laboratory:

1. ‘Tg32’ (B6.Cg-Fcgrt^tm1DcrTg(FCGRT)32Dcr/DcrJ) mice express hFcRn under the endogenous human FcRn promoter. These mice demonstrate the longest half-life of the hFcRn models due to higher transgene copy numbers and are considered the best for modeling human PK of therapeutic mAbs.

2. ‘Tg32-SCID’ (B6.CgFcgrt^tm1DcrPrkdcscidTg(FCGRT)32Dcr/DcrJ) have the scid allele, making them immunodeficient, thereby providing the ability to test potential immunogenic molecules without the impact of ADA.

3. ‘Tg276’ (B6.Cg-Fcgrt^tm1DcrTg(CAG-FCGRT)276Dcr/DcrJ) mice express hFcRn under the ubiquitous CAG (chicken β-actin). These mice have lower transgene copy numbers due to expression via CAG, resulting in lower and non-physiological FcRn expression tissue patterns. These mice are best suited for discriminating subtle PK differences among lead candidates due to the lower FcRn expression.

So far, literature suggests the Tg32 mice are best suited for selecting lead candidate IgG from pools of candidates. Each transgenic strain is available with either one or two transgenic alleles of hFcRn, also known as hemizygous and homozygous, respectively. Avery and colleagues compared the PK of antibodies in the Tg32 hemi- and homozygotes and demonstrated similar serum clearances. The Tg276 mouse expresses hFcRn under the CAG (chicken β-actin) promoter and is also available as hemizygous or homozygous and is recommended for discrimination between antibodies with very small differences in clearance or very long half-life. Avery et al. also compared the differences between Tg32 strains and Tg276 homozygous mice. One tested mAb demonstrated approximately a 9-fold greater clearance in the Tg276 homozygous strain than in the Tg32 strains. The Tg276 strain was used to compare the half-lives of a number of different clinically approved mAbs and fusion proteins with published human and primate data.¹⁶⁹ Analysis comparing half-life in all three species revealed Tg276 mice had greater correlation with humans (r = 0.80) than with wild-type mice (r = 0.66) or primates (r = 0.75).¹⁷⁰ It is important to note the differences observed in the FcRn tissue expression of these various strains and the comparison to humans.^35,171,172

The varying degree of hFcRn expression between these two strains could result in different IgG or Fc-fusion protein therapeutics PK profiles. For example, Petkova et al. demonstrated trastuzumab and two HLE versions all exhibited greater clearance in Tg276 strains (hemizygous and homozygous) compared to Tg32 hemizygous mice. A meta-analysis of Pfizer antibodies concluded that Tg32 homozygous mice provided good prediction of human PK for antibodies following allometric scaling of CL with a 0.9 exponent.¹⁷³ However, a meta-analysis to understand the predictive ability of Tg276 mice is lacking. Case study 20 provides evidence to further investigate Tg276 mice.

Another Tg32 variant, Tg32-hFc (B6.Cg-Tg(FCGRT)32Dcr Fcgrt^tm1Dcr Ighg1^{em2(IGHG1)Mvw}/MvwJ), contains a knock-out of the IGHG1 locus that is responsible for the expression of the mouse hinge and knock-in for human IgG1 Fc domain. PK studies in this model may be more instructive since the endogenous IgG in this mouse model will bind and occupy hFcRn as per human.¹⁷⁴ More recently, a new knock-in mouse strain, designated the Scarlett mouse¹⁶⁴ that expresses hFcRn, all hFcγRs, β2-microglobulin and human IgG transgenes has been developed. It thus, recapitulates all the key processes relevant to human mAb persistence in circulation, namely: 1) physiological expression of hFcRn, 2) the impact of hFcγRs on antibody clearance, and 3) the role of competing endogenous IgG. An IgG containing the L309D/Q311H/N434S (DHS) substitutions exhibited markedly improved PK in both conventional hFcRn transgenic mice and in these new strains. One challenge is that these mice are not readily available to drug developers since they are not distributed by any of the most commonly used mouse providers due to intellectual property issues. See case studies 6 and 7, where studies in hFcRn KI mice were performed.

5. Clinical assessment of Fc-engineered antibodies

5.1. First in human-dosing considerations

In general, approaches for guiding first in human (FIH) dose for Fc-engineered antibodies are similar to those of conventional antibodies. FIH approaches based on toxicology studies such as no adverse event level (NOAEL)/highest non-severely toxic dose (HNSTD), pharmacological active dose (e.g., receptor occupancy (RO) and/or target engagement (TE)), and minimal anticipated biologic effect level (MABEL) are generally used depending on the risk assessment on target and molecular properties.¹⁷⁵ For example, a silenced Fc is expected to reduce unwanted engagement of FcγRs and thereby improve its safety profile; thus, FIH dose based on toxicology studies or pharmacologically active dose that generally results in a higher starting dose may be appropriate. Also, Fc-engineering to extend half-life leads to antibodies expected to have similar pharmacological and toxicological effects as that of conventional Fc; thus, approaches considered for a conventional antibody are appropriate. Moreover, in the case of enhanced Fc activity, given the expected higher potency, one must carefully review the safety window in the preclinical setting to determine an appropriate starting dose. A more conservative approach may be necessary depending on the target properties and mechanism of action.

Bioanalytical strategies to support clinical studies for Fc-engineered antibodies are generally similar to those of conventional antibodies. Measurements should typically include assessment of free/active antibody, ADA, RO, and TE when applicable. Often, RO and TE are measured in the peripheral blood, and the results may need to be extrapolated to the tissue (site of action) with the help of PK/PD and/or quantitative systems pharmacology models.

5.2. Dosing regimen and clinical monitoring

Once advanced into the clinic, any special considerations around dosing regimen and clinical monitoring for Fc-engineered mAbs is largely dictated by the properties of each engineered mAb. For example, for extended half-life mAbs, dosing intervals are typically prolonged with more infrequent dosing required to maintain desired serum concentrations. Similarly, for mAbs engineered with enhanced effector function, such as afucosylation, which induces rapid and prolonged depletion of target cells, infrequent dosing (e.g., every 8 weeks) may be adequate to maintain cell depletion.

Specialized clinical monitoring of Fc-engineered mAbs is not typically required. However, for half-life extended mAbs, the safety follow-up period may be prolonged (6 months to 1 year) compared to standard mAbs because of the duration of exposure and slow clearance of these mAbs. Regarding immunogenicity assessment of Fc-engineered mAbs, the bioanalysis strategy for ADA formation may include the development and implementation of specialized assays intended to characterize the ADA specificity against any novel sequences engineered into the Fc region. This is because mAbs with highly engineered Fc regions may be recognized as foreign, leading to more frequent ADA formation.

5.3. Clinical experience with Fc-engineered IgG

The maturation of Fc engineering is such that a significant number have entered clinical trials, and some have reached regulatory approval and marketing. Approved Fc-engineered mAbs are summarized in Table 9, whilst we summarize case studies provided by the authors of this publication, drawing on their organizations’ direct experience in monoclonal antibody development.”, some of which have entered the clinic, in section 6.

Table 9.

Approved Fc-engineered mAbs (adapted from the antibody society. Therapeutic monoclonal antibodies approved or in review (31 March 2025); https://www.Antibodysociety.org/antibody-therapeutics-product-data/).

Modification	INN			Target	Fc modification	Function		Indication (1st approval)	Approval Year (US)
Enhanced Fc effector function	Amivantamab			EGFR, cMET	Low-fucose; K409R; F405L	Enhanced ADCC/ADCP		NSCLC w/EGFR exon 20 insertion mutations	2021
	Belantamab mafodotin (belantamab mafodotin-blmf)			BCMA	Afucosylated	Enhanced ADCC/ADCP		Multiple myeloma	2020
	Benralizumab			IL-5 R α	Afucosylated	Enhanced ADCC/ADCP		Asthma	2017
	Inebilizumab (inebilizumab- cdon)			CD19	Afucosylated	Enhanced ADCC/ADCP		Neuromyelitis optica spectrum disorders	2020
	Margetuximab- cmkb			HER2	F243L; R292P; Y300L; V305I; P396L	Enhanced ADCC/ADCP		HER2+ metastatic breast cancer	2020
	Mogamulizumab (mogamulizumab- kpkc)			CCR4	Afucosylated	Enhanced ADCC/ADCP		Mycosis fungoides or Sézary syndrome	2018
	Cipterbin			HER2	Not found	Enhanced ADCC/ADCP		HER2-positive metastatic breast cancer	NA
	Obinutuzumab			CD20	Low-fucose	Enhanced ADCC/ADCP		Chronic lymphocytic leukemia	2013
	Tafasitamab (tafasitamab-cxix)			CD19	S239D; I332E	Enhanced ADCC/ADCP		Diffuse large B-cell lymphoma	2020
	Ublituximab			CD20	Low fucose	Enhanced ADCC/ADCP		Multiple sclerosis	2022
Half-life extension	Amubarvimab +			SARS-CoV-2	M252Y, S254T,	Half-life		SARS-CoV-2 infection	NA
	Romlusevimab				T256E	extension
	Crovalimab			Complement C5	L235R, G236R, S239K, A327G,	Half-life extension		Paroxysmal nocturnal hemoglobinuria (PNH)	2024
					A330S, P331S;
					M428L, N434A,
					Q438R, S440E
	Netakimab			IL-17	M252Y/S254T/T256E	Half-life extension		Plaque psoriasis	NA
	Nirsevimab			RSV	M252Y, S254T,	Half-life		Prevention of respiratory	2023
					T256E	extension		syncytial virus disease
	Pucotenlimab			PD-1	S228P, S254T,	Half-life		Metastatic microsatellite	NA
					V308P, N434A	extension		instability-high (MSI-H) or
								mismatch repair deficient
								(dMMR) advanced solid
								tumors,
	Ravulizumab (ravulizumab- cwvz)			Complement C5	M428L; N434S	Half-life extension		Paroxysmal nocturnal hemoglobinuria	2018
	Sotrovimab			SARS-CoV-2	M428L, N434S	Half-life extension		COVID-19	NA
	Recaticimab			PCSK9	M252Y, S254T,	Half-life		Hypercholesterolemia	NA
					T256E	extension
	Tixagevimab,			SARS-CoV-2	L234F L235E M252Y	Half-life		COVID-19	EUA#
	cilgavimab				S254T T256E P331S	extension
	Levilimab	IL-6 R			E233P/L234V/L235A and M252Y/S254T/T256E	Half-life extension	Inflammation due to COVID-19 infection			NA
Reduced or silenced Fc effector function	Adebrelimab	PD-L1			S228P, F234A, L235A	Remove Fc effector functions	Small cell lung cancer			NA
	Anifrolumab, anifrolumab-fnia	IFNAR1			L234F; L235E; P331S	Remove Fc effector functions	Systemic lupus erythematosus			2021
	Atezolizumab	PD-L1			Aglycosylated (N297A)	Remove Fc effector functions	Bladder cancer			2016
	Batoclimab	FcRn			L234A, L235A	Reduce Fc effector functions	Generalized myasthenia gravis (gMG)			NA
	Cadonilimab	PD-1, CTLA4			L234A/L235A/G237A	Reduce Fc effector functions	Cervical cancer			NA
	Crovalimab	Complement C5			L235R, G236R, S239K, A327G, A330S, P331S; M428L, N434A, Q438R, S440E	Reduce Fc effector functions	Paroxysmal nocturnal hemoglobinuria (PNH)			2024
	Durvalumab	PD-L1			L234F; L235E; P331S	Remove Fc effector functions	Bladder cancer			2017
	Envafolimab	PD-L1			C220S/D265A/P331S	Remove Fc effector functions	Microsatellite instability- high or deficient Mismatch Repair advanced solid tumors			NA
	Epcoritamab		CD20, CD3		L234F x K409R; both chains L234F, L235E, D265A	Reduce Fc effector functions	Diffuse large B-cell lymphoma			2023
	Eptinezumab (eptinezumab- jjmr)		CGRP		Aglycosylated (N297A)	Reduce Fc effector functions	Migraine prevention			2020
	Faricimab, faricimab-svoa		VEGF-A, Ang-2		S354C + T366W x Y349C + T366S + L368A + Y407V; L234A L235A P329G; I253A H310A H435A	Remove Fc effector functions	Neovascular age-related macular degeneration, diabetic macular edema			2022
	Fremanezumab (fremanezumab- vfrm)		CGRP		A330S; P331S	Remove Fc effector functions	Migraine prevention			2018
	Galcanezumab (galcanezumab- gnlm)		CGRP		S228P; F234A; L235A	Remove Fc effector functions	Migraine prevention			2018
	Glofitamab		CD20, CD3e		S354C-T366W x Y349C-T366S-L368A- Y407V; L234A, L235A, P329G	Remove Fc effector functions	Diffuse large B-cell lymphoma			2023
	Levilimab		IL-6 R		E233P/L234V/L235A and M252Y/S254T/T256E	Reduce Fc effector functions	Inflammation due to COVID-19 infection			NA
	Mirikizumab		IL-23p19		S228P; F234A, L235A	Reduce Fc effector functions	Ulcerative colitis			2023
	Mosunetuzumab		CD20, CD3		T366W x T366S- L368A-Y407V; N297G	Remove Fc effector functions	Follicular lymphoma			2022
	Nemolizumab		IL-31 Rα		C222S and H268Q	Reduce Fc effector functions	Pruritus with atopic dermatitis			NA
	Penpulimab		PD-1		L234A L235A G237A	Reduce Fc effector functions	Metastatic nasopharyngeal carcinoma			In review
	Prolgolimab		PD-1		L234A/L235A	Reduce Fc effector function	Melanoma			NA
	Risankizumab (risankizumab- rzaa)		IL-23 p19		L234A; L235A	Reduce Fc effector functions	Plaque psoriasis			2019
	Spesolimab		IL-36 R		L234A/L235A	Reduce Fc effector functions	Generalized pustular psoriasis			2022
	Sutimlimab (sutimlimab- jome)		Complement C1s		S228P; L235E	Remove Fc effector functions	Cold agglutinin disease			2022
	Tagitanlimab		PD-L1		L234A L235A G237A	Reduce Fc effector function	Nasopharyngeal carcinoma, solid tumor			NA
	Talquetamab (talquetamab- tgvs)		GPCR5D, CD3		F405L-R409K x WT; S228P, F234A-L235A	Reduce Fc effector functions	Multiple myeloma			2023
	Teclistamab		BCMA, CD3		F405L-R409K x WT (R409); S228P and F234A, L235A	Reduce Fc effector functions	Multiple myeloma			2022
	Teplizumab		CD3		L234A; L235A	Reduce Fc effector functions	Type 1 diabetes			2022
	Tislelizumab		PD-1		S228P; E233P, F234V, L235A, D265A	Remove Fc effector functions	Esophageal squamous cell carcinoma			In review
	Tixagevimab, cilgavimab		SARS-CoV-2		L234F L235E M252Y S254T T256E P331S	Reduce Fc effector functions	COVID-19			EUA#
	Vedolizumab		α4β7 integrin		L235A, G237	Remove Fc effector functions	Ulcerative colitis, Crohn disease			2014
	Odronextamab		CD20, CD3		WT x H435R-Y436F; E233P, F234V, L235A, G236del; S228P	Remove Fc effector functions	Relapsed/refractory (R/R) follicular lymphoma or R/R diffuse large B-cell lymphoma			In review
	Ivonescimab		PD-1, VEGF		L234A, L235A	Reduce Fc effector functions	Lung cancer			NA
Reduced FcgRIIb binding	Margetuximab- cmkb		HER2		F243L; R292P; Y300L; V305I; P396L	Reduce FcγRIIB (CD32B)	HER2+ metastatic breast cancer			2020
Systemic clearance/half- life	Faricimab, faricimab-svoa		VEGF-A, Ang-2		S354C + T366W x Y349C + T366S + L368A + Y407V; L234A L235A P329G; I253A H310A H435A	systemic clearance/half-life	Neovascular age-related macular degeneration, diabetic macular edema			2022

6. Fc-enhanced mAb case studies

6.1. Fc-enhanced mAbs for oncology

6.1.1. Case study 1: non-clinical safety testing of twoFc-enhanced depleting mAbs; mab-W and Mab-X

MAb-W is an IgG1 mAb against a tumor target that blocks ligand binding and receptor dimerization and signaling. It is both afucosylated for enhanced ADCC and IgG1/IgG3 chimeric for enhanced CDC (Accretamab® technology). MAb-W had 16- and 18-fold increased affinity for FcγRIIIA F158 and V158, respectively, and 16-fold increased affinity for C1q relative to parental wild-type IgG1 that corresponded to enhanced ADCC and CDC of tumor antigen-positive target cells in vitro. MAb-W bound epithelium and peripheral nerves in multiple tissues in a tissue cross-reactivity (TCR) study, consistent with tumor antigen expression in healthy tissue. One-month toxicology studies were conducted in rat and cynomolgus monkeys, both of which were considered relevant because of tumor antigen-binding affinity comparable to human. Tissue cross-reactivity and ADCC activity in the monkey were comparable to human, but enhanced effector function of mAb-W in rat was uncertain. In rats, mAb-W was given once weekly for 4 weeks as an intravenous (IV) bolus injection at 0 (vehicle), 5, 50, or 500 mg/kg/week. mAb-related findings included lobuloalveolar atrophy of the male mammary gland, decreased prostate weight, and decreased serum testosterone and luteinizing hormone, all non-adverse and likely pharmacology-related. In cynos, mAB-W was given once weekly for 4 weeks as an IV bolus injection at 0 (vehicle), 3, 30 or 300 mg/kg/week. Findings included abnormal fecal consistency at ≥30 mg/kg/week and mucosal ulcer in the cecum colon at 300 mg/kg. ADA were detected in some rats and monkeys and were associated with reduced mAb-W plasma concentrations, but not immunopathology findings. Even though the tumor antigen epitope is widely distributed throughout the body based on IHC results, particularly in nerve and epithelial tissues, no toxicity was observed in these tissues, aside from those described above. The NOAEL was 500 mg/kg/week in rat, and 30 mg/kg/week in monkey. mAb-W was well-tolerated in a FIH study in tumor antigen-positive cancers.

MAb-X is an afucosylated IgG1 mAb specific for a cell surface protein on activated T cells and other activated immune cell subsets, with nearly 19-fold higher ADCC activity than the wild-type mAb. MAb-X bound monkey target with approximately 10-fold greater affinity than human target, based on SPR assessment. However, the affinity for the respective FcγR IIIa and ADCC activity in an in vitro blood assay was comparable between monkey and human. Tissue cross-reactivity studies did not identify any potential off-target binding, and no cytokine changes were detected in an in vitro cytokine release assay. In a 4-week repeat dose (once weekly) cynomolgus toxicology study, a, adverse marked decrease in neutrophils was observed at 100 mg/kg/week. Non-adverse neutrophil decreases were observed at 30 mg/kg/week and were associated with adaptive granulopoieses responses in the bone marrow. In a 26-week repeat dose cynomolgus toxicology study, non-adverse transient (reversed with continued dosing) neutrophil decreases were observed at 30 or 100 mg/kg/week. While target expression was not detected in neutrophils, weak binding of mAb-X was observed predominantly through FcγRII or FcγRIII on monkey and human neutrophils, respectively. Non-adverse findings from either the 4 week or 26-week toxicology included injection site reactions in subcutaneous (SC) dose groups, minimal to mild Kupffer cell activation, and decreases in NK cells and platelet count. Non-adverse mild nonspecific changes in other hematologic or clinical chemistry parameters were observed in individual animals in the 4-week study but not the 26-week study. Target-related PD findings included in the studies are not described here.

6.1.2. Case study 2: assessing safety of an Fc-enhanced B cell depletor in huFcγR Tg mice

NVS32b is an afucosylated IgG1 mAb specific for FcγRIIB (CD32B). CD32B expression on malignant B cells is known to provide a mechanism of resistance to rituximab.¹⁷⁶ Cynomolgus monkeys were not appropriate for non-clinical safety assessment because the in vitro binding and depletion profile was not comparable to human (NVS32b did not deplete cyno B cells but instead cyno monocytes). Human FcγRIIB transgenic mice were inadequate because NVS32b did not activate mouse FcγRIV, the orthologue to human FcγRIIIA, resulting in no depletion of mouse B cells. Hence, human FcγR transgenic mice (huFcγR (‘Ravetch’) mice) were used to evaluate NVS32b toxicity after extensive characterization. These mice express all human FcγR similar to humans,¹⁵⁸ but deviations from human were observed, such as high huFcγR expression was observed on huFcγR mice myeloid cells (especially high frequency of CD32b+ monocytes and granulocytes opposed to low frequency in human); constitutive FcγRI on neutrophils opposed to inducible FcγRI on human neutrophils, low FcγRII expression observed on T cells (absent on human T cells). HuFcγR transgene copy number varied highly and unintended integration of additional human genes was observed. The predictive value of this model for human risk assessment may be affected by the deviations and therefore should be carefully considered. NVS32b showed pharmacological activity, as observed by B cell depletion. The number of neutrophils and monocytes were also decreased upon NVS32b dosing in huFcγR tg mice, which was not observed in human, assessed using in vitro whole-blood depletion assays.¹⁷⁷ Unexpectedly, NVS32b induced severe systemic thrombosis in most animals. These thrombotic events observed in huFcγR transgenic mice were caused by NVS32b binding to platelet CD32a (found to be expressed more strongly in these transgenic mice compared to human) in an Fc- and complementary-determining region (CDR)-dependent fashion, causing activation and aggregation of platelets in vitro. On the contrary, in humans, NVS32b only bound to platelets if pre-activated, engaging an unknown epitope only exposed upon activation via the Fc and CDR. Therefore, this mouse model appears to be over-predictive, but was useful to identify this important vascular safety liability and to trigger further evaluations to understand the human risk.¹⁷⁸

6.1.3. Case study 3: non-clinical safety testing of aFc-enhanced mAb that depletes tumor cells

MEDI-531 is an IgG1 mAb specific for EphA2, a receptor tyrosine kinase that is overexpressed in a variety of human tumor types. MEDI-531 is an agonist that induces EphA2 internalization and degradation, thereby abolishing its oncogenic effects.¹⁷⁹ MEDI-531 contains a triple mutation (3 M) in the Fc domain to increase the binding affinity to human FcRγIIIA. MEDI-531 demonstrates potent ADCC activity in vitro and enhanced anti-tumor effect in vivo. Single and repeat dose toxicity studies were performed in cynomolgus monkeys in which MEDI-531 was highly immunogenic. Premature deaths were observed, consistent with hypersensitivity/anaphylaxis. High frequency and very high titer ADA responses were seen which included IgG, IgM, and IgE isotypes. Other toxicity findings included acute phase inflammatory response and chronic reticulo-endothelial system stimulation. The 3 M Fc modification, which was intended to enhance ADCC function by increasing affinity to FcRγIIIA, also increased binding to other FcγR subtypes such as FcγRIIB and FcγRI. This indiscriminate FcγR-binding caused increased binding to a broad range of cell types in blood, including NK cells, neutrophils and monocytes compared to the unmodified Fc parental mAb. The Fc modification may have exacerbated the cynomolgus monkey ADA response, leading to the immune-related adverse effects observed. This project was terminated prior to clinical development for inadequate risk-benefit.

6.2. Fc-enhanced mAbs for non-oncology indications

6.2.1. Case study 4: non-clinical safety testing of aFc-enhanced mAb that depletes eosinophils

Benralizumab (Fasenra®) is an afucosylated IgG1 mAb against interleukin-5 receptor alpha (IL-5 Rα), which is expressed on eosinophils and basophils. It is approved for the treatment of eosinophilic asthma.¹⁸⁰ It has high affinity for FcγRIIIA, leading to ADCC-mediated killing and apoptosis by NK cells. Additionally, by targeting the receptor, benralizumab also blocks IL-5 binding.¹⁸¹ Although its binding affinity for IL-5 Rα and its potency to inhibit IL-5–induced cell proliferation were indistinguishable from those of benralizumab, the fucosylated parental mAb did not induce target-cell ADCC/apoptosis.¹⁸¹ The binding affinity of benralizumab to cynomolgus monkey IL-5 Rα was within 3-fold of the binding to human IL-5 Rα. Five nonclinical safety studies were conducted in cynomolgus monkeys following IV and SC administration of benralizumab. These include a 9-week repeated-dose IV study with an 18-day recovery, a single-dose SC study, a 15-week repeated-dose SC study with a 12-week recovery, a 9-month IV and SC study with a 12-week recovery, and an enhanced pre- and post-natal development (ePPND) reproductive toxicology study. As expected, eosinophil counts were markedly decreased in benralizumab-treated animals throughout the dosing and recovery periods. In the 9-week repeat-dose IV study, transient decreases in leukocytes due to differential decreases in neutrophil counts were seen in 2 of 10 animals (one male and one female) treated with the highest dose (30 mg/kg). Decreased neutrophils were not observed in subsequent studies of longer duration. In the ePPND study, the only benralizumab-related effect was depletion of eosinophils (an expected pharmacological effect), which resolved by 180 days of life in all but 1 infant in the 30 mg/kg dose group.¹⁸² No other immune parameters assessed were affected. The PK of benralizumab in cynos was dose-proportional and the PK parameter values were typical of an IgG1 mAb without an antigen sink. Upon repeat dosing, some animals developed ADA, resulting in reduced exposure. Tissue cross-reactivity studies with cyno and human tissues showed similar patterns of staining with benralizumab. This nonclinical safety profile is similar to the molecule’s clinical safety profile. In both species, eosinophil depletion was evident, with recovery of depletion observed over time. Additionally, though the non-human primate is not generally considered to be predictive of immunogenicity in humans, benralizumab was immunogenic in some monkeys and this has been reflected in the human immunogenicity profile.

6.2.2. Case study 5: non-clinical safety testing of a Fc-enhanced mAb that depletes B and T cell sub-sets

11G2 is an afucosylated IgG1 mAb specific for CXC chemokine receptor type 5 (CXCR5), a G protein-coupled receptor expressed by B cells, T follicular helper (Tfh) cells, and circulating Tfh-like (cTfh) cells. CXCR5 plays an important role in the formation of autoreactive germinal centers important in driving the pathology of autoimmune diseases.¹⁸³ 11G2 demonstrated potent antagonism of CXCL13-dependent signaling in a cAMP reporter assay and robust ADCC of B and cTfh cells in human PBMC cultures at much lower concentrations than the fucosylated parental CXCR5 mAb. Cynomolgus monkey was a relevant toxicology species since 11G2 showed enhanced target cell killing compared to the fucosylated parent mAb. In cyno toxicology studies, a single dose of afucosyl 11G2 caused dose-dependent depletion of B and cTfh cells from the peripheral blood, and decreased B cells and bona fide Tfh cells in the spleen and other lymphoid tissues. Depletion of peripheral blood B cells and Tfh occurred early (within 4 hours) and potently with approximately 50% decrease of peripheral B cells and Tfh observed at 0.001 mg/kg. These decreases were accompanied by smaller decreases in NK cells. No cytokine release was observed. The depletion of CXCR5+ cells was profound and needed up to 10 months for the target cells to rebound to close to baseline levels in the monkeys. 11G2 induced a decrease in the T cell-dependent antibody response, consistent with a depletion of Tfh cells. No infections were noted despite prolonged B and Tfh cell depletion.

6.3. Fc silencing case studies

6.3.1. Case study 6: use of human FcRn transgenic mouse models to compare Fc-engineered (silenced) mAbs and predict their human PK

Conventional pharmacology studies in species such as rats and monkeys cannot accurately predict human PK due to the differences in affinity between the human IgG-FcRn interactions and those of other species (e.g., rodents have ~10× greater affinity, while monkeys have ~3× greater affinity). Therefore, more accurate models, such as hFcRn transgenic mouse models, must be used to predict human PK more precisely. This case study describes efforts to find suitable animal models to predict the PK of Fc-engineered antibody molecules in humans. Furthermore, it describes the use of these models to explore the effect of Fc silencing mutations on the stability of a clinically relevant mAb. The models used for this testing were Tg32 (B6.Cg-Fcgrt^tm1Dcr Tg(FCGRT)32Dcr/DcrJ (JAX strain 014565)), Tg32-SCID (B6.Cg-Fcgrt^tm1Dcr Prkdc^scid Tg(FCGRT)32Dcr/DcrJ (JAX strain 018441)), and Tg276 (B6.Cg-Fcgrt^tm1Dcr Tg(CAG-FCGRT)276Dcr/DcrJ (JAX strain 004919)).

To evaluate the performance of the different models, we selected an Fc-silenced hIgG1 mAb with HLE mutations introduced within the Fc region (mAb Y) as a tool molecule. mAb X was evaluated in humans, hFcRn Tg mouse models and monkeys. Since mAb Y does not cross-react with the murine target, its PK is expected to be linear and unperturbed by TMDD in the hFcRn Tg mice. Projection and evaluation of human PK parameters from the mouse and monkey models was carried out in a three-step process: 1) PK parameters were estimated by fitting a linear two-compartment model to Tg mouse data, 2) allometric scaling was carried out on CL and V parameters, and then 3) PK parameters were compared to monkey and human population PK parameters and exposures. mAb Y was administered at a 10 mg/kg IV dose where serum concentrations were evaluated for 60 days.

Consistent with prior literature observations,¹⁸⁴ mAb Y demonstrated more rapid clearance (CL) in Tg276 mice than Tg32 mice and similar CL between the two Tg32 mouse strains. A scaling exponent of 1 was used for V, while the scaling exponents used for the four CL cases were as follows: 0.93 for Tg32 mouse to human,¹⁷⁰ 0.91 for Tg32 mouse-to-monkey,¹²¹ 0.85 for Tg276 mouse to human, and 0.85 for monkey-to-human.¹⁸⁵ Table 10 summarizes the general findings of the PK analysis. These data suggested allometrically scaling PK parameters from Tg276 and applying a CL scaling exponent of 0.85 aligned with the human population PK parameters for mAb Y. Furthermore, simulations determined allometrically scaling resulted in a 59% overestimation of the area under the curve (AUC) using Tg32 mice, 33% overestimation for Tg32-SCID mice, 8.3% overestimation for Tg276 mice, and 72% overestimation for monkeys. Based on these findings, Tg276 was a more appropriate model for predicting human PK parameters for mAb X. Although it was found that Tg276 was a better-suited model for mAb Y, these results contradict prior literature observations that suggested Tg32 is a better animal model for human PK projection. Considering these results, investigations are ongoing to assess the performance of Tg276 vs. Tg32 mice in projecting human PK parameters.

Table 10.

Key biophysical parameters assessed for a panel of eight Fc-silenced mutant antibodies.

Candidate	USP	DSP	HMW (e.g., Aggregation)	Biophys. Properties	Clipping/Truncation/Integrity	Liquid mini-stability	Cumulative DAS Rating	FcgR-binding	C1q-binding
Z1	*	**	**	**	**	**	**	**	***
Z2	**	*	*	*	***	*	***	***	*
Z3	*	*	*	*	*	*	*	*	*
Z4	*	*	*	*	*	*	*	*	*
Z5	*	*	**	**	**	**	**	***	*
Z6	*	**	***	**	***	***	***	***	**
Z7	**	**	**	**	*	*	*	*	*
Z8	*	**	*	**	*	*	**	*	*
WT	*	*	*	*	*	*	*	***	***

* Indicates acceptable; **Indicates caution; ***Indicates no-go.

6.3.2. Identification of function-silenced Fc variants and testing their PK parameters in hFcRn transgenic mice

Several clinically approved therapeutic IgG1 antibodies utilize the LALA (L234A/L235A) Fc variants. However, data has emerged that, while they retain FcRn binding to maintain normal half-life, they still bind with low affinity to Fcγ receptors (mainly FcγR1).¹⁰⁶ To identify mutations that abolish the Fc function completely, a stringent screening assay based on the in vitro binding to different Fcγ receptors was used to identify an Fc variant, L234S/L235T/G236R (STR), which was assembled with the specificities to CD20, CD3, and CD52. Using multiple in vitro assay methods, we demonstrated that the STR variant for all 3 targets showed lack of binding to human FcγRs immobilized to a surface or expressed by Jurkat cells, no binding to immobilized C1q, no in silico-predicted immunogenicity signal, no in vitro immune stimulation of naive human T cells when co-cultured with APCs (low predicted immunogenicity risk) and no cytokine release in a functional assay with human PBMCs. In addition, there was good thermal stability, no aggregation by size exclusion chromatography (SEC), and retention of FcRn binding measured by SPR assay. A remaining question was the impact of the new STR Fc variant (L234S/L235T/G236R) on the serum half-life in vivo. This was addressed by comparing the half-life of three different variants of an anti-CD20 antibody (WT, LALA, STR) in mouse models proven to model human in vivo antibody PK accurately.

The models used for these experiments were described earlier in this article. In addition, we used Tg32-hFc (B6.Cg-Tg(FCGRT)32Dcr Fcgrt^tm1Dcr Ighg1^{em2(IGHG1)Mvw}/MvwJ (JAX strain 029686) that also contains a knock-in genetic modification of the Ighg1 (immunoglobulin heavy constant gamma 1 (G1m marker)) locus that replaces the mouse hinge and Fc domain with the equivalent human IgG1 domains (Ighg1^{em2(IGHG1)Mvw}). The Tg32-hFc mice produce mouse IgG Fab2-human IgG1 Fc- chimeric antibodies in serum concentrations at physiologically relevant levels, providing an endogenous source of competing human IgG1. When tested in the Tg32 human FcRn-bearing mice, all the anti-CD20 antibody variants showed half-lives between 9 and 11 days, with the WT version having a slightly longer half-life. These data indicate that the STR mutations do not reduce the stability of the antibody in vivo.

The Tg32 mouse lacks the expression of human antibodies that might compete against the binding of engineered antibodies to FcRn. Without any competition from human IgGs, slight decreases in the affinity to FcRn might affect the stability of an engineered antibody in humans. To evaluate the potential effects of human IgG1 competition, we used the Tg32-hFc. As expected, the presence of human immunoglobulins decreased the half-life of all test articles from about 10 days to approximately 5 days. Notably, the STR displayed a similar PK to WT and LALA variants of the anti-CD20 mAb, suggesting that the silencing variant mutations do not affect the affinity parameters of the interaction with FcRn in vivo.

Finally, the STR and LALA variants of the anti-CD20 antibody were compared to the WT antibody in Tg276 mice. The unique expression pattern of the human FcRn protein in the Tg276 model decreases the measured antibody plasma half-lives and exacerbates minor differences between the PK of similar antibodies. Using the Tg276 model, we could detect a significant half-life difference between the mutant antibodies and WT rituximab. Remarkably, the STR mutation half-life was identical to the LALA variant.

In conclusion, using a clinically tested engineered monoclonal antibody (mAb X), we demonstrated that FcRn-humanized animal models can effectively predict the human PK of therapeutic mAbs. Among the transgenic lines evaluated, the Tg276 model, which expresses human FcRn at higher levels, outperformed both the well-characterized Tg32 model and non-human primates in predicting human PK. Although Tg276 mice exhibit faster clearance than Tg32 mice, ongoing studies aim to further investigate the reasons behind this observation. Additionally, we tested the use of the different transgenic lines to explore the effect of silencing mutations on the half-life of engineered IgGs. Using a combination of models, we demonstrated that a new Fc-silencing modification, the STR mutation, does not dramatically affect the plasma half-life of a clinically relevant IgG1 and is at least as stable as the well-known LALA variant.

6.3.3. Case study 7: developability and characterization of Fc-silenced mAbs in vitro and in vivo in hFcRn transgenic mice

Modification of Fc regions can potentially lead to developability, stability, PK and immunogenicity complications. This case study outlines a workflow to select novel FcγR silencing mutation motifs utilizing the tool antibody, mAb Z. Eight IgG1 mAbs (Z1–8) were produced with the various FcγR silencing mutations and compared to WT Fc with no silencing and a positive control mAb with a common silencing motif (DAPA). First, developability and stability were assessed. Upstream/Downstream process development (USP and DSP) focuses on mAb expression from CHO cells compared to WT and DAPA mAbs. Drug substance production focuses on recovery and impurities (i.e., HMW and LMW) as measured by SEC following an initial purification and polishing process. Quality Attributes are focused on the analysis of impurities in the final material by SEC. LC-MS profiling examines the mass, glycosylation, clipping, sequence variants, and free thiols (if applicable). Biophysical properties assessed are diffusion interaction parameters, viscosity, hydrophobicity by HIC, melting temperature by Differential Scanning Calorimetry (DSC), unfolding onset by DSC, isoelectric point (pI) by capillary isoelectric focusing (cIEF), and acidic and basic peaks by (cIEF). The results are summarized in Table 11. In addition, binding to hFcγR, C1q, and FcRn was evaluated using SPR. Notably, molecules Z2, Z5, and Z6 still demonstrated high-affinity binding to FcγRI while other silencing motifs demonstrated no detectable binding to FcγRs (Table 12). Modest C1q binding was observed with X1 while X4 and X6 had residual binding. The novel silencing motifs did not influence FcRn binding kinetics. Following the initial screening process, the top four candidates Z3, Z4, Z7 and Z8 were further tested in an in vitro functional assay to assess crosslinking of luciferase reporter gene target cells with THP-1 cells, which naturally express FcγRI, FcγRII, and FcγRIII. Based on the EC50, the DAPA positive control was 3-fold more silent than WT, while Z3, Z4, Z7, and Z8 were 4000-, 2000-, 1000-, and 200-fold more silent than WT, respectively. Furthermore, all molecules demonstrated no measurable difference in ADCC and CDC activity in vitro. Finally, in an in vivo PK study in the Tg32-SCID hFcRn transgenic mice, these four candidates showed similar PK compared to the WT antibody. The Fc-silencing motif, Z7, was selected as the internal lead candidate based on the above results. Currently, in vitro studies are being conducted to assess the immunogenicity risk of the Z7 motif.

Table 11.

Assessment of binding to hFcγR and C1q and influence of FcRn of the eight Fc mutants.

Candidate	Binding to hFcγRs	Binding to C1q	Influence on hFcRn
Z1	N.B.	moderate affinity	no influence
Z2	high affinity	no binding	no influence
Z3	N.B.	no binding	no influence
Z4	N.B.	residual binding	no influence
Z5	high affinity	no binding	no influence
Z6	high affinity	residual binding	no influence
Z7	N.B.	no binding	no influence
Z8	N.B.	no binding	no influence
WT	standard binding	standard binding	no influence

Table 12.

Summary of health authority feedback received by one company relating to Fc-modified antibodies.

Target type (ROA)	MOA	Mutation/Fc Engineering	HA/Comments	Feedback After Response
Soluble (IV) Membrane (IV)	Antagonist	IgG1 LALA	Several authorities requested ADCC/CDC and C1q binding. Data submitted was using a platform approach. Provide molecule specific information. Molecule specific data had to be generated. Provide data demonstrating absence of binding to FcγRs	HAs did not accept a “platform” approach. Molecule specific data was accepted. Data submitted.
Membrane (IVT)	Agonist	IgG1 LALA	Provide a detailed explanation to explain your conclusion that testing Fcg binding. ADCC, CDC is not warranted to characterize potential toxicity.	FcγR and C1q binding, ADCC, CDC data was being generated. Project terminated before data submitted.
Membrane (IVT)	Antagonist	IgG1 LALA	In the IND, please explain your lack of concern (prior to availability of product characterization data) for the potential for Fc-mediated activity, ADCC and CDC.	Data generated.
Soluble oligomer (IVT)	Antagonist	IgG1 LALA	Address the potential for Fc-mediated activity, particularly ADCC and CDC. We infer that the designation “IgG1KO” is intended to denote modification of the antibody structure to prevent Fc functions. From a nonclinical perspective, previous data for the antibody backbone demonstrating lack of activity is sufficient. Otherwise, we recommend testing ADCC and CDC. The Fc modification contains a lysine to alanine mutation to minimize Fc effector function. Provide characterization data to support the lack of binding to relevant FcγR and C1q.	Provided written rationale. Data generation in progress.
Soluble oligomer (sub- cutaneous)	Neutralizing	IgG1 WT	No comment.	-
Membrane bound (IV)	Antagonist	IgG4 Pro	FcRn should be considered in characterization and comparability studies (ADCC/CDC had been originally submitted).	FcRn binding was characterized.
Membrane/Membrane (IV)	Agonist	IgG1 LALA	Consider performing Fc segment affinity and in-vitro ADCC/CDC activity validation study during the clinical trial stage.	Project terminated before data submitted.
			The molecule has a mutated Fc portion to minimize Fc receptor function. However, Fc binding activities (FcγR1, FcγRRII, FcvRIII, FcRn or C1q binding) are not reported and no ADCC/ADCP, CDC assay results are included. The sponsor should provide these data.	FcγR binding data, ADCC and CDC binding data submitted. Response accepted.
Membrane/Membrane (IV)	Cell killing	IgG LALA	Provide data during development to support the lack of binding to FcγR and C1q.	ADCC/CDC assays performed and submitted.
Membrane/Membrane (sub-cutaneous)	Dual agonist	Hybrid	Data submitted for ADCC. Request for ADCP data.	Planned.
Soluble monomer (IV)	Blocking antibody	IgG WT	Regarding potential Fc-mediated effect, the IB states that with the target being a soluble circulating target, ADCC or CDC response is not anticipated. The Applicant should provide a more elaborate justification and risk assessment, also in view of the tissue cross reactivity results.	Written response: The molecule is highly specific to the target and does not bind to its closest related membrane bound protein, thus it should remain in circulation and ADCC/CDC risk is low. An assay to evaluate a soluble target that has no membrane forms would be highly artificial and not relevant. ADCC/CDC experiments in the relevant tissue would be technically challenging
				and we would have to confirm expression through the assay as the epithelial cells are regenerative in vivo. Also, observed staining in tissue cross reactivity was low intensity and frequency.
Membrane/Membrane (IV)	Cell killing	IgG1 LALA	Confirm absence of effector function (binding to FcγR and complement), according to the replacements done to the Fc region.	Written response contained SPR data, reference to published data using IgG1 LALA mutations that show strongly reduced/abolished ADCC/CDC activity. Data needed to be generated and was submitted.

6.3.4. Case study 8. Functional characterization and safety assessment (target cross-linking/agonism risk) of Fc-silenced biotherapeutic in vitro and in vivo in huFcγR Tg mice

Targeting of a highly validated protein (Target X; linked to the pathogenesis of multiple autoimmune diseases), with a mAb containing a wild-type IgG4-Fc (Anti-X#1-IgG4) exhibited signs of systemic inflammatory response in early clinical trials, which were not observed in cyno toxicity studies. This wild-type human IgG4-Fc biologic did not have cyno cross reactivity at the level of the Target X binding to antigen. Toxicity studies were thus conducted using a ‘cyno-surrogate’ of this wild-type IgG4-Fc in which the antigen-binding portion of the molecule was substituted with a cyno cross-reactive version. No toxicity-related finding was observed in these studies. To understand the trigger behind this observation in the clinic and potentially de-risk it in back-up candidates with Fc-silencing mutations, a thorough and systematic investigation was conducted to identify in-vitro signals of immune activation. Thirty Fc-silencing mutants were designed and tested as singles and combinations, including the commonly employed mutations N297A, D265A, LALA, IgG4-S238P. All these variants were evaluated individually in a panel of assays including SPR (binding to FcγR, FcRn), ELISA (binding C1q), FACS, immature dendritic cell (iDC) activation assays, ADCC, ADCP, CDC, in-silico and in-vitro immunogenicity, and developability assays. Target X was selected for this exercise, as it was known to be highly sensitive to clustering-mediated agonism. In iDCs derived from ~30% of donors stimulated with Anti-X#1-IgG4, there were signals of increased immune cell activation (IL-6, increased CD80/86, CD54). Additionally, to further increase the sensitization of the iDC activation assay, an excess of CD32 (FcγRII)-expressing CHO cells were added to the iDC assay to test the potential for FcγR-mediated target clustering. Under these conditions 100% of donors exhibited iDC activation with a greater increase in cytokine production and CD80/CD86 upregulation was observed, highlighting the potential importance of FcγR-mediated clustering-based agonism of this target. Interestingly, the same antigen-binding moiety paired with a different silenced Fc (Anti-X#4 Fc-IgG1-A) did not trigger this response. Expectedly, an agonist antibody also produced a similar response in the iDC activation assays. In additional experiments, activation of ADCP by macrophages was also noted. Also, consistent with the potential for FcγR-mediated Target X signaling, Anti-X#1- Fc-IgG4 was able to induce activation of Target X in an NFκB response in a reporter cell line only under conditions of co-expression of CD32.

Our studies also revealed that FcγR-mediated agonism of Target X was context-dependent, that is, it is influenced by both the Fc and the antigen targeted by the mAb. In additional iDC experiments, iDC activation was observed as measured by CD54 upregulation with Anti-X#3 with an IgG1-LALA Fc mutation, but not with N297A (aglycosylated) mutation. Interestingly, the same LALA-Fc mutation was silent in Anti-X#2-IgG1-Fc, but not in Anti-X#3-IgG1-Fc. Other data showed that with highly sensitive targets, even typically ‘silent’ Fc mutations such as IgG1.1-Fc and IgG1.3-Fc (internal company mutations) can yield clustering-driven agonism, despite greatly reduced binding to FcγRs, particularly under conditions of a high level of CD32-mediated clustering. However, we also noted that there were other mutations which could produce silent molecules.

The unexpected observation that typically Fc silent mutations were insufficient for Target X molecules triggered additional Fc-mining work to identify next-generation silent Fc for this target, which led to the selection of mutant isotype IgG1-Fc-M1. In SPR studies, IgG1- Fc-M1 was found to be ~100× weaker than WT IgG1-Fc in binding to FcγR1, and, remarkably, undetectable binding to all the other all FcγRs. In the same set of studies, our standard Fc-silenced mutant IgG1.3-Fc was found to bind weakly but measurably to FcγRI and to both variants of FcγRIIA at high concentrations. To further de-risk IgG1-Fc-M1, a novel cross-linking (avidity) assay was established, in which crosslinking was triggered in the presence of antigen. While wt-IgG1-Fc and IgG1.3-Fc showed enhanced FcγR binding when cross-linked through binding to an antigen, IgG1-Fc-M1 was inert. Finally, in iDC activation assays, we were able to identify antibodies to target X, which were exquisitely silent. In a single donor study, representative of more than 10 donors, it was clearly demonstrated that IgG1-Fc-M1 provided a truly silent molecule with no iDC activation, either alone or with the addition of FcγRII expressing-CHO cells.

Following these in-vitro studies, conclusive evidence for the potential for in-vivo Fc-mediated cross-linking came from studies in huFcγR × hu Target X transgenic mice. We conducted acute studies in these mice dosed at a single high dose of 100 mg/kg, ensuring 100% RO with all drug molecules: Anti-X#1 IgG4- Fc original molecule, Fc mutant molecules: Anti-X#2 IgG1-Fc-LALA, Anti-X#3 IgG1-Fc-N297A, or novel inert Fc molecule: Anti-X#4 IgG1-Fc-M1. Exposures were comparable for all the evaluated molecules, when dosed at 100 mg/kg. As expected from the cellular studies, for Anti-X#1 IgG4-Fc (original molecule), within 24 h, clear signs of toxicity were observed, such as decreased platelets, lymphocytes, and eosinophils and increased neutrophils with red/dark discoloration of spleen that correlated microscopically with congestion, an increased size of gallbladder that corresponded microscopically with distention or mucosal edema, drug-related increase in spleen weights (20–49%), hematopoietic necrosis, lymphoid necrosis, hepatocellular single-cell necrosis/coagulative necrosis and fibrin thrombosis in multiple tissues. On the other hand, no drug-related clinical observations or changes in hematology, organ weights, gross lesions, or microscopic histopathology (no evidence of thrombi) was seen for Anti-X#2 IgG1-Fc-LALA, Anti-X#3 IgG1-FcN297A, or novel inert Fc molecule: Anti-X#4 IgG1-Fc-M1.

To summarize, in this case-study we inferred that FcR-mediated clustering could potentially be responsible for the systemic immune response observed in clinic. Retrospectively, as the proposed IgG4 FcyR-mediated clustering Target X activation was not observed in cyno, we can hypothesize that in cyno there may be subtle differences in the cellular expression and localization of either FcyR or Target X receptors that prevented the observation of an inflammatory response. Of note, in our clinical study, there was subject variability of the observed inflammatory response, indicating the potential requirement for specific circumstances for the response to be observed. Additionally, our thorough and systematic investigation revealed that commonly used silent Fc sequences may not be silent enough for some targets, i.e., FcγR-mediated agonism is both target epitope- and Fc sequence-dependent. Specialized human cell-based in-vitro models, SPR studies, as well as alternative in-vivo models such as huFcγR transgenic mice can be useful tools in selecting and de-risking potential candidates.

6.3.5. Case study 9: functional characterization and safety assessment (platelet activation risk) of an Fc-silenced mAb in vitro and in vivo in huFcγR Tg mice

CFZ533 (iscalimab) is a fully human, non-depleting (Fc-silenced) IgG1 anti-CD40 antagonist mAb for the treatment of inflammatory diseases. CD40 is constitutively expressed on B cells and other APCs, including monocytes, macrophages, DCs, as well as by platelets, and inflamed parenchyma. Binding of CD40 by its ligand CD154 (CD40-L) results in DC maturation, monocyte survival, and cytokine secretion by many cell types and plays a key role in germinal center (GC) formation, memory B cell development, antibody production, immunoglobulin (Ig) isotype switching, and affinity maturation.¹⁸⁶ Since CFZ533 is required to block the proinflammatory functions of CD40 on immune cells, the activation of CD40-expressing cells (e.g., through receptor cross-linking) and/or the depletion of these cells, and the associated pro-inflammatory effects, should be avoided. Hence, an N297A mutation (aglycosyl) was incorporated in the Fc domain of CFZ533 to abolish FcγRIIIA and C1q binding,¹⁰¹ thereby minimizing the risk of ADCC- and CDC-mediated killing of CD40-expressing cells. A silenced Fc also minimizes binding to FcγRIIB that could promote CD40 agonism, which occurs through enhanced receptor cross-linking following Fc binding to FcγRIIB. A silenced Fc also prevents CFZ533 from binding to and activating FcγRIIA on platelets and so circumvents the potential risk of platelet aggregation and associated thromboembolism (TE), observed with anti-CD154 antibodies.^53,187,188

In vitro studies confirmed that CFZ533 was unable to bind to the recombinant or cell surface FcγRIIIA in contrast to the parental (non-Fc silenced) mAb HCD122 (wild-type IgG1). CFZ533 did not mediate ADCC of CD40-expressing human cells, including B cells, in whole blood, or mediate CDC of Raji B cells in the presence of rabbit complement, both of which were observed with the depleting anti-CD20 mAb rituximab. Blood lymphocyte counts, including CD20-positive B cells, were normal in all NHP toxicity studies throughout the treatment phase with CFZ533, although moderate and recoverable reduction of CD20-positive B cells in the blood was observed at high dose levels after longer-term dosing. In contrast, the parental mAb HCD122 led to a prominent decrease in CD20-positive B cells. CFZ533 completely suppressed primary and secondary antibody responses to immunization with the T cell-dependent antigen Keyhole limpet hemocyanin and abrogated GC formation in lymphoid organs in the absence of significant B cell depletion.^189,190 CFZ533 inhibited CD154-induced activation and proliferation of human leukocytes in vitro, but, unlike soluble CD154, it had no agonist activity, failing to stimulate proliferation nor CD69 upregulation of human PBMCs, either alone or in combination with other co-stimulatory signals, or when cross-linked by human FcγRs.¹⁸⁹ Unlike sCD154, CFZ533 did not stimulate cytokine production from human monocyte-derived DC or induce ICAM-1 upregulation or MCP-1 release from CD40-expressing human endothelial cells and cyno toxicology studies failed to demonstrate any evidence of CD40 pathway activation such as systemic cytokine release.

Effects on platelet function (whole-blood aggregometry, WBA) and blood hemostasis (thromboelastography, TGE) were investigated in human and cyno blood. CFZ533, either alone or as preformed complexes of CD40/anti-CD40 mAb did not induce platelet aggregation but rather displayed some mild inhibitory effects on platelet aggregation at a high concentration, whilst CD154/anti-CD154 mAb complexes did induce platelet aggregation in vitro. In addition, huFcγR transgenic mice¹⁵⁸ were used to investigate the ability of CFZ533 to induce TEs in vivo in comparison to an anti-CD154 mAb (with active Fc) as a positive control (note: these mice were used because CFZ533 does not bind mouse CD40 and normal mouse platelets, do not express FcγRIIA, unlike human platelets which do). Mice injected with soluble CD154/anti-CD154 mAb immune complexes (ICs) showed severe signs of sickness, a significant decrease in platelets and had to be sacrificed. Post-mortem histology revealed thrombi formation in the lung, confirming previously published data.¹⁹¹ In contrast, mice injected with soluble CD40/anti-CD40 mAb ICs were not affected and showed no signs of TE. Further, no evidence of CFZ533-mediated effects on WBA and TGE were observed using blood samples from kidney-transplanted cynos treated with CFZ533 and Cyclosporin A. In addition, toxicology studies with CFZ533 in either cyno or rhesus monkeys (more sensitive to anti-CD40-mediated platelet effects¹⁹² did not reveal any significant toxicities, including no TEs and blood coagulation parameters and platelet counts were normal with no evidence of platelet activation.¹⁹⁰ In the clinic, CFZ533 was safe and well-tolerated with CD40-expressing target cell depletion, nor evidence of T cell activation, cytokine release nor risk of TEs.¹⁹³

6.3.6. Case study 10: the potential for platelet activation should be considered when a therapeutic antibody targets a platelet expressed receptor

Mechanistically, platelet activation by a therapeutic antibody may occur through: (1) direct receptor activation (i.e., agonism), (2) receptor clustering, or (3) Fc receptor crosslinking. Platelets are known to express the activating FcγRIIA, CD32a, and antibody-mediated activation of this receptor is thought to occur through formation of a bridge between the targeted antigen (i.e., through the Fab region of the antibody) and the FcγRIIA either on the same platelet or adjacent platelets. In our studies, we were interested in understanding the contribution of different Fc backbones to FcγRIIA-mediated platelet activation in both human and cynomolgus monkey (cyno) platelets. For this purpose, we generated antibodies with different Fc backbones (i.e., IgG1 wt, IgG1 LALA, IgG4(P)) against CD9, as well as Target X, Y, and Z (all targets are known to be expressed by platelets). We evaluated platelet activation using flow cytometry methods, by gating for platelets via CD41 (GP IIb/IIIa), CD42a (GPIX) and CD61 (vitronectin receptor) and determining platelet activation in human and cyno via PAC-1 (a mouse monoclonal antibody binding to activated GPIIb/IIIa only) in ex vivo whole blood. In some experiments with human platelets, activation was instead measured via detection of upregulation of CD62P (P-selectin) in ex vivo whole blood. To evaluate surface expression of the selected platelet receptor and binding of the antibodies to the expected targets in both human and cyno, we used fluorophore labeled commercial and in house antibodies including respective isotype controls.

The first example evaluated CD9 targeting antibodies. Similar levels of CD9 expression were demonstrated in both human and cyno platelets. In both human and cyno whole blood, significant activation of platelets was observed with anti-CD9 IgG1 WT, while platelets were not activated in the presence of an anti CD9 antibody with an IgG1 LALA mutated backbone that reduces interaction with FcγRIIA. Significant platelet activation was also observed following exposure to an anti-CD9 IgG4(P) antibodies in both human and cyno whole blood. Blocking with an FcγRIIA antibody abolished the activation observed with the IgG4(P) antibody in human whole blood in vitro. The second group of case studies highlights two examples of targets expressed on platelets in both human and cyno whole blood at similar levels, but species differences in activation patterns, dependent on antibody target. Exposure to an IgG4(P) antibody targeting X, an antagonist monoclonal antibody against an immunoregulatory receptor, resulted in human platelet activation, but not cyno platelet activation in vitro. In contrast, an IgG1 WT antibody targeting Y, an antagonist monoclonal antibody against an immunoregulatory receptor resulted in platelet activation in both human and cyno ex vivo whole blood, while changing the backbone of the antibody to IgG4(P) diminished activation in human, but not cyno platelets. In an additional case, Target Z, an antagonist monoclonal antibody against a transmembrane glycoprotein, was expressed to similar levels in both human and cyno platelets. A wild-type IgG1 antibody against Target Z resulted in activation of human and cyno platelets, although to a lesser extent in the cynomolgus monkey. In this case an IgG1 LALA mutated antibody eliminated platelet activation in both species. Interestingly, the observed species differences in activation did not seem to be consistently missing or weakened in one species over the other, suggesting that the observed differences are unlikely due to a general difference in the biology of Fc-mediated platelet activation between cyno and human.

In the examples presented here, the levels of target expression between the species appeared similar via flow cytometry analysis, but more sensitive methods may be able to detect differences. Differences in clustering cannot be excluded. Antibody affinity to the respective targets were within 5-fold and may have contributed to the observed differences. No known differences in target biology have been reported. Nevertheless, the observed species differences in platelet activation may be due to undetected differences in target expression or clustering, affinity differences of the antibodies to target and/or FcγRIIa, or perhaps target biology. Thus, these case studies highlight the importance to evaluate in vitro platelet binding and activation in both human and the relevant tox species on a case-to-case basis to understand possible gaps in the safety assessment. Additionally, these case studies demonstrate that the platelet activation risk through therapeutic antibodies with platelet expressed targets can be minimized most consistently through LALA mutations in the IgG1 Fc backbone.

6.3.7. Case study 11: health authority (HA) requests regarding characterization of Fc-silenced mAbs

6.3.7.1. Agonist mAb targeting a cell surface receptor with a well-characterized DAPA silencing mutation

In this case, the HA questioned the fact that the DAPA (D265A and P329A) mutations led to a reduced (but not absence of) binding to FcγRs which, together with the broad expression of the target, was a cause for concern. The company was asked to provide ADCC and CDC data. The company provided further background information on mAb-A, supplementary in vitro study results and non-clinical and clinical safety data to argue that these assays were not warranted. The DAPA Fc mutations are known to significantly reduce binding affinity to all FcγRs, and binding to, and subsequent activation of FcγRs is required for ADCC and CDC.^61,194 These mutations were introduced into mAb-A, not to mitigate an active safety concern of targeting it’s antigen with an active Fc, but to reduce any potential for unexpected cytotoxicity, which was supported by the robust nonclinical package and FIH clinical data. The reduction of immunostimulatory potential was confirmed by SPR binding studies, in which mAb-A showed undetectable binding to FcγRIIAH131, FcγRIIAR131, FcγRIIB, FcγRIIIAF158, FcγRIIIAV158 and FcγRIIIB, and reduced binding affinity by several orders of magnitude to FcγRI, compared to WT IgG1. Consequently, no ADCC activity is anticipated for mAb-A. The binding of C1q to CH2 domain of the Fc region of mAb A was evaluated in an ELISA-based assay and compared to rituximab, an IgG1 known to mediate complement activation by its Fc portion. Compared to rituximab, no significant binding to C1q was detected with mAb-A, hence CDC activity is unlikely. In 13-week GLP IV and SC toxicity studies in rats and cynos, mAb-A was well-tolerated and showed no immune stimulatory alterations in anatomic and clinical pathology endpoints (i.e., globulins, coagulation parameters, differential leukocytes, or organ weights and histologic changes, including lymphoid organs) that could be indicative of ADCC and CDC activity. Clinical data also supports the safety of mAb-A with no indication of cytotoxicity. Hence, collectively, these data support a lack of ADCC and CDC activity, and so directly assessing ADCC/CDC activity of mAb-A was not warranted, which the HA accepted.

6.3.7.2. Antagonist mAb targeting a soluble cytokine and silenced with a well-characterized LALA silencing mutation

In this case, the HA wanted to see data for complement binding, ADCC and CDC activities, cytokines in vitro in human blood, as well as their interpretations and clinical relevance. The company indicated that they had not performed these studies because mAb-Z inhibits the function of a soluble cytokine and mAb-Z cannot bind to its target already bound to its receptor on cells. Hence, mAb-Z does not have the capacity to bind to immune or other cells and potentially directly activate them, or induce ADCC, CDC or cytokine release through engagement of the Fc domain with FcγR on immune cells or activation of complement C1q. A recent US Food and Drug Administration guidance states that cytokine release assays are not warranted for molecules that do not target surface receptors on immune cells.¹⁹⁵ Despite this very low risk of immune activation, since mAb-Z is based on an IgG1 isotype, an L234A/L235A (LALA) silencing mutations were introduced into the Fc domain to further minimize the risk of mAb-Z binding to complement or FcγRs.^28,61 mAb-Z showed no binding to human or cynomolgus monkey tissues in cross-reactivity studies and showed no evidence of histopathological or immunological changes as a result of ADCC, complement activation or cytokine release in a 13-week toxicity study in cynomolgus monkeys (after repeated SC and IV dose levels of up to 100 mg/kg weekly). Hence, collectively, these data support a lack of complement activation, ADCC and CDC activity, and cytokine release potential and so directly assessing these activities with mAb-X is not warranted, which the HA accepted.

6.3.7.3. Summary of HA feedback for multiple Fc-silenced mAbs from one company (see Table 12)

The majority of these mAbs were LALA mutants targeting both cell surface and soluble molecules. As with the two cases presented above most HA questions related to showing further data that the molecules were truly silent and lacking FcγR binding, complement activation and ADCC, ADCP and CDC data. The majority of the mAbs targeted membrane targets. In some cases, the data were provided and in other cases a written rationale was provided. Of interest, as with case b. above, even for soluble targets, where the mAb would not interact with cells, on some occasions HAs have requested cell cytotoxicity data. Of note, arguing that silencing mutations have been shown to be effective at silencing one antibody cannot always be used to support the silencing of another mAb against a different target (a “platform” approach). Data may need to be generated for the specific molecule in question, which is prudent at least for some targets when considering the results with the mAb presented in case study 11.

6.4. Half-life extension case studies

6.4.1. Case study 12: mAb with novel mutation for enhanced binding to FcRn where half-life extension in NHPs was not observed in humans

MTRX1011A is a humanized anti-CD4 Fc silenced IgG mAb derived from a previously described TRX1 mAb¹⁹⁶ that targets CD4 on helper T cells. MAb binding to CD4 leads to down-modulation of surface CD4 and blocking of its interaction with MHC II. MTRX1011A and TRX1 bind to the D1 domain of CD4+ T cells, in baboons and in humans (does not bind macaque CD4), with similar and high affinity (KD) <1 nmol/l). Both mAbs harbor the N297A mutation (aglycosyl), impairing their binding to FcγRs. PK/PD modeling and simulation based on clinical data of TRX1 suggested that at the maximum feasible SC dose, weekly administration would not maintain CD4 down-modulation in the majority of patients due to target-mediated clearance (TMDD). Therefore, MTRX1011A was engineered with a single amino acid substitution (N434H) to enhance its binding to FcRn and so reduce its in-vivo CL; although it was understood that only the nonspecific CL would be reduced, and target-mediated CL would still be the main driver of total CL at sub-saturating mAb concentrations. PK/PD modeling was done to understand the level of reduction needed in nonspecific clearance that would translate to meaningful reduction in total clearance at clinically feasible doses.

While both MTRX1011A and TRX1 bind CD4 cells with similar affinity in humans and in baboons, MTRX1011A demonstrated an approximately three times higher affinity for FcRn (at pH 6) relative to TRX1, suggesting a potential reduction in in vivo CL. Following four twice-weekly IV doses of MTRX1011A or TRX1 at a saturating dose of 40 mg/kg, the total CL of MTRX1011A decreased by 49.9% (95% CI 38–59) compared with TRX1. For MTRX1011A and TRX1, rapid and near complete (>98%) CD4 RO was observed on blood T cells. The duration of CD4 RO and down-modulation correlated with mAb exposure. In TRX1-treated baboons, full CD4 RO was lost by day 29, whereas in animals treated with MTRX1011A, this occurred by day 47. PK/PD modeling estimated a reduction in the nonspecific elimination rate for MTRX1011A of 48.7% (compared to TRX1) and predicted a reduction of 30–40% in total CL in patients at the maximum feasible SC dose.

In a Phase 1 study in rheumatoid arthritis (RA) patients at single or multiple weekly IV or SC doses of 0.3–7 mg/kg, nonlinear kinetics was observed across the dose range, which was expected and attributed to saturable CD4-mediated elimination, also reported for other anti-CD4 therapeutics.^196,197 Because of the nonlinear PK, a direct comparison of total CL values may not reflect the difference between MTRX1011A and TRX1 with regard to nonspecific elimination (Kel). Instead, the mean concentration–time profile of MTRX1011A at 7 mg/kg IV was compared with that of TRX1 at 10 mg/kg IV from the single-dose phase I study. There was no obvious difference between the slopes of the two profiles in the linear (i.e., high concentration) range, thereby suggesting that the nonspecific elimination rates of the two molecules are similar. However, a more robust analysis was required to quantify the difference in the nonspecific elimination rates, and so population PK/PD modeling was conducted using the MTRX1011A Phase 1 data. Based on the fitted PK–PD model, the estimated population mean nonspecific elimination rate (Kel) of MTRX1011A is comparable to that reported for TRX1 (0.078/day). Additionally, covariate modeling of the pooled phase I data related to both MTRX1011A and TRX1, with the molecule type as a covariate on Kel, confirmed that there was no significant difference between the Kel values for TRX1 and MTRX1011A. Inter-patient variability was higher for MTRX1011A, which could reflect differences in patient population (healthy volunteers for TRX1 vs RA patients for MTRX1011A). Interestingly, high levels of preexisting IgM antibodies were detected in ~70% of the RA patients in Phase 1, which appeared to correlate with the levels of RF. Immunodepletion experiments suggest that these preexisting IgM antibodies recognize the N434H mutation in the Fc region of MTRX1011A, which is located in the binding loop as a contact residue on IgG for RF.^198–200 There was no apparent relationship between the presence of the IgM anti-MTRX1011A and the PK of MTRX1011A in the patients who were tested.

These preclinical and clinical studies provided intriguing insights into the limitations of modifications to increase FcRn binding in the presence of substantial target-mediated CL. However, the nonspecific elimination (Kel) in humans did not appear significantly decreased, opposing the initial hypothesis and observations in baboon studies. It is speculated that the presence of anti-MTRX1011A IgM antibodies in RA patients could interfere with its ability to bind to FcRn. Although there was no observed impact of the presence of these antibodies on PK, there was increased inter-patient variability on the estimation of the Kel, which may have contributed to the IgM antibodies. There was also a difference in the trial participants with healthy volunteers used in TRX1 Phase 1 study and RA patients in the MTRX1011A Phase 1 study. Additionally, the RA patients had soluble CD4 at baseline, which was stabilized upon dosing MTRX1011A. This study shows the complexities faced while translating results from animal studies to humans. Factors including high inter-subject variability in clinical settings, heterogeneity in human populations, and potential interference by ADA antibodies might have contributed to the differences noticed between baboons and humans. More robust clinical trials might be required to understand the effects of FcRn-directed mutations on the population PK of mAbs.

6.4.2. Case study 13: sweeping mAb with enhanced binding to FcRn and FcγRIIB that used cyno surrogate mAb in for PK and safety studies

GYM329 is an anti-latent myostatin sweeping mAb that has enhanced hFcRn binding at both pH 6 and 7.4 and is also engineered to dissociate from myostatin once internalized in the endosome, thus targeting myostatin alone for lysosomal degradation. It is also Fc-engineered to have increased affinity for huFcγRIIB to promote its uptake and clearance within cells (sweeping). Mouse and cyno were used for nonclinical safety assessments, since GYM329 binds mouse and cyno myostatin with similar affinity as human myostatin. No toxicological changes were observed in pivotal repeat-dose toxicity studies in mouse and cyno (up to 200 mg/kg weekly for mouse and 100 mg/kg every 2 weeks for cyno for up to 13 weeks). However, GYM329 does not have enhanced binding to cyno FcγRIIB compared to a conventional IgG1, and thus a decrease in latent myostatin plasma concentrations was not observed in the pivotal cyno toxicity studies. Therefore, a cyno surrogate mAb, a prototype mAb with enhanced binding affinity to cyno FcγRIIB and with the same Fab region as GYM329, was assessed in a supplemental 2-month repeat-dose cyno study. There were no significant GYM329-related toxicological changes up to a pharmacologically active dose of 5 mg/kg administered IV every 4 weeks.²⁰¹ Since the Fc region of GYM329 was modified to enhance the binding affinity to huFcγRIIB, the potential effect of GYM329 on phosphorylation of ITIM of FcγRIIB in PBMCs and on apoptosis in B cells was assessed using human and cyno blood samples. GYM329 at a concentration of 100 μg/mL with or without equimolar amount of antigen (human or cyno latent myostatin, respectively) induced no to very few changes in both phosphorylation of FcγRIIB and B cell apoptosis. In contrast, the CD19-targeting mAb, which has strongly enhanced binding affinity to human or cyno FcγRIIB, showed significant changes as expected.²⁰² These data suggest that the potential risk of GYM329 inducing immunosuppression through FcγRIIB signaling would be low.²⁰³ No relevant safety risks were detected for potential infusion-related reaction/cytokine release syndrome upon first infusion (human whole-blood assay), for potential platelet aggregation (in vitro human platelet aggregation assay) and for potential tissue cross-reactivity (in human, cyno and mouse tissues).

6.5. FcγR blockade case studies

6.5.1. Case study 14: assessing in vitro safety of Fc multimers designed to block FcγRs

Hexameric Fc fragments were created from human IgG1 and IgG4 with the aim of effecting pan-FcγR blockade by use of a C-terminal ‘tailpiece’ sequence derived from human IgM.¹³¹ Since these molecules are designed to bind and block FcγRs, immune safety studies to assess the potential for unwanted immune cell activation are important. Stark differences were found in their in vitro safety profiles, which were examined by combining Fc-engineering and cell biology assays. These data highlighted critical differences between human whole blood-based and PBMC-based cytokine release assays, primarily due to the presence or absence of platelets, neutrophils and complement components. Both IgG1 and IgG4 forms of Fc-hexamers looked relatively inactive in PBMC safety assays, whereas the IgG1 elicited very strong responses in human whole blood assays. Studies showed that cytokine release was primarily due to differential binding to FcγRIIIB on neutrophils, but key Fc residues were also highlighted for the activation of human platelets via FcγRIIA.¹³² For example, amino acid 234 (L/F) alone was dominant in determining platelet activation (via FcγRIIA), but both 234 (L/F) and 327 (A/G) were involved in IFN-γ release (from neutrophils) and 331 (P/S) was the major contributor to C1q binding and complement activation. Both 234 and 327 were definitively shown to be involved in binding to FcγRIIIB expressed on HEK cells and by SPR, but the single mutants only partially reduced binding to isolated neutrophils. Hence, 234 contributes to binding to FcγRIIA and FcγRIIIB whereas 327 only contributes to binding to FcγRIIIB. The observations were fundamentally enabled due to the hexameric valency of the study proteins: monomeric Fc/IgG do not elicit any response from human cells and SPR affinities are very low even before mutations, which reduce binding further. Since mouse platelets do not express FcγRIIA and neither mouse nor primate neutrophils express FcγRIIIB, these data question the validity of these animals for some in vivo safety studies with Fc-modified molecules. HuFcγR transgenic mice might offer additional utility over human in vitro cell assays in certain situations. The uniformity of the form and reproducibility of activity for these hexameric Fc molecules enabled a ~ 100 donor cytokine release assay using both whole blood and PBMCs. These data confirmed the greater sensitivity of the WB-CRA in detecting immune cell activation and was able to pull apart subtle responses due to FcγR allotypic differences and gene copy number, including interactions between two pairs of allotypes, which are not possible to detect in a statistically meaningful way using smaller group sizes.²⁰⁴ For example, the FcγRIIA allotype HH, when compared to the RR allotype, showed statistical difference in IFN-γ release with both Campath (anti-CD52 IgG1) and IgG1 Fc-hexamer test proteins, whereas the FcγRIIIA variants VV and FF did not show any differences. Subjects who carried both HH/VV (n = 5) still showed significance over RR/FF (n = 8). Data also showed that the FcγRIIIB variant AA, AB and BB were statistically indistinguishable, suggesting that all subjects could be equally responsive to the neutrophil-related cytokine stimulations through FcγRIIIB stimuli.

6.6. FcRn blockade case study

6.6.1. Case study 15: pharmacology and safety testing of a FcRn blocker in cyno and humans

Efgartigimod is an engineered human IgG1 Fc fragment with five amino acid substitutions M252Y, S254T, T256E, H433K, N434F (MST-HN) in the CH2-CH3 region of the IgG Fc.^137,205 The HLE mutations M252Y, S254T and T256E (MST/’YTE’)²⁰⁶ and the structurally adjacent H433K, N434F (HN/’KF’) substitutions²⁰⁵ separately increase affinity to FcRn mainly at pH 6.0. The combination of these substitutions further increases the affinity to human FcRn at both pH 7.4 and pH 6.0, while the Fc retains a pH-dependent binding profile,²⁰⁵ hence efgartigimod is recycled in an FcRn-dependent manner.¹³⁷ Importantly, in vivo, IgG1-MST-HN was found to antagonize the IgG-FcRn interaction and specifically reduce IgG levels and showed efficacy in both passive and active models of antibody-mediated autoimmune disease.^207–209 The principle was called ‘antibody that enhances the degradation of IgG’ or short ABDEG²⁰⁵ for which defining the most suitable format (Fc only or full-length IgG with irrelevant Fabs) and understanding the biological consequences thereof⁴⁹ was a critical task. Early structural studies of human FcRn with rat IgG-Fc suggested an orientation of the CH2 domains toward the cellular membrane.²¹⁰ We observed striking differences in cellular trafficking: Fc-MST-HN (efgartigimod) entered cells and occupied FcRn more efficiently than comparator IgG, which correlated with more efficient blocking of IgG recycling in vitro and in vivo in cynomolgus monkeys.⁴⁹

In cynomolgus monkeys, a single administration of efgartigimod at 20 mg/kg resulted in maximum transient reductions of endogenous IgG by about 55% compared to pre-dose levels.¹³⁷ Four consecutive administrations of 20 mg/kg every 4 days resulted in deeper and longer-lasting reductions of endogenous IgG than 4 consecutive daily administrations. In a FIH study, single-dose administrations of 0.2, 2, 10, 25, and 50 mg/kg efgartigimod resulted in substantial reductions in all IgG subclasses, with IgG1, IgG2 and IgG3 being cleared more efficiently than IgG4 with a maximum percentage of reductions (Emax) of 49.1%, 43.5%, and 54.0% compared to 36.6% at the 10 mg/kg dose, respectively. Importantly, Emax and area under the percentage of reduction curve (AUEC) showed no statistical differences for the doses administered higher than 10 mg/kg, indicating a maximum PD effect was already reached at this dose in humans. Based on these data, 10 mg/kg and 25 mg/kg every 4 days (q4d) and every 7 days (q7d) and 25 mg/kg every 7 days was tested. Reductions of the endogenous IgG1 compared to the baseline ranged from 73% to 78% 1 week after the last (4th) injection and so were more pronounced than in the single-dose phase. Neither the q4d regimen of 10 mg/kg nor the q7d regimen of 25 mg/kg indicated superior effects comparing Emax and AUEC to 10 mg/kg of the q7d regimen.¹³⁷ Hence, the efgartigimod reaches its near maximum PD effect at a dose of 10 mg/kg in a q7d regimen. Efgartigimod has shown efficacy in several registrational trials in IgG-mediated autoimmune diseases,^134,211 supporting registration in generalized myasthenia gravis (gMG).²¹²

Data from a wealth of clinical studies has proven efgartigimod’s benign safety profile with no significant increased risk of infections as a result of IgG decreases.^{134,137,211,213} Importantly, while vaccine-induced IgG was reduced during efgartigimod treatment, antimicrobial antibody titers remained above minimally protective thresholds for the majority of patients, and titers returned to baseline when treatment was stopped. Patients treated with efgartigimod were able to mount antigen-specific immune responses both upon natural infection as well as vaccine-induced, suggesting a specific effect on circulating IgGs while sparing cellular immune components required for adaptive immune responses.²¹⁴ Mice and humans lacking functional FcRn are markedly deficient in both albumin and IgG because of the rapid degradation of these proteins. Efgartigimod specifically inhibits IgG recycling without reducing albumin.¹³⁷ The hypoalbuminemia observed in FcRn-deficient mice and humans^215,216 did not occur in toxicity studies or in clinical trials with efgartigimod.^{134,137,211,213}

7. Discussion and recommendations

Fc modification of human IgG antibodies has resulted in mAb therapeutics with enhanced potency and efficacy, as well as mAbs with more favorable safety profiles. The decision to modify the Fc region of therapeutic antibodies should be grounded in scientific evidence that considers the intended therapeutic effect, the biology of the target, and the safety profile of the molecule. Additionally, considerations around the molecule’s ‘novelty’, such as its manufacturability, potential immunogenicity, stability, PK, and the regulatory landscape. Many mutants are now well validated in the clinic and understood by regulators. However, new mutants or combinations of mutants might be understood from multiple case studies to carry new and unforeseen risks. Hence, their selection and adoption should be taken ‘wisely’ and in full understanding of the potential benefits and challenges.

Fc-enhanced molecules, and molecules designed to bind and block activating FcγRs (with the risk of unintended activation), are likely to require a more extensive in vitro and in vivo characterization than Fc-silenced ones. Enhanced molecules should be assessed for both the intended enhanced pharmacological effects and any associated safety concerns due to the higher risk of unwanted immune system activation and associated sequelae. Care should be given to differences between PBMC and whole blood assays due to the absence of numerate and FcγR-expressing platelets and neutrophils from PBMCs and for inter-species and transgenic animal differences of FcγR and FcRn expression and function. For Fc-silenced mAbs, it is important to confirm reduced effector function for cell-targeting mAbs in the presence and absence of competing IgG/IVIG. The number of studies required will depend on the novelty and complexity of the silencing mutations due to their ability to also impact FcγR, C1q, FcRn and RF binding as well as immunogenicity risk profile. Practical considerations for the characterization of Fc-modified mAbs are discussed below.

For Fc-enhanced molecules, confirming the specificity for the desired activating FcγR(s) that will drive efficacy (e.g., target cell depletion or agonism) is key. More promiscuous FcγR-binding profiles can lead to a risk of broad and perhaps unwanted immune cell activation. Activation can be witnessed as cytokine release, cell proliferation, cell reduction and/or cell aggregation, resulting in infusion-related events such as fever/chills or thrombocytopenia/neutropenia and associated thrombotic risks. Case study 1 showed that systemic inflammation and cytopenias are potential concerns with some mAbs with enhanced Fc effector functions, even when not readily predictable for these mAbs based on target distribution, suggesting that enhanced Fc effector functions may have contributed to the findings. However, neither generalized systemic inflammation nor neutropenia was observed with one of the mAbs or with benralizumab (case study 4), or an ADCC-enhanced anti-CD20 mAb, where findings in monkey toxicology studies corresponded only to B-cell depletion or ADA-mediated hypersensitivity reactions.²¹⁷ Other factors that likely contribute to the toxicity potential, include expression patterns and levels of the target, and pharmacology and mAb disposition. Potential Fc-enhancement safety concerns from findings identified in relevant toxicology studies should be considered in assessing benefit/risk, and appropriate measures built into the clinical protocols to manage the possible risk to patients. Case study 4 highlights the importance of in vitro studies in assessing FcγR specificity of Fc-enhanced mAbs, since a broader level of FcγR binding led to inflammatory effects and higher levels of ADA in animals. A retrospective analysis concluded afucosylated antibodies are associated with a higher incidence of neutropenia in cynomolgus monkeys, primarily due to ADA-induced inflammation, and highlights the importance of monitoring ADA responses and their pathological effects in preclinical studies.²¹⁸

For Fc-silenced mAbs, one must confirm silencing/lack of effector function in in vitro and in vivo studies (lack of cell killing, cytokine release, platelet activation). Standard silencing mutations may not be sufficient for some targets. The modified mAb may still bind the high-affinity FcγRI (CD64), and FcγRIIA (CD32A) at high concentrations and induce target cell agonism or immune cell/platelet activation. Certain targets may require confirmation of a silent Fc to derisk these effects using highly sensitive in vitro assays and potentially also in huFcγR Tg mice when standard toxicology species or assay systems are not sensitive enough to assess target agonism (case study 8) or CD32A-mediated platelet activation (case study 9). The majority of HA feedback for Fc-silenced mAbs relates to confirming silencing using in vitro studies, even for mAbs that bind soluble targets (case study 10). We recommend that organizations might choose to question requests relating to most soluble targets. In vitro studies are more relevant for mAbs targeting cell surface targets, but even in such cases, one can argue against in vitro studies for well-characterized mutants (e.g., LALA, LALA-PG) unless it is applied to novel cell surface targets, with potential safety liabilities. Perhaps case study 8 suggests that, for novel cell surface targets it would be prudent to confirm the silencing of even previously well-characterized mutants that provided silencing against other targets.

For HLE mAbs, one must consider the complexities faced while translating PK/exposure data from animal studies to humans. Factors including high inter-subject variability in clinical settings, heterogeneity in human populations, and potential interference by ADA antibodies might contribute to the lack of NHP-to-human translation. It should be considered whether standard animal models are still useful in modeling the exposure, PK/PD relationships and safety of these mAbs in humans and when to use hFcRn transgenic mice, which can be used not only to confirm half-life extension but also to show a lack of impact on PK of Fc-silenced (cases 6 and 7) and Fc-enhanced mAbs. These transgenic mouse models prove helpful for PK/PD translation and estimation of human exposures and are increasingly accepted by HAs as a replacement for NHP PK studies, especially when TMDD is not expected and NHP are not a pharmacologically relevant species. Additionally, it is important to consider when to utilize immunocompromised versions of the hFcRn transgenic mice to test highly engineered molecules (such as bispecific antibodies or Fc-fusion proteins), as immunogenicity could impact the interpretation of any generated PK/PD data.

When selecting a pharmacologically relevant species for pharmacology and toxicology studies, binding to both target and FcγR and FcRn should be considered. Do the paradigms for wild-type IgG hold true for Fc-engineered mAbs? What is the degree of enhancement and silencing in animals compared to humans? A key activity is demonstrating similarly enhanced effector function in humans and in cynomolgus monkeys (and rodents for mAbs that cross-react). For the majority of the Fc-enhanced mAbs that are approved to date (mainly afucosylated), these molecules showed potent target cell depletion in cynomolgus monkeys (where the FcγR biology and affinity for human IgG is closest to humans) supporting that human and cynomolgus FcγRIIIA and NK biologies are meaningfully matched. Mutations exploring biology beyond FcγRIIIA/NK cells will each need to make their own ‘first time’ discovery of how closely the human and cynomolgus receptors and biologies match. Consider when differences in FcγR biology/expression/function make rodent and or cynomolgus monkeys a non-relevant safety species. If an Fc-enhanced mAb showed a strong depletion of target cells or other effector functions in cynomolgus monkey but it was not enhanced over WT, then one might still use the cynomolgus monkey for toxicology assessment of the human therapeutic, but evaluate supplementing with data using a cynomolgus surrogate mAb that does show enhanced effector function in cynomolgus monkeys, as in case 16 for a sweeping mAb. For mAbs that are cross-reactive in both cynomolgus monkeys and rodents, consider how much variance in effector function in rodents would be required to justify not using it as a second toxicology species. If an Fc-enhanced mAb can still deplete the target cells in rodents better than the WT mAb, despite a significantly lower potency in rodents than human, it might be hard to argue that the rodent has no relevance as a second species doe toxicology studies, especially for a novel target where it could provide valuable in vivo safety data (e.g., the impact of and duration of depleting or agonizing cells and the recovery of these effects). If the enhanced Fc effects over the WT mAb are only observed in cynomolgus monkeys, then one might argue that cynomolgus monkey toxicology studies only should be sufficient.

When interspecies differences in Fc-effector function are substantial, evaluate the need for alternative models such as surrogate antibodies or human FcγR transgenic mice. This is particularly relevant when the toxicology species does not reflect the human in vitro binding or depletion profile, shows no enhanced effector activity relative to human cells, or lacks sensitivity to FcγRIIA-mediated effects. In such cases, in vitro studies with human cells may serve as a more appropriate or complementary approach. For example, if an Fc-modified mAb is specific for an established target that is already in the clinic and the Fc mutation is not novel, a toxicology study with surrogate molecules may not add any additional value.

When use of a surrogate mAb is deemed to have value, the preceding question is should an Fc-enhanced surrogate mAb try to mimic the same Fc engineering approaches (e.g., mutations in the same regions of the Fc, similar afucosylation) as for the human therapeutic (which may not necessarily result in the same effect across species), or is it more important for the surrogate to mimic the human half-life/exposure and effector function, (e.g., ADCC and CDC activity), expected with the human mAb in humans? When does the use of huFcγR mice add translational value? For novel enhancing mutations with perceived higher safety risks, one may want to use these mice, perhaps in combination with cynomolgus monkey data. On the contrary, would cynomolgus in vivo data combined with human in vitro studies be sufficient? HuFcγR mice were shown to be important in derisking a certain safety risk, i.e., platelet activation/TE, which is not easily assessed in cynomolgus/rhesus monkeys (case studies 8 and 9). However, one needs to consider the translatability of these huFcγR mice and how the expression of the huFcγRs on mouse cells relates to their expression and function in humans. If expression is different from humans, then effects seen in these mice may not translate to human (Case 7).

7.1. Key practical considerations in characterizing the activity of Fc-modified mAbs

7.1.1. FcγR-binding studies

Binding of monovalent IgG to FcγR and FcRn is inherently of low affinity and is complicated by natural differences between IgG’s1–4 and FcγR allotypic variants. Determinations of binding affinities have differed between methods and studies. HEK or CHO cell lines expressing a single FcγR and FcRn ‘retarded binding’ column and transcytosis studies may be useful, but the results could be complicated by receptor glycosylation differences. FcγR activation is normally a function of IgG-driven clustering, and so some researchers have demonstrated valuable utility from the study of antigen-driven immune complexes of different sizes²⁴ or ordered Fc complexes.¹³² Case study 14 illustrates radical differences in activity/safety of Fc-hexamers, derived from subtle binding differences which were essentially or virtually undetectable in monovalent binding studies. Researchers should consider if testing of multivalent IgG will inform their studies. ‘Fc-null’ mutant IgGs face a significant challenge in the demonstration of ‘no-binding’ vs ‘undetectable’ binding. In practice, some form of controlled multimerization is essential to understand the level and type of binding reduction. If possible, assays may be applied that address the avidity in Fc receptor binding. In addition, some form of sensitive cell assay might be instructive. For example, case studies 9 and 10 illustrate the sensitivity of platelet-focused assays and case study 11 shows the value of target clustering/agonism assays.

7.1.2. Cell binding studies

Binding to cells is complicated in several respects. Some cells express only one receptor (platelets, NK, B cells) whilst others express two or more. Hence, isolated cells, PBMC’s and whole blood assays offer context-dependent differential binding in response to any particular Fc variant, which can additionally be dynamically competitive in mixed cell systems in that receptors can compete with each other for IgG. IgG and C1q are further present as competitors, and the inclusion of exogenous IgGs in assays further adds some aspect of reality to a binding study. If allotypic differences might be important, use of pre-typed blood donors or a sufficient number of random donors to deal with individual variations should be considered. Similar considerations apply to the study of purified or HEK/CHO presented animal FcγR and the in vitro study of WT mice, transgenic mice, or cyno isolated cells, PBMCs, and whole blood. Finally, any mutations designed to influence FcγRIIA/B or FcγRIIIA/B might consider focused platelet and/or neutrophil studies to understand identifiable acute in vivo risks ahead of rodent or primate studies.

7.1.3. FcRn-related studies

Binding studies to FcRn need to be conducted at neutral and acidic pHs. Depending on the intended application, particular attention should be given to any signs of increased binding at neutral pH, as this has resulted in decreased, rather than increased, in vivo half-life. Some researchers have used FcRn affinity column studies, where IgGs are assessed using a pH elution gradient. Others have also found value in studying the dissociation rate and fraction bound remaining, but these assays can be technically challenging. Altogether, in vivo studies are essential to understand the real effect of mutations on PK, as certain molecular properties beyond FcRn binding can impact PK, such as effects caused by the Fab/fusion protein(s) surface properties (e.g., charge/pI). Mice transgenic for human FcRn (and knocked out for mouse FcRn) have become very useful tools and critically reduce the use of non-human primates. Case study 6 shows that subtle differences between various strains of mice should be considered. Since mouse IgGs binds less well to human FcRn than human IgG, mice are also being created that express humanized IgGs. Alternatively, mice can be pre-dosed with IVIG or recombinant IgG. Since FcRn also binds albumin, it will ultimately be useful to have mice expressing both human albumin and human IgG. One recent observation is that since the FcRn binding site on the Fc ‘elbow’ is also the binding site of many RFs, it can be prudent to check for enhanced or decreased RF binding, using patient sera. Finally, researchers may want to consider whether FcRn binding enhancement alters the placental transit of IgG to the unborn fetus.

Overall, we recommend that Fc mutations be considered holistically. The intended and unintended consequences should be evaluated and tested in a manner relevant to the target biology and its innate toxicology risks. Researchers should thoroughly examine all binding interactions to C1q, FcγRs, and FcRn and consider changes to CDC, ADCC/ADCP, PK, and RF binding properties. Studies may be less complex for Fc-null or FcRn-binding enhancement mutations in the context of neutralizing soluble cytokines, compared to FcγR-binding enhancement mutations affecting FcγRIIA/B or FcγRIIIA in the context of target agonism, cell depletion, or combined mutations for target ‘sweeping’.

Selection of pharmacologically relevant toxicology species always faces a challenge over the matching of target binding affinity/epitope/function and the impact of ADA in longer-term, multi-dose studies. Fc engineering brings additional species selection challenges as discussed earlier. NHPs have the closest match of receptor type and function to humans (excepting FcγRIIIB), but they should only be used when they have the most human-translatable value. Wild-type or FcγR and FcRn transgenic mice may each offer utility depending upon the kind of engineering performed and how much is known quantitively about differences in performance in human in vitro systems. The major gap in using WT mice is the lack of FcγRII on platelets and the lack of any equivalent of FcγRIIIB on neutrophils. Relatively little is known about FcγR and FcRn biology of other potential toxicology species, such as dog, or mini-pig. Use of a ‘surrogate’ molecule, for example, using a mouse Fc engineered in a way to attempt to mimic the action of the human Fc, can be fraught with complication. Therefore, species selection should be determined from a case-by-case approach.

Finally, Fc-engineering should take full consideration of the potential impact on the developability profile of the resulting drug. The potential for enhanced immunogenicity risk favors fewer conservative mutations over multiple non-conservative ones, at least until they have been safely tested in human studies by ‘originators’, but then used by others as ‘followers’. There may be some value in in silico and in vitro human APC assays to identify at least the most strikingly high-risk motifs. Fc mutations can affect expression levels, protein A binding, and thermal and chemical stability, and hence the ability to pass through harsher process steps such as low pH hold steps, concentration or filtration. Mutations can also affect RF and FcRn binding, aggregation potential, chemical modification upon long-term liquid storage, ability to be concentrated, viscosity and SC bioavailability.

In summary, Fc-engineering can and has offered clinical benefit and value to patients, but it invokes a series of complex and overlapping considerations that can be unwise to overlook. As outlined in this review, advances in understanding these complexities – together with accumulating clinical experience and evolving perspectives on therapeutic applications – have driven the expanded use of key Fc modifications such as ‘LALA’, ‘LS’, ‘YTE’, and their derivatives. Significant steps in novelty outside these and other core mutations likely require a deep understanding of the function and biology, potential safety risks involved and hence a balance against the perceived advantage they offer. No doubt, novel mutations will continue to proceed to the clinic. We hope that innovators will continue to share details of successes, failures, and lessons learned, as shown by the spirit of the case studies herein, in order to advance collective learnings and hence continually improve drug discovery.

Funding Statement

This HESI scientific initiative is primarily supported by in-kind contributions (from public and private sector participants) of time, expertise, and experimental effort. These contributions are supplemented by direct funding (that largely supports program infrastructure and management) that was provided by HESI’s corporate sponsors. A list of supporting organizations is available at www.hesiglobal.org.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Employment: Each author is employed by a distinct institution or organization. These affiliations are as follows:

• Frank R. Brennan and Shirley J. Peters: Discovery Research, UCB Pharma, Slough, UK

• J. Ryan Polli and Melissa Ramones: Novartis Biomedical Research, Cambridge, MA, USA

• Babette Wolf (formerly employed): Novartis Biomedical Research, Basel, Switzerland

• Tilman Schlothauer: Roche Pharmaceutical Research and Early Development, Roche Innovation Center Munich, Penzberg, Germany

• Curtis C. Maier: GSK, PA, USA

• Jean Sathish and Changhua Ji: Comparative Medicine and Drug Safety R&D, Pfizer, Pearl River, NY, and La Jolla, CA, USA

• David L. Wensel: Viiv Healthcare, Branford, CT, USA

• Derrick Witcher: Eli Lilly and Company, Indianapolis, IN, USA

• Patricia C. Ryan and T. Scott Manetz: Immune Safety, Clinical Pharmacology and Safety Sciences, Biopharmaceutical R&D, AstraZeneca, Gaithersburg, MD, USA

• Adriano Flora and Brian Soper: The Jackson Laboratory, USA

• Birgit Fogal and Lindsey Dzielak: Boehringer Ingelheim Pharmaceuticals Inc., Nonclinical Drug Safety, Ridgefield, CT, USA

• Xiaoting Wang: Amgen Research, Amgen Inc., Thousand Oaks, CA, USA

• Prathap Nagaraja Shastri and Karen Price: Johnson & Johnson Innovative Medicine, Spring House, PA, USA

• Michael Doyle, Nidhi Sharda, and Mary Struthers: Bristol-Myers Squibb, Princeton, NJ, USA

• Maximilian Brinkhaus and Bianca Balbino: Argenx, Ghent, Belgium

• Eric Stefanich: Genentech, South San Francisco, CA, USA

• Masaki Honda: Chugai Pharmaceutical Co., Ltd, Kanagawa, Japan

• David P. Humphreys: Discovery Research, UCB Pharma, Slough, UK

The views expressed in this manuscript are those of the authors and do not necessarily represent the views, activities, or policies of their respective employers.

Shareholding

Authors employed by for-profit institutions hold shares/stock options in their respective employing institution. These shareholdings are personal investments and do not influence the research process, findings, or decision-making in the preparation of this manuscript.

The authors assert that these interests have not compromised the scientific integrity of the research process, or the findings presented in this manuscript. All authors have participated in (a) conception and design; (b) drafting the article or revising it critically for important intellectual content; and (c) final approval of the version to be published.

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/19420862.2025.2505092.