750
Bioconjugate Chem. 2001, 12, 750−756
Prolonged in Vivo Residence Times of Antibody Fragments Associated with Albumin Bryan J. Smith,* Andrew Popplewell, Dee Athwal, Andrew P. Chapman, Sam Heywood, Shauna M. West, Bruce Carrington, Andrew Nesbitt, Alastair D. G. Lawson, Pari Antoniw, Alison Eddelston, and Amanda Suitters Celltech R and D Ltd, 208 Bath Road, Slough, Berks SL1 4EN, UK. Received January 8, 2001; Revised Manuscript Received April 12, 2001
Antibody fragments can be expressed at a high level in microbial systems, but they may have limited therapeutic value because they are rapidly eliminated from the body. We demonstrate here that sitespecific conjugation or binding of bacterially derived Fab′ to the long-lived protein serum albumin allows full retention of the antibody’s binding characteristics while imparting the albumin’s longevity in vivo. In rats the area under the curve for Fab′ conjugated to rat serum albumin was 17-fold greater than for the control of Fab′ conjugated to cysteine. Again, a bispecific F(ab′)2 with specificity for rat serum albumin showed an area under the curve about 8-fold greater than did a F(ab′)2 without specificity to albumin. Genetic fusions of scFv to albumin were similarly long-lived and could be expressed in yeast to provide the basis of a cost-effective production system.
INTRODUCTION
Antibodies are useful therapeutic agents because of their great specificity in binding to other molecules. Large volumes of low cost antibody are required to satisfy the large markets of common or chronic diseases. The ability of cultured cells or transgenic organisms to supply a large market with immunoglobulin at low cost is unproven so far. Expression of whole immunoglobulin G (IgG1)1 has been reported in fungi (1, 2), but not at high level or on an industrial scale. Furthermore, the nonmammalian glycosylation pattern of IgG produced in fungi would be expected to show problems of immunogenicity and compromised function when injected into man. Bacterial systems are a source of low-cost immunoglobulin fragments, such as Fab′, F(ab′)2, or single chain Fv (scFv) (3). These molecules lack the Fc domain and its functions, for example, the interaction with FcRn that gives a long lifetime in vivo (4): the half-life in circulation in mammals of Fab′ or F(ab′)2 is about 1% that of whole IgG (5), and the β-phase half-life (the time taken for half of the molecules in circulation to be eliminated) of Fab′ is about 5% that of whole IgG (6). This rapid elimination of immunoglobulin fragments from the body can be disadvantageous where prolonged presence of the molecule is required, for instance, in order to provide prolonged therapeutic cover for inhibition of inflammatory cytokines. Extension of the half-life of antibody fragments such as Fab′ or F(ab′)2 has been achieved by in vitro conjugation to one or more molecules of poly(ethylene glycol) (6, 7). In the present work, we have investigated an alternative approach to improving in vivo half-life that employs a natural polymer, namely serum * Corresponding author. Tel: +(44) 1753 534655. Fax: +(44)1753 536632. e-mail:
[email protected]. 1 Abbreviations: AUC, area under the curve; BMH, 1,6bismaleimidohexane; FcRN, neonatal Fc receptor; HSA, human serum albumin; IgG, immunoglobulin G; RSA, rat serum albumin; SDS PAGE, sodium dodecyl sulfate polyacrylamide electrophoresis; scFv, single chain Fv; TNF, tumour necrosis factor.
albumin. Albumin is abundant in both vascular and extravascular compartments and has a half-life of about 19 days (8) in man (similar to that of IgG1 at about 21 days (5)), though it is less in other species (about 2 days in rats, for example (8)). Albumin does not possess the noted ability of antibodies to specifically bind ligands, particularly those of high molecular weight. We have investigated the potential of association of immunoglobulin fragments with albumin in three different ways: chemical cross-linking of the separate, lowcost components, albumin, and Fab′; genetic fusion, which would allow a one-step synthesis of the combination; noncovalent binding of an antibody species to endogenous albumin, an approach that could lead to a relatively small, low-cost protein therapeutic. The concept was validated since all three approaches were successful: each albumin-associated immunoglobulin fragment retained its antigen-binding capability while showing extended in vivo half-life. EXPERIMENTAL SECTION
Preparation of Proteins. Rat serum albumin (RSA) was obtained from Sigma, St. Louis, MO (code A6272). Anti-RSA IgG F(ab′)2 was generated by pepsin (Sigma, St. Louis, MO) digestion (9) of IgG that was prepared from rabbit antiserum (ICN Biomedicals, Costa Mesa, CA) by affinity chromatography on (i) Gammabind plus Sepharose (following the instructions of the manufacturer, Amersham Pharmacia Biotech, Little Chalfont, UK, with elution of adsorbed protein by a buffer of acetic acid, 0.5 M, made to pH 3 by addition of sodium hydroxide), and then (ii) RSA-Sepharose (prepared by linking RSA to CNBr-activated’s instructions). For RSASepharose affinity chromatography, the sample (at pH 7) was loaded onto the column, which was preequilibrated in phosphate-buffered saline, and the RSA-binding component eluted in acetate buffer (0.5 M acetic acid, brought to pH 3 by addition of sodium hydroxide). Fractions eluted from each column were immediately neutralized by addition of trizma (2 M, pH 8). Control F(ab′)2 was
10.1021/bc010003g CCC: $20.00 © 2001 American Chemical Society Published on Web 08/16/2001
Antibody−Albumin Conjugates
made by the same method, from rabbit anti-keyhole limpet haemocyanin antiserum. Recombinant anti-human Tumor Necrosis Factor R (TNF) and anti-cell surface marker F(ab′)2 and Fab′ molecules were produced in E. coli, as described previously (6). Thiols were assayed as described previously (6). Preparation of Conjugates. Bispecific F(ab′)2. Freshly reduced anti-RSA Fab′ was incubated for 2 h at 21 °C in PE buffer (sodium phosphate, 0.1 M, pH 6, 2 mM ethylenediaminetetraacetate) with an 80-fold molar excess of 1,6-bismaleimidohexane (BMH, Perbio Science UK Ltd, Tattenhall). Excess BMH was removed by Sephadex G25M size exclusion chromatography (“PD10”, Amersham Pharmacia Biotech, Little Chalfont, UK) in PE buffer, as per manufacturer’s instructions. The derivatized anti-RSA Fab′ was then incubated for 20 h at 21 °C with freshly reduced anti-TNF Fab′ at a molar ratio of anti-RSA Fab′:anti-TNF Fab′::1:3.1. Purification of the bispecific anti-RSA-anti-TNF F(ab′)2 heterodimer was purified by affinity chromatography on RSA immobilized on Sephadex (as described above), followed by size exclusion chromatography in a buffer of 0.2 M sodium phosphate, pH 7, on Zorbax GF250 (Agilent Technologies, Palo Alto, CA), using an HP1090 HPLC (formerly HewlettPackard, currently Agilent Technologies, Palo Alto, CA), to remove anti-RSA Fab′. Overall yield (nonoptimized) was about 9%. Control monospecific anti-TNF F(ab′)2 homodimer was made in an analogous way and purified by gel filtration on Zorbax GF250 in a buffer of 0.2 M sodium phosphate, pH 7. RSA-Fab′. RSA was reduced either by (a) incubation for 50 min at 37 °C in 5 mM 2-mercaptoethylamine (Sigma, St Louis, MO) in PE buffer, or (b) incubation for 40 min at 37 °C in a 3-fold molar excess of dithiothreitol (Sigma, St Louis, MO) in sodium acetate, 0.1 M, pH 5.9. Reduced RSA was reacted in PE buffer with a 21-fold excess of BMH in PE buffer for 100 min, 21 °C. Approximately 80% of reduced RSA reacted with BMH under these (nonoptimized) conditions. Derivatized RSA was reacted for 2 h at 21 °C with freshly reduced Fab′ in the molar ratio, derivatized RSA:reduced Fab′::1:1.4. The reaction mixture was first subjected to chromatography on Gammabind plus Sepharose (described above), preequilibrated in PE buffer. Adsorbed protein was eluted by acetic acid, 0.5 M, made to pH 3 by addition of sodium hydroxide. Eluted fractions were immediately neutralized by addition of 2 M trizma, pH 8. As expected of this protein G matrix, nonconjugated Fab′ bound and was eluted by the pH 3 buffer, whereas the nonconjugated RSA did not bind at all and emerged in the flow-through. Conjugation of a single Fab′ to one RSA molecule clearly affected its binding to the protein G on the matrix, for the conjugate emerged in the flow-through, just slightly later than (and overlapping with) the unconjugated RSA. More free RSA was removed from the conjugate by both: (i) a further cycle of chromatography on Gammabind plus (as above); (ii) gel permeation chromatography on a Zorbax GF250 HPLC column of size (described above). Overall yield (nonoptimized) was about 5%. Control anti-TNF Fab′-cys conjugate was prepared by a similar approach: (i) derivatization of Fab′ by BMH; (ii) incubation with a 22-fold molar excess of cysteine (Sigma, St Louis, MO). Purification was by chromatography on a 1 mL volume Mono S column using an FPLC apparatus (Amersham Pharmacia Biotech). Adsorbed protein were then eluted in a gradient of 0 to 250 mM sodium chloride in sodium acetate, 50 mM, pH 4.5. Overall yield (nonoptimized) was about 28%.
Bioconjugate Chem., Vol. 12, No. 5, 2001 751
Figure 1. Structure of trimaleimide cross-linking agent.
Bivalent Conjugate RSA-F(ab′)2. This conjugate was made by use of the trimaleimide cross-linking agent NR-[1,4,7-tris(carboxymethyl)-1,4,7-triaza-2-cyclononylbutylaminosuccinoyl]-Nω-(maleimidopropanoyl)lysyl-Nω-(maleimidopropanoyl)lysyl-Nω-(maleimidopropanoyl)lysinamide, illustrated in Figure 1. It was synthesized according to the method described previously (10), starting from 2-(4aminobutyl)-1,4,7-triazacyclononane-1,4,7-triyltriacetic acid (11) and N6-[3-(2,5-dihydro-2,5-dioxo-1H-pyrrol-1-yl)-1oxopropyl]-L-lysyl-N6-[3-(2,5-dihydro-2,5-dioxo-1H-pyrrol1-yl)-1-oxopropyl]-L-lysyl-N6-[3-(2,5-dihydro-2,5-dioxo1H-pyrrol-1-yl)-1-oxopropyl]-L-lysinamide mono(trifluoroacetate). Anti-cell surface marker Fab′ PE buffer was reacted with this trimaleimide cross-linking agent for 3 h at 21 °C at a molar ratio cross-linker:Fab′::1:2.56. Monomer, dimer, and trimer Fab′conjugates were resolved by chromatography on SP-Sepharose HP (Amersham Pharmacia Biotech, Little Chalfont, UK), being eluted by a gradient of 0 to 250 mM sodium chloride in 50 mM acetate, pH 4.5. The dimer (approximately 30% of the total material) was then reacted with freshly reduced RSA at molar ratio of 1:1, for 24 h at 21 °C. The RSA-F(ab′)2 was finally prepared by affinity chromatography on Gammabind plus Sepharose to remove free albumin, then on Blue Sepharose, to remove unconjugated F(ab′)2. The yield of this (nonoptimized) procedure was about 12%. A control molecule of anti-cell surface marker F(ab′)2 (containing no RSA) was prepared by conjugating two freshly reduced Fab′ molecules with BMH, in a manner analogous to conjugation of anti-TNF Fab′ (see above). Preparation of scFv-HSA Fusion Proteins. Oligonucleotides were designed that coded for fusions of scFv with the N-terminus of mature Human Serum Albumin (HSA), either directly (scFv-HSA) or via linkers positioned to physically separate the scFv from the albumin moiety. The linkers were a sequence of 3 repeats of (glycine)4-serine (scFv-G4S-HSA); a sequence derived from the flexible upper (N-terminal) end of the human IgG1 hinge region, (scFv-UHL-HSA); or the C-domain from human Ig-kappa light chain (scFv-CK-HSA). The orientation of variable domains in the scFv was VL-VH, except for scFv-CK-HSA, in which the order was VH-VL. The scFv sequence was immediately preceded by a spacer of (glutamic acid-alanine)2 as recommended by Sreekrishna et al. (12) to alleviate possible problems of steric hindrance of pro-sequence-processing protease. An “EasySelect Pichia Expression” kit (Invitrogen, Carlsbad), with the plasmid pPIC, was used as per manufacturer’s
752 Bioconjugate Chem., Vol. 12, No. 5, 2001
instructions to express each type of fusion protein in shake flask culture. Each fusion protein was secreted into 30-40 mL of medium. Supernatant from each culture was adjusted to pH 7 by addition of 2 M sodium hydroxide, and applied to Blue Sepharose (Amersham Pharmacia Biotech, Little Chalfont, UK). Each construct bound and was eluted by sodium thiocyanate, 0.2 M in phosphate-buffered saline. Each protein was concentrated and the buffer exchanged to phosphate-buffered saline by use of an Amicon stirred cell concentrator (Millipore, Bedford, MA). Analytical Procedures. Sodium dodecyl sulfate polyacrylamide electrophoresis (SDS PAGE) utilized precast SDS gels (Novex, San Diego, CA), run as per manufacturer’s instructions. Protein bands were quantified after staining by Coomassie BBG (13). Protein sequencing was performed on an Applied Biosystems (Warrington, UK) model 470A or Procise 492 protein sequencer. A combination of approximations of molecular weight (by SDS PAGE) and quantity (by N-terminal sequencing) allowed determination of the stoichiometry of components in conjugates. Surface plasmon resonance study of interactions with ligand was performed a BIACORE 2000 (Biacore AB, Stevenage, UK) on a CM5 Sensor Chip, with kinetic parameters calculated by use of BIAevaluation 3.0 software. scFv-HSA fusion protein was bound to the chip surface by amine coupling chemistry to approximately 400 Response Units. The ligand, recombinant human TNFR, was passed across the immobilized molecule at concentrations between 0 nM to 120 nM. RSA-Fab′ conjugate was captured by goat anti-human IgG F(ab′)2 fragment specific antibody (Jackson ImmunoResearch Labs Inc, West Grove) that was immobilized on the chip surface, prior to application of ligand. For the bispecific conjugate, anti-RSA-anti-TNF, affinity for TNF was assayed by capture of the conjugate by AffiniPure F(ab′)2 fragment, goat anti-rabbit IgG, F(ab′)2 fragment specific (Jackson ImmunoResearch Labs Inc, West Grove) that was immobilized on the chip surface, prior to application of the TNFR ligand. To assay the affinity of the bispecific conjugate for RSA, the ligand RSA was immobilized on the chip surface, and conjugate solution was passed over that. RSA-F(ab′)2 and F(ab′)2 control were fluoresceinated prior to Scatchard analysis of binding, in which they were incubated with cells for 3 h at 4 °C for 3 h, before analysis by fluorescence activated cell sorting, using a FACScan flow cytometer (Becton Dickinson, Franklin Lakes). Pharmacokinetic Analysis of the Conjugate. Proteins were labeled at the -amino groups of lysyl residues, using 125I-labeled Bolton and Hunter reagent (Amersham International, Little Chalfont, UK) as follows. A 300 µg amount of protein dissolved in 300-370 µL of 0.1 M borate, pH 8, was mixed with 20 µL of Bolton and Hunter solution in propan-2-ol (containing 9 MBq of 125I) and incubated at 21 °C for 15 min. The reaction was then quenched by addition of 60 µL solution of glycine, 1 M, in borate, 0.1 M, pH 8. After approximately 5 min reaction at 21 °C, the labeled protein was separated from other reagents by chromatography on Sephadex G25M using a PD10 column (Amersham Pharmacia, used as per manufacturer’s instructions). In doing so, the buffer was exchanged for phosphate buffered saline. The specific activity of each preparation was calculated from estimates of protein concentration (estimated by absorption at 280 nm) and of radioactivity and was typically in the
Smith et al.
range 0.45 to 0.54 µCi/µg. The radiolabeled samples were used directly after labeling. A 20 µg amount of 125I-labeled or 180 µg of unlabeled protein in solution in phosphate-buffered saline was injected into a tail vein of male Wistar rats (approximately 250 g each). Blood samples were taken periodically from the tail artery, with plasma prepared by heparinization and centrifugation. Radioactivity was detected in whole blood by gamma counting (using a Cobra II auto gamma, from Canberra Packard, Ontario). The percent injected dose (%ID) was calculated for each individual rat, based on standards and expressed as %ID/ mL total blood volume. Data were analyzed by WinNonlin software (Pharsight Corp., Mountain View, CA). For this, a two-compartment model was used, except for Fab′cys monitored by ELISA, where too few data were acquired due to rapid elimination and where a onecompartment model was used instead. Phosphorimaging and quantification of radiolabeled samples on SDS PAGE utilized a Cyclone system with Optiquant software (Canberra Packard, Ontario). This enabled examination of the quality of radiolabeled protein in plasma throughout the experiments. ELISA’s used microtiter plates coated with the antigen recombinant human TNFR (Strathmann Biotech GMBH, Hannover). Detection was by rabbit anti-RSA (Cappel, obtained from ICN, Costa Mesa, CA) followed by goat anti-rabbit immunoglobulin Fc-peroxidase conjugate (Jackson ImmunoResearch Labs Inc, West Grove), or by goat anti-human kappa light chain (Southern Biotechnology Associates, Inc., Birmingham, AL) followed by donkey anti-goat immunoglobulin (H+L)-peroxidase conjugate (Jackson ImmunoResearch Labs Inc, West Grove). Quantification was by absorption at 630 nm, generated from 3,2′5,5′-tetramethylbenzidine (R and D Systems Europe, Abingdon, UK; 120 µM in 10 mM acetate, pH 6) as a result of the activity of peroxidase. Anti-TNF activity was assayed by the inhibition of killing of monolayered mouse L929 cells by recombinant human TNFR (Strathmann Biotech GMBH, Hannover, NH) in normal RPMI 1640 plus glutamine and 10% (v/ v) foetal calf serum (both from Life Technologies, Paisley, UK). Cell survival was quantified as follows: the brown color produced by surviving cells from 50 µg/mL 3-[4,5dimethylthiazol-2-yl]-2,5-diphenyltetrazolium bromide (Sigma, St Louis, MO) in the culture medium at 37 °C for 4 h was extracted (by a solution of SDS, 20%, w/v, in 50% v/v in water, adjusted to pH 4.7 by addition of 50% acetic acid in 1 M hydrochloric acid) and quantified by absorbance at 570 nm; absorption at 630 nm was subtracted, and compared to standards from known amounts of TNF added to samples of cells, in the absence of TNF-neutralizing activity. RESULTS
Characterization of Conjugates. The cysteinyl residue at position 34 of albumin is commonly engaged in formation of a mixed disulfide with molecules such as glutathione or free cysteine, but suitable reduction conditions can liberate that thiol without disrupting disulfides present elsewhere in the molecule (8). Reduction of albumin in such conditions increased the amount of free thiol from 0.15 in the original albumin preparation to between 0.95 and 0.97 mol of thiol per mol of albumin. Albumin dimers were not formed during derivatization by BMH, because the free cysteinyl 34 with which the BMH is presumed to react lies within a cleft in the albumin structure, and the 16.1 angstrom spacer arm of
Antibody−Albumin Conjugates
Bioconjugate Chem., Vol. 12, No. 5, 2001 753 Table 1. Ligand-Binding Characteristics of Conjugates and Fusion Proteinsa
IgG F(ab)2 RSA-Fab′ conjugate scFv-HSA scFv-G4S-HSA scFv-UHL-HSA scFv-CK-HSA anti-RSA-anti-TNF F(ab′)2 vs RSA vs TNF F(ab′)2 (i) (ii) RSA-F(ab′)2 (i) (ii)
Figure 2. Reduced SDS PAGE of conjugates and fusion proteins. Stain: Coomassie Brilliant Blue G in perchloric acid. Approximate apparent molecular weights (indicated) were determined by co-electrophoresis of standard proteins. Samples: a ) RSA-anti-TNF Fab′ conjugate; b ) RSA-anti-TNF F(ab′)2 conjugate; c ) bispecific anti-TNF-anti-RSA F(ab′)2; d ) scFv-HSA fusion protein; e ) scFv-G4S-HSA fusion protein; f ) scFv-UHL-HSA fusion protein; g) scFv-Cκ-HSA fusion protein.
the bifunctional BMH was too short to reach from within the cleft of one albumin molecule and penetrate the depth of the equivalent cleft in a second albumin molecule. Again, the reducing conditions used here produced Fab′ that contained only one free cysteinyl residue, near the C-terminus of the Fd chain (12). It was therefore possible to cross-link specific thiols in albumin and Fab′ molecules, to make conjugates with defined stoichiometry. Methods for production of the various conjugates were not optimized, but provided sufficient material for subsequent experiments. SDS PAGE (Figure 2) and N-terminal sequencing of preparations and of protein bands on blots from SDS PAGE confirmed that 1:1 molar ratios were achieved in conjugates of anti-TNF Fab′ with RSA, and anti-TNF Fab′ with anti-RSA Fab′. Similarly, one anti-TNF F(ab′)2 was covalently attached to one RSA molecule. Fab′ molecules were conjugated through their Fd chain, as seen by generation of free light chain by reducing conditions in SDS PAGE, seen as a band of about 30 kDa apparent molecular weight. The H chain dimer from the bispecific F(ab′)2 appeared as a band of 40 to 50 kDa apparent molecular weight or as a species of very approximately 150 kDa apparent molecular weight when linked to RSA. The H chain-RSA from the Fab-RSA conjugate migrated as an approximately 100 kDa band but was accompanied by lesser a quantity of H chainRSA dimer material (as can sometimes be seen upon electrophoresis of preparations of albumin alone - not shown), and some (putative) free albumin at an apparent molecular weight of about 65-70 kDa. Four types of scFv-HSA fusion proteins were secreted into the medium of Pichia pastoris cultures. A single step of Blue Sepharose chromatography of culture supernatant provided fusion proteins of between 75% and 91% purity (see Figure 2), with mature HSA as the other component (presumably arising from proteolysis of linker). Yields of prepared proteins as µg/mL of culture supernatant (mean of two experiments) were: scFv-HSA, 9.0 µg/mL; scFv-G4S-HSA, 8.5 µg/mL; scFv-UHL-HSA, 7.5 µg/mL; scFv-CK linker, 7.8 µg/mL. Neither the nature of the linker nor the order of the variable domains in the scFv portion significantly affected levels of expression of
ka, 105 M-1 s-1
kd, 10-4 s-1
KD, 10-10 M
3.63 2.79 3.88 7.67 6.58 6.64 7.09
1.41 0.56 1.65 1.11 1.32 1.52 0.95
3.88 2.01 4.25 1.45 2.29 2.29 1.34
0.23 5.05 -
6.08 0.40 -
260.00 0.78 120 30 80 30
a Data were generated by surface plasmon resonance analysis, except for F(ab′)2 and RSA-F(ab′)2, which were determined by Scatchard analysis of binding to ligand on cell surfaces, which was biphasic. (i and ii were the data for the two phases.)
Table 2. Pharmacokinetics of Albumins, Conjugates, Fusions, and Other Immunoglobulins in Rats assay RSA F(ab′)2 Fab′-cys RSA-Fab′ conjugate bispecific F(ab′)2 HSA scFv-HSA scFv-G4S-HSA scFv-UHL-HSA Fab′-cys RSA-Fab′ conjugate RSA-Fab′ conjugate RSA-Fab′ conjugate
125I 125I 125I 125I 125I 125I 125I 125I 125I
ELISA1 ELISA1 ELISA2 L929
t1/2R (h) t1/2β (h) 4.1 1.4 1.0 4.6 9.2 0.8 2.3 1.0 3.0 nd 1.3 0.8 2.7
49.1 15.9 31.4 39.6 42.5 14.8 16.6 15.2 16.5 0.7 21.6 19.5 24.7
AUC (0-∞) 2637 h*%id 213 h*%id 94 h*%id 1609 h*%id 1672 h*%id 1272 h*%id 1164 h*%id 1282 h*%id 1225 h*%id 4 h*µg/mL 824 h*µg/mL 780 h*µg/mL 975 h*µg/mL
a Data were analyzed using WinNonlin and a two-compartment model, except for Fab′-cys when a one-compartment model was used. AUC (0-∞): area under the plasma concentration curve, or plasma residence time. 125I: proteins monitored by gamma counting. ELISA1: TNF-binding material detected by anti-L chain. ELISA2: TNF-binding material detected by anti-albumin. L929: quantified by neutralization of TNF activity. For 125I assay, n ) 5-7, others n ) 2. Data shown for RSA and F(ab′)2 are means of four and two experiments, respectively.
fusion protein. N-Terminal protein sequencing revealed that the (Glu-Ala)2 sequence lying N-terminal to the scFv domain was not removed by aminopeptidase activity (cf. ref 12). Anti-TNF Fab′ retained ligand binding activity in all conjugates and fusions (Table 1, and ELISA and L929 assay data, below). Similarly, affinity for cell surface ligand was retained by both the RSA-F(ab′)2 and the corresponding control F(ab′)2 conjugate. Scatchard analysis showed the presence of both low and high affinity binding sites (Table 1) with 40 000 and 20 000 RSAF(ab′)2 binding sites per cell, respectively, and 125 000 and 50 000 F(ab′)2 binding sites per cell, respectively. Thus, association of Fab′ with albumin did not decrease ligand binding affinity, though the 67 kDa albumin moiety hindered access to some binding sites present on a cell surface. Pharmacokinetic Analysis of the Conjugate in Rat Plasma. As judged by phosphorimaging, all 125Ilabeled conjugates were stable in vivo (that is, the quality of the labeled protein remained constant throughout the experiment), with one exception. The exception was the 125I-RSA-Fab′ conjugate, in which some instability was
754 Bioconjugate Chem., Vol. 12, No. 5, 2001
Smith et al.
Figure 3. Pharmacokinetics of 125I-labeled RSA, conjugates, and other immunoglobulins in rats. 125I detected in plasma samples by gamma-counting. (9) RSA; (() RSA-Fab′; (b) Fab′-cys; (2) F(ab′)2. For each, n ) 6. Error bars: sem.
Figure 4. Pharmacokinetics of unlabeled RSA-Fab′ conjugate and Fab′-cys in rats. Proteins quantified in plasma samples by ELISA in which TNF-binding protein was detected by anti-L chain: (2) RSA-Fab′; (1) Fab′-cys. Proteins quantified by ELISA in which TNF-binding protein was detected by anti-albumin: (b) RSA-Fab′. Protein quantified by assay of neutralization of L929 cell killing by TNFR: (9) RSA-Fab′; (() Fab′-cys. Results for the three types of assay of RSA-Fab′ overlay each other, and results for the two types of assay for Fab′-cys likewise overlay each other. For each, n ) 2. Error bars ) sem.
noted in vivo, whereby a species was generated with an apparent molecular weight similar to that of RSA on SDS PAGE. Experiments (not shown) indicated that this instability was artificial, coinciding with the presence of the covalently attached 125I or to some other aspect of the labeling process. Nevertheless, phosphorimage analysis showed that even after 144 h in vivo the intact conjugate still represented about 30% of all labeled protein present. Data from phosphorimage analysis of rat plasma samples were used to correct gamma counting data from in vivo experiments with 125I-labeled RSA-Fab′, to reflect the quantity of 125I in intact, labeled conjugate only. In rat plasma the intact 125I-RSA-Fab′ conjugate had a residence time, or an area under the curve (AUC), that was 17-fold greater than the Fab′-cys control in the same experiment in vivo (Figure 3 and Table 2). The longevity of the RSA-Fab′ conjugate in vivo was confirmed by experiments using conjugate that was not labeled, by ELISA format or by L929 survival assay (Figure 4, Table
2). By these methods, the RSA-Fab′ AUC was approximately 200-fold greater than that of the Fab′-cys, which was very rapidly eliminated. Similarly, binding to endogenous albumin by 125Ibispecific F(ab′)2 caused a 6-fold increase in AUC over that of the control in the same experiment, unispecific anti-TNF F(ab′)2, which did not bind albumin (Figure 5, Table 2). The residence time for the bispecific F(ab′)2 was approximately 70% that of RSA in the same experiment, though their t1/2β values were similar. The loss of bispecific F(ab′)2 in the first few minutes of the experiment was probably not due to failure to bind to endogenous albumin, for experiments (not shown) with anti-RSA Fab′ showed that in vitro incubation with plasma prior to injection did not alter the residence time in vivo. The residence times of three 125I-labeled fusion proteins tested (scFv-HSA, scFv-G4S-HSA, and scFv-UHL-HSA) were similar to those of HSA and approximately 12-fold greater than those of 125I-Fab′-cys control in the same
Antibody−Albumin Conjugates
Figure 5. Pharmacokinetics of 125I-labeled RSA, bispecific and unispecific F(ab′)2 in rats. 125I detected in plasma samples by gamma counting. (9) RSA; (2) anti-RSA-anti-TNF bispecific F(ab′)2; (1) anti-TNF unispecific F(ab′)2. For each, n ) 5-7. Error bars ) sem.
Figure 6. Pharmacokinetics of 125I-labeled HSA and fusion proteins in rats. 125I detected in plasma samples by gamma counting. (() HSA; (b) scFv-HSA; (9) scFv-G4S-HSA; (2) scFvUHL-HSA. For each, n ) 6. Error bars ) sem, but are small and not readily visible on the plot. The four curves overlay each other.
experiment (Figure 6, Table 2). In rats the residence time of scFv-HSA fusions was shorter than of RSA-Fab′ conjugate, but this was taken to be because the albumin moiety dictates the longevity, and the heterologous HSA was only about half as persistent as the homologous RSA. DISCUSSION
Data presented here show that association of immunoglobulin fragments with serum albumin, whether by conjugation, fusion or noncovalent binding, results in extended persistence in plasma in vivo, approaching that of albumin itself. The albumin moiety determined longevity, and in man the t1/2β of serum albumin is about 19 days in man (8) (compared to 2 to 2.5 days in rat (8)). Accordingly, in man an immunoglobulin-albumin molecule might persist for about the same time as would whole IgG1 (t1/2β ) about 21 days in man (5)). Extended residence time of an agent in vivo can be beneficial in providing extended therapeutic cover. This goal of extending the half-life of Fab′ has previously been achieved by attachment of poly(ethylene glycol), which gave 7- to 21-fold increase in AUC, depending on the amount of the
Bioconjugate Chem., Vol. 12, No. 5, 2001 755
polymer that was attached (7). We show here that albumin provides an alternative to poly(ethylene glycol) as a conjugation partner. Any effect due to the albumin moiety, administered as part of a therapeutic agent, would be expected to be negligible: the role of albumin is in transportation of various molecules (such as steroid hormones, bile pigments and fatty acids) in circulation and in extravascular fluids and is normally present in plasma at about 50 mg/mL, in great excess over any administered therapeutic. In the present work defined conjugates were made in vitro by cross-linking through a specific site in each component, particularly a site in the Fab′ molecule near its C-terminus, such that the Fab′ ligand-binding properties are not adversely affected (6). Both Fab′(s) and albumin components can be produced at low cost by microbial systems and therefore provide the basis of an economic therapeutic molecule. Additionally, we have illustrated an alternative possibility, namely that of genetically engineering novel albumin-immunoglobulin fusion proteins and expressing them in yeast cells. This approach has the extra advantage, compared to conjugation of immunoglobulin to albumin or PEG, that no in vitro conjugation step is required in the production process, which is therefore simpler. Expression of these fusion molecules in yeast was not fully optimized here, but it has considerable potential if it can approach the level of albumin production in yeast, which can be at a level of grams per liter (15, 16, 17). This level of expression makes industrial production of pharmaceutical grade protein economically feasible (16). The potential problems of antigenicity that surround yeast oligosaccharides when injected into mammals are avoided by excluding glycosylation sites from the fusion protein sequencesglycosylation is not required for prolonged serum half-life in this case, unlike for whole IgG (18). Bispecific immunoglobulin that can bind to endogenous serum albumin provides a third strategy. Once injected, anti-albumin Fab′ binds rapidly to endogenous albumin, whose concentration in plasma is about 0.6 mM. From the observed association and dissociation rates for antiRSA (Table 1) it may be calculated that at equilibrium in serum only 4.4 × 10-3 % of bispecific F(ab′)2 would remain unbound to albumin. We have exemplified the bispecific molecule approach with a molecule made by chemical conjugation of components produced economically in microbial systems, but other forms of bispecific antibodies such as those reviewed by Cao and Suresh (19) would be expected to function in the same way. Possibilities exist for further elaboration of the principles described here. These include use of alternative proteins to albumin, generation of both polyvalent and polyspecific immunoglobulins (by either chemical crosslinking or genetic fusion, or a combination of both), and engineering of additional function into the protein or the cross-linking agents. For instance, the trimaleimide cross-linker used here includes a macrocycle moiety capable of chelating any of a range of metal ions such as might be used for therapeutic or diagnostic purposes. The macrocycle illustrated (Figure 1) is able to chelate Indium. In summary, we have shown that association of fragments or derivatives of immunoglobulin with albumin provides the basis of a system for production of a potentially low cost but long-lived therapeutic immunoglobulin.
756 Bioconjugate Chem., Vol. 12, No. 5, 2001 LITERATURE CITED (1) Horwitz, A. H., Chang, C. P., Better, M., Hellstrom, K. E., Robinson, R. R. (1988). Secretion of functional antibody and Fab fragment from yeast cells. Proc. Natl. Acad. Sci. U.S.A. 85, 8678-82. (2) Ogunjimi, A. A., Chandler, J. M., Gooding, C. M., Recinos, A., III, Choudary, P. V. (1999). High-level secretory expression of immunologically active intact antibody from the yeast Pichia pastoris. Biotechnol. Lett. 21, 561-567. (3) Better, M., Chang, C. P., Robinson, R. R., Horwitz, A. H. (1988). Escherichia coli secretion of an active chimeric antibody fragment. Science 240, 1041-1043. (4) Medesan, C., Matesoi, D., Radu, C., Ghetie, V., Ward, E. S. (1997) Delineation of the amino acid residues involved in transcytosis and catabolism of mouse IgG1. J. Immunol. 158, 2211-2217. (5) Waldmann, T. A., and Strober, W. (1969). Metabolism of immunoglobulins. Prog. Allergy 13, 1-110. (6) Chapman, A. P., Antoniw, P., Spitali, M., West, S., Stephens, S., and King, D. J. (1999). Therapeutic antibody fragments with prolonged in vivo half-lives. Nat. Biotechnol. 17, 780783. (7) Kitamura, K., Takahashi, T., Yamaguchi, T., Noguchi, A., Noguchi, A., Takashina, K., Tsurumi, H., Inagake, M., Toyokuni, T., Hakomori, S. (1991). Chemical engineering of the monoclonal antibody A7 by poly(ethylene glycol) for targeting cancer chemotherapy. Cancer Res. 51, 4310-4315. (8) Peters, T., Jr. (1985). Serum albumin. Adv. Prot. Chem. 37, 161-245. (9) Parham, P. (1983). On the fragmentation of monoclonal IgG1, IgG2a, and IgG2b from BALB/c mice. J. Immunol. 131, 2895-2902. (10) King, D. J., Turner, A., Beeley, N. R. A., Millican, T. A. (1992). Tri- and tetra-valent monospecific antigen-binding proteins. International Patent WO 92/22583. (11) Cox, J. P. L., Craig, A. S., Helps, I. M., Jankowski, K. J., Parker, D., Eaton, M. A. W., Millican, T. A., Millar, K., Beeley, N. R. A., Boyce, B. A. (1990). Synthesis of C- and Nfunctionalised derivatives of 1,4,7-triazacyclononane-1,4,7-
Smith et al. triyltriacetic acid (NOTA), and 1,4,7,10-tetra-azacyclododecane1,4,7,10-tetrayltetraacetic acid (DOTA), and diethylenenetriaminepenta-acetic acid (DTPA): bifunctional complexing agents for the derivatisation of antibodies. J. Chem. Soc., Perkin Trans. 1 2567-2576. (12) Sreekrishna, K., Brankamp, R. G., Kropp, K. E., Blankenship, D. T., Tsay, J. T., Smith, P. L., Wierschke, J. D., Subramaniam, A., Birkenberger, L. A. (1997). Strategies for optimal synthesis and secretion of heterologous proteins in the methylotrophic yeast Pichia pastoris. Gene 190, 55-62. (13) Smith, B. J. (1996) Quantification of proteins by staining in polyacrylamide gels, in The Protein Protocols Handbook (Walker, J. M., Ed.) pp 167-172, Humana, Totowa, NJ. (14) King, D. J., Turner, A., Farnsworth, A. P., Adair, J. R., Owens, R. J., Pedley, R. B., Baldock, D., Proudfoot, K. A., Lawson, A. D., Beeley, N. R. Millar, K., Millican, T. A., Boyce, B. A., Antoniw, P., Mountain, A., Begent, R. M. J., Shocat, D., Yarranton, G. T. (1994). Improved tumor targeting with chemically cross-linked recombinant antibody fragments. Cancer Res. 54, 6176-6185. (15) Barr, K. A., Hopkins, S. A., Sreekrishna, K. (1992). Protocol for efficient secretion of HSA developed from Pichia pastoris. Pharm. Eng. 12, 48-51 (16) Fleer R., Yeh, P., Amellal, N., Maury, I., Fournier, A., Bacchetta, F., Baduel. P., Jung, G., L′Hote, H., Becquart, J., Fukuhara, H., Mayaux, J. F. (1991) Stable multicopy vectors for high-level secretion of recombinant human serum albumin by Kluyveromyces yeasts. Bio/Technology 9 968-75. (17) Cregg, J. M., Vedvick, T. S., Raschke, W. C. (1993). Recent advances in the expression of foreign genes in Pichia pastoris. Bio/Technology 11, 905-10. (18) Wawrzynczak, E. J., Cumber, A. J., Parnell, G. D., Jones, P. T., Winter, G. (1992). Blood clearance in the rat of a recombinant mouse monoclonal antibody lacking the N-linked oligosaccharide side chains of the CH2 domains. Mol. Immunol. 29, 213-220. (19) Cao, Y., Suresh, M. R. (1998) Bispecific antibodies as novel bioconjugates. Bioconjugate Chem. 9, 635-644.
BC010003G