CHAPTER 1
BIOPHYSICAL METHODS APPLIED IN EARLY DISCOVERY OF A BIOTHERAPEUTIC: CASE STUDY OF AN EGFR-IGF1R BISPECIFIC ADNECTIN

Michael L. Doyle, James W. Bryson, Virginie Lafont, Zheng Lin, Paul E. Morin, Lumelle A. Schneeweis, Aaron P. Yamniuk, and Joseph Yanchunas, Jr.

Protein Science and Structure Department, Bristol-Myers Squibb Research and Development, Princeton, NJ, USA

1. 1.1 Introduction
2. 1.2 Target Identification
3. 1.3 Target Generation
   1. 1.3.1 Multiple Constructs Strategy
4. 1.4 Hit Evaluation
   1. 1.4.1 Qualitative and Rapid Self-Association Check
   2. 1.4.2 Qualitative and Rapid Thermal Stability Check
   3. 1.4.3 Confirmation of Binding
5. 1.5 Lead Selection
   1. 1.5.1 Self-Association
   2. 1.5.2 Thermal Stability
   3. 1.5.3 Binding Affinity, Kinetics, and Epitope
6. 1.6 Lead Optimization
7. 1.7 Lead Formatting
   1. 1.7.1 Solubility
   2. 1.7.2 Thermal Unfolding Behavior
8. 1.8 Final Development Candidate Selection
9. 1.9 Concluding Remarks
10. Acknowledgment
11. References

1.1 INTRODUCTION

Biophysical characterization of protein therapeutics and associated reagents in drug discovery is critical to selection and optimization of molecules that have the desired biological activity and to selection of drug candidates that can be efficiently developed and manufactured. Protein therapeutic molecules are larger and more complex than small-molecule drugs. Consequently, analytical strategies for determining whether a protein therapeutic is pure, stable, and homogeneous require that a larger number of physical properties be investigated, including characterization of tertiary and quaternary structures. Furthermore, several physical properties of protein therapeutics, for example, aggregation state, require multiple, orthogonal methods to confidently define them (Table 1.1).

Table 1.1 Biophysical and biochemical methods used to characterize targets, reagents, and drug candidates for protein therapeutic discovery programs in terms of identity, purity, stability, oligomeric status, binding activity, and molecular binding mechanism

In addition to production and characterization of hundreds or thousands of drug candidates during drug discovery, a large number and diversity of protein reagents must also be produced and characterized. To begin with, the biological target must be produced in a form that is well behaved and representative of the functional form to be targeted in vivo. There are a multitude of other protein reagents needed to run the program as well (e.g., multiple affinity-tagged forms of the target for use in a variety of assays, truncated forms of the target for structural studies, counter-targets, co-targets, and nonhuman species ortholog variants of the target; Figure 1.1; see also Kim and Doyle [1] for a detailed listing). Target reagents that are aggregated or misfolded confound the drug discovery process during hit identification and downstream assays. The famous admonition “garbage in, garbage out” is often cited as a reminder that biophysically well-behaved reagents generally lead to higher success rates during lead identification and optimization of protein therapeutics. Biophysical methods thus play a wide variety of roles in the characterization of biotherapeutic candidates and protein reagents during the early discovery stages of biotherapeutics.

Figure 1.1 Scheme showing the different classes of protein reagents and drug candidates produced and characterized by biophysical methods from the initiation of a drug discovery program through selection of a final molecule for subsequent progression into development. Initially the protein production and biophysical characterization efforts are focused on the target(s) and reagents. As the program progresses, the amount of protein chemistry increases and shifts toward the production and characterization of protein therapeutic candidate molecules. The type and extent of biophysical characterization done for each class of protein and for each stage of discovery is different as described in Table 1.1.

Biophysical characterization is a central part of the selection and optimization process. But how much biophysical characterization is optimal for each type of reagent or biotherapeutic candidate molecule, and how does the extent of biophysical characterization change during each stage of the discovery process? The goals of this chapter are to describe the types of biophysical methods that are used in a stage-dependent manner throughout discovery for reagent and drug candidate production of protein therapeutics and to discuss how the application of these methods in discovery help to de-risk the potential costly challenges later in the development and manufacturing phases.

The discovery process is described in this chapter by several stages: target generation, hit evaluation, lead selection, lead optimization, lead formatting, and final lead candidate selection of a molecule to progress into development. We note that the types and extent of biophysical characterization will depend to some degree on the molecular class of the protein therapeutic (monoclonal antibody, Adnectin, antibody fragments, non-antibody fragments, etc.) and the technology used for selecting lead candidates (immunizations, phage display, RNA display, etc.). The purpose here is to present a case study of biophysical applications during the discovery of a bispecific Adnectin against epidermal growth factor receptor (EGFR) and insulin-like growth factor-1 receptor (IGF1R). Many of the details for this system have been reported elsewhere [2].

1.2 TARGET IDENTIFICATION

Identification of a drug's biological target is a critically important part of a biotherapeutic discovery program. One of the expanding areas in biotherapeutics research is the design of bispecific biotherapeutics that bind to two different, already validated biological targets. The proposed benefits for the bispecific-targeting approach include improved efficacy and lower cost of goods than developing two drugs independently.

Drug targets may also be identified from genetic validation studies (correlation between mutation of target and disease state) or pharmacological validation studies (utilizing a surrogate molecule such as a natural ligand to demonstrate efficacy in a non-clinical setting). The Holy Grail for identification of completely novel targets is to utilize the growing information from genomic, proteomic, and interactomic studies to draw correlations between specific drug targets, or sets of drug targets, and treatment of disease.

This chapter describes a case study for discovery of an Adnectin [3] bispecific biotherapeutic that targets inhibition of both EGFR and IGF1R (Emanuel et al. [2]). EGFR is a clinically validated target for cancer therapy, and there are both small-molecule kinase inhibitors and biotherapeutic inhibitors of the extracellular domains presently available as marketed drugs. IGF1R is also an attractive target for cancer therapy and there are several small-molecule and biotherapeutic inhibitors in preclinical and clinical studies [4].

1.3 TARGET GENERATION

Once a target has been identified, it is usually produced recombinantly to provide sufficient material to enable selection of biotherapeutic candidate “hits” through a screening or selection process. There are several technologies commonly used for generating biotherapeutics hits, including in vivo immunization, phage display, mRNA display, and yeast display [5,6]. All of these technologies rely on the production of biophysically well-behaved target molecule. Biophysical methods thus play a critical role as “gate-keeper” at this phase of discovery, to ensure the quality of the target being used for screening or selections is suitable for generating the best candidates.

The first step in producing the target reagent is to engineer a form of the target molecule that will be expressed well and has acceptable biophysical behavior when purified. Sometimes the design is fairly straightforward. For instance, the construct design, expression, and purification for some targets may be well described in the literature. Construct design may also be straightforward if the protein target itself is structurally small and simple. An example would be a soluble target such as a cytokine. The construct design of a simple small protein could be as straightforward as expressing the entire native protein. On the other hand, construct design of large membrane-spanning protein targets can be much more challenging since the membrane-spanning and intracellular regions usually need to be deleted in order to make well-behaved soluble extracellular fragment(s) of the target. Whether or not some or all of the extracellular domains extracted from the full-length protein can be expressed, purified, and well behaved biophysically is often not known in advance.

1.3.1 Multiple Constructs Strategy

Given significant uncertainties and risk surrounding the production of critical target molecules, it is prudent to approach the problem with the design of multiple constructs in parallel, at least through DNA expression vector or small-scale expression screening stages. There are several reasons for designing multiple constructs up front...