N-Glycosylation as Novel Strategy to Improve Pharmacokinetic Properties of Bispecific Single-chain Diabodies*

The therapeutic efficacy of recombinant antibodies such as single-chain Fv fragments and small bispecific or bifunctional molecules is often limited by rapid elimination from the circulation because of their small size. Here, we have investigated the effects of N-glycosylation on the activity and pharmacokinetics of a small bispecific single-chain diabody (scDb CEACD3) developed for the retargeting of cytotoxic T cells to CEA-expressing tumor cells. We could show that the introduction of N-glycosylation sequons into the flanking linker and a C-terminal extension results in the production of N-glycosylated molecules after expression in transfected HEK293 cells. N-Glycosylated scDb variants possessing 3, 6, or 9 N-glycosylation sites, respectively, retained antigen binding activity and bispecificity for target and effector cells as shown in a target cell-dependent IL-2 release assay, although activity was reduced ∼3–5-fold compared with the unmodified scDb. All N-glycosylated scDb variants exhibited a prolonged circulation time compared with scDb, leading to a 2–3-fold increase of the area under curve (AUC). In comparison, conjugation of a branched 40-kDa PEG chain increased AUC by a factor of 10.6, while a chimeric anti-CEA IgG1 molecule had the longest circulation time with a 17-fold increase in AUC. Thus, N-glycosylation complements the repertoire of strategies to modulate pharmacokinetics of small recombinant antibody molecules by an approach that moderately prolongs circulation time.

sites in total) with a ϳ2-fold prolonged half-life and an improved in vivo activity (19). In other studies, glycosylated analogs of follicle stimulating hormone (FSH) with prolonged circulation time and increased activity have been generated by adding N-glycans through an N-terminal extension of the ␣ subunit (20) or through a peptide linker joining the ␤and the ␣-subunit (21), demonstrating the feasibility of this approach.
As yet, no studies have been described that successfully applied N-glycosylation to prolong the circulation time of small recombinant antibodies. Here, we performed a comparative analysis of N-glycosylated bispecific single-chain diabody (scDb) molecules with other modifications including PEGylation. N-Glycosylation at 3, 6, or 9 Asn-X-Thr sites introduced at the flanking linker sequences and C-terminal extensions prolonged half-life ϳ2-fold leading to a 2-3-fold increase of the AUC. In contrast, site-directed chemical coupling of a branched PEG molecule of 40 kDa to a cysteine residue introduced in one of the linker sequences prolonged circulation time ϳ5-fold resulting in a 11-fold increase of the AUC, similar to fusion of the scDb molecule with human serum albumin or a albumin binding domain of streptococcal protein G.

EXPERIMENTAL PROCEDURES
Materials-Horseradish peroxidase-conjugated anti-His tag antibody was purchased from Santa Cruz Biotechnology, unconjugated anti-His tag antibody from Dianova (Hamburg, Germany) and anti-mouse IgG-FITC or PE-conjugated antibody as well as goat anti-rabbit IgG-FITC-conjugated antibody from Sigma (Taufkirchen, Germany). Rabbit antiserum recognizing PEGylated scDb was produced by Pineda Antikörper-Service (Berlin, Germany). Horseradish peroxidase-conjugated anti-rabbit IgG was purchased from Roche Applied Sciences (Mannheim, Germany). Carcinoembryonic antigen was obtained from Europa Bioproducts (Cambridge, UK). Branched mPEG 2 -Mal was purchased from Nectar Therapeutics (Birmingham, AL). The human colon adenocarcinoma cell line LS174T was purchased from ECACC (Wiltshire, UK) and cultured in RPMI 1640 supplemented with 5% FBS. HEK293 were cultured in RPMI 1640, 5% FBS. Buffy coat from a healthy human donor was kindly provided by Prof. G. Multhoff (München, Germany). IL-2 was purchased from Immunotools (Friesoythe, Germany) and phytohemagglutinin-L (PHA-L) from Roche Applied Sciences (Germany). CD1 mice were purchased from Elevage Janvier (Le Genest St. Isle, France).
Construction of scDb Molecules-scDbCEACD3 (4) was used as starting material for the construction of scDb variants. Cysteine residues were introduced into linker A or the C terminus by PCR (scDb-AЈ, scDb-CЈ). Similarly, N-glycosylation sites were introduced into the linkers A and B as well as at the C terminus (scDbABC 1 , scDbABC 4 , scDbABC 7 ) (see also Fig. 1). The N-glycosylation sites for scDbABC 4 and scDbABC 7 were first optimized using the NetNglyc-Software, 4 which allowed an in silico prediction of N-glycosylation efficiency. All constructs were cloned into pSecTagA (Invitrogen, Karlsruhe, Germany) via SfiI/EcoRI for eukaryotic expression.
Protein Expression and Purification-Plasmid DNA encoding the respective recombinant antibody was transfected with Lipofectamine TM 2000 (Invitrogen) into HEK293 cells. Stable transfectants were generated by selection with zeocin (300 g/ml). For protein production, cells were first expanded and grown in RPMI 1640, 5% FBS to 90% confluence and subsequently cultured in Opti-MEMI (Invitrogen) replacing media every 3 days 3-4 times. Supernatants were pooled, and proteins were concentrated by ammonium sulfate precipitation (60% saturation), before loading onto a Ni-nitrilotriacetic acid column (Qiagen, Hilden, Germany). Purification by immobilized metal ion affinity chromatography (IMAC) was performed as described elsewhere (4).
Chimeric Anti-CEA IgG1-The regions encoding the human ␥1 heavy chain constant region and the human C domain were amplified from cDNA generated from RNA isolated from human PBMCs and combined with DNA encoding the V H and V L domain of anti-CEA antibody MFE23 (24) cloned previously into vector pSecTagA. DNA encoding the complete heavy and light chain was then cloned into pEE6.4 or pEE12.4, respectively (Lonza Biologics, Berkshire, UK). The two plasmids were then combined according to the manufacturer's protocol resulting in a bicistronic expression plasmid (pEE12.4-IgG1-CEA). The linearized plasmid was transfected into CHO-K1 cells and stably transfected clones were selected in glutamine-free RPMI 1640 medium (Invitrogen) in the presence of 25 M methionine sulfoximine (MSX). IgG1 was purified from cell culture supernatant by protein A chromatography. In brief, proteins were precipitated from supernatant with 60% saturated ammonium sulfate, resuspended in PBS, pH 8 and loaded onto a protein A-Sepharose CL-4B column (Sigma). Bound protein was eluted with 100 mM glycine, pH 3.0 and protein-containing fractions were dialyzed against PBS.
ELISA-Carcinoembryonic antigen (CEA) (300 ng/well) was coated overnight at 4°C. After 2 h blocking with 2% (w/v) dry milk/PBS, recombinant antibody fragments were titrated in duplicates and incubated for 1 h at room temperature. Detection was performed either with mouse horseradish peroxidaseconjugated anti-His tag antibody or rabbit antiserum and horseradish peroxidase-conjugated goat anti-rabbit antibody using TMB substrate (1 mg/ml TMB, sodium acetate buffer, pH 6.0, 0.006% H 2 O 2 ). The reaction was stopped with 50 l of 1 M H 2 SO 4 . Absorbance was measured at 450 nm in an ELISA reader.
Flow Cytometry-5 ϫ 10 5 cells/well were incubated with recombinant antibodies (10 g/ml) for 2 h at 4°C. After washing, cells were incubated for 1 h at 4°C with mouse anti-His tag antibody or rabbit antiserum followed by washing and 30 min of incubation with PE-labeled anti-mouse IgG or FITC-conjugated goat anti-rabbit antibody. Wash cycles and incubation steps were performed with PBS, 2% FBS, 0.02% azide. Finally, cells were analyzed by flow cytometry using an EPICS XL-MCL (Beckman Coulter, Krefeld, Germany).
Carbohydrate Analysis-Glycoproteins (10 -30 g) were denatured in the presence of SDS and 2-mercaptoethanol prior to addition of Nonidet P-40 and 100 units of PNGaseF (NEB Biolabs) according to the supplier's instructions. After incubation for 18 h the sample was desalted on graphitized carbon columns (Carbograph, Alltech) (25). The samples were dried by vacuum centrifugation and in a desiccator in the presence of P 2 O 5 /KOH. To the dry sample, 50 l of base (NaOH/Me 2 SO) was added and incubated for 30 min at room temperature with occasional shaking. Finally, an aliquot of 25 l of methyl iodide was added to the reaction mixture followed by incubation for another 30 min at room temperature. After neutralization with dilute acetic acid the methylated glycans were extracted with chloroform-water. The chloroform phase was dried under nitrogen and the glycans solubilized in methanol. Matrix-assisted laser desorption ionization mass spectrometry (MALDI-MS) was performed on a Bruker Reflex IV instrument (Bruker Daltonics, Bremen, Germany). The methylated glycan samples (ϳ500 pmol per l) in methanol were applied to the stainless steel target by mixing a 0.5-l aliquot with 1.0 l of matrix (saturated solution of 2,5-dihydroxybenzoic acid in acetonitrile/0.1% trifluoroacetic acid, 1:2). Analyses were performed by positive ion detection in the reflectron mode as described previously (26).
PBMCs had been thawed the day before and seeded in RPMI, 10% FBS on a culture dish, to remove monocytes by attachment to the plastic surface. Only cells that remained in suspension were used for the assay. After addition of PBMCs, the 96-wellplate was incubated for 24 h at 37°C, 5% CO 2 . Plates were centrifuged, and the cell-free supernatant collected. The concentration of human IL-2 in the supernatant after T-cell retargeting was determined by an IL-2-Sandwich-ELISA. Antihuman IL-2 antibodies as well as the standard of recombinant human IL-2 was provided by DuoSet IL-2 ELISA Development System kit (R&D Systems, Nordenstadt, Germany), and the assay was performed following the manufacturer's protocol.
Pharmacokinetics-Animal care and all experiments performed were in accordance with federal guidelines and have been approved by university and state authorities. CD1 mice (female, 18 -30 weeks, weight between 31-44 g, 3-6 mice/ group) received an i.v. injection of 25 g of recombinant antibody in a total volume of 200 l. In time intervals of 3, 10, 30, 60, 120, 360 min and 24-h blood samples (100 l) were taken from the tail and incubated on ice. Clotted blood was centrifuged at 10,000 ϫ g for 10 min, 4°C and serum samples stored at Ϫ20°C. Serum concentrations of CEA-binding recombinant antibodies were determined by ELISA (as described above), interpolating the corresponding calibration curves. For comparison, the first value (3 min) was set to 100%. Pharmacokinetic parameters AUC, t1 ⁄ 2␣ and t1 ⁄ 2␤ were calculated with Excel using the first 3 times points to calculate t1 ⁄ 2␣ and the last three time points to calculate t1 ⁄ 2␤. For statistics, the Student's t test was applied.

RESULTS
N-Glycosylated scDb Molecules-To generate N-glycosylated scDb molecules 2 N-glycosylation sequons (NGT) were introduced in the flanking linkers (A, B) and 1 to 7 sequons at C-terminal extensions (Fig. 1). Thus, scDb-ABC 1 contains 3 sequons, scDb-ABC 4 6 sequons, and scDb-ABC 7 9 sequons. These modified scDb molecules were expressed in stably transfected HEK293 cells and purified by IMAC with yields of 8 -17 mg/liter supernatant. SDS-PAGE analysis revealed reduced mobility of these three molecules compared with unmodified scDb depending on the number of sequons ( Fig. 2a and Table  1). N-Glycosylation of these proteins was confirmed by enzymatic deglycosylation with PNGase F, which reduced the apparent molecular mass to that calculated for the unmodified protein (Fig. 2a). Some heterogeneity in the degree of N-glycosylation was observed. In size exclusion chromatography (SEC), proteins eluted with a major peak corresponding in size to apparent molecular masses of 36 kDa for unmodified scDb, 38 kDa for scDb-ABC 1 , 65 kDa for scDb-ABC 4 , and 78 kDa for scDb-ABC 7 ( Fig. 2b and Table 1). As expected from SDS-PAGE analysis scDb-ABC 4 and scDb-ABC 7 eluted over a broad range, further indicating size heterogeneity of these proteins. Binding to CEA was not or only marginally impaired by N-glycosylation as shown in ELISA with immobilized CEA and flow cytometry with CEA-expressing cell line LS174T as well as CD3-positive PBMCs (Fig. 2, c and d). MALDI-MS further revealed the structural heterogeneity of attached N-glycans, which were of the bi-, tri-, and tetra-antennary complex type (Table 2). No highmannose or hybrid N-glycans were detected. The degree of sia-lylation of the N-glycans was found to be 34% for scDb-ABC 1 , 23% for scDb-ABC 4 , and 22% for scDb-ABC 7 as deduced from a semi-quantitative analysis of the MALDI-MS data.    MARCH 21, 2008 • VOLUME 283 • NUMBER 12

JOURNAL OF BIOLOGICAL CHEMISTRY 7807
PEGylated scDb Molecules-For site-directed PEGylation we generated 2 scDb variants having an additional cysteine residue either in linker A (scDb-AЈ) or at the C terminus (scDb-CЈ) (Fig.  1). Both variants were expressed in stably transfected HEK293 cells and purified by IMAC with yields of 8 and 10 mg/liters supernatant, respectively. SDS-PAGE analysis showed under reducing conditions a single band with an apparent molecular mass of 59 kDa and under non-reducing conditions two bands corresponding to monomeric protein (apparent molecular mass 57 kDa) and disulfide-linked proteins (apparent molecular mass of 170 kDa for scDb-AЈ and 112 kDa for scDb-CЈ) (Fig.  3a). Both proteins were PEGylated by chemical conjugation of a branched 40-kDa mPEG chain (mPEG 2 -Mal) to reduced scDb-AЈ and scDb-CЈ (Fig. 3b). PEGylation resulted in an increase of the molecular mass to ϳ230 kDa as shown by SDS-PAGE analysis. Coupling efficiency were 94% for scDb-AЈ and 79% for scDb-C' as determined by SEC (Fig. 3c). Both proteins eluted with a major peak corresponding to an apparent molecular mass of ϳ650 kDa and a hydrodynamic radius of 7.9 nm (Table 1). Interestingly, PEGylated scDb-AЈ (scDb-AЈ-PEG 40k ) and scDb-CЈ (scDb-CЈ-PEG 40k ) were not or only weakly recognized by an anti-His tag antibody (Fig. 3b). Thus, for further analysis a polyclonal rabbit antiserum was produced by immunization with scDb-AЈ-PEG 40k , which recognized both unPEGylated and PEGylated scDb variants in ELISA (not shown). Using this anti-serum similar binding of scDb and scDb-AЈ-PEG 40k to immobilized CEA was found (Fig. 3d). Furthermore, PEGylation did not influence binding to CEA-expressing cell line LS174T or CD3positive PBMCs as shown by flow cytometry (Fig. 4e). Because of the higher PEGylation efficiency all further studies were performed with scDb-AЈ-PEG 40k .
Chimeric Anti-CEA IgG1-A chimeric anti-CEA IgG1 was generated by fusing murine anti-CEA V H and V L (24) to human constant regions (C ␥1 , C k ). The antibody was purified from the supernatant of stably transfected CHO-K1 cells with yields of 3.7 mg/liter supernatant. In SDS-PAGE analysis, a single band with an apparent molecular mass of Ͼ250 kDa was observed under non-reducing conditions (Fig. 4a). After reduction, two bands with apparent molecular masses of 56 kDa and 29 kDa were observed. By size exclusion chromatography an apparent molecular mass of 280 kDa and a hydrodynamic radius of 5.7 nm was determined (Fig. 4b) ( Table 1). The discrepancy between calculated and apparent molecular mass of the intact IgG molecules was also observed for other recombinant and natural whole antibodies using an established set of standard proteins (see "Experimental Procedures") as molecular weight markers. This discrepancy is best explained by the Y-shaped structure of IgG molecules, which discerns it from other globular proteins. The purified antibody recognized CEA in ELISA  4) conditions. 3 g of purified protein was loaded per lane, and the gel was stained with Coomassie Blue. b, SDS-PAGE analysis of PEGylated scDb-AЈ-PEG 40k (lanes 1, 3, 5) and scDb-CЈ-PEG 40k (lanes 2, 4, 6). Gels were stained with Coomassie Blue (lanes 1, 2), barium iodide (lanes 3, 4), or immunoblotted with an anti-His tag antibody (lanes 5, 6). c, size exclusion chromatography of scDb-AЈ-PEG 40k and scDb-CЈ-PEG 40k . d, ELISA of binding of unmodified and PEGylated scDb molecules to immobilized CEA (n ϭ 2). e, flow cytometry analysis of binding of unmodified and PEGylated scDb molecules to CEA-expressing tumor cells (LS174T), CD3-positive PBMCs, or HEK293 cells included as negative control. Gray-filled, cells incubated with immune serum and secondary antibody alone; black line, scDb; black dotted line, scDb-AЈ-PEG 40k ; gray line, scDb-CЈ-PEG 40k . (Fig. 4c) and on CEA-expressing tumor cell line LS174T in flow cytometry (Fig. 4d).
Target Cell-dependent Effector Cell Activation-The potential of the glycosylated and PEGylated bispecific anti-CEACD3 scDb to activate effector cells (PBMCs) in a target cell-specific manner was determined by an IL-2 release assay. In this assay, CEA-expressing cells are incubated with PBMCs in the presence of increasing concentrations of bispecific antibody, and stimulation of PBMCs is determined by measuring IL-2 release after 24 h. This assay showed that glycosylated scDb (determined for scDb-ABC 1 and scDb-ABC 7 ) as well as scDb-AЈ-PEG 40k are able to stimulate PBMCs. However, both modifications reduced stimulatory efficacy of the scDb (Fig. 5a). Unmodified scDb mediated a maximal release of ϳ1.2 ng/ml IL-2 with an EC 50 of 0.14 nM and an optimum between 0.3 and 3 nM. ScDb-AЈ-PEG 40k reached a similar maximal release although with an EC 50 of 1.3 nM. ScDb-ABC 1 and scDb-ABC 7 reached only a maximal IL-2 release of ϳ0.8 ng/ml with an EC 50 of 0.4 and 0.9 nM, respectively. Selectivity of target cell-dependent stimulation was shown with control experiments using CEA-negative HT1080 cells stably transfected to express fibroblast activation protein (FAP). All scDb CEACD3 molecules (at a concentration of 25 nM) mediated no or only marginal stimulation of PBMCs, while a control scDb33CD3 directed against FAP and CD3 showed strong stimulation of PBMCs (Fig. 5b).

DISCUSSION
We have generated N-glycosylated scDb by genetically introducing several N-glycosylation sites into linker sequences connecting the variable domains as well as a C-terminal extension of varying length. All scDb variants containing 3, 6, or 9 sequons were expressed in glycosylated form in HEK293 cells. Yields were similar to that obtained for the unmodified scDb indicating that the modifications do not interfere with translation and secretion into the cell culture supernatant. SDS-PAGE analysis revealed heterogeneity in the degree of N-glycosylation. This finding indicates that not all sites are glycosylated with the same efficiency, especially in the scDb-ABC 7 construct containing 7 sequons in the C-terminal extension in close proximity to each other.
All N-glycosylated constructs were biologically active and recognized CEA and CD3, similar to unmodified scDb as shown by ELISA and flow cytometry analysis. Furthermore, a target cell-dependent IL-2 release assay demonstrated the capability to simultaneously bind both antigens on cells and thus to stimulate effector cells. However, compared with unmodified scDb the N-glycosylated scDb variants showed an ϳ3-5-fold reduced stimulatory activity. Similar observations were made with a PEGylated scDb as well as for fusion proteins of scDb and human serum albumin or an albumin-binding domain of streptococcal protein G (4,15). These findings indicate that modifications of the scDb structure and size have a negative effect on stimulation of PBMCs, which might be caused by interfering with the formation of close contacts between target and effector cells. To investigate if this is a general phenomenon we have initiated studies investigating other scDb molecules with different target cell specificity.
Glycosylation of scDb led to a prolonged circulation with a 2.3-2.9-fold increase of the AUC. Interestingly, pharmacokinetic properties of the three N-glycosylated scDb variants did not differ significantly. This finding indicates that the addition of 3 N-glycosylation sites is sufficient to prolong circulation time. These three sites increase the molecular mass by ϳ6 kDa as determined by SDS-PAGE analysis leading to a small increase of the Stokes radius from 2.7 to 2.9 nm. For hyperglycosylated erythropoietin (darbepoetin alpha) with a molecular mass of 37 kDa an ϳ2-fold increase of the terminal half-life was observed after i.v. injection into rats and a 3-fold increase of the terminal half-life from 8.5 h for normally glycosylated epoetin alpha to 26.3 h for darbepoetin alpha in dialysis patients (19). Glycosylated FSH analogs containing 2 or 4 N-glycosylation sites and a molecular mass of ϳ55 kDa exhibited a 2-4-fold increase of the AUC after i.v. injection into rats (20,21). Thus, for hyperglycosylated erythropoietin and glycosylated FSH a similar increase in half-lives was observed as for our glycosylated scDb variants. These findings indicate that the addition of a few N-glycans improves the pharmacokinetic properties of a protein leading to a moderate increase in circulation time by a factor of 2-4.
Comparison of an anti-CEA FabЈ molecule with a natural N-glycosylation site in the V L domain obtained from hybridoma-produced IgG (thus being glycosylated) and a bacterially expressed non-glycosylated FabЈ revealed faster elimination from circulation and reduced accumulation in the tumor of the non-glycosylated FabЈ (27). Accelerated clearance was also described for the non-glycosylated forms of various other glycoproteins, including antithrombin, lymphotoxin, interferon-␥, and granulocyte macrophage colony-stimulating factor (GM-CSF), further underlying the importance of N-glycosylation for pharmacokinetics (28 -31).
The type of N-glycosylation plays an important role in determining the circulation time of glycoproteins. Thus, an engineered N-glycosylated scFv containing high-mannose carbohydrate chains at a C-terminal extension was more rapidly cleared from circulation than the unmodified scFv (32). Similar observations were made for interferon-␥ (29). We found that the three N-glycosylated scDb variants all contained N-glycans of the complex type (bi-, tri-, and tetra-antennary). No hybrid or high-mannose N-glycans could be detected. These complex type N-glycans were composed of a mixture of sialylated and non-sialylated chains. The predominant N-glycans found in all three N-glycosylated scDb variants are FH4N5, F2H4N5, and FH5N5, which account for ϳ40 -50% of the N-glycans. These N-glycans differ from those found in the Fc-regions of antibodies, which bear mainly FH3N4 (G0F) oligosaccharides (33). The  N-glycosylation pattern of the scDb variants is, however, similar to that found in other recombinant glycoproteins expressed in HEK293 cells (34). Glycosylation of recombinant antibodies and other proteins is influenced by the cell line and conditions used for expression (18,35). Thus, it was found that expression of glycoproteins in CHO, NSO, and Sp2/0 cells can produce abnormally glycosylated antibodies that lack potency and are potentially immunogenic (36). For example, murine NSO cells express more heterogeneous oligosaccharide patterns, e.g. than CHO cells, with a tendency for ␣-linked galactose residues and N-glycolylneuraminic acid absent in glycoproteins from primate cell lines (37). Furthermore, experimental data indicate that the N-glycosylation pattern are also protein-dependent, e.g. as shown by studies of the glycosylation pattern of recombinant osteonectin, TSC-36, and bone sialoprotein expressed in HEK293 cells (34,38,39).
In vivo experiments have shown that the extent and type of sialylation influences serum half-lives of recombinant proteins (40). Hence, the pharmacokinetic properties of our N-glycosylated scDb molecules might be further improved using expression systems leading to increased production of fully sialylated N-glycans with ␣-2,6-linked terminal sialic acid residues (41).
PEGylation of scDb strongly improved pharmacokinetics of scDb, similarly to fusion of scDb to serum albumin or an albumin-binding domain as shown in previous studies (see also  Tables 1 and 3) (4,15). We found that conjugation of a 40-kDa branched mPEG chain strongly increased the hydrodynamic radius of the protein from 2.7 to 7.9 nm as determined by SEC. Similar observations were made for other PEGylated recombinant antibodies (42). Thus, the increased circulation time is likely caused by a reduced renal clearance. The introduction of a cysteine residue into one of the flanking linkers or at the C terminus allowed for a site-directed and efficient coupling of a single PEG molecule. Several studies showed that random PEGylation of whole antibodies or Fab fragments can reduce antigen-binding activity depending on number and length of conjugated PEG chains (summarized in Ref. 10). However, reduced antigen binding was also described for PEGylated antibody molecules generated by site-directed PEGylation mainly caused by reduced target association rates (43). We could not observe a reduced binding of the modified scDb molecules to each antigen separately as shown in ELISA and flow cytometry experiments, although for all modified scDb reduced activity was observed in IL-2 release assays where simultaneous binding is required.
None of the applied strategies reached the pharmacokinetics observed for a chimeric IgG1, which exhibited a 17-fold higher AUC than scDb. Similar to the albumin-based strategies the long circulation time of IgG is based on a reduced renal clearance and on FcRn-mediated recycling processes (44). However, currently we do not know to which extent the latter contributes to pharmacokinetics of IgG and the albumin-utilizing modifications. A comparison of the pharmacokinetics in wild-type and FcRn knock-out mice revealed a reduction of terminal halflife of albumin (from 39 to 24 h) and IgG (from 95 to 19 h), while the half-life of IgA was not affected (45). This finding indicates that FcRn-mediated recycling contributes to different extents to pharmacokinetics of albumin and IgG in mice.
In summary, we have completed our comparative analysis of various strategies to prolong circulation time of a bispecific single-chain diabody including fusion to albumin (4), fusion to an albumin-binding domain (ABD) of streptococcal protein G (15), PEGylation and N-glycosylation. These studies established that in mice longest circulation time is obtained by fusion of scDb to an ABD (14-fold increase of AUC), followed by a scDb-HSA fusion protein and PEGylated scDb (ϳ10-fold increase of AUC), while N-glycosylation in HEK293 cells resulted only in a moderate increase in circulation time (2-3fold increase of AUC) (summarized in Table 3). Further studies have now to demonstrate how these increases in circulation time translate into improved efficacy in vivo.