The antigen-binding entity of an antibody, reduced in size to one single domain, is referred to as a “single-domain antibody.” Various strategies have been explored with variable success to arrive at functional single-domain antibodies. The potential of single-domain antibodies, as research tools or in medicine, is reflected by the three companies – founded in Europe – with a mission to bring these molecules to the market. Domantis using human VH-derived single-domain antibodies started in 2000 and was bought by GSK for £300M in December 2007. Haptogen employing shark single-domain antibodies was acquired by Wyeth, and Ablynx focusing on llama-derived single-domain antibodies received over €70M in three rounds of venture capitalist investments and another €80M on the Euronext stock market in November 2007. Regarding therapeutic applications, Arana Therapeutics in Australia entered a Phase 2 clinical trial with its single-domain antibody derivative. In this chapter, we will review (1) the various antibodies used for generating single-domain antibodies, (2) the properties of single-domain antibodies that create an added value for use in immunotherapy, and (3) a number of therapeutic applications.

THE DEVELOPMENT OF SINGLE-DOMAIN ANTIBODIES

Antibodies comprise two identical heavy chain polypeptides (H) carrying chains of carbohydrates and two identical light chain proteins (L). Their ability to bind specifically to an antigen is dictated by the paired variable regions of the heavy (VH) and the light (VL) chain (Figure 16.1).

During evolution, antibodies have acquired several invaluable properties that are now being exploited for clinical applications. First, they can bind a wide variety of target molecules with exquisite specificity. This property can be used to block the action of ligands such as TNFα in patients with rheumatoid arthritis or the Her-2 receptor in patients with breast cancer. In contrast to this mode of action, antibodies can also imitate ligand binding and stimulate various signaling pathways. Antibodies binding to CD20, for example, can induce apoptotic signals in the malignant cells of patients with non-Hodgkin's lymphoma. Additional effector functions are provided by the Fc domains, which can induce cell lysis by binding to complement (CDC) or by binding to Fc receptors on natural killer cells and macrophages (ADCC). An additional binding domain for the neonatal receptor on endothelial cells facilitates their uptake and recycling, enabling antibody therapeutics to remain in the circulation for many weeks.

To optimize the properties of an antibody for a particular indication or for use as a diagnostic, it would be preferable to improve or even delete particular characteristics. For example, to achieve better tumor penetration or a better tumor-to-blood ratio for visualizing metastases, it would be preferable to have a relatively small antibody fragment with a fairly short half-life. On the other hand, the antibody should not be too small in order to avoid a rapid clearance immediately after its application. It would also be very advantageous for certain clinical applications to improve the effector functions.

The immune system creates binding sites of high specificity and affinity in the variable domains of antibodies by generating sequence and consequently structural diversity in the complementarity-determining region (CDR) loops, which are located at the N-terminal ends of these domains. Sequence variations in the CDR loops of an antibody generally do not have a significant influence on the overall structure of the variable domain that carries them. This feature of variable domains is actually observed in a more general sense in immunoglobulin-like domains, which are known to have a similar general shape in the core beta-barrel and high structural variability in the loops. Furthermore, overall sequence similarity of domains with an immunoglobulin fold is mainly below 25%, while their structural similarity is high, with a root mean square (rms) deviation of Cα atoms always below 3.9 Å (Halaby et al., 1999).

We therefore set out to explore whether this inherent stability and conservation of the immunoglobulin fold allows loops of immunoglobulin domains other than the CDR loops to accommodate sequence variation without negatively impacting the overall structure and stability of the protein. As shown in Figure 17.1, the candidate loops of an IgG1 for this kind of engineering are manifold, including the N- and C-terminal loops of the constant domains as well as C-terminal loops of the variable domains.

In the examples described in this chapter, we engineered the AB and the EF loops of the third constant domain (CH3) of human IgG1 by randomizing a number of residues and also by inserting random sequences in the loops, thereby generating new binding sites.

Since its reemergence following the discovery of monoclonal antibodies in the early 1980s, the field of antibody therapy in cancer has progressed in leaps and bounds. From murine to chimeric, through humanized to fully human, we are now in a situation where, with over 200 antibodies having passed through some kind of clinical testing (Reichert & Valge-Archer, 2007), the monoclonal is now an accepted form of treatment for malignancy. In fact, for some malignancies, most notably non-Hodgkin's lymphoma, monoclonals are routinely used as frontline therapy. As such, we are past the point of asking whether monoclonal therapy works and into the more expansive territory of asking how it works and how we can make it work better.

While antibodies can function to combat a tumor in a number of ways – for example, sequestration of factors essential to survival or growth and stimulation of the immune response – one of the best-studied mechanisms of action is direct tumor cell killing. Here we will begin by looking in detail at the mechanisms by which antibodies can mediate cell killing, and which of these mechanisms is likely to be most important. Subsequently, we will review briefly the possible ways that this cell killing can be increased through the process of protein engineering, several of which will be expanded upon by the authors of subsequent chapters.

With over 20 therapeutic antibodies currently approved by the Food and Drug Administration (FDA) and close to 100 leads in clinical trials, therapeutic antibodies are responsible for a considerable part of the therapeutic proteins sales worldwide. The observation of immunogenicity with the early therapeutic antibodies did not come as a surprise, as many of them were murine antibodies or chimeric variants, consisting of murine variable parts in conjunction with a human constant domain. Over time, there has been a strong evolution toward the development of humanized and fully human antibodies, thereby reducing the observed immunogenicity to a significant extent. General side effects such as anaphylaxis and allergy against protein therapeutics are also less prevalent but this is due to better manufacturing processes giving more homogeneous products.

However, some of the currently available fully human antibodies have induced significant immunogenic responses over time. This has led to the regulatory instances in Europe and the United States supporting the development of guidelines to assess the likelihood of observing immunogenicity and its potential severity and side effects.

IMMUNOGENICITY DRIVERS

Several factors contribute to the potential immunogenicity of a protein therapeutic:

• Homology to human or endogenous proteins: The degree of “foreignness” of a protein to the host is one of the major contributors to an immune response. Indeed, the likelihood to observe immunogenicity related to a bacterium-derived protein therapeutic, such as staphylokinase, is higher than against proteins that show high homology to endogenous proteins, such as erythropoietin (EPO) and insulin. The overall immunogenicity of antibody therapeutics has been severely reduced by the development of fully human and humanized therapeutic antibodies as compared to the first-generation murine and chimeric antibodies (Hwang and Foote, 2005).

• […]

PEGylation of proteins has been performed for over 30 years (Abuchowski et al., 1977a,b). Although the details such as polyethylene glycol (PEG) size, structure, synthesis, purification, and reactive chemistries have changed, the basic aims of the method remain the same. These aims are to improve the biophysical and pharmaceutical characteristics of proteins by modifying pharmacokinetics (circulating serum half-life); increasing resistance to proteolysis; reducing antigenicity and immunogenicity; and in some instances, increasing solubility and reducing propensity to aggregate. These improvements have been demonstrated successfully in the clinic with a variety of proteins including enzymes, cytokines, and antibodies. In this chapter we will introduce the aspects of PEGylation common to all proteins before dealing with their specific application to antibodies and antibody fragments.

POLYMERS FOR PROTEIN CONJUGATION

Many potential therapeutic proteins have characteristics that can be improved by conjugation to large water-soluble polymers. Tailoring of these characteristics is required in order to generate the most effective therapeutic. Alteration of a protein's characteristics may also expand its use, for example, from single use in acute indications to repeat dosing in chronic indications. Conjugation of both small molecule and protein-based drugs to a diverse range of polymers has been investigated in order to improve their therapeutic profile.

Recombinant human antibody repertoires are now used routinely for the identification of individual antibodies with defined specificities to any conceivable antigen. The generation of large libraries (>1010) has been reported from many commercial and academic laboratories, along with a growing number of examples of isolated antibodies in clinical development for a range of therapeutic applications. Our laboratory has constructed nonimmunized libraries of human scFv antibody fragments with a combined size of >1011 transformants that have been used for more than ten years to successfully isolate antibodies suitable for clinical development.

LIBRARY DESIGN CONSIDERATIONS

Natural Antibody Diversity

The common aim of all nonimmunized recombinant antibody libraries is to mirror the immune system's ability to provide binding specificity to any antigen. For naïve human antibody libraries this is achieved by capturing the full spectrum of antibody sequences available from the human B cell repertoire.

The primary repertoire of variable heavy and light chain DNA sequences is generated by the recombination of V, J, and in case of the heavy chain, also D gene segments, which can recombine to give 7,650 (16,218 considering the use of multiple reading frames for the D segments) different VH and 324 different VL sequences (Corbett et al., 1997; Nossal, 2003).

Antibodies were discovered in 1890 but remained on the periphery of the pharmaceutical industry for more than 100 years. Yet within the last 15 years, a succession of antibodies has been approved for therapy by the United States Food and Drug Administration (FDA). Unlike natural antibodies which are polyclonal and directed against infectious disease, almost all those approved by the FDA are monoclonal antibodies directed against human self-antigens and used for treatment of cancer and diseases of the immune system.

Two major breakthroughs proved necessary to launch this antibody revolution. The first breakthrough was rodent hybridoma technology in the 1970s. Antibodies could now be made against single antigens in complex mixtures and used to identify the molecular targets of disease. In some cases this allowed disease intervention by blocking the antigen or by killing a class of cells (such as cancer cells) bearing the antigen. However, hybridoma technology provided only part of the solution; the rodent antibodies proved immunogenic and often did not trigger human effector functions efficiently. The second breakthrough, in the 1990s, was protein engineering; its application allowed the creation of chimeric and humanized antibodies from rodent monoclonal antibodies; not only were these less immunogenic than rodent antibodies, but they more efficiently triggered human effector functions. These chimeric and humanized antibodies now account for the majority of the currently approved therapeutic antibodies.

Nevertheless the field continued to embrace new technologies and to spawn new approaches, most notably the development of genuine human antibodies in the 1990s.

The discovery of the monoclonal antibody technology by Milstein and Kohler paved the way for antibodies of desired specificity to be made in quantities that could enable large clinical trials, and heralded the start of the antibody targeted-therapy era. Numerous clinical trials were conducted using murine antibodies derived from the spleen cells of immunized mice and myeloma cells. A major drawback to the use of these murine, xenogeneic antibodies in man was the development of a human anti-murine antibody response (HAMA) against both the constant and variable regions of the antibody. This response rarely led to anaphylactic or other hypersensitivity reactions but did severely limit the number of administrations that could be made, and hence it often negated the therapeutic efficacy of these antibodies.

Studies in a number of laboratories paved the way to humanizing these murine antibodies (see chapter by Saldanha) and, as the advances in antibody technology increased, fully human antibodies with high affinity have been developed for clinical use. Today, antibodies are by and large combined with chemotherapeutics, and in this setting, have been shown to improve both the time to disease progression and survival in patients with a wide spectrum of tumors. Combination therapy in oncology is an established protocol, as it is necessary to target various molecular events of the tumor cell as well as antigens preferentially expressed by such tumor cells.

In antibody-directed enzyme prodrug therapy (ADEPT), an antibody is used to target an enzyme to tumor. After tumor localization and deactivation or clearance of enzyme from blood and other normal tissue, a prodrug is given. The prodrug is converted into a toxic chemotherapeutic by the pretargeted enzyme at the tumor site (Figure 22.1). The ADEPT system, originally conceived in 1987, has a number of potential advantages over standard chemotherapy or the use of antibody-toxin conjugates. If a relatively nontoxic prodrug is used and there is no significant conversion of prodrug in nontarget organs, toxicity is restricted to the tumor site, allowing highly potent and specific treatments. Moreover, since one enzyme is able to turn over many prodrug molecules, the tumor essentially becomes a factory for generating its own means of destruction. Importantly, active drug can also diffuse to nearby cells, creating a local bystander effect where antigen negative cells and tumor-supportive stromal elements are destroyed.

ADEPT is a complex system that can be influenced by many components. These components, outlined inFigure 22.2, have been investigated by various workers over the last 2 decades and the results provide a platform of understanding for future applications of the treatment. Here we describe the progress of ADEPT since the first proofs-of-principle to recent advances in the clinic.

Unconjugated, target-cell killing antibodies of the human IgG1 isotype are now established as successful therapeutic agents, as demonstrated by the use of rituximab and trastuzumab for the treatment of B cell malignancies and Her2-overexpressing breast cancer, respectively. While both Fc-dependent and independent mechanisms can contribute to the efficacy of these drugs, it is clear that for both rituximab and trastuzumab, significant in vivo target cell depletion requires the Fc portion of the antibody. In vivo, the Fc region may either engage complement activation and/or interact with Fcγ receptors that are important for cellular immune effector functions such as antibody-dependent cell-mediated cytotoxicity (ADCC), which can be mediated by various effector cells such as natural killer (NK) cells and macrophages.

Increasing evidence indicates an important role for the interaction of antibodies with FcγRIIIa. In particular, retrospective studies have correlated superior objective response rates and progression-free survival with being homozygous for the higher affinity allele of FcγRIIIa encoding a valine residue at position 158.– Only approximately 15% of the population is homozygous for this form of the receptor. Therefore, it may be valuable to generate therapeutic antibody variants that bind to all forms of this receptor with at least as high affinity as current IgG1 antibodies bind to FcγRIIIa-158V.

Both the polypeptide chain and the oligosaccharide component may be engineered in order to increase affinity for FcγRIII. We have chosen the latter path and first demonstrated that recombinant engineering of the glycosylation pattern of antibodies generates antibody glycosylation variants with increased FcγRIII binding affinity and increased ADCC. As explained in more detail below, this was achieved by overexpression of a glycosyltransferase gene in Chinese hamster ovary (CHO) cells, which are the preferred and established cell host for the commercial production of therapeutic antibodies.