Deciphering the Interaction between Neonatal Fc Receptor and Antibodies Using a Homogeneous Bioluminescent Immunoassay

Therapeutic Abs and Fc fusion proteins are effective against a variety of diseases because of their exquisite specificity, ability to activate an immune response through effector functions, and their long serum half-life (∼20 d). The long half-life of IgG and Fc fusion proteins is attributed to the interaction of the Fc domain with neonatal Fc receptor (FcRn) (1–3). FcRn is an intracellular protein located in acidic (pH 6) endosomal vesicles of a wide variety of cells, including endothelial cells, hepatocytes, monocytes, dendritic cells, and macrophages, among others. Abs, albumin, and other proteins in circulation are internalized through pinocytosis by endothelial cells lining the blood vessels and transported to the acidic (pH 6) endosomal compartment. In endosomes, Abs and albumin bind to the FcRn and are recycled back to the cell membrane and dissociate at neutral pH and are released back into circulation. Proteins that do not bind to the FcRn are trafficked to lysosomes for degradation.

Because of the critical role of IgG/FcRn interaction (IgG in this context includes Abs and Fc fusion proteins) for Ab recycling, this interaction is routinely targeted to achieve desired therapeutic goals. For example, modifications in amino acid sequences in the Fc domain have been shown to significantly increase the IgG/human FcRn (hFcRn) binding affinity, which correlated with improved serum half-life of Abs (4, 5). Further evidence of leveraging IgG/FcRn interaction for therapeutic purposes is in the treatment of autoimmune diseases like systemic lupus erythematosus using IVIG (6, 7). Injected IVIG at high concentrations (500–800 mg/kg body weight) overwhelms the hFcRn and forces the endogenous autoreactive IgG to lysosomes for degradation. Alternate strategies like Abs that enhance IgG degradation (ABDEGs) with engineered Fc that bind strongly to hFcRn or anti-FcRn Abs are being investigated as alternates for IVIGs (8–11). Finally, IgG/FcRn affinity changes depending upon the animal species and will impact the preclinical toxicity and pharmacokinetics studies of Ab therapeutics in which animal models (e.g., mouse, rat, and cynomolgus) are routinely used as surrogate for humans.

Although there is a clear consensus about the importance of the IgG/FcRn interaction, surprisingly, the biosensor platforms used to measure these interactions have been shown to introduce artifacts if experiments are not properly designed. Recent investigations of these platforms (12–14) have identified assay format, surface chemistry of the sensor chip, and immobilization method as key factors contributing to contradictory data. Two different formats are typically used for binding assay: first, in which FcRn is immobilized on the sensor surface, and IgG in the solution is injected over the sensor; or second, in which IgG is immobilized, and FcRn in solution is injected over the chip. In a recent study, two recombinant human Abs with 99% sequence homology were tested using these two formats on a surface plasmon resonance (SPR) platform. The affinity value of one Ab was independent of the assay format, but the second Ab gave two different affinity values, depending on which component was immobilized on the surface. Upon further investigation, it was found that a positive patch in the Fab region of the second Ab was involved in nonspecific interaction with the charged surface of the chip used in the assay (14), and the authors recommended use of multiple assay formats, which can be onerous. A separate study identified additional factors impacting affinity values including the density of FcRn or Ab immobilized on the surface, method of immobilization, and architecture of the chip surface (planar or hydrogel) (12). These observations are not unique to IgG/FcRn interactions and have been noted for other biomolecular interaction studies performed on biosensor platforms (15). These same studies also offer guidelines to minimize artifacts and obtain accurate results, but the time and resources required to pursue all these options are significant, and such guidelines may not be generalizable. The later aspect is evident in the different recommendations made by these studies, which were specific to the model Ab tested in the study. ELISA and AlphaLISA are alternatives to biosensor platforms, but these methods also involve solid surfaces and may suffer from similar limitation, although no detailed study has been performed yet.

Surprisingly, not much attention has been given to orthogonal biochemical assays that measure binding in solution (homogeneous assays) and may alleviate surface- and immobilization-related problems. Occasionally, a reformatted SPR protocol to measure solution phase affinity values and isothermal calorimetry have been used (12, 16). Other solution-based methods frequently employed in small molecule drug discovery research are rarely used for IgG/FcRn interactions. Some of these methods include fluorescence polarization, time-resolved fluorescence resonance energy transfer, and homogeneous time-resolved fluorescence (17). We can only speculate, but possibly, these methods may lack sensitivity to reliably detect relatively weak IgG/FcRn affinity constants (100–1000 nM). Given the complexity of optimizing assays on biosensor platforms and limited availability of solution-based methods, we developed a new (to our knowledge) homogeneous immunoassay that integrates elements of protein-fragment complementation assay (PCA) and competitive binding to measure IgG/FcRn affinity.

PCAs are routinely used to monitor protein–protein interactions within intact cells and in vivo (18–22). In PCAs, an enzyme (e.g., β-galactosidase or firefly luciferase) or a fluorescent protein is split into two or three fragments with low affinity for each other. When these fragments are recombinantly fused to two interacting proteins, then, upon interaction, the fragments come into close proximity and reassemble into the functional enzyme or fluorescent protein and “report” on the interaction. Recently, a new reporter called NanoBiT was added to the list of enzymes used in complementation assays (23). NanoBiT is based on NanoLuc luminescent enzyme, which is ∼100-fold brighter than the Firefly and Renilla luciferases. NanoBiT is composed of a small 11-aa peptide called small BiT (SmBiT) and a large 18-kDa polypeptide called large BiT (LgBiT) fragment that have low affinity (K_D > 100 μM) for each other. However, when brought into close proximity, they reconstitute to form an extremely bright bioluminescent enzyme. The small size of NanoBiT fragments minimizes any steric impact on interacting proteins, and because of its bright bioluminescent signal, NanoBiT can detect protein interactions at low endogenous expression levels (23–25). More recently, NanoBiT technology has been used to develop immunoassays for detection of cell signaling pathways (26), SARS-CoV-2 Abs in patient serum and plasma (27, 28), and small mycotoxin molecules in food (29). NanoBiT immunoassays were enabled by chemical labeling or making genetic fusion of detection reagents (Abs, proteins, and small molecules) and performing the assays in sandwich or competitive assay format. These biochemical immunoassays have several advantages, including simple, no-wash format, high sensitivity and specificity, require low sample volume, and can be run in a high-throughput format.

In the current study, we used the NanoBiT to develop a competition immunoassay for rapid and reliable measurement of IgG/FcRn interaction (Fig. 1). NanoBiT FcRn assays were developed for hFcRn, cynomolgus FcRn (cFcRn), mouse FcRn (mFcRn), and rat FcRn (rFcRn). Optimized assays were used to measure IgG/FcRn interaction with a panel of Abs that include seven different therapeutic Abs and one National Institute of Standards and Technology (NIST) mAb reference standard. As we demonstrate, an orthogonal method such as NanoBiT FcRn assay can be useful in validating data generated on other platforms. Alternatively, because of the simple homogeneous format, this assay can be used for screening large libraries of Abs, followed by validation of Abs with desirable properties on alternative platforms.

Recombinant hFcRn, cFcRn, mFcRn, and rFcRn with C terminus biotin were from ACROBiosystems (Newark, NJ). Human IgG1 used for labeling with LgBiT was produced by GlycoScientific (Atlanta, GA) and has a sequence similar to that of adalimumab. Streptavidin was from Sigma-Aldrich. All the therapeutic Abs were purchased from distributors. NIST mAb was purchased from NIST (Bethesda, MD). IgG1 with the same sequence as rituximab but with M252Y/S254T/T256E (YTE) mutation in Fc region was made by Absolute Antibody (Boston, MA). IdeS protease, magnetic His tag beads, and magnetic protein A beads were from Promega. To create NanoBiT-labeled proteins, the SmBiT and LgBiT recombinant proteins (available from Promega) were expressed in Escherichia coli as genetic fusions with HaloTag by combining existing NanoBiT and HaloTag sequences. Labeling was performed using Lumit Immunoassay labeling kit (Promega) and include HaloTag-LgBiT and HaloTag-SmBiT. Buffers were prepared in-house, and superblock was purchased from ScyTek (Logan, UT).

The labeling of IgG and streptavidin with LgBiT and SmBiT, respectively, was done using a modified methods reported earlier (30). Briefly, solutions of human IgG1 and streptavidin at 1.0 mg/ml were pH adjusted (pH 8.3) with 1/10th volume of 1.0 M bicarbonate buffer and reacted with 20-fold molar excess of amine reactive HaloTag ligand [HaloTag Succinimidyl Ester (O4) Ligand; Promega] for 1 h. Unreacted HaloTag ligand was removed with Zeba desalting column (Thermo Fisher Scientific). Two columns were used in sequence to ensure complete removal of unreacted HaloTag ligand. Concentrations of HaloTag ligand–labeled IgG1 and streptavidin were calculated by measuring absorbance at 280 nm. A280 of 1.4 and 3.2 corresponds to 1.0 mg/ml of IgG and streptavidin, respectively.

IgG1 and streptavidin activated with HaloTag ligand were incubated overnight with 4-fold molar excess of HaloTag-LgBiT or HaloTag-SmBiT fusion protein, respectively. This step allows the HaloTag fusion to form covalent attachment with the HaloTag ligand on the proteins. Any unreacted HaloTag-LgBiT or HaloTag-SmBiT were removed by incubating the samples with HaloLink beads (Promega). HaloLink beads are nonmagnetic agarose beads activated with HaloTag ligand and will bind to any unreacted HaloTag-LgBiT or HaloTag-SmBiT in the solution. Nonreducing SDS-PAGE gel was used to confirm the labeling of streptavidin and IgG.

NanoBiT FcRn assay involves four steps. In the first step, 25 µl (or 10 µl) of IgG-LgBiT solution was mixed with 25 µl (or 10 µl) of the IgG or Fc fusion proteins in a white, 96-well, nonbinding plate (Corning). In the second step, FcRn–biotin tag and streptavidin–SmBiT were mixed at 1:1 molar ratio to form FcRn–biotin–streptavidin–SmBiT (FcRn-SmBiT). In the third step, 50 µl (or 20 µl) of FcRn-SmBiT is added to each well containing IgG and IgG-LgBiT and plate incubated for 30–60 min. In the final step, furimazine substrate (Promega) is diluted 1:50-fold, and 25 µl (or 10 µl) is added to plate, and signal is allowed to stabilize for 3 min, and the bioluminescence signal (relative light units [RLU]) is measured in a GloMax Discover Luminometer (Promega). All the dilutions were made in PBS containing 10% superblock (pH 6). Before running the assay, pH of the Abs was adjusted to pH 6 by addition of 1/10th the volume of 0.5 M citrate buffer containing 50% superblock (pH 5.7). All the experiments were run in duplicate or triplicate. Normalized RLU data are generated by assigning 100% to the maximum bioluminescent signal obtained in absence of the IgG and then calculating percentage drop in signal in the presence of an IgG. Inhibition curves were generated by plotting normalized RLU as a function of log value of the IgG concentration (in nM) and fitted to a four-parametric equation with 1/y² weightage using GraphPad Prism. IC₅₀ values that are equivalent to apparent affinity values (K_D) are used when comparing data generated using biosensor and reported in the literature.

Abs (infliximab, adalimumab, and etanercept) were cleaved into Fc and F(ab)₂ fragments by incubating with IdeS protease (1.0-unit enzyme per 1.0 µg of IgG) for 1.5 h at 37°C. IdeS has a His tag; therefore, after digestion, enzyme was removed by using magnetic His beads following vendor provided protocol. An aliquot of the cleaved mixture was saved, and magnetic protein A beads was added to the rest to bind the Fc domain (vendor protocol was followed), and flowthrough containing F(ab)₂ fragment was collected and used for studying interaction with various FcRn molecules. Magnetic protein A beads containing bound Fc were washed, and bound Fc was eluted using glycine buffer (pH 2.7). The solution was neutralized with 2 M Tris buffer (pH 7.5). F(ab)₂ and Fc fragments were dialyzed into PBS buffer before running the assay as described above.

Ninety-microliter aliquots of each Ab (panitumumab, etanercept, and infliximab) were pipetted into six separate 1.5-ml microcentrifuge tubes. Hydrogen peroxide (H₂O₂) of 0, 1, 3, 5, 10, and 30% solutions were prepared from stock solution, and 10 µl of each solution was added to the Abs so that final H₂O₂ concentrations were 0, 0.1, 0.3, 0.5, 1.0, and 3.0%. Samples were incubated for 2 h at room temperature, followed by buffer exchange (PBS [pH 7.2]) using Zeba desalting column (7K) to remove H₂O₂. Ab concentrations were determined using NanoDrop. A280 of 1.4 corresponds to 1.0 mg/ml of Ab.

Binding of three anti-TNF Abs (etanercept, adalimumab, and infliximab) to biotinylated FcRn was measured using biolayer interferometry (BLI) on Octet Red system (Pall Bio). FcRn–biotin at 1.0 µg/ml in PBS (pH 6) was loaded onto the streptavidin-coated biosensor to a final surface density of ∼0.3 nm. Samples of three anti-TNF Abs were diluted to 0, 31.25, 62.5, 125, 250, 500, and 1000 nM with PBS containing 0.02% Tween 20 (pH 6 and pH 7.2). Samples were incubated with FcRn-coated biosensor for 60 s to measure the association rate, followed by 60-s wash in PBS containing 0.02% Tween 20 to measure dissociation. Biosensor were regenerated with PBS (pH 7.4) before the next cycle. All data were plotted using instrument software, and steady-state data were plotted on GraphPad Prism and fitted to a one-site binding curve to calculate equilibrium K_D.

Our goal was to be able to determine the K_D of IgG/FcRn interaction using the NanoBiT FcRn assay, and this can be achieved from IC₅₀ values under specific conditions as proposed by the Cheng–Prusoff equation (31).

L* is the concentration of tracer, K_D* is tracer-target dissociation constant, and K_D is the dissociation constant of the analyte and target. At tracer concentration much lower than the dissociation constant, IC₅₀ can be approximated to the apparent dissociation constant of the analyte and target. The competition assay as described (Fig. 1) is used routinely in receptor-ligand binding assays for high-throughput screening for small molecule drug discovery. Specific criteria have been proposed for setting up successful competition assays and include the following: 1) specific binding of tracer with low nonspecific binding that leads to a wide assay window; 2) tracer concentration lower than the expected equilibrium dissociation constant, which in the case of IgG/hFcRn, is in 100 nm range (see below); and 3) target concentration lower than tracer concentration to minimize tracer depletion. It is worth noting that several alternates to Cheng–Prusoff with claims of more accurate estimation of affinity have been proposed and include linearized Cheng–Prusoff equation (32), Schild method (33), and Gaddum Equation (34).

In the NanoBiT FcRn assay, the “target” is FcRn-SmBiT, the “tracer” is IgG-LgBiT, and analyte is IgG. Therefore, the first step in the assay development is to determine the concentrations of IgG-LgBiT and FcRn-SmBiT that meet the criteria described above and to ensure that interaction between the two is specific. Starting with hFcRn, a positive interaction between IgG-LgBiT and hFcRn-SmBiT (2.5 nM) at pH 6 was confirmed by concentration-dependent increase in bioluminescent signal resulting from LgBiT and SmBiT complementation (Fig. 2). The interaction is pH dependent, as minimal interaction is seen at pH 7.2. To verify that bioluminescent signal is due to a specific binding interaction between the IgG-LgBiT and FcRn-SmBiT and not from the random proximity events, a competing polyclonal human IgG was added at 25-fold molar excess to the IgG-LgBiT. The excess IgG outcompetes the IgG-LgBiT for binding to the target, resulting in a drop in bioluminescent signal indicative of specific binding of IgG-LgBiT and FcRn-SmBiT. When excess human serum albumin (HSA), which is known to bind to FcRn but at a distal site and does not compete with IgG, was added instead of IgG, no drop in signal was seen (data not shown). Finally, no bioluminescent signal is observed when no FcRn is added again, confirming that specific IgG/FcRn interaction is needed for the assay to work. Similar assays were performed for rFcRn, mFcRn, and cFcRn (data not shown).

A combination of 2.5 nM hFcRn-SmBiT and 20 nM IgG-LgBiT was selected for subsequent competition assays as it satisfied the assay conditions required by Cheng–Prusoff equation. Moreover, we obtained a good assay window (∼20-fold) and sufficiently high absolute RLU such that the assay can be performed on a wide variety of luminometers. For NanoBiT FcRn assays involving rFcRn, mFcRn, and cFcRn, same concentrations were used to keep the experimental setup simple. As shown below, even though rFcRn and mFcRn have higher affinity for human IgG, the conditions set by Cheng–Prusoff equation were met.

Optimized NanoBiT FcRn assays were used to measure affinities of a panel of seven therapeutic Abs and one NIST mAb standard. The panel of selected Abs is diverse, with three different isotypes (IgG1, IgG2, and IgG4), three Ab types (chimeric, humanized, and human), four target types (TNF-α, CD20, PD1, and EGFR), and one Fc fusion protein (Fig. 3). As expected in a competition assay, a concentration-dependent decrease in bioluminescent signal was observed with all IgG/FcRn interactions and allowed for the determination of the IC₅₀ values for 32 individual IgG/FcRn interactions (Fig. 3).

Several interesting observations can be made from the data. First, the IC₅₀ values of eight Abs for individual FcRn varied within a relatively narrow range. The average ± SD for eight Abs for hFcRn, cFcRn, mFcRn, and rFcRn were 199.0 ± 86.4, 210.9 ± 95.7, 36.9 ± 14.6, and 28.3 ± 7.7, respectively. Second, the IC₅₀ values are similar for two primate FcRn proteins, and same is true for two rodent FcRn proteins. Third, binding of the Abs to rodent FcRn is around 5-fold stronger than binding to the primate FcRn. Finally, it is interesting to note that three anti-TNF Abs (etanercept, adalimumab, and infliximab) had similar IC₅₀ values.

To ensure that decrease in bioluminescence signal seen with the Abs is due to true competition between Abs and the IgG-LgBiT, we tested HSA as a negative control. Albumin is known to bind to FcRn but at a site distal from the Fc binding site, and binding is not competitive with Ab binding. When HSA was used in the assay, no signal inhibition was seen, indicating that the binding of IgG to FcRn is specific (Fig. 4). To further validate the assay and ensure that NanoBiT FcRn is indeed capturing expected binding, we performed binding of mouse IgG1 and mouse IgG2b to all four FcRn (Fig. 4). mFcRn is known to be promiscuous and binds both human and mouse IgG equally well, whereas hFcRn is more specific and is known to bind only weekly to mouse Abs, as also seen in our data.

To compare the data from the NanoBiT IgG/FcRn assay with the traditional biosensor platform, we used BLI platform and measured the affinities of three anti-TNF Abs. For the BLI assay, we used a streptavidin sensor and immobilized FcRn–biotin, which is followed by incubation with Abs in solution. This format was selected because it has been identified as an acceptable format to get an artifact-free affinity measurement (12). Affinities for three anti-TNF Abs were similar, as was observed with NanoBiT FcRn; however, the absolute numbers were lower than that obtained using NanoBiT method (Fig. 5). Differences in absolute numbers is not surprising, as it has been noted before, but it was encouraging to see that, qualitatively, results matched on these two platforms.

Because of the importance of IgG/FcRn binding, it is important that the NanoBiT FcRn assay should be able to detect any changes in the binding affinities. It is known that oxidation of two methionine residues (Met²⁵² and Met⁴²⁸) in the Fc region of Abs decrease the IgG/FcRn affinity, and therefore, we used this as model system with NanoBiT FcRn assay. Three different Abs, panitumumab, etanercept, and infliximab, were oxidized with varying concentrations of hydrogen peroxide, and IgG/FcRn affinities were measured. Dose-response curves for panitumumab on hFcRn and mFcRn are shown in (Fig. 6, and as expected, a rightward shift in IC₅₀ values was seen, indicating decrease in the affinity for IgG/FcRn binding. Similar data were generated with all three Abs interacting with four different FcRn (n = 72 interactions curves). Fold changes in IC₅₀ values compared with untreated Abs are shown in (Fig. 6, and a steady increase in IC₅₀ values was observed with ratio of IC₅₀ values between Abs treated with 3% H₂O₂ and untreated Abs in the range of ∼5–10-fold for primate FcRn and 10–16-fold for rodent FcRn.

Although NanoBiT FcRn can capture the decrease in affinity of IgG/FcRn binding, we asked if it can also detect increase in the IgG/FcRn binding affinity. A well-documented method for improving the IgG/FcRn affinity is to introduce the mutation YTE in the Fc fragment of the IgG, which has been shown to improve the IgG/FcRn affinity by 11-fold (4, 5). An Ab with sequence of rituximab, except for YTE mutation in Fc region, was designed, expressed in HEK293 cells, and purified using affinity chromatography (work was done by Absolute Antibody). YTE-modified Ab was tested using four FcRn (Fig. 7), and as expected, affinities for YTE variant were around 9–18-fold higher than the rituximab on four FcRn molecules.

An ongoing discussion in the IgG/FcRn interaction is the role played by the F(ab)₂ domain, with some studies finding no influence (12, 35, 36) and other finding significant impact of the domain (13, 37–39) on IgG/FcRn interaction. SPR has been predominantly used for such studies, and we reasoned that an orthogonal platform like NanoBiT FcRn assay may provide new insights into this interaction. We focused on three anti-TNF Abs because they share the same Fc fragment sequence but differ in their Ag binding domain. For adalimumab and infliximab, Ag binding domain is traditional F(ab)₂, whereas in the case of etanercept, Ag binding domain is TNFR. For simplicity, Ag binding domain of all three anti-TNF molecules are noted as F(ab)₂. To clearly delineate the roles of different fragments in the FcRn binding, we decided to measure and compare the affinities of F(ab)₂ domain, Fc domain, a mixture of the two fragments, and the full-length IgG. All three molecules were digested with IdeS protease, which cleaves Abs at a site below the hinge region and yields a mixture of Fc domain and F(ab)₂ domain as shown in the gel image (Fig. 8).

Full-length IgG, mixture of Fc and F(ab)₂ domain, purified Fc, and purified F(ab)₂ domains were analyzed using the NanoBiT FcRn assay and shown for hFcRn and mFcRn (Fig. 9). Similar data were generated for cFcRn and rFcRn (data not shown). Data for two primate FcRn proteins were similar, so were the data for rodent FcRn proteins. Affinities for full-length molecule mixture of Fc and F(ab)₂ were similar to purified Fc fragment, which implies that F(ab)₂ domain does not play a major role in FcRn interaction. Interestingly, however, when only F(ab)₂ domain was used in the assay, we did observe an interaction between F(ab)₂ domain and FcRn. For adalimumab and infliximab, inhibition with F(ab)₂ alone was <50%, whereas etanercept showed somewhat stronger binding with >50% inhibition on both hFcRn and mFcRn.

This study describes a homogeneous bioluminescent immunoassay for measuring IgG/FcRn binding. The assay is based on protein complementation technology and designed using the extremely bright NanoBiT bioluminescent reporter. Compared with PCAs with fluorescent or absorbance signal, bioluminescent detection has negligible background, which translates into a high signal-over-background ratios, superior sensitivity, and wide dynamic range. More importantly, the assay is performed in solution phase and minimizes surface- and immobilization-related artifacts introduced in traditional biosensor affinity measurement methods. In addition, homogeneous assay formats are attractive because of their simple workflow, reduced hands-on time, scalability, and high-throughput screening capabilities.

While developing the NanoBiT FcRn assay, we intentionally decided to take an approach that can be easily reproduced. First, we selected a chemical method to label IgG and streptavidin with LgBiT and SmBiT, respectively. This approach is similar to the widely used method of labeling proteins and Abs with fluorophore, biotin, and enzymes and is therefore easy to implement. A disadvantage of chemical labeling method is its stochastic nature, in which the number and position of label cannot be controlled. Therefore, we made a robust protocol and generated multiple batches of IgG-LgBiT as well as FcRn-SmBiT and obtained very reproducible results (data not shown). Our observations are not surprising considering that three Food and Drug Administration–approved Ab drug conjugates are prepared using chemical labeling of drugs to Abs (40), and the therapeutic efficacy of multiple batches are reproducibly maintained by strict process controls. Second, our modular choice of making the FcRn-SmBiT by mixing site specific biotinylated FcRn and streptavidin–SmBiT. This approach allowed us to rapidly create four different assays simply by switching to FcRn from different animals. Biotinylated proteins are widely available commercially, and therefore, this approach will allow additional assays to be developed easily including for other Fc receptors. It is worth noting that FcRn is a heterodimeric membrane protein, but we used a soluble extra cellular domain of the protein expressed in mammalian cell expression system. Truncated soluble fragment of the FcRn is routinely used in traditional biochemical assays (e.g., SPR and BLI) to measure binding affinities, and such measurements have been shown to correlate well with Ab half-life (4, 12).

Initial evaluations of the NanoBiT FcRn assay were done with a panel of seven therapeutic Abs and one NIST mAb. The panel covers a wide variety of Abs and includes human Abs of IgG, IgG2, and IgG4 isotypes, a mix of chimeric, humanized, and human IgG and targeted to different Ags. A common theme among all these Abs is their similar CH2–CH3 domain sequence, and specifically, the amino acid sequences known to interact with FcRn are identical (Fig. 10). In addition, except for etanercept, they all share high amino acid sequence homology in the CH1 domain. Etanercept has TNFR instead of traditional F(ab)₂ domain.

All eight Abs were evaluated for binding to four FcRn molecules. Several interesting observations can be made from this extensive dataset of 32 individual interactions (Fig. 3). The first observation is that two rodent FcRn proteins have very similar affinities for the Abs as do the two primate FcRn proteins. This is not surprising considering that amino acid sequences of rFcRn and mFcRn show 89 and 83% homology in the FcRn H chain and β₂-microglobulin, respectively (41). Similarly, hFcRn and cFcRn sequences are very similar, and among the few sequence differences between the two species, none are known to be involved in IgG/FcRn interaction (42). The second observation that rodent FcRn binds human Abs with higher affinity (∼5-fold) compared with primate FcRn has been noted before (12, 43) and gave us the confidence that NanoBiT FcRn assay is able to capture the predicted binding affinities. Site-directed mutagenesis and crystallographic studies have shown that Ile²⁵³, Arg²⁵⁵, His³¹⁰, and His⁴³⁵ on human Fc play a critical role in the interaction with FcRn. These residues interact with anionic residues on FcRn and stabilize the interaction by making salt bridges. Residue on rodent FcRn involved in the interactions are Glu¹¹⁷, Glu¹¹⁸, Glu¹³², Trp¹³³, and Asp¹³⁷ (on rat), or Glu¹³⁷ (on mouse). The respective residues on primate FcRn are Glu¹¹⁷, Glu¹¹⁸, Asp¹³², Trp¹³³, and Leu¹³⁷. The presence of Asp¹³² and Leu¹³⁷ in primate FcRn are responsible for weaker binding of human IgG to primate FcRn compared with rodent FcRn. These 2-aa differences also make rodent FcRn very promiscuous in its binding to IgG from various species, whereas primate FcRn is much more stringent in cross-species binding and bind mouse and rat IgG very weakly (Fig. 3). Strong binding of rodent FcRn with human IgG makes a mouse model unsuitable for preclinical pharmacokinetics experiments and transgenic mouse with hFcRn are preferred for such studies.

The third point to note is that eight different Abs have IC₅₀ values within a relatively narrow range for any one FcRn. Given that FcRn predominantly interacts with Fc domain, and the amino acid sequences in that domain are fairly conserved for all eight Abs, our observations are not surprising. Furthermore, the fact that eight Abs have widely different F(ab)₂ domains, but similar IgG/FcRn affinities, argues against the involvement of F(ab)₂ domains in IgG/FcRn interaction. Our result differs from some recent reports that F(ab)₂ domain plays a key role in IgG/FcRn binding (13, 37–39) but agrees with other studies (12, 35, 36) that found no evidence of F(ab)₂ domain interacting with FcRn. It is worth noting that all these experiments were performed on SPR, and the contradictory conclusions may be the result of surface-related artifacts. Our result also contradicts a recent report in which 7-fold differences in affinities of IgG/FcRn were observed in a panel of 53 Abs when measured using BLI (44). This discordant observation may be due to smaller number of Abs in our case, but there is also potential of surface-induced artifact in BLI platform, which were not explored in detail.

A fourth interesting observation is the similar IgG/FcRn affinity values for etanercept, infliximab, and adalimumab, even though etanercept has a much lower half-life of ∼4 d compared with adalimumab (10–20 d) and infliximab (8–10 d). In this context, our observations are in agreement with the study by Foss et al. (36), which reported similar affinities of etanercept (780 nM) and adalimumab (940 nM) measured by SPR and concluded no role of Fab arms on FcRn binding. They further confirmed their results by cell-based transcytosis assay. Indeed, these results are expected because the Fc domain of the three Abs have identical amino acid sequence. Our, as well as Foss et al. (36), observations are, however, in contrast with other reports in which a 10-fold–weaker binding of etanercept to FcRn was observed compared with the binding of adalimumab and infliximab to hFcRn (39, 45). In the first study by Suzuki et al. (39), affinities of adalimumab, infliximab, and etanercept were 672, 727, and 3612 nM, respectively, whereas Porter et al. (45) reported affinities of 225, 132, and 1500 nM, respectively. Differences in the relative affinities between three anti-TNF molecules as well as differences in absolute affinity values may be due to different assay format used for SPR experiments. Interestingly, when Suzuki et al. (39) digested the Abs with papain and measured the affinities of only the Fc fragments, they found affinities to be comparable for all three anti-TNF Abs. Based on their observations, authors attributed the 10-fold–weaker binding of etanercept, which is a fusion of Fc domain with TNFR, to the TNFR interfering with the binding of the Fc domain with FcRn. Adalimumab and infliximab, which have traditional F(ab)₂ domains, were hypothesized to not impact the FcRn binding. No additional validations were done in the studies to confirm the results, and it is possible that some of the known artifacts introduced by biosensor platforms (12, 14, 15) may have played a role. In view of this conflicting data, we decided to measure the affinities of three anti-TNF Abs on a BLI platform using an assay format known to reduce surface-related artifacts (see Materials and Methods). We found affinities of three anti-TNF Abs to be similar and qualitatively match the data obtained using NanoBiT FcRn assay and that reported by Foss et al. (37). These results clearly demonstrate that Fc domain is the key driver for IgG/FcRn interaction and highlights the importance of using multiple method for measuring affinity values. Later in Discussion, we provide additional evidence of importance of Fc domain in this interaction.

Finally, a desirable goal for IgG/FcRn binding assays is to be able to accurately predict the Ab half-life in vivo, but as can be seen from (Fig. 3, that is not the case, and correlation between IC₅₀ values and the reported half-life is poor. This is not unexpected and has been observed when using other biochemical methods for measuring affinities, especially for Abs against different targets and with different mechanism of action (44). Ab half-life is a complex interplay of target-mediated and nontarget-mediated clearance mechanisms (1–3). Among the nontarget-mediated clearance mechanism, FcRn binding is one among the many factors that include immunogenicity, biophysical characteristics of the Abs, and off-target binding. Within FcRn binding, affinity of the IgG/FcRn interaction at pH 6 is certainly important, but association and dissociation kinetics at pH 6 and binding characteristic at neutral pH can also influence the Ab half-life. Hence, it is unlikely that any one method can accurately predict the Ab half-life in vivo. However, modulation of IgG/FcRn affinity by Fc engineering has been clearly shown to correlate strongly with the in vivo data and therefore easy to use, and reliable methods for measuring this interaction remain important. It is important to note that SPR/BLI can determine the association and dissociation kinetics both at pH 6 and pH 7.2 all of which are relevant for mechanistic understanding of IgG/FcRn interaction. Such measurements are not feasible with NanoBiT FcRn assays, which is a limitation of the method.

The importance of the Fc domain is evident when the changes made in that region, for example, by exposing Abs to oxidizing conditions or mutating the amino acid sequences, causes significant changes in IgG/FcRn affinity, which then impacts in vivo half-life. Fc domain contains two conserved methionine residues (Met²⁵² and Met⁴²⁸) that are not in direct contact with the FcRn but are structurally close to the FcRn binding site. Oxidation of these two methionine residues changes the conformation of the Fc domain and decreases the binding to the FcRn and are known to reduce the Ab half-life in vivo by up to 2.3-fold (46). In our study with panitumumab, etanercept, and infliximab, we saw oxidation significantly decreased the IC₅₀ values of IgG/FcRn binding with larger impact seen with rodent FcRn compared with primate FcRn. Among three different Abs with different F(ab)₂ domains, the change in IC₅₀ values were fairly consistent, indicating that confirmational changes and their impact on FcRn binding are localized to Fc domain. It is surprising that not more attention has been paid to modulating the IgG/FcRn interaction by inducing confirmation changes in Ab structure. IgG/FcRn affinities can also be increased by introducing mutations within Fc domain and once such mutation is YTE (4, 5). Introduction of this mutation in IgG with sequence similar to rituximab resulted in ∼13-fold increase in affinity for primate FcRn when measured using NanoBiT FcRn assay, which compares favorably to 11-fold enhancement reported in the literature (5). IgG/FcRn affinity improvement upon YTE modification has been shown to improve serum half-life by nearly 4-fold.

Even though the evidence so far strongly points to the Fc domain as the key driver of IgG/FcRn interaction, we further decided to individually compare the binding of full-length IgG, Fc domain, F(ab)₂ domain, and the mixture of Fc and F(ab)₂ domains. For this study, our choice of three anti- TNF gave us an opportunity to not only look at interactions of FcRn with two traditional F(ab)₂ domain with differences in CDR region but also with TNFR, which has a completely different sequence and structure. To our surprise, F(ab)₂ domain in the absence of Fc did show interaction with FcRn (Fig. 9), although it is much weaker than the full-length IgG or Fc alone. This is probably the first time that interaction of F(ab)₂ with FcRn is being reported, but the study does raise the possibility that F(ab)₂ may, in some subtle form, alter the interaction with FcRn and impact the half-life of the Ab.

Finally, we would like to address the question of widely different absolute IgG/FcRn affinity values reported in the literature. Differences in assay platform and format have been noted before as a source of discordant results, but we found that switching the source of FcRn protein also changes the absolute affinity values, although relative binding affinities remain unchanged. In the absence of a reference FcRn protein or a standard assay method, we suggest incorporating NIST mAb in IgG/FcRn binding studies as a reference standard to normalize for experimental variations. In fact, we use NIST mAb to monitor the reproducibility of our assay and over a roughly 2-y period with multiple batches of reagents IC₅₀ values changed <20% (IC₅₀ = 129.6 ± 25.8). We also tested different sources of human IgG1 for use as tracer and obtained similar IC₅₀ values, although absolute luminescence signal may change.

In conclusion, we developed a sensitive, robust, solution-based assay for measuring IgG/FcRn interactions using NanoBiT protein complementation technology. Method is well suited for high-throughput screening and rank ordering of Abs and, when used in combination with existing assay platforms, will allow accurate measurement of Fc/FcRn interaction and accelerate development of biologic drugs.

We thank Darrell R. McCaslin and Dan Stevens at Biophysics Instrumentation Facility, University of Wisconsin, Madison for help in running experiments on Octet.

All the authors are employees of Promega Corp., and the NanoBiT technology and products based on NanoBiT are being commercialized.