Soluble Monomeric IgG1 Fc

Background: The Fc region of an antibody is a homodimer of two CH2-CH3 chains.

Results: Monomeric IgG1 Fcs (mFcs) were generated by using a novel panning/screening procedure.

Conclusion: The mFcs are highly soluble and retain binding to human FcRn comparable with that of Fc.

Significance: The mFcs are promising for the development of novel therapeutic antibodies of small size and long half-lives.

Abstract

Antibody fragments are emerging as promising biopharmaceuticals because of their relatively small size and other unique properties. However, compared with full-size antibodies, these antibody fragments lack the ability to bind the neonatal Fc receptor (FcRn) and have reduced half-lives. Fc engineered to bind antigens but preserve interactions with FcRn and Fc fused with monomeric proteins currently are being developed as candidate therapeutics with prolonged half-lives; in these and other cases, Fc is a dimer of two CH2-CH3 chains. To further reduce the size of Fc but preserve FcRn binding, we generated three human soluble monomeric IgG1 Fcs (mFcs) by using a combination of structure-based rational protein design combined with multiple screening strategies. These mFcs were highly soluble and retained binding to human FcRn comparable with that of Fc. These results provide direct experimental evidence that efficient binding to human FcRn does not require human Fc dimerization. The newly identified mFcs are promising for the development of mFc fusion proteins and for novel types of mFc-based therapeutic antibodies of small size and long half-lives.

Introduction

Therapeutic monoclonal antibodies have been improved gradually during the past two decades (1–10). Most of the mAbs approved for clinical use are full-size (7, 9, 10). However, full-size antibodies exhibit poor penetration into tissues, especially solid tumors, and also poor or absent binding to regions of some antigens that are occluded and can only be accessed by molecules of smaller size (11–13). Engineering of a variety of antibody fragments of smaller size such as Fab, Fv, scFv, heavy chain variable domain (VH), and other antibody fragment formats are under development (9, 10, 14–18). However, to date, these antibody fragments have been of limited therapeutic applications because they usually display greatly reduced half-lives compared with full-size IgG. One approach to increase half-lives is by fusion with Fc or by engineering binding sites to the neonatal Fc receptor (FcRn).²

The Fc contributes to the long half-life of IgG through its unique pH-dependent association with the neonatal Fc receptor (FcRn) (19, 20). The IgG Fc can bind to FcRn in the acidic environment of the endosome after internalization and then be recycled into the cell surface and released into circulation. This protects IgG from degradation and increases its serum half-life. A technology to produce Fc fusion proteins has been proven effective in extending half-lives of therapeutic molecules (21, 22). A typical Fc fusion protein contains two effector molecules because the Fc fragment of the IgG consists of a tightly packed homodimer, and each molecule is attached to one chain of the Fc dimer. Recently, so-called “monomeric Fc fusion proteins” were generated by fusing a single active protein to dimeric wild-type Fc (23–25). Such smaller molecules have been shown to possess extended half-lives compared with the dimeric version and are promising for therapeutic applications. Despite this advancement, the Fc domain in a fusion protein is still dimeric and of relatively large size (∼50 kDa).

It has been debated whether the dimeric Fc binds to two molecules FcRn or only one and whether the dimeric state of Fc is required for efficient binding to FcRn (20, 26–30). Several experiments showed a 2:1 FcRn/Fc binding stoichiometry (26, 28, 29), whereas the 1:1 FcRn-Fc complex was also observed in some studies (27, 29). A better understanding of these interactions could be helped by generating and using monomeric Fc (mFc).

Here, we describe the identification and characterization of highly soluble functional mFcs of half the Fc size and preserved binding to FcRn. Fc dimerization is mediated mainly by a large hydrophobic interface in its CH3 domain, which involves at least 16 residues in each polypeptide chain that make intermolecular interactions (31, 32). Disruption of this large interaction interface would cause exposure of a hydrophobic surface resulting in poor solubility, instability, and/or aggregation. To solve this problem, a large phage library was constructed in which >10⁹ human IgG1 Fc individual molecules were displayed with extensive mutations in the CH3 dimer interface. This library was first selected by protein G to screen for soluble and well folded variants, and then functional clones were further enriched by panning directly against human FcRn for several rounds.

Using these strategies, we generated three mFc proteins that are >99% monomeric as indicated by size exclusion chromatography. These novel antibody formats are relatively stable and also represent, to the best of our knowledge, the smallest fragments (∼27 kDa) of IgG that retain binding to FcRn comparable with that of Fc. A VH-mFc fusion protein was still monomeric and functional, suggesting that using mFc to replace dimeric wild-type Fc for generating fusion proteins is promising. Moreover, mFcs could be efficiently expressed in Escherichia coli, with a yield of 15–20 mg of purified proteins per liter culture. The mFc bound to FcRn similarly to Fc, suggesting that dimeric form of Fc is not required for efficient binding to FcRn. These results also suggest the possibility of using monomeric Fc as scaffolds for antibody engineering. The desired binders would have small size, long half-lives, and antigen-binding ability, which could make them promising candidate therapeutics and diagnostics for certain purposes.

EXPERIMENTAL PROCEDURES

Library Construction and Selection of Monomeric Fc Clones

A large phage display library (∼1.3 × 10⁹ individuals) was constructed by randomly mutating seven residues located at the CH3 dimer interface of human IgG1 Fc (Leu-351, Thr-366, Leu-368, Pro-395, Phe-405, Tyr-407, and Lys-409). Amplified libraries with 10¹² phage-displayed Fc mutants were applied to a pre-equilibrated protein G column (Roche Applied Science). The resins were washed extensively with PBS, and the bound phages were eluted by 0.1 m HCl glycine (pH 2.2). The elution was then neutralized with 1 m Tris base, mixed with TG1 cells for 1 h at 37 °C, and the phages capable of binding to protein G were amplified from the infected cells and used in the biopanning against FcRn. Human FcRn, containing both β and α chains in a 1:1 molar ratio, was expressed in mammalian cells and purified as a soluble protein as described previously (33). Libraries with 10¹² phages were mixed with PBS (pH 6.0) and incubated in ELISA wells coated with FcRn for 2 h at 37 °C. After incubation, the wells were washed 10 times for the first round and 20 times for the later rounds with PBS (pH 6.0) containing 0.05% Tween 20. The bound phages were eluted with PBS (pH 7.4), amplified by infecting TG1 cells along with helper phage M13KO7 (Invitrogen). 80 clones were randomly picked from the fifth selection round, transferred into HB2151 cells, inoculated into 3 ml of 2YT medium containing 100 μg/ml ampicillin, and incubated for 2 h at 37 °C with shaking at 250 rpm. After the addition of 1.5 μl of isopropyl-1-thio-β-d-galactopyranoside, bacteria were grown for 3 additional h, harvested by centrifugation. The pellet was resuspended in PBS buffer containing 5 microunits of polymixin B (Sigma-Aldrich), incubated at room temperature for 30 min, and centrifuged at 10,000 × g for 10 min. The supernatant was separated by non-reducing SDS-PAGE without boiling and then analyzed by Western blot. The clones that did not show evidence of dimeric Fc via Western blot (∼54-kDa bands) were selected for further characterization.

Expression and Purification of Monomeric Fc Proteins

The selected clones (mFc.1, mFc.23, and mFc.67) were sequenced, and plasmids extracted from these clones were used for transformation of HB2151 cells. A single and freshly transformed colony was inoculated into 200 ml of SB medium with 100 μg/ml ampicillin and incubated at 37 °C with vigorous shaking at 250 rpm. When optical density of the culture at 600 nm reached ∼0.6, expression was induced by the addition of isopropyl-1-thio-β-d-galactopyranoside to a final concentration of 1 mm, and the culture was further incubated at 30 °C for 6 h. Cells were then harvested by centrifugation at 6000 rpm for 15 min and resuspended in PBS buffer. Polymixin B (Sigma-Aldrich) (0.5 milliunits/ml) was added to the suspension (1:1000). After 30 min of incubation with rotation at 50 rpm at room temperature, the culture was centrifuged at 12,000 rpm for 15 min at 4 °C. The supernatant was used for further purification by nickel-nitrilotriacetic acid resin (Qiagen, Valencia, CA) according to the manufacturer's protocols. Protein purity was estimated as >95% by SDS-PAGE, and protein concentration was measured spectrophotometrically (NanoVue, GE Healthcare).

Size Exclusion Chromatography

The molecular composition of purified proteins was analyzed by size exclusion chromatography using an FPLC AKTA BASIC pH/C system (GE Healthcare) with a Superdex 75 10/300 GL column (GE Healthcare). PBS (pH 7.4) was selected as the running buffer, and a flow rate of 0.5 ml/min was used. Eluting protein was monitored at 280 nm. The molecular mass standards used were ribonuclease A (13.7 kDa), chymotrypsinogen A (25 kDa), ovalbumin (44 kDa), bovine serum albumin (67 kDa), and aldolase (158 kDa).

Reverse Mutation

The reverse mutation assay was conducted using a selected monomeric Fc clone, mFc.23, as template, in which the dependent residues were mutated back to their original counterparts in wild-type Fc. Six mFc.23 mutants, S351L, R366T, H368L, R366T/H368L, K395P, and E405F/K407Y/A409K, were all generated by the overlap extension PCR method. Each mutation was confirmed by automated DNA sequencing. The reverse mutations were expressed and purified using a similar procedure as monomeric Fc proteins described above.

Circular Dichroism (CD)

The circular dichroism spectra of CH2, Fc, and mFc proteins were collected from 190 to 250 nm (0.1-cm path length), with an AVIV model 202 spectropolarimeter (Aviv Biomedical). The protein samples were dissolved in PBS (pH 7.4) at a final concentration of 0.4 mg/ml. Spectra were first recorded at 25 °C for native structure measurements. For evaluation of the refolding, the samples were heated slowly to 90 °C (1 °C/min), kept at 90 °C for 10 min, and then rapidly cooled down to 25 °C (10 °C/min), and the spectra were recorded again. For evaluation of thermal stability, CD signals at 216 nm were recorded, and the instrument was programmed to acquire spectra at 1 °C intervals over the range 25–90 °C.

Spectrofluorometry

Fluorescence spectra were measured using a Fluoromax-3 spectrofluorometer (HORIBA Jobin Yvon, Inc., Edison, NJ). For urea denaturation tests, Fc and mFc proteins were dissolved in urea-containing buffers (50 mm Tris-Cl, 450 mm NaCl, pH 8.0, 0 to 8 m urea), to give a final protein concentration of 10 μg/ml. The samples were kept overnight at 4 °C, and fluorescence measurements were performed with the excitation wavelength at 280 nm. The emission spectra were recorded from 320 to 380 nm, and fluorescence intensity at 343 nm was used for quantitative evaluation of urea unfolding.

Serum Stability Assay

Normal human serum was collected from healthy human donors approved by the NCI-Frederick Research Donor Program. Wild-type human Fc (30 μg) and mFc proteins (15 μg) were incubated with normal human serum in PBS at 37 °C. An aliquot was taken out at each time point and immediately stored at −80 °C. Western blot and ELISA assays were applied to check the serum stability. For Western blot, samples were electrophoresed through SDS-PAGE and transferred onto a 0.2-μm nitrocellulose membrane (Bio-Rad). The membrane was blocked with 3% milk in PBS for 1 h at room temperature and then incubated with anti-His Tag monoclonal antibody (ABM, Vancouver, Canada) for 1 h. Washing with PBST was followed by incubation with anti-mouse IgG-alkaline phosphatase antibody (Sigma-Aldrich). The BCIP/NBT substrate solution (Sigma-Aldrich) was used for detection. For the ELISA test, wells were coated with 50 μl of anti-Fc Fab (Sigma-Aldrich) and then blocked in 100 μl of protein-free blocking buffer (Thermo Scientific) for 1 h at 37 °C. After five washes with PBST, wells were incubated with samples for 2 h at 37 °C. Following six washes with PBST, 50 μl of HRP-conjugated anti-FLAG tag antibody (Sigma-Aldrich) were added and incubated for 1 h at 37 °C. The assay was developed with ABTS substrate (Roche Applied Science) and monitored at 405 nm.

Binding ELISA

Antigens were coated on ELISA plate wells at 50 ng per well in PBS overnight at 4 °C and blocked with protein-free blocking buffer for 1 h at 37 °C. 3-Fold serially diluted protein was added and incubated at 37 °C for 2 h. The plates were washed with PBST, and HRP-conjugated anti-FLAG tag antibody in PBS was incubated with wells for 1 h at 37 °C. After extensive washes with PBST, binding was detected with the addition of ABTS substrate, and signals were read at 405 nm.

Surface Plasmon Resonance Experiments

The interaction of samples with immobilized FcRn was monitored by surface plasmon resonance detection using a BIAcore X100 instrument (GE Healthcare). Purified human FcRn was diluted in 10 mm sodium acetate buffer (pH 5.0) and immobilized on a CM5 biosensor chip using an amine coupling kit. The running buffer was PBS with 0.005% Tween 20 for binding at pH 7.4 or PBS pH 6.0 with 0.005% Tween 20 for binding at pH 6.0. The proteins diluted with running buffer were allowed to flow through the cells, at concentrations ranging from 125 nm to 2000 nm. After 10 min of dissociation, the chip was regenerated with pH 8.0 buffer (100 mm Tris, pH 8.0, 50 mm NaCl). Another test with a protein concentration of 500 nm was repeated to monitor the regeneration efficiency. As a negative control for the FcRn binding assay, the Fc I253A/S254A/H435A/Y436A mutant was constructed using the overlap extension method.

Generation of m36 Fusion Proteins

The following primers were used: Omp, 5′-AAGACAGCTATCGCGATTGCAG-3′; gIIIF, 5′-ATCACCGGAACCAGAGCCACCAC-3′; m36R, 5′-TGAGGAGACGGTGACCAGGGTGCCCTG-3′; mFcF, 5′- CTGGTCACCGTCTCCTCAGCACCTGAACTCCTGGG-3′; and mFcCH3F, 5′-CTGGTCACCGTCTCCTCACCCCGAGAACCACAGGTGTAC-3′. The m36 gene was amplified by PCR (primer, Omp and m36R) with the m36-encoding plasmid pCom36 (13) as a template. The mFc.67 gene (primer, mFcF and gIIIF) and mFc.67 CH3 domain gene (primer, mFcCH3F and gIIIF) were amplified from the mFc.67 plasmid, which was constructed in a pComb3x vector. The m36 fragment was joined to mFc.67 or the CH3 domain of mFc.67 by overlap extension PCR, and the products were digested with SfiI enzyme and cloned into a pComb3x vector. The fusion proteins were expressed and purified using a similar procedure as for monomeric Fc proteins described above.

RESULTS

Identification of mFc from Library of Fc Mutants

To identify mFcs, we utilized a combination of structure-based rational protein design combined with multiple screening strategies of Fc mutant libraries. Fc dimerization is mainly mediated by a large (1000 Å buried surface), tightly packed interface between the two CH3 domains (Fig. 1A). This interface is composed of multiple regions containing at least 16 residues in each chain, most of which are hydrophobic (31, 32). We identified four regions in the CH3 domain of human IgG1 contributing to the interchain interactions with the following critical contact residues: Leu-351, Thr-366, and Thr-368, Pro-395, Phe-405, Tyr-407, and Lys-409 (Fig. 1A). This is in agreement with previous studies, e.g. significant destabilizing effects were found by mutation of the above seven residues in human IgG1 CH3 (34). Two problems must be solved to generate the soluble Fc monomer: disruption of the strong interactions between these residues and prevention of protein aggregation due to exposure of the hydrophobic residues.

FIGURE 1.

A, structure of human IgG1 Fc CH3 domain, showing key residues in its dimerization interface (Protein Data Bank code 2WAH). B, schematic of the multiple panning/screening strategies. C, amino acid sequence alignment of human IgG1 Fc, mFc.1, mFc.23, and mFc.67. Here, the numbering begins with the IgG1 Fc; thus residues 121, 136, 138, 165, 175, 177, and 179 correspond to residues 351, 366, 368, 395, 405, 407, and 409 in the IgG1 numbering system, respectively.

We hypothesized that functional and highly soluble Fc monomers could be produced by panning and screening of a large Fc library with extensive mutations in the hydrophobic interface for proper folding, FcRn binding, and solubility. Thus, a phage library was constructed by randomly mutating the above seven residues (Leu-351, Thr-366, Leu-368, Pro-395, Phe-405, Tyr-407, and Lys-409) in human IgG1 Fc. Initially, this library was panned directly against human single-chain soluble FcRn (33). Buffer with pH 6.0 was used for washing and buffer with pH 7.4 was used for elution to select pH-dependent binders. However, enrichment was not observed after two rounds of panning. It is likely that functional binders were masked by ubiquitous misfolded molecules in the library. Thus, the library was first panned against protein G resulting in a library enriched for phage-displayed soluble and well folded mFcs. This library was further panned against human FcRn for five rounds as described above. To select for highly expressed monomers, 80 clones from the final enriched library were expressed in E. coli and screened by non-reducing SDS-PAGE and Western blot. This panning/screening procedure is schematically depicted in Fig. 1B.

Using this procedure, three mFc proteins, mFc.1, mFc.23, and mFc.67, were identified (Fig. 1C). The mFc.1 and mFc.23 contain seven mutations in the CH3 dimer interface, whereas mFc.67 contains six. Although contain different mutations, all of them were expressed in E. coli with high efficiency. Purified mFcs were obtained with yields of 15–20 mg liter⁻¹ bacterial culture. All proteins were monomeric with molecular masses of ∼27 kDa as demonstrated by size exclusion chromatography (Fig. 2A).

FIGURE 2.

Size exclusion chromatography of wild-type IgG1 Fc, mFc.1, mFc.23, and mFc.67 (A) and the S351L, R366T, H368L, R366T/H368L, K395P, and E405F/K407Y/A409K mutants (B) of mFc.23. In both panels, the insets show a standard curve for the gel filtration standards. MW, molecular weight.

Identification of Mutations in mFc That Are Essential and Sufficient to Maintain Soluble Monomer

To determine which residues/mutations are essential for generation of a soluble mFc, we reversed certain mutations in mFc.23. Six mFc.23 mutants, S351L, R366T, H368L, R366T/H368L, K395P, and E405F/K407Y/A409K were generated. They were expressed and purified using a similar procedure as used for the other mFc proteins and then analyzed by size exclusion chromatography (Fig. 2B). Shoulder peaks appeared near the monomer peaks for the reverse mutants S351L, R366T, and R366T/H368L, suggesting formation of a small proportion of dimers. Although only single peaks were observed for H368L and K395P, their shapes were broad and distorted. The E405F/K407Y/A409K mutant of mFc.23 appeared completely monomeric, but its expression was relatively low (5 mg protein/liter). Taken together, these results suggest that four specific mutations are essential to produce mFc and that the number of mutations needed for formation of soluble mFc can be reduced but results in lower expression.

Stability of mFc

To establish the thermal stability of the mFcs, their CD ellipticity at 216 nm was measured as a function of temperature (Fig. 3A). The stability of the wild-type dimeric Fc and the isolated monomeric CH2 domain were also examined. The midpoint transition temperatures (T_m) for Fc, mFc.1, mFc.23, mFc.67, and CH2 were 75.1 ± 0.5 °C, 45.0 ± 0.6 °C, 45.2 ± 0.6 °C, 51.0 ± 0.5 °C, and 54.0 ± 0.8 °C, respectively. The thermal refolding was reversible as indicated by the fact that CD spectra of the samples after cooling from 90 °C were similar to the original measurements (Fig. 3B).

FIGURE 3.

A, plots of the change in fraction folded (calculated from CD molar ellipticity at 216 nm) for Fc, CH2, mFc.1, mFc.23, and mFc.67. B, CD spectra of Fc, mFc.1, mFc.23, and mFc.67 at 25 °C (black line) and refolding samples at 25 °C after heating to 90 °C (dotted line).

The stability was further tested by urea denaturation experiments monitored by fluorescence spectroscopy (data not shown). The 50% unfolding of Fc occurred at higher urea concentrations (5.8 m) than that of mFc.1 (4.1 m), mFc.23 (4.1 m), mFc.67 (4.3 m), and CH2 (4.2 m).

We next evaluated the serum stability of the mFc proteins. Samples were incubated with human serum at 37 °C, and an aliquot was taken out at each time point and stored at −80 °C before Western blot (Fig. 4A) or ELISA (Fig. 4B) analyses. Fig. 4A shows that the band for mFc.1 disappeared after 3 days incubation, whereas the bands for Fc, mFc.23, and mFc.67 were not evidently diminished even after an 11-day incubation. We further assessed protein degradation by ELISA. Anti-Fc Fab was coated on ELISA plates to capture Fc and mFcs, and anti-FLAG-HRP conjugate were used for detection. From Fig. 4B, it is evident that Fc, mFc.23, and mFc.67 were degraded more slowly than mFc.1, in agreement with the Western blot results. These data suggest that Fc, mFc.23, and mFc.67 have high serum stability. It is possible that some mutations in mFc.1 make it more accessible to proteases in human serum.

FIGURE 4.

Serum stability of Fc, mFc.1, mFc.23, and mFc.67 measured by Western blot (A) and ELISA (B). For ELISA, plates were coated with anti-Fc F(ab′)₂, and anti-FLAG HRP conjugate was used for detection.

mFcs Bind to FcRn

To test whether these mFcs are functional and behave in a similar manner as wild-type Fc, we first checked whether they could still bind protein A and protein G. A monomeric Fc, mFc.1, was applied to a pre-equilibrated protein A and protein G columns, respectively. The flow-through and eluted factions from the columns were collected and analyzed by SDS-PAGE. As shown in Fig. 5, all proteins were bound to the columns, indicating mFc could bind strongly to protein A and protein G.

FIGURE 5.

The flow-through and elution of mFc.1 from protein A and protein G columns were analyzed by SDS-PAGE.

We next examined whether monomeric Fc could functionally bind FcRn. Surface plasmon resonance experiments were used to validate the pH-dependent FcRn binding and to obtain reliable binding constants. Recombinant human FcRn was immobilized on a CM5 biosensor chip followed by analysis of the interaction using a BIAcore X100 as described under “Experimental Procedures.” The binding was performed under pH 6.0 and pH 7.4, respectively, and the chip was regenerated with pH 8.0 buffer. As shown in Fig. 6, the mFcs displayed a similar behavior to that of the wild-type Fc. At pH 6.0, the calculated binding affinities (K_D) of wild-type Fc, mFc.1, mFc.23, and mFc.67 to human FcRn were 126, 204, 59, and 111 nm, respectively (Fig. 6, A–D). The binding of an Fc mutant (Fc I253A/S254A/H435A/Y436A mutant), in which four residues in the FcRn binding interface of Fc were mutated and served as a negative control in this analysis, was too weak to quantify (Fig. 6E). At pH 7.4, neither wild-type Fc nor the mFcs showed detectable binding to FcRn (Fig. 6F). These results demonstrate that the mFcs we generated maintained characteristic pH-dependent FcRn binding.

FIGURE 6.

FcRn binding of mFc.1 (A), mFc.23 (B) mFc.67 (C), Fc (D), the Fc I253A/S254A/H435A/Y436A mutant (E) at pH 6.0, and mFc.1 at pH 7.4 (F) measured by BIAcore.

Fusion Proteins with mFcs

In a final series of experiments, we explored whether our monomeric Fc can be used to produce fusion proteins. m36, an engineered human antibody VH targeting the HIV-1 envelope glycoprotein (Env) (13), was used to construct fusion proteins with a monomeric Fc, mFc.67. Two versions of fusion proteins were generated. As shown in Fig. 7A, one is m36 fused with mFc.67 (molecular mass ∼ 39 kDa), and the other is m36 fused with only the CH3 domain of mFc.67 (molecular mass ∼ 27 kDa). We found that both of these can be expressed efficiently in E. coli, with yields of ∼20 mg liter⁻¹ for m36-CH3^mFc.67 fusion protein and 5–10 mg liter⁻¹ for m36-mFc.67 fusion protein. Size exclusion chromatography results indicated that they are both in pure monomeric forms, with molecular mass of ∼27 kDa for m36-CH3^mFc.67 and ∼39 kDa for m36-mFc.67 (Fig. 7B).

FIGURE 7.

A, schematic of different antibody fragments and monomeric Fc fusion proteins. B, size exclusion chromatography of m36-CH3^mFc.67 and m36-mFc.67 fusion proteins. C, binding of m36 and fusion proteins to gp120_Bal-CD4 measured by ELISA.

We then assessed whether the effector molecules in these fusion proteins are still active. m36 exhibits potent broadly neutralizing activity against HIV-1 by targeting a highly conserved CD4 binding-induced structure on the Env gp120 (13). As shown in Fig. 7C, m36 and the two fusion proteins, m36-mFc.67 and m36-CH3^mFc.67, exhibited comparable binding to gp120_Bal-CD4, a single-chain fusion protein of gp120_Bal with soluble two-domain human CD4 (35), as measured by ELISA. Wild-type Fc and mFc.67 monomer did not show any binding ability. Taken together, these results confirmed that monomeric Fc can be used to develop functional monomeric fusion proteins.

DISCUSSION

In the present study, several mFcs were generated using a combination of rational design and multiple panning/screening methods. In nature, many proteins are in oligomeric states, and some are “compulsory complexes” in which free monomers are not available (36, 37). Although much effort has been put into disrupting such oligomers, most of the engineered monomers were insoluble or structured poorly. To address this challenge, large libraries could be constructed, and desired monomers could be selected by directed evolution. Besides, a multiple screening strategy is very valuable because it provides multiple evolutionary pressures. Support for this view is provided by the fact that the screening of our library directly against FcRn did not produce mFcs, but an additional round against protein G before FcRn screening successfully produced some highly soluble and functional monomers. We expect that our strategy could be extended to other cases of monomer development and expand the arsenal of protein engineering.

A significant application of monomeric Fc is for the so-called monomeric Fc fusion technology where a monovalency of the active protein is presented, but currently, it is fused to dimeric wild-type Fc (23–25). This “monomeric” technology is an upgraded version of the traditional dimeric Fc fusion molecules, which contain two effector molecules. It has been shown to be more robust due to better tissue penetration offered by the smaller size, and a reduced steric hindrance, which can make effector protein and/or Fc part more effective. For instance, it has been found that monomeric factor IX-Fc fusion protein not only has an extended half-life compared with the factor IX-Fc dimer but also greatly enhanced pharmacokinetics, with a 10-fold increase in C_max and more than a 12-fold increase in area under the curve (25). These advantages are valuable because they provide cheaper therapeutics and enhance the delivery of therapeutic proteins by non-invasive routes. In this study, we have shown that mFc, which is only half the size of the Fc dimer but retains FcRn binding ability, can replace dimeric Fc and generate mFc fusion proteins. Such constructs consist of only one effector molecule and one Fc monomer; thus, the size is reduced largely compared with the current monomeric Fc fusion and may lead to a new type of promising therapeutics.

Moreover, mFc represents a novel antibody format in the field of therapeutic antibodies. Currently, monoclonal antibodies have enjoyed widespread therapeutic applications and represent the largest class of biological drugs. However, antibody therapeutics have demonstrated difficulty in penetrating tissues due to their large size. A variety of small antibody formats, such as Fab, Fv, scFv, VH, and V_HH, have been developed but at the expense of their in vivo half-lives (16, 17). Thus, it is emerging that the generation of small engineered antibodies with appropriately long half-lives may lead to a new therapeutic revolution. Our work establishes the development of functional fusion proteins of antibody fragments with monomeric Fc that have molecular masses of ∼30 and 40 kDa, only one-fourth the size of IgG.

We also hypothesize that mFcs themselves could serve as novel antibody formats and be used as scaffolds for construction of libraries containing diverse binders to various antigens. They are relatively stable, can bind FcRn in a pH-dependent manner, and can be produced in large quantities in bacteria. Importantly, compared with wild-type Fc, a large surface area is exposed in the Fc monomers due to the disruption of the CH3 dimerization interface, providing more access for protein engineering by designed point mutations and CDR-grafting onto an mFc framework. It is anticipated that binders generated from such a design would have molecular masses of ∼27 kDa, similar to that of scFv but possess much longer in vivo half-lives.

Immunogenicity is a potential problem for any protein-based therapeutics. Many factors could affect immune responses elicited by therapeutic proteins such as molecular size, structural features, storage conditions, T-cell epitopes, and others (5). Although currently, immunogenicity cannot be predicted, and only human clinical trials can definitely show whether these new proteins are immunogenic, it is likely that their immunogenicity is relatively low compared with similar foreign proteins for several reasons. First, mFcs are fully human proteins, and the probability that several additional mutations will dramatically increase immunogenicity is relatively low. Second, they are smaller in molecular size than the wild-type Fc, and perhaps more importantly, they are monomeric and possible avidity effects may not exist. Third, mFcs seems to retain the structural characteristics of the wild-type Fc, as suggested by their similar CD spectra (Fig. 3B). Fourth, some of the mFcs have slightly less aromatic amino acids (which may contribute more to immunogenicity than non-aromatic) than wild-type Fc (20 aromatic amino acids in Fc and mFc.1, 18 in mFc.23, and 19 in mFc.67). Further engineering of these proteins to reduce the number of mutations compared with wild-type Fc could also contribute to a potential decrease in immunogenicity if successful.

Finally, our results provide direct evidence that mFc can bind FcRn comparably with Fc, and therefore, dimerization is not required for effective binding to FcRn. The stoichiometry of the interaction between FcRn and its ligand has been a matter of debate (20, 26–29). In the cocrystals, a long, repeating oligomeric ribbon with a 2n:n stoichiometry was observed, in which FcRn dimers are bridged by Fc molecules so that every Fc dimer interacts with two receptors (38). Studies also suggested that the 2:1 FcRn-Fc complex is formed in solution, although FcRn is not capable of dimerization in these conditions (28, 29). However, in some studies, 1:1 FcRn/Fc stoichiometry was also observed under non-equilibrium conditions (27, 29). It is not clear which of the complexes are physiologically relevant, but clearly, the stoichiometry of the FcRn-Fc interaction is essential to further our understanding of the mechanism by which FcRn transports IgG in vivo and also important to facilitate antibody engineering to modify the IgG-FcRn interactions in vitro. Our mFcs presented a unique opportunity to access the Fc-FcRn interactions; the fact that an Fc monomer is capable of binding FcRn offers direct evidence for the 2:1 FcRn/Fc stoichiometry. It is noteworthy to point out that some discrepancies among different assay systems have been reported in the FcRn binding studies (20); however, our data definitely show that mFc binds one molecule FcRn and does not require avidity effects. It also is interesting to note that mFcs have an affinity (K_D) similar to that of the dimeric Fc for FcRn binding (126 nm mFc.1, 204 nm mFc.23, 59 nm mFc.67, and 111 nm wild-type Fc). Two distinct FcRn-Fc binding mechanisms could account for this similarity. In one binding mechanism, the two FcRn molecules bind symmetrically to two Fc monomers in a dimeric Fc, and in the other, the two FcRn bind asymmetrically: the first FcRn molecule binds to an Fc monomer in a similar manner to the FcRn-mFc binding, and the binding of the second FcRn to Fc must have a low affinity, to give an overall comparable K_D to the one-step FcRn-mFc binding. The latter mechanism seems more likely because a bent, asymmetrical Fc structure was found in the 1:1 FcRn/hdFc (heterodimeric Fc that can only bind one FcRn) cocrystal (38). The low affinity of the second FcRn binding is probably due to the conformational change in Fc caused by the binding of the first FcRn. The low affinity also explains the observation of the 1:1 FcRn/Fc stoichiometry under non-equilibrium conditions.

In addition to binding FcRn to protect antibodies from degradation, the Fc also mediates effector functions such as antibody-dependent cellular cytotoxicity by binding Fcγ receptors and complement-dependent cytotoxicity by binding the complement factor C1q. The N-linked glycosylation on Asn-297 of the Fc CH2 domain plays an important role in these functions. The mFcs described in this study are expressed in E. coli and are not glycosylated. In addition, they are monomeric, and therefore, their binding to Fcγ receptors and C1q, if any, is expected to be significantly weaker compared with glycosylated wild-type dimeric Fc. In many cases, e.g. when developing binders against soluble monomeric ligands, effector functions may not be needed. In cases when such functions are needed, they could be engineered, and the proteins could be expressed in mammalian cells. It is reasonable to expect that mFc would have similar glycosylation profile as wild-type Fc because the mutations in mFc are in the CH3 domain, whereas the CH2 domain is the same.

In conclusion, we have generated several mFcs using a novel panning/screening strategy. The monomers could be expressed in E. coli with high efficiency, are relatively stable, and retain the ability to bind FcRn in a pH-dependent manner. These proteins are promising as tools to explore mechanisms of Fc interactions with FcRn and for the development of mFc fusion proteins; they could also represent robust building blocks for mFc-based antibody engineering, which could produce a new type of therapeutic antibodies with small size, long half-lives, and antigen-binding capabilities.