Biolayer interferometry and its applications in drug discovery and development

A typical BLI experiment begins with an equilibration step where the sensor tip is dipped into an assay buffer to establish a baseline. The next step, called loading, is performed to immobilize molecules on the biosensor tip; the biosensor is dipped into a solution of biotinylated/His-tagged ligand. After immobilization, biosensor tip is dipped into the buffer solution to assess baseline drift and the ligand loading level. Then, an association step takes place where the sensor tip is dipped into an analyte solution. Interactions between the immobilized ligand and its binding partner are measured. In the dissociation step, the sensor tip is dipped into an analyte-free buffer so that the molecular complexes can dissociate. Binding responses are recorded and shown in real time in the form of a sensorgram (Fig. 3). Besides providing information about binding affinity and kinetics, the shape of the sensorgram can disclose the stoichiometry of molecular interaction, whereas the signal intensity can be used for analyte quantitation.

Several biophysical methods have emerged in the last 25 year and are used to interrogate molecular interactions, they provide quantitative data as the equilibrium dissociation constant K_d and stoichiometric ratios. Beside K_d SPR and BLI provide also kinetic parameters of binding, isothermal titration microcalorimetry (ITC) provides thermodynamic parameters of binding (enthalpy and entropy changes) and microscale thermophoresis (MST) can provide affinity data in highly complex matrices [5] The BLI technology was developed to overcome issues related to SPR-based analysis kinetic parameters of biding. In SPR, microfluidics is used to deliver samples to a biosensor surface with immobilized molecules of interest, whereas BLI uses the ‘dip-and-read’ approach where biosensor tips are immerged into the microtiter wells containing samples. Microfluidic channel clogging is a common SPR issue that requires regular machine maintenance. Open, fluidics-free system in theory allows unlimited length of association and dissociation phases that in practice are only limited by evaporation. In contrast, the SPR's closed system is limited by the microfluidics volume. With BLI, usually several analyte concentrations are run in parallel in 96- or 384-well plates. This means that the entire analyte concentration series can be recorded in a single experimental run which significantly reduces assay time and allows for a higher throughput. Microplate format also allows sample volumes to be as small as 40 μL. Biosensors and sample solutions are not destroyed or consumed during experiments. Indeed, biosensors can often be recovered with regeneration buffers and reused which reduces the experimental costs. Only the molecules bound to the biosensor surface affect measurements (Fig. 1) which means that neither BLI nor SPR technologies are affected by the refractive index or the viscosity of the sample. Thus, both are compatible with crude sample analysis, while BLI is also suitable for directly analysing samples containing relatively high percentages of DMSO or glycerol which can present a problem in SPR measurements.

Both SPR as well as BLI are regarded as mid-to high-throughput methods capable of measuring thousands of analytes per day. Capabilities are constantly increasing with the instrumentation automatization and miniaturization, whereas the actual throughput depends on the details of the assay setup.

Advantages and disadvantages of BLI are even more evident when the technique is compared to other biophysical techniques used in the characterisation of compound:target interactions. Micro-scale thermophoresis (MST) is a biophysical method which monitors the changes in the thermophoretic motion of a fluorescently labelled or intrinsically fluorescent protein upon ligand binding. A thermal gradient is created by an infrared laser in a focused zone within a glass capillary containing the analyte:macromolecule complex. While MST does not require ligand immobilization, it often does require protein labelling (as in many cases the protein of interest is not intrinsically fluorescent). Additional information on the binding mode is limited, and K_d only is obtained. MST is sometimes employed in screening campaigns, after assay development using SPR or BLI failed [25]. BLI and MST can, on the other hand, be used as orthogonal methods for characterization of binding and confirming target specificity, as exemplified by Esposito et al. [26] for RNA aptamer binders of DNA methyltransferase 1.

A study comparing the Octet BLI-based platform to the well-established Biacore SPR-based platform was conducted in 2008 in which interactions between a murine monoclonal antibody and its cognate antigens human calcitonin gene-related peptide (CGRP), its rat ortholog and the truncated isoforms thereof were monitored [27]. Similar kinetic rate constants for monoclonal antibodies interacting with several immobilized peptides were obtained with both methods. A direct comparison study from 2016 employing a panel of ten high-affinity monoclonal antibodies against proprotein convertase substilisin kexin type 9 antigen showed that SPR technology exhibited excellent data quality and consistency, whereas BLI technology demonstrated high flexibility and throughput with compromises in data accuracy and reproducibility [28].

One way to overcome sensitivity problems is to perform in-solution affinity determination where neither of the binding partners are immobilized. Instead, the interacting partners are preincubated to reach equilibrium binding, and the solution is then exposed to biosensors coated with a molecule capable of selectively capturing the free target protein from the solution, resulting in a binding response. Quantification of unbound protein from mixtures containing increasing concentrations of binders enables determination of its affinity for the protein of interest [29]. However, this method has its limitations as only interactions at or below K_d can be measured.

Despite BLI's limited sensitivity with declining molecular mass of analytes, its other advantages such as fluidic-free system, high-throughput analysis, compatibility with crude sample analysis and the possibility to recover precious samples makes the technique a good complement to SPR and other biophysical and/or biochemical methods used in drug discovery [27].

BLI is well-suited for characterizing strong interactions between macromolecule binding partners, such as antibody:antigen pairs. However, as discussed in section 2 (BLI in small molecule drug discovery), with low-molecular-weight analytes, especially if affinity to the macromolecular target is moderate, achieving a measurable signal can be challenging. To enhance sensitivity, biosensors with a thicker layer of biocompatible polymer (sometimes referred to as the 3D platform) for ligand immobilization were developed. In general, increased thickness of the sensor layer together with high ligand density greatly improves sensitivity. Yet, dense packing of ligand molecules on a biosensor layer can in some instances be counterproductive. For example, Vignon et al. [30] have reported a BLI-based aptasensor for monitoring l-tyrosinamide (L-Tym), a compound with molecular mass of 180 Da. L-Tym-specific biotinylated aptamer was coupled to either conventional (i.e., 2D) or 3D streptavidin-coated sensors at different densities. The authors observed a decrease in apparent l-Tym affinity with increasing aptamer density. The phenomenon was attributed to steric hindrance as the analyte binding induces folding of the aptamer. Crowding was more pronounced on 2D compared to 3D sensors, since the same amount of immobilized aptamer led to higher ligand density on 2D sensors, being reflected in decreased k_on values, whereas the k_off values were not affected.

Non-specific sensor binding is another issue commonly encountered when analysing low affinity binding events regardless of the analyte size [31]. Quantitative characterization of interactions requires relatively high analyte concentrations (up to 10 × the K_d value or even higher) at which small molecule analyte colloidal aggregates may form [32] resulting in promiscuous binding to the sensor surface. A separate reference sensor lacking the ligand (or ideally one to which an inactive ligand analogue is coupled) is therefore used to account for non-specific analyte binding to the sensor surface. Still, specific binding signals can be relatively small and thus get lost in the signal subtraction error. Inclusion of certain additives in assay buffers was shown to effectively suppress the non-specific binding in BLI experiments [31]. Non-ionic detergents, such as polysorbate 20, are used to re-solubilize colloidal aggregates to monomers and to limit hydrophobic interactions, whereas increasing ionic strength or adjusting the pH of the binding buffer can help to limit electrostatic interactions. On the other hand, bovine serum albumin or casein are commonly added to shield protein analytes from engaging in non-specific interactions, and recently saccharide osmolytes, especially sucrose, were shown to be highly effective in suppressing non-specific binding by enhancing protein solvation and hence attenuating protein analyte aggregation [31]. Furthermore, a negative control (binding buffer only) is regularly included to correct for the signal drift resulting from ligand leaching, which may obscure weak but specific binding events. Importantly, promiscuous binders can often be differentiated from true ones by inspecting sensorgram profiles [24], a valuable feature when conducting BLI (or SPR) screening campaigns. In an early paper on the use of a BLI sensor for small molecule fragment screening, Wartchow et al. [24] described the behaviour of the so-called ‘atypical binders’. Such analytes may show high binding signals relative to a well-characterized positive control, very slow off-rates, and/or a non-saturating response (i.e., a linear correlation between analyte concentration and response). Of note, the mere fact that an analyte reaches signal plateau at high concentrations does not necessarily indicate specific binding. Rather than saturation of ligand binding sites, such behaviour may be attributable to surpassing the solubility limit resulting in analyte precipitation [24]. Conversely, unusually high responses are not exclusively associated with non-specific small analyte binding, as exemplified by the specific interaction between ADP and the N-terminal domain of p97 AAA + ATPase [28]. Here, conformational change of the ligand imposed by analyte binding was reported to produce unusually strong BLI signals.

Drug discovery is a complex, multidisciplinary process. In most cases it starts with target identification and validation. Once this step is confirmed with sufficient evidence, scientists can move on to the next step, finding modulators of the target, which will alter its function in such a way that it cures the disease or prevents its further development. Here, different screening techniques [34,35] are applied to narrow down binders from large libraries. BLI's ability to differentiate between non-ideal and ideal binding interactions makes it advantageous over various biochemical methods [36].

BLI's fluidics-free system and the fact that measurements are not affected by refractive index and viscosity enables screening of complex libraries such as microbial [37] and plant extracts [38] or large synthetic peptide libraries [39]. Biosensor-captured compounds can be identified by chromatographic isolation and chemical analysis by MS and/or NMR, and their binding validated with in vitro and in vivo bioassays or other biophysical methods such as ITC. BLI proved to be an effective high-throughput method for screening complex libraries.

Many researchers opt for fragment screening as it better covers chemical space in comparison to screening drug-like compounds. BLI was demonstrated to be a feasible method for fragment screening using the Maybridge RO3 library against the carbonic anhydrase. Screening of chemical space using fragments is especially useful in design of small-molecule protein-protein interaction (PPI) inhibitors since many proteins possess large binding surfaces and poorly defined binding pockets, which limits computational approaches and standard structure-based drug discovery. For example, BLI was used to screen fragment libraries for PPI inhibitors targeted against BCL-2, eIF4E and JNK1. Considering the number of binding sites on each target, 1.4 % hit rates are reasonable. Additionally, kinetic profiles obtained by BLI were similar to those obtained by SPR. Furthermore, BLI and SPR were able to weed out hits with non-ideal binding profiles and identify novel hits that were not picked up by biochemical assays [36].

Bothe et al. [40] reported the first crystal structure of AAA + ATPase p97 N-terminal domain in complex with a small molecule which was detected by BLI-based fragment screening. The mean molecular weight of fragments in the library was 202 Da, demonstrating that BLI is an appropriate method for screening low-molecular weight compounds. High-scoring hits according to Score_BLI (consensus ranking of hits based on obtained rate constants and sensorgram quality) were selected for further analysis. Low-scoring hits were eliminated as they possess undesired properties such as promiscuous binding. Binding of selected hits was confirmed with saturation-transfer difference (STD)-NMR and further studied with mixed-solvent molecular dynamics simulation. Similarly, Chimenti et al. [41] screened a fragment library against the ND1 domain of AAA + ATPase p97 using SPR. The hits were evaluated with STD-NMR and further analysed with virtual screening.

Comparison of the results from different studies [36,37] confirmed that BLI and SPR were able to identify fragments occupying similar chemical space. In addition, the BLI-based screen identified new fragments with distinct properties, making it a viable alternative to SPR.

In the optimization of phenoxyethylthio benzimidazole and phenoxyethylthio indole types of androgen receptor inhibitors, the effect of different substituents on the benzene ring and substitution of oxygen and sulfur atoms in the linker region was studied. SAR analysis was conducted using the eGFP transcriptional assay in a reporter cell line as well as a BLI binding study of drug analogues interacting with recombinant androgen receptor. BLI was used to confirm direct reversible binding of selected key compounds (sized 270–340 Da) [43].

BLI assays have also been used to validate SAR data obtained with other biochemical or biophysical methods such as fluorescence polarization (FP) binding assays. A BLI assay was used to confirm the direct binding of compounds to the Mcl-1 and Bfl-1 proteins [44]. Biochemical assay may be inappropriate for measuring IC₅₀ values of highly potent compounds. BLI was used to determine the activity and selectivity of highly selective miniprotein inhibitors of integrins α_vβ₆ and α_vβ₈, with all miniproteins having subnanomolar binding affinity [45].

In an attempt to develop competitive inhibitors of lysine acetyltransferases (i.e., histone acetytransferases), Simon et al. [46] employed a FP-based assay combined with BLI competition analysis for compound library screening and profiling of the hits, respectively. BLI was used to validate the hits obtained in the FP competition assay. To study compounds as competitive binders, the chosen hits were immobilized on a sensor surface and dipped into solutions of the target protein containing increasing concentrations of the known inhibitor acetonyl-CoA. The degree of competition was quantified by comparing signals at the end of the association phase for each acetonyl-CoA concentration. Plotting respective values against the ligand concentrations gave K_d value of 1.00 ± 0.28 μM for acetyltransferase protein binding to immobilized biotinylated CoA analogue compounds, which was in close agreement with K_d determined by FP (0.71 ± 0.08 μM) and ITC (1.39 ± 0.23), meaning that BLI is a viable method for validating competitive ligands. When testing the developed assay platform on other lysine acetyltransferase family members, one of them (p300 (1284–1673)) showed slow, tight binding to CoA analogues in FP assay. The same was observed in the BLI experiment, so the method had to be adjusted. This time, ligands of interest were not immobilized on biosensors but rather incubated with the target protein in solution, either in presence or absence of the competitive inhibitor NM. Association to biosensors from the solution was measured. Although sensorgrams did not provide kinetic data, they proved to be useful for demonstrating the competitive nature of binders with regard to NM.

A common aim in drug optimization is to maximize drug affinity for its biological target. However, off-target interactions can occur especially with proteins that are structurally and/or functionally related. Unwanted interactions lead to adverse effects, thus binding selectivity of the drug for its intended target must be confirmed before entering clinical trials [47]. One of the methods to monitor selectivity in vitro in the early stages of drug discovery is BLI.

Over 40 bromodomain-containing proteins, considered the ‘epigenetic readers’ of acetylated lysine residues on histones, in our body are classified into eight subfamilies. Bromodomain and extra terminal domain (BET) subfamily contains four members, each with two bromodomains. The use of BET inhibitors in oncology is restricted due of gastrointestinal toxicities and thrombocytopenia, caused by non-selective inhibition of different bromodomains. Design of C-terminal domain-selective BET inhibitors was supported by BLI to monitor their binding to different bromodomains. Results showing 200-fold selectivity for one of the designed compounds were consistent with those of thermal shift assay and homogeneous time-resolved fluorescence assay [48]. In another study, the selectivity of designed inhibitors over seven non-BET bromodomain-containing proteins and selectivity within the BET-subfamily was confirmed using BLI [49]. BLI was also used to confirm selective binding of novel tetrahydroquinoline inhibitors to one of the three NLRP3 inflammasome domains [50] and was used for providing selectivity data for amidobenzimidazoles analogues targeting stimulator of interferon gene (STING) receptor variants [51].

Additionally, BLI can be used to study site selectivity when labelling macromolecules. Zang et al. [52] confirmed that the monoclonal antibody trastuzumab retained its activity when a cysteine within a 7-residue tag (termed DBCO-tag), fused to the C-terminus of heavy chain was site-selectively labelled with a model aza-dibenzocyclooctyne (DBCO) reagent. The labelled DBCO-tag-trastuzumab showed full binding affinity to recombinant HER2 protein in a BLI assay (K_d = 39 pM). BLI was also used to validate and evaluate a Sirt2-selective mid-nanomolar affinity probe by showing a dose-dependent binding of Sirt2 to the immobilized SirReal probe [53].

Protein-protein interactions (PPI) interruption by small molecules is usually evaluated in a functional assay (e.g., GI₅₀ determination on cancer cell lines), while for the structure-based affinity optimization measuring binding kinetics parameters of designed PPI small molecule inhibitor with one of the interacting proteins can be used [13,[54], [55], [56]]. When the PPIs are inhibited by other protein molecules, interactions can directly be measured. This was shown for mammalian carbonic anhydrases, where the protein murine inhibitor of carbonic anhydrase acted as PPI disruptor [57]. Orthwein et al. [58], on the other hand, described an approach of overcoming relatively poor sensitivity in directly monitoring the binding of small moleculesto a sensor surface functionalized by a ligand involved in PPIs. Here, one of the proteins engaging in PPI was immobilized onto a sensor tip and a fixed amount of the other protein was preincubated in solution with increasing concentrations of the small molecule presumed to be a PPI inhibitor. The sensor tip was dipped into solution and the binding response was recorded. As protein:protein complexes are larger in size compared to protein:small molecule complexes, they gave signals exceeding instrument's noise and enabled determination of binding kinetics parameters and IC₅₀ of PPI inhibitors.

The robust setup of BLI experiments is valued when studying completely novel biomolecules. Computationally de novo designed protein binders are one of the challenges which raised a lot of interest. Cao et al. [59] evaluated nanomolar affinity of de novo designed miniprotein binders (all smaller than 65 amino acid residues) for 12 diverse protein targets with different shapes and surface properties using BLI, while Gainza et al. [60] computationally designed, produced and evaluated protein-protein interactions for a set of de novo proteins (40–120 amino acid residues) with four protein targets (SARS-CoV-2 spike, PD-1, PD-L1 and CTLA-4) using BLI.

BLI proved to be useful in development of proteolysis-targeting chimeras (PROTACs) [[61], [62], [63], [64]]. For their design and optimization, step by step profiling along ubiquitin-proteasome degradation pathway is needed to fully understand their structure-activity relationship. BLI is able to recognise secondary (protein A – PROTAC or protein B – PROTAC) and ternary complex (protein A – protein B – PROTAC) formation [65] and determine their affinity value which is essential for PROTAC characterization [66]. Until recently, PROTAC optimization was focused on improving affinity between the PROTAC warhead and the cognate target protein. This approach disregards the importance of ternary complex formation which began to be recognized in recent years [65]. Various biophysical methods were optimised for the purpose of studying ternary complexes [67], one of them being BLI. Schiemer et al. [68] conducted a study of cellular inhibitor of apoptosis 1 (cIAP1)-mediated degradation of Bruton's tyrosine kinase (BTK), where cIAP was immobilized onto streptavidin-coated biosensors and dipped into PROTAC solution. Kinetics of binary complex formation was measured and after a brief wash the binary complex was dipped into the BTK solution. The ternary complex was formed, and the binding constants of the binary cIAP1-PROTAC complex binding to BTK were measured. BLI facilitated determination of the system cooperativity through comparison of affinities in binary and ternary complexes (K_d (binary)/K_d (ternary)). Additionally, the authors provided valuable insight how a PROTAC linker modifications (PEG and pyrazine linker) result in a noticeable (4-fold) slower off rate and thus different affinities.

To treat diseases with unmet medical needs completely novel approaches are sought. One of the forefront approaches is targeting RNA with small molecules [69]. In the absence of biochemical assays, biophycial methods are being employed, especially BLI. For example, BLI was used as a tool for confirming the binding of positive hits obtained from the qFRET primary screen, namely the fluoroquinolones binding and stabilizing an RNA stem-loop [70]. Wang et al. [71] used a BLI assay to study affinity and selectivity of a hit compound for the binding to GG internal loops in the hairpin form of an expanded G4C2 repeat, which is associated with amyotrophic lateral sclerosis and frontotemporal dementia. Similarly, BLI was used to study affinity and selectivity of a small molecule binding to the expanded nucleotide repeats (r (CGG)) found in the 5′ untranslated region of the fragile X mental retardation 1 gene causing fragile X-associated tremor ataxia syndrome [72].

An interesting application of BLI technology in drug discovery and development is the so-called target fishing. It is used for target identification which helps to elucidate unknown molecular mechanisms and adverse drug effects. Dai et al. [73] used affinity chromatography coupled with mass spectrometry to identify the target of 1,2,4-oxadiazole derivative DDO-7263, a nuclear factor erythroid 2-related factor 2 (Nrf 2)/antioxidant response element (ARE) pathway activator. Proteomic analysis indicated that the 26 S proteasome non-ATPase regulatory subunit 11, known as Rpn6, directly interacted with the biotinylated DDO-7263 analogue, and the binding affinity was verified with BLI analysis. The assay was conducted in a way to either immobilize the target protein or the biotinylated DDO-7263 analogue to the streptavidin-coated sensor surface, and yielded K_d of 8.53 μΜ when the immobilized protein was dipped into the small-molecule solution, and K_d of 1.38 μΜ when immobilized small molecule was dipped into protein solution, confirming that the perceived interaction was real.

BLI is being intensively used in conjunction with mass spectrometry [74]. For example, Jung et al. [75] designed a protocol to elute proteins from the sensor surface directly after the BLI association step and analysed them by LC-MS/MS. Very recently, Luo and Yang presented an even more integrated approach; they directly coupled BLI with MS. An analyte was first captured on the sensor surface and subsequently directly eluted from the sensor inside an electrospray emitter and analysed by MS. The method named microprobe-capture in-emitter elution electrospray ionization mass spectrometry (MPIE-ESI-MS) [76] was used for the elucidation of the proteo- and glycoforms of β1-and β2-transferrin from the cerebrospinal fluid [77].

There are a number of studies describing BLI as a technique for examining binding affinities of small molecules to diverse macromolecules for various therapeutics fields, from antitumor [[78], [79], [80], [81], [82], [83]], cardiovascular [84] to antiviral [85,86] therapy. For the latter, the assays were conducted in a manner differing from conventional assays for monitoring small molecule-target protein interactions (where the immobilized protein is dipped into a small-molecule drug candidate, or vice versa). Here, cell receptors responsible for virus binding were immobilized onto streptavidin-coated biosensors. The virus was incubated with increasing inhibitor concentrations and interactions between cell receptors and viruses were monitored. Inhibition of viral envelope protein binding was quantified indirectly through cell protein-virus interactions.

Additionally, dual activity of designed viral inhibitors was analysed using BLI. Their potency was evaluated in either absence or presence of a molecule (oseltamivir carboxylate, OC) that is known to bind strongly to influenza A virus neuraminidase but not hemagglutinin. Calculations showed that in contrast to OC, the designed divalent inhibitors bound to both proteins which led to increased antiviral activity in cell assays [86].

BLI can be used to better understand already established high potency in cell-based assays. However, Han et al. [87] encountered some problems when analysing tight-binding compounds (K_d values in picomolar range) by BLI. For some of them, K_d values could not be calculated as their dissociation rates were too low to measure. As an alternative method they used differential scanning fluorimetry that was able to rank high affinity compounds.

Complex biological interactions can be accurately measured using BLI. Affinity of 20 kDa protein dimer (Max-Max 6) binding to immobilized DNA was accurately determined with K_d = 50 nM [88].

In the process of cell line development, host cells are transfected with an expression plasmid harboring the gene for the protein of interest. The expression cassette is randomly integrated into host cell genome, leading to formation of distinct cell clones that can differ considerably in production yield and quality. The compatibility with crude sample analysis combined with reasonable throughput [89] makes BLI suitable for comparing drug titers of individual production clones to select the ones affording highest yields. Since the response is proportional to the number of molecules bound to the biosensor surface, quantification of recombinant proteins using a standard curve method is possible [90]. Commercial capture sensors (e.g., functionalized with immunoglobulin-binding proteins A, G or L, and anti-human or anti-murine IgG antibodies) are available for antibody or Fc-fusion protein binding, and a number of chemistries can be used for custom immobilization of binding partners of non-immunoglobulin proteins (Table 1). The same principles can be applied to monitoring production yield during fermentation in a bioprocess optimization, and to characterize matrices with respect to dynamic binding capacity of affinity chromatographic columns in downstream process development [89]. Looking beyond simple quantification, BLI has been applied to assessing product quality with respect to post-translational modifications (also see section 3.3). Wallner et al. [91] have used a galactose-specific lectin (Ricinus communis agglutinin 120) as a ligand attached to the BLI sensor surface to determine galactosylation levels of antibodies produced in different mammalian cell culture processes. The extent of galactosylation as determined by the BLI platform correlated well with data from LC-MS analyses, enabling the use of lectin-functionalized biosensor for the study of how diverse bioprocess parameters affect antibody quality. It was shown that galactosylation levels were higher when the amount of ammonium per antibody produced was lower, indicating that BLI can be efficiently used as a tool for process development as well as process monitoring.

Monoclonal antibodies recognizing similar epitopes often share the same function (e.g., capability of neutralizing soluble antigens, or receptor antagonistic activity), so high-throughput methods of classifying antibodies from large libraries with regard to the same or adjacent antigenic determinants (so-called ‘epitope bins’) are warranted as they can disclose numerous leads early in the process of mAb drug discovery. BLI-based competitive assays represent a powerful tool for epitope binning of mAbs [[93], [94], [95]]. One of the antibodies is immobilized on a sensor surface and complexed with the cognate antigen, and then the biosensor is immersed in solution of a different antibody. Pairs of mAbs can be systematically analysed in parallel for antigen binding competition, signifying recognition of the same epitope or steric hindrance in case of overlapping/adjacent epitopes.

The same approach is useful for monitoring the breadth of humoral immune responses in animal immunization experiments for mAb development using hybridoma technology [96] and for vaccine immunogen design [94]. Binning polyclonal antibodies (pAbs) with regard to antigenic determinants can be especially challenging, and requires previously well-characterized antibodies with known epitopes as controls. In addition, pAb pools can be very heterogeneous in binding avidity, which led Li et al. [97] to develop the so-called ‘polyclonal antibodies avidity resolution tool’ or PAART for monitoring pAb:antigen interactions using BLI to measure dissociation rate constants for defining avidity. PAART is based on a sum-of-exponentials mathematical model to fit dissociation time-courses of the interactions and resolve multiple k_off constants defining the overall dissociation, wherein individual k_off values correspond to a group of antibodies with similar avidity. As a proof of principle, the authors used PAART to interrogate pAbs heterogeneity with respect to avidity for post-vaccination serum IgG antibodies from phase 2 clinical studies of vaccines, and for pAbs from HIV-1-positive individuals that naturally control viral loads. Typically, pAbs could be classified in two or three groups based on avidity differences. The method enabled tracking of the fine-resolved affinity maturation of humoral response against different malaria vaccine antigens over time.

A more detailed knowledge of antigenic determinants recognized by mAbs is often desirable, however, exact epitope mapping is all but trivial, and methods such as X-ray crystallography and hydrogen/deuterium exchange MS are laborious and time-consuming [98]. Therefore, Guo et al. [99] have developed a relatively simple but high-resolution epitope mapping approach consisting of site-directed mutagenesis to alter the surface-exposed residues on influenza A virus major antigens hemagglutinin and neuraminidase, followed by analysis of antibody:mutein interactions using BLI. Here, the His-tagged muteins were immobilized on anti-His biosensors, which were then dipped in solutions of mAbs to monitor both affinity and kinetics of interaction. In this setting, purification of neither the recombinant antigens nor the antibodies were required, i.e., screening was compatible with crude cell culture supernatants.

Biopharmaceuticals can display considerable chemical and physical instability [100] and recombinant protein drugs are subjected to a plethora of host cell- and production process-dependent post-translational modifications that might affect their potency and/or biological half-life [101,102]. In this context, BLI technology has been used for in vitro prediction of pharmacodynamic and pharmacokinetic properties, as well as for simple stability-indicating assays.

Monoclonal antibodies are glycoproteins with complex glycans attached to the Asn 297 residue located on the CH2 domains of the Fc region [103]. Glycosylation is considered a critical quality attribute of therapeutic antibodies since it impacts stability and conformation of the CH2 domains, in turn affecting binding to FcγR receptors. Thus, antibody effector functions conveyed via these receptors on immune cells are indirectly affected by the N-glycan composition. In addition, antibodies of the IgG class bind to the neonatal Fc receptor (FcRn, also known as the Brambell receptor) in the acidified endosomes in endothelial cells and monocytes/macrophages following pino- or phagocytosis. This facilitates the rescue of IgG from lysosomal degradation and their recycling back to the extracellular space, in turn extending the biological half-life of IgG. The interaction of therapeutic antibodies and Fc-fusion proteins with FcRn and FcγR receptors can be strongly influenced by post-translational modifications and chemical degradation of the Fc region. For example, afucosylation augments antibody-dependent cellular cytotoxicity (ADCC) [104] mediated via FcγRIIIa expressed on natural killer cells, and galactosylation levels correlate positively with antibody-dependent cellular phagocytosis (ADCP) potency [105] conveyed via FcγRIIa on phagocytic cells. Furthermore, chemical degradations, such as Met 252 oxidation, were reported to negatively affect FcRn [106] and FcγRIIa binding [107], decreasing plasma half-life and ADCP activity, respectively. Cell-based assays can be replaced by more convenient in vitro assays to monitor Fc-receptor binding for characterization and release of antibody drug products. Yet, the presence of antibody aggregates in samples may increase Fc receptor binding on account of avidity effects, and mask the potential decrease of antibody monomer affinity due to suboptimal post-translational modifications and/or degradations. To overcome the limitations of the bead-based bridging assay AlphaScreen, Bajardi-Taccioli et al. [108,109] have developed BLI-based assays for monitoring mAb binding to FcRn and FcγRIIa. The authors ranked samples according to affinity based on k_off values and binding responses at the end of association step relative to the reference standard, respectively, rather than calculating K_d values. They have shown that the assays were fairly resilient to aggregate interference, while being sensitive to glycosylation patterns and structural changes in the Fc region due to oxidation stress. Importantly, in the case of FcγRIIa binding, the BLI results for antibody samples containing up to 5 % aggregates (a content much higher than typically deemed acceptable for therapeutic antibodies) correlated well with the data from cell-based ADCP assay indicative of antibody potency. In contrast, the AlphaScreen assay suggested artificially strong binding for the same set of samples (with the 5 % aggregate sample displaying potency of ∼200 % relative to that of the reference standard). Similarly, FcRn binding was not significantly affected when aggregates were present in up to 5 %. Furthermore, Neuber et al. [110] have previously demonstrated that BLI technology can discriminate between monomeric antibodies with minute differences in FcRn binding affinities due to oxidation, and Souders et al. [111] were able to predict the experimentally verified half-life in vivo (from transgenic mice expressing human FcRn and from phase 1 human clinical study) using an in vitro model established on FcRn-based BLI biosensor data. In all studies described herein, biotinylated Fc receptors were coupled on streptavidin-coated biosensors, as the indirect immobilization method allows for better control of ligand density, which is an important parameter in minimizing avidity effects.

Another mode of action of mAbs is the complement-dependent cytotoxicity (CDC), where antibody clustering upon binding to cell surface antigens triggers C1 complex assembly, eventually leading to formation of the cytolytic effector known as the membrane attack complex. The first step of the classical complement pathway entails C1q component binding to multiple antibody Fc regions. To study antibody-C1q interactions, Zhou et al. [112] have therefore immobilized (biotinylated) IgGs of different subclasses on either protein L- or streptavidin-coupled biosensors, creating a densely packed ligand surface, and immersed them into C1q solution. As expected, IgG1 therapeutic antibodies bound C1q with K_d in the low nanomolar range, whereas IgG2 and IgG4 only displayed low micromolar or no affinity for C1q, respectively. Interestingly, however, when the two IgG1 antibodies were complexed with the cognate antigens, adalimumab:TNF-α retained affinity to C1q, whereas trastuzumab:HER2 no longer bound C1q. This is consistent with the observed lack of trastuzumab's CDC activity [113], and strong adalimumab's CDC activity [114] despite the two antibodies having the same Fc sequence, indicating that carefully planned BLI assays might be a valid alternative also to cell-based CDC activity assays. Indeed, the US Pharmacopoeia National Formulary now cites BLI in its chapter ‘<1108> Assays to Evaluate Fragment Crystallizable (Fc)-Mediated Effector Function’.

Finally, protein therapeutics are inherently immunogenic, and induction of anti-drug antibodies (ADAs) in patients can have various consequences, from (slight) increase of clearance to drug neutralization, in turn potentially affecting pharmacokinetics, efficacy and/or safety [115]. Monitoring ADAs during (and sometimes even beyond) clinical trials is therefore required [116]. However, lack of standardized ADA quantification methods and especially human ADA reference standards makes ADA characterization difficult. Panels of low- and high-affinity anti-drug monoclonal antibodies recognizing diverse epitopes are being developed to validate ADA immunoassays [[117], [118], [119]]. For example, Shibata et al. [117] have compared the analytical performances of three immunoassays (SPR, BLI and bridging electrochemiluminescence (ECL) assay) for detection of anti-epoetin antibodies using a panel of 9 monoclonal antibodies (developed by the World Health Organization) spiked into pooled human serum. They reported considerable differences in sensitivity for the three methods; dose-dependent increase in binding was observed for all 9 antibodies using SPR and BLI, whereas the ECL failed to detect low-affinity IgG antibodies with high k_off values. Presumably, the binding of such antibodies cannot be maintained in the washing steps of the ECL assay, while in SPR and BLI there are no washing steps. Similar results were reported by Tada et al. [119] for their in-house developed panel of human-rat chimeric anti-rituximab IgG1 mAbs. Comparing detectabilities determined based on the assay cut point (calculated from the response measurements by individual serum samples from healthy human subjects), the ECL assay had highest sensitivity for both low-/mid-affinity but high-avidity IgM and high-affinity IgG anti-epoetin antibodies [117]. The BLI assay was more sensitive than SPR for detecting low-affinity ADAs, while for sensing high-affinity ADAs the assays displayed similar sensitivities.