PEDS Advance Access originally published online on July 25, 2006
Protein Engineering Design and Selection 2006 19(10):461-470; doi:10.1093/protein/gzl031
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
A highly stable polyethylene glycol-conjugated human single-chain antibody neutralizing granulocyte-macrophage colony stimulating factor at low nanomolar concentration
1 Micromet AG Staffelseestr. 2, 81477 Munich, Germany 2 Enzon Pharmaceuticals 20 Kingsbridge Road, Piscataway, NJ 08854-3969, USA 3 LEO Pharma A/S, Industriparken 55 DK-2750 Ballerup, Denmark
4To whom correspondence should be addressed. E-mail: tobias.raum{at}micromet.de
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and methodsResultsDiscussionAcknowledgementsReferences |
|---|
GM-CSF (granulocyte-macrophage colony stimulating factor) plays a central role in inflammatory processes. Treatment with antibodies neutralizing murine GM-CSF showed significant therapeutic effects in mouse models of inflammatory diseases. We constructed by phage display technology a human scFv, which could potently neutralize human GM-CSF. At first, a human VL repertoire was combined with the VH domain of a parental GM-CSF-neutralizing rat antibody. One dominant rat/human scFv clone was selected, neutralizing human GM-CSF with an IC50 of 7.3 nM. The human VL of this clone was then combined with a human VH repertoire. The latter preserved the CDR 3 of the parental rat VH domain to retain binding specificity. Several human scFvs were selected, which neutralized human GM-CSF at low nanomolar concentrations (IC50
2.6 nM). To increase serum half-life, a branched 40 kDa PEG-polymer was coupled to the most potent GM-CSF-neutralizing scFv (3077) via an additional C-terminal cysteine. PEG conjugation had a negligible effect on the in vitro neutralizing potential of the scFv, although it caused a significant drop in binding affinity owing to a reduced on-rate. It also significantly increased the stability of the scFv at elevated temperatures. In mouse experiments, the PEGylated scFv 3077 showed a significantly prolonged elimination half-life of 59 h as compared with 2 h for the unconjugated scFv version. PEGylated scFv 3077 is a potential candidate for development of a novel antibody therapy to treat pro-inflammatory human diseases.
Keywords: GM-CSF/neutralization/PEGylation/phage display/single chain antibody
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionAcknowledgementsReferences |
|---|
GM-CSF (granulocyte-macrophage colony stimulating factor) was originally described as a potent stimulus of granulocyte and macrophage precursor cells in vitro (Sheridan and Metcalf, 1973
). The cytokine is an 
23 kDa glycoprotein with a four 
-helical bundle structure (Diederichs et al., 1991). It binds to a heterodimeric receptor composed of subunits belonging to the type 1 cytokine receptor family (Shibuya et al., 1991; Hercus et al., 1994). Further experiments have demonstrated that GM-CSF also stimulates mature granulocytes and macrophages as well as antigen-presenting dendritic cells (Inaba et al., 1992) to increase their anti-infective potential. Gene knockout experiments in mice suggested that the major physiological role of GM-CSF is to maintain or stimulate the functional activity of mature macrophages and granulocytes rather than to control hematopoesis (Zhan et al., 1998).
GM-CSF is produced by activated T lymphocytes, endothelial cells, macrophages and stromal cells. In humans, several chronic inflammatory diseases such as rheumatoid arthritis (RA), multiple sclerosis (MS) and asthma are associated with systemically or locally increased GM-CSF levels. One study showed elevated GM-CSF serum levels in 87 RA patients compared with healthy volunteers (Fiehn et al., 1992). Synovial fluid from inflamed arthritic joints of RA patients also showed significantly increased levels of GM-CSF (Alvaro-Gracia et al., 1991). In MS patients, elevated GM-CSF levels were found in the cerebral fluid (Perrella et al., 1993). Furthermore, GM-CSF plays a key role in respiratory diseases such as asthma and chronic obstructive pulmonary disease (COPD) (Balbi et al., 1997; Adkins et al., 1998; Culpitt et al., 2003). In allergic asthma, GM-CSF produced by bronchial epithelial cells is involved in enhancing survival of infiltrated inflammatory cells and, thereby, maintains the inflammatory response (Esnault and Malter, 2002).
Several animal models have demonstrated a high therapeutic potential of neutralizing GM-CSF by antibodies. A study in mice with collagen-induced arthritis (CIA) showed amelioration of the severity and speed of limb deformation by neutralizing murine GM-CSF with antibody 22E9 (Cook et al., 2001), while administration of the cytokine increased the severity of CIA in DBA/1 mice (Campbell et al., 1997). In mice with experimental autoimmune encephalo-myelitis (EAE), administration of antibody 22E9 has been shown to not only prevent the onset of EAE but also to ameliorate established disease (McQualter et al., 2001). Likewise, therapeutic effects of GM-CSF neutralization have been observed in various models of lung inflammation and asthma (Bozinovski et al., 2004), and a psoriasis model (Schon et al., 2000).
The potent anti-inflammatory activity of GM-CSF-neutralizing antibodies in a variety of animal models prompted us to develop a monoclonal antibody fragment with high potential to neutralize human GM-CSF. Because the antibody fragment would be administered in conditions of enhanced immune reactivity, we sought to give it the lowest possible immunogenicity. For high-level production by Escherichia coli or yeast fermentation, we decided to produce it as an scFv protein. Lastly, for extension of serum half-life of the scFv and further reduction of potential immunogenicity, the scFv was conjugated with 40 kDa PEG (polyethylene glycol). Here, we describe the generation and in vitro characteristics of a PEGylated fully human scFv that neutralizes the cell proliferation activity of human GM-CSF at low nanomolar concentrations.
|
Materials and methods |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionAcknowledgementsReferences |
|---|
Sources of recombinant human GM-CSF
Escherichia coli-derived: The hGM-CSF DNA fragment was subcloned from the pORF-hGM-CSF (Novagen, USA) into pET22b(+) (Novagen, USA). Expression of the human (h)GM-CSF was performed in E.coli BL21(DE3) periplasm according to standard methods. Purification of hGM-CSF was done by metal affinity chromatography (IMAC) using a Qiagen Ni-NTA Superflow column, followed by gel filtration on a Superdex 200 Prep Grade column (Pharmacia, Sweden).
Yeast-derived: Leukine® Liquid (Sargramostim) was obtained from Berlex, USA.
Biotinylation of hGM-CSF
Biotinylation of E.coli-derived hGM-CSF was done in phosphate-buffered saline (PBS)/5% DMSO (Sigma, Germany) with a 5-fold molar excess of EZ-Link Sulfo NHS-LC-LC Biotin (Pierce, USA) for 1 h at room temperature. Biotinylated hGM-CSF was purified by anion exchange chromatography according to standard protocols.
Fluorescein labeling of hGM-CSF
Conjugation of E.coli-derived hGM-CSF with Fluorescein-NHS (Fluka, Switzerland) was performed in borate buffer (0.05 M boric acid, 0.1 M NaCl, pH 8.5, 17.5% DMSO) with a 5-fold molar excess of Fluorescein-NHS for 1 h at room temperature. The conjugate was purified via gel filtration on Sephadex G25 medium (Amersham Biosciences, Germany).
Cloning of parental scFv
A rat hybridoma cell line was identified, which produced a monoclonal antibody neutralizing human GM-CSF. Its VL and VH gene fragments were amplified as described in Raum et al. (2001) and subcloned as functionally arranged VH-(G4-S)3-VL scFv into the pBluescript derived periplasmic expression vector pMicHis (Micromet AG, Germany).
Lymphocyte cDNA preparation for human VL and VH repertoires
PBMCs were isolated by Ficoll density centrifugation from five healthy donors. IgD+ B-cells were separated from PBMC using magnetic beads (CELLectionTM Pan Mouse IgG Kit, Dynal, Germany) coated with mouse anti-human IgD antibody (PharMingen, Germany). Subsequently total RNA was prepared from IgD+ B-cells or PBMCs (RNeasy® Midi Kit, Qiagen, Germany) and cDNA was synthesized by random hexamer priming (Roche, Germany).
PCR amplification of variable regions
For PCR amplification of the V
regions from IgD+ B-cell cDNA, each primer of the 5'-huV primer set (described in Raum et al., 2001) was combined with the following 3'-huV primers: 3'-hu-V-J1-SpeI-BsiWI 5'-GACGACACTAGTTGCAGCCACCGTACGTTTGATTTCCACCTTGGTCC-3', 3'-hu-V-J2/4-SpeI-BsiWI 5'-GACGACACTAGTTGCAGCCACCGTACGTTTGATCTCCASCTTGGTCC-3', 3'-hu-V-J3-SpeI-BsiW 5'-GACGACACTAGTTGCAGCCA CCGTACGTTTGATATCCACGTTGGTCC-3' and 3'-hu-V-J5-SpeI-BsiWI 5'-GACGACACT AGTTGCAGCCACCGTACGTTTAATCTCCAGTCGTGTCC-3'. V fragments were purified from an agarose gel and pooled according to their germline distributions, which were defined through sub-group-specific primers. VH fragments were PCR-amplified from cDNA of PBMC combining each primer from the 5'-huVH primer set (described in Raum et al., 2001) with the following 3'-primers: 3'-huVH-J1-BstEII 5'-CTGAGGAGACGGTG ACC-3' and 3'-huVH-J3-BstEII 5'-CTGAAGAGACGGTGACC-3'. The fragments were gel purified and pooled at a J1:J3 ratio of 3:1. Starting from this pool, a CDR3-truncated intermediate was generated by PCR amplification using the primers from the 5'-huVH primer set in combination with primers annealing in the FR3 region of the 3'-huVH-FR3 primer set. In a third PCR reaction, the parent VH CDR3 and the human FR4 region (JH3) were added to these fragments by PCR.
Construction of scFv libraries
VL-selection: The V pools were cloned via SacI/SpeI into the phagemid vector pMic5BHis (Micromet) containing the sequence of the parental rat VH. Transformation of the ligation products into E.coli Xl-1 blue was accomplished by electroporation (2.5 kV, 0.2 cm gap cuvette, 25 µF, 200 
. Determination of the library size and culture conditions after electroporation was performed as described (Burton et al., 1991). The bacterial culture was infected with 1 x 1012 pfu of helper phage VCSM13 resulting in secretion of M13 phage displaying scFv protein as a translational fusion to coat protein III on its surface.
VH-selection: The pooled VH fragments were digested with XhoI and BstEII. These digested fragments were then ligated into phagemid vector pMic5BHis, which had been previously restricted with the same two enzymes. Vector pMic5BHis contained a sequence encoding a human VL (VL 5–306) isolated in the VL-selection to form, together with the VH of the parental rat antibody, a hGM-CSF-specific scFv fragment. VL 5–306 was inserted in vector pMic5BHis via the restriction enzymes SacI and SpeI. Transformation and rescue with VCSM13 helper phage was carried out as described above.
Phage display selection
The phage library was harvested from the culture supernatant according to Burton et al. (1991). For selection 1–10 x 1011 phage particles were incubated with biotinylated hGM-CSF for 1 h in a total volume of 0.5 ml PBS/0.5% BSA. Then 6 x 107 Streptavidin-coated magnetic beads (Dynabeads® M-280 Streptavidin, Dynal, Germany) were added for another0.5 h. After repeating up to 10 washes with PBS/0.1% BSA, phages were eluted by HCl-glycine, pH 2.2. After neutralization with 2 M Tris, pH 12, the eluate was used for infection of a 2 ml culture of E.coli Xl-1 blue (OD600 = 1). Elution was repeated with HCl-glycine, pH 1 for selection of human VL. In the human VH selection, the second elution step was performed by resuspending the selection beads in 100 µl E.coli Xl-1 blue (OD600 = 1). After incubation for 15 min, beads were removed and the culture was added to the E.coli culture infected with the first eluate. Infected E.coli Xl-1 blue cultures were selected for carbenicillin resistance and were subsequently infected with VCSM13 helper phage to start the next selection round.
Expression and purification of soluble scFv
The pool of fragments obtained after four and five rounds of panning were subcloned into pMicFlag/His plasmid (pBluescript derivative for periplasmic expression of antibody fragments, Micromet). Expression of multiple different clones was performed in E.coli TG-1 in 96-well format. One hundred microlitre LB/0.1% glucose were inoculated with 10 µl of an overnight culture of single clones and grown for 4 h at 37°C. After addition of IPTG to a final concentration of 1 mM, the culture was grown at 30°C for another 18–20 h. Per well, 40 µl of BEL-buffer (400 mM boric acid, 320 mM NaCl, 4 mM EDTA pH 8.0 + 2.5 mg/ml lysozyme) was added and shaken at room temperature for 1 h. Cellular debris was eliminated by centrifugation and supernatants were tested by ELISA. For small-scale expression of soluble scFv in E.coli, preparation of periplasmic extracts was performed as described in Raum et al. (2001). ScFv was collected for further examination by ELISA. Alternatively, one-step purification of the scFv was performed in NiNTA spin columns (Qiagen, Germany) according to the manufacturer's instructions. For large-scale production, scFv was either expressed in E.coli BL21(DE3) from pMicFlag/His or in E.coli BL-21 AI from pBAD peri-Kan (a derivative of pBAD; Invitrogen, Germany) for periplasmic expression of antibody fragments. Single colonies were grown in selective medium to an OD600 = 0.8. Then, IPTG (pMicFlag/His) or L-arabinose (pBAD peri-Kan) was added to a final concentration of 1 mM or 0.8% (w/v), respectively, and the cultures were grown overnight at 30°C. Following harvest, bacteria were resuspended in 100 ml PBS and periplasmic preparation was performed as described in Raum et al. (2001). For purification of scFv produced on a large-scale as described above, a two-step purification process of IMAC (Ni-NTA Superflow, Qiagen, Germany) followed by gel filtration (HiLoadTM 16/60 Superdex 75 Prep Grade column, Pharmacia, Sweden) was applied. Gel filtration resulted in clearly distinguishable peaks signifying monomer and dimer fractions.
DNA sequencing
DNA sequencing was performed at SequiServe (Germany). Antibody sequences were analyzed according to V-base (http://vbase.mrc-cpe.cam.ac.uk).
ELISA
Detection of scFv fragments from periplasmic extracts: One µg/ml hGM-CSF was immobilized on microtiter plates (Maxisorb, Nunc, Germany). Wells were blocked with PBS/3% BSA. After incubation of periplasmic extracts for 1 h at room temperature, bound scFv was detected by an anti-FLAG M2 antibody (Sigma, Germany) followed by a POD-conjugated goat-anti-mouse IgG antibody (Dianova, Germany). The ELISA was developed with ABTS substrate (Roche, Germany) and measured at 405 nm.
Quantification of scFv 3077 or scFv 3077-PEG40 plasma concentrations: One µg/ml hGM-CSF was immobilized on microtiter plates (Maxisorb, Nunc, Germany). After blocking the wells with PBS/3% BSA, samples were added and incubated for 1 h at room temperature. Samples were prediluted to appropriate concentrations in PBS/50% mouse plasma. Anti-3077 scFv goat immunoserum was used for detection of bound scFv, followed by a POD-conjugated rat anti-goat antibody (Dianova, Germany). The ELISA was developed with TMB substrate (Sigma, Germany) and measured at 450 nm. Quantification of scFv 3077 and scFv-PEG40 was done by means of a standard curve for scFv 3077 and scFv-PEG40, respectively; calculation of the scFv concentrations was performed from optical readings within the linear range.
TF-1 proliferation inhibition assay
TF-1 cells (DSMZ ACC 334) were cultivated in RPMI 1640/10% FCS in the presence of 2.5 ng/ml hGM-CSF. For the inhibition assay TF-1 cells were washed twice with 1x PBS and seeded at a density of 0.9 x 104 cells/well in RPMI 1640 (containing 10% FCS, but lacking hGM-CSF) in a 96-well flat bottom microtest plate. A final concentration of 0.3 ng/ml hGM-CSF was added to stimulate TF-1 cell proliferation. A dilution series of scFv with final concentrations ranging from 0.4 to 4000 nM was added to inhibit proliferation. After incubation for 72 h at 37°C in 5% CO2, the proliferative status was determined by adding WST-1 reagent (Roche, Germany). Viable cells were quantitated by measuring the absorbance at 450 nm. The data were analyzed and fitted for half-maximal inhibition of proliferation (IC50) using the GraphPadPrism4 software.
FACS competition assay
Ten µg/ml parent antibody or small-scale purified scFv was incubated with 0.4 µg/ml hGM-CSF-FITC conjugate in PBS. The protein samples were left to equilibrate at 25°C for 1 h prior to addition of a suspension of TF-1 cells. TF-1 cells were starved from hGM-CSF overnight, washed twice with PBS/1% FCS/0.05% NaN3 and resuspended in 100 µl of pre-equilibrated protein sample containing the hGM-CSF-FITC and scFv. After 1 h at 4°C, cells were analyzed by flow cytometry on a FACScalibur (Becton Dickinson, Germany).
Surface plasmon resonance
Association and dissociation rate constants were determined by surface plasmon resonance on the BIAcoreTM 2000 (Biacore AB, Switzerland). The surface of the CM5 sensor chip was activated with NHS/EDC. The hGM-CSF was coupled by injection of 10 µg/ml hGM-CSF in 0.01 M sodium-acetate, pH 4.7. Approximately 800 response units were immobilized. Binding experiments were performed by injecting serially diluted samples at a flow rate of 5 µl/min and HBS-EP (0.01 M HEPES, pH 7.4, 0.15 M NaCl, 3 mM EDTA, 0.005 % surfactant P20) as running buffer at 25°C. Dissociation was monitored for 100 s. The association and dissociation rates (ka [1/Ms] and kd [1/s], respectively) were calculated using the BIAevaluation software (3.2 RC1) with a 1:1 Langmuir binding equation and the equilibrium dissociation constant KD [M] was calculated from kd/ka. Standard deviations were calculated from at least three independent fits in dilution experiments.
Site-specific PEGylation
Two subcloning steps were performed to provide an scFv 3077 variant that displayed a C-terminal cysteine suitable for site-specific PEGylation (Yang et al., 2003). The resulting fragment was ligated into a pBAD peri-Kan vector for periplasmic expression. Expression of the 3077_His6_Cys construct in E.coli BL21-AI was carried out as described above. PEGylation was performed as in Yang et al. (2003). In brief, the 3077_His6_Cys construct was purified from periplasmic extracts by a two-step process of IMAC (Ni-NTA Superflow, Qiagen, Germany) followed by gel filtration (HiLoadTM 16/60 Superdex 75 Prep Grade column, Pharmacia, Sweden). After IMAC purification and a reducing step with Dithiothreitol (DTT), gel filtration was carried out. Immediately prior to PEGylation, a PD10 SEC column (Amersham Biosciences, Germany) was used to remove the DTT used in the gel filtration step. Protein-containing fractions were determined with a Bradford assay (Bio-Rad, Germany) and pooled. mPEG2 Maleimide of 40 kDa molecular weight (Nektar Therapeutics, USA) was added at a 10-fold molar excess over scFv polypeptide. Conjugation was performed for 2 h at room temperature in the dark. Separation of the PEGylated scFv from free PEG and scFv molecules was performed by cation exchange chromatography (ResourceS, CV = 1 ml), using a linear NaCl gradient for elution. The running buffer contained 20 mM citrate pH 4 and the gradient elution buffer was 1 M NaCl and 20 mM citrate, pH 4. PEGylated scFv eluted from the column at 140 mM NaCl. Protein concentration was monitored via absorption of the respective fractions at a wavelength of 280 nm. Protein containing fractions were pooled, dialyzed against PBS and analyzed for purity and PEGylation efficiency using SDS–PAGE.
Determination of thermal stability
PEGylated and non-PEGylated scFv 3077 were diluted to a final concentration of 25 µg/ml (1 µM). Aliquots (30 µl) were heated in a water bath at temperatures varying from 40 to 100°C. The actual temperature was monitored in a reference vial. After 5 min incubation the protein was snap-cooled on ice. To test the neutralizing activity of a respective sample, a TF-1 proliferation-inhibition assay was performed in duplicate. TF-1 cells (0.9 x 104 per well) were seeded in a 96-well microtest plate and 10 µl of the heated protein solution was added. A final concentration of 0.3 ng/ml hGM-CSF was used to stimulate the proliferation of cells. After incubation at 37°C, 5% CO2 for 72 h, the proliferative activity of the TF-1 cells was quantified by addition of WST-1 and by measurement of the absorbance at 450 nm. As a control for neutralization, the untreated scFv 3077 and the untreated PEGylated scFv 3077 were included in the proliferation experiments. Data were normalized defining the maximal neutralizing effect obtained with the respective untreated scFvs as 0% proliferation and the minimal neutralizing effect of scFv heated at 100°C as 100% proliferation of TF1 cells. The data were fitted using the GraphPadPrism4 software.
Guanidinium hydrochloride treatment
scFv 3077 was mixed with the denaturant to a final protein concentration of 5 µg/ml and guanidinium hydrochloride (GdnHCl) concentrations varying from 0 to 5 M. After overnight equilibration at 10°C, the fluorescence emission spectra from 320 to 370 nm were measured at an excitation wavelength of 280 nm. Maximum emission wavelength was determined for each spectrum. Data from two experiments were included, normalized to the fraction of unfolded scFv protein and analyzed in accordance with Pace (1990).
Pharmacokinetics of scFv 3077 and scFv 3077-PEG40
C57BL/6 mice (female, 9–10 weeks old, three animals per group) were injected intravenously with 1.5 mg per kg of scFv 3077 or scFv 3077-PEG40, respectively. Different groups were alternatingly bled at 0.08, 0.25, 0.5, 1, 1.5, 2, 3, 4, 6, 8 and 24 h after scFv 3077 injection and after 0.08, 0.25, 0.5, 1, 3, 8, 24, 48, 72, 96 and 168 h after scFv 3077-PEG40 injection. Serum concentrations of scFv 3077 and scFv 3077-PEG40 were quantified by ELISA.
Pharmocokinetic calculations were performed by the pharmacokinetic software package WinNonlin Professional 4.1 (Pharsight Corporation, Mountain View, CA, 2003). Parameters were determined by non-compartmental analysis based on the model for intravenous bolus injection. The distribution half-life (T1/2-alpha) was calculated using a log-linear regression of the first three to four sample time points whereas the terminal elimination half-life (T1/2-beta) was calculated by the last three to four sample time points with detectable scFv concentration.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionAcknowledgementsReferences |
|---|
Isolation of parental VH and VL-fragments and expression of the parental scFv
Total RNA was isolated from a rat hybridoma cell line (MT-G-02C) producing an antibody neutralizing hGM-CSF. This RNA was transcribed into cDNA, and variable heavy (VH) and light chain (VL) gene segments were amplified by PCR, and sequenced. Gene segments for VH and VL domains were re-amplified with primers introducing suitable restriction sites followed by sub-cloning into a vector for periplasmic production in E.coli. Periplasmic extracts from four transformants were tested for the presence of functional scFv that would bind to hGM-CSF in ELISA. However, none of the clones showed a positive signal. By producing the scFv in E.coli cytoplasm as inclusion bodies, small amounts of functional scFv could be obtained after a refolding procedure (data not shown), demonstrating that the scFv was expressed but not produced in a functional form in E.coli periplasm or cytoplasm.
Selection of a human light chain
To overcome the lack of functional expression in E.coli and to convert the parental rat scFv into a human equivalent of reduced immunogenicity, a two-step phage display guided selection strategy was applied as schematically outlined in Figure 1.
|
In a first step, the parental heavy chain was sub-cloned into a phage display vector and combined with a human kappa light chain repertoire from B cells of five healthy donors. Each plasmid in the library contained the parental rat VH combined with a distinct human VL gene. After expression on phage, a pool of scFvs was generated in which rat VH and human VL are joined via a (Gly4Ser)3 flexible linker. The resulting library consisted of 3 x 108 independent clones. Five rounds of phage display selection were carried out using soluble biotinylated hGM-CSF as antigen. To select for strong binders, the stringency was increased during selection by decreasing antigen concentration from 100 nM (rounds 1–3) to 10 nM (round 4), and finally to 1 nM (round 5).
Phages selected after four and five rounds were expressed as scFvs in the periplasm of E.coli. The scFv molecules were affinity-purified by spin columns, followed by testing in ELISA for specific binding to hGM-CSF. Human GM-CSF used for phage selection had been produced in prokaryotes and, therefore, lacked N-glycosylation, an attribute that, however, allowed for more effective biotinylation. To ignore binders that specifically recognized the non-glycosylated form of the cytokine, a glycosylated form of hGM-CSF produced in yeast was used for screening of scFvs in ELISA. A total of five clones showed strong binding to the antigen (4–301, 4–306, 5–301, 5–306, 5–310) (Figure 2A). All three binders identified from the fifth selection round (5–301, 5–306, 5–310) and clone 4–306 were identical in sequence. These clones are in the following collectively referred to as clone 5–306. Clone 4–301 had a different sequence but showed homology to clone 5–306 at the amino acid level. Both sequences were derived from the V kappa 1 germline subfamily gene Vk1-O12.
|
GM-CSF neutralizing activity of half-human scFvs
Cytokine binding is necessary but not sufficient for neutralizing GM-CSF. ELISA-positive clones 5–306 and 4–301 were, therefore, tested for the ability to inhibit the binding of hGM-CSF to its high-affinity receptor expressed on TF-1 cells. Binding of fluorescently-labeled hGM-CSF-FITC to TF-1 cells was detected by flow cytometry. Purified scFvs from clones 5–306 and 4–301 were pre-incubated with hGM-CSF-FITC and the mixture used for staining of TF-1 cells. ScFv from the dominant clone 5–306 as well as the parental rat antibody, inhibited binding of FITC-labeled GM-CSF, while the scFv from clone 4–301 still allowed binding of labeled GM-CSF to TF-1 cells (Figure 2B and C). Clone 5–306 was used for further analysis and guided selection.
An assay was developed that monitored the effect of scFvs on neutralizing the proliferating activity of hGM-CSF on cell line TF-1. ScFv from clone 5–306 was purified from periplasmic preparations by IMAC. Monomeric and dimeric isoforms were separated by gel filtration and both tested for bioactivity. Proliferation of TF-1 cells was induced by hGM-CSF in the presence of a serial dilution of scFv and viable cells quantitated after 72 h by addition of WST-1. ScFv 5-306 inhibited proliferation of GM-CSF-stimulated TF-1 in a dose-dependent form. Monomeric and the dimeric variants were similarly active. IC50 values of 7.3 nM and 3.5 nM were determined for monomeric and dimeric scFvs, respectively (Figure 3).
|
Selection of a human heavy chain
For the final generation of a hGM-CSF-neutralizing scFv, the parental rat VH domain in clone 5–306 was complemented by a human VH repertoire that preserved the 10 amino acid-long CDR3 from the parental VH domain. The resulting phage library was then used for phage selection with soluble biotinylated hGM-CSF as antigen. Combination of the pre-selected VL 5–306 with the human VH repertoire resulted in an antibody phage library of 1.6 x 108 independent clones. Antigen concentration was reduced for selection from 100 nM in the first round to 10 nM for the following three rounds. Highly enriched phages were harvested and VH/VL fragments were sub-cloned into a vector for periplasmic expression in E.coli. ScFv proteins from 160 clones were prepared and tested for binding to immobilized hGM-CSF in ELISA. Over 80% of scFvs showed strong antigen binding. Sequence analysis of the 13 strongest binders led to the identification of four different sequences. ScFv proteins expressed by representative clones encoding these four sequences (clones nos 3035, 3039, 3077 and 3080) gave strong binding signals to hGM-CSF in ELISA (Figure 4A). All VH sequences were derived from the human germline sequence VH-1 1-O2.
|
GM-CSF neutralizing activity of human scFvs
ScFvs from all four clones were produced and monomeric fractions analyzed for their neutralizing potential in the TF-1 cell proliferation assay. All proteins inhibited hGM-CSF-dependent cell proliferation in a 72 h assay (Figure 4B). IC50 values were 130 nM for scFv 3039, 19.1 nM for scFv 3080, 3.2 nM for scFv 3025 and 2.6 nM for scFv 3077. Repetitions of the assay showed considerable inter-assay variance of IC50 values, but the relative potency of the molecules remained constant.
Binding constants of scFvs 3035, 3039, 3077 and 3080 were determined by surface plasmon resonance analysis using immobilized hGM-CSF. As shown in Table I, dissociation constants (KD) ranged from 10–6 to 10–9 M. With one exception, binding affinities were in line with the neutralizing activity of scFvs in the TF-1 cell proliferation assay. Although scFvs from clones 3035 and 3077 showed a very similar neutralizing efficacy in the TF-1 assay, the KD of scFv 3077 (KD = 1.1 x 10–9 M) was seven times lower than for scFv 3035 (KD = 9 x 10–9 M) due to a high association rate (ka = 1.7 x 106 M s–1). The better kinetic properties of 3077 in this low nanomolar affinity range may have been concealed in the neutralization assay because of the long incubation period of 72 h. Based on highest affinity and biological activity, scFv 3077 was selected for further analysis and modification.
|
Thermodynamic stability of scFv 3077
The selected scFv 3077 was investigated for its thermodynamic stability using GdnHCl equilibrium denaturation. The GndHCl-induced unfolding of the protein was measured by the change in emission maximum at an excitation wavelength of 280 nm after equilibration with varying concentrations of the denaturing reagent. Figure 5 shows a co-operative two-step unfolding behavior of the scFv with a first transition and midpoint of unfolding at 2.05 M GdnHCl and a second transition and midpoint of unfolding at 4.23 M GdnHCl. This reflects a high thermodynamic stability of scFv 3077.
|
PEGylation of scFv 3077
Owing to their small size, scFv proteins are rapidly cleared from the circulation through the kidneys. Pharmacokinetic studies have demonstrated that the circulating half-life of scFvs and of other small proteins and chemicals can be substantially prolonged by conjugation with PEG (Lee et al., 1999). Owing to its favorable binding properties and stability, scFv from clone 3077 was chosen for PEGylation. To this end, a variant of scFv 3077 was constructed containing a C-terminal sulfhydryl group for site-directed conjugation by a maleimide-PEG polymer (Yang et al., 2003). The coding sequence of scFv 3077 was amplified with primers introducing histidine and cysteine codons at the C-terminus and restriction sites for ligation into a new expression vector.
Production of scFv 3077 yielded 2.8 mg monomeric scFv from a 5 l shaking flask culture. The proportion of monomer to dimer was approximately 62% to 38%. In small-scale experiments, we observed a reduction of monomeric scFv portion bearing a C-terminal cysteine and, therefore, aimed to produce larger quantities of monomeric scFv 3077_His_Cys. For improved expression, the scFv gene was subcloned into a prokaryotic expression vector based on pBAD containing a signal peptide coding region for efficient transport of the protein into the periplasm and produced in a 2 l fed batch fermenter run. The E.coli pellet from 750 ml culture volume was used for periplasmic extraction and purification. We obtained 5.2 mg monomeric scFv 3077_His_Cys at a proportion of 42% monomer and 58% dimer. The resulting monomeric scFv 3077_His_Cys was then exposed to reducing conditions by addition of DTT for dissociation of C-terminally linked scFv dimers and to generate reactive C-terminal sulfhydryl groups. Reducing conditions were maintained during separation of monomeric scFv by gel filtration. DTT was removed right before conjugation of branched 40 kDa PEG-maleimide. These experimental conditions allowed selective conjugation of the 40 kDa PEG-Mal to the free C-terminal cysteine residue by formation of a thioether bond between maleimide and sulfhydryl groups.
Figure 6 shows that PEGylation of scFv 3077_His_Cys was very effective and yielded a homogeneous product with an apparent molecular size of 100 kDa. In the elution profile of PEGylated scFv 3077_His_Cys from a cation exchange chromatography column, the signal for residual non-PEGylated scFv was at the limit of detection (Figure 6A). Analysis of column fractions A9 and A10 by SDS–PAGE (Figure 6B, lane 4) showed that cation exchange chromatography completely removed non-PEGylated scFv left after coupling reaction (compare with lanes 2 and 3). Non-PEGylated, maleimide-activated scFv 3077_His_Cys had a molecular size of 31 kDa (Figure 6B, lane 1), which increased to 100 kDa upon reaction with a 40 kDa PEG moiety. Under the present reaction conditions, more than 90% of the scFv was conjugated with PEG, as can be estimated from Figure 6A and B.
|
Affinity and activity of naked and PEGylated scFv
Purified PEGylated scFv 3077_His_Cys (3077-PEG40) was compared with scFv 3077 for kinetic parameters and the potential to neutralize the activity of hGM-CSF. For 3077-PEG40, an association rate of 2.7 x 104 Ms–1 and a dissociation rate of 1.1 x 10–3s–1 were determined by surface plasmon resonance analysis, resulting in a KD of 4.1 x 10–8 M. This corresponded to a 60-fold reduction in the association rate of 3077-PEG40 compared with its non-PEGylated counterpart, while the dissociation rate remained essentially unaffected by PEGylation. Overall, PEGylation led to a 36-fold decrease in affinity from 1.1 x 10–9 M for scFv 3077 to 4.1 x 10–8 M for 3077-PEG40.
We next tested the influence of PEGylation on GM-CSF neutralization in the TF-1 cell proliferation assay. Surprisingly, 3077-PEG40 exhibited the same neutralizing efficacy as non-PEGylated scFv 3077 (Figure 7). Apparently the decrease in affinity caused by PEGylation did not result in a corresponding reduction of the biological activity of the molecule in a long-term assay.
|
Thermal stability of naked and PEGylated scFv 3077
In order to investigate the impact of PEGylation on the scFv with regard to its thermal stability, purified scFv 3077 and its PEGylated counterpart were exposed to temperatures varying from 40°C up to 100°C. After heating the compounds for 5 min at the respective temperatures, samples were cooled down on ice and assayed for their capability to neutralize GM-CSF dependent TF-1 cell proliferation. As shown in Figure 8, a significant difference in the sensitivity of the respective PEGylated and non-PEGylated scFvs to increasing temperature was observed. scFv 3077 reached half-maximal temperature inhibition of its activity at 73°C, whereas the PEGylated scFv 3077 reached half-maximal temperature inhibition at 83°C. Complete loss of neutralizing activity was seen after incubation at 85°C for scFv 3077 and at 95°C for the PEGylated scFv 3077.
|
Pharmacokinetic properties of scFv 3077 and scFv 3077-PEG40
C57BL/6 mice were intravenously injected with 1.5 mg/kg scFv 3077 or scFv 3077-PEG40, respectively, and mice bled at different time points after injection. ScFv 3077 and scFv 3077-PEG40 serum concentrations were quantified by specific ELISAs and serum concentrations versus time profiles generated (Figure 9). Peak serum concentrations of the two constructs were detected 5 min after bolus i.v. injection and reached the lower limit of quantification after 6 h for scFv 3077 and after 168 h for the scFv3077-PEG40. Elimination rate constants were determined and resulted in distribution half-lives T1/2-alpha of 0.12 ± 0.01 for scFv 3077 and 0.39 ± 0.04 for scFv 3077-PEG40, respectively, and terminal elimination half-lives T1/2-beta of 1.99 ± 0.49 h for scFv 3077 and 59.36 ± 0.31 for scFv 3077-PEG40. It was observed that the prolonged half-life of the PEGylated version correlated with an increased area under the curve (AUClast) of 156.80 ± 9.11 h µg/ml as compared with 4.91 ± 0.37 h µg/ml for scFv 3077. Accordingly, a significantly reduced clearance of 8.52 ± 0.38 mg/h kg was observed for the PEGylated molecule in comparison with 301.22 ± 22.47 mg/h kg of the unconjugated scFv.
|
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionAcknowledgementsReferences |
|---|
The present work describes the construction and characterization of a PEGylated human scFv neutralizing hGM-CSF. As a starting point, the scFv of a rat monoclonal antibody neutralizing the biological activity of hGM-CSF was constructed and produced in the periplasm of E.coli. In this manner, no functional scFv could be obtained despite the fact that the protein, when expressed in inclusion bodies, regained biological activity after refolding. Deficient activity of scFvs is frequently observed when these are constructed from VH and VL domains of monoclonal antibodies originating from hybridoma cells (e.g. Rojas et al., 2004
). After replacing the original rat kappa light chain with a human kappa repertoire and selecting for hGM-CSF binding using phage display technology, rat/human scFvs could be isolated that bound specifically to the antigen and could be well expressed in a functional form in the periplasm of E.coli. This finding demonstrates that exchange of the light chain can foster functional producibility of an scFv, which is otherwise not expressed in a soluble form in the periplasm of prokaryotes. A similar observation was published very recently. An scFv derived from a monoclonal antibody binding to N-glycolyl GM3 ganglioside could only be expressed functionally in the periplasm of E.coli after the exchange of the original light chain against a light chain selected from human and murine VL repertoires by phage display (Rojas et al., 2004). Light chain selection can thus be a general means of stabilizing scFvs while preserving their binding specificity.
One of the two human light chains allowing functional expression of an hGM-CSF-specific scFv in E.coli was dominantly selected. Both human sequences are derived from the human Vkappa 1 subgroup, in particular the O12 Vkappa segment. The human Vkappa 1 subgroup is known to be preferentially selected in phage display experiments (e.g. Xu et al., 2004), probably due to its high intrinsic stability (Ewert et al., 2003). This potential gain in intrinsic stability could also be a reason that the selected human kappa light chains are able to stabilize the original rat VH upon expression in E.coli. The difference between the original rat VL and the selected human VLs is also illustrated by the observation that the two selected human sequences share only 60% identity to the original rat VL. Analysis of the selected human VLs and the parental rat VL regarding their CDR structures according to Al-Lazikani et al. (1997) revealed the same canonical structures in CDR1 and CDR2 indicating a common architecture of the CDRs. Apparently, owing to epitopic guidance by retaining the parental VH (‘guided selection’), the selected rat/human scFvs still showed GM-CSF neutralizing activity in a low nanomolar range. In most cases, guided selection approaches maintain the epitopic specificity of the parental antibody (Jespers et al., 1994; Figini et al., 1998; Raum et al., 2001), although differences in the fine specificity have been observed (Ohlin et al., 1996).
Since human-derived proteins, especially antibodies and fragments thereof, are thought to have reduced immunogenic potential in man compared with non-human proteins (Qu et al., 2005), the rat/human scFv 5–306 was subjected to a further round of phage display selection. In this second round, the rat VH was exchanged with VH domains from a human repertoire. On the notion that especially the CDR3 of VH is dominant for antigenic fine specificity (Kabat and Wu, 1991; Collet et al., 1992; Xu and Davis, 2000; Sidhu et al., 2004), the original VH CDR3 was maintained in combination with one human FR4 (from JH3) and a diverse human repertoire of FR1-CDR1-FR2-CDR2-FR3 derived from peripheral B cells (Rader et al., 1998; Steinberger et al., 2000). Owing to the generally high sequence diversity within the CDR3 of VH (Sanz, 1991), the 10 amino acid CDR3 sequence in our scFv cannot be identified as species-specific. Consequently, because the CDR3 is embedded in a human VH domain and combined with a human VL domain, the complete scFvs may be considered to be fully human.
Using soluble hGM-CSF as antigen for phage display selection, four human scFvs could be isolated, which specifically bound to hGM-CSF. All four human scFvs also showed GM-CSF-neutralizing activity, indicating that the guided selection approach preserved the binding specificity of the parental rat monoclonal antibody. An IC50 range from 2 to 130 nM was observed for the GM-CSF-neutralizing activity of human scFvs. IC50 values correlated with binding affinities that ranged from 1 to 1700 nM, indicating that increased binding affinity may translate into prolonged retention of the antigen by the antibody. A 10-fold higher on-rate of scFv 3077 compared with scFv 3035 could apparently not produce a significant difference in neutralizing activity under the experimental conditions of the TF-1 neutralization assay, because IC50 values were comparable (2.6 and 3.2 nM, respectively). The molecule chosen for further development, scFv 3077, was characterized and exhibited properties comparable with stability engineered scFv proteins (Worn and Pluckthun, 1998).
Serum half-life for scFvs has been determined to be on the order of a few hours in humans (Larson et al., 1997; Fitch et al., 1999) probably due to renal clearance (Hamilton et al., 2002). Strategies have been established to prolong the half-life of such molecules by coupling to serum protein (Smith et al., 2001), or to polymeric organic compounds of high molecular weight. It could be demonstrated that PEGylation of scFv molecules dramatically increases their serum half-life (Lee et al., 1999). To generate a more long-lived GM-CSF neutralizer, a version of scFv 3077 was constructed with a C-terminal cysteine for site-directed and quantitative coupling of 40 kDa PEG. When compared side-by-side, non-PEGylated scFv 3077 and PEGylated 3077_His_Cys showed identical biological activity with respect to neutralization of hGM-CSF. This was despite the fact that PEGylated scFv 3077 suffered a distinct loss in affinity due to a reduced on-rate. Such an effect has previously been observed (Yang et al., 2003). The reduced on-rate may be attributable to several circumstances. One is a reduced diffusion rate of PEGylated scFv 3077, which has double the size of unconjugated scFv 3077. The other could be steric hindrance for GM-CSF access and binding by the large hydrated PEG polymer. The equal biological activity of naked and PEGylated scFv in the TF-1 assay suggests that over longer periods of time only off-rate matters for neutralizing activity. The same may be true for the in vivo situation. The high biological activity of the molecules in a cell culture system running for 72 h also suggests that the scFvs are sufficiently stable and resistant to proteases in fetal calf serum or released by TF-1 cells.
To further investigate the stabilizing effect of PEG described for other proteins (Diwan and Park, 2001), scFv 3077 and its PEGylated version were exposed to temperatures from 40 to 100°C and then tested in an activity-based assay. We observed a significant increase in thermal stability of the PEGylated scFv of 10°C. This effect may be due to the conformational shielding of the scFv by the surrounding PEG molecule that potentially reduces the ‘breathing’ of the VH and VL domain and, therefore, stabilizes the respective scFv. Besides an increase in molecular size, this stabilizing effect of the PEG may contribute to the prolonged half-life of PEGylated molecules in vivo by protection against serum proteases.
C57BL/6 mice i.v. bolus injection experiments were performed to address the pharmacokinetic profile of the unconjugated scFv3077 versus its PEGylated counterpart. In this experiment a 30-fold prolongation in in vivo half-life of the PEGylated scFv 3077 could be observed, probably due to a reduced renal clearance. Comparable data were observed previously using PEGylated scFv fragments in mouse experiments (Yang et al., 2003).
The human scFv 3077 and its PEGylated counterpart 3077-PEG40 are able to neutralize the proliferating activity of the hGM-CSF molecule at low nanomolar levels with off-rates that predict a durable sequestration of hGM-CSF preventing the cytokine from binding to immune cells bearing high-affinity GM-CSF receptor. This long-term sequestration of GM-CSF may have clinical benefits in inflammatory and autoimmune diseases comparable with the in vivo effects reported for the monoclonal anti-mouse GM-CSF antibody 22E9. Owing to its human origin, scFv 3077 and derivatives are expected to exhibit very low immunogenicity in patients. This is particularly important when treating pro-inflammatory disease presenting with an enhanced reactivity of immune cells. The non-PEGylated scFv may be useful in clinical settings where short-term GM-CSF-neutralizing activity is desirable, or administration to a compartment where clearance is reduced. Use of the PEGylated version of scFv 3077 may be more desirable in clinical settings where long half-life and low immunogenicity may matter.
|
Footnotes |
|---|
E.-M.Krinner and J.Hepp contributed equally to this work
Edited by Prof Dr Hennie Hoogenboom
|
Acknowledgements |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionAcknowledgementsReferences |
|---|
We thank Steve Zeman for his assistance and critical review in the preparation of this manuscript.
|
References |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionAcknowledgementsReferences |
|---|
Adkins K.K., Levan T.D., Miesfeld R.L., Bloom J.W. (1998) Am. J. Physiol. 275:L372–378.
Al-Lazikani B., Lesk A.M., Chothia C. (1997) J. Mol. Biol. 273:927–948.[CrossRef][Web of Science][Medline]
Alvaro-Gracia J.M., Zvaifler N.J., Brown C.B., Kaushansky K., Firestein G.S. (1991) J. Immunol. 146:3365–3371.[Abstract]
Balbi B., Bason C., Balleari E., Fiasella F., Pesci A., Ghio R., Fabiano F. (1997) Eur. Respir. J. 10:846–850.[Abstract]
Bozinovski S., Jones J., Beavitt S.J., Cook A.D., Hamilton J.A., Anderson G.P. (2004) Am. J. Physiol. Lung Cell Mol. Physiol. 286:L877–885.
Burton D.R., Barbas C.F. III, Persson M.A., Koenig S., Chanock R.M., Lerner R.A. (1991) Proc. Natl Acad. Sci. USA 88:10134–10137.
Campbell I.K., Bendele A., Smith D.A., Hamilton J.A. (1997) Ann. Rheum. Dis. 56:364–368.
Collet T.A., Roben P., O'Kennedy R., Barbas C.F. III, Burton D.R., Lerner R.A. (1992) Proc. Natl Acad. Sci. USA 89:10026–10030.
Cook A.D., Braine E.L., Campbell I.K., Rich M.J., Hamilton J.A. (2001) Arthritis Res. 3:293–298.[CrossRef][Web of Science][Medline]
Culpitt S.V., Rogers D.F., Fenwick P.S., Shah P., De Matos C., Russell R.E., Barnes P.J., Donnelly L.E. (2003) Thorax 58:942–946.
Diederichs K., Boone T., Karplus P.A. (1991) Science 254:1779–1782.
Diwan M. and Park T.G. (2001) J. Control Release 73:233–244.[CrossRef][Web of Science][Medline]
Esnault S. and Malter J.S. (2002) Arch. Immunol. Ther. Exp. (Warsz) 50:121–130.[Medline]
Ewert S., Huber T., Honegger A., Pluckthun A. (2003) J. Mol. Biol. 325:531–553.[CrossRef][Web of Science][Medline]
Fiehn C., Wermann M., Pezzutto A., Hufner M., Heilig B. (1992) Z. Rheumatol. 51:121–126.[Web of Science][Medline]
Figini M., Obici L., Mezzanzanica D., Griffiths A., Colnaghi M.I., Winter G., Canevari S. (1998) Cancer Res. 58:991–996.
Fitch J.C., et al. (1999) Circulation 100:2499–2506.
Hamilton S., Odili J., Wilson G.D., Kupsch J.M. (2002) Melanoma Res. 12:373–379.[CrossRef][Web of Science][Medline]
Hercus T.R., Cambareri B., Dottore M., Woodcock J., Bagley C.J., Vadas M.A., Shannon M.F., Lopez A.F. (1994) Blood 83:3500–3508.
Inaba K., Steinman R.M., Pack M.W., Aya H., Inaba M., Sudo T., Wolpe S., Schuler G. (1992) J. Exp. Med. 175:1157–1167.
Jespers L.S., Roberts A., Mahler S.M., Winter G., Hoogenboom H.R. (1994) Biotechnology (N Y) 12:899–903.[CrossRef][Medline]
Kabat E.A. and Wu T.T. (1991) J. Immunol. 147:1709–1719.[Abstract]
Larson S.M., et al. (1997) Cancer 80:2458–2468.[CrossRef][Web of Science][Medline]
Lee L.S., Conover C., Shi C., Whitlow M., Filpula D. (1999) Bioconjug. Chem. 10:973–981.[CrossRef][Web of Science][Medline]
McQualter J.L., Darwiche R., Ewing C., Onuki M., Kay T.W., Hamilton J.A., Reid H.H., Bernard C.C. (2001) J. Exp. Med. 194:873–882.
Ohlin M., Owman H., Mach M., Borrebaeck C.A. (1996) Mol. Immunol. 33:47–56.[CrossRef][Web of Science][Medline]
Pace C.N. (1990) Trends Biotechnol. 8:93–98.[CrossRef][Web of Science][Medline]
Perrella O., Carrieri P.B., De Mercato R., Buscaino G.A. (1993) Eur. Neurol. 33:152–155.[Web of Science][Medline]
Qu Z., Griffiths G., Wegener W., Chang C., Govindan S., Horak I., Hansen H., Goldenberg D. (2005) Methods 36:84–95.[CrossRef][Web of Science][Medline]
Rader C., Cheresh D.A., Barbas C.F. III. (1998) Proc. Natl Acad. Sci. USA 95:8910–8915.
Raum T., Gruber R., Riethmuller G., Kufer P. (2001) Cancer Immunol. Immunother. 50:141–150.[CrossRef][Web of Science][Medline]
Rojas G., Talavera A., Munoz Y., Rengifo E., Krengel U., Angstrom J., Gavilondo J., Moreno E. (2004) J. Immunol. Methods 293:71–83.[CrossRef][Web of Science][Medline]
Sanz I. (1991) J. Immunol. 147:1720–1729.[Abstract]
Schon M., Denzer D., Kubitza R.C., Ruzicka T., Schon M.P. (2000) J. Invest. Dermatol. 114:976–983.[CrossRef][Web of Science][Medline]
Sheridan J.W. and Metcalf D. (1973) J. Cell Physiol. 81:11–23.[CrossRef][Web of Science][Medline]
Shibuya K., Chiba S., Miyagawa K., Kitamura T., Miyazono K., Takaku F. (1991) Eur. J. Biochem. 198:659–666.[Web of Science][Medline]
Sidhu S.S., Li B., Chen Y., Fellouse F.A., Eigenbrot C., Fuh G. (2004) J. Mol. Biol. 338:299–310.[CrossRef][Web of Science][Medline]
Smith B.J., et al. (2001) Bioconjug. Chem. 12:750–756.[CrossRef][Web of Science][Medline]
Steinberger P., Sutton J.K., Rader C., Elia M., Barbas C.F. III. (2000) J. Biol. Chem. 275:36073–36078.
Worn A. and Pluckthun A. (1998) Biochemistry 37:13120–13127.[CrossRef][Medline]
Xu J.L. and Davis M.M. (2000) Immunity 13:37–45.[CrossRef][Web of Science][Medline]
Xu M.Y., Xu X.H., Chen G.Z., Deng X.L., Li J., Yu X.J., Chen M.Z. (2004) World J. Gastroenterol. 10:2619–2623.[Medline]
Yang K., et al. (2003) Protein Eng. 16:761–770.
Zhan Y., Lieschke G.J., Grail D., Dunn A.R., Cheers C. (1998) Blood 91:863–869.
Received February 16, 2006; revised June 1, 2006; accepted June 12, 2006.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||