A Fab-Selective Immunoglobulin-Binding Domain from Streptococcal Protein G with Improved Half-Life Extension Properties

Background Half-life extension strategies have gained increasing interest to improve the pharmacokinetic and pharmacodynamic properties of protein therapeutics. Recently, we established an immunoglobulin-binding domain (IgBD) from streptococcal protein G (SpGC3) as module for half-life extension. SpGC3 is capable of binding to the Fc region as well as the CH1 domain of Fab arms under neutral and acidic conditions. Methodology/Principal Findings Using site-directed mutagenesis, we generated a Fab-selective mutant (SpGC3Fab) to avoid possible interference with the FcRn-mediated recycling process and improved its affinity for mouse and human IgG by site-directed mutagenesis and phage display selections. In mice, this affinity-improved mutant (SpGC3FabRR) conferred prolonged plasma half-lives compared with SpGC3Fab when fused to small recombinant antibody fragments, such as single-chain Fv (scFv) and bispecific single-chain diabody (scDb). Hence, the SpGC3FabRR domain seems to be a suitable fusion partner for the half-life extension of small recombinant therapeutics. Conclusions/Significance The half-life extension properties of SpGC3 can be retained by restricting binding to the Fab fragment of serum immunoglobulins and can be improved by increasing binding activity. The modified SpGC3 module should be suitable to extend the half-life of therapeutic proteins and, thus to improve therapeutic activity.


Introduction
With more and more small recombinant protein therapeutics advancing into the clinic, a major task in protein engineering is to overcome the limitation of the rapid clearance from a patient's bloodstream and to generate efficient and long-lasting pharmaceuticals, e.g. by implementation of half-life extension strategies [1][2][3]. One approach is to increase the hydrodynamic radius in order to prevent the protein from rapid renal filtration. Measures to circumvent this are for example coupling of hydrophilic polymers such as polyethylene glycol (PEG) and hydroxyethyl starch, and genetic engineering to introduce (additional) glycosylation sites or hydrophilic and flexible polypeptide chains [4][5][6]. Furthermore, immunoglobulins and also albumin are serum proteins that show the longest circulation time at all and are therefore interesting fusion or interaction partners [7,8]. Their astonishing long circulation time is mediated by the binding, after intracellular uptake, to the neonatal Fc receptor (FcRn) under acidic conditions in a recycling endosome [9][10][11]. Hence, the proteins are transferred back to the cell surface and upon reaching neutral pH released into the bloodstream, thus being salvaged from lysosomal degradation. Fusion of the Fc part or albumin is known to elongate the half-life of therapeutic proteins, but also transient interactions through IgG-and albumin-binding peptides and protein domains are able to mediate a long circulation time [1,12,13].
Recently, we found that amongst several tested immunoglobulin-binding domains (IgBD) from staphylococcal protein A (SpA) and streptococcal protein G (SpG), the domain C3 of protein G (SpG C3 ) had the best properties in improving the pharmacokinetic profile of a bispecific single-chain diabody (scDb) fusion protein [13,14]. This domain, composed of 56 amino acids, has binding sites for the Fc and the Fab fragment of long-circulating IgG molecules. Its primary binding site is formed by the CH2 and CH3 domain of the Fc part, overlapping with the binding site of the neonatal Fc receptor [15,16]. Distinct from this site, the second binding site is located on the CH1 domain of the Fab arms [17]. The SpG C3 domain was able to extend the plasma half-life of the 53 kDa large scDb approximately 18-fold due to binding to serum immunoglobulins after intravenous injection, primarily enhancing the hydrodynamic radius of the complex and thus preventing rapid renal clearance. Of note, the domain was able to bind to IgG also at acidic conditions [14]. Thus, in addition to reduced renal clearance, IgBDs may also be co-recycled via the FcRn as cargo molecule in an acidified endosome.
To investigate the effects of the Fab binding site of the SpG C3 domain on half-life extension, we generated a variant of the SpG C3 domain (SpG C3Fab ) with strongly reduced Fc binding activity. This variant was characterized for its binding properties in vitro and their capacity to extend the half-life of recombinant antibody molecules. Applying phage display selection of SpG C3Fab libraries, we further identified a variant with improved binding and half-life extension properties.

Generation of a Fab-selective SpG-C3
The three immunoglobulin-binding domains (IgBD) of Streptococcus protein G (SpG C1 , SpG C2 , SpG C3 ) possess two distinct binding sites on IgG molecules. One site is located in the CH1 domain of the Fab fragment and one site in a region formed by the CH2 and CH3 of the Fc fragment. The latter site overlaps with the binding site of the neonatal Fc receptor (FcRn) [15] and could, therefore, potentially interfere with FcRn-mediated recycling. Hence, in a first approach we attempted to eliminate this binding site in SpG C3 to restrict binding to the Fab fragment. Three residues in the helical structure of SpG C3 (aa Glu27, Lys28, Lys31; numbering as in Fig 1) involved in binding to the Fc region, as deduced from the crystal structure of the homologous domain C2 bound to the Fc region of human IgG [18], were substituted by alanines (SpG C3Fab ) (Fig 1). SpG C3 and SpG C3Fab were fused C-terminally to a 55 kDa bispecific single-chain diabody directed against CEA and CD3 endowed with a hexahistidyl-tag at the very C-terminus. Fusion proteins were produced in stably transfected HEK293T cells and purified from the cell culture supernatant by immobilized metal affinity chromatography. Protein purity was confirmed by SDS-PAGE analysis and by size exclusion chromatography (Fig  2a and 2b). All proteins showed single bands corresponding to their calculated molecular mass of 60 kDa (Table 1) with hydrodynamic radii between 3.0 and 3.2 nm.
Binding Properties of Fab-selective SpG C3Fab fused to a single-chain Diabody IgBD-mediated binding of the fusion proteins to immobilized IgG and fragments thereof was investigated by quartz crystal microbalance (Fig 3). Binding curves had to be fitted assuming biphasic kinetics, presumably due to heterogeneity of the immunoglobulin preparations obtained from human and mouse serum, respectively. Unmodified scDb-SpG C3 showed strong binding to human and mouse IgG as well as their Fab and Fc fragments, both at neutral pH  [17,19]. Positions randomized in the phage library are indicated by orange triangles (residues 33, 35, 36, and 37). b) Structure of SpG C3 (pdb entry 1IGC [17]). The Fc binding site is shown in red/yellow, the major Fab binding site in blue, and the minor Fab binding site in yellow, overlapping with the Fc binding site. Residues mutated to alanines are indicated as red spheres, positions mutated in the phage library are highlighted as orange spheres. c) Superposition of SpGC3 bound to an Fc fragment (pdb entry 1FCC) and a Fab fragment (1IGC) (variable domains are not shown). Residues are marked as in b).
doi:10.1371/journal.pone.0139838.g001  (pH 7.4) and acidic pH (pH 6) ( Table 2). In contrast, for scDb-SpG C3Fab binding was only observed for human and mouse IgG and the Fab fragments, while no or only negligible binding was detected for the Fc fragments at neutral pH (Fig 3). Importantly, compared to scDb-SpG C3 , the affinity of the SpG C3Fab domain towards the Fab fragment was preserved or even slightly improved at neutral and at acidic pH for both, human and mouse IgG, while a weaker binding of scDb-SpG C3Fab to full-length IgG was observed at neutral pH (Table 2). Interestingly, at acidic pH SpG C3Fab was able to bind to the Fc fragment to a certain extent, especially mouse Fc. These experiments established that the introduced alanine substitutions at positions Glu27, Lys28, and Lys31 specifically deleted the Fc binding site in scDb-SpG C3Fab , at least at neutral pH, without affecting Fab binding. Next, a functional assay was performed to analyze the effector cell recruiting and activating activity of the scDb moiety capable of crosslinking tumor cells and T-cells, thereby leading to  the activation of the effector T-cells. Adherent CEA-expressing LS174T cells were incubated with human PBMCs as effector cells together with the scDb fusion protein and activation of the effector T cells was determined by quantification of induced IL-2 release. All scDb-fusion proteins were able to activate the IL-2 secretion of PBMCs in a dose-dependent manner, reaching the maximal stimulating concentration at 3 to 10 nM in accordance with previous findings [12] (Fig 4a-4c). In the absence of human IgG, all scDb fusion proteins behaved as the scDb control. However, in the presence of IgG, scDb-SpG C3 exhibited strongly reduced activation levels. In contrast, scDb-SpG C3Fab -mediated IL-2 release was not affected in the presence of IgG with levels of IL-2 that were similar to the scDb control. Binding of scDb-SpG C3 and scDb-SpG C3Fab fusion proteins to CEA-expressing LS174T cells and isolated PBMCs analyzed by flow cytometry was not or only marginally affected in the presence of IgG, indicating that the reduced activation of effector cells by scDb-SpG C3 is not caused by reduced target and effector cell binding (data not shown).

Pharmacokinetics of Fab-selective SpG C3Fab fused to a single-chain Diabody
The pharmacokinetic properties of scDb-SpG C3 and scDb-SpG C3Fab were investigated in female CD1 mice after intravenous injection of the fusion proteins. The serum concentration was determined by ELISA to detect CEA-binding molecules. The unmodified scDb showed rapid clearance from the bloodstream with a terminal half-life of 1.3 h ( Table 1, Fig 5). Compared to the unmodified scDb, the scDb-SpG C3Fab fusion protein exhibited a strongly augmented terminal half-life (15.1 h), which was also reflected by a strongly increased AUC (Table 1). However, scDb-SpG C3Fab did not reach the half-life (24.3 h) of scDb-SpG C3 .

Affinity maturation of SpG C3Fab
The reduced half-life of scDb-SpG C3Fab compared to scDb-SpG C3 might be caused by a reduced binding to serum IgG (see Table 2). In order to improve the binding to the CH1 domain of the IgG Fab arms, a phage display library of mutated SpG C3Fab variants was generated. By site-directed mutagenesis, positions 33, 35, 36 and 37, described to be involved in Fab binding [17], were randomized (Fig 1a). After 3 rounds of selection against biotinylated soluble mouse Fab fragments, a polyclonal phage ELISA revealed an enhanced binding of the enriched phage to human and mouse IgG and Fab fragments (data not shown). One clone with two mutations at positions 35 and 36 (N35R, D36R) and wild-type residues at positions 33 and 37 (SpG C3Fab RR) showed the best binding properties and was subjected to further analysis as scDb-SpG C3Fab RR fusion protein. QCM measurements demonstrated an increased binding to mouse Fab fragment, especially under acidic pH reflected also by a slightly improved affinity towards mouse IgG (Table 2). Similarly, the affinity for human IgG was increased, which was also seen for human Fab fragments at acidic pH, although only marginal differences were determined for binding to human Fab fragments at neutral pH. SpG C3Fab RR did not bind to human and mouse Fc fragments at neutral pH, but retained binding to human and mouse Fc fragments at acidic pH, as observed for SpG C3Fab . Similar to scDb-SpG C3Fab , scDb-SpG C3Fab RR showed strong activation of PBMCs in the recruitment experiment, and this activity was not affected in the presence of excess amounts of IgG (Fig 4d).

Pharmacokinetics of affinity matured SpG C3Fab fused to single-chain Diabody and an scFv
Finally, we determined the pharmacokinetic properties of scDb-SpG C3Fab RR in CD1 mice.
Here, we included also an anti-CEA scFv with only half the size of the scDb as fusion partner for the various SpG C3 variants. In addition, we tested also SpG C2 , which differs at 4 positions from SpG C3 , fused to scDb.

Discussion
Immunoglobulins of the IgG class and albumin exhibit exceptionally long serum half-lives and have therefore gained increasing interest for half-life extension strategies [1]. For example, fusion of moieties with affinity for IgG or albumin to therapeutic proteins allows binding to the serum protein in vivo resulting in an extended half-life. While various albumin-binding moieties have been extensively studied for half-life extension, little efforts have been made to use immunoglobulin-binding moieties for this purpose [13,14,[19][20][21][22][23]. Several factors might influence the capacity of these immunoglobulin-binding moieties to extend the serum half-life of fusion partners [14]. Amongst those are the selectivity for long-circulating antibody classes and subclasses (especially IgG1, IgG2, IgG4) [24,25], the stability of the complexes in the blood stream (i.e. at neutral pH) and after internalization and during endosomal recycling (i.e. at acidic pH), and the interference with FcRn binding under acidic conditions. Furthermore, the size of the IgBD fusion protein itself might affect the pharmacokinetic properties. The SpG C3 domain can bind to the Fc and Fab fragments of immunoglobulins reflecting the binding specificities of protein G [26], with a preference for IgG [27]. In a previous study, we demonstrated already that SpG C3 binds stronger to the Fc fragment than to the Fab fragment [14], which was confirmed here using Fc and Fab fragments from human and mouse serum IgG. By mutating three key residues of the Fc binding site of SpG C3 to alanines we were able to generate variants with selectivity for Fab fragments. However, the half-life of an scDb-SpG C3Fab fusion protein was markedly reduced compared to the corresponding SpG C3 fusion protein, which was also seen for an scFv-SpG C3Fab fusion protein. Presumably, this is caused by a reduced affinity for serum IgG of SpG C3Fab fusion protein. For that reason, we increased the affinity towards the whole IgG and in particular the Fab fragment by phage display, resulting in the SpG C3Fab RR domain.
Using this affinity-improved domain, we were able to increase the terminal half-life of the scDb fusion protein around 40-fold. Of note, the half-life of scDb-SpG C3Fab RR (47.8 h) is much longer compared to an IgG that was tested in FcRn knockout mice (19 h compared with 95 h in wild-type mice), indicating a contribution of FcRn-mediated recycling [28]. Mechanistically, it is essential for the IgBDs to retain the binding affinity towards the IgG molecule and especially the Fab fragment under neutral and acidic conditions to avoid dissociation of the bound IgBD fusion protein from the recycling IgG molecule. Both, SpG C3Fab and SpG C3Fab RR fulfill this requirement, suggesting that the complexes with the Fab arms are stable also during the endosomal recycling process. The SpG C3Fab RR exhibits a better affinity towards the Fab fragment than the SpG C3Fab domain, which should stabilize the complex in the neutral blood as well as during the recycling in the FcRn. It was already described that the SpG C3 domain changes its affinity towards the Fc part depending on the pH, probably due to the protonation of histidine residues in the Fc part that are in close proximity to the bound SpG C3 domain [18,29,30]. Our data suggest that despite SpGC3FabRR having an increased affinity for the Fcdomain at pH 6 compared to SpGC3Fab, which could inhibit IgG recycling by the FcRn, the half-life of SpGC3FabRR is considerably longer than SpGC3Fab, presumably due to its higher affinity binding to Fab.
Our studies also showed that the size of the IgBD fusion proteins influences the pharmacokinetic properties. Compared with scDb-IgBD fusion proteins, the scFv-IgBD fusion proteins with only half the size exhibited a reduced half-life and AUC. Dissociation of the bound fusion proteins from immunoglobulins in serum and subsequent renal clearance might be responsible. Further affinity maturation of SpG C3Fab RR might therefore be beneficial to further extend the half-lives of small size molecules.
SpG C3 when fused to the bispecific scDb, designed to retarget effector T cells to tumor cells, reduced the effector cell stimulating activities mediated by binding to CD3 in the presence of IgG. Of note, SpG C3Fab and SpG C3Fab RR did not show this reduction. Similar findings have previously been observed for various other half-life extension strategies, e.g. PEGylation and fusion to an albumin-binding domain [5,12]. Our findings indicate that in this application, where effector cells have to come in close contact with the target cells, differences in binding activity of the IgBD have strong effects on the T-cell retargeting and triggering activity. Thus, SpG C3Fab and SpG C3Fab RR might be especially useful for the half-life extension of small bispecific T cell engagers [31,32].
In summary, we were able to generate new immunoglobulin-binding domains that bind specifically to the Fab fragment of IgG molecules. The reduced half-life extension properties of SpG C3Fab were compensated by mutations leading to improved affinity. Hence, the SpG C3Fab RR domain seems to be a suitable fusion partner for the half-life extension of small recombinant therapeutics.

Materials and Methods Materials
Horseradish peroxidase-conjugated (HRP) anti-His-tag antibody was purchased from Santa Cruz Biotechnology (Santa Cruz, CA, USA). Carcinoembryonic antigen (CEA) was obtained from Antikörper-online.de (Aachen, Germany). Human and mouse serum IgG were purchased from Sigma, Fab and Fc fragments from Emelca Bioscience (Breda, Netherlands). The human colon adenocarcinoma cell line LS174T was purchased from ECACC (Wiltshire, UK) and cultured in RPMI, 5% fetal bovine serum (FBS), 2 mM glutamine (Invitrogen, Karlsruhe, Germany). Buffy coats from anonymous healthy human donors were obtained from blood donation center of the Klinikum Stuttgart (Stuttgart, Germany; http://www.klinikum-stuttgart. de/). The DuoSet IL-2 ELISA Development System kit was purchased from R&D Systems (Nordenstadt, Germany). CD1 mice were purchased from Charles River (Sulzfeld, Germany).

Cloning and production
The DNA encoding the SpG C3Fab sequence, including the His-tag, was purchased from Geneart (Regensburg, Germany) with a NotI restriction site at the 5' end and an EcoRI site at the 3' end to exchange the SpG C3 domain from the previously used vector pSecTagA-scDbCEACD3-SpG C3 or pSecTagA-scFvCEA-SpG C3 [14]. For phage display, degenerate primers were used for PCR with a SalI restriction site at the 5' end and a NotI site at the 3' end. NNK triplets were used at the sites that encode for amino acids binding to the CH1 domain (forward primer: 5'-TCG AGT CGA CGC CGA AAC AGC CGC CGC TGC CTT TGC CCA GNN KGC CNN KNN KNN KGG CGT GGA CGG CGT GTG GAC C-3', reverse primer: 5'-ATT TGC GGC CGC CTC GGT CAC GGT GAA GGT-3'). The digested PCR product was ligated into pHEN3-SpG C3Fab ΔN (SalI), a modified version of SpG C3Fab with a stop codon at the 5' region generated via cloning into the phagemid vector pHEN3 (based on pHENIX [33]), with an additional His-tag via SalI and NotI. The insert was created by PCR (Forward primer: 5'-TCGAGTCGACAGTACGCCAACGATAATGGCG-3' and reverse primer from above). The ligated library was transformed into E.coli TG1 by electroporation. The SpG C3Fab RR domain was amplified by PCR from the pHEN3 vector and then cloned into pSecTagA-scDb-SpG C3Fab with NotI and BamHI (Forward primer: 5'-ATT TGC GGC CGC AGG CGG ATC TGG CGG AAC AAC ATA CAA GCT CGT G-3', reverse primer: 5'-GAT CGG ATC CGC CCT CGG TCA CGG TGA AGG T-3'). All pSecTag plasmids were stably transfected into HEK293T cells and the resulting fusion proteins were purified from the supernatant by immobilized metal affinity chromatography as described previously [19].

Phage display
Purified murine Fab fragment was biotinylated with EZ-link™Sulfo-NHS-SS-Biotin from Thermo Scientific (Rockford, USA) according to manufacturer's protocol. Phages were rescued from library stock or preselected rounds with VSC-M13 helper phages. In the presence of 1% bovine serum albumin (BSA) as blocking reagent in PBS, the phages were incubated with biotinylated Fab fragment (100 nM to 1 nM over 3 rounds) for 1 h at room temperature and then added to streptavidin-coupled M-280 Dynabeads (Thermo Scientific, Rockford, USA) for 15 minutes to allow the complex to bind to the beads. A magnetic separator was used to separate the beads. After washing 4 times with PBST (PBS, 0.1% Tween 20) and twice with PBS, bound phages were eluted with 10 mM dithiothreitol for 5 min and subsequently inoculated to logphase E. coli TG1 for further selection rounds. Starting from the second round, washing was performed at pH 6.0.

Size exclusion chromatography
Purity and Stokes radii of the recombinant fusion proteins were determined by HPLC size exclusion chromatography with a Yarra SEC-2000 column (Phenomenex, Aschaffenburg, Germany). 20 μl of proteins (0.5 mg/ml) were injected at a flow rate of 0.5 ml/min. The following standard proteins were used to calculate the apparent molecular mass and hydrodynamic radius: thyroglobulin (669 kDa), apoferritin (443 kDa), alcohol dehydrogenase (150 kDa), BSA (67 kDa), carbonic anhydrase (29 kDa) and FLAG peptide (1 kDa). ELISA CEA (3 μg/ml) or immunoglobulins (1 μg/ml) were immobilized on ELISA plates over night at 4°C and remaining binding sites were blocked with 2% (w/v) non-fat dry milk/phosphate buffered saline (MPBS). Purified recombinant proteins, serum samples or phage display supernatants were diluted in MPBS, titrated in duplicates and incubated for 1 h at room temperature. Detection was performed with HRP conjugated anti-His-tag antibody or anti-M13 antibody for the phage ELISA using 100 μl 3,3',5,5'-tetramethylbenzidine (TMB) substrate (0.1 mg/ml TMB, 100 mM sodium acetate buffer pH 6.0, 0.006% H 2 O 2 ). The reaction was stopped with 50 μl of 1 M H 2 SO 4 and absorbance was measured at 450 nm. Data were fitted with GraphPrism software (La Jolla, CA, USA) from three independent binding curves. From these three individual EC 50 values the mean and standard deviation were calculated.

Flow cytometry
Binding to CEA-expressing LS174T cells and IL-2-stimulated T cells from peripheral blood mononuclear cells (PBMCs) was determined by flow cytometry [34]. 2.5 x 10 5 cells were incubated with dilution series of antibodies for 2 h at 4°C. Cells were then washed with PBS, 2% FBS, 0.02% NaN 3 (PBA) and bound antibodies were detected with PE-conjugated mouse anti-His-tag antibody.

Interleukin-2 release assay
Peripheral blood mononuclear cells (PBMCs) from healthy donors were isolated from buffy coat as described before [19]. 10 5 LS174T cells/100 μl/well were seeded in 96-well plates. The next day, supernatant was removed and 150 μl of recombinant antibody was added. After 1 h preincubation at 37°C, 2 x 10 5 PBMC/50 μl/well were added. PBMCs had been thawed the day before and seeded on a culture dish. Only cells that remained in suspension were used for the assay. After addition of PBMCs, the 96-well plate was incubated for 24 h at 37°C, 5% CO 2 . All incubations were performed in RPMI supplemented with 10% FBS. Plates were centrifuged and cell-free supernatant was collected. Concentration of human interleukin (IL)-2 in the supernatant was determined using the DuoSet IL-2 ELISA kit (R&D Systems) following the manufacturer's protocol.

Affinity determination
Affinities of IgBD fusion proteins for human and mouse serum IgG as well as Fab and Fc fragments thereof at neutral or acidic pH were determined by quartz crystal microbalance (QCM) measurements (A200 C-Fast system; Attana, Stockholm, Sweden). The ligands were chemically immobilized on a low nonspecific binding carboxyl sensor chip according to the manufacturer's protocol at a density resulting in a signal increase of 70-90 Hz. Binding experiments were performed in PBST (0.1% Tween 20), pH 7.4 or pH 6.0, at a flow rate of 25 μl/min. The chip was regenerated twice with 10 mM glycine-HCl, pH 3.0 for 30 seconds. Before each measurement, a base line was measured which was subtracted from the binding curve. Proteins were injected in a random order in two-fold dilutions, starting at a concentration of 1 μM. Data were analyzed with the Attana evaluation software (Version 3.3.4) and TraceDrawer 1.6, using a one-to-two binding model to determine the binding parameters.

Pharmacokinetics
Animal care and all experiments performed were in accordance with federal and european guidelines and have been approved by Regierungspräsidium Stuttgart. Female CD1 mice (6-12 weeks, 25-35 g) received an intravenous injection of 25 μg of the IgBD fusion proteins in a total volume of 150 μl. In time intervals of 3 min, 30 min, 1 h, 2 h, 6 h, 1 day, 3 days, and 7 days, blood samples (50 μl) were taken from the tail and incubated on ice. Clotted blood was centrifuged at 13,000 g for 30 min at 4°C, and serum samples were stored at -20°C. Serum concentrations of CEA-binding recombinant antibodies were determined by ELISA (as described above). For comparison, the first value (3 min) was set to 100%. Half-lives (t 1⁄2 α, t 1⁄2 β) and area-under-the-curve (AUC) were calculated with Excel. Terminal half-lives (t 1⁄2 β) were calculated using the last three or four serum concentrations shown in Fig 5. For statistics, One-way-Anova was applied with Tukey's post-test. Results are shown as mean value ± standard deviation.