ABfocus on THERAPEuTIc AnTIBoDIEs

REVIEWS
Therapeutic antibodies for autoimmunity and inflammation
Andrew C. Chan* and Paul J. Carter‡

Abstract | The development of therapeutic antibodies has evolved over the past decade into a mainstay of therapeutic options for patients with autoimmune and inflammatory diseases. Substantial advances in understanding the biology of human diseases have been made and tremendous benefit to patients has been gained with the first generation of therapeutic antibodies. The lessons learnt from these antibodies have provided the foundation for the discovery and development of future therapeutic antibodies. Here we review how key insights obtained from the development of therapeutic antibodies complemented by newer antibody engineering technologies are delivering a second generation of therapeutic antibodies with promise for greater clinical efficacy and safety.
Effector functions
Fc-mediated antibody properties that are involved in target cell destruction: antibody-dependent cell-mediated cytotoxicity (ADCC), antibodydependent cellular phagocytosis (ADCP) and complement-dependent cytotoxicity (CDC).

Half-life
The time taken for the plasma concentration of a drug to fall to half of its original value. Initial half-life and terminal half-life refer to the first (distribution) and second (elimination) phase for bi-exponential pharmacokinetics, respectively.

*Department of Immunology, Genentech, Inc.,1 DNA Way, South San Francisco, California 94080, USA. ‡ Department of Antibody Engineering, Genentech, Inc., 1 DNA Way, South San Francisco, California 94080, USA. e-mails: acc@gene.com; pjc@gene.com doi:10.1038/nri2761

Antibodies have rapidly become a clinically important drug class: more than 25 antibodies are approved for human therapy and more than 240 antibodies are currently in clinical development worldwide for a wide range of diseases, including autoimmunity and inflammation (the focus of this Review), cancer (the focus of the Review by Weiner and colleagues in this issue, REF. 165), organ transplantation, cardiovascular disease, infectious diseases and ophthalmological diseases1. The clinical success of antibodies has led to a major commercial impact, with rapidly growing annual sales that exceeded US$27 billion in 2007, including 8 of the 20 top-selling biotechnology drugs2. This modern era of therapeutic antibodies originated with the invention of hybridoma technology to generate mouse monoclonal antibodies in 1975 (REF. 3). Major limitations of mouse antibodies as therapeutic agents — immunogenicity, lack of effector functions and short serum half-life — were subsequently identified and largely overcome by the advent of antibody chimerization and, later, humanization4 technologies in the mid-1980s. Many antibody drugs, including those for autoimmunity and inflammation (TABLE 1), are chimeric or humanized versions of rodent antibodies. Several approved antibody drugs and an increasing proportion of antibodies entering clinical trials are of human origin5. These human antibodies are typically derived from large phage display libraries expressing human antibody fragments or transgenic mice engineered with human immunoglobulin genes6. Human antibodies from transgenic mice are commonly developed as therapeutic agents without prior optimization6. By contrast, phage-derived antibodies may require

improvements in binding affinity for antigen or biological potency that are routinely obtained by additional selection for desired phenotypes from phage display libraries7. Evolving technologies, including yeast, ribosome, mRNA, mammalian and Escherichia coli display libraries, as well as the direct cloning of human antibodies from human blood or bone marrow-derived cells, might also contribute to future therapeutic antibodies7. Beyond improved generation and optimization technologies, the development of antibody drugs has benefited from better choices and matching of target, antibody and patients8, as well as advances in the industrialization of recombinant antibody production and purification processes9. Current therapeutic antibodies provide much experience to guide future antibody drug development — strengths on which to build, limitations to overcome and new opportunities to seize (BOX 1). The quest for better therapeutic antibodies has gained great momentum in recent years, motivated by a convergence of clinical, scientific, technological and commercial factors10. First, there is a strong desire to improve on the clinical benefits for patients that were achieved by first-generation therapeutic antibodies. Second, there is a growing understanding of the mechanisms of action of antibody-based drugs and, in some cases, their limitations, including mechanisms of resistance. Third, technological advances are available to optimize and overcome existing biophysical, functional and immunogenic limitations of antibodies, as well as to confer antibodies with new activities10–12. Fourth, the major commercial success of antibodies2 has fuelled strong competition between companies that will probably intensify as many approved and developing volUME 10 | MAy 2010 | 301

NATURE REviEWS | Immunology © 2010 Macmillan Publishers Limited. All rights reserved

REVIEWs
Table 1 | Monoclonal antibodies and Fc fusion proteins for autoimmunity and inflammation* generic name (trade name; sponsoring companies)
Natalizumab (Tysabri; Biogen Idec/Elan)

Format

Targets

Development stages
Approved

Diseases

Proposed mechanisms of action
Receptor binding and antagonism; inhibits leukocyte adhesion to their counter receptor (or receptors) Receptor binding and antagonism; inhibits leukocyte adhesion to their counter receptor (or receptors) Ligand binding and neutralization

Humanized IgG4

α4 subunit of α4β1 and α4β7 integrins α4β7 integrin

MS and Crohn’s disease

Vedolizumab (MLN2; Millennium Pharmaceuticals/ Takeda) Belimumab (Benlysta; Human Genome Sciences/ GlaxoSmithKline) Atacicept (TACI–Ig; Merck/Serono)

Humanized IgG1

Phase III

UC and Crohn’s disease

Human (phage-produced) IgG1

BAFF

Phase III

SLE

TACI ECD–Fc (IgG1) fusion protein, modified Fc to eliminate effector functions LFA3 ECD–Fc (IgG1) fusion protein Chimeric light chain, humanized heavy chain, IgG1 aglycosyl Fc Humanized IgG1 with mutated Fc Chimeric IgG1

BAFF and APRIL

Phase II/III

SLE

Ligand binding and neutralization; blocks activation of TACI

Alefacept (Amevive; Astellas) Otelixizumab (TRX4; Tolerx/ GlaxoSmithKline) Teplizumab (MGA031; MacroGenics/Eli Lilly) Rituximab (Rituxan/ Mabthera; Genentech/ Roche/Biogen Idec)

CD2 CD3

Approved Phase III Phase III

Plaque psoriasis GVHD T1D

Inhibits LFA3–CD2 interaction and blocks lymphocyte activation Modulates T cell function

CD3 CD20

Phase III Approved

T1D Non-Hodgkin’s lymphoma, RA (in patients with inadequate responses to TNF blockade) and CLL CLL RA RA and SLE

Modulates T cell function Sensitizes cells to chemotherapy; induces apoptosis, ADCC and CDC

Ofatumumab (Arzerra; Genmab/ GlaxoSmithKline) Ocrelizumab (2H7; Genentech/Roche/ Biogen Idec) Epratuzumab (hLL2; Immunomedics/UCB) Alemtuzumab (Campath/ MabCampath; Genzyme/Bayer) Abatacept (Orencia; Bristol-Myers Squibb)

Human (mouse-produced) IgG1 Humanized IgG1

CD20

Approved Phase III

CDC and ADCC

CD20

Phase III

ADCC and CDC

Humanized IgG1 Humanized IgG1

CD22 CD52

Phase III Approved Phase III

SLE and non-Hodgkin’s lymphoma CLL MS RA and JIA UC and Crohn’s disease SLE Paroxysmal nocturnal haemoglobinuria

ADCC and downregulation of B cell receptor ADCC

CTLA4 ECD–Fc, mutated IgG1 Fc

CD80 and CD86

Approved Phase III Phase II/III Approved

Inhibits T cell activation by binding to CD80 and CD86, thereby blocking interaction with CD28 Binds C5, inhibiting its cleavage to C5a and C5b and preventing the generation of the terminal membrane attack complex C5b–C9 Ligand binding and receptor antagonism, reduces release of allergic response mediators from mast cells and basophils Ligand binding and receptor antagonism

Eculizumab (Soliris; Alexion pharmaceuticals)

Humanized IgG2 and IgG4

C5 complement protein

Omalizumab (Xolair; Genentech/Roche/ Novartis) Canakinumab (Ilaris; Novartis)

Humanized IgG1

IgE

Approved

Moderate to severe persistent allergic asthma Cryopyrin-associated periodic syndromes Systemic JIA, neonatalonset multisystem inflammatory disease and acute gout

Human (mice) IgG1

IL-1β

Approved Phase III

302 | MAy 2010 | volUME 10 © 2010 Macmillan Publishers Limited. All rights reserved

www.nature.com/reviews/immunol

f o c u s o n T H E R A P E u T I c A n T I E V ID W s R Bo EIE
Table 1 (cont.) | Monoclonal antibodies and Fc fusion proteins for autoimmunity and inflammation* generic name (trade name; sponsoring companies)
Mepolizumab (Bosatria; GlaxoSmithKline) Reslizumab (SCH55700; Ception Therapeutics) Tocilizumab (Actemra/ RoActemra; Chugai/ Roche) Ustekinumab (Stelara; Centocor)

Format

Targets

Development stages
Phase III Phase III Approved Phase III

Diseases

Proposed mechanisms of action

Humanized IgG1 Humanized IgG4 Humanized IgG1

IL-5 IL-5 IL-6R

Hyper-eosinophilic syndrome Eosinophilic oesophagitis RA JIA Plaque psoriasis Psoriatic arthritis Crohn’s disease Psoriasis and plaque psoriasis RA, JIA, psoriatic arthritis, AS and plaque psoriasis Crohn’s disease, RA, psoriatic arthritis, UC, AS and plaque psoriasis

Ligand binding and receptor antagonism Ligand binding and receptor antagonism Receptor binding and ligand blockade Ligand binding and receptor antagonism

Human (mice) IgG1

IL-12 and IL-23

Approved Phase III Phase II/III Phase III

Briakinumab (ABT-874; Abbott) Etanercept (Enbrel; Amgen/Pfizer) Infliximab (Remicade; Centocor/Merck)

Human (phage-produced) IgG1 TNFR2 ECD–Fc (IgG1) fusion protein Chimeric IgG1

IL-12 and IL-23 TNF

Ligand binding and receptor antagonism Neutralizes TNF activity by binding soluble and transmembrane TNF and inhibiting binding to TNFRs Neutralizes TNF activity by binding soluble and transmembrane TNF and inhibiting binding to TNFRs; induction of activated T cell and macrophage apoptosis Neutralizes TNF activity by binding soluble and transmembrane TNF and inhibiting binding to TNFRs; lyses TNF-expressing cells by CDC; induction of activated T cell and macrophage apoptosis Neutralizes TNF activity by binding soluble and transmembrane and inhibiting binding to TNFRs Neutralizes TNF activity by binding soluble and transmembrane TNF and inhibiting binding to TNFRs

Approved

TNF

Approved

Adalimumab (Humira/ Trudexa; Abbott)

Human (phage-produced) IgG1

TNF

Approved

RA, JIA, psoriatic arthritis, Crohn’s disease, AS and plaque psoriasis

Certolizumab pegol (Cimzia; UCB) Golimumab (Simponi; Centocor)

Humanized Fab, PEG conjugate Human (mouse-produced) IgG1

TNF

Approved

Crohn’s disease and RA

TNF

Approved

RA, psoriatic arthritis and AS

ADCC, antibody-dependent cellular cytotoxicity; APRIL, a proliferation-inducing ligand; AS, ankylosing spondylitis; BAFF, B cell activating factor; CDC, complementdependent cytotoxicity; CLL, chronic lymphocytic leukaemia; CTLA4, cytotoxic T lymphocyte antigen 4; ECD, extracellular domain; GVHD, graft-versus-host disease; IL, interleukin; JIA, juvenile idiopathic arthritis; LFA3, lymphocyte function-associated antigen 3; MS, multiple sclerosis; PEG, polyethylene glycol; R, receptor; RA, rheumatoid arthritis; TACI, transmembrane activator and calcium-modulating cyclophilin ligand interactor; SLE, systemic lupus erythematosus; T1D, type 1 diabetes; TNF, tumour necrosis factor; UC, ulcerative colitis. *Data sources include drug prescribing information, www.ClinicalTrials.gov, REF.1, REF. 163 and company web sites. Antibodies and Fc fusion proteins that have reached at least Phase II or III clinical trials are included, many are in less advanced stages for additional indications. The αL integrin-specific antibody efalizumab (Raptiva/Xanelim; Genentech/Roche/Merck–Serono) was voluntarily withdrawn from the market by its manufacturers in April 2009 because of its association with an increased risk of progressive multifocal leukoencephalopathy, a rare and usually fatal disease of the central nervous system.

Phage display
Technology for displaying a protein, such as an antibody fragment, on the surface of a bacteriophage that contains the gene (or genes) encoding the displayed protein (or proteins), thereby physically linking the genotype and phenotype.

Binding affinity
For two interacting molecules this is the ratio of their association (ka) and dissociation (kd) rate constants: Kd = kd ÷ ka.

drugs target the same antigen and/or diseases. Together, these major factors of target biology modulated by antibody properties can greatly affect the degree of clinical efficacy achieved and the ultimate successful adoption and use of a drug. The main functions of igG are determined by their interaction with four classes of naturally occurring binding partners: antigen, Fc receptors for igG (FcγRs), complement and the neonatal FcR (FcRn) (FIG. 1). Highly selective antigen binding is a signature and functionally crucial property of antibodies that is mediated by their variable domains. Several igG functions are dependent on interaction of the Fc region with other proteins: FcγRs for antibody-dependent cell-mediated cytotoxicity (ADCC) and antibody-dependent cellular phagocytosis (ADCP), complement components for complement-dependent

cytotoxicity (CDC) and FcRn for long serum persistence. Here, we first discuss the diverse mechanisms by which therapeutic antibodies modulate the functions of their target antigens to affect their biological behaviours and therapeutic profile. These include binding of soluble cytokines and growth factors, receptor blockade, cellular depletion through ADCC, ADCP, CDC and induction of apoptosis, receptor modulation and/or receptor signalling. Thereafter, we discuss the strategies that are being applied to create the next generation of antibody drugs.

Mechanisms of therapeutic antibody action Cytokine and growth factor blockade. The first therapeutic antibody for the treatment of inflammatory diseases was infliximab (Remicade; Centocor/Merck), in 1998, for the treatment of Crohn’s disease. infliximab, volUME 10 | MAy 2010 | 303

NATURE REviEWS | Immunology © 2010 Macmillan Publishers Limited. All rights reserved

REVIEWs
Box 1 | Strengths, limitations and future opportunities of therapeutic antibodies
Strengths • The availability of several well-established and broadly applicable methods for antibody generation and optimization. • A growing repertoire of technologies to redesign antibodies with modified or new properties to enhance their clinical potential; for example, pharmacokinetic properties such as terminal half-life can be tuned from minutes up to several weeks10–12. • A high success rate compared to other drugs: 17% for humanized antibodies from the first human trial to regulatory approval5. • Well-established and broadly applicable production technologies9. • Often, but not invariably, well tolerated by patients. • Broad experience and expanding knowledge of antibody drug development facilitates the generation of future therapeutic antibodies. limitations • Expensive, reflecting high production costs and commonly large doses9, potentially limiting patient access or clinical applications. • Clinical applications currently limited to cell surface or extracellular targets. • Cannot be orally administered. • The large size, particularly of the IgG format (~150 kDa), may limit tissue penetration. • There is limited penetration of the central nervous system by IgG owing to inefficient penetration of the blood–brain barrier. Future opportunities • Greater efficacy. • Higher affinity or potency. • Enhanced effector functions — antibody-dependent cell-mediated cytotoxicity (ADCC), antibody-dependent cellular phagocytosis (ADCP), complement-dependent cytotoxicity (CDC). • Target multiple disease mediators from distinct signalling pathways or that have redundant roles in the same pathway using bispecific antibodies. • Greater safety. • Reduced effector functions. • More ready reversibility of antibody action; for example, shorter half-life. • Greater selectivity for target. • Avoid activation of the target. • Improved delivery. • Lower or less frequent dosing using antibodies that are more potent or have a longer serum half-life. • Increased bioavailability. • Ability to cross the blood–brain barrier to reach the central nervous system. • More facile ocular delivery, including sustained release, ocular penetration and half-life. • Oral delivery.

Neonatal FcR
(FcRn). An Fc receptor that is structually related to MHC class I molecules and protects IgG from degradation, resulting in long serum half-life. Additionally, FcRn mediates IgG transfer from a mother to her fetus, thereby providing passive immunity.

Immunoadhesin
A fusion protein that combines the functional domain of a binding protein, such as a receptor or ligand, with an immunoglobulin Fc domain. Such Fc fusion proteins can endow binding proteins with antibody-like properties including long serum half-life and effector functions.

a chimeric antibody with mouse variable domains and human constant domains, binds both soluble and membrane-associated tumour necrosis factor (TNF)13. over the past 12 years, four additional TNF antagonists have been approved by the US Food and Drug Administration (FDA) for human use, each with distinguishing features 14. These include etanercept (Enbrel; Amgen/Pfizer), a TNF receptor ii (TNFRii) immunoadhesin fused to the Fc domains of human igG1 that neutralizes both TNF and lymphotoxin-α (also known as TNFβ); adalimumab (Humira/Trudexa; Abbott), a fully human igG1κ antibody derived from phage display technology that inhibits TNF; certolizumab pegol (Cimzia; UCB), a humanized Fabʹ fragment, chemically conjugated to polyethylene glycol (PEGylated), that neutralizes both membrane-associated and soluble human TNF; and golimumab (Simponi; Centocor), a fully human igG1κ TNF-specific antibody

created using genetically engineered mice (FIG. 2). The TNF antagonists are presently the most successful class of biological drug for inflammatory diseases, with total worldwide sales of $16.4 billion in 2008, and they are approved not only for the treatment of Crohn’s disease but also for the treatment of rheumatoid arthritis, juvenile rheumatoid arthritis, psoriatic arthritis, ulcerative colitis, ankylosing spondylitis and plaque psoriasis (see http://www.pipelinereview.com). These five TNF antagonists show the evolution of antibody technologies over the past two decades, beginning with the initial challenges of producing industrial quantities of the chimeric antibody infliximab and the immunoadhesin etanercept followed by the introduction of humanization protocols, phage display technologies, use of genetically modified mice expressing human immunoglobulins and application of PEGylation to increase the half-life of antibody fragments. www.nature.com/reviews/immunol 304 | MAy 2010 | volUME 10 © 2010 Macmillan Publishers Limited. All rights reserved

f o c u s o n T H E R A P E u T I c A n T I E V ID W s R Bo EIE
IgG binding partner VL Fab VH CH1 CL CH2 Fc region CH3 Glycosylation Sialic acid Galactose N-acetylglucosamine Mannose Core Variable Fucose Asn297 Glycosylation strategies for modifying FcγR and complement interactions Aglycosylation Bisecting N-acetylglucosamine Non-fucosylation
↓ ADCC, ↓ ADCP ↑ ADCC ↑ ADCC

Protein strategies for modifying interactions Mutate V domain sequences using display libraries and/or rationale design

Potential impact of modifying interaction Altered binding affinity or specificity

Antigen

FcγR Complement FcRn

Mutate Fc sequence using display libraries and/or rationale design; select IgG isotype Mutate Fc sequence using display libraries and/or rationale design

↑ ↑ ↑ ↑

or ↓ ADCC or ↓ ADCP or ↓ CDC or ↓ half-life

Bisecting N-acetylglucosamine Antibody fragment lacking Fc
↓ Half-life, ↓ CDC, ↓ ADCC and ↓ ADCP

and ↓ CDC

Human anti-chimeric antibodies
(HACAs). The immune system can develop an antibody response to chimeric antibodies. Binding of HACAs to chimeric antibodies forms immune complexes that can shorten the half-life of the therapeutic antibodies and compromise clinical effectiveness. Additionally these immune complexes can deposit in organs such as skin and kidney causing adverse events, including rashes and glomerulonephritis.

Figure 1 | Engineering Igg structure and function. IgGs are ~150 kDa tetramers comprising pairs of identical heavy and light chains linked by disulphide bonds (yellow bars). The heavy chains contain a variable domain, VH, and three constant domains, CH1, CH2 and CH3, whereas the light chains contain a variable domain, VL, linked to a single constant domain, CL. A signature property of antibodies is highly selective antigen binding mediated by their variable domains. Nature Reviews | Immunology Human IgG, particularly IgG1 and IgG3, bound to an antigen on a target cell surface can interact with Fc receptors for IgG (FcγRs) on effector cells and may support the destruction of target cells by antibody-dependent cell-mediated cytotoxicity (ADCC) or antibody-dependent cellular phagocytosis (ADCP), whereas interaction with the complement component C1q may support killing by complement-dependent cytotoxicity (CDC). IgG antibodies, like other circulating proteins, are taken up by vascular endothelial cells and other cells by pinocytosis. Subsequently, IgG can interact with the salvage receptor, neonatal FcR (FcRn), in a pH-dependent manner with binding occurring in endosomes at pH 6.0–6.5, followed by recycling and release at the cell surface at pH 7.0–7.4. Fc interaction with FcRn is mainly responsible for the long serum half-life of IgG108. Engineering of IgG variable domain sequences provides the means to tailor their antigen-binding affinity or specificity10. Fc amino acid sequence modification allows modulation of effector functions (ADCC, ADCP and CDC) and/or half-life10–12,71. Effector functions can be also be tuned by modifying Fc glycosylation. For example, mutating the conserved aspargine, Asn297, prevents glycosylation to generate aglycosylated antibodies lacking effector functions. ADCC can be enhanced by increasing the bisecting N-acetylglucosamine in the Fc carbohydrate or by eliminating fucose11. This figure is modified from Nature Reviews Immunology REF. 10.

Human anti-human antibodies
(HAHAs). The immune system can also develop an immune response to human therapeutic antibodies. HAHAs tend to be less common and less severe than HACAs.

Evolution of these technologies partially solved one of the historic obstacles of antibodies — immunogenicity. Patients with Crohn’s disease treated with infliximab develop human anti-chimeric antibodies (HACAs) in 8–61% of patients15,16. These HACAs can bind to the therapeutic antibody limiting its half-life and clinical effectiveness, as well as causing infusion-related anaphylaxis in some patients15,17. The incidence of HACA development can be attenuated by co-administration of immunosuppressive agents in patients with Crohn’s disease or with rheumatoid arthritis15,18. The advent of humanization protocols and development of human antibodies have minimized this immunogenicity. However, patients treated with humanized or human antibodies still develop human anti-human

(HAHAs). Adalimumab, a phage-derived human antibody, has an incidence of 12% neutralizing HAHAs in patients with rheumatoid arthritis receiving monotherapy that is decreased to ~1% in patients also treated with the immunosuppressant methotrexate19. in a prospective observational cohort of 121 patients with rheumatoid arthritis treated with adalimumab, HAHAs were detected in 34% of non-responders compared with 5% of responders20. Similarly, 6.3% of patients with early onset rheumatoid arthritis treated with golimumab, a human antibody derived from mice expressing human immunoglobulin genes, developed neutralizing HAHAs with the incidence of HAHAs higher in non-methotrexate-treated (13.5%) than methotrexate-treated patients (1.9–3.7%)21. antibodies volUME 10 | MAy 2010 | 305

NATURE REviEWS | Immunology © 2010 Macmillan Publishers Limited. All rights reserved

REVIEWs
Ligand blockade TNF TNF TNFRI TNFRI Therapeutic antibody Fab fragment Immunoadhesin Infliximab* Adalimumab* Golimumab Certolizumab pegol Canakinumab Briakinumab Ustekinumab Omalizumab* Belimumab Eculizumab Mepolizumab Reslizumab Etanercept‡ Atacicept‡ Alefacept‡

Therapeutic antibody

Receptor blockade IL-6R IL-6 gp130

Soluble IL-6R gp130 Tocilizumab Efalizumab* Natalizumab Vedolizumab Abatacept‡

Receptor downregulation αLβ2 integrin Efalizumab* Omalizumab* Otelixizumab* Teplizumab* Epratuzumab*

Depletion Effector cell FcγR CD20 Lysis Complement components
MAC

Rituximab* Ofatumumab Ocrelizumab GA101* Alemtuzumab Muromonab* Epratuzumab*

Signalling induction

TCR–CD3 complex

Otelixizumab* Teplizumab* Muromonab* GA101* Infliximab* Adalimumab* Rituximab*

Figure 2 | mechanisms of action of therapeutic antibodies. Five non-overlapping mechanisms of action are depicted. Examples of therapeutic antibodies are listed for each mechanism of action depicted. Ligand blockade with full length IgG therapeutic Nature Reviews | Immunology antibodies (for example, infliximab, adalimumab or golimumab), antibody fragments (for example, certolizumab pegol) or receptor immunoadhesins (for example, etanercept and those indicated with ‡) can prevent ligands from activating their cognate receptors. Binding of ligands (for example, interleukin-6 (IL-6)) to receptors (for example, IL-6R) can also be blocked by antibodies directed to their cognate receptors and inhibit receptor activation or function. Binding of cell surface receptors by antibodies can also result in their internalization and downregulation to limit cell surface receptors that can be activated by the ligand. Note that binding of cell surface receptors by antibodies (for example, αL integrin by efalizumab) or binding of a ligand (for example, free serum IgE by omalizumab) can indirectly also result in downregulation of cell surface receptors available for cellular activation. Binding of cell surface receptors can result in depletion of antigen-bearing cells through complement-mediated lysis and opsonization, as well as Fc receptor for IgG (FcγR)-mediated clearance. Therapeutic antibodies can also induce active signals that alter cellular fates. Binding of the T cell receptor (TCR)–CD3 complex by teplizumab can induce TCR-mediated signals and alter T cell functions and differentiation. MAC, membrane attack complex; TNF, tumour necrosis factor; TNFRI; TNF receptor I. *Antibodies with several mechanisms of action.
306 | MAy 2010 | volUME 10

Half-lives of each of the TNF-specific antibodies also define dosing schedules and modes of delivery. infliximab has the shortest half-life (~8–10 days) among the TNF-specific antibodies and is administered intravenously every 4–8 weeks. By contrast, adalimumab and golimumab, both with half-lives of ~2 weeks, are administered subcutaneously every 2–4 weeks, respectively. Subcutaneous administration is more convenient for patients, particularly those with chronic diseases requiring long-term therapy, as intravenous administration requires visits to physicians’ offices or infusion centres. Finally, the different formats of TNF-specific antibodies have also revealed some interesting aspects concerning the therapeutic mechanism (or mechanisms) of efficacy. A proposed mechanism by which infliximab, and not etanercept, was thought to be efficacious in Crohn’s disease was the ability of infliximab to bind membrane-associated TNF and induce apoptosis of activated T cells and macrophages22–24. However, the therapeutic success of certolizumab pegol and the preliminary clinical efficacy of golimumab in Crohn’s disease, agents that do not induce apoptosis of activated T cells or macrophages, suggest that other mechanisms probably contribute to the efficacy of TNF-specific agents in Crohn’s disease14,25–27. The lessons learnt from the development of TNFneutralizing therapeutic agents are rapidly being applied to other classes of therapeutic antibodies. Canakinumab (ilaris; Novartis), an interleukin-1β (il-1β)-specific antibody approved for the treatment of cryopyrin-associated periodic syndrome (CAPS), builds on the clinical experience of anakinra (Kineret; Amgen/Biovitrum), a recombinant non-glycosylated form of the human il-1 receptor antagonist (il-1Ra) approved for the treatment of rheumatoid arthritis28,29. The long half-life of canakinumab (~26 days) allows subcutaneous dosing every 8 weeks for children with CAPS. By contrast, anakinra has a half-life of only 4 to 6 hours, necessitating daily subcutaneous injections for patients with rheumatoid arthritis. Ustekinumab (Stelara; Centocor), a human igG1κ antibody that binds the p40 subunits of il-12 and il-23, which has been approved for the treatment of moderate to severe plaque psoriasis, has a median halflife of ~21.6 days30,31. Administration of ustekinumab to patients with psoriasis requires only quarterly subcutaneous injections following an initial loading dose. interestingly, more frequent dosing seems to be required for patients with Crohn’s disease and may reflect differences in the biology of il-12 and il-23 in these two immunologically and genetically related diseases32. Receptor blockade and receptor modulation. in addition to ligand blockade, therapeutic antibodies can also block ligand–receptor interactions by targeting the receptor (FIG. 2). These include antibodies that target the il-6 receptor (tocilizumab (Actemra/RoActemra; Chugai/Roche)), αl integrin (also known as CD11a and lFA1) (efalizumab (Raptiva/Xanelim; Genentech/ Roche/Merck–Serono)), the α4 subunit of α4β1 and α4β7 integrins (natalizumab (Tysabri; Biogen idec/Elan)) and α4β7 integrin (vedolizumab (MlN2; Millennium www.nature.com/reviews/immunol © 2010 Macmillan Publishers Limited. All rights reserved

f o c u s o n T H E R A P E u T I c A n T I E V ID W s R Bo EIE
Pharmaceuticals/Takeda)). Targeting of receptors adds a secondary level of mechanistic activity as a subset of these therapeutic antibodies not only blocks ligand binding but also downregulates the cell surface expression of the targeted receptors. However, this antibody-induced internalization of the target also results in antigen-induced clearance of the therapeutic antibody and decreases its serum half-life. For example, efalizumab not only binds αl integrin to block interactions between αl integrin and intercellular adhesion molecule 1 (iCAM1) but also downregulates αl integrin expression by T cells, a crucial co-receptor for T cell activation, as well as the expression of other T cell co-stimulatory molecules. Hence, efalizumab interferes with both T cell homing by interrupting αl integrin–iCAM1 interactions and T cell activation by downregulating co-stimulatory molecule expression33. Downmodulation of cell surface receptor expression can also be achieved indirectly through ligand targeting. omalizumab (Xolair; Genentech/Roche/Novartis), which is approved for the treatment of moderate and severe allergic asthma, binds and prevents the secreted form of igE from occupying the high affinity receptor for igE (FcεRi) on mast cells and basophils34. Whereas binding sequesters allergic igE from allergen-mediated activation of mast cells and basophils, the unoccupied surface FcεRi undergoes internalization and results in a decrease in cell surface FcεRi expression on dendritic cells, mast cells and basophils35–37. Hence, binding of igE by omalizumab probably operates through two distinct mechanisms: sequestration of the allergic igE and indirect downregulation of cell surface FcεRs on mast cells and basophils38. Receptor downregulation might serve as the basis for the efficacy of omalizumab in the treatment of chronic idiopathic urticaria and angioedema (CiU)39–41. Approximately 40% of patients with CiU have autoantibodies that recognize and can crosslink the FcεRi α-subunit, triggering mast cell and basophil activation42,43. Downregulation of cell surface FcεRs by omalizumab in these patients may attenuate the ability of these autoantibodies to activate mast cells and basophils to provide clinical improvement. A theoretical concern for targeting cell surface receptors, as opposed to their corresponding soluble ligands, is a greater potential for inducing an immunogenic response. Therapeutic antibodies that induce antigendependent internalization through the endocytic pathway can theoretically be processed as foreign antigens and presented on MHC class ii molecules to initiate a CD4+ T cell-dependent humoral response. However, although this is possible, we are not aware of any data that support this target class differential in immunogenicity of therapeutic antibodies. ongoing development of therapeutic antibodies that target dendritic cell surface antigens (such as DEC205; also known as ly75) may be informative in this regard. Depleting and signalling antibodies. Another class of therapeutic antibodies binds to cell surface antigens — for example, CD20, CD22 and CD52 — and depletes antigen-bearing cells (FIG. 2). Rituximab (Rituxan/Mabthera; Genentech/Roche/Biogen idec) and alemtuzumab
NATURE REviEWS | Immunology © 2010 Macmillan Publishers Limited. All rights reserved

(Campath/MabCampath; Genzyme/Bayer), approved first for the treatment of cancer, were subsequently tested in autoimmune and inflammatory conditions. Rituximab, a chimeric CD20-specific antibody, was approved in 2005 for the treatment of patients with rheumatoid arthritis who had inadequate responses to TNF blockade44; rituximab and alemtuzumab (a humanized CD52-specific antibody) have shown clinical activity in Phase ii clinical trials of the relapsing–remitting form of multiple sclerosis45,46. ocrelizumab (2H7; Genentech/Roche/Biogen idec), which is a humanized CD20-specific antibody, is being studied in Phase iii clinical trials of rheumatoid arthritis47, and epratuzumab (hll2; immunomedics/UCB), which is a humanized CD22-specific antibody, is being studied in Phase iii clinical trials of systemic lupus erythematosus48. This class of therapeutic antibodies operates mainly through FcγR-mediated clearance of antigen-bearing cells; although CDC can also contribute to their therapeutic effect under certain circumstances49–52. Patients with non-Hodgkin’s lymphoma treated with rituximab have better prognosis if they express the FcγRiiiA val158 encoding allele, which has high affinity for igG1, than patients expressing the FcγRiiiA Phe158 encoding allele, which has lower affinity for igG1 (REFS 53,54). Similarly, in a small study of patients with SlE treated with rituximab, B cell depletion was superior in patients with FcγRiiiA val158 than those with FcγRiiiA Phe158 (REF. 55). Hence, FcγR-mediated functions have an important role for this class of therapeutic antibody and offer an opportunity to design more efficacious therapies. in addition to effector functions, the nature of the antigen also contributes to the depleting potential of a therapeutic antibody. Although antibodies such as efalizumab and natalizumab bind to αl and α4 integrins expressed by T cells, these antibodies do not deplete T cells. Many antigen-related factors, including cell surface density, probably contribute to the depleting potential of a therapeutic antibody. CD20 is highly expressed (~90,000 molecules) on normal B cells, but its expression level is highly variable in B cell malignancies56,57. lower levels of CD20 expression can inhibit the ability of CD20-specific antibodies to mediate CDC58. in addition, expression of complement regulatory proteins (such as CD59, CD46 and CD55) on the target cell can compromise complement-mediated killing of target cells59; however, the expression of these complement regulatory proteins does not predict clinical outcome following rituximab treatment in patients with non-Hodgkin’s lymphoma60. Finally, antibody-mediated internalization can also decrease the number of accessible cell surface antibody–antigen complexes to affect both CDC- and ADCC-mediated depletion of target cells. These mechanistic data provide clues for the generation of future therapeutic antibodies with improved effector function. As the clearance of cells is mediated, in part, by mobilization of cells from organs into the blood and clearance of antibody-bound targets by myeloid-derived cells (such as Kupffer cells in the liver), non-circulating antigen-bearing cells saturated with the therapeutic antibody seem to be more resistant to antibody-mediated clearance50. Biopsies of the synovium of patients with rheumatoid volUME 10 | MAy 2010 | 307

REVIEWs arthritis following treatment with rituximab contain residual CD20-bearing B cells61,62. in FvB mice, marginal zone B cells in the spleen are not completely depleted by CD20-specific antibody therapy 50. However, these marginal zone B cells are not intrinsically resistant to CD20specific antibody killing as forced mobilization of these cells into the blood with blockade of αl integrin and α4 integrin renders them sensitive to depletion mediated by CD20-specific antibody. Hence, strategies to mobilize cells from their microenvironments may enhance the depleting potential of therapeutic antibodies. Therapeutic antibodies that bind cell surface antigens can also transmit intracellular signals required for clinical efficacy and/or contribute to adverse events63. The mouse CD3-specific antibody muromonab-CD3 (oKT3; Janssen-Cilag), the first FDA-approved therapeutic antibody for the treatment of acute allograft rejection in renal transplantation64, crosslinks the T cell receptor (TCR)–CD3 complex to induce transient cytokine release, resulting in an acute and severe flu-like syndrome and T cell depletion64. Muromonab-CD3 has an additional limitation in that ~50% of patients develop neutralizing antibodies that diminish its therapeutic effect65. The beneficial activities of muromonab-CD3 are mediated by its variable domains, whereas the undesirable mitogenic activity — stimulation of T cell proliferation and cytokine production — requires interaction between the muromonab-CD3 Fc region and FcγRs66. A second generation of CD3-specific antibodies (teplizumab (MGA031; MacroGenics/Eli lilly) and otelixizumab (TRX4; Tolerx/GlaxoSmithKline)) that are under evaluation for the treatment of type 1 diabetes are humanized antibodies with mutations that impair interaction with FcγRs (FIG. 2). These non-mitogenic antibodies do not induce the same degree of T cell activation and adverse events as observed with muromonab-CD3. However, both these antibodies still activate T cells, although their mechanistic effects are complex. First, these antibodies seem to downregulate the level of cell surface TCR–CD3 complexes. Although the TCR–CD3 complex is continually internalized and recycled, internalization of even unbound TCR–CD3 increases ~10-fold in the presence of otelixizumab without change in the exocytic rate67. Second, in vitro crosslinking of T cells with teplizumab induces TCR-mediated increases in free cytoplasmic Ca2+ and an altered pattern of cytokine production favouring il-10 rather than interferon-γ expression68. Patients treated with teplizumab developed increased numbers of il-10-producing CD4+ T cells and an even greater expansion of CD8+CD25+CTlA4+FoXP3+ regulatory T cells69. Hence, this class of antibody seems to modulate T cell function by altering T cells from an activating phenotype to potentially a tolerizing or regulatory phenotype. disease mediator blockade — that seem particularly promising for the generation of antibody drugs for autoimmunity and inflammation. We focus on igG as the format most widely adopted for therapeutic antibodies. Antibody fragments are also discussed as they are widely used in engineering antibody properties, they provide the building blocks to construct many new formats (including bispecific antibodies) and they are becoming important therapeutic agents in their own right 70. Optimization of Fc-mediated antibody functions. Modification of Fc-mediated activities is emerging as one of the most promising ways to further increase the clinical potential of antibodies10–12. Modulation of effector functions (ADCC, ADCP and CDC) and pharmacokinetic half-life are reviewed here in the context of antibody development for autoimmunity and inflammation. Engineering igG–binding partner interactions is routinely accomplished through amino acid sequence alterations in the antibody identified by selection from display libraries of antibody fragments and/or structureguided design. Tailoring Fc glycosylation offers a powerful additional method to modify some igG functions, particularly enhancement of ADCC11. Fc binding to another ligand, namely protein A, is ubiquitously exploited for chromatographic purification of antibodies, including for therapeutic use9. Effector function minimization. For some antibody therapies, antigen binding may be sufficient for achieving efficacy, and effector functions may be unnecessary and a potential source of adverse events in patients. Moreover, monovalent target binding may be necessary to avoid unwanted activation of the target. Several different strategies are available to generate antibodies that have minimal or no effector functions and, if need be, that are also non-activating 71. indeed many antibody drugs are designed with minimal effector functions, as exemplified below with antibodies that have reached Phase iii clinical trials or beyond in autoimmunity and inflammatory diseases (TABLE 1). As discussed previously, the clinical potential of CD3-specific antibodies has been greatly improved by Fc engineering to attenuate FcγR interactions and overcome mitogenic activity and the development of a flu-like syndrome, in conjunction with humanization to reduce immunogenicity. indeed, two such humanized CD3specific antibodies — teplizumab and otelixizumab — have advanced to Phase iii clinical trials (TABLE 1). These CD3-specific antibodies use different means to minimize FcγR binding: igG1 isotype with leu234Ala and leu235Ala mutations for teplizumab72,73 and aglycosylated igG1 with an Asn297Ala mutation for otelixizumab74. The igG4 isotype has also been selected for a few therapeutic antibodies for which effector functions are not desired75, including natalizumab and reslizumab (SCH55700; Ception Therapeutics) (TABLE 1). igG4 molecules are prone to exchange Fabʹ arms in vivo to become functionally monovalent 76. However, this seemingly undesirable attribute of igG4 for antibody drugs can be readily prevented by mutation of the hinge www.nature.com/reviews/immunol © 2010 Macmillan Publishers Limited. All rights reserved

Bispecific antibodies
Antibodies capable of binding to two different antigens or two distinct epitiopes on the same antigen are known as bispecific.

Next-generation therapeutic antibodies Strategies for new antibody generation can be broadly classified as modification of existing antibody properties or the endowment of antibodies with new capabilities (FIG. 1). Here we discuss examples of these two distinct categories — modulation of Fc-mediated functions and the generation of bispecific antibodies (FIG. 3) for dual

308 | MAy 2010 | volUME 10

f o c u s o n T H E R A P E u T I c A n T I E V ID W s R Bo EIE a Antigen-binding building blocks
Single variable domain binding sites Antigen 1
VH VL

Two variable domain binding sites
VL VH VL VL VH VL Fv VH scFv

VL VH VL VL VH VH VH

Antigen 2
VH VL

VH

VL Diabody

b Bispecific antibody fragments
Tandem scFv Tandem scFv–Fc scFv–Fc knobs-into-holes scFv–Fc–scFv F(ab′)2 VL CH1 CL VH

Fab–scFv

(Fab′scFv)2

Diabody

scDiabody–Fc

scDiabody–CH3

scDiabody

Figure 3 | modular Igg architecture supports numerous bispecific antibody formats. a | Fv fragments, comprising a VL and VH domain pair, are the most common antigen-binding building block in IgG. Fv fragments are routinely engineered into a single-chain (sc) format using a short peptide linker (~15 residues) to connect variable domains in either VH–VL or VL–VH topology156,157. Construction of scFv fragments with shorter linkers (typically 1–10 residues) allows inter-chain, but not intra-chain, domain pairing to form a dimer — the diabody158. Single variable domain antigen-binding sites are found in some IgG molecules from camelids and certain shark species, and this format is now being used for constructing therapeutics159. b | Antigen-binding building blocks, particularly scFv and diabodies, have been extensively permutated with other antibody domains and protein domains (not shown) to create an increasing repertoire of alternative bispecific antibody fragment formats. Fc regions are included in cases in which specific effector functions and/or long serum half-life are desired. By contrast, CH3 domains have been used for dimerization in cases in which Fc-mediated functions are not needed. c | Hybrid hybridoma technologies provided the first direct route to bispecific IgG by co-expression of two different IgG specificities160, albeit commonly with low yield and purity as, beyond the desired bispecific IgG, co-expression of two different antibody heavy and light chains can give rise to up to nine unwanted antibody molecules (not shown)161. Efficient construction of bispecific human IgG was achieved using phage-derived antibodies with a common light chain, to circumvent light chain mispairing, and knobs-into-holes engineering, to direct the assembly of two different heavy chains162. Phage display technology has endowed a monospecific antibody with the ability to bind a second antigen while maintaining high affinity for the first antigen — ‘two-in-one’ IgG141. The binding sites are partially overlapping so they can bind only one antigen at a time. Several bispecific antibody formats are both bispecific and bivalent for each antigen. For example, the dual variable domain IgG format was created by appending VL and VH domains with similar domains of a second antigen-binding specificity147,148. Bispecific antibodies have also been generated by appending either the amino or carboxyl terminus of an IgG heavy or light chain with a scFv or single variable (V) domain of a second binding specificity. scFv–IgG c IgG-based bispecific antibodies
Hybrid hybridoma Knobs-into-holes with common light chain IgG–scFv

Two-in-one IgG

Dual V domain IgG

IgG–V

V–IgG

NATURE REviEWS | Immunology © 2010 Macmillan Publishers Limited. All rights reserved

volUME 10 | MAy 2010 | 309

Nature Reviews | Immunology

REVIEWs sequence (Cys-Pro-Ser-Cys to Cys-Pro-Pro-Cys) or the CH3 domain76. igG4 can bind to some FcγRs, including FcγRi77,78 — an interaction that can be weakened by point mutations to abolish ADCC activity 79. Antibodies lacking effector functions can also be created with hybrid isotype igG molecules that mimic the desired attributes of their component isotypes. For example, the complement component C5-specific antibody eculizumab (Soliris; Alexion pharmaceuticals) (TABLE 1) uses a hybrid igG2–igG4 constant region that includes the CH1 domain and hinge region of a human igG2 fused to the CH2 and CH3 domains of a human igG4. Eculizumab lacks the ability to bind FcγRs and activate complement, which are common attributes for human igG2 and igG4, respectively 80. Strategies to eliminate effector functions are readily extended to Fc fusion proteins. For example, the cytotoxic T lymphocyte antigen 4 (CTlA4)–Fc fusion protein abatacept (orencia; Bristol-Myers Squibb) has five mutations compared with human igG1: lys215Gln, Cys220Ser, Cys226Ser, Cys229Ser and Pro238Ser 81 (Eu numbering 82). Multiple functional consequences of these mutations include improved protein production, elimination of inter-chain disulphide bonds and O-linked glycosylation by virtue of the mutated hinge (Cys-ProSer-Cys to Ser-Pro-Pro-Cys), inefficient binding to FcγRs and lack of ADCC and CDC activity 12,81. A seemingly small risk in making Fc mutations to ablate effector functions is that they might lead to deleterious structural perturbations. introducing the triple mutation lys234Phe:lys235Glu:Pro331Ser (‘FES’) into an igG1 molecule caused a substantial decrease in binding to FcγRs and the complement component C1q164. The structure of FES Fc is similar to other human Fc structures164. Thus the FES mutations designed to ablate effector functions probably do so without causing undue structural perturbation of the Fc domain. Many different engineered antibody fragments have been developed including several that have entered clinical development70. Many antibody fragments lack the igG Fc region and its associated properties — effector functions and extended half-life. As discussed previously, certolizumab pegol, a humanized TNF-specific Fabʹ fragment, is PEGylated to extend its pharmacokinetic half-life and is the first FDA-approved PEGylated antibody fragment for clinical use (TABLE 1). Whereas Fab fragments have a terminal half-life of ~20 minutes, certolizumab pegol has a half-life of ~14 days in patients, similar to its parent igG83. Enhancing effector functions. For some antibodies, potent effector functions, as described previously for rituximab, may be required for optimal clinical activity. Antibodies with efficient effector functions can be identified directly by screening panels of antibodies84, engineering to improve the activities of existing antibodies10–12 or a combination of both of these approaches. Potential benefits from enhancing effector functions include greater therapeutic efficacy, lower or less frequent dosing and decreased cost of goods. Potential risks from enhanced effector functions include more frequent or severe adverse events in patients.
310 | MAy 2010 | volUME 10 © 2010 Macmillan Publishers Limited. All rights reserved

ADCC activity can be enhanced by engineering the Fc protein sequence or tailoring the Fc carbohydrates to strengthen binding to FcγRs, most importantly FcγRiiiA11,12. Detailed mutational analysis of igG1 identified the functionally important Fc residues for binding to the different FcγRs (FcγRi, FcγRiiA, FcγRiiB and FcγRiiiA) and FcRn, as well as igG1 mutants with increased binding to FcγRiiiA and increased ADCC activity85. ADCC-enhancing mutants have also been identified from yeast display libraries86. Computational design based on the structures of Fc–FcγR complexes together with functional screening has identified several mutants with up to ~100-fold higher affinity binding for FcγRiiiA. one such mutant, Ser239Asp:Glu330leu:ile332Glu, gave a ~100-fold increase in ADCC potency, with greater maximal target cell killing in some cases87. Fc protein engineering to enhance ADCC is broadly applicable as shown with variants of several therapeutically relevant antibodies: trastuzumab (Herceptin; Genentech/Roche), cetuximab (Erbitux; imClone Systems), rituximab and alemtuzumab87. ADCC-enhancing antibodies can kill cells expressing lower levels of antigen, as shown with trastuzumab variants87 — a potential benefit for improved efficacy, albeit with unknown risk of killing non-target cells expressing low levels of antigen. ADCP activity can also be enhanced through Fc protein engineering. For example, a Gly236Ala mutation increases igG1 affinity for FcγRiiA 70-fold, and it has a 15-fold improvement in the selectivity of binding to this activating receptor over the related inhibitory receptor FcγRiiB88. The Gly236Ala mutation enhances phagocytosis of antibody-coated target cells by macrophages. ADCP and ADCC were simultaneously enhanced by combining Gly236Ala and Ser239Asp:ile332Glu mutations that enhance ADCP and ADCC, respectively 88. Aglycosylated igG molecules bind poorly to FcγRs and do not support ADCC. By contrast, bacterial display selection was used to identify a double mutant, Glu382val:Met428ile, that endows aglycosylated igG1 with the ability to bind FcγRi with high (nanomolar) affinity and support ADCC89. Moreover, the binding was remarkably selective for FcγRi with no detectable binding to the other FcγRs. The Glu382val:Met428ile mutations did not affect FcRn binding in vitro or serum persistence in vivo. Thus, aglycosylated igG can be engineered with selectivity profiles for FcγRs that are distinct from those of glycosylated igG89. A second widely applicable and comparably successful method for increasing the igG ADCC activity is through tailoring of Fc glycosylation11,90. Antibody production in a Chinese hamster ovary (CHo) cell line transfected to overexpress N-acetylglucosaminyltransferase iii resulted in antibodies linked to carbohydrate with a bisecting N-acetylglucosamine and more potent ADCC 91. Decreasing the fucose content of the carbohydrates in the Fc region can also enhance ADCC92,93. Production of antibodies devoid of fucose is readily achieved using CHo cells that lack expression of fucosyltransferase 8 (REF. 94). Total elimination of fucosylated antibodies is important to avoid competition with, and reduction of, the activity of the more potent non-fucosylated www.nature.com/reviews/immunol f o c u s o n T H E R A P E u T I c A n T I E V ID W s R Bo EIE antibodies95. other host organisms are used to produce non-fucosylated antibodies, including extensively engineered strains of Pichia pastoris 96. Most licensed therapeutic antibodies are extensively fucosylated11, and their in vitro ADCC activity is decreased by the presence of serum, presumably because of competition for FcγRiiiA with high concentrations of irrelevant igG. By contrast, non-fucosylated variants of rituximab and trastuzumab have increased affinity for FcγRiiiA and ~100-fold more potent ADCC activity that is minimally affected by serum igG95. The increases in ADCC potency achieved from Fc mutations87 and non-fucosylation95 seem to be similar as judged by corresponding variants of trastuzumab and rituximab. The binding site on human igG1 for complement component C1q has been identified by mutational analysis97. This led to the generation of a rituximab variant with improved complement binding 97 and ~twofold greater CDC activity, albeit with slightly impaired ADCC activity 98. Enhancement of antibody CDC activity has also been achieved using hinge-region mutations99. increased CDC activity has recently been achieved by creating a hybrid isotype igG in which some igG1 residues were replaced with corresponding igG3 sequences100. A few additional igG1 residues were required to restore efficient binding to protein A to allow purification100. An antibody enhanced in both CDC and ADCC activities was created by combining Fc protein mutations with non-fucosylation100. The efficacy and safety of antibodies with enhanced effector functions are challenging to assess preclinically because of species differences between the immune systems of mice and humans. CD20-specific antibodies with high ADCC activity have been evaluated in cynomolgus macaques using B cell depletion as a surrogate for efficacy87,101. Several antibodies with potent effector functions have progressed into clinical development. For example, human CD20-specific antibodies were obtained from transgenic mice with potent CDC activity even against tumour cell lines expressing high levels of complement regulatory proteins, CD55 and CD59, that are resistant to CDC by rituximab84. one of these human CD20specific antibodies, ofatumumab (Arzerra; Genmab/ GlaxoSmithKline), is now licensed for the treatment of chronic lymphocytic leukaemia and is under clinical evaluation for rheumatoid arthritis, multiple sclerosis and other cancers (TABLE 1). Another CD20-specific antibody, ocrelizumab (2H7; Genentech/Roche/Biogen idec), has in vitro ADCC and CDC activities that are higher and lower than rituximab, respectively 102. ocrelizumab has advanced to Phase iii clinical trials for rheumatoid arthritis and systemic lupus erythematosus (TABLE 1). At least eight effector function-enhanced antibodies have progressed into clinical development. For example, GA-101 (Roche/GlycArt) is a humanized type ii CD20specific antibody that is glyco-engineered (bisected and non-fucosylated) to enhance ADCC and is in Phase ii clinical trials for non-Hodgkin’s lymphoma103. other clinical stage glyco-engineered antibodies include a non-fucosylated humanized il-5R-specific antibody MEDi-563 (BioWa/Medimmune). A single intravenous dose of MEDi-563 of only 30 μg/kg — much lower than for most antibody drugs — was well tolerated and induced robust and reversible blood eosinopenia in all six mildly asthmatic patients treated11,104. ADCCimproved antibodies containing Fc mutations have also entered clinical development, including the humanized CD30-specific antibody Xmab2513 (Xencor) that is in a Phase i clinical trial for Hodgkin’s and anaplastic large cell lymphomas. A distinct mode of depleting CD20-specific antibodies has emerged for the treatment of B cell malignancies that has implications potentially for future depleting therapies in autoimmune and inflammatory diseases. CD20-specific antibodies can be defined as either type i (exemplified by rituximab) or type ii (such as tositumomab (Bexxar; Corixa/GlaxoSmithKline))105. Type ii, but not type i, CD20-specific antibodies induce an actindependent homotypic adhesion and aggregation of CD20bearing cells106. Through an unknown mechanism, type ii antibodies trigger lysosome destabilization and release of lysosomal contents to mediate a caspase- and B cell lymphoma 2 (BCl-2)-independent form of cell death. Should these mechanisms prove effective in human malignancies, they offer a distinct therapeutic option to deplete all CD20-bearing cellular niches. Pharmacokinetic half-life extension. The terminal half-life for chimeric, humanized and human antibody drugs in the circulation varies widely (~3–27 days)107. Engineering antibodies to increase their serum halflife and exposure offers the potential benefits of greater efficacy, lower or less frequent dosing, lower cost and enhanced localization to the target. The interaction between the Fc region of igG and the recycling receptor FcRn has a key role in igG homeostasis and is largely responsible for the long half-life of igG108: ~21 days in humans for igG1 (REF. 109). igG, like other circulating proteins, is taken up by vascular endothelial cells and other cells by pinocytosis. Subsequently, igG can interact with the FcRn, in a pH-dependent manner with binding occurring in endosomes at pH 6.0–6.5, followed by recycling and release at the cell surface at pH 7.0–7.4. Fc engineering to modulate the FcRn interaction and extend igG half-life is discussed briefly here and in greater depth elsewhere10,12. An important early milestone was the identification of Fc mutations from phage display libraries that strengthen Fc binding to FcRn and prolong the half-life of mouse igG in mice110. Subsequently, Fc mutations (Thr250Gln:Met428leu; ‘Ql’) were identified that increase the binding affinity of a human igG2 for human FcRn at pH 6.0 but not at pH 7.4 (REF. 111). These Ql mutations confer a similar pH-dependent increase in binding of a human igG2 to rhesus FcRn and a ~twofold increase in serum half-life in rhesus macaques111. The Ql mutations had a similar effect in extending the half-life of a human igG1 in rhesus macaques112. The Ql mutations also increased the binding affinity of a TNF-specific igG1 to FcRn from cynomolgus macaques, albeit without extending half-life for unknown reasons113. Alternative mutations have been identified that extend the half-life of antibodies in non-human primates. For example, the volUME 10 | MAy 2010 | 311 © 2010 Macmillan Publishers Limited. All rights reserved

Exposure
In the pharmacokinetic sense, the area under the curve for a plot of drug concentration versus time.

NATURE REviEWS | Immunology

REVIEWs triple Fc mutation, Met252Tyr:Ser254Thr:Thr256Glu (‘yTE’) increases the half-life of a humanized respiratory syncytial virus (RSv)-specific igG1 in cynomolgus macaques by three–fourfold114. The X-ray crystallographic structure of the yTE Fc is similar to other Fc structures, strongly suggesting that the yTE mutations do not significantly perturb the Fc structure115. Fc mutations that extend pharmacokinetic halflife can also enhance in vivo efficacy, which has been recently shown for the first time116. The double mutation Met428leu:Asn434Ser (‘lS’) engineered into the vascular endothelial growth factor (vEGF)-specific antibody bevacizumab (Avastin; Genentech) extended serum half-life by ~threefold in cynomolgus monkeys and by ~fourfold in engineered mice expressing human, but not mouse, FcRn116. Moreover, the lS mutations increased the antitumour activity of bevacizumab in a xenograft study in mice. in addition, a humanized version of the chimeric epidermal growth factor receptor (EGFR)-specific antibody cetuximab engineered with the lS mutations out-performed cetuximab in vivo, extended half-life in cynomolgus monkeys (~threefold) and in engineered mice (~fivefold), and had superior antitumour activity in mice. Studies in primates with engineered antibodies (see above) support the feasibility of extending the plasma half-life of therapeutic antibodies by tailoring the interaction of Fc with FcRn. indeed, the humanized RSv-specific igG1 with yTE mutations (MEDi-557) is currently in Phase i clinical trials to explore this possibility. The halflife-extending yTE mutations have also been successfully combined with mutations that enhance ADCC as judged by potent in vitro activity in cytotoxicity assays114. Species differences in FcRn–Fc interactions between mice and humans add much complexity to the preclinical analysis of antibodies engineered for extended halflife117. Transgenic mice expressing human FcRn may offer a partial solution to this challenge118. An engineered igG that binds to FcRn at both neutral and slightly acidic pH is more rapidly cleared than wild-type igG119. Thus, maintaining the strict pH dependence of Fc binding to FcRn is apparently crucial for supporting the long halflife of igG. An igG engineered to bind FcRn with higher affinity and reduced pH dependence potently inhibits the interaction of FcRn with endogenous igG and rapidly lowers the concentration of circulating igG in mice120. These FcRn-blocking antibodies known as ‘Abdegs’ may have therapeutic potential for reducing pathogenic igG levels in antibody-mediated autoimmune diseases120. Pharmacokinetic half-life reduction. For some therapeutic antibodies, it may be advantageous to decrease their terminal half-life to decrease exposure, improve safety or, in the case of some imaging applications, improve the target/non-target localization ratios. A large reduction in antibody half-life — from weeks to hours or less — is readily achieved by using small antibody fragments susceptible to renal elimination and that lack a functional Fc region to preclude recycling by FcRn. indeed, two unmodified Fab fragments are licensed for human therapy: the integrin-specific Fab abciximab
312 | MAy 2010 | volUME 10 © 2010 Macmillan Publishers Limited. All rights reserved

(ReoPro; Eli lilly) for the prevention of platelet-mediated clot in coronary angioplasty and the vEGF-specific Fab ranibizumab (lucentis; Genentech) for the treatment of neovascular (wet) age-related macular degeneration. it may be desirable to tune the pharmacokinetic properties of some antibody drugs to achieve half-lives that are intermediate between small antibody fragments and igG. This can be accomplished by Fc mutations that attenuate the interaction with FcRn and decrease antibody half-life, as shown by pharmacokinetic optimization of single-chain (sc)Fv–Fc antibody fragments for tumour imaging 121. Antibody fragments with a wide range of terminal half-lives — from a few hours to days — can be generated through site-specific PEGylation122. Several PEGylated proteins are licensed for human therapy 123 including one PEGylated antibody fragment, certolizumab pegol, that has igG-like pharmacokinetic behaviour. The pharmacokinetic properties of PEGylated proteins can be tuned by varying the size and extent of branching of the PEG molecules, sites of attachment and number of PEG molecules attached per protein123. Several alternative strategies have been developed to extend the half-lives of antibody fragments and other proteins including genetic fusion to albumin124, igG Fc125 or an unstructured recombinant polypeptide (XTEN)126, engineering them with the ability to bind long-circulating proteins such as albumin127 or igG128, or by creating new glycosylation sites129. Fc mutations that impair interaction with FcRn can decrease antibody half-life as exemplified by pharmacokinetic optimization of scFv–Fc antibody fragments for tumour imaging 121.

Optimization of antigen-binding domains Engineering of antibody variable domains to increase the binding affinity for antigen or tune the binding specificity has been extensively reviewed elsewhere10 and is not discussed further here. Two major strategies gaining momentum for endowing antibodies with new activities are antibody–drug conjugates130 and bispecific antibodies131,132, which are mainly used for targeting cytotoxic drugs and effector cells, respectively, to kill tumour cells. Bispecific antibodies have additional potential applications including dual blockade of disease mediators133, as discussed here.
Bispecific antibodies. The concept of bispecific antibodies was first suggested nearly 50 years ago134, and it is finally coming of age for the generation of therapeutic agents131,132. Historically, a major obstacle to developing bispecific antibodies as therapeutic agents has been the difficulty in designing these complex molecules with favourable drug-like properties and producing them in sufficient quantity and quality to support drug development. Bispecific antibodies as potential therapeutic agents have undergone a renaissance in recent years, facilitated by the advent of numerous alternative formats and improved methods for their more efficient generation, optimization and production132,133,135 (FIG. 3). Moreover, greater understanding of disease pathogenesis is leading to better choices of target antigen and antibody pairs. Here we propose ideal attributes of bispecific antibodies www.nature.com/reviews/immunol f o c u s o n T H E R A P E u T I c A n T I E V ID W s R Bo EIE for drug development based on experience with antibodies and other biological drugs. Additionally, advances in selected technologies for generating bispecific antibodies for therapeutic applications are presented (FIG. 3). The most successful clinical application of bispecific antibodies to date has been in oncology by redirecting cytotoxic T cells to kill tumour cells. This concept was originally proposed in 1985 (REF. 136) and has been explored in the context of several different bispecific antibody formats. The most successful format identified to date for this application is a tandem scFv (FIG. 3) known as a ‘BiTE’ (bispecific T cell engager). Blinatumomab (MT103; Micromet/Medimmune), a BiTE specific for CD19 and CD3, has been used to treat patients with nonHodgkin’s lymphoma137. Complete and partial responses were observed with blinatumomab administered at doses as low as 0.15 μg per m2 per day137, that is, doses at least 10,000 times lower than those typically used for therapeutic igG1 antibodies for targeting tumours. BiTEs are rapidly cleared, necessitating administration by a portable mini-pump; however a possible benefit is fine dosing control. Adverse events with blinatumomab include neurological symptoms that are seemingly reversible. This early clinical success with BiTEs has encouraged further pursuit of this format, including targeting of many different tumour antigens131. Several additional applications of bispecific antibodies have been investigated133,135, including blockade of several disease mediators as discussed here. Designing bispecific antibodies as therapeutic agents. An ideal bispecific antibody format for therapeutic applications would be broadly applicable to different combinations of target antigens without requiring extensive customization for individual antibody pairs. Routine highlevel expression (gram per litre) of bispecific antibody will probably be needed to support manufacturing for clinical development, preferably in well established and widely available host production systems such as CHo cells or E. coli. Bispecific antibodies should preserve the antigenbinding affinity and beneficial biological activities of their component monospecific antibodies. Bispecific antibody design preferably facilitates purification to homogeneity and recovery in high yield. Simultaneous binding of both antigens to the bispecific antibody is desirable and may be essential depending on the application. Selection of the bispecific antibody format to match the valency of each component antibody to the biology of the target antigens may be necessary, and is readily achievable from the many alternative formats (FIG. 3). Bispecific antibodies need favourable physicochemical properties such as high solubility and stability plus low propensity to aggregate, as suggested by successful biological drugs. Bispecific antibodies have pharmacokinetic properties that are suitable for their intended clinical application — long serum half-life in many cases.
Valency
For antibodies, including bispecific antibodies, the number of binding sites for each cognate antigen.

Dual blockade of disease mediators with bispecific antibodies. The pathogenesis of many human diseases involves several mediators that function in distinct signalling pathways or that have redundant roles in the same pathway. Simultaneous blockade of several different

disease mediators may lead to greater therapeutic efficacy and/or benefit more patients than targeting individual disease mediators133. This dual targeting concept was initially explored in oncology with bispecific antibodies that bind vEGFR1 and vEGFR2 (REF. 138), EGFR and insulin-like growth factor receptor 1 (iGFR1)139,140. More recently the human epidermal growth factor receptor 2 (HER2)-specific antibody trastuzumab was engineered using phage display technology to create a ‘two-in-one’ antibody (FIG. 3c) that can bind either HER2 or vEGF with high affinity; it was shown to inhibit both HER2- and vEGF-mediated cell proliferation in vitro and tumour progression in mouse models141. Although this combinatorial strategy is currently being pursued in many cancers, combinatorial therapy has been limited, to date, by significant safety issues when used in inflammatory diseases. Combination of il-1β and TNF blockade by anakinra and etanercept, respectively, as well as dual targeting of TNF (by etanercept) and CD80 and CD86 (by abatacept (orencia; BristolMyers Squibb)) did not result in any significant additive or synergistic clinical effects, but rather in a substantial increase in infectious complications142,143. Nonetheless, dual targeting of more appropriate pathways may provide greater efficacy without significantly increased toxicities. in turn, bispecific antibodies seem preferable in these cases, as opposed to developing two unapproved therapeutic antibodies, for reasons that include lower costs, simpler clinical drug development schemes and more traditional regulatory pathways. The pro-inflammatory cytokines il-1α and il-1β have key and seemingly redundant roles in the pathogenesis of various diseases144. indeed, blockade of both il-1α and il-1β in a mouse model of collagen-induced arthritis using a combination of monospecific neutralizing antibodies was more efficacious than either individual antibody 145,146. These antibody combination studies provide a strong rationale for developing therapeutic agents that neutralize both il-1α and il-1β, ideally without interfering with endogenous il-1Ra147,148. identifying a single antibody that binds to both il-1α and il-1β has proved elusive, as these cytokines share only ~20% identity. By contrast, a new bispecific tetravalent igG-like molecule known as dual variable domain igG (DvD-igG; FIG. 3c) was generated that neutralizes both mouse il-1α and il-1β and has similar in vivo potency to a combination of the two monospecific igGs from which it was derived148. A DvD-igG binding human il-1α and il-1β was then created, allowing clinical evaluation147. A further DvD-igG, comprising il-12-specific and il-18-specific antibodies, suggests that this format may be broadly applicable148. DvD-igG can match or at least approach the binding affinity of the parent antibodies, although some optimization may be required; for example, antibody selection, order of antigen-binding variable domains and design of linkers connecting variable domains147,148. DvD-igG can be generated with many attributes conducive to drug development, including robust expression in mammalian cells and facile purification by protein A chromatography, as well as favourable physicochemical and pharmacokinetic properties147,148. volUME 10 | MAy 2010 | 313

NATURE REviEWS | Immunology © 2010 Macmillan Publishers Limited. All rights reserved

REVIEWs
Functional blockade of additional cytokine pairs has been achieved with several different bispecific antibody formats. T helper 17 (TH17) cells may be associated with the pathogenesis of some allergic diseases such as allergic contact dermatitis, atopic dermatitis and asthma149. Dual targeting of il-23, a TH17 cell growth and survival factor, and il-17A, a pro-inflammatory cytokine produced by TH17 cells, may be more effective than blockade of either il-23 or il-17A alone in inflammatory diseases involving TH17 cells150. Selection from phage display libraries was used to generate and optimize stable and high affinity scFv neutralizing either il-17A or il-23. Bispecific antibodies binding to both il-17A and il-23 were then generated by incorporating these monospecific scFv into a several previously described bispecific antibody formats150: tandem scFv–Fc151, scFv–Fc–scFv 152 and igG– scFv 153 (FIG. 3). Bispecific antibodies were identified that potently neutralize both il-17A and il-23 with good stability and pharmacokinetic properties. These findings support the design principle of constructing bispecific antibodies from stable monomeric building blocks154. first decade of therapeutic antibodies in autoimmunity and inflammation as well as a subset of the technological advances that will enhance the capabilities of therapeutic antibodies. The experiences from this second generation of therapeutic antibodies will undoubtedly inform our next generation of therapeutic antibodies. Although not discussed in this Review, there is much to be learnt about how antibody delivery, as well as other factors, can alter the distribution of antibody in the body. These include oral delivery of antibodies, development of sustained release platforms that can be used in the eye to avoid frequent intravitreous injections and improved access of therapeutic antibodies into the brain. Although rituximab and alemtuzumab have shown preliminary evidence of efficacy in the relapsing–remitting form of multiple sclerosis, steady state cerebrospinal fluid levels of rituximab in relapsing–remitting multiple sclerosis are low (