Therapeutics

Adv ances in analyt ical biophys ical to ols to ac cel erate the dis covery of t hera peutic a ntibodies

The innate ability of to bind their targets with high affinity and exquisite specificity is leveraged in the discovery of therapeutic antibodies to treat a broad range of diseases including cancer, autoimmune disease and heart disease. In addition, antibodies are commonly used as companion diagnostics and reagents to support the therapeutic pipeline and to probe antigenic surfaces to inform vaccine design. Discovering antibodies with appropriate characteristics for a given purpose relies on their proper selection from the prolific number of clones that are routinely produced by modern generation methods, including hybridoma technology in both normal and transgenic animals, phage display and synthetic libraries. Triaging large panels of antibodies to a few leads is normally followed by substantial engineering to optimise their binding affinity, minimise immunogenicity and improve developability. This article reviews the current and emerging label-free biosensor tools that are used to characterise the binding interactions of antibodies in terms of their kinetics, affinities and specificities from early stage screening to the clinic, with emphasis on throughput.

ithin the animal kingdom, antibodies ism and their classical presentation is as a Y-shaped By Dr Yasmina have evolved in jawed vertebrates glycoprotein comprised of heavy and light Noubia Abdiche W (gnathostomes) to protect these organ - polypeptide chains belonging to the immunoglobu - isms against pathogens and parasites 1. Antibodies lin (Ig) superfamily of proteins. However, some bind their targets with high affinity and high speci - species naturally produce other antibody formats, ficity, which makes them appealing as both thera - such as the small heavy chain-only fragments peutic agents and companion diagnostic reagents, known as VHH or nanobodies in camelids 2 and and these products are generating lucrative sales in the Ig new receptor (IgNAR) in sharks 3. the pharmaceutical industry. Antibodies are a class Theoretical possibilities for unique-sequence anti - of specialised molecules produced and secreted by bodies are almost unlimited due to their architec - B lymphocytes in the immune system of an organ - ture, where stretches of conserved framework

Drug Discovery World Summer 2018 55 Therapeutics

& $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $

' (

Figure 1: SPR kinetics on a 384-array (a) tile view with each tile representing a discrete mAb-coated spot and (b) close-up view of three mAbs with diverse kinetic profiles (App Note, Carterra)

References residues descended from a limited set of germlines interactions in terms of their kinetic rate constants 1 Klein, Jan and Nikolaidis, alternate with hypervariable ‘complementary- and affinities 4. Knowledge of an interaction’s affin - Nikolas. The descent of the determining regions’ (CDRs), which are responsi - ity is significant because it dictates the dose or con - antibody-based immune ble for the enormous sequence and structural diver - centration at which an antibody will be effica - system by gradual evolution. 5 PNAS January 4, 2005. 102 (1) sity of an organism’s antibody repertoire. While cious . Oftentimes, selecting an appropriate affini - 169-174. antibody generation in the pharmaceutical indus - ty for a given application is an empiric process and 2 Arbabi-Ghahroudi, M. try is highly commoditised, with modern in vivo can be optimised by engineering. Dissecting a bind - Camelid Single-Domain and in vitro libraries typically producing vast num - ing affinity into its constituent kinetic rate con - Antibodies: Historical bers of clones, the analytical tools used to charac - stants provides useful information that can guide Perspective and Future Outlook Front Immunol. 2017 terise their binding properties in terms of kinetics, the evaluation of an antibody’s performance Nov 20;8:1589. affinity and specificity, which are key parameters throughout screening, optimisation and manufac - 3 Feige, Matthias J, Gräwert, for assessing their quality and functional activity, ture. Label-free methods obviate the need to label Melissa A, Marcinowski, Moritz, lag orders of magnitude behind in throughput. or conjugate either of the interacting species and Hennig, Janosch, Behnke, Julia, With therapeutic antibodies being the largest class the binding event is monitored in real-time. This Ausländer, David, Herold, Eva M, Peschek, Jirka, Castro, of biotherapeutic proteins that are in clinical trials, makes it ideally suited for screening crude antibody Caitlin D, Flajnik, Martin, there is an ever-increasing demand for higher samples that are produced in early-stage research, Hendershot, Linda M, Sattler, throughput analytical methods that can match the where numerous clones are available, but in limit - Michael, Groll, Michael and capacity of antibody production and guide the ed quantities. Buchner, Johannes. The library-to-leads triage. Since taking an antibody Commonly-used commercially-available biosen - structural analysis of shark IgNAR antibodies reveals from bench to the market is estimated to cost sor platforms differ in their throughput, sample evolutionary principles of about $1 billion, there is a need to make antibody consumption and ease of use. They are versatile immunoglobulins. PNAS June screening more efficient and comprehensive to cut tools and are applied to the study of a broad range 3, 2014. 111 (22) 8155-8160. costs and timelines. of biomolecular interactions, from small molecules 4 Hearty, S, Leonard, P, to antibodies. Unlike small molecule analysis, O’Kennedy, R (2012). Measuring Antibody – Antigen Using SPR to characterise the where binding interactions are generally transient Binding Kinetics Using Surface binding kinetics and affinities with weak affinities, as characterised by equilibri - Plasmon Resonance. In: of antibody interactions um dissociation constants (or KD values) in the Chames P. (eds) Antibody Label-free biosensors such as those based on opti - micromolar range, antigen/antibody binding inter - Engineering. Methods in cal detection principles such as surface plasmon actions often exhibit affinities that are a million - Molecular Biology (Methods and Protocols), vol 907. resonance (SPR) or biolayer interferometry (BLI) fold tighter, with KD values in the picomolar range. Humana Press. are routinely employed in the pharmaceutical Accurately measuring the kinetic rate constants of Continued on page 57 industry to characterise antigen/antibody binding high affinity interactions requires long binding

56 Drug Discovery World Summer 2018 Therapeutics

cycles to enable sufficient information to be gath - of several wash steps may also compromise the Continued from page 56 ered for both the association and dissociation ability to identify weak affinity binders, which may phases 6. Therefore, biosensor platforms that have desirable . 5 Rudnick, Stephen I and Adams, Gregory P. Affinity and increase the number of interactions monitored in Another option is the Octet HTX platform from Avidity in Antibody-Based parallel can significantly accelerate the analysis of Pall/ForteBio, which is a 96-channel BLI biosensor Tumor Targeting. Cancer large panels of antibodies. offering versatility in throughput from 8, 16, 32, Biotherapy And A recent study benchmarked the Mass-1 system 48 to 96-channel modes. Since each interaction Radiopharmaceuticals Volume from Sierra Sensors against the Biacore 4000 sys - necessitates the use of single disposable sensors, 24, Number 2, 2009. 6 Katsamba, PS, Navratilova, I, tem from GE Healthcare and found comparable running the system in 96-channel mode is costly on Calderon-Cacia, M, Fan, L, kinetic rate constants for a large panel of sensors and sample consumption scales with the Thornton, K, Zhu, M, Bos, TV, antigen/antibody interactions when chip types size of the assay. Also, the lack of microfluidics Forte, C, Friend, D, Laird- were matched 7. However, with the throughput of affects the accuracy with which binding kinetics Offringa, I, Tavares, G, Whatley, these two state-of-the-art instruments limited to can be measured, because they are often contami - J, Shi, E, Widom, A, Lindquist, KC, Klakamp, S, Drake, A, monitoring only eight interactions at once via par - nated by rebinding artifacts, which confound a Bohmann, D, Roell, M, Rose, L, 8 allel flow channels, these analyses not only con - reliable and detailed analysis . Indeed, a study by Dorocke, J, Roth, B, Luginbühl, sume significant quantities of samples but necessi - Estep et al in 2013 stated that, “When measuring B, Myszka, DG. Kinetic analysis tate long run times when addressing hundreds of affinities based on currently available methods, of a high-affinity interactions, making the analysis of more than 100 one must compromise either on throughput or antibody/antigen interaction performed by multiple Biacore clones rather tedious. As such, SPR is often used as accuracy”, which unfortunately has posed a bottle - users. Anal Biochem. 2006 May 9 a secondary tool, after large antibody panels have neck in analytics for decades . In that paper, they 15;352(2):208-21. Epub 2006 been whittled down to a subset of clones via pre - compared the affinities of antigen/antibody inter - Feb 23. liminary screening by higher throughput, but lower actions measured by orthogonal techniques, name - 7 Kamat, Vishal, Rafique, information content assays, such as ELISA, which ly the BLI-based 16-channel Octet RED384 plat - Ashique. Exploring sensitivity & throughput of a parallel flow is an end-point analysis lacking real-time kinetic form and low and high throughput solution affini - SPRi biosensor for information. ELISA also is prone to a high false hit ty methods, respectively the Kinetic Exclusion characterization of antibody- & $ $ $rate, espec$ially$ whe$n targets a$re ads$orbed $directly$ $ Assay (Kin$ExA) and Mes$o Scale Discovery (MSD$). antigen interaction. Analytical to plastic which may expose irrelevant, denatured The authors reported an excellent correlation Biochemistry 525 (2017) 8e22. epitopes or mask native ones. In addition, the use between KinExA and MSD affinity determinations 8 Abdiche, Y, Malashock, D, Pinkerton, A, Pons, J. Determining kinetics and affinities of protein interactions using a parallel real-time label-free biosensor, the Octet. Anal Biochem. 2008 Jun 15;377(2):209-17. 9 Patricia Estep, Felicia Reid, Claire Nauman, Yuqi Liu, Tingwan Sun, Joanne Sun and Yingda Xu. High throughput solution-based measurement of antibody-antigen affinity and epitope binning. mAbs 5:2, 270–278; March/April 2013. 10 Salimi-Moosavi, H et al. Rapid affinity measurement by Gyrolab. Anal. Biochem. 426 (2012) 134-141.

Figure 2: Epitope binning a panel of 29 mAbs via BLI into 4 distinct clusters 17 Continued on page 58

Drug Discovery World Summer 2018 57 Therapeutics

Continued from page 57

11 Adler, Adam S, Bedinger, Daniel, Adams, Matthew S, Asensio, Michael A, Edgar, Robert C, Leong, Renee, Leong, Jackson, Mizrahi, Rena A, Spindler, Matthew J, Rao Bandi, Srinivasa, Huang, Haichun, Tawde, Pallavi, Brams, Peter and Johnson, David S (2018). A natively paired antibody library yields drug leads with higher sensitivity and specificity than a randomly paired antibody library, MAbs. 2018 Jan 29:1-13. 12 Ching, KH, Collarini, EJ, Abdiche, YN, Bedinger, D, Pedersen, D, Izquierdo, S, Harriman, R, Zhu, L, Etches, RJ, van de Lavoir, MC, Harriman, WD, Leighton, PA. Chickens with humanized immunoglobulin genes generate antibodies with high affinity and broad epitope coverage to conserved targets. MAbs. 2018 Jan;10(1):71-80. 13 Kaplon, Hélène and Reichert, Janice M. Antibodies to watch in 2018. (2018), mAbs, 10:2, 183-203. 14 Pellesi, L, Guerzoni, S, Pini, LA. Spotlight on Anti-CGRP Monoclonal Antibodies in Migraine: The Clinical Evidence to Date. Clin Pharmacol Drug Dev. 2017 Nov;6(6):534-547. 15 Van Blarcom, T, Rossi, A, Foletti, D, Sundar, P, Pitts, S, Bee, C, Melton, Witt J, Melton, Z, Hasa-Moreno, A, Shaughnessy, L, Telman, D, Zhao, L, Cheung, WL, Berka, J, Zhai, W, Strop, P, Chaparro-Riggers, J, Shelton, DL, Pons, J, Rajpal, A. Precise and efficient antibody epitope Figure 3: Epitope binning a panel of 106 mAbs via Array SPR into several distinct clusters 12 . Network plots are determination through library coloured by orthogonal data to provide an even more comprehensive analysis; (a) to the target’s design, yeast display and next- subdomains (labelled alphabetically), (b) mAb library, and (c) cross-reaction to the mouse ortholog generation sequencing. J Mol Biol. 2015 Mar 27;427(6 Pt B):1513-34. when sufficient time was allowed for the interact - interaction being measured is performed in solu - 16 Rojas, Gertrudis, Tundidor, Yaima and Cabrera Infante, ing samples to reach equilibrium. In contrast, the tion phase. Affinity-ranking can help triage clones, Yanelys (2014). High throughput correlation with BLI was rather poor, as it consis - but oftentimes, dissecting affinities into their kinet - functional epitope mapping: tently underestimated the affinity of low-picomo - ic components, as provided by real-time biosensor Revisiting phage display platform lar antibodies, reaching an apparent affinity limit analyses is important in evaluating clones for a to scan target antigen surface, at 100pM, likely due to a combination of loading given MOA. mAbs, 6:6, 1368-1376. level limitations and rebinding artefacts. While To accelerate the kinetic screening of a large binding affinities determined via both MSD and panel of antibodies, Array SPR has been gaining Gyrolab platforms 10 have shown excellent correla - popularity as evidenced by several publications tion with KinExA measurements, these equilibri - reporting kinetic and affinity measurements col - um-based assays rely on secondary detection, so lected on single arrays comprising 96 or more Continued on page 59 are not considered label-free, even though the clones 11,12 . These highly parallel analyses employ

58 Drug Discovery World Summer 2018 Therapeutics

a ‘one-on-many’ ‘analyte-on-ligand’ assay format practical methods include epitope mapping using Continued from page 58 that not only speeds up kinetic measurements, but libraries of antigenic variants produced by display is also very efficient on antigen consumption, methods 15,16 . 17 Sapparapu, G et al. Neutralizing human antibodies necessitating only a single aliquot per analyte con - A facile and quick way of identifying antibodies prevent Zika virus replication centration to perform a kinetic analysis on a large that cluster into epitope families is provided by epi - and fetal disease in mice. array of antibodies. Indeed, measurements per - tope binning experiments which can be performed Nature 2016 Nov; 540(7633): formed on Carterra’s Array SPR biosensor support in high throughput and with low sample consump - 443-447. 384 antibody arrays as standard and deliver high tion on biosensors. Epitope binning is a competi - 18 Abdiche, YN, Miles, A, Eckman, J, Foletti, D, Van Blarcom, quality binding data, as shown in Figure 1 . tive assay where antibodies are tested in a pairwise TJ, Yeung, YA, Pons, J, Rajpal, A. and combinatorial manner for their ability to bind High-throughput epitope binning Using SPR to characterise the binding simultaneously to their specific antigen. If both assays on label-free array-based specificity of antibody interactions antibodies can bind at the same time, they are pre - biosensors can yield exquisite The region of an antigen that is recognised by an sumed to target distinct, non-overlapping epitopes, epitope discrimination that facilitates the selection of antibody is referred to as the ‘epitope’. Knowing whereas if one appears to block the other, it is monoclonal antibodies with an antibody’s epitope is highly relevant to the suc - inferred that they compete for overlapping or sim - functional activity. PLoS One. cess of a therapeutic antibody programme because ilar epitopes. With the now routine implementa - 2014 Mar 20;9(3):e92451. it largely dictates an antibody’s biological function tion of label-free biosensors for epitope binning, 19 Sivasubramanian, Arvind, or mechanism of action (MOA). An antibody’s epi - there is a need for higher throughput platforms Estep, Patricia, Lynaugh, Heather, Yu, Yao, Miles, Adam, Eckman, tope is an innate property that can neither be pre - that enable more antibodies to be tested in a single Josh, Schutz, Kevin, Piffath, dicted nor rationally designed by in silico methods, assay to explore the depth and breadth of an anti - Crystal, Boland, Nadthakarn, and so selecting an antibody with an appropriate body campaign in a comprehensive manner. BLI- Hurley Niles, Rebecca, Durand, epitope to fulfil a given therapeutic goal is an based platforms are commonly used for epitope Stéphanie, Boland, Todd, empirical process. Furthermore, since an anti - binning but since sample consumption scales with Vásquez, Maximiliano, Xu, Yingda and Abdiche, Yasmina (2016). body’s epitope cannot be shifted or optimised by the size of the antibody panel investigated, these Broad epitope coverage of a 17 engineering in a rational way, it is advantageous to assays are often limited to rather small panels human in vitro antibody library. survey the epitope landscape of an antibody library (Figure 2 ). In contrast, Array SPR methods enable mAbs, 9:1, 29-42. at the earliest possibility to identify and move for - larger panels to be analysed on a single chip with ward only those clones exhibiting the most appro - published examples demonstrating these analyses priate epitopes or a variety of epitopes that can be on 96- 18 , 192- 12 (Figure 3 ), and 384-antibody tested in functional assays to converge upon leads. arrays 19 . From an intellectual property perspective, claiming The demand for higher throughput analytics is novel epitopes enables companies to have an edge needed to support the paradigm shift that is in the fiercely competitive market of therapeutic emerging in the industry towards a deeper appre - antibody discovery, with multiple companies often ciation of the epitope diversity offered within an working on the same targets, as evidenced by the antibody library by applying higher resolution and clinical pipeline. For example, the immune check - higher throughput methods for epitope character - point modulators PD-1 (programmed death recep - isation at the earliest stages of research. This tor-1) and its ligand (PD-L1) remain popular tar - enables the identification of uniquely-suited clones gets for treating cancer, despite five antibodies tar - and minimises the need for their extensive engi - geting the PD-1 pathway already on the market. neering, which streamlines library-to-leads and Currently, in clinical development there are 21 ultimately cuts costs to patients by delivering molecules targeting PD-1, including five in late- medicines faster. DDW stage clinical studies and nine antibodies targeting PD-L1 13 . Another example is in the prevention of migraine, where multiple companies have antibod - ies targeting CGRP (calcitonin gene-related pep - tide) or its receptor in late-stage clinical trials 14 . There are many analytical techniques available Dr Yasmina Noubia Abdiche is Chief Scientific for characterising an antibody’s epitope, with the Officer at Carterra. Dr Abdiche joined Carterra as gold standard for precise epitope determination CSO in 2016 after 12.5 years’ experience in the provided by atomic-level structural data produced pharmaceutical industry at Rinat-Pfizer, where she by x-ray crystallography and cryo-electron led a group of analytical scientists that applied microscopy. However, these methods are labour- label-free biosensors to the discovery of therapeu - intensive and slow, and therefore used as confirma - tic antibodies. Dr Abdiche is co-inventor of several tory tools, rather than research-based ones. More therapeutic antibodies in clinical trials.

Drug Discovery World Summer 2018 59