Glycoengineering Approach to Half-Life Extension of Recombinant

Jun 9, 2012 - Copyright © 2012 American Chemical Society ... in human embryonic kidney cells engineered to express human polysialyltransferase, and t...
0 downloads 0 Views 5MB Size
Download PDF
Article pubs.acs.org/bc

Glycoengineering Approach to Half-Life Extension of Recombinant Biotherapeutics Chen Chen,†,# Antony Constantinou,†,# Kerry A. Chester,‡ Bijal Vyas,† Kevin Canis,† Stuart M. Haslam,† Anne Dell,† Agamemnon A. Epenetos,† and Mahendra P. Deonarain*,† †

Department of Life Sciences, Faculty of Natural Sciences, Imperial College London, Exhibition Road, London, United Kingdom, SW7 2AZ ‡ UCL Cancer Institute, Paul O’Gorman Building, 72 Huntley Street, London, United Kingdom, WC1E 6BT S Supporting Information *

ABSTRACT: The potential for protein-engineered biotherapeutics is enormous, but pharmacokinetic modulation is a major challenge. Manipulating pharmacokinetics, biodistribution, and bioavailability of small peptide/protein units such as antibody fragments is a major pharmaceutical ambition, illustrated by the many chemical conjugation and recombinant fusion approaches being developed. We describe a recombinant approach that leads to successful incorporation of polysialic acid, PSA for the first time, onto a therapeutically valuable protein. This was achieved by protein engineering of the PSA carrier domain of NCAM onto single-chain Fv antibody fragments (one directed against noninternalizing carcinoembryonic antigen-CEA and one against internalizing human epidermal growth factor receptor-2-HER2). This created novel polysialylated antibody fragments with desired pharmacokinetics. Production was achieved in human embryonic kidney cells engineered to express human polysialyltransferase, and the recombinant, glycosylated product was successfully fractionated by ion-exchange chromatography. Polysialylation was verified by glycosidase digestion and mass spectrometry, which showed the correct glycan structures and PSA chain length similar to that of native NCAM. Binding was demonstrated by ELISA and surface plasmon resonance and on live cells by flow cytometry and confocal immunofluorescence. Unexpectedly, polysialylation inhibited receptor-mediated endocytosis of the anti-HER2 scFv. Recombinant polysialylation led to an estimated 3-fold increase in hydrodynamic radius, comparable to PEGylation, leading to an almost 30-fold increase in blood half-life and a similar increase in blood exposure. This increase in bioavailability led to a 12-fold increase in tumor uptake by 24 h. In summary, recombinant polysialylation of antibody fragments in our system is a novel and feasible approach applicable for pharmacokinetic modulation, and may have wider applications.

■

INTRODUCTION The clinical and commercial success of therapeutic antibodies has led to a proliferation of alternative formats1 and nonantibody alternative scaffolds.2 This has been driven, in part, by the limitations of whole immunoglobulin in some targeted applications. Their relatively large size and poor tumor vascularization impedes tissue penetration.3,4 Furthermore, in immuno-compromised patients and within the immunosuppressed microenvironment of the tumor the Fc effector/ recruitment function of whole antibodies becomes largely redundant.5 Finally, the Fc domain can cross-react with normal tissues, which can also lead to unwanted side effects, particularly when cytotoxin-loaded antibodies are used. Binding ligands devoid of Fc-regions such as nanobodies, scFvs, anticalins, and DARPins are being developed.1,2 Being smaller, they are capable of more rapid tissue penetration but clear from the system rapidly thus decreasing the therapeutic window. There is a need for noninvasive technologies to increase small protein half-life, without the disadvantages associated with the Fc-domain. © 2012 American Chemical Society

Polymer conjugation using poly(ethylene glycol) (PEGylation) has been one of the most successful approaches to increasing protein serum half-life. PEG is a neutral polymer that can bind water molecules, forming an “aqueous cloud” around the protein.6,7 This gives the conjugate a larger hydrodynamic volume compared to its true molecular weight, affecting its pharmacokinetics and pharmacodynamics in the body. Increasing protein size to above 60−70 kDa (or around 50 Å in diameter) excludes it from glomerular filtration and maintains its bioavailability.8 In addition to its size, the protein surface is modified and biological epitopes are shielded from potential immune responses or degradation.6−8 However, the use of synthetic nonbiodegradable polymers has raised concerns. The primary reason for this is the discovery that, when PEG is used for chronic conditions, PEGylated peptides or byproducts have been reported to Received: November 17, 2011 Revised: June 2, 2012 Published: June 9, 2012 1524

dx.doi.org/10.1021/bc200624a | Bioconjugate Chem. 2012, 23, 1524−1533

Bioconjugate Chemistry

Article

nude mice (6−8 weeks old) were from Harlan UK and all in vivo work was carried out under Home Office Project License: 70/6982 under Dr. Deonarain, Imperial College London. Cell Transfections, Selections, and Expression. HEK cell transfections were done using Fugene HD (Roche) in serum-free DMEM media using 3 × 106 cells in 10 cm tissueculture dishes and 10 μg/40 μL of Fugene HD. Double transfection was achieved by mixing both plasmid DNA (10 μg each) and 80 μL of Fugene HD. For stable cell selection, cells were washed with PBS, trypsinized, and reseeded in complete DMEM media with Zeocin (1 mg/mL) or G418 sulfate solution (1 mg/mL). The surviving transfected cells formed monoclonal cell colonies, which were expanded and maintained in 0.25 mg/mL antibiotic. Selected colonies on the plate were picked by trypsin-soaked 5 mm sterilized paper and transferred into a 24-well plate where they were expanded and stored. Recombinant protein-expressing cells were grown in Freestyle293 expression medium, supplemented with 4 mM L-glutamine. Valproic acid (1 mM) was included in the culture media to enhance polysialyl-transferase (PST) enzyme expression as suggested by previous researchers.20 Clones were identified by protein expression level using dot-blots and ELISAs. Prior to HEK293 cells, initial experiments were carried out in a rat neuronal cell line (NB2a) which was expected to express high levels of polysialylated NCAM. However, none of the derived clones displayed significant polysialylation. Successful polysialylation was also achieved in CHO and HeLa cells coexpressing PST. Enzyme-Linked Immunosorbent Assay (ELISA). Antigens were immobilized on the surface of 96-well Nunc immunosorb plates overnight at 4 °C in Tris-buffered salineTBS, pH 7.8 (100 ng/well for biotinylated N-A1 and hErbB2Fc). Biotinylation was carried out at a minimal 1:1 ratio using SulphoNHS-LC-LC-Biotin (Pierce). All incubations were carried out in PBS+3% Marvel milk protein and all washes in PBS+0.01% Tween-20 followed by PBS. All binding incubations were at 37 °C for at least 1 h. For biotinylated antigens, plates were precoated in 20 μg/mL avidin. A serial dilution of samples (scFvs or scFv fusion proteins) was prepared with a typical range between 5000 and 0.05 nM in triplicate. Detection was achieved using the relevant horseradish peroxidase conjugated secondary antibodies. Development was with BM Blue POD substrate. Absorbance at 450 nm was measured on a Spectramax 340pc plate reader. The equilibrium dissociation constant (KD) determinations were fitted, using Sigmaplot v 11, to a standard three-parameter sigmoid curve. Surface Plasmon Resonance. Real-time binding analysis of MFE23 scFv and its fusions was performed on a BIAcore 3000 instrument. The running buffer was 100 mM Tris buffer pH 7.4 containing 300 mM NaCl, 0.005% P20 surfactant, and 0.005% sodium azide. All experiments were carried out using double referencing. A streptavidin (SA) sensor chip was immobilized with 700−900 resonance units (RU) of biotinylated N-A1. For kinetic measurement experiments, a serial dilution containing a concentration range of 50 to 0.4 nM of each MFE23-based fusion protein sample was prepared in the running buffer. The standard kinetic wizard was used including blank buffer injections. Regeneration was with 10 mM HCl for 60 s. Evaluation of the kinetic data was with the BIAevalution software (v 4.1) employing a global fit to the 1:1 (Langmuir) binding model correcting for spikes and bulk shifts. Chromatography. Cobalt metal affinity resin (TALON) run in TBS was used to purify the proteins with 2 mL TALON

accumulate in tissues and cause unforeseen toxic effects (e.g., kidney cells vacuolizaton9) and/or generate an immune response.10 As such, a new generation of biodegradable biopolymers are now being investigated as candidates to supersede PEG technology by addressing these concerns. These technologies include HEPylation (heparosan polymer), HESylation (hydroxyethyl starch7), and the most extensively studied, polysialylation (polysialic acid). Polysialic acid (PSA) is a highly hydrophilic and negatively charged biopolymer that is extensively found in nature. On pathogenic bacteria, PSA (columinic acid) is a capsular virulence factor and is involved in protecting against phagocytosis and immune clearance.11 PSA has similar properties to PEG, including the ability to associate with 5−10 times its mass of water increasing its hydrodynamic volume, hiding immunogenic epitopes, and increasing overall conjugate solubility and stability.12−14 These properties have led to commercialized use of chemically linked PSA,15 and early clinical data have shown that polysialylated erythropoietin is well-tolerated in patients.12 We have recently examined the chemical conjugation of PSA onto antibody fragments, showing that their systemic half-life and bioavailabilty can be significantly increased while maintaining better tumor/blood ratios than whole immunoglobulins.12−14 However, limitations such as loss of antibody binding, long processing times, and poor conjugate yields were observed. In humans, PSA is found predominantly on NCAM where it has a role in lubricating interactions and cell migration and promoting the plasticity of neuronal cells in early brain development.16 We hypothesized that a recombinant humanbased approach to heterologous protein polysialylation would have many biotechnological advantages over chemical approaches (e.g., retention of function and product yields). This could lead to the target therapeutic being made in its final form with no further manipulations necessary in a fully human form. Although glyco-engineering of biotechnologically important proteins is an advanced and evolving field, with applications ranging from increased potency17 to longer halflife,18 there was no existing means to recombinantly attach PSA onto heterologous proteins. Therefore, we set out to develop such a technology based on NCAM which is naturally polysialylated and shares homology with antibody immunoglobulin domains.

■

EXPERIMENTAL SECTION Reagents. N-A1 (CEA subdomain which bears the epitope for MFE-23) was expressed from Pichia pastoris and purified by IMAC.19 Recombinant HER2 (hErbB2/Fc) was from R&D system Ltd. All standard antibodies were from Sigma, but specific clones were as follows: Anti-NCAM (CD56, mouse monoclonal clone 123C3 recognizing NCAM epitope between exons 11−13, first fibronectin type-III, FN1, Abcam); Anti-PSA (rat monoclonal IgM clone-12F8, recognizing the polysialic acid on NCAM, with sialic acid residue number >7, BD Pharmingen); Endo-Neuraminidase (Endo-N from Phage K1 0.7, recognizing α2−8 linked sialyl residues, requiring a minimum of 5 sialyl residues for activity (Abcys)); Acetylneuraminyl hydrolase (Exo-N, Clostridium perfringens α2− 3,α2−6 sialidase, New England Biolabs); N-Glycosidase F (PNGase F- Flavobacterium meningosepticum amidase that cleaves between the innermost GlcNAc and asparagine residues of high mannose, hybrid, and complex oligosaccharides from glycoproteins; New England Biolabs). Healthy female BALB/c 1525

dx.doi.org/10.1021/bc200624a | Bioconjugate Chem. 2012, 23, 1524−1533

Bioconjugate Chemistry

Article

Figure 1. Schematic illustration of PSA fusion protein structures. (A) Glycan structure of the polysialylated core on NCAM Ig5 for reference. (B) Cartoon illustrations of the fusion proteins with the scFv-Iq5-FN1 protein attaining recombinant polysialylation (blue line) after enzymic processing (scheme 4). The controls cannot achieve polysialylation (schemes 1−3).

glycans were released by PNGase F digest and extracted using a Sep-pak C18 reverse-phase column. Aliquots of purified native oligosaccharides were spotted onto a MALDI TOF-MS target and allowed to dry. The dried samples were lactonized by addition of 1 μL of a 0.1% ortho-phosphoric acid solution. Lactonized samples were dissolved in 1 μL of the ATT matrix (20 mg/mL 6-aza-2-thiothymine in 50:50 acetonitrile/50 mM ammonium citrate) and allowed to dry. Matrix assisted laser desorption ionization/time of flight (MALDI-TOF) MS analyses were performed using a Voyager DE-STR (Applied Biosystems) mass spectrometer in the linear negative mode with delayed extraction. For the N-glycan cores and PSA, aliquots of native N- and O-linked oligosaccharide samples were subjected to digestion by endoneuraminidase. Permethylated glycans were extracted in chloroform and washed several times with dH2O, prior to a C18 purification step of the derivatized glycans. Permethylated glycans were dissolved in 10 μL of 1:1 methanol/water. One μL was mixed with 1 μL of the DHB matrix (10 mg/mL of 2,5-dihydroxybenzoic acid in 50:50 methanol/water) and was spotted onto a MALDI-TOF/TOF plate and allowed to dry under vacuum. MS and MS/MS analyses were performed using a 4800 MALDI TOF/TOF mass spectrometer (Applied Biosytems) in the reflectron positive mode with delayed extraction. In MS/MS mode, the collision energy was set to 1 kV and argon was used as collision gas. The instrument was calibrated using [Glu1]-fibrinopeptide G human as external calibrant. All data were acquired using the 4000 Series Explorer Instrument Control Software and were processed using Data Explorer MS processing software. Full technical details and conditions are described in the Supporting Information. Mouse Pharmacokinetic and Biodistribution Studies. Protein samples were dialyzed PBS buffer prior to iodination. Pierce Pre-Coated Iodination tubes were used to radio-iodinate the protein samples. Direct iodination was carried out inside the tubes according to the manufacturer’s protocol, but this led to overiodination and oxidation of the PSA. A more gentle iodination was carried out by removing the activated 125I and radiolabeling in a clean tube, To stop the reaction, 50 μL of scavenging buffer (10 mg/mL tyrosine) in PBS was added. All samples were dialyzed into PBS were sterilized through a 0.2 μm PES syringe filter before injecting to the mice. Up to 200 μL (∼10 μg) of each radiolabeled fusion protein was injected intravenously into the mouse tail vein. For blood pharmacokinetic studies, groups of 4−6 BALB/c mice were used for each protein sample test. At the appropriate time points after

resin was used for 500 mL dialyzed cell culture medium. Protein was eluted in cold 200 mM imidazole in TBS. Further purification of proteins with different degrees of polymerization (DP) after IMAC purification was achieved by anion exchange chromatography based on the negative charge nature of sialic acid. A 1 mL AcroSep Chromatography column with strong anion exchanger Q-Ceramic HyperD F was used connected to a BIO-RAD BioLogic DuoFlow liquid chromatography system with QuadTec UV−vis detector for simultaneous wavelengths monitoring. Tris buffer (100 mM, pH 7.0) was used as the chromatography buffer with a NaCl step elution. Low salt (0.1 M NaCl) buffer eluted the non-PSA containing scFv-Ig5-FN1, whereas 0.2 M NaCl eluted low-PSA containing proteins. A long wash was used to remove all of these species before a high salt (1 M NaCl) wash eluted the highly polysialylated scFv fusion protein. Pure proteins were analyses by SDS-PAGE and quantified by Nanodrop using extinction coefficients of 97 750 M−1 cm−1, which was determined from the protein amino acid sequence using ProtParam on the ExPASy Server (http://www. expasy.ch/tools/protparam.html). Size exclusion chromatography (SEC) was used for the determination of protein apparent hydrodynamic volume. HiLoad Superdex-200 prep grade column (MW range 10−600 kDa, bed volume 120 mL) was used on a BIO-RAD BioLogic DuoFlow system with Quadtec. The running buffer was TBS at 1 mL/min. Glycosidase Treatments. Typically, excess amounts of each were used to hydrolyze the oligosaccharides from the various proteins (less than 10 μg, incubated overnight incubation at 37 °C). For Endo-Neuraminidase, 2 units of the enzyme was used in PBS buffer. For Exo-N, 25 units was used with 1×G1 Reaction Buffer (50 mM sodium citrate, pH 6.0). Before PNGase F hydrolysis, proteins were denatured with 1× “Glycoprotein denaturing buffer” (which contains a final concentration of 0.5% SDS and 40 mM DTT) in a total volume of 10 μL at 100 °C for 10 min. Together with the addition of 1% NP-40 and 1× G7 Reaction Buffer (50 mM sodium phosphate, pH 7.5), 20 units of PNGase F was used. Confocal Laser Scanning Immunofluorescent Microscopy. Confocal laser scanning microscopy (CLSM) was used for tracking live cell (LS174T-CEA expressing and SKOV3HEr2 expressing) surface antigen binding and internalization processes. Secondary antibodies, such as anti-myc, anti-6×His, anti-FN1, or anti-PSA antibodies were used. Full details are described in the Supporting Information. MALDI-TOF Mass Spectrometry Analyses. Carboxymethylated tryptic fragments were prepared and purified. The 1526

dx.doi.org/10.1021/bc200624a | Bioconjugate Chem. 2012, 23, 1524−1533

Bioconjugate Chemistry

Article

Figure 2. Production and purification of scFv-NCAM fusion proteins. (A) Chromatogram trace of the purification of various polysialylated species based on degree of negative charge, using increasing NaCl steps to elute proteins. Inset is an antimyc blot (to detect entire protein) and an anti-PSA blot (to detect polysialylated protein) showing that the 1 M NaCl buffer eluted the highly polysialylated protein. Around 70% of the applied material eluted as highly polysialylated. (B) SDS-PAGE (a, Coomassie-stained) and Western blotting (b, anti-Myc; c, anti-MFE23; and d, anti-PSA) all confirmed that the correct full-length proteins had been produced and isolated bearing the correct molecular weight. The polysialylated protein appeared as a more diffuse band, especially when detected with the anti-PSA antibody which detects longer chains of PSA. The SDS-PAGE (and calculated) molecular weights were as follows: Lane 1, scFv 28 kDa (29.2 kDa); Lane 2, scFv-FN1 48 kDa (43.7 kDa); Lane 3, scFv-Ig5 49 kDa (44.3 kDa); Lane 4, scFv-Iq5-FN1 75 kDa (53.6 kDa + (0.29 × PSA DP). (C) An 11 kDa bacterial PSA, chemically attached to a scFv, is shown for comparison showing a greater range of heterogeneity.

injection, a cut was made in the mouse tail tip, and 20 μL of blood was taken from each mouse by using microhematocrit capillary tubes. Blood radioactivity was determined by gamma counting and values converted to percentage of injected dose per gram of tissue. For analyzing the in vivo blood clearance, pharmacokinetics data were fitted to the two-compartmental intravenous model of clearance, which takes into account the biexponential clearance pattern of systemic molecules. Tumor and normal tissue uptake was determined in groups of 4 mice culled at two time points using proteins radiolabeled as above injected into BALB/c nude mice bearing subcutaneously implanted LS174T tumors.14 All tissues were weighed and counted and the uptake expressed as the percentage per gram tissue of injected sample.

results; not shown). Individual clones expressing recombinant proteins of all types were selected, cultured, and screened for binding activity by ELISA. All clones were screened for binding with anti-His6 or antimyc secondary antibodies. ScFv-fusion clones that were designed to be polysialylated (polysialylated scFv: scFv-Ig5-FN1) were additionally screened with an antiPSA antibody which detects PSA chain lengths (degree of polymerization, DP) of greater than 5 sialic acid units.25 Positive clones were identified, expanded, adapted to serumfree media, and purified by metal affinity chromatography. Polysialylated scFvs were further purified to separate non-PSA, low-PSA, and high-PSA glycoforms by anion exchange chromatography (Figure 2a). No carry-over of contaminants such as autopolysialylated polysialyltransferase enzymes was seen. Full protein chemical and glycosylation characterization was confirmed by Western blotting (Figure 2b) demonstrating scFv polysialylation and full-length protein production (Figure 2b). All proteins were expressed and purified well (yields were around 2−5 mg/L cell culture) and migrated as expected on SDS-PAGE (Figure 2b). The polysialylated scFv displayed a polydisperse migration pattern (around 70 kDa) confirming various degrees of polysialylation. SDS-PAGE, although indicative of molecular weight, is likely to be misleading due to the steric hindrance of PSA influencing migration. The polysialylated scFv dispersity was lower than chemically produced polysialylated antibodies as judged by immunoblotting (Figure 2c) indicating a higher-quality, better polysialylated product. Characterization of the Glycans on scFv-NCAM-Based Fusion Proteins. Glycosidase treatment was used to verify that various glyco-species were present (Figure 3a). Three scFv glyco-isoforms differing by the degree of polysialylation showed corresponding band shifts by different sialic acid hydrolyzing

■

RESULTS Construction and Production of scFv-NCAM Based Fusion Proteins. Colley and co-workers have described substantial and elegant research mapping out the important features of NCAM polysialylation.21,22 They discovered that the N-linked core glycosylation motif on the fifth immunoglobulin (Ig5) domain was the acceptor for sialic acid polymerization21,22 (Figure 1a), which was catalyzed by polysialyltransferase enzymes docking onto the juxtaposed first fibronectin (FN1) domain, which lies C-terminal to the five immunoglobulin domains of the polymeric glycoprotein.22 We proposed that the Ig5 domain could be used as a carrier for PSA on alternative proteins, such as antibody fragments. We constructed a series of anticarcinoembryonic antigen23 singlechain Fv-fusion proteins (Supporting Information Figure 1 and Figure 1b) to test this idea and transfected the DNA constructs into human embryonic kidney (HEK293) cells which were already transgenically expressing a human polysialyltransferase enzyme, PST/ST8SiaIV24 (STX/ST8SiaII24 also gave similar 1527

dx.doi.org/10.1021/bc200624a | Bioconjugate Chem. 2012, 23, 1524−1533

Bioconjugate Chemistry

Article

Figure 3. Glycosylation characterization of the scFv-NCAM fusion proteins. (A) Glycosylation status of the scFv-fusion proteins was verified by exoN (which cleaves 2,3- or 2,6-linked sialic acids), endo-N (which cleaves 2,8- or 2,9-linked sialic acids) and PNGase F (which cleaves the entire Nlinked glycan core). Protein band shifts were revealed by both Coomassie staining and anti-Myc immunoblotting. The Endo-N treatment results in a significant shift in molecular weight for the highly polysialylated scFv only. (B) Characterization of PSA levels were determined by MALDI-TOF mass spectrometry. N-Glycans released from the scFv-Iq5-FN1 protein by PNGase F digestion were lactonized prior to analyses. The linear negative ion mode MS spectrum shows the detection of a population of 273 Da unit oligomers corresponding to lactonized neuraminic acid chains up to about 35 residues long. In order to confirm the nature of these oligomers, the PSA extensions were digested by an endoneuraminidase followed by permethylation and analyses by MS/MS. The resulting reflectron positive mode MALDI-TOF/TOF MS/MS spectrum (inset) provides a diagnostic fragmentation pattern undoubtedly confirming the nature of the neuraminic acid chains.

(273 Da) was detected up to about 35 residues long, confirming the high polysialylation level of the scFv glycoprotein. In addition, treatment of the native glycans by endoneuraminidase followed by permethylation and MS and MS/MS analyses on a MALDI-TOF/TOF instrument confirmed the presence of a linear oligomer of sialic acids and revealed the nature of the N-glycan cores (Figure 3b and Supporting Information Figure 2). In contrast, the same analyses undertaken on the post PNGase F fraction focusing on potential O-glycans did not provide any signals, showing that polysialylation, very similar to the level found on human NCAM,26 occurs only on N-glycans of the scFv.

glycosidases and eventually reached the same molecular weight level after N-glycan removal. Conventional mass spectrometry glycomics was able to determine the core N-glycosylation (Supporting Information Figure 2). However, the distribution of the oligo-sialylation status was not determinable. The degree of terminal polysialylation of the highly polysialylated scFv was also not possible by conventional techniques. A modification of the procedure reported by Galuska26 was carried out utilizing PSA lactonization to stabilize the labile sugars during MALDITOF mass spectrometry. Analyses of N-glycans released from the polysialylated scFv by PNGase F digestion were performed in the linear negative ion mode (Figure 3b). A heterogeneous population oligomers comprising lactonized sialic acid units 1528

dx.doi.org/10.1021/bc200624a | Bioconjugate Chem. 2012, 23, 1524−1533

Bioconjugate Chemistry

Article

Table 1. Binding Constants of the Various MFE-23 Based scFv-NCAM Fusion Proteinsa scFv sample

ELISA Kd (nM)

scFv scFv-Ig5 scFv-FN1 scFv-Ig5-FN1(Sia) scFv-Ig5-FN1(low-PSA) scFv-Ig5-FN1(high-PSA, CM3) scFv-Ig5-FN1(high-PSA, CM5)

± ± ± ± ± ± ±

4.3 9.4 17.5 13.9 8.3 6.0 5.9

0.6 1.9 3.9 1.5 0.7 0.6 0.5

SPR Kd (nM) 5.7 ND ND 7.4 6.1 13.8 141

SPR kon (M−1 s−1) (6.3 ND ND (4.7 (9.2 (2.1 (3.9

± 0.2) × 10

4

± ± ± ±

0.6) 0.4) 0.3) 0.3)

× × × ×

104 104 104 103

SPR koff (s−1) (3.6 ND ND (3.5 (5.6 (2.9 (5.5

± 0.2) × 10−4

± ± ± ±

0.4) 0.5) 0.1) 0.9)

× × × ×

10−4 10−4 10−4 10−4

a

Equilibrium dissociation constants determined by ELISA on immobilized NA1 antigen and kinetic constants/equilibrium dissociation constants determined by BIACore SPR on low-dextran, CM3 chip-immobilized NA1 antigen. The SPR data was modeled to a 1:1 Langmuir binding equation which resulted in a good fit with χ2 values all less than 5% of Rmax. The anti-CEA Kd values determined by ELISA were all consistent with the published affinity of the MFE23 scFv (4 nM).34 The Kd value determined by SPR correlated with the ELISA data, but diffusion effects were seen when using a more dense and charged CM5 surface. This resulted in a lower association rate and hence Kd. ND-Not determined.

Antigen Binding Characterization. The various scFvfusion proteins were tested for both binding function by ELISA and surface plasmon resonance (SPR) kinetic analyses (Table 1). ELISA binding studies showed that the affinities as measured by equilibrium dissociation constants were similar (Kd = 3−10 nM), suggesting little or no detrimental effect due to the fusion domains and resulting glycosylation (Table 1). This was consistent with our previous work with C-terminal chemically conjugated PSA.14 BIAcore SPR using a low-density matrix surface gave similar affinities and component on- and off-rates (Table 1). However, a noticeable decrease in association rate of the polysialylated species was observed when using a high-density carboxy-methyl dextran/antigen matrix, suggesting some biophysical interactions (see Discussion). Cell Binding Studies. With the “front end” (antigen binding) and “back end” (effector function-polysialylation) characterized, we tested the ability of the various scFv derivatives to bind to live cells in vitro and its behavior in vivo. To make a comparison, we additionally made a series of similar fusion proteins with an anti-HER2 internalizing scFv, C6.5.27 This was expressed/purified and characterized in the same way as the anti-CEA MFE23 scFv proteins. Live cell binding was determined by confocal immunofluorescent microscopy on antigen-expressing and antigen-negative cell lines. The anti-CEA based scFv-NCAM fusion proteins all bound to CEA-expressing LS174T cells with no significant binding to antigen negative KB cells (Supporting Information Figure 3). A similar observation was made using flow cytometry, where all the anti-CEA scFv-NCAM fusion proteins bound similarly to LS174T cells, with negligible binding to KB cells (Supporting Information Figure 4). Anti-PSA antibody successfully detected the polysialylated scFv only. These observations also support the previous findings of Constantinou et al. using chemically conjugated PSA14 where we observed a lack of nonspecific binding to normal tissues in vivo. Interestingly, the anti-HER2 based polysialylated fusion protein was able to bind but not internalize to any significant level at 37 °C. This was in contrast to the nonpolysialylated anti-HER2 scFv (Figure 4). This suggested that polysialylation, in some instances, could interfere with internalization. Hydrodynamic Volume, Pharmacokinetic and Tumor Uptake Effects. The low and high levels of polysialylation resulted in an increased hydrodynamic volume as determined by size-exclusion chromatography (Supporting Information Figure 5). This confirmed that the PSA sugars were functioning as a water-retaining biopolymer. The highly polysialylated scFv

had an apparent molecular weight of over 700 kDa. Compared to BSA (hydrodynamic diameter of around 35 Å, molecular weight of 66 kDa), which migrates true to its molecular weight by SEC,28 the highly polysialylated scFv was estimated to have a radius of up to 78 Å, an almost 3-fold increase compared to that predicted for the nonpolysialylated proteins. This is clearly above the renal exclusion limit29 and comparable to the effects of protein PEGylation.6 This increase in hydrodynamic volume and consequently increased apparent mass led to an increase in blood half-life in vivo (Figure 5a; Table 2). This pharmacokinetic modulatory function led to a moderate increase in distribution half-life (t1/2 α), a 29.6-fold increase in blood elimination half-life (t1/2 β) and thus a 28.8-fold increase in overall blood bioavailability. The longer half-life in the blood allowed greater tumor exposure resulting in 19.5% injected dose/gram tissue being taken up by the tumor by 6 h and 21.4% by 24 h, a 7-fold and 12-fold increase, respectively (Figure 5b). Concomitantly, there were higher levels in polysialylated scFv in all the other organs, but there was no suggestion of nonspecific cross reactivity. As expected and also seen for chemically polysialylated scFvs, the tumor/blood ratios were lower due to the longer presence of the proteins in blood. Therefore, the achieved recombinant polysialylation technology was able to extend the half-life of an antibody fragment as well as chemical polysialylation and approaching PEGylation.

■

DISCUSSION In many technological applications, recombinant approaches supersede chemical methods due to the direct and simplified systems used. However, the commercial value of chemical polysialylation or PEGylation has driven its development to be a highly efficient and established process. A direct comparison showed that, for the recombinant polysialylation, protein yields were on average 5 mg/L/day, representing a yield of at least 50% of the expressed material (10 mg/L). This is substantially higher than the 1 mg obtained from a 10% yielding chemical process (10 mg starting material) in our hands.14 The polysialylated scFv dispersity was lower than chemically produced polysialylated antibodies as judged by immunoblotting, indicating a better polysialylated product. Recombinant protein could be isolated within a day from expression, compared to 2−3 days for a chemical approach.14 This reduced association rate “phenomenon” observed on a BIACore chip surface was explored by Pluckthun’s group30 investigating the binding of PEGylated scFvs by SPR. They similarly observed reduced association rates, which they attributed to the SPR system rather than an actual reduction 1529

dx.doi.org/10.1021/bc200624a | Bioconjugate Chem. 2012, 23, 1524−1533

Bioconjugate Chemistry

Article

Figure 4. Internalization properties scFv-NCAM fusion proteins. A comparison of the binding and internalization properties of anti-CEA and antiHER2 based scFv-Iq5-FN1 fusion proteins: (a) anti-CEA scFv-Iq5-FN1(PSA) on LS174T cells at 37 °C, (b) anti-HER2 IqG on SKOV3 cells at 4 °C, (c) anti-HER2 IgG on SKOV3 cells at 37 °C, (d) anti-HER2 scFv on SKOV3 cells at 37 °C, (e) anti-HER2 scFv-FN1 on SKOV3 cells at 37 °C, (f) anti-HER2 scFv-Iq5-FN1(Low PSA) on SKOV3 cells at 37 °C, (g) anti- HER2 scFv-Iq5-FN1(high PSA) on SKOV3 cells at 37 °C. Panels (a) and (b) show cell surface binding with little significant internalization, whereas panel (c) shows complete and significant internalization. Panels (d) to (f) show lower levels of internalization, whereas the highly polysialylated scFv (panel g) shows cell surface only binding comparable to (a) and (b).

the recombinantly applied PSA chain appears to be around 5−8 kDa. It is possible that the recombinantly applied PSA is longer and underestimated by the mass spectrometry due to the difficulty and complexities in obtaining a spectrum. It is clear that the PSA chain is the major determining factor in modulating the half-life. The increase in protein molecular weight (around 24 kDa) did not seem to affect the half-life as seen in the blood clearance studies. It is possible that there were minor effects due to the changes in protein pI which were not investigated. One should consider the possibility of refining the NCAM fusion proteins to possess the minimal protein sequences needed to achieve recombinant polysialylation. This would require a lot of protein design, but toward this we attempted to engineer a protease cleavage site between the Ig5

of affinity. The large hydrodynamic volume of PEG (and here, PSA) could impede scFv-antigen association by sterically blocking epitopes/paratopes, slowing diffusion, or, potentially significant for PSA, negative dextran surface−PSA repulsion.30 This finding also supports the biophysical data suggesting increased hydrodynamic volume, but it may have implications for the type of application best suited to such charged polymer−conjugates. The increased hydrodynamic volume achieved using the recombinant polysialylation technology was able to extend the half-life of an antibody fragment more effectively than the chemical polysialylation methods used previously in our laboratory. Here, a 10- to 15-fold increase in half-life/ bioavailability is seen with a 11 kDa PSA chain),14 whereas 1530

dx.doi.org/10.1021/bc200624a | Bioconjugate Chem. 2012, 23, 1524−1533

Bioconjugate Chemistry

Article

Table 2. Pharmacokinetic Characterization of Selected MFE23 Based scFv-NCAM Fusion Proteinsa scFv sample

t1/2α (h)

t1/2β (h)

t1/2β foldchange

scFv scFv-FN1 scFv-Ig5-FN1 (HighPSA) desialylated scFv-Ig5-FN1 (high-PSA) scFv-Ig5-FN1 (high-PSA) oxidized

0.353 ± 0.07 0.383 ± 0.09 0.301 ± 0.19

4.78 ± 1.1 4.58 ± 1.5 2.66 ± 0.4

1 1 0.6

30.86 31.26 25.79

1 0.9 0.7

1.474 ± 2.32

78.75 ± 7.3

29.6

890.14

28.8

2.19 ± 0.12

24.75 ± 2.1

5.2

132.43

4.3

AUC (% h/g)

AUC foldchange

a

Radiolabeled proteins were injected IV (t=0 h) and blood sampled at 1, 2, 6, 24, and 48 h post injection. The percentage of the injected dose per gram of blood was determined, plotted, and modeled to a biexponential decay equation to estimate the clearance pharmacokinetic parameters. The plots were also integrated to estimate the total blood exposure (bioavailability) for t = 0 to t = 48 h. Polysialylated scFv had a significantly slower blood clearance and correspondingly higher bioavailability (as measured by area under the curve-AUC of dose vs time).

receptor dimerization and hence internalization. Normally, size is not a limitation to internalization since 750 kDa IgMs can internalize. The other possible mechanism is that the high negative charge makes it unfavorable to internalize. Similar observations have been made in the nanoparticle research field. For example, the biodegradable poly(ethylene glycol-graf tmethyl methacrylate) polymer designed for drug release was found to internalize less effectively when the zeta potential was more negative.32 Further experiments are needed to explore this observation further, but this observation points to alternative ways to inhibit receptor function. This technology, suitably developed, could result in a simple fusion tag that will extend the half-life of antibody fragments or any pharmaceutically important protein. This is a powerful approach with immense potential: The highly hydrophilic and negatively charged structure could have a better pharmacokinetic effect than PEG due to repulsion from the negatively charged glomeruli membranes.29 This same structure could also act as a nanoparticle framework for use as a scaffold for conjugating a variety of additional effectors. Further work is needed to understand how internalization is prevented, but this opens up intriguing and novel ways to antagonize receptormediated signaling. Yet to be explored would be the ability of sialic acids to confer anti-inflammatory effects.33 The next step is to apply recombinant polysialylation to a therapeutic model, such as TNFα or VEGFα antagonists, and more detailed studies regarding the microdistribution of such highly charged species within target tissues is needed. Recombinant polysialylation represents an attractive alternative to the Fc-domain of immunoglobulins, which may be less useful in effector recruitment within the immuno-suppressed microenvironment of the tumor.3 Additionally, the Fc domain can cross-react with normal tissues, leading to unwanted side effects. Hence, our work adds to the recently described peptidebased approaches of homoamino acid polymers (HAPylation34), the improved derivative, Pro-Ala-Ser (PASylation, being developed by the company called XL-protein), and the hydrophilic unstructured sequence called XTEN.35 Sialic acids, we believe, present additional benefits already outlined. Given

Figure 5. Blood pharmacokinetic analyses. Blood pharmacokinetic analyses. (a) Pharmacokinetic analyses of selected scFv-NCAM fusion proteins in BALB/c mice. Blood levels were monitored over time for radiolabeled polysialylated and nonpolysialylated scFvs The scFvs ± FN1 domain had near-identical pharmacokinetic profiles (rapid clearance). The highly polysialylated scFv demonstrated a slow blood equilibrium (α-) phase and a slow (β-) elimination phase, properties which were lost upon enzymic desialylation. The low-level polysialylated scFv, a result of periodate oxidation, showed intermediate pharmacokinetic properties. See Table 2 for pharmacokinetic constants. (b) Bar graph (±SEM) of mean tumor and normal organ uptake at two selected time points (6 h and 24 h) for scFv-Iq5FN1 (high PSA) and scFv-Iq5-FN1 (desialylated) fusion proteins.

and FN1 domains to remove the FN1 docking domain once it had performed its function. This subtle sequence change led to unexpected but significantly reduced polysialylation and structural changes (data not shown). This supports the observations by Colley et al.22 relating to the key structural features required for enzyme recognition and NCAM polysialylation. The longer half-life led to a similar increase in tumor uptake as observed before with chemical polysialylation.14 However, this was a model system to exemplify the technology, and tumor localization may not be the best use for this technology due to poor specificity ratios and a lack of understanding of other factors such as tumor perfusion. The antibody blocking of soluble growth factors is likely to benefit from this type of halflife extension. Our previous work with fluorescent dye conjugates of C6.5 scFv showed that these conjugates were able to successfully internalize31 whereas the same polysialyated scFv was unable to enter cells. It is possible that the increased hydrodynamic volume resulted in a steric effect which physically blocks 1531

dx.doi.org/10.1021/bc200624a | Bioconjugate Chem. 2012, 23, 1524−1533

Bioconjugate Chemistry

Article

(9) Gaberc-Porekar, V., Zore, I., Podobnik, B., and Menart, V. (2008) Obstacles and pitfalls in the PEGylation of therapeutic proteins. Curr. Opin. Drug Discovery Devel. 11, 242−50. (10) Armstrong, J. K., Hempel, G., Koling, S., Chan, L. S., Fisher, T., Meiselman, H. J., and Garratty, G. (2007) Antibody against poly(ethylene glycol) adversely affects PEG-asparaginase therapy in acute lymphoblastic leukemia patients. Cancer 110, 103−11. (11) Muhlenhoff, M., Eckhardt, M., and Gerardy-Schahn, R. (1998) Polysialic acid: three-dimensional structure, biosynthesis and function. Curr. Opin. Struct. Biol. 8, 558−64. (12) Constantinou, A., Chen, C., and Deonarain, M. P. (2012) Polysialic acid and polysialylation to modulate antibody pharmacokinetics. In in Pharmacokinetic modulation of therapeutic proteins (Kontermann, R, Ed) Wiley, London (in press). (13) Constantinou, A., Epenetos, A. A., Hreczuk-Hirst, D., Jain, S., and Deonarain, M. P. (2008) Modulation of antibody pharmacokinetics by chemical polysialylation. Bioconjugate Chem 19, 643−50. (14) Constantinou, A., Epenetos, A. A., Hreczuk-Hirst, D., Jain, S., Wright, M., Chester, K. A., and Deonarain, M. P. (2009) Site-specific polysialylation of an antitumor single-chain Fv fragment. Bioconjugate Chem. 20, 924−31. (15) Fernandes, A. I., and Gregoriadis, G. (1997) Polysialylated asparaginase: preparation, activity and pharmacokinetics. Biochim. Biophys. Acta 1341, 26−34. (16) Mü h lenhoff, M., Oltmann-Norden, I., Weinhold, B., Hildebrandt, H., and Gerardy-Schahn, R. (2009) Brain development needs sugar: the role of polysialic acid in controlling NCAM functions. Biol Chem. 390, 567−74. (17) Umaña, P., Jean-Mairet, J., Moudry, R., Amstutz, H., and Bailey, J. E. (1999) Engineered glycoforms of an antineuroblastoma IgG1 with optimized antibody-dependent cellular cytotoxic activity. Nat. Biotechnol. 17, 176−80. (18) Elliott, S., Lorenzini, T., Asher, S., Aoki, K., Brankow, D., Buck, L., Busse, L., Chang, D., Fuller, J., Grant, J., Hernday, N., Hokum, M., Hu, S., Knudten, A., Levin, N., Komorowski, R., Martin, F., Navarro, R., Osslund, T., Rogers, G., Roger,s, N., Trail, G., and Egrie, J. (2003) Enhancement of therapeutic protein in vivo activities through glycoengineering. Nat. Biotechnol. 21, 414−21. (19) Sainz-Pastor, N., Tolner, B., Huhalov, A., Kogelberg, H., Lee, Y. C., Zhu, D., Begent, R. H., and Chester, K. A. (2006) Deglycosylation to obtain stable and homogeneous Pichia pastoris-expressed N-A1 domains of carcinoembryonic antigen. Int. J. Biol. Macromol. 39, 141− 50. (20) Beecken, W. D., Engl, T., Ogbomo, H., Relja, B., Cinatl, J., Bereiter-Hahn, J., Oppermann, E., Jonas, D., and Blaheta, R. A. (2005) Valproic acid modulates NCAM polysialylation and polysialyltransferase mRNA expression in human tumor cells. Int. Immunopharmacol. 5, 757−69. (21) Nelson, R. W., Bates, P. A., and Rutishauser, U. (1995) Protein determinants for specific polysialylation of the neural cell adhesion molecule. J. Biol. Chem. 270, 17171−9. (22) Colley, K. J. (2010) Structural basis for the polysialylation of the neural cell adhesion molecule. Adv. Exp. Med. Biol. 663, 111−26. (23) Begent, R. H., Verhaar, M. J., Chester, K. A., Casey, J. L., Green, A. J., Napier, M. P., Hope-Stone, L. D., Cushen, N., Keep, P. A., Johnson, C. J., Hawkins, R. E., Hilson, A. J., and Robson, L. (1996) Clinical evidence of efficient tumor targeting based on single-chain Fv antibody selected from a combinatorial library. Nat. Med. 2, 979−84. (24) Foley, D. A., Swartzentruber, K. G., and Colley, K. J. (2009) Identification of sequences in the polysialyltransferases ST8Sia II and ST8Sia IV that are required for the protein-specific polysialylation of the neural cell adhesion molecule, NCAM. J. Biol. Chem. 284, 15505− 16. (25) Perera, A. D., Lagenaur, C. F., and Plant, T. M. (1993) Postnatal expression of polysialic acid-neural cell adhesion molecule in the hypothalamus of the male rhesus monkey (Macaca mulatta). Endocrinology 133, 2729−35.

the growth and antibody fragments with some entering advanced clinical trials,36 and other binding ligands, recombinant polysialylation could be a valuable technology.

■

CONCLUSIONS This technology, suitably developed, could result in a simple fusion tag that will extend the half-life of antibody fragments. It is proposed that the recombinant attachment of PSA onto heterologous proteins would offer versatility in the modulation and enhancement of the pharmacokinetics of therapeutically useful proteins and demonstrates a new capability of transferring post-translational modifications to heterologous proteins.

■

ASSOCIATED CONTENT

S Supporting Information *

Additional figures. This material is available free of charge via the Internet at http://pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Author Contributions #

Both authors contributed equally to the work.

Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS We thank Daniel Wells from the CBS unit for excellent technical assistance and Imperial College for funding. We wish to thank Prof. K. Colley for polysialyltransferase clones and useful comments on the manuscript, Dr. J. Saffell for guidance and NCAM clones, and Dr. J. Silva for technical help. This work was also supported by the Biotechnology and Biological Sciences Research Council (BBSRC), studentship 01/A2/C/ 07249 (to A.C. and M.P.D.), grant BBF0083091 (to A.D. and S.M.H.) and by the British Heart Foundation (FS/06/069/ 21490).

■

REFERENCES

(1) Holliger, P., and Hudson, P. J. (2005) Engineered antibody fragments and the rise of single domains. Nat. Biotechnol. 23, 1126−36. (2) Gebauer, M., and Skerra, A. (2009) Engineered protein scaffolds as next-generation antibody therapeutics. Curr. Opin. Chem. Biol. 13, 245−55. (3) Fukumura, D., and Jain, R. K. (2007) Tumor microenvironment abnormalities: causes, consequences, and strategies to normalize. J. Cell Biochem. 101, 937−49. (4) Graff, C. P., and Wittrup, K. D. (2003) Theoretical analysis of antibody targeting of tumor spheroids: importance of dosage for penetration, and affinity for retention. Cancer Res. 63, 1288−96. (5) Dougan, M., and Dranoff, G. (2009) The immune response to tumors. In Current Protocols in Immunology, Chapter 20, Unit 20.11, John Wiley & Sons, New Jersey. (6) Jevsevar, S., Kunstelj, M., and Porekar, V. G. (2010) PEGylation of therapeutic proteins. Biotechnol. J. 5, 113−28. (7) Chen, C., Constantinou, A., and Deonarain, M. (2011) Modulating antibody pharmacokinetics using hydrophilic polymers. Expert Opin. Drug Delivery 8, 221−36. (8) Deen, W. M., Bohrer, M. P., and Brenner, B. M. (1979) Macromolecule transport across glomerular capillaries: application of pore theory. Kidney Int. 16, 353−65. 1532

dx.doi.org/10.1021/bc200624a | Bioconjugate Chem. 2012, 23, 1524−1533

Bioconjugate Chemistry

Article

(26) Galuska, S. P., Geyer, R., Mühlenhoff, M., and Geyer, H (2007) Characterization of oligo- and polysialic acids by MALDI-TOF-MS. Anal. Chem. 79, 7161−9. (27) Schier, R., Bye, J., Apell, G., McCall, A., Adams, G. P., Malmqvist, M., Weiner, L. M., and Marks, J. D. (1996) Isolation of high-affinity monomeric human anti-c-erbB-2 single chain Fv using affinity-driven selection. J. Mol. Biol. 255, 28−43. (28) Müller, D., Karle, A., Meissburger, B., Höfig, I., Stork, R., and Kontermann, R. E. (2007) Improved pharmacokinetics of recombinant bispecific antibody molecules by fusion to human serum albumin. J. Biol. Chem. 288, 12650−60. (29) Deen, W. M., Lazzara, M. J., and Myers, B. D. (2001) Structural determinants of glomerular permeability. Am. J. Physiol. Renal Physiol. 281, 579−96. (30) Kubetzko, S., Sarkar, C. A., and Plückthun, A. (2005) Protein PEGylation decreases observed target association rates via a dual blocking mechanism. Mol. Pharmacol. 68, 1439−54. (31) Kuimova, M. K., Bhatti, M., Deonarain, M., Yahioglu, G., Levitt, J. A., Stamati, I., Suhling, K., and Phillips, D. (2007) Fluorescence characterisation of multiply-loaded anti-HER2 single chain Fvphotosensitizer conjugates suitable for photodynamic therapy. Photochem. Photobiol. Sci. 6, 933−9. (32) Scheler, S., Kitzan, M., and Fahr, A. (2011) Cellular uptake and degradation behaviour of biodegradable poly(ethylene glycol-graftmethyl methacrylate) nanoparticles crosslinked with dimethacryloyl hydroxylamine. Int. J. Pharm. 403, 207−18. (33) Anthony, R. M., and Ravetch, J. V. (2010) A novel role for the IgG Fc glycan: the anti-inflammatory activity of sialylated IgG Fcs. J. Clin. Immunol. 30 (Suppl 1), S9−14. (34) Schlapschy, M., Theobald, I., Mack, H., Schottelius, M., Wester, H. J., and Skerra, A. (2007) Fusion of a recombinant antibody fragment with a homo-amino-acid polymer: effects on biophysical properties and prolonged plasma half-life. Protein Eng. Des. Sel. 20, 273−84. (35) Schellenberger, V., Wang, C. W., Geething, N. C., Spink, B. J., Campbell, A., To, W., Scholle, M. D., Yin, Y., Yao, Y., Bogin, O., Cleland, J. L., Silverman, J., and Stemmer, W. P. (2009) A recombinant polypeptide extends the in vivo half-life of peptides and proteins in a tunable manner. Nat. Biotechnol. 27, 1186−90. (36) Nelson, A. L., and Reichert, J. M. (2009) Development trends for therapeutic antibody fragments. Nat. Biotechnol. 27, 331−7.

1533

dx.doi.org/10.1021/bc200624a | Bioconjugate Chem. 2012, 23, 1524−1533