Properties of Human IgG1s Engineered for Enhanced Binding to the Neonatal Fc Receptor (FcRn)*

We describe here the functional implications of an increase in IgG binding to the neonatal Fc receptor. We have defined in a systematic fashion the relationship between enhanced FcRn binding of a humanized anti-respiratory sincytial virus (RSV) monoclonal antibody (MEDI-524) and the corresponding biological consequences in cynomolgus monkeys. The triple mutation M252Y/S254T/T256E (YTE) was introduced into the Fc portion of MEDI-524. Whereas these substitutions did not affect the ability of MEDI-524 to bind to its cognate antigen and inhibit RSV replication, they resulted in a 10-fold increase in its binding to both cynomolgus monkey and human FcRn at pH 6.0. MEDI-524-YTE was efficiently released from FcRn at pH 7.4 in both cases. We show that MEDI-524-YTE consistently exhibited a nearly 4-fold increase in serum half-life in cynomolgus monkeys when compared with MEDI-524. This constituted the largest half-life improvement described to date for an IgG in a primate. For the first time, we demonstrate that these sustained serum levels resulted in an up to 4-fold increase in lung bioavailability. Importantly, we also establish that our non-human primate model is relevant to human. Finally, we report that the YTE triple substitution provided a means to modulate the antibody-dependent cell-mediated cytotoxicity (ADCC) activity of a humanized IgG1 directed against the human integrin αvβ3. Therefore, the YTE substitutions allow the simultaneous modulation of serum half-life, tissue distribution and activity of a given human IgG1.

We describe here the functional implications of an increase in IgG binding to the neonatal Fc receptor. We have defined in a systematic fashion the relationship between enhanced FcRn binding of a humanized anti-respiratory sincytial virus (RSV) monoclonal antibody (MEDI-524) and the corresponding biological consequences in cynomolgus monkeys. The triple mutation M252Y/S254T/T256E (YTE) was introduced into the Fc portion of MEDI-524. Whereas these substitutions did not affect the ability of MEDI-524 to bind to its cognate antigen and inhibit RSV replication, they resulted in a 10-fold increase in its binding to both cynomolgus monkey and human FcRn at pH 6.0. MEDI-524-YTE was efficiently released from FcRn at pH 7.4 in both cases. We show that MEDI-524-YTE consistently exhibited a nearly 4-fold increase in serum half-life in cynomolgus monkeys when compared with MEDI-524. This constituted the largest half-life improvement described to date for an IgG in a primate. For the first time, we demonstrate that these sustained serum levels resulted in an up to 4-fold increase in lung bioavailability. Importantly, we also establish that our non-human primate model is relevant to human. Finally, we report that the YTE triple substitution provided a means to modulate the antibodydependent cell-mediated cytotoxicity (ADCC) activity of a humanized IgG1 directed against the human integrin ␣ v ␤ 3 . Therefore, the YTE substitutions allow the simultaneous modulation of serum half-life, tissue distribution and activity of a given human IgG1.
Conclusive evidence indicates that the neonatal Fc receptor (FcRn) 3 plays a central role in the regulation of serum IgG levels in mammals (reviewed in Ref. 1). The Fc-binding FcRn is a heterodimer, which comprises ␤ 2 -microglobulin and a membrane-anchored ␣ chain that is related to the ␣ chain of major histocompatibility complex class I molecules. The corresponding sequences and overall structures are well conserved across species such as in human, mouse, rat, and cow (2,3,4,5,6,7), as well as in non-eutherian mammals (8). However, differences in their binding specificity to various IgG exist (9), indicating the presence of species-specific binding mechanisms.
The IgG-FcRn interaction exhibits a remarkable pH dependence, varying from strongest at slightly acidic pH to marginal under neutral and basic conditions (10,11). This crucial characteristic is intricately linked to the IgG salvage mechanism, which involves recycling of FcRn-bound molecules from within acidic lysosomes back to the general circulation (1). As a result, recycled IgGs exhibit a significantly prolonged serum half-life when compared with other serum proteins. It is thought that this rescue takes place in the endothelial cells of adult tissues and particularly in small arteries and capillaries (12,13). IgGs that retain significant binding to FcRn at neutral pH have a significantly decreased serum persistence (14).
At present, no direct evidence exists for a correlation between increased IgG binding to FcRn at pH 6.0 and longer serum half-life in human. However, previous work carried out in mouse (15,16) and non-human primates (17,18) suggest that engineering IgGs for better binding to FcRn may represent a viable strategy to modulate their half-lives in human. Therapeutic antibodies that exhibited longer half-lives likely would be of benefit with increased efficacy because of sustained serum concentrations, decreased dosing frequency and/or lower cost of goods.
While the judiciousness of engineering the Fc-FcRn interaction to prolong serum half-lives seems to be well-established, much remains to be known about the potential biological consequences related to the IgG-transportation role of FcRn. Most notably, this function includes the transcytosis of IgGs through human placenta (19) and intestine (20,21), as well as the reabsorption of IgG in the human kidneys (22). FcRn has also been reported to be actively involved in the transport of IgG from the lumen of the lungs to the systemic circulation in both mice (23) and cynomolgus monkeys (24). However, the in vivo consequences of increasing the affinity of IgGs to FcRn on their distribution to FcRn-expressing tissues are still unknown.
We describe here the behavior of a humanized IgG1 Fc variant dosed in a non-human primate. The Fc changes consisted of a triple substitution (M252Y/S254T/T256E; EU numbering as reported in (25); referred to as YTE thereafter) in the C H 2 domain of MEDI-524 (26,27). We have shown previously that these mutations increase the binding of palivizumab to human FcRn by about 10-fold at pH 6.0 while allowing efficient release at pH 7.4 (14). However, the in vivo behavior of such a mutated human IgG in a relevant model (non-human primate) was unknown. Thus, detailed pharmacokinetics studies in cynomolgus monkeys were specifically designed in an effort to further clarify the relationship between the affinities of the IgG/ FcRn interaction and the corresponding IgG serum half-lives and lung distribution properties. For the first time, we describe here the consequence of an increased binding to FcRn on the distribution of IgGs from the systemic circulation to respiratory airways. We also report the largest serum half-life increase to date in a primate. Finally, the ability of YTE to modulate the antibody dependent cell-mediated cytotoxicity (ADCC) activity of an anti-␣ v ␤ 3 humanized IgG1 (MEDI-522, formerly known as Vitaxin®; Ref. 28) is described, and new insights into the molecular basis of the Fc-FcRn interaction revealed. Based on the established relevance of our animal model to human, we propose that our set of substitutions will be useful in the development of the next generation of therapeutic antibodies.

EXPERIMENTAL PROCEDURES
Reagents-All chemicals were of analytical grade. Restriction enzymes and DNA-modifying enzymes were purchased from New England Biolabs, Inc. (Beverly, MA). Oligonucleotides were purchased from Invitrogen (Carlsbad, CA).
Generation of MEDI-524-YTE-The heavy and light chains of the humanized anti-RSV monoclonal antibody MEDI-524 were cloned into a mammalian expression vector encoding a human cytomegalovirus major immediate early (hCMVie) enhancer, promoter and 5Ј-untranslated region (29). In this system, a human ␥1 chain is secreted along with a human chain (30). A combination of three mutations (M252Y/S254T/T256E; EU numbering, (25) was introduced into the heavy chain of MEDI-524. Generation of these mutations was carried out by site-directed mutagenesis using a QuickChange XL mutagenesis kit (Stratagene) according to the manufacturer's instructions, and the primers: 5Ј-GCATGTGACCTCAGGTTCCCGAGTGATATA-GAGGGTGTCCTTGGG-3Ј and 5Ј-CCCAAGGACACCCTC-TATATCACTCGGGAACCTGAGGTCACATGC-3Ј. This generated MEDI-524-YTE. The sequence was verified using an ABI 3100 sequencer. NS0 (murine myeloma) cells were then stably transfected with the corresponding antibody constructs, and the secreted immunoglobulins were purified using protein A and standard protocols.
14. An additional purification step was included, whose aim was to remove any FcRn molecule potentially containing endogenously produced human ␤ 2 microglobulin. This consisted in applying the IgG Sepharose-purified cynomolgus monkey FcRn to a 1-ml nitrilotriacetic acid (Ni 2ϩ -NTA) affinity column according to the manufacturer's instructions (Qiagen, Valencia, CA). Purified cynomolgus monkey FcRn (Ͼ95% homogeneity as judged by SDS-PAGE) was dialyzed against phosphate-buffered saline (PBS), flash frozen, and stored at Ϫ70°C.
Determination of the Dissociation Constants (K D ) of the Various IgG/FcRn Pairs-The interaction of soluble human and cynomolgus monkey FcRn with immobilized MEDI-524 and MEDI-524-YTE was monitored by surface plasmon resonance detection using a BIAcore 3000 instrument (Biacore International AB, Uppsala, Sweden). Protein concentrations were calculated by the bicinchoninic acid (BCA) method for both human and cynomolgus monkey FcRn or using the 1% extinction coefficient at 280 nm of 1.47 for MEDI-524 and MEDI-524-YTE Both IgGs were coupled to the dextran matrix of a CM5 sensor chip (Biacore International AB) using an Amine Coupling Kit as described (33) at a surface density of ϳ1000 RU. Excess reactive esters were quenched by injection of 70 l of 1.0 M ethanolamine hydrochloride, pH 8.5. Human and cynomolgus monkey FcRn were buffer-exchanged against 50 mM PBS, pH 6.0 containing 0.05% Tween 20 and used in equilibrium binding experiments. All measurements were performed at 25°C with FcRn concentrations typically ranging from 2.86 M to 6 nM at a flow rate of 5 l/min; data were collected for ϳ50 min and three 1-min pulses of PBS, pH 7.4 containing 0.05% Tween 20 were used to regenerate the surfaces. Human and cynomolgus monkey FcRn were also flowed over an uncoated cell and the sensorgrams from these blank runs subtracted from those obtained with IgG-coupled cells. Dissociation constants were determined by fitting the binding isotherms using Graph-Pad Prism (GraphPad Software, Inc.).
The interaction of soluble human FcRn with immobilized MEDI-522, MEDI-522-S239D/A330L/I332E, MEDI-522-YTE, and MEDI-522-YTE/S239D/A330L/I332E was monitored using a BIAcore 3000 instrument as described above. IgGs were coupled to the surface of a CM5 sensor chip at a surface density of ϳ8000 RU. Human FcRn was buffer-exchanged against 50 mM PBS, pH 6.0 containing 0.05% Tween 20 and used in equilibrium binding experiments at concentrations typically ranging from 2.86 M to 8 nM at a flow rate of 5 l/min. Data were collected as described in the previous paragraph. Values for K D were determined by fitting the binding isotherms using Graph-Pad Prism.
Determination of the FcRn pH Dependence of Binding toward Various Fc Variants-Comparison of the interaction of soluble human and cynomolgus monkey FcRn with immobilized MEDI-524 and MEDI-524-YTE at acidic and neutral pH was carried out using a BIAcore 3000 in 50 mM PBS, pH 6.0 containing 0.05% Tween 20 or 50 mM PBS, pH 7.4 containing 0.05% Tween 20, respectively. IgGs were coupled to the surface of a CM5 sensor chip at a surface density of ϳ1000 RU whereas human and cynomolgus monkey FcRn were used at a typical concentration of 500 nM. Analysis of the pH dependence of binding for MEDI-522, MEDI-522-S239D/A330L/I332E, MEDI-522-YTE, and MEDI-522-YTE/S239D/A330L/I332E was carried out in a similar fashion after coupling the IgGs at a surface density of ϳ8000 RU and flowing human FcRn at a typical concentration of 250 nM. In all cases, FcRn from both species was also flowed over an uncoated cell, and the sensorgrams from these blank runs subtracted from those obtained with IgG-coupled cells.
Microneutralization Assays-The microneutralization assays were carried out essentially as described (30). Briefly, dilutions of MEDI-524 or MEDI-524-YTE were made in quadruplicate in a 96-well plate. RSV Long (ATCC, Manassas, VA) was added to each well and incubated for 2 h at 37°C in 5% CO 2 . 2 ϫ 10 4 HEp-2 cells (ATCC, Manassas, VA) were then added to each well and incubated for 5 days at 37°C in 5% CO 2 . Cells were then washed three times with PBS containing 0.1% Tween 20 and fixed with acetone. Viral replication was quantified by successive incubations with a mouse anti-RSV monoclonal antibody (Chemicon, Temecula, CA) and a horseradish peroxidase conjugate of a goat anti-mouse IgG (TAGO, Burlingame, CA). Peroxidase activity was detected with 3,3Ј,5,5Ј-tetramethylbenzidine (TMB), and the reaction was quenched with 2 M H 2 SO 4 . The absorbance was read at 450 nm and means plotted for each antibody concentration.
Pharmacokinetics Studies in Cynomolgus Monkeys-A first non-GLP (Good Laboratory Practice) pharmacokinetics study (Study A) was approved by GeneLogic's Institutional Animal Care and Use Committee (IACUC) and conducted at Gene-Logic (GeneLogic Laboratories, Gaithersburg, MD). Twenty male cynomolgus monkeys were randomized using computergenerated random numbers and assigned to one of two study groups. Each animal received a single intravenous (i.v.) dose of MEDI-524 or MEDI-524-YTE at 30 mg/kg. Blood samples were drawn prior to dosing on day 0, at 1 and 4 h after dosing, and at 1, 2, 3, 4, 6,8,10,12,14,16,20,24,31,41, and 55 days after dosing. The serum concentrations of MEDI-524 or MEDI-524-YTE were determined using the anti-MEDI-524 ELISA described in the next section. For each i.v. infusion, a non-compartmental model was fitted for the serum concentration data of each animal using SAS 8.0 (SAS Institute, Cary, NC). Descriptive statistics for several pharmacokinetics parameters were then calculated.
A second non-GLP pharmacokinetics study (Study B) was approved by the GeneLogic IACUC and conducted at Gene-Logic (GeneLogic Laboratories, Gaithersburg, MD). Twelve male and twelve female cynomolgus monkeys were randomized separately using computer-generated random numbers and assigned to one of two study groups (6 males and 6 females per group). Each animal received a single i.v. dose of MEDI-524 or MEDI-524-YTE at 30 mg/kg. Bronchio-alveolar lavages (BALs) were conducted at 4 and 24 days after dosing. More precisely, 6 animals (3 males and 3 females) of each group were sacrificed by barbiturate overdose and exsanguinations at each time point. Chests were opened and the lungs were removed. The distal trachea of each animal was opened to visualize the bifurcation of the bronchi. Lavages were then carried out using a syringe fitted with an appropriate tube. A fresh syringe and tubing was used for each collection. Approximately 8 ml of physiological saline warmed for at least 1 h prior in a water bath set to maintain 39 Ϯ 3°C was introduced into the main bronchus of the right lung until a slight expansion of the lung was noted. Aspirates were transferred in a tube and the procedure was carried out again on the right lung. The procedures were then repeated for the left lung. Blood samples were also drawn on surviving animals prior to dosing on day 0 and at 1, 4, 8, 14, and 24 days after dosing. The concentrations of MEDI-524 or MEDI-524-YTE in serum and BAL samples were determined using the anti-MEDI-524 ELISA described in the next section. Pharmacokinetics parameters were calculated as described above for Study A.
Analysis of the Various Cynomolgus Monkey Biological Samples-The concentrations of MEDI-524 or MEDI-524-YTE in the serum and BAL samples derived from Study A and B were determined using an anti-human IgG ELISA and appropriate standards. In this assay, MEDI-524 and MEDI-524-YTE were captured by a goat anti-MEDI-524 antibody (anti-idiotype, MedImmune, Inc.) coated to a microtiter plate. Any bound MEDI-524 or MEDI-524-YTE was detected using a goat anti-human IgG antibody linked to biotin. Streptavidin conjugated to horseradish peroxidase followed by TMB (KPL, Gaithersburg, MD) as substrate were used for the colorimetric reaction. Total protein levels in the BAL samples from Study B were also determined by the BCA method.
The possible presence of anti-MEDI-524 antibodies in the serum samples derived from Study A (pre-dose and at days 6, 14, 31, and 55 post-infusion) and B (pre-dose and at days 4, 8, and 24 post-infusion) was determined using an ELISA as follows: typically, 15 ng of MEDI-524 or MEDI-524-YTE were coated onto the wells of a microtiter plate, which were then blocked with 0.1% Tween 20/0.5% BSA/PBS and incubated with appropriate dilutions of serum samples. Plates were then incubated with horseradish peroxidase-conjugated MEDI-524. Horseradish peroxidase activity was detected with TMB substrate, and the reaction quenched with stop solution (KPL). Plates were read at 450 nm.
Comparison of the Binding of Human and Cynomolgus Monkey IgG to Human and Cynomolgus Monkey FcRn-The interaction of soluble human and cynomolgus monkey IgG with immobilized human and cynomolgus monkey FcRn was monitored as follows: briefly, human and cynomolgus monkey FcRn were coupled to the surface of a CM5 sensor chip at a surface density of ϳ2500 RU. MEDI-524, MEDI-524-YTE and purified cynomolgus monkey IgG (Antibodies, Inc., Davis, CA) were used in equilibrium binding experiments at concentrations typically ranging from 3 M to 16 nM at a flow rate of 5 l/min. Dilutions were made in 50 mM PBS, pH 6.0 containing 0.05% Tween 20. Data were collected for ϳ50 min and three 1-min pulses of PBS, pH 7.4 containing 0.05% Tween 20 were used to regenerate the surfaces. IgGs were also flowed over an uncoated cell and the sensorgrams from these blank runs subtracted from those obtained with FcRn-coupled cells.
Comparison of the Binding of MEDI-524 and MEDI-524-YTE to Human and Cynomolgus Monkey Sera-The interaction of MEDI-524 and MEDI-524-YTE with human and cynomolgus monkey sera was monitored after coupling the IgGs to the surface of a CM5 sensor chip at a surface density of ϳ2100 RU. Human (MedImmune, Inc.) and cynomolgus monkey (Antibodies, Inc., Salem, NH) sera were then used in binding experiments at dilutions of 1:10 and 1:2500 at a flow rate of 5 l/min. Dilutions were made in 50 mM PBS, pH 7.4 containing 0.05% Tween 20. Data were collected for ϳ50 min and three 1-min pulses of PBS, pH 7.4 containing 0.05% Tween 20 were used to regenerate the surfaces. Sera were also flowed over an uncoated cell and the sensorgrams from these blank runs subtracted from those obtained with IgG-coupled cells.
Cloning, Expression, and Purification of FLAG-tagged Human Fc␥RIIIA-The C-terminal end of the extracellular domain of human Fc␥RIIIA (F158 allotype) was fused with a FLAG tag. Cloning and expression were carried out as follows: briefly, the EA1/EA2 (see below) oligonucleotides combination was used to PCR amplify the extracellular domain of human Fc␥RIIIA using a cDNA library of human bone marrow (Clontech, Mountain View, CA) as the template and standard protocols. The PCR product was then cloned as an XbaI/NotI fragment into the mammalian cell expression vector described earlier for IgG and FcRn expression. HEK-293 cells were transiently transfected with this construct using Lipofectamine and standard protocols. Supernatants were harvested three times at 72, 144, and 216 h post-transfection and pooled. The secreted, soluble human FLAG-tagged Fc␥RIIIA was purified from the conditioned medium directly on an anti-FLAG M2 agarose column according to the manufacturer's instructions (Sigma). The FLAG-tagged Fc␥RIIIA was then dialyzed against PBS buffer and stored at Ϫ70°C. EA1: 5Ј-TCCACAGGTGTCCACTCC- Analysis of Human Fc␥RIIIA Binding to MEDI-522 and Its Fc Variants-The interaction of soluble human FLAG-tagged Fc␥RIIIA (F158 allotype) with immobilized MEDI-522, MEDI-522-YTE, MEDI-522-S239D/A330L/I332E, and MEDI-522-YTE/S239D/A330L/I332E was monitored by surface plasmon resonance detection using a BIAcore 3000 instrument. Protein concentrations were calculated by the BCA method. The different humanized IgG1s were coupled to the dextran matrix of a CM5 sensor chip at a surface density of between 7745 and 10426 RU. FLAG-tagged human Fc␥RIIIA was used in equilibrium binding experiments at concentrations ranging from 16 M to 31.3 nM at a flow rate of 5 l/min. Dilutions and binding experiments were carried out in 50 mM HBS buffer containing 0.01 M HEPES, pH 7.4, 0.15 M NaCl, 3 mM EDTA and 0.005% P-20. Data were collected for about 50 min, and one 30-s pulse of 5 mM HCl was used to regenerate the surfaces. FLAG-tagged human Fc␥RIIIA was allowed to flow over an uncoated cell, and the sensorgrams from these blank runs subtracted from those obtained with IgG-coupled cells. Values for K D were determined from the corresponding binding isotherms using GraphPad Prism.
ADCC Assays-Typically, the ADCC activity of MEDI-522 and of its Fc variants was assessed as follows: human blood samples were collected from independent healthy volunteers using heparinized syringes, diluted with twice the volume of PBS buffer, layered onto a Lymphoprep gradient (ICN, Irvine, CA), and centrifuged at 400 ϫ g for 30 min at room temperature. Peripheral blood mononuclear cells (PBMCs) were harvested from the interface, washed three times with PBS, and resuspended in Roswell Park Memorial Institute (RPMI) 1640 medium with L-glutamine (Invitrogen) supplemented with 10% fetal bovine serum (FBS). 40 ng/ml of recombinant human IL-2 (R&D Systems, Minneapolis, MN) was then added to the PBMCs followed by incubation overnight of these cells at 37°C in T-175 flasks (BD Biosciences, Bedford, MA). Cultured M21 (mouse melanoma) cells, which exhibit high surface expression of ␣ v ␤ 3 , were harvested the following day and resuspended in RPMI 1640 supplemented with 5% FBS (assay buffer) at a density of 2 ϫ 10 5 cells/ml. These were then added to a 96-well round bottom tissue culture plate (BD Biosciences, Bedford, MA) at 50 l/well along with various concentrations of antibody at 50 l/well in assay buffer (see above), and the mixtures were preincubated at 37°C for 30 min. PBMCs were then harvested from their overnight incubation and resuspended at 5 ϫ 10 6 cells/ml (for an Effector (E):Target (T) ratio of 50:1) and 2.5 ϫ 10 6 cells/ml (for an E:T ratio of 25:1) in assay buffer (see above) and added at 100 l/well to the assay plate. 25 l/well of 9% Triton X-100 (Promega, Madison, WI) was added as a control for complete lysis. The plates were centrifuged at 300 ϫ g for 3 min, and incubation at 37°C was continued for 4 h. Plates were then centrifuged at 300 ϫ g for 10 min, and 50 l of supernatant from each well was transferred to MaxiSorp 96-well plates (BD Biosciences, Bedford, MA). 50 l of reconstituted substrate mix (CytoTox 96 Non-Radioactive Cytotoxicity assay kit, Promega) was then added to all wells and incubated in the dark at room temperature for 30 min. 50 l of stop solution (Promega) was added to each well and lactate dehydrogenase (LDH) release was quantified by measuring the absorbance at 490 nm. % cytotoxicity was calculated using Equation 1. Experimental corresponds to the signal measured in one of the conditions of interest described above, Effector Spontaneous corresponds to the signal measured in the presence of PBMCs alone, Target Spontaneous corresponds to the signal measured in the presence of M21 cells alone and Target Maximum corresponds to the signal measured in the presence of detergentlysed M21 cells.

Characterization of the IgG Variants-Binding of MEDI-524
and MEDI-524-YTE to both cynomolgus monkey and human FcRn was analyzed by BIAcore as described under "Experimental Procedures." Results are shown in Table 1. The K D for the interaction of MEDI-524 with human FcRn agrees well with previous values determined for other human IgG1s (14,19). Similarly, the affinity determined for the binding of MEDI-524-YTE to human FcRn is in complete agreement with that measured for the same triple mutation in a different humanized IgG1 background (namely palivizumab, Ref. 14). Interestingly, the observed affinities for the binding of MEDI-524 and MEDI-524-YTE to cynomolgus monkey FcRn were ϳ2-fold higher than seen for human FcRn. However, when compared with the unmodified IgG, the introduction of YTE into MEDI-524 resulted in a very similar affinity increase of IgG binding to both cynomolgus monkey and human FcRn at pH 6.0 (9-and 11-fold, respectively). No such increase was detected at pH 7.4. In the latter condition, both MEDI-524 and MEDI-524-YTE did not exhibit any significant binding to either cynomolgus monkey or human FcRn (Fig. 1, A and B).
An RSV microneutralization assay was conducted to investigate the effect of YTE on antibody function. As seen in Fig. 2, both MEDI-524 and MEDI-524-YTE exhibited undistinguishable RSV microneutralization properties, indicating that the triple Fc mutation did not result in major structural changes of the IgG molecule or in significant alteration of its functional activity. This is in complete agreement with data showing that the affinity of MEDI-524-YTE for its cognate antigen (RSV F protein) is not significantly different to that 4 W. F. Dall'Acqua, unpublished observations. of MEDI-524 (data not shown). The lack of significant YTErelated structural effects in the Fc region is further confirmed by the ability of MEDI-524-YTE to bind and be purified by protein A (see "Experimental Procedures, Generation of MEDI-524-YTE").
Pharmacokinetics Studies-The in vivo consequences of increasing the binding of MEDI-524 to FcRn were investigated by two separate pharmacokinetics studies in cynomolgus monkeys. More precisely, Study A was designed to examine changes in IgG serum half-life, whereas Study B proposed to investigate changes in both IgG serum half-life and distribution to the lungs following systemic administration.
The serum pharmacokinetics profiles derived from Study A for MEDI-524 and MEDI-524-YTE are shown in Fig. 3. MEDI-524-YTE exhibited a significantly increased serum persistence when compared with MEDI-524. Indeed, although the mean serum concentration of MEDI-524 and MEDI-524-YTE were similar during the first day post-dose, the serum levels of MEDI-524 decreased rapidly from ϳ300 g/ml at day 2 to below the limit of detection at the last day of the study (day 55). In contrast, the serum levels of MEDI-524-YTE gradually decreased from ϳ380 g/ml at day 2 to ϳ60 g/ml at day 55. This was further confirmed by the determination of the corresponding pharmacokinetics parameters reported in Table 2. As shown in this table, the serum elimination (␤ phase) half-life of MEDI-524-YTE was nearly four times greater than that of MEDI-524. Likewise, the area under the curve (AUC) for MEDI-524-YTE was nearly five times larger than for MEDI-524. The Wilcoxon test (34) was used to compare these two parameters between the two treatment groups. It suggested that the group differences in half-lives and AUCs were statistically significant ( p Ͻ 0.001). The mean maximum serum antibody concentration (C MAX ) was very similar for MEDI-524 and MEDI-524-YTE and was achieved by 1-h post-infusion in most of the animals.
The serum pharmacokinetics profiles and corresponding parameters derived from Study B are shown in Fig. 4 and Table  2, respectively. The same trends as in Study A were observed. Specifically, the serum elimination (␤ phase) half-life and AUC of MEDI-524-YTE were over three and eight times, respectively, as much as that of MEDI-524. Here again, the mean maximum serum antibody concentrations were not significantly different between MEDI-524 and MEDI-524-YTE.
To determine whether changes in affinity for FcRn would alter the distribution of antibodies to the lungs, BALs were collected at two time points for both MEDI-524 and MEDI-524-YTE and the corresponding human IgG levels measured. Data are reported in Fig. 5A and show that the MEDI-524-YTE levels in BALs were significantly increased at days 4 (2.6-fold) and 24 (4.1-fold) post-infusion when compared with MEDI-524. A similar increase was observed when the BAL IgG levels were normalized for total protein concentration (Fig. 5B). Finally, [BAL]/[serum] ratios at days 4 and 24 post-dose were very similar between MEDI-524 and MEDI-524-M252Y/S254T/T256E (Fig. 6), indicating that the inherent ability of MEDI-524-YTE to traffic to the lungs was not altered.
None of the animals from Studies A and B which were infused with MEDI-524 or MEDI-524-YTE tested positive for the presence of anti-MEDI-524 antibody (at a 1:640 serum dilu-    AUGUST 18, 2006 • VOLUME 281 • NUMBER 33 tion; data not shown). Although the ELISA used for this purpose could not detect antibodies specifically recognizing the YTE motif, this result suggested that MEDI-524-YTE was not immunogenic in cynomolgus monkeys. This conclusion is in good agreement with the corresponding individual pharmacokinetics profiles in which no precipitous clearance of MEDI-524 or MEDI-524-YTE was observed in any of the animals up to the last day post-infusion (day 55 and 24 for study A and B, respectively). Our data also agree with previous reports showing that human and humanized IgGs are only slightly immunogenic (if at all) in cynomolgus monkeys and other non-human primates (35)(36)(37).

Increase in Human IgG1 Binding to the Neonatal Fc Receptor
Comparative Binding of Human/Cynomolgus Monkey IgG to Human/Cynomolgus Monkey FcRn-To investigate the influence of endogenous IgG on the clearance rate of human IgGs in cynomolgus monkeys, binding of purified MEDI-524 and cynomolgus monkey IgG to cynomolgus monkey FcRn was studied. As shown in Fig. 7A, human and cynomolgus IgGs bound sim-ilarly well to cynomolgus monkey FcRn. More specifically, the ratio of the human over cynomolgus IgG concentrations at R eq50% (corrected equilibrium response corresponding to 50% of the maximum signal) was estimated at 1.2. Thus, our data indicated that the strength of the endogenous IgG/FcRn interactions in cynomolgus monkeys was similar to that found in human, thereby validating our choice of this particular nonhuman primate model for pharmacokinetics studies. The same observations were made when binding of human and cynomolgus IgGs to human FcRn was studied (Fig. 7B). In this situation, the ratio of the human over cynomolgus IgG concentrations at R eq50% was estimated at 0.9.
Serum Binding Abilities of MEDI-524 and MEDI-524-YTE-Comparison of the serum binding abilities of MEDI-524 and MEDI-524-YTE was assessed using BIAcore. As shown in Fig. 8, A and B, both IgGs exhibited no significant binding to cynomolgus monkey and human sera, respectively. Although the sensorgrams corresponding to the interaction of MEDI-524 and MEDI-524-YTE with human serum did not perfectly overlap, we did not consider this difference to be significant. Indeed, complex mixtures such as serum typically exhibit a substantial bulk effect; therefore, corrections to remove this nonspecific component often result in signal variations spanning several RUs (as exemplified by the negative corrected signal exhibited by MEDI-524-YTE; Fig. 8B). Thus, our data indicated that: (i) the introduction of YTE into MEDI-524 did not modify its binding properties toward human or cynomolgus monkey serum components, and (ii) the extended serum half-life seen for MEDI-524-YTE in cynomolgus monkeys was not because of a serum protein carrier effect.

DISCUSSION
Several crucial points of intervention exist to engineer the potency of therapeutic antibodies. Most notably, these include serum half-life, tissue distribution and effector functions. The ability to modulate these different properties, individually or together, could lead to the generation of novel, improved therapeutics. First, multiple benefits could be derived from immunoglobulins exhibiting very long serum persistence. Among those benefits are the possibility of decreasing their administration frequency, while maintaining or improving their efficacy. Second, the optimization of the tissue distribution of a given therapeutic antibody would also potentially enhance its efficacy. Indeed, the latter is directly linked to the efficiency with which the molecule reaches its site of action, such as the respiratory tract (38) and rheumatoid joints (39). Third, the ability to up-regulate effector functions could be an important tool in increasing the potency of various therapeutic antibodies, to the   Human and cynomolgus monkey sera were injected at a 1:2500 dilution on a surface onto which 2103 RU of MEDI-524 and 2124 RU of MEDI-524-YTE had been coupled. When sera were used at a 1:10 dilution, no significant difference in the binding activity of both IgGs was observed (data not shown). a Affinity measurements were carried out by BIAcore as described under "Experimental Procedures." b Errors were estimated as the S. D. of at least two independent experiments for each interacting pair. c Residue numbering is according to EU (25). extent that their killing properties toward tumors can often be correlated with efficient ADCD/CDC activities (40,41). Fourth and last, the ability to decrease ADCC and/or CDC could be beneficial in terms of minimizing toxicity in a therapeutic setting (42,43). The triple Fc mutation M252Y/S254T/T256E (YTE) was previously shown by us to increase the binding of a humanized anti-RSV monoclonal antibody to both murine and human FcRn by ϳ10-fold (14). In an effort to further characterize this set of mutations and modulate the different properties discussed above, we describe here the behavior of two correspondingly mutated humanized IgG1s, first in terms of serum pharmacokinetics and lung penetration in cynomolgus monkeys, and second in terms of effector functions.
We have introduced YTE into the Fc portion of MEDI-524, a humanized monoclonal antibody (IgG1, ) derived by in vitro affinity maturation of palivizumab. MEDI-524 exhibits potent anti-RSV neutralizing activity (26,27) and is currently undergoing clinical trials in human for prevention of severe RSV infection in high-risk infants. This particular background was chosen so that the interpretation of the in vivo data were not complicated by significant binding of the IgG to an endogenous antigen. In this study, we have shown that: (i) MEDI-524 and MEDI-524-YTE both exhibited a similar affinity to human and cynomolgus monkey FcRn (within 2-fold of each other), (ii) the YTE triple mutation resulted in the same binding affinity increase of MEDI-524 to human and cynomolgus monkey FcRn (ϳ10-fold), and (iii) MEDI-524-YTE retained its pH dependence of binding, showing no significant interaction with human and cynomolgus monkey FcRn at pH 7.4. This last point was of particular importance since it was previously noted that a lack of efficient IgG release at neutral pH lead to their decreased serum persistence in mouse (14). Because of these similarities between human and cynomolgus monkey FcRn in terms of binding to MEDI-524 and MEDI-524-YTE, we concluded that this particular non-human primate was a relevant and predictive pharmacokinetics model for studying the functional effects of increasing binding of IgG to FcRn. Importantly, we have also shown that the interaction of cynomolgus monkey FcRn with endogenous (cynomolgus) immunoglobulin G was similar to the corresponding interaction with human IgG (see "Results" and Fig.  7A). This further established that the cynomolgus monkey in vivo context closely resembles that found in human; therefore, the pharmacokinetics properties, which we describe here for our YTE-modified IgG, are likely to be seen in the human.
We have shown here that the serum elimination half-life of MEDI-524-YTE was increased by a factor of up to ϳ4-fold when compared with the unmodified version of MEDI-524. The serum half-life of MEDI-524 was typical of that seen for a human IgG1 in cynomolgus monkeys (44,45). To our knowledge, this constitutes the largest serum half-life increase for a human IgG in a primate. The mean maximum serum antibody concentrations were indistinguishable between MEDI-524 and MEDI-524-YTE, indicating that the increased serum levels seen for MEDI-524-YTE were not due to altered distribution properties in the general circulation. MEDI-524-YTE also exhibited a significant increase in its total exposure, as its area under the curve extrapolated to infinity (AUC) was 5-8-fold larger than seen for MEDI-524. Interestingly, Hinton et al. (17,18) have described a different set of mutations (T250Q/M428L) which, when introduced into the Fc portion of human IgG1 and IgG2, result in a significant increase in both their binding to rhesus FcRn and serum half-life in rhesus monkeys (nearly 40and 2.5-fold, respectively, for IgG1 and 25-and 2-fold, respectively, for IgG2). These author's data, when compared with ours, seem to indicate the lack of a direct correlation between increase in FcRn binding and increase in serum-half-life. This may be attributable to inherently different IgG recycling mechanisms across primates. Alternatively, T250Q/M428L and/or YTE may invoke alternate, FcRn-independent IgG rescue or clearance strategies. To this end, it is worth noting that MEDI-524-YTE did not exhibit significantly enhanced binding to cynomolgus monkey serum when compared with MEDI-524, suggesting that its enhanced serum half-life was not mediated through specific interaction with serum components. The same lack of reactivity with human serum will be important in the interpretation of future in vivo data in human. It is also possible that endogenous IgGs affected the clearance rate of the Fcmodified human IgGs to a different extent in cynomolgus and rhesus monkeys. Therefore, it would be interesting to compare the relative strength of the interactions between human IgG/ rhesus FcRn and rhesus IgG/rhesus FcRn, as was done in our study for the cynomolgus monkeys. Conceivably, a large difference in the corresponding affinities could result in an under-or over estimation of the predicted serum half-life enhancement of the test human IgGs in human because of varying competition levels between endogenous IgGs and endogenous FcRn.
It has been reported that reduced binding of an immunocytokine to mouse Fc␥R could lead to its increased serum half-life in mice (46). When compared with MEDI-524, the binding affinity of MEDI-524-YTE to recombinant human Fc␥RIIA, Fc␥RIIB, and Fc␥RIIIA was decreased by 3-, 2-, and 3-fold, respectively, whereas binding to human Fc␥RI was not significantly affected (data not shown). Therefore, in light of the high similarity between human and cynomolgus monkey Fc␥ receptors (Ͼ85% identity at the amino acid level for Fc␥RI, Fc␥RIIA, Fc␥RIIB, and Fc␥RIIIA; US patent number 6,911,321), we cannot rule out that this phenomenon might play, along with FcRn, a role in increasing the serum half-life of MEDI-524-YTE. However, we note that the model used by Gillies et al. (46) significantly differs from ours because of the potential ability of immunocytokines to co-cross-link Fc␥ receptors with cytokine receptors and trigger their own internalization. A decrease in the ability of the fusion protein to bind Fc␥R would decrease co-cross-linking of the receptors and thus alter its rate of clearance.
The increased persistence of MEDI-524-YTE in serum was reflected by up to 4-fold higher lung levels as measured in bronchio-alveolar lavages. Essentially similar results were observed when the IgG levels in BALs were normalized for total protein concentration, suggesting no particular bias during collection of the corresponding samples. Importantly, the [BAL]/[serum] ratios were similar between MEDI-524 and MEDI-524-YTE, suggesting that the higher BAL levels exhibited by MEDI-524-YTE were a direct reflection of its increased serum persistence as opposed to a specific accumulation of this variant. It is worth noting that FcRn has been implicated in the active transport of Fc-containing molecules from the lumen of the central airways to the systemic circulation in cynomolgus monkeys (24). Thus, the BAL concentrations of MEDI-524-YTE may represent the snapshot of a phenomenon in which an enhanced passive transport from the serum to the lumen is counterbalanced by an active transport pathway in the opposite direction. To date however, little is known about the role of FcRn in a potential IgG active transport mechanism from the general circulation to the lungs. Therefore, the BAL levels observed for MEDI-524-YTE could conceivably be the result of an equilibrium between two active, opposite transport pathways.
Our study provided new insights on the nature of the Fc/FcRn interaction. Although the binding affinities of palivizumab (14), MEDI-524, and MEDI-522 (Table 1) to human FcRn were similar, some significant differences were nevertheless observed. Whereas palivizumab and MEDI-524 exhibited undistinguishable dissociation constants (2.2-2.5 M), MEDI-522 bound more tightly by a factor of about 2-fold (1.3 M). We primarily attribute this result to binding variations caused by different lots of human FcRn. However, because these IgGs all had identical Fc regions, it may also suggest that the Fab portion plays a discernible (though minor) role in the Fc-FcRn interaction. It is likely that such a role would be mediated through long range interaction(s) as the FcRn binding site is located at the interface between the IgG C H 2 and C H 3 domains (5,47).
We have also studied the effect of YTE on effector functions. To this end, we introduced YTE into the Fc portion of MEDI-522, a humanized, affinity-optimized monoclonal antibody (IgG1, ) directed against the human ␣ v ␤ 3 integrin complex (28). The indications of this molecule include the treatment of rheumatoid arthritis and of solid tumors because of its antiinflammatory and anti-angiogenic properties (48,49). This particular IgG1 was chosen because of its ability to mediate efficient ADCC. We have shown that YTE significantly reduced the binding of MEDI-522 to human Fc␥RIIIA (F158 allotype) as well as its ADCC activity. Our results are in good agreement with the fact that Fc positions 254 and 256 have previously been shown to be implicated in Fc␥RIIIA/F158 binding (50). It is likely that MEDI-522-YTE would exhibit different binding affinities toward other Fc␥RIIIA allotypes, such as V158 whose inherent affinity to human IgGs is higher than F158 (51). Likewise, the various donors used here were not genotyped in terms of their Fc␥RIIIA allotypic diversity; therefore, it is possible that the ADCC activities we derived did not exactly reflect our binding data in which the F158 allotype was used. Importantly, the ADCC activity of MEDI-522-YTE as well as its binding to Fc␥RIIIA can be restored at a level significantly higher than the unmodified MEDI-522 through the introduction of an ADCCenhancing triple substitution (S239D/A330L/I332E; Ref. 31). Thus, YTE provides a reversible way to modulate the ADCC function of a human IgG1. In contrast, the serum half-life-enhancing double mutation T250Q/M428L described by Hinton et al. (17,18) was shown to minimally impact the CDC and ADCC activities of a humanized anti-HLA-DR IgG1.
In summary, we have dissected the functional consequences of a pH-dependent increase in IgG binding to FcRn and shown that the YTE triple mutation is generally applicable to human IgG1s for this purpose. It will be interesting to know if our approach can be used in conjunction with the strategy described by Hinton et al. (17,18) to further increase serum half-life. Additional characterization of our Fc mutations will require a more detailed analysis, such as the determination of the pharmacokinetics properties of YTE-containing IgGs in the context of their binding to an endogenously expressed antigen. Likewise, the impact of YTE on IgG distribution to the general circulation following aerosol administration into the airways will provide new insights on the nature of FcRn-mediated transport pathways. Conceivably, this could also constitute a novel method to increase the serum bioavailability of IgG or Fc fusion proteins after deposition into the upper respiratory tract. We believe that our set of mutations will be a valuable addition to the antibody therapy field.