Abstract
It is well established that the neonatal Fc receptor (FcRn) plays a critical role in regulating IgG homeostasis in vivo. As such, modification of the interaction of IgG with FcRn has been the focus of protein-engineering strategies designed to generate therapeutic antibodies with improved pharmacokinetic properties. In the current work, we characterized differences in interaction of IgG between mouse and primate receptors using three humanized anti-tumor necrosis factor α antibodies with variant IgG1 Fc regions. The wild-type and variant IgG showed a differential combination of improved affinity, modified dissociation kinetics, and altered pH-dependent complex dissociation when evaluated on the primate and murine receptors. The observed in vitro binding differences within and between species allowed us to more completely relate these parameters to their influence on the in vivo pharmacokinetics in mice and cynomolgus monkeys. The variant antibodies have different pharmacokinetic behavior in cynomolgus monkeys and mice, which appears to be related to the unique binding characteristics observed with the murine receptor. However, we did not observe a direct relationship between increased binding affinity to the receptor and improved pharmacokinetic properties for these molecules in either species. This work provides further insights into how the FcRn/IgG interaction may be modulated to develop monoclonal antibodies with improved therapeutic properties.
The neonatal Fc receptor (FcRn) is a heterodimeric protein consisting of a soluble light chain [β2-microglobulin (β2m)] and a transmembrane anchored heavy chain (α-FcRn). FcRn functions as a “salvage receptor” regulating levels of circulating IgG in rodents (Ghetie et al., 1996; Junghans and Anderson, 1996). FcRn is expressed ubiquitously in endothelial cells of human tissues (Ghetie and Ward, 2000), and it has been speculated that FcRn also plays a key role in IgG homeostasis in humans (Ghetie et al., 1996; Kim et al., 1999).
Earlier studies show that the FcRn/IgG interaction is highly pH-dependent. IgG binds to FcRn via the Fc region at pH 6.0 (Rodewald, 1976; Raghavan and Bjorkman, 1996) and does not bind to FcRn at neutral pH, and the dissociation of the FcRn/IgG complex is facilitated at pH 7.4 (Rodewald, 1976; Raghavan et al., 1995; Raghavan and Bjorkman, 1996). In endothelium, FcRn appears to have an exclusively intracellular localization within acidified endosomes (Ober et al., 2004a). These observations and cellular studies have suggested of a model of IgG homeostasis, which involves the uptake of IgG into the cell via fluid-phase pinocytosis with subsequent binding to FcRn in endosomes (Ward et al., 2003; Ober et al., 2004b). Unbound IgG is directed down a degradative pathway resulting in proteolysis in lysosomes, whereas FcRn-bound antibody is recycled to the cell surface where the neutral pH facilitates dissociation and release into the circulation (Ward et al., 2003).
As monoclonal antibodies continue to show promise as therapeutic agents, there have been a number of studies that have attempted to characterize the in vivo influence of Fc mutations that affect the affinity or pH dependence of this receptor interaction (Kim et al., 1994; Ghetie et al., 1997; Dall'Acqua et al., 2002). Optimizing this interaction may lead to the development of superior therapeutic antibodies through modulation of their pharmacokinetic and/or pharmacodynamic properties. Studies have suggested that mutations that increase the affinity of the IgG/FcRn interaction at pH 6.0 result in an improvement in the terminal phase half-life (t1/2β) of an antibody in vivo (Kim et al., 1994; Ghetie et al., 1997; Medesan et al., 1998; Hinton et al., 2004, 2006), although recent studies have questioned this observation (Dall'Acqua et al., 2002; Gurbaxani et al., 2006). Other investigations suggest that modulation of the affinity of IgG for FcRn interaction at neutral pH is an important consideration (Dall'Acqua et al., 2002; Hinton et al., 2004). It is likely that optimization of the pharmacokinetics through FcRn will need to consider a combination of these properties.
In the current work, we have developed a series of humanized anti-tumor necrosis factor α (TNFα) IgG1 Fc variants that show a combination of in vitro FcRn binding properties. The variants were built on a TNFα backbone, with the intent of eliminating/limiting the influence of antigen binding on the kinetics or distribution of the antibody. The mutations made in the wild-type (WT) Fc sequence were P257I/Q311I, P257I/N434H, and D376V/N434H. [Residues are numbered according to the Eu numbering system. P257I/Q311I indicates mutation of both Pro257 to Ile257 and Glu311 to Ile311. P257I/N434H indicates mutation of both Pro257 to Ile257 and Asn434 to His434. D376V/N434H indicates mutation of both Asp376 to Val376 and Asn434 to His434.] These variants showed differential in vitro binding properties against human (H-FcRn), cynomolgus monkey (C-FcRn), and murine FcRn (M-FcRn), which allowed us to correlate in vitro binding properties to the in vivo pharmacokinetic behavior of these molecules in mice and cynomolgus monkeys. Each of the variants bound with greater affinity (∼15- to 197-fold) to H-FcRn, C-FcRn, and M-FcRn at pH 6.0 but showed no direct binding at neutral pH (7.4). After formation of the antibody/M-FcRn complexes at pH 6.0, we observed a decrease in the proportion of the variant IgG dissociating from M-FcRn with increasing pH, which was not evident after formation of the C-FcRn/IgG or H-FcRn/IgG complexes. In cynomolgus monkeys, the molecules showed similar pharmacokinetic behavior suggesting that increased binding affinity at pH 6.0 is not enough to result in improved pharmacokinetic properties for these molecules. After administration to mice, we observed an increased clearance of each of the variant IgG from the circulation relative to the WT antibody. The pharmacokinetics in mice suggest the proportion of each variant IgG released from M-FcRn at the neutral pH of the cell surface is less than the WT molecule, consistent with the differential pH-dependent release we observed in vitro. These findings suggest that even subtle influences on the pH dependence of the FcRn/IgG dissociation of these variant molecules may influence how an antibody is processed by the recycling pathway. Our results show that a systematic assessment of multiple parameters of the FcRn/IgG interaction both within and between species allows one to assess the relationship of FcRn-mediated antibody trafficking to in vivo IgG performance.
Materials and Methods
Cell Culture. 293EBNA cells were maintained at 37°C under 5 to 8% CO2 conditions in Dulbecco's modified Eagle's medium/Ham's F-12 medium (Invitrogen, Carlsbad, CA) supplemented with 20 mM HEPES (Invitrogen), 5 μg/ml nucellin (Eli Lilly and Company, Indianapolis, IN), 0.4 μg/ml tropolone (Sigma-Aldrich, St. Louis, MO), 0.075% (w/v) F68 (Invitrogen), and 50 μg/ml Geneticin (Sigma-Aldrich).
Chinese hamster ovarian (CHO)-K1 cells were maintained at 37°C under 5 to 8% CO2 conditions in Dulbecco's modified Eagle's medium/Ham's F-12 medium (Invitrogen) supplemented with 10% fetal bovine serum.
Cloning, Expression, and Purification of H-FcRn, C-FcRn, and M-FcRn. The genes encoding for the soluble portion of the heavy subunit of M-FcRn (residues 1–276), H-FcRn (residues 1–268), murine β2m, and human β2m were constructed by Geneart GmbH (Regensburg, Germany) and subsequently subcloned into cytomegalovirus expression vectors. The heavy subunits and β2m-containing plasmids were cotransfected at a 3:1 ratio, respectively, into 293EBNA cells using XtremeGene (Roche, Basel, Switzerland). H-FcRn and M-FcRn were purified from supernatants using pH-dependent binding to separate columns packed with IgG Sepharose 6 Fast Flow (GE Healthcare, Little Chalfont, Buckinghamshire, UK). Before loading, the pH of the supernatants was adjusted to 6.0, and the columns were equilibrated with 50 mM NaPO4, pH 6.0, to capture the receptors. After loading, the IgG columns were washed with 50 mM NaPO4, pH 6.0, until the absorbance returned to baseline, and the bound receptors were eluted with 50 mM NaPO4, pH 7.4. Fractions containing the receptors were pooled and further purified by size exclusion chromatography on a Superdex 200 column (GE Healthcare) with 50 mM NaPO4, pH 7.4, as the running buffer. Fractions containing H-FcRn or M-FcRn were pooled and characterized by SDS-polyacrylamide gel electrophoresis (PAGE) and mass spectrometry. Samples were dialyzed into phosphate-buffered saline (PBS) (1 mM potassium phosphate, 3 mM sodium phosphate, and 0.150 M NaCl, pH 7.4), dispensed into single-use aliquots, and stored at –20°C.
The genes encoding for the heavy subunit of C-FcRn (residues 1–299) and cynomolgus monkey β2m were cloned by polymerase chain reaction using cDNA from cynomolgus monkey peripheral blood mononuclear cells. A His6 tag was added to the 3′-end of heavy chain, and each gene was subcloned into cytomegalovirus expression vectors. The heavy subunit and cynomolgus monkey β2m were cotransfected at a 1:1 ratio into CHO-K1 cells using Lipofectamine 2000 (Invitrogen), and cells expressing the receptor were selected in serum-free Ultraculture medium (Cambrex Bio Science Walkersville, Inc., Walkersville, MD) containing 25 μM methotrexate. C-FcRn was purified from cell supernatants by immobilized metal-affinity chromatography (IMAC) using a HisTrap column (GE Healthcare). Before loading, the pH of the supernatant was adjusted to 7.0, and the columns were equilibrated with 20 mM NaPO4, 500 mM NaCl, and 200 mM arginine, pH 7.0 (buffer A), to capture the receptor. After loading, the IMAC column was washed with buffer A until the absorbance returned to baseline, and the bound C-FcRn was eluted using a linear gradient of 0 to 0.5 M imidazole (dissolved in buffer A). Fractions containing C-FcRn were pooled and further purified by anion exchange chromatography using a HiPrep Sepharose column (GE Healthcare) with 50 mM NaPO4, pH 7.4, as the running buffer. Fractions containing C-FcRn were pooled and characterized by SDS-PAGE. Samples were dialyzed into PBS, pH 7.4, dispensed into single-use aliquots, and stored at –80°C.
Construction of the IgG Fc Library. Humanized IgG1 Fc variant libraries were created using a Kunkel-based strategy (Kunkel et al., 1987) with oligonucleotides designed to mutate a single Fc residue in the clone to each of the 18 non-WT amino acids (cysteine was excluded). In brief, in a 96-well polymerase chain reaction plate, the oligonucleotide for each variant was individually annealed to an uridinylated IgG1 heavy chain (γ1) template at a 50:1 ratio by denaturing at 85°C for 1 min, cooling to 45°C over 30 min, and then chilling to 4°C. After addition of T7 DNA polymerase, T4 DNA ligase, and synthesis buffer (31), the synthesis reaction was carried out at 37°C for 2 h. A small aliquot of the reaction was then transformed into Escherichia coli XL-1 cells (Stratagene, La Jolla, CA). Individual colonies were picked for plasmid preparation and DNA sequencing to identify clones with designed Fc mutations.
The IgG were expressed from transiently transfected CHO-K1 cells. The concentration of IgG in the culture supernatant was determined by an enzyme-linked immunosorbent assay (ELISA) using light and heavy chain detection antibodies.
After IgG quantification, the concentration of Fc variant in the supernatant was normalized with mock medium, and the pH was adjusted to 6.0 with FcRn-binding buffer (100 mM NaPO4, 0.05% Tween 20, pH 6.0) and assayed in duplicate for binding to recombinant H-FcRn at pH 6.0. Single-mutation variants that displayed an enhanced H-FcRn binding at pH 6.0 relative to the WT IgG were identified and further characterized for concentration-dependent H-FcRn binding at pH 6.0 and 7.4.
Data from the interaction of H-FcRn with the single mutants were used to guide the synthesis of IgG variants with two amino acid mutations in the Fc region by Kunkel mutagenesis (Kunkel et al., 1987). The double mutants were expressed as described above and assayed for H-FcRn binding at pH 6.0 and 7.4. Three variants that displayed enhanced H-FcRn binding at pH 6.0 (relative to the WT molecule) and minimal impact on the direct binding interaction at pH 7.4 were converted to a humanized anti-TNFα IgG1 by Fab-Fc gene ligation and fully characterized as described below. The anti-TNFα IgG1 platform allowed the study of pharmacokinetics in normal primates or mice with minimal or no influence of antigen-driven clearance.
Expression and Purification of IgG. The WT anti-TNFα antibody and Fc variants P257I/N434H, P257I/Q311I, and D376V/N434H were expressed in 293EBNA cells and purified using one of the two following methods.
Expression media from 293EBNA cells were concentrated and loaded directly onto an IMAC column (GE Healthcare) at a flow rate of 5 ml/min. The column was washed with buffer A (1 mM potassium phosphate, 3 mM sodium phosphate, and 0.150 M NaCl, pH 7.4) until the absorbance returned to baseline, and the bound antibodies were eluted with a linear gradient from 0 to 0.225 M imidazole (dissolved in buffer A) over 60 min. Fractions containing the anti-TNFα antibody variants were pooled and dialyzed against 25 mM phosphate and 50 mM NaCl, pH 7.4 (buffer B). The dialysates were then concentrated and loaded onto a Superdex 200 (26/60, GE Healthcare) sizing column equilibrated with buffer B at a flow rate of 3 ml/min. Fractions containing antibody were pooled and characterized by SDS-PAGE and mass spectrometry. Samples were sterile-filtered (0.22 μm) and stored at 4°C.
For the second purification method, expression media from 293EBNA cells containing the expressed anti-TNFα antibodies were concentrated and loaded onto a recombinant protein-A Sepharose prepacked column (GE Healthcare) at a rate of 5 ml/min. The column was washed with buffer A until the absorbance returned to baseline. The antibodies were eluted with 100 mM glycine, pH 3.2, and fractions were neutralized using 40 μl of Tris (pH 8.0) per milliliter of elution buffer. Fractions containing the antibodies were pooled and concentrated and loaded onto a Superdex 200 (26/60, GE Healthcare) sizing column equilibrated with buffer A at a flow rate of 3 ml/min. Fractions containing antibody were pooled and characterized by SDS-PAGE and mass spectrometry. Samples were sterile-filtered (0.22 μm) and stored at 4°C.
Labeled FcRn Preparation. Fluorescently labeled M-FcRn and C-FcRn for anisotropy measurements were obtained by reaction of the purified soluble proteins with Alexa 488 (Invitrogen) using the vendor-recommended reaction conditions. The number of moles of Alexa 488 per mole of each FcRn was determined spectrophotometrically as 2.0 and 1.3 mol for C-FcRn and M-FcRn, respectively. Biotinylated H-FcRn, M-FcRn, and C-FcRn for ELISA assays were produced by reacting each purified soluble protein with EZ-Link Sulfo-NHS-Biotin (Pierce Chemical, Rockford, IL) using the conditions supplied by the vendor. The FcRn/biotin ratio for H-FcRn, M-FcRn, and C-FcRn was measured as 1.1, 1.2, and 1.1, respectively, using the EZ Biotin Quantitation Kit (Pierce Chemical).
Surface Plasmon Resonance (BIAcore) Measurements. Surface plasmon resonance (SPR) measurements were performed on a BIAcore 2000 instrument using a CM5 sensor chip (Biacore Inc., Uppsala, Sweden). H-FcRn, C-FcRn, and M-FcRn were immobilized to flow cells 2, 3, and 4, respectively, of the sensor chip using amine-coupling chemistry. The flow cells were activated for 7 min with a 1:1 mixture of 0.1 M N-hydroxysuccinimide and 0.1 M 3-(N,N-dimethylamino)propyl-N-ethylcarbodiimide at a flow rate of 5 μl/min. H-FcRn (4 μg/ml in 10 mM sodium acetate, pH 5.5), C-FcRn (5 μg/ml in 10 mM sodium acetate, pH 5.5), and M-FcRn (3 μg/ml in 10 mM sodium acetate, pH 5.5) were injected over flow cells 2, 3, and 4, respectively, for 15 min at 5 μl/min, which resulted in a surface density of 200 to 300 response units for each receptor. Surfaces were blocked with a 7-min injection of 1 M ethanolamine-HCl, pH 8.5. Flow cell 1 was used as a control surface without FcRn and was prepared similarly to sample flow cells. The data from this “blank” flow cell were subtracted from the sample data.
Kinetic SPR experiments for each antibody were carried out at a flow rate of 100 μl/min and a sampling rate of 1 Hz. Sensorgrams were collected for all the antibodies dissolved in PBS, pH 6.0, and 0.005% (v/v) Tween 20 at 25°C over a concentration range of 0.0033 to 4 μM. The running buffer used in the experiments was the same as the antibody diluent buffer. Signals were monitored as flow cell 2 minus flow cell 1 for H-FcRn/antibody interactions, flow cell 3 minus flow cell 1 for C-FcRn/antibody interactions, and flow cell 4 minus flow cell 1 for M-FcRn/antibody interactions. One 30-s pulse of PBS, pH 7.4, was used to regenerate the surfaces. Kinetic binding constants for the antibody/FcRn interactions were determined from data recorded at various antibody concentrations by using the program BIAevaluation, version 3.1. Global fits were determined from an average of two datasets collected on separate days. The kon (association rate) and koff (dissociation rate) were simultaneously fit to a heterogeneous binding model (Martin and Bjorkman, 1999) to determine KD (equilibrium dissociation constant) values. Curve fits for both association and dissociation phases of the sensorgrams showed low χ2 values and low residuals.
Fluorescence Anisotropy pH-Dependent Dissociation Assay for the WT, P257I/N434H, P257I/Q311I, and D376V/N434H Anti-TNFα Antibodies. Anisotropy measurements of Alexa 488 (Invitrogen) labeled M-FcRn and C-FcRn were recorded at 25°C on an ISS PC-1 fluorometer by using excitation and emission wavelengths of 494 and 519 nm, respectively. Alexa 488-FcRn was premixed with antibody to 75% saturation in PBS, pH 6.0. For the displacement of Alexa 488-FcRn from antibody, increasing amounts of 0.25 M NaOH were added while monitoring the pH with an electrode. Displacement curves were analyzed by a four-parameter nonlinear regression fit (Sigma Plot v8, SPSS Inc., Chicago, IL) to determine the midpoint (pH50) of the titration curve (the pH at which 50% of the Alexa 488-FcRn/antibody complexes dissociates).
pH-Dependent Dissociation ELISA for the WT, P257I/N434H, P257I/Q311I, and D376V/N434H Anti-TNFα Antibodies. Greiner Microlon ELISA plates were coated with 0.005 μg/well of NeutrAvidin (Pierce Chemical Co.) diluted in 50 mM carbonate buffer, pH 9.3, at 4°C overnight. After washing and blocking, 0.5 μg of biotinylated H-FcRn, C-FcRn, or M-FcRn was added to each well and allowed to bind the NeutrAvidin-coated wells for 1 h at 25°C. Wells were washed after FcRn binding, and 0.05 μg/well of anti-TNFα antibody, dissolved in 100 mM NaPO4, pH 6.0, 0.05% Tween 20 (v/v), and 0.1% ovalbumin (m/v), was added to each well and incubated for 1 h at 25°C. After antibody binding, wells were washed three times with 10- or 30-min incubations at 25°C between each wash using 100 mM NaPO4/0.05% Tween 20 (v/v) buffer having pH values from 6.0 to 8.0 in 0.2-pH increments. The remaining bound anti-TNFα antibodies were detected with a horseradish peroxidase (HRP)-conjugated goat (Fab′)2 anti-human-Fab (The Jackson Laboratory, Bar Harbor, ME). Optical density data were analyzed by the same four-parameter nonlinear regression fit as the fluorescence anisotropy data. The amount of antibody that remained bound at each pH was expressed as a percentage of the total antibody bound at pH 6.0.
Cynomolgus Monkey Pharmacokinetic Study. Nine male cynomolgus monkeys (2.8–3.8 kg) were assigned to one of three study groups. Each animal received a single i.v. dose of anti-TNFα WT, P257I/N434H, or D376V/N434H dissolved in PBS, pH 7.4, at 0.5 mg/kg. Blood samples were collected from the femoral vein before dosing and at 0.25, 0.5, 1, 3, 6, 12, 24, 48, 72, 96, 120, 168, 240, 312, 384, 456, and 528 h after administration of the dose. Blood samples were allowed to clot at ambient temperature before centrifugation to obtain serum.
Murine Pharmacokinetic Studies. Pharmacokinetic analyses for anti-TNFα WT, P257I/Q311I, P257I/N434H, and D376V/N434H were conducted in male CD-1 (20–30 g) (Harlan, Indianapolis, IN) and male B6 (C57BL/6J; 15–20 g) (The Jackson Laboratory) mice. A single i.v. dose of the anti-TNFα WT, P257I/Q3111I, P257I/N434H, or D376V/N434H dissolved in PBS, pH 7.4, was administered via the tail vein at a dose level of 1 mg/kg. Blood samples were collected from three or four animals per treatment group per time point at 0.08, 0.25, 1, 6, 12, 24, 72, 96, 144, 168, 192, 216, 288, 360, and 432 h after administration. The samples were collected by saphenous vein or by tail clip into tubes containing potassium EDTA as anticoagulant and processed to plasma.
Bioanalytical Assays. Concentrations of the WT anti-TNFα and the P257I/Q311I, P257I/N434H, and D376V/N434H variants in cynomolgus monkey and murine samples were determined using a validated anti-human IgG1 and/or antigen capture (with TNFα) (R&D Systems, Minneapolis, MN) ELISA. In brief, for the antigen capture ELISA, each well of a Immulon 4 microtiter plate (Thermo Electron Corporation, Waltham, MA) was coated with 0.1 μg of TNFα at 4°C overnight; for the anti-human IgG1 ELISA, each well was coated with 0.2 μg of a goat anti-human kappa light chain antibody. After washing and blocking, standards and samples were added to the wells in a volume of 0.1 ml and incubated for 1 h at room temperature. Standards (anti-TNFα WT, P257I/Q311I, P257I/N434H, and D376V/N434H) were prepared in either cynomolgus monkey serum or mouse plasma (EDTA), and study samples were diluted in the appropriate matrix. After washing, the bound antibodies were detected with an HRP-conjugated mouse anti-human light chain antibody (Southern Biotechnology Associates, Birmingham, AL). The standard curve range in both assay formats for anti-TNFα WT and D376V/N434H was from 0.78 to 50 ng/ml, and the lower limit of quantitation was defined as 2 ng/ml. For P257I/Q3111I and P257I/N434H, the standard curve range in both assay formats was from 1.56 to 100 ng/ml, and the lower limit of quantitation was defined as 4 ng/ml.
Pharmacokinetic Data Analysis. Pharmacokinetic parameters were calculated using WinNonlin Professional (version 3.2) software package (Pharsight Corporation, Mountain View, CA). Serum concentration-time data were calculated using a model-independent approach based on the statistical moment theory. The parameters calculated included the area under the curve (AUC0-∞), clearance (CL), volume of distribution (Vss) and elimination half-life (t1/2β).
Statistical Analyses. Statistical differences in pharmacokinetic parameters for the WT and Fc variants in cynomolgus monkey were determined using standard analysis of variance methods. In vitro binding data (KD and pH50) were also analyzed using analysis of variance. For mouse plasma concentration data, a linear mixed effect model was used that included time-by-group and quadratic terms for time-by-group as fixed effects.
Results
Binding Affinities and Interaction Kinetics of the Anti-TNFα Antibodies with FcRn. The IgG library carrying single or double Fc mutations was initially screened for binding to H-FcRn at pH 6.0 and 7.4 using a FcRn ELISA. It is possible that mutation of residues in the Fc critical to maintaining the pH dependence of receptor binding at pH 6.0 and release at pH 7.4 could generate variants with enhanced affinity at both pH values. The IgG/FcRn interaction was measured at both pH values to ensure that the variants displayed the characteristic receptor affinity at pH 6.0 and minimal interaction at pH 7.4. The variants that displayed the greatest binding enhancement to H-FcRn at pH 6.0 relative to WT with insignificant interaction at pH 7.4 were selected for further interaction characterization with C-FcRn and M-FcRn. The variants that displayed improved FcRn affinity at pH 6.0 compared with WT with no significant binding at pH 7.4, across the three species of receptors, included P257I/Q311I, P257I/N434H, and D367V/N434H. Whereas the binding ELISA was used as a screen for ranking the relative affinity improvements, a more complete understanding of the interaction with FcRn requires consideration of the equilibrium dissociation constants (KD values) and binding kinetics (kon and koff values) of each of the antibodies. We measured these parameters for our WT and three humanized Fc variant anti-TNFα IgG1 antibodies by SPR using high flow rates (100 μl/min) and a low density of surface immobilized FcRn (∼250 response units) to avoid mass transport effects on the receptor/IgG interaction. Sensorgrams for the interactions of the WT anti-TNFα and P257I/N434H variant with H-FcRn, C-FcRn, and M-FcRn at pH 6.0 are representative of the other variants (Fig. 1). Because the association and dissociation phases of the IgG interactions with human FcRn were too fast for an accurate calculation of KD values through global fitting, equilibrium binding data were used. A noninteracting two-site binding model that has been reported previously to describe the rat FcRn/Fc interaction (Martin and Bjorkman, 1999) was used to determine the rate constants and KD values for the interaction of the IgG with C-FcRn and M-FcRn.
The binding affinity (KD) of the WT antibody for H-FcRn, C-FcRn, and M-FcRn was 37 ± 5, 209 ± 11, and 118 ± 7, respectively (Tables 1, 2, 3). Compared with the WT antibody, the mutants display enhanced binding to H-FcRn, C-FcRn, and M-FcRn at pH 6.0 in the SPR assays (Tables 1, 2, 3). Importantly, none of the variants tested bound at detectable levels to the FcRn of any species tested at pH 7.4 when a concentration of 4 μM IgG (the highest concentration measured for pH 6.0 binding) was flowed over the FcRn chip (data not shown). The variants thereby maintained the characteristic pH dependence of the receptor/IgG interaction critical for regulating the persistence and levels of IgG in circulation (Dall'Acqua et al., 2002). The trend for the enhanced stabilities of H-FcRn/, C-FcRn/, and M-FcRn/variant antibody complexes at pH 6.0 shows that mutations in the CH2 and CH3 Fc regions affect FcRn binding. These results are consistent with previous reports indicating the involvement of these regions in interactions with FcRn (Burmeister et al., 1994a; Kim et al., 1994, 1999; Raghavan et al., 1994, 1995; Medesan et al., 1996; Vaughn et al., 1997; Vaughn and Bjorkman, 1998; West and Bjorkman, 2000; Martin et al., 2001; Shields et al., 2001; Dall'Acqua et al., 2002).
Overall, the variants we generated displayed affinity enhancements to the human, murine, and monkey receptors of ∼15- to 197-fold (Tables 1, 2, 3). These increases are larger than observed in earlier studies, which reported FcRn binding affinity enhancements with variant humanized IgG from ∼2- to 30-fold (Shields et al., 2001; Dall'Acqua et al., 2002; Hinton et al., 2004, 2006). The increases in binding affinities were more marked for the interaction of the Fc variants with M-FcRn than H-FcRn, analogous with previous reports of the interaction of human IgG with murine and human receptors (Dall'Acqua et al., 2002). The interaction of the WT anti-TNFα and the variants with C-FcRn was similar to H-FcRn (Fig. 1, A and B). Each of the Fc variant antibodies have comparable affinities for C-FcRn (KD values of ∼3–4 nM) and H-FcRn (KD values of ∼2–3 nM) at pH 6.0 (Tables 1 and 2), indicating these mutations in the IgG backbone make similar free energy contributions to H-FcRn and C-FcRn binding. The association rate (kon values) of the Fc variant antibody/C-FcRn complexes is ∼20- to 30-fold faster than that of the WT antibody (Table 2). The dissociation rates (koff values) of the three mutant antibody/C-FcRn complexes are similar to that (within ∼2- to 3-fold) of the WT antibody/C-FcRn complexes at pH 6.0 (Table 2). The results indicate that the changes in affinities of the variants are influenced more by the rate of association with C-FcRn.
The P257I/Q311I, P257I/N434H, and D376V/N434H Fc variants have affinities (KD values) for the murine receptor of ∼5 nM, 1 nM, and 7 nM, respectively (Table 3). Mutation of position 434 to a titratable residue (histidine) in the cases of D376V/N434H and P257I/N434H increases the affinity of each IgG for M-FcRn ∼17- and 197-fold, respectively, relative to the WT antibody. In crystal structure, the center of the FcRn/IgG binding interface includes a core of hydrophobic residues surrounded by a network of salt bridges. The Fc residues Q311 and N434 are at the border of the interface (Burmeister et al., 1994a,b; Vaughn and Bjorkman, 1998; West and Bjorkman, 2000). Mutations at these positions may increase the number of noncovalent interactions within the binding interface, which increases the affinity for FcRn. The Fc residues P257 and D376 [residues are numbered according to the Eu numbering system. P257 indicates residue Pro257; D376 indicates residue Asp376] are located in the CH2-CH3 interdomain region but are not in direct contact with FcRn (Deisenhofer, 1981). Substitutions at these positions may indirectly affect interactions at the M-FcRn/Fc binding interface through differential influence on the interdomain angle. This is evident in the interactions of the variants D376V/N434H and P257I/N434H with M-FcRn. Although both variants contain the N434H1 mutation, the P257I/N434H variant shows a more marked affinity enhancement for the murine receptor (∼12-fold more) than D376V/N434H. Enhanced IgG/M-FcRn binding affinity was caused by either an increased association rate, slowed dissociation rate, or a combination of both (Table 3). Each of the Fc variants showed a ∼3- to 6-fold increase relative to the WT antibody in their rate of association with M-FcRn and also displayed ∼3- to 30-fold slower rates of dissociation from the M-FcRn/antibody complexes compared with the WT molecule.
pH-Dependent Dissociation of H-FcRn/, C-FcRn/, and M-FcRn/Anti-TNFα Antibody Complexes. FcRn-mediated IgG salvage begins with the interaction of IgG with FcRn, which is facilitated by the acidic pH of the endosomal compartment (Ober et al., 2004b). However, the association and dissociation kinetics of the complex at various points in the intracellular recycling pathway and when the complex is exposed to the neutral pH of blood may also affect the pharmacokinetic properties of monoclonal antibodies. To measure these parameters in vitro, we formed FcRn complexes with our anti-TNFα antibodies at pH 6.0 and determined the extent of dissociation of the preformed complexes as the pH was increased using two independent methods: fluorescence anisotropy and ELISA. Fluorescence anisotropy is a solution-based method, in which preformed IgG/FcRn complexes were titrated with base and exposed to each pH change in tandem. The ELISA is a solid-phase method in which FcRn was immobilized to a plate surface. In contrast to the fluorescence anisotropy method, preformed IgG/FcRn complexes were exposed to a single pH change in the ELISA format.
The data from the pH-dependent dissociation of IgG/FcRn complexes were consistent between both assay formats (Figs. 2 and 3). After formation of the H-FcRn/IgG and C-FcRn/IgG complexes at pH 6.0, the dissociation of the WT and variant molecules from the complex as the pH increased was very similar and characteristic of the Fc/FcRn interaction (Fig. 2, A and B). The H-FcRn/ and C-FcRn/anti-TNFα antibody complexes dissociated with pH50 values of ∼6.3 to 6.6 (Tables 1 and 2). At pH 7.4, ≥87% of the WT and each Fc variant antibody were dissociated from H-FcRn and C-FcRn (Fig. 2, A and B; Tables 1 and 2). In contrast, the pH at which preformed M-FcRn/IgG complexes dissociated varied significantly, with each of the variants showing decreased dissociation from preformed receptor complexes at neutral pH relative to the WT (Figs. 2C and 3B). The pH50 at which P257I/Q311I/M-FcRn, D376V/N434H/M-FcRn, and P257I/N434H/M-FcRn complexes dissociate is pH 7.3, 7.2, and 7.6, respectively, whereas the WT/M-FcRn complex has a pH50 value of 6.6. The majority of the WT IgG (>90%) dissociates from M-FcRn at pH 7.4, whereas ∼40 to 80% of the mutant antibodies remained bound to M-FcRn (Fig. 2C; Table 3).
Pharmacokinetics of Anti-TNFα WT and Mutant Antibodies in Cynomolgus Monkeys. To understand how the in vitro affinity differences observed for the variant IgG/C-FcRn interactions correlated to antibody kinetics in vivo, we investigated the pharmacokinetics of the WT and two of the variants (D376V/N434H and P257I/N434H) in cynomolgus monkeys after i.v. administration. The third variant (P257I/Q311I) was not studied in vivo because its binding characteristics were similar to the D376V/N434H and P257I/N434H variants. The WT and variants were cleared from the serum in a biphasic manner with similar serum profiles (Fig. 4). No statistically relevant differences were observed between the WT and variant antibodies in total exposure, clearance, volume of distribution, or t1/2β (Table 4). The WT antibody had a mean elimination t1/2β of 6 days, whereas the P257I/N434H and D376V/N434H variants had a mean t1/2β of 4.3 and 4.8 days, respectively (Table 4). There were no differences observed in the serum concentrations determined by the antigen capture or human IgG1 ELISA for the WT anti-TNFα or the variants, indicating the antibodies were intact.
Pharmacokinetics of Anti-TNFα WT and Mutant Antibodies in CD-1 Mice. Because of the differences observed in the in vitro binding with M-FcRn and C-FcRn, we also examined the kinetics of the WT and three variants in CD-1 mice after i.v. administration. The WT antibody was cleared slowly (clearance value of 0.6 ml/h/kg), having an elimination t1/2β estimated at approximately 11.6 days (Table 5). Conversely, the D376V/N434H, P257I/N434H, and P257I/Q311I variants distributed from the circulation extremely rapidly (Fig. 5). The clearance of P257I/N434H appeared to be more extensive than for P257I/Q311I and D376V/N434H (Fig. 5; Table 5). The t1/2β of D376V/N434H, P257I/N434H, and P257I/Q311I were estimated as 27, 8.6, and 28 h (Table 5), respectively. Statistical analyses of the plasma concentrations of the Fc variants at 24 h postdose and beyond showed significant differences (P < 0.01) from WT. Analyses of the kinetics of the antibodies in B6 mice also showed the same rank order in clearance, exposure, and t1/2β, indicating the behavior of the molecules is not unique to the CD-1 murine strain (data not shown).
Discussion
In this report, we characterize differences in the IgG/FcRn interaction using H-FcRn, C-FcRn, and M-FcRn and relate the observations to pharmacokinetics in mice and monkeys using a series of humanized antibody Fc variants on an anti-TNFα IgG1 antibody backbone. The in vitro FcRn binding properties of these molecules within and between species provided a unique opportunity to systematically assess how multiple binding parameters, either independently or in combination, influenced their in vivo kinetics.
In endothelial cells, FcRn diverts IgG from intracellular degradative pathways to a recycling compartment, resulting in increased IgG serum persistence in vivo (Ward et al., 2003; Ober et al., 2004b). Simplistically, increasing the affinity of an antibody for FcRn at pH 6.0 would shift the equilibrium toward recycling in endosomal compartments and thereby translate to improved in vivo kinetics. Reports in the literature have suggested that there is a correlation between the serum t1/2β and binding affinity of an IgG for FcRn (Kim et al., 1994; Ghetie et al., 1997; Medesan et al., 1998; Hinton et al., 2004, 2006). Although these investigations showed that the FcRn binding affinity of an IgG correlates with increased serum t1/2β, they presented fairly limited assessments and did not develop systematic in vitro binding property relationships for prediction of in vivo pharmacokinetics. For instance, in a rhesus monkey study, there is a fundamental gap in the direct relationship between the magnitude of receptor binding affinity enhancement and increased IgG serum persistence (Hinton et al., 2004). This discrepancy has also been observed in mice (Ghetie et al., 1997; Medesan et al., 1998; Dall'Acqua et al., 2002; Gurbaxani et al., 2006). These discrepant observations are probably related to other aspects of the IgG/FcRn interaction, which were not characterized or controlled. To address these issues, multiple FcRn binding characteristics need to be measured. In addition to KD, the interaction kinetics, relative binding at neutral and acidic pH, and the effect of increasing pH on complex dissociation are factors that may need to be considered and balanced to translate in vitro observations to in vivo performance. The influences of the various in vitro binding parameters on in vivo kinetics may be different for IgG1 molecules other than anti-TNFα, as well as other IgG subtypes and other specific Fc variants.
Because the only difference in the in vitro C-FcRn binding properties of our IgG was affinity at pH 6.0 (Table 2), we were able to evaluate the effect of this parameter on in vivo performance independent of other influences. In cynomolgus monkeys, we observed that the clearance and t1/2β of all the variant antibodies were similar to that of the WT antibody after i.v. administration (Fig. 4). In rhesus monkeys, recent reports show humanized variant IgG1 and IgG2 molecules with ∼30-fold increases in FcRn binding in vitro have 2-fold increased serum persistence (Hinton et al., 2004, 2006). Although we observed enhancements in binding affinity that were greater than or comparable with those of Hinton et al. (2004, 2006) (Table 3), our mutations do not translate to an increased serum persistence. Because none of our variants bound to C-FcRn at pH 7.4 and all were released in a pH-dependent fashion (Fig. 2B), it is very unlikely that cell surface binding at neutral pH could have offset the benefit provided by the improved pH 6.0 binding. Others have suggested that the kinetics of dissociation at endosomal pH (pH 6.0) may influence antibody pharmacokinetics (Ghetie et al., 1997). The increased affinity of our variants for C-FcRn was driven predominantly by an increase in association rate, with each displaying rates of dissociation from complexes similar to that of the WT molecule (Table 2). Because we did not observe improved in vivo properties of our variants in primates (Table 4; Fig. 4), one could speculate that, at the cellular level, the rate of dissociation of the FcRn/IgG complexes at endosomal pH may be an additional factor determining whether a preponderance of antibody is processed through the recycling or degradative pathways. Although Hinton et al. (2004, 2006) did not report on the kinetics of binding of their molecules with FcRn, it is plausible that the affinity increase imparted by their mutation is driven by off-rate kinetics, which in turn leads to the observed pharmacokinetic benefit. Because of the similarity in binding interactions with C-FcRn and H-FcRn, it would be anticipated that our mutations on the anti-TNFα molecule would also not translate to pharmacokinetic benefit in humans (Tables 1 and 2). Overall, the results suggest that simply measuring the in vitro binding affinity at pH 6.0 is not the sole predictor of in vivo pharmacokinetics of these anti-TNFα variants.
The differential interactions of the anti-TNFα antibodies with M-FcRn allowed us to systematically examine the combined effect of these parameters using mice as the in vivo model. The P257I/N434H/M-FcRn and P257I/Q311I/M-FcRn complexes displayed rates of dissociation slower than the WT complex at pH 6.0 (Table 3), whereas the D376V/N434H complex has a koff value similar to that of the WT IgG (Table 3). Based on our speculation around the influence of this parameter on intracellular trafficking and pharmacokinetics in primates, we expected that the slower rates of dissociation at pH 6.0 may drive a greater proportion of the IgG in these complexes (P257I/N434H and P257I/Q311I) to the recycling pathway compared with those having faster dissociation kinetics (D376V/N434H and WT). This mechanism would result in the P257I/N434H and P257I/Q311I variants having a slower clearance and/or longer serum half-life in mice. However, after i.v. administration, the P257I/N434H, P257I/Q311I, and D376V/N434H variants were cleared ∼9 to 30 times faster than the WT antibody in mice, with the levels not measurable in plasma 24 to 96 h after dosing. Others have also observed a rapid clearance of humanized Fc variants from circulation in mice that displayed enhanced affinity for M-FcRn at pH 6.0 in vitro (Dall'Acqua et al., 2002). In these studies, variants that had increased affinity for M-FcRn at pH 6.0 also displayed an uncharacteristic concomitant increase in affinity for the murine receptor at pH 7.4 (Dall'Acqua et al., 2002). It was speculated that the benefit of improved affinity at pH 6.0 was likely offset by the direct pH 7.4 binding, which allows IgG to rebind to the receptor at the cell surface, resulting in inefficient release into the circulation. Our humanized antibody variants are not acting through a rebinding mechanism at pH 7.4 because they showed no direct binding to M-FcRn at pH 7.4. In contrast to the C-FcRn observations, ∼40 to 80% of our variant IgG do not dissociate from preformed M-FcRn/variant complexes even after 30 min of incubation at neutral pH (Fig. 2C). It is likely that the inability to dissociate from FcRn at neutral pH traps our variant anti-TNFα IgG in the cell as the receptor complex recycles with the IgG still bound. This mechanism yields the poor pharmacokinetic profile of these molecules in mice in a process that is distinct from the rebinding phenomenon at pH 7.4 (Dall'Acqua et al., 2002).
In summary, there are a number of parameters of the IgG/FcRn interaction that alone or in combination affect the in vivo clearance of our IgG. Our studies clearly show that changes in the pH dependence of complex dissociation appear to negatively affect the clearance of these variant anti-TNFα antibodies in vivo. Under either of these conditions it appears that the residence time of FcRn on cell surfaces is not long enough to allow significant proportions of the variants to dissociate from the complexes into the circulation. Because of their residence within FcRn-containing tissues, such as the vasculature lining, these complexes may be capable of immune surveillance (Yoshida et al., 2004). Mechanistically, our variants may act as FcRn blockers that facilitate the degradation of endogenous IgG in vivo (Vaccaro et al., 2005). Alternatively, the consequence of enhanced retention of the variants within the tissues could also lead to antibody degradation. Even without an influence on the pH dependence of the interaction as we observed in the primate, it is plausible that the acidic microenvironment created by Na+-H+ exchange close to the endothelial cell surface (Claypool et al., 2004) results in slower release of high binding affinity IgG from FcRn, offsetting the advantage of higher affinity binding observed at pH 6.0 (Ober et al., 2004a; Lencer and Blumberg, 2005). The rank order in the increased clearance of the variants in mice parallels their dissociation behaviors from M-FcRn at pH 6.0 and 7.4, suggesting both of these phenomena influence the pharmacokinetics of the variants in mice. These studies have shown that modification of FcRn/antibody interactions for optimization of pharmacokinetic properties in vivo more than likely involves the consideration of multiple factors in tandem, including binding affinities, rates of interaction, and pH-dependent dissociation. Understanding the relative influence of each of these factors on IgG kinetics, as well as in vitro approaches that integrate these parameters, would be valuable to directing therapeutic antibody engineering.
Acknowledgments
We thank Don McClure and Joseph Brunson for the expression of FcRn and anti-TNFα antibodies, Peng Luan for help with purification of the anti-TNFα antibodies, Fabian Tibaldi, Kevin Guo, and Justin Recknor for performing the statistical analyses of the data, and Nancy Goebl, Mary-Ann Campbell, Elaine Conner, and Brian Ondek for helpful discussions.
Footnotes
-
Article, publication date, and citation information can be found at http://dmd.aspetjournals.org.
-
doi:10.1124/dmd.106.011734.
-
ABBREVIATIONS: FcRn, neonatal Fc receptor; β2m, β2-microglobulin; t1/2β, terminal phase half-life; TNFα, tumor necrosis factor α; WT, wild-type; H-FcRn, human FcRn; C-FcRn, cynomolgus monkey FcRn; M-FcRn, murine FcRn; CHO, Chinese hamster ovarian; PAGE, polyacrylamide gel electrophoresis; PBS, phosphate-buffered saline; IMAC, immobilized metal-affinity column; ELISA, enzyme-linked immunosorbent assay; SPR, surface plasmon resonance; kon, association rate; koff, dissociation rate; KD, equilibrium dissociation constant; pH50, pH at which 50% of the FcRn-antibody complexes dissociate; HRP, horseradish peroxidase; AUC0-∞, area under the plasma concentration curve from zero to infinity; CL, clearance; Vss, volume of distribution at steady state.
- Received June 29, 2006.
- Accepted October 11, 2006.
- The American Society for Pharmacology and Experimental Therapeutics