PEDS Advance Access published online on November 19, 2008
Protein Engineering Design and Selection, doi:10.1093/protein/gzn066
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Functional humanization of an anti-CD16 Fab fragment: obstacles of switching from murine 
to human or 
light chains
Lehrstuhl für Biologische Chemie, Technische Universität München, 85350 Freising-Weihenstephan, Germany
1 To whom correspondence should be addressed. E-mail: skerra{at}wzw.tum.de
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and methodsResultsDiscussionFundingAcknowledgementsReferences |
|---|
An
CD30xCD16 bispecific monoclonal antibody (MAb) was previously shown to induce remission of Hodgkin’s disease refractory to chemo- and radiotherapy through specific activation of natural killer (NK) cells, but the appearance of a human anti-mouse antibody (HAMA) response prevented its use for prolonged therapy. Here, we describe an effort to humanize the Fab arm directed against Fc
RIII (CD16), which—in context with the previously humanized CD30 Fab fragment—provides the necessary component for the design of a clinically useful bispecific antibody. Thus, the CDRs of the anti-CD16 mouse IgG1/
MAb A9 were grafted onto human Ig sequences. In a first attempt, the murine V domain was converted to a humanized chain, which led, however, to complete loss of antigen-binding activity and extremely poor folding efficiency upon periplasmic expression in Escherichia coli. Hence, its CDRs were transplanted onto a human 
light chain in a second attempt, which resulted in a functional recombinant Fab fragment, yet with 100-fold decreased antigen affinity. In the next step, an in vitro affinity maturation was performed, wherein random mutations were introduced into the humanized VH and V domains through error-prone PCR, followed by a filter sandwich colony screening assay for increased binding activity towards the bacterially produced extracellular CD16 fragment. Finally, an optimized Fab fragment was obtained, which carries nine additional amino acid exchanges and exhibits an affinity that is within a factor of 2 identical to that of the original murine A9 Fab fragment. The resulting humanized Fab fragment was fully functional with respect to binding of the recombinant CD16 antigen in enzyme-linked immunosorbent assay and in cytofluorimetry with CD16-positive granulocytes, thus providing a promising starting point for the preparation of a fully human bispecific antibody that permits the therapeutic recruitment of NK cells.
Keywords: affinity maturation/CD16/immunoglobulin subclass/light chain/natural killer cells
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionFundingAcknowledgementsReferences |
|---|
Natural killer (NK) cells are a major component of the cellular immune response and can directly destruct foreign or infected tissue (Trinchieri, 1989
). They possess a large reservoir of cytolytic granules, i.e. secretory lysosomes containing perforin and granzymes. Upon encounter with a susceptible target cell, these granules traffic to the contact zone and their content is released to effect lysis.
FcRIII (CD16), the low-affinity receptor for the immunoglobulin G (IgG) Fc region expressed on human large granular lymphocytes, has been found to be an efficient trigger of NK cell-mediated cytotoxicity (Fanger et al., 1989). Human FcRIII receptors exhibit an extracellular Ig-like region of 190 amino acids that folds into two extensively glycosylated C2-type domains, each containing one disulfide bridge (Sondermann et al., 2000). Owing to the heterogeneous glycosylation pattern, the apparent molecular weight lies in the range between 50 and 80 kDa. FcRIII binds IgG in the form of immune complexes, with a specificity for human IgG1 and IgG3 but minimal binding of IgG4 and IgG2 (Gessner et al., 1998).
Two separate genes encoding FcRIII, III-A and III-B have been identified, showing a high mutual sequence identity of 97% (Ravetch and Perussia, 1989; Gessner et al., 1995). FcRIIIa and FcRIIIb show amino acid differences responsible for distinct post-translational modifications such as N-linked glycosylation and membrane anchoring (Gessner et al., 1998). The presence of Ser at position 187 of the mature FcRIIIb (numbering according to Swiss-Prot accession no. O75015
[GenBank]
, with the putative signal peptide of 16 amino acids omitted) is essential for attachment of a glycosylphosphatidylinositol anchor on neutrophils. FcRIIIa, which is expressed on NK cells and monocytes, contains Phe at the same position, which results in a transmembrane receptor isoform including a cytoplasmic tail of 25 amino acids.
Several polymorphisms of FcRIII that influence the binding of IgG have been described. On NK cells and monocytes, three polymorphisms are known for FcRIIIa. The first represents a triallelism in the N-terminal membrane-distal Ig domain, leading to Leu, Arg or His at position 50 of the mature protein (numbering according to Swiss-Prot accession no. P08637
[GenBank]
) (de Haas et al., 1996). The second and third polymorphisms are localized in the membrane-proximal Ig domain close to the Ig-binding site, comprising Phe or Val at positions 141 and 160 (Koene et al., 1997; Wu et al., 1997). Notably, the Val160 isoform shows higher affinity for IgG1, IgG3 and IgG4 compared with Phe160 (Gessner et al., 1998).
Multiple interaction of antigen-specific IgG bound to the surface of a target cell with the FcRIIIa receptor on NK cells induces receptor cross-linking. This results in increased intracellular Ca2+ levels, which trigger a signaling cascade similar to the one activated by the T-cell receptor (Mandelboim et al., 1999), thus culminating in the destruction of the target cell through antibody-dependent cellular cytotoxicity (Lanier, 1998). Interestingly, certain monoclonal antibodies (MAbs) that were raised against CD16 itself can also effect NK cell activation, which has been successfully used to construct ‘armed’ bispecific antibodies for immunotherapy (Carter, 2001). Such bispecific antibodies typically bind with one arm to the tumor cell and with the second arm they trigger the receptor on an immune effector cell.
A promising example for this strategy is the CD30xCD16 bispecific mouse quadroma antibody HRS3/A9 (Hombach et al., 1993), which recognizes the tumor marker CD30 on Hodgkin–Reed–Sternberg (HRS) cells and CD16 on NK cells. Binding of the A9 arm to an epitope of the CD16 antigen directly activates the NK cell without need for pre-stimulation. Thus, in this case monovalent binding seems to be sufficient to induce ADCC by NK cells, whereas binding of antibodies directed against other epitopes on CD16 requires additional signals (Hombach et al., 1993). Data from in vitro and in vivo experiments as well as from a clinical study impressively demonstrated the general applicability of this approach for the treatment of Hodgkin’s disease (Hombach et al., 1993; Renner et al., 1994; Renner et al., 1997; Hartmann et al., 2001).
Unfortunately, however, a pronounced human anti-mouse antibody (HAMA) response occurred in these patients, which prevents prolonged application of the murine bispecific antibody and poses demand for alternative reagents with less immunogenic potential. For example, the construction of a corresponding diabody was described (Arndt et al., 1999), which comprises just the variable domains of the CD30 and CD16 MAbs. This diabody retained the ability to activate and target NK cells both in cellular cytoxicity experiments in vitro and in mice with xenotransplanted Hodgkin’s lymphoma in vivo. Yet, this antibody fragment still carries the framework sequences of murine origin.
Immunogenicity of antibodies can be effectively reduced by CDR-grafting (Jones et al., 1986; Winter and Harris, 1993; Almagro and Fransson, 2008). This strategy was already successfully applied to the Fab fragment of the CD30-binding MAb HRS3 (Schlapschy et al., 2004). Here we describe the humanization of the MAb A9, which provides the CD16-binding arm of the clinically validated CD30xCD16 bispecific mouse quadroma antibody, thereby converting the murine V into a human V domain.
|
Materials and methods |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionFundingAcknowledgementsReferences |
|---|
Materials
Escherichia coli K-12 strain JM83 (Yanisch-Perron et al., 1985) was used for cloning and production of recombinant Fab fragments, whereas W3110 (Bachmann, 1972) was employed for the fermenter production of the CD16 fragment. The vectors pASK102-D1.3 (Schlapschy et al., 2004), pASK111 (Vogt and Skerra, 2001), pASK88 (Schiweck and Skerra, 1995) and pASK106 (Fiedler et al., 2002) were all based on the generic expression plasmid pASK75 (Skerra, 1994b), carrying the chemically inducible tetracycline promoter/operator. The chromatography material with immobilized engineered streptavidin (Voss and Skerra, 1997) was prepared by I. Theobald. The anti-CD30 MAb (Engert et al., 1990) and the cloned FcRIIIa cDNA (pCRII-CD16-2) were kindly provided by Prof. Dr. C. Renner.
Construction of pASK111–CD16 and pASK–IBA4–CD16–His6
The coding sequence of CD16, corresponding to amino acids 21 to 194 of the gene product (Swiss-Prot accession no. P08637
[GenBank]
), was amplified from pCRII-CD16-2 using a published procedure (Skerra, 1992). The primer sequences were 5'-GCATGCCGGCCGAAGATCTCCCAAAGGCTG-3' and 5'-GCTACCTTGAGTGATGGTGATGTTCAC-3', introducing, respectively, a recognition sequence for EagI and a blunt end compatible with Eco47III (both underlined). The unique amplification product was cut with EagI and ligated with the vector fragment of pASK111, which had been cut with BsaI and Eco47III. On the resulting plasmid pASK111-CD16, the extracellular region of CD16 was fused with the OmpA signal peptide at its N-terminus and with the Strep-tag II (Schmidt and Skerra, 2007) at the C-terminus. The insert was checked by DNA sequencing using an ABI-PrismTM 310 Genetic Analyzer (Perkin-Elmer Applied Biosystems, Weiterstadt, Germany) with the BigDyeTM terminator kit.
For the production of a CD16 fusion protein with both an N-terminal Strep-tag II and a C-terminal His6-tag, the CD16 gene was amplified from pASK111-CD16 with the primers 5'-GTAGCGGGCGCCCGGACTGAAGATCTCCCAAAGGCTG-3' and 5'-CAGGTCAAGCTTAATGGTGATGGTGGTGATGAGCGCTACCTTGAGTGATG-3', carrying recognition sequences for NarI and HindIII (both underlined). The digested amplification product was ligated with the likewise cut vector pASK-IBA4 (IBA, Göttingen, Germany), yielding pASK-IBA4-CD16-His6.
Cloning of the variable genes of the antibody A9
The genes for the variable (V) domains of the heavy and light chains of the murine (mu) MAb A9 were cloned on pASK88, a vector permitting the production of a corresponding chimeric Fab fragment with human (hu) constant domains of subclass IgG1/, which also provides an His6-tag at the C-terminus of the heavy chain (Skerra, 1994a; Schiweck and Skerra, 1995). Total cellular RNA of the hybridoma A9 (IgG1/) (Hombach et al., 1993) was isolated according to a standard guanidinium-isothiocyanate protocol (Kingston et al., 2005). First-strand cDNA was synthesized from 
5 µg purified, ethanol-precipitated RNA using oligo(dT)12–18 as primer and 400 units SuperScript RNase H– reverse transcriptase (Life Technologies, Gaithersburg, MD, USA) at 42°C in a total volume of 50 µl according to the supplier’s protocol. An aliquot (1 µl) of the cDNA solution was used in polymerase chain reactions (PCRs) for the separate amplification of the variable domains VH and V of the heavy and light chains.
Polymerase chain reaction was performed in a total volume of 50 µl using Pfu DNA Polymerase (Stratagene, La Jolla, CA, USA) according to a published procedure (Skerra, 1992). VH was amplified with the primers VHc3-T, 5'-GAGGTSMARCTGCAGSAGTCWGG-3' (the PstI restriction site is underlined), and VHr2-T, 5'-TGAGGAGACGGTGACCGTGGTSCCTTGGCCCC-3' (the BstEII site is underlined). V was amplified with the primers VLc2-T, 5'-GACATTGAGCTCACCCAGTCTCCAGCAATCATGKCTGC (the SstI site is underlined), and VLr1-T, 5'-CCGTTTCAGCTCGAGCTTGGTSCCWSCWCCDAACGT-3' (the XhoI site is underlined). Unique PCR products with 351 and 329 bp, respectively, were obtained and isolated from agarose gels, followed by phosphorylation using T4 polynucleotide kinase (New England Biolabs, Ipswich, MA, USA). First, the VH and V gene fragments were separately cloned on pASK75, which was cut to yield blunt ends using StuI and desphosphorylated with shrimp alkaline phosphatase (GE Healthcare Buchler, Braunschweig, Germany), resulting in the intermediate plasmids pASK75-IntVL-A9 and pASK75-IntVH-A9.
After sequencing, the V gene was excised with XhoI and, by partial digest, with SstI and ligated with the similarly cut expression vector pASK102-D1.3 (Schlapschy et al., 2004), yielding pASK102-D1.3-IntVL-A9. Then, the VH gene was excised from pASK75-IntVH-A9 with BstEII and, by partial digest, with PstI and ligated with the likewise cut vector pASK102-D1.3-IntVL-A9 from above, yielding pASK102-muA9. For bacterial production of a similar muA9 Fab fragment carrying the His6-tag—instead of the modified Strep-tag II with an additional C-terminal Cys—the VH and V gene sequences were subcloned on pASK88 utilizing the conserved XbaI/BstEII and NcoI/XhoI restriction sites (Skerra, 1994a), yielding pASK88-muA9.
Synthesis of the humanized variable genes by PCR assembly
The humanized variable gene regions were each assembled from six partially overlapping oligodeoxynucleotides (see Results) by PCR according to a previously published strategy (Schlapschy et al., 2004). In both cases, the unique amplification product was cut with PstI/BstEII (VH) and SacI/XhoI (VL), respectively, and separately cloned on pASK88. Inserts with a correct sequence were combined on the same plasmid utilizing the singular XbaI and NcoI restriction sites, yielding pASK88-huA9/Kol or pASK88-huA9/HCS.
The VH and VL encoding regions were individually amplified from pASK88-huA9/HCS under mutagenic conditions (Casson and Manser, 1995) using the flanking primer pairs VHx/VHy (see text) and VLx/VLy according to a published procedure (Schlapschy et al., 2004). Alternatively, the shorter primers VHxII (5'-GAAGTTCAACTGCAGGAG-3') and VHyII (5'-CCTCCTTTGCCACTGGTC-3') were used for VH, with VHxII introducing a Gln6Glu substitution (Arndt et al., 1999). After digest with PstI and BstEII or with SacI and XhoI, the DNA fragments were ligated with the appropriately cut pASK106-huA9 vector, which coded for the wild-type huA9 Fab fragment as a fusion with a bacterial albumin-binding domain (König and Skerra, 1998).
Filter sandwich colony screening assay
The colony screening assay for antigen-binding activity of the bacterially produced Fab fragments was performed as described previously (Fiedler and Skerra, 2001). To probe antigen-binding activity of the Fab variants, which had been secreted by E.coli colonies supported on a first membrane and captured on a hydrophobic second membrane, purified recombinant CD16 was labeled with digoxigenin (DIG) at a molar ratio of 2:1 (Schlehuber et al., 2000). After washing the membrane three times with PBS/T [PBS, 4 mM KH2PO4, 16 mM Na2HPO4, 115 mM NaCl, pH 7.4, containing 0.1% (v/v) Tween 20], incubation was performed for 1 h with 10 ml PBS/T containing either 1 µM or 500 nM of the CD16–DIG conjugate. After washing, bound antigen was detected with anti-DIG Fab fragment conjugated with alkaline phosphatase (Roche Diagnostics, Penzberg, Germany), and the signals were developed in a chromogenic reaction. The colonies corresponding to the most intense signals were identified on the first membrane and propagated for subsequent analysis.
Recombinant protein production and purification
The CD16 fragment was produced in E.coli W3110 harboring pASK111-CD16 (for CD16-StrepII) or pASK-IBA4-CD16-His6 (for StrepII-CD16-His6) using an 8 l bench top fermenter with a synthetic glucose mineral medium supplemented with 30 mg/l chloramphenicol or 100 mg/l ampicillin (Amp), respectively, in a similar manner as described previously for the production of Fab fragments (Schiweck and Skerra, 1995; Fiedler and Skerra, 2001). Each time the recombinant protein was purified through the Strep-tag II (Schmidt and Skerra, 2007). In case of the double-tagged protein, an immobilized metal affinity chromatography (IMAC) on Zn(II)-charged IDA Sepharose (Skerra, 1994a) was performed as second purification step. To this end, 0.5 M betain was added and two elution steps with 50 and 300 mM imidazole were performed in order to first remove unspecifically bound host cell proteins and then recover the StrepII-CD16-His6 protein, respectively. To avoid metal-dependent aggregation, EDTA was added immediately after elution to a final concentration of 10 mM.
The chimeric A9 Fab fragment and its humanized variants were produced at 22°C in E.coli JM83 using 2 l shake flask LB cultures (Sambrook et al., 1989) supplemented with 100 mg/l Amp as described earlier (Fiedler and Skerra, 1999). The recombinant protein was purified through IMAC either on Zn(II)-charged IDA sepharose (Skerra, 1994a) or on a POROS MC column (Perseptive Biosystems, Framingham, MA, USA). Fractions were analyzed for purity through SDS–PAGE (Fling and Gregerson, 1986), appropriately pooled, and finally dialyzed against PBS.
Protein concentrations were determined according to the absorption at 280 nm using calculated extinction coefficients (Gill and von Hippel, 1989) of 44 500 M–1 cm–1 for the StrepII–CD16–His6 antigen fragment, and of 81 146 M–1 cm–1 and 82 800 M–1 cm–1 for the chimeric A9 and humanized huA9 Fab fragments, respectively.
Enzyme-linked immunosorbent assay
ELISA experiments were performed according to a published protocol (Schlapschy et al., 2004). Shortly, a 96-well microtitre plate (12 x 8 well ELISA strips with high binding capacity; Greiner, Frickenhausen, Germany) was coated with 50 µl of purified StrepII-CD16-His6 at a concentration of 20 µg/ml in 5% (w/v) NaHCO3. After blocking with BSA, anti-CD16 MAbs or Fab fragments were applied in a dilution series. Bound MAbs or Fab fragments were detected by anti-mouse Fc-specific IgG/AP or anti-human C-light-chain IgG/AP conjugate (Sigma, St. Louis, MO, USA), respectively, followed by chromogenic reaction. Data were fitted by non-linear least squares regression according to the Law of Mass Action using KaleidaGraph software (Voss and Skerra, 1997).
Real-time biomolecular interaction analysis
The affinities of the anti-CD16 Fab fragments for the recombinant CD16 antigen fragment were measured through surface plasmon resonance (SPR) (Jonsson et al., 1991) on a BIAcore X instrument (BIAcore, Uppsala, Sweden) as described earlier (Schlapschy et al., 2004). Shortly, after buffer exchange to 10 mM sodium acetate, pH 3.85, purified CD16 was diluted to 20 µg/ml and immobilized on a ‘research grade’ CM5 sensor chip through the amine coupling kit (BIAcore), resulting in ca. 1700 response units (RU). The Fab fragments were applied in PBS/P [PBS containing 0.005% (v/v) surfactant P-20; BIAcore] at appropriate concentrations. Complex formation was observed under continuous buffer flow of 5 µl/min. For determination of KD, an equilibrium analysis was performed using the BIAevaluation software, V 3.0 (BIAcore) (Huber et al., 1999).
Flow cytometric analyses were performed with CD16-positive granulocytes using a FACScanTM Calibur cytofluorimeter (Becton Dickinson, Mountain View, CA, USA). A total of 5 x 105 granulocytes isolated from the blood of a healthy donor by centrifugation on a discontinuous density gradient of Histopaque-1077 and Histopaque-1119 (Sigma-Aldrich) were incubated for 30 min at 4°C in a 96-well cell-culture plate (Greiner Bio-One) with 5 µg purified Fab fragment in a total volume of 50 µl PBS containing 2% (w/v) BSA. Cells were pelleted at 400 g and 4°C for 10 min and washed twice with 200 µl PBS/BSA. Then the cells were resuspended in the same buffer containing 5 µl mouse anti-human C PE F(ab)2 conjugate (0.13 µg/µl; Dianova, Hamburg, Germany) and incubated for 30 min at 4°C. After pelleting and washing as described above, the cells were resuspended in 500 µl PBS containing 0.05% (w/v) NaN3 and counted in the FACS instrument. The muHRS3c Fab fragment directed against the CD30 antigen (Schlapschy et al., 2004) was applied under the same conditions, serving as negative control.
Database analysis and molecular modeling
The most appropriate human CDR acceptor framework was selected by homology search of immunoglobulin sequences from the Kabat database (Johnson and Wu, 2001) and the RSCB Protein Data Bank (Berman et al., 2000). The Fv moieties from the crystal structures of the Fab fragments KOL (human; light chain; PDB code 2FB4
[PDB]
), HC19 (murine; light chain; PDB code 1GIG), and hu4D5 (humanized; kappa light chain; PDB code 1FVC) were superimposed through the C coordinates of the framework residues and graphically displayed using PyMol (www.pymol.org).
|
Results |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionFundingAcknowledgementsReferences |
|---|
Preparation of the soluble recombinant CD16 antigen
On the basis of the published crystal structure of the Fc receptor III (Sondermann et al., 2000), the coding sequence for residues 21–194 of the CD16 gene product, comprising the two Ig-like extracellular domains, was cloned on the vector pASK111 (Vogt and Skerra, 2001) in a similar manner as previously described for the CD30 antigen (Schlapschy et al., 2004). E.coli W3110 harboring pASK111-CD16 was used for the laboratory fermenter production of the CD16 antigen fragment, followed by purification from the periplasmic cell fraction through the Strep-tag II (Schmidt and Skerra, 2007). However, the resulting CD16 antigen fragment was still contaminated with host cell proteins and showed aggregation within a few hours at 4°C as well as upon freezing.
Thus, an alternative construct was designed with the Strep-tag II attached to the N-terminus and the His6-tag appended at the C-terminus. Again, the recombinant CD16 protein was produced in a bench-top fermenter, using E.coli strain W3110 haboring pASK-IBA4-CD16-His6. The antigen fragment was first purified from the periplasmic extract by streptavidin affinity chromatography and then by IMAC. In this case the protein turned out to be less prone to aggregation. SDS–PAGE revealed (Fig. 1) that, in the reduced state, the recombinant StrepII-CD16-His6 appeared as a single homogeneous band with an apparent molecular size of 31 kDa, which is significantly larger than the calculated mass of 22 310 Da. When not reduced prior to gel electrophoresis, the mobility was significantly increased, thus indicating complete formation of the two disulfide bonds.
|
The functionality of the recombinant CD16 fragment was first investigated in an ELISA (Fig. 2). The bispecific mouse quadroma antibody A9/HRS3 (Hombach et al., 1993
) gave rise to a pronounced binding signal with a typical saturation curve, whereas no signal was detected with the anti-CD30 MAb or with the secondary anti-mouse antibody AP-conjugate alone. Hence, the epitope recognized by the anti-CD16 arm of the bispecific MAb was also present on the cloned fragment of CD16. The affinity of the bispecific, monovalent CD30xCD16 antibody for the recombinant CD16 antigen was determined by SPR, revealing a rather moderate dissociation constant (KD) of 319 ± 42 nM (cf. Table I).
|
A405/
|
Cloning of the variable gene regions of muA9 and production as a recombinant Fab fragment
The genes for the VH and VL domains of muA9 were cloned from cDNA of the A9 hybridoma cell line. After subcloning on pASK88, which encodes the His6-tag at the C-terminus of the human heavy chain constant region (Schiweck and Skerra, 1995), the chimeric Fab fragment was produced in E.coli JM83 and purified from the periplasmic cell fraction by IMAC (Skerra, 1994a). The yield was astonishingly low with 0.04 mg/l shake flask culture.
The antigen-binding activity of the muA9 Fab fragment was investigated in an ELISA. A pronounced binding signal with a typical saturation behavior was observed (see below), indicating that the Fab fragment recognized the epitope on the recombinant CD16 antigen, similarly as the corresponding murine bispecific quadroma antibody. Measurement of its antigen affinity by SPR revealed a KD value of 49.1 ± 2.8 nM, which corresponds quite well with the reported affinity of 37.7 nM for the bispecific, monovalent A9/HRS3 MAb towards the native antigen produced in cell culture (Renner et al., 2001). However, this KD value was by a factor six better than the affinity measured for the bispecific, monovalent A9/HRS3 MAb for the recombinant CD16 fragment (cf. Table I).
Humanization of the muA9 Fab fragment using the human IgG1/ antibody KOL as template
The variable domains of muA9 were humanized through CDR grafting (Riechmann et al., 1988). As the VL domain of muA9 belongs to the murine class, a search for similar human variable domains was performed in the RCSB Protein Data Bank (Berman et al., 2000) as well as in the Kabat database (Johnson and Wu, 2001). Only a few human antibodies showed reasonable sequence identity, among those the human myeloma protein KOL (Huber et al., 1976) as the most promising one. The framework of the muA9 V domain had 39 mismatches (51% identity for 79 aligned residues) with the KOL V domain while the muA9 VH framework showed 40 mismatches (54% identity for 87 aligned residues) compared with the KOL VH domain. Framework residues likely to interfere with the CDR conformations were identified by assessing their structural influence on the basis of the KOL crystal structure (PDB code 2FB4
[PDB]
) in comparison with the crystal structure of the anti-influenza virus hemagglutinin murine Mab HC19 (Bizebard et al., 1994), which showed high structural similarity in the Fv framework regions and served as model for the mouse antibody (see Discussion). As a result, several residues of the human KOL sequence were replaced by the original murine A9 amino acids, in line with established strategies (Schlapschy et al., 2004).
In the case of V, GlyL49 [numbering according to Kabat et al. (1991)] of the murine sequence is part of the ‘vernier’ region (Foote and Winter, 1992) and may contact CDR-L2 and influence its orientation. At positions L66 and L67, the murine residues Leu and Ile were retained (instead of Lys and Ser) because they are part of a loop that flanks and stabilizes the conformations of CDR-L1 and CDR-L2, whereby L66 again represents a vernier position. Furthermore, owing to practical considerations with respect to the cloning of the V gene on the expression vector pASK88, ValL3 of the KOL V sequence was replaced by Glu, thus allowing the introduction of an SacI restriction site. Likewise, at the C-terminus of V, the Val and Thr residues at positions L104 and L105 were replaced by Leu and Glu, respectively, allowing the introduction of an XhoI restriction site. The amino acids at the very N-terminus (Asp–Ile instead of Gln–Ser) and C-terminus (Ile–Lys instead of Val–Leu) of the final construct were plasmid encoded (Skerra, 1994a).
In case of the VH domain, the residue 71 was previously proposed to be critical for the conformation of CDR-H2 (Tramontano et al., 1990; Carter et al., 1992). Hence, the murine Ala was kept at this position instead of Arg present in the KOL VH domain. Again, for cloning of the VH gene on pASK88, a PstI restriction site was introduced close to the 5' end, generating a codon for Gln at position H5 instead of Val present in KOL. The three amino acids at the N-terminus (Glu–Val–Lys) of the final construct were plasmid encoded, leading to a Lys at position H3 instead of Gln present in KOL. The BstEII restriction site could be introduced close to the 3' end of the VH gene without changing the KOL amino acid sequence.
Most amino acids at framework positions of the VL/VH interface (Chothia et al., 1985) were identical between the muA9 and the KOL sequence except for residues L36, L38, L44, L46 and L87. Thus, in the case of V, all the murine residues, Val36, Glu38, Phe44, Gly46 and Phe87, were substituted with the human counterparts (Tyr36, Gln38, Pro44, Leu46 and Tyr87) to ensure correct association with the humanized VH domain.
The amino acid sequences of both humanized variable domains were backtranslated into a nucleotide sequence (Fig. 3) using preferred codon frequencies of E.coli (Bennetzen and Hall, 1982), and potential mRNA secondary structure was minimized. For each variable gene, a set of six overlapping oligodeoxynucleotides with lengths between 54 and 85 bp was designed following a published strategy (Essen and Skerra, 1994), and gene synthesis was performed in a PCR assembly. In both cases, a single amplification product was obtained, which was cut with appropriate restriction enzymes (Fig. 3A) and separately inserted into the Fab expression vector pASK88. Inserts with the correct nucleotide composition from one VL clone and from one VH clone were combined on a single plasmid, yielding pASK88-huA9/KOL.
|
Expression analysis of E.coli JM83 harboring pASK88-huA9/KOL showed that both chains were detectable in the whole cell protein, but only the heavy chain seemed to be secreted into the periplasm. Subsequent analysis using two hybrid Fab fragments with murine/human variable domain pairing as described for an anti-CD30 Fab fragment (Schlapschy et al., 2004
)—both carrying human constant domains as in the original humanized construct—showed that only the combination muV/huVH could be successfully produced in the periplasm of E.coli and recovered in a soluble form. These findings indicated that the humanized V domain was deficient in terms of protein folding, probably resulting in premature protein aggregation. Humanization of the muA9 Fab fragment using human consensus sequences as template
In a second approach, human framework consensus sequences of VH subgroup III and V subgroup I (Kabat et al., 1991) were used as acceptor frameworks for CDR grafting. The same sequences were already successfully used for the humanization of several other murine antibodies directed against HER2 (hu4D5ver8) (Carter et al., 1992), against CD18 (Eigenbrot et al., 1994) and against CD30 (huHRS3) (Schlapschy et al., 2004). Superposition of the Fv moieties of the crystal structures of anti-HER2 (PDB code 1FVC) and anti-CD18 (PDB code 1FGV) humanized antibody fragments revealed high similarity of the human framework region, indicating that its conformation is essentially independent of the CDRs supported (Eigenbrot et al., 1993, 1994). Thus, the crystal structure of the Fv fragment hu4D5ver8 (PDB code 1FVC) (Eigenbrot et al., 1993) was used to guide another attempt to humanize the VH and V domains of A9.
To this end, the three CDR sequences of each A9 variable domain were grafted onto the corresponding human consensus sequence (Fig. 3B). The muA9 V domain had 35 mismatches in its framework segments (corresponding to 56% sequence identity for 79 aligned residues) compared with the V subgroup I (see Discussion). On the other hand, the muA9 VH framework showed 34 mismatches in the framework segments (corresponding to 61% sequence identity for 87 aligned residues) compared with the human consensus sequence of VH subgroup III. Framework residues likely to interact with the CDRs were identified by assessing their structural influence on the basis of the hu4D5ver8 crystal structure and, again, several residues of the human sequence were restored by the original murine A9 amino acids.
In the case of V, GlyL46 could interact with CDR-H3 of VH and its missing side chain seems also to be important for the contact with the VH framework. An affinity improvement by a factor of 6 was described for a humanized anti-EGF antibody (Baca et al., 1997) upon substitution of the human Leu with Val at this position, as it was present in the original murine antibody. GlyL49 seems to directly interact with CDR-L2. Therefore, at positions L46 and L49 of the vernier region (Foote and Winter, 1992), Gly was kept instead of Leu and the voluminous Tyr residue, respectively, which occur in the human consensus sequence. V residue 66 is usually a Gly both in human and murine -chain sequences (Carter et al., 1992). However, Leu occupies this position in the muA9 light chain. The side chain of this residue is likely to affect the conformations of CDR-L1 and CDR-L2 as well as the hairpin turn at residues 68–69 that connects β-strands D and E. Hence, Leu was kept at this vernier position. The vernier residue L71 may interact with CDR-L1. Consequently, the small amino acid Ala was kept at this position instead of the more voluminous aromatic Phe residue of the human consensus sequence. Furthermore, due to practical considerations with respect to the cloning of the VL gene on pASK88, Glu and Leu were introduced at positions L3 and L4 (instead of Gln and Met from the human consensus sequence) and Leu was introduced at position L104 instead of Val, thus allowing introduction of SacI and XhoI restriction sites as mentioned before.
In case of the VH domain, the human VHIII consensus sequence showed nine differences with muA9 at positions of the vernier region, which were clustered especially at the N-terminal side of CDR-H1 and at the C-terminal side of CDR-H2. The positions 27–30 are part of a larger loop comprising CDR-H1 and thus influencing its correct positioning (Foote and Winter, 1992). Except for H27 (see below) and H30, the amino acids of the murine sequence were kept in this region. The second cluster of differing amino acids occurred between positions 66–80, in a loop that flanks and possibly stabilizes CDR-H2 (Baca et al., 1997). In the human VHIII consensus sequence the positions H71, H73 and H78 are occupied with Arg, Asn and Leu, respectively, whereas the A9 VH domain reveals the corresponding residues Ala, Thr and Ala. It is known that the residue at position H71 is crucial for the conformation of both CDR-H2 and CDR-H3 (Tramontano et al., 1990). CDR-H2 can contact CDR-H3 in different ways, depending on the side chain at this position. A large side chain, such as Arg, keeps the two CDRs apart, whereas a small one, like Ala, enables a close contact. Thus, the murine sequence was kept in this region. Again, for cloning of the VH gene on pASK88, a PstI restriction site was introduced close to the 5' end, generating a codon for Gln at position H5 instead of Val of the human consensus sequence. The three amino acids at the N-terminus (Glu–Val–Lys) of the final construct were plasmid encoded, leading to a Lys at position H3 instead of Gln. The BstEII restriction site could be introduced close to the 3' end of the VH gene without changing the amino acid sequence.
Finally, the correct association of the humanized VH and VL domains was checked because it is important for the gross geometry of the antigen-binding site. Twenty highly conserved positions in the VL/VH interface have been described with six positions (L44, L96, L98, H45, H100 and H103) forming the inner core of the interface (Chothia et al., 1985). Of these 20 residues, the amino acid sequences for the human and the murine domain differ in the framework regions at positions L36, L38, L44, L46 and L87 while the VH domains differ just at position H91. During humanization on the basis of the consensus sequences, all positions of the VL/VH interface, except the ones in the CDRs (L89, L91, L96, H35, H95 and H100) and GlyL46 mentioned above, were changed to the conserved residues according to Bendig and Jones (1996). The fact that there was 90% sequence identity at these 20 positions of the VL/VH contact region and even 100% identity at the 6 core positions (whereby CDR residue H100 was already occupied in muA9 by the conserved Phe side chain) should ensure that the mode of association of the two humanized variable domains remains unchanged.
Taken together, for humanization of the muA9 V domain on the basis of the VH subgroup III and V subgroup I consensus sequences altogether 28 amino acid exchanges were required (excluding the three residues exchanged for cloning purposes) while four amino acids of the murine framework segments were retained. In the muA9 VH domain 27 amino acids had to be exchanged in the murine framework (excluding the two residues exchanged for cloning purposes) while five of the original residues were kept.
Following the methodology described above, the amino acid sequences of both humanized variable domains were backtranslated into a nucleotide sequence, which was synthesized in a PCR assembly (Fig. 3B) and cloned, yielding pASK88-huA9/HCS.
The humanized A9 Fab fragment was produced in E.coli JM83 under the same conditions as for the chimeric Fab fragment carrying the original murine variable domains. The yield was ca. 0.3 mg/l culture (optical density at 550 nm, OD550 = 1) and thus significantly increased. SDS–PAGE under reducing conditions revealed two bands corresponding to sizes around 25 kDa (>98% purity), consistent with the expected masses of the light and heavy chains (see below). Under non-reducing conditions, a single band was observed with the expected mass of ca. 50 kDa. The affinity of the humanized Fab fragment toward the recombinant CD16 antigen was analyzed by SPR, revealing a KD of 4.5 ± 0.8 µM under equilibrium conditions. Thus, antigen-binding activity was in principle retained, but there was a ca. 100-fold decrease in affinity compared with the chimeric muA9 Fab fragment upon humanization (Table I).
Affinity maturation of the IgG1/-humanized A9 Fab fragment
To elucidate which one of the variable domains had the greater contribution to the observed loss in affinity, we produced two hybrid Fab fragments with murine/human variable domain pairing, similarly as described before for the anti-CD30 antibody HRS3 (Schlapschy et al., 2004), and analyzed their binding for the recombinant CD16 fragment by qualitative SPR measurements. As result, both partially humanized Fab fragments showed higher binding activity compared with the fully humanized A9 Fab fragment, whereby a slightly stronger decrease in antigen affinity could be attributed to the humanized VH domain.
Thus, we decided to first improve the huVH domain by combinatorial engineering in context of the original murine V domain. To this end, random amino acid substitutions were introduced into the huVH domain through error-prone PCR. Experimental conditions were chosen such as to achieve approximately one base substitution per 100 bp, which corresponds to the frequency of mutations in B-cells (Berek and Milstein, 1987). The resulting library of mutated huVH domains was cloned and subjected to a filter sandwich colony assay (Skerra et al., 1991; Schlapschy et al., 2004) to screen for hybrid Fab fragments having enhanced binding activity for the recombinant CD16 antigen.
Colonies giving rise to most intense staining signals were recovered and their plasmids encoding corresponding A9 Fab variants were isolated and sequenced (Fig. 4). The mutated VH gene cassettes were then subcloned together with the humanized VL domain on pASK88-huA9/HCS for the production of the corresponding Fab fragments. After purification by IMAC, their antigen-binding activities were compared in an ELISA with the huA9 Fab fragment. The best mutant Fab fragment from this first cycle of affinity maturation, huA9-VH-EP1/1, which carried, among others, a SerH49Gly amino acid substitution, showed just slightly improved binding compared with the original huA9 Fab fragment.
|
Consequently, a second cycle of affinity maturation was performed for VH, this time starting from pASK88-huA9-VH-EP1/1 as template, under the same experimental conditions as before. The best variant from this cycle, huA9-VH-EP2/2, revealed a significantly improved binding activity compared with huA9, with a KD = 1.5 ± 0.2 µM as determined by SPR (Table I). Interestingly, although the mutated variable gene was amplified with the PCR primer VHx (cf. Figure 3), a Gly to Asp amino acid substitution was found at position H16 in huA9-VH-EP2/2, i.e. located within the primer-encoded region.
Further cycles of affinity maturation were performed starting from huA9-VH-EP2/2 using the shorter primers VHxII and VHyII (cf. Figure 4), this time screening the library in context with the humanized VL domain. In this way the variants huA9-VH-EP9/5 and huA9-VH-EP9/7 were isolated, which carried the additional mutations ValH37Ile and MetH82Val, respectively. They revealed clearly improved binding activity compared with huA9-VH-EP2/2 (cf. Table I), yet not approaching that of the original muA9 Fab fragment. Notably, each of these two variants carried one new favorable amino acid substitution. As positive effects may be additive, we combined both substitutions using the unique EcoRV and HindIII restriction sites that had been introduced into the synthetic gene. Indeed, the resulting variant huA9-VH-EP9.5+7 exhibited a significantly better KD = 196 ± 40 nM.
One final cycle of affinity maturation was performed, this time introducing random mutations into the huVL domain and screening the library in context of the huA9-VH-EP9.5+7 domain. After several attempts, the variant huA9-VH-EP9.5+7/VL-EP7.5 was isolated, which showed the triple substitution AsnL34Ser, ThrL69Ala and PheL83Leu (Fig. 4). This engineered Fab fragment exhibited comparable binding activity with muA9 in the ELISA (Fig. 5). The affinity of huA9-VH-EP9.5+7/VL-EP7.5 for the recombinant CD16 antigen was determined by SPR (Fig. 6), revealing a KD = 88± 9 nM (Table I). This value is identical within a factor of 2 with the one measured for the muA9 Fab fragment in the same assay.
|
|
Finally, FACScan experiments were performed with CD16+ granulocytes (Fig. 7). In this experiment, the binding signal of the mutant huA9-VH-EP9.5+7/VL-EP7.5 Fab fragment was almost as high as the one observed with the original muA9 Fab fragment. Thus, the resulting humanized antibody fragment is similarly active in terms of recognizing its native antigen as the original muA9 Fab fragment and, remarkably, it shows significantly higher affinity than the corresponding bispecific quadroma antibody (cf. Table I).
|
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionFundingAcknowledgementsReferences |
|---|
In this study we have successfully humanized the variable region of the CD16-binding MAb A9 through CDR grafting. In this strategy, one usually chooses a framework sequence that is most similar to the murine donor sequence in order to avoid differences in core packing and framework conformation, which could have a deleterious effect on CDR conformation and thus on antigen affinity (Wörn et al., 2000
). Consequently, as the light chain of this MAb belongs to the murine class, we initially intended to retain the same Ig class in the humanized antibody.
In spite of its rather low sequence identity (51% for VL, 54% for VH), the framework of the human IgG1/ antibody KOL (Huber et al., 1976) was chosen as the closest acceptor framework with available crystal structure for grafting of the muA9 CDRs. Up to now, only the VH framework of KOL was successfully employed for the humanization of an anti-FcRI antibody in context with its V domain being humanized on the basis of the myeloma protein REI (Graziano et al., 1995). In this case, the sequence identity between the murine anti-FcRI MAb and the KOL VH domain was significantly better with 79%. A sequence identity of at least 65% has been suggested (Tsurushita et al., 2005) for a successful humanization process, as the antigen affinity of the humanized antibody correlates with the degree of homology between the human and murine framework sequences.
Apart from the overall homology, key positions play a crucial role for successful humanization, for example, at the VH/VL dimer interface. Despite high conservation in this region, uncommon interface residues in the loop donor may have to be retained to achieve full antigen-binding activity after CDR grafting. Differences in the relative orientation of VL and VH by up to 15° among various antibodies have been reported, and considerable deviations in the domain orientation between free and antigen-bound crystal structures of the same antibody fragment have demonstrated the flexibility of this interface (Ewert et al., 2004).
Hence, mutations of these interface residues can have a significant impact on antigen binding. This is of special importance for grafts involving murine V domains, which show several amino acid substitutions at the dimer interface compared with the closest human homologue (Ewert et al., 2004). This is also the case for the V domains of KOL and A9: L36 is predominantly Tyr in human V, but is exchanged to Val in murine V, and similarly at positions L38: Gln versus Glu, L41: Gly versus Asp, L43: Ala versus Leu, L44: Pro versus Phe and L46: predominantly large hydrophobic residues in human V versus Gly in mouse.
A striking example that the dimer interface in a humanized antibody plays a crucial role for the stability and antigen-binding activity was reported with an attempt to humanize the murine anti-GCN4 antibody, which carries a light chain, by grafting its CDRs onto the V-type hu4D5 framework (Wörn et al., 2000). In this case the sequence identity of 43% on the VL side was even lower than in our study. Notably, the resulting ‘-graft’ intrabody had a 105-fold reduced antigen affinity.
In our first attempt to humanize the anti-CD16 MAb A9, we used the KOL framework for CDR grafting (Fig. 8). Thus, we retained the subclass of the murine light-chain framework donor and we reconstituted the dimer interface according to the KOL acceptor. However, insufficient structural compatibility between the huV domain and the grafted CDR loops may have resulted in a loss of folding stability as the humanized A9 V did not lead to the bacterial biosynthesis of a soluble recombinant Fab fragment.
|
So far, examples of successful antibody humanization with a human
chain serving as acceptor are rare. This may partly be due to the fact that chains are more frequent in human antibodies than chains (Mole et al., 1994). In one example the CDRs of a anti-IL-12 antibody from chicken were grafted onto a human germline sequence (Tsurushita et al., 2004). To retain antigen-binding activity altogether 11 murine framework residues had to be reconstituted in the humanized antibody, 6 in V and 5 in VH. In another example, a ‘resurfacing’ approach was used for humanization of a rat anti-CD30 antibody (Routledge et al., 1991). Thus, problems associated with chain association and/or CDR constraints imposed by the new framework where circumvented by keeping the entire VL/VH dimer interface and murine presentation platform.
Alternatively, a rodent V domain can be humanized using a human framework as acceptor of the murine CDRs as shown in our second grafting attempt with the A9 Mab in the present study. Under stability aspects, this approach was already investigated in the example mentioned above, where the CDRs of the anti-GCN4 antibody were grafted onto a human consensus V subgoup I sequence as present in the hu4D5 V framework (Wörn et al., 2000). Indeed, the resulting scFv fragment had a higher folding stability compared with the framework donor hu4D5 and with the original murine anti-GCN4 scFv. However, there was a dramatic loss in antigen affinity as mentioned above, which rendered this humanized antibody fragment useless for therapeutic applications. The authors claimed that this effect was due to the framework causing perturbations of the CDR conformations or changing the relative domain orientation. Even after reconstitution of several residues at the VL/VH interface, the resulting scFv fragment showed still a 10-fold reduced affinity.
When considering these problems, the question arises which would be the best strategy to humanize a rodent antibody by CDR grafting. In fact, a comparison of the amino acid sequence relations reveals that murine framework segments have quite a low sequence identity with their human counterparts while they are similarly related to human consensus framework sequences (Fig. 9). In particular, in the framework segments, the A9 variable domain even shows a slightly higher sequence identity with the human consensus sequence (56%) than with KOL (51%). Notably, this sequence identity with the human consensus sequence was 25% higher than in the example mentioned above, where the CDRs of the anti-GCN4 V where grafted onto the hu4D5 framework (Wörn et al., 2000).
|
Our initial version of the correspondingly humanized A9 Fab fragment showed a 100 times lower affinity toward the recombinant CD16 antigen than the original murine A9 Fab fragment. Probably, this was caused by residual deviating amino acid residues within the framework that have an important structural influence on the correct spatial arrangement and conformation of the CDRs. However, due to the many changes in the framework during the humanization process, their effect is generally difficult to predict (Baca et al., 1997
).
In the first two cycles of the following affinity maturation, the randomized huVH gene libraries were cloned in the context of the murine V domain. Two further affinity maturation procedures, one performed with huVH in the context of huVL and one with huVL in context of the best version of the huVH domain led to the isolation of the optimal variant so far, huA9-VH-EP9.5+7/VL-EP7.5. In this humanized Fab fragment, the huVH and huVL domains carry six and three additional amino acid substitutions, respectively. Analysis of these mutations revealed that altogether three side-chain exchanges occurred at positions that are relevant for the VH/VL domain association: ValH37Ile and LeuH45Pro in huVH and AsnL34Ser in CDR-L1 of huVL. The LeuH45Pro substitution is located in the core of the contact region, where the imino acid Pro should influence the backbone conformation.
Furthermore, two key positions of the vernier region were substituted: SerH49Gly in VH and ThrL69Ala in VL. VH position H49 directly flanks CDR-H2 and seems to influence its conformation. Interestingly, the Ser present in the original huVH domain was back-exchanged with Gly present in the murine VH domain. This result of the combinatorial screening experiment supports our rational strategy otherwise during the humanization. VL position 69 is in a loop in direct neighborhood to the CDR-L1 (Eigenbrot et al., 1993). Substitution of the hydrophilic and more voluminous Thr by the smaller hydrophobic Ala residue may favorably influence the conformation of this segment.
Residues 66–80 in huVH form a loop adjoining the antigen-binding site. Here, the hydrophilic Thr77 of the initially humanized VH domain was substituted by the more voluminous and hydrophobic Ile. This should have an influence on the conformation of CDR-H2. Additionally, the resulting huVH carries two side-chain replacements in framework segments: GlyH16Asp in FR-H1 and MetH82Val in FR-H3. Their effects on binding activity are difficult to rationalize. However, deliberate back-substitution of AspH16 to Gly starting from huA9-VH-EP9.5+7/VL-EP7.5 revealed a clear drop in antigen-binding activity, thus emphasizing the positive influence of Asp at this position (data not shown). The substitution MetH82Val reconstituted the original residue present in muA9.
The extracellular domains of the FcRIII receptors of type A and B differ at altogether nine amino acid positions in the region 21–194, which comprises the recombinant CD16 fragment employed in this study for the screening and binding assays. As some of these amino acid differences could be part of epitopes that are recognized by the humanized A9 Fab fragment, it may be possible that in FACS experiments with FcRIIIb+ granulocytes only cells of certain donors are well recognized. The same may be possible for NK cells due to the known polymorphism of FcRIIIa.
In conclusion, an anti-CD16 Fab fragment was functionally humanized involving a to chain conversion. By introducing six amino acid substitutions in the VH and three in the VL domain, the affinity could be increased by almost a factor 50 with respect to the initial design. Considering other clinically approved antibodies (Newsome and Ernstoff, 2008), an even higher affinity might be desirable. However, in the present case this would only be possible by focusing mutagenesis at the CDRs themselves, because one would have to surpass the affinity of the original murine A9 Fab fragment. On the other hand, the affinity realized here should be sufficient to achieve a therapeutic effect, as the original bispecific murine antibody had shown excellent performance in clinical studies without significant adverse reactions, except for HAMA. In fact, measurement of the dissociation constants by SPR revealed an affinity that was by a factor three higher for our improved humanized huA9-VH-EP9.5+7/VL-EP7.5 Fab fragment if compared with a sample of the clinically tested bispecific HRS3/A9 MAb.
|
Funding |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionFundingAcknowledgementsReferences |
|---|
The authors wish to thank Biotest Pharma GmbH, especially Dr. Matthias Germer and Dr. Michael Kloft, for advice and financial support in frame of a grant by the Bundesministerium für Bildung und Forschung (BMBF project no. 0311759).
|
Footnotes |
|---|
Edited by Dario Neri
|
Acknowledgements |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionFundingAcknowledgementsReferences |
|---|
We thank Dr. Hendrik Gille for performing the FACS measurements and Prof. Dr. Cristoph Renner for providing us with the muA9 cDNA and general advice.
|
References |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionFundingAcknowledgementsReferences |
|---|
Almagro J.C., Fransson J. Front. Biosci. (2008) 13:1619–1633.[Web of Science][Medline]
Arndt M.A., Krauss J., Kipriyanov S.M., Pfreundschuh M., Little M. Blood (1999) 94:2562–2568.
Baca M., Presta L.G., O’Connor S.J., Wells J.A. J. Biol. Chem. (1997) 272:10678–10684.
Bachmann B.J. Bacteriol. Rev. (1972) 36:525–557.
Bendig M.M., Jones T. Antibody Engineering. A Practical Approach.—McCafferty J., Hoogenboom H.R., Chiswell D.J., eds. (1996) Oxford: Oxford University Press. 147–168.
Bennetzen J.L., Hall B.D. J. Biol. Chem. (1982) 257:3026–3031.
Berek C., Milstein C. Immunol. Rev. (1987) 96:23–41.[CrossRef][Web of Science][Medline]
Berman H.M., Bhat T.N., Bourne P.E., Feng Z., Gilliland G., Weissig H., Westbrook J. Nat. Struct. Biol. (2000) 7(Suppl.):957–959.[CrossRef][Medline]
Bizebard T., Daniels R., Kahn R., Golinelli-Pimpaneau B., Skehel J.J., Knossow M. Acta Crystallogr. D Biol. Crystallogr. (1994) 50:768–777.[CrossRef][Medline]
Carter P. Nat. Rev. Cancer (2001) 1:118–129.[CrossRef][Medline]
Carter P., Presta L., Gorman C.M., Ridgway J.B., Henner D., Wong W.L., Rowland A.M., Kotts C., Carver M.E., Shepard H.M. Proc. Natl Acad. Sci. USA (1992) 89:4285–4289.
Casson L.P., Manser T. J. Exp. Med. (1995) 182:743–750.
Chothia C., Novotny J., Bruccoleri R., Karplus M. J. Mol. Biol. (1985) 186:651–663.[CrossRef][Web of Science][Medline]
de Haas M., Koene H.R., Kleijer M., de Vries E., Simsek S., van Tol M.J., Roos D., von dem Borne A.E. J. Immunol. (1996) 156:3948–3955.
Eigenbrot C., Randal M., Presta L., Carter P., Kossiakoff A.A. J. Mol. Biol. (1993) 229:969–995.[CrossRef][Web of Science][Medline]
Eigenbrot C., Gonzalez T., Mayeda J., Carter P., Werther W., Hotaling T., Fox J., Kessler J. Proteins (1994) 18:49–62.[CrossRef][Web of Science][Medline]
Engert A., Martin G., Pfreundschuh M., Amlot P., Hsu S.M., Diehl V., Thorpe P. Cancer Res. (1990) 50:2929–2935.
Essen L.O., Skerra A. J. Mol. Biol. (1994) 238:226–244.[CrossRef][Web of Science][Medline]
Ewert S., Honegger A., Plückthun A. Methods (2004) 34:184–199.[CrossRef][Web of Science][Medline]
Fanger M.W., Shen L., Graziano R.F., Guyre P.M. Immunol. Today (1989) 10:92–99.[CrossRef][Web of Science][Medline]
Fiedler M., Skerra A. Protein Exp. Purif. (1999) 17:421–427.[CrossRef][Web of Science][Medline]
Fiedler M., Skerra A. Gene (2001) 274:111–118.[CrossRef][Web of Science][Medline]
Fiedler M., Horn C., Bandtlow C., Schwab M.E., Skerra A. Protein Eng. (2002) 15:931–941.
Fling S.P., Gregerson D.S. Anal. Biochem. (1986) 155:83–88.[CrossRef][Web of Science][Medline]
Foote J., Winter G. J. Mol. Biol. (1992) 224:487–499.[CrossRef][Web of Science][Medline]
Gessner J.E., Grussenmeyer T., Kolanus W., Schmidt R.E. J. Biol. Chem. (1995) 270:1350–1361.
Gessner J.E., Heiken H., Tamm A., Schmidt R.E. Ann. Hematol. (1998) 76:231–248.[CrossRef][Web of Science][Medline]
Gill S.C., von Hippel P.H. Anal. Biochem. (1989) 182:319–326.[CrossRef][Web of Science][Medline]
Graziano R.F., Tempest P.R., White P., Keler T., Deo Y., Ghebremariam H., Coleman K., Pfefferkorn L.C., Fanger M.W., Guyre P.M. J. Immunol. (1995) 155:4996–5002.[Abstract]
Hartmann F., et al. Clin. Cancer Res. (2001) 7:1873–1881.
Hombach A., Jung W., Pohl C., Renner C., Sahin U., Schmits R., Wolf J., Kapp U., Diehl V., Pfreundschuh M. Int. J. Cancer (1993) 55:830–836.[Web of Science][Medline]
Hu G. DNA Cell Biol. (1993) 12:763–770.[Web of Science][Medline]
Huber A., Demartis S., Neri D. J. Mol. Recognit. (1999) 12:198–216.[CrossRef][Web of Science][Medline]
Huber R., Deisenhofer J., Colman P.M., Matsushima M., Palm W. Nature (1976) 264:415–420.[CrossRef][Web of Science][Medline]
Johnson G., Wu T.T. Nucleic Acids Res. (2001) 29:205–206.
Jones P.T., Dear P.H., Foote J., Neuberger M.S., Winter G. Nature (1986) 321:522–525.[CrossRef][Medline]
Jonsson U., et al. Biotechniques (1991) 11:620–627.[Web of Science][Medline]
Kabat E.A., Wu T.T., Perry H.M., Gottesman K.S., Foeller C. Sequences of Proteins of Immunological Interest (1991) 5th edn. Washington, DC: US Department of Health and Human Services.
Kingston R.E., Chomczynski P., Sacchi N. Current Protocols in Molecular Biology—Ausübel F.M., Brent R., Kingston R.E., Moore D.D., Seidman J.G., Smith J.A., Struhl K., eds. (2005) New York, NY: John Wiley & Sons, Inc.
Koene H.R., Kleijer M., Algra J., Roos D., von dem Borne A.E., de Haas M. Blood (1997) 90:1109–1114.
König T., Skerra A. J. Immunol. Methods (1998) 218:73–83.[CrossRef][Web of Science][Medline]
Lanier L.L. Adv. Exp. Med. Biol. (1998) 452:13–18.[Web of Science][Medline]
Mandelboim O., Malik P., Davis D.M., Jo C.H., Boyson J.E., Strominger J.L. Proc. Natl Acad. Sci. USA (1999) 96:5640–5644.
Mole C.M., Bene M.C., Montagne P.M., Seilles E., Faure G.C. Clin. Chim. Acta (1994) 224:191–197.[CrossRef][Web of Science][Medline]
Newsome B.W., Ernstoff M.S. Br. J. Clin. Pharmacol. (2008) 66:6–19.[CrossRef][Web of Science][Medline]
Ravetch J.V., Perussia B. J. Exp. Med. (1989) 170:481–497.
Renner C., Jung W., Sahin U., Denfeld R., Pohl C., Trumper L., Hartmann F., Diehl V., van Lier R., Pfreundschuh M. Science (1994) 264:833–835.
Renner C., Hartmann F., Pfreundschuh M. Cancer Immunol. Immunother. (1997) 45:184–186.[CrossRef][Web of Science][Medline]
Renner C., et al. Cancer Immunol. Immunother. (2001) 50:102–108.[CrossRef][Web of Science][Medline]
Riechmann L., Clark M., Waldmann H., Winter G. Nature (1988) 332:323–327.[CrossRef][Web of Science][Medline]
Routledge E.G., Lloyd I., Gorman S.D., Clark M., Waldmann H. Eur. J. Immunol. (1991) 21:2717–2725.[Web of Science][Medline]
Sambrook J., Fritsch E.F., Maniatis T. Molecular Cloning: A Laboratory Manual (1989) 2nd edn. Cold Spring Harbor, NY: Cold Spring Harbor Laboratory Press.
Schiweck W., Skerra A. Proteins (1995) 23:561–565.[CrossRef][Web of Science][Medline]
Schlapschy M., Gruber H., Gresch O., Schäfer C., Renner C., Pfreundschuh M., Skerra A. Protein Eng. Des. Sel. (2004) 17:847–860.
Schlehuber S., Beste G., Skerra A. J. Mol. Biol. (2000) 297:1105–1120.[CrossRef][Web of Science][Medline]
Schmidt T.G., Skerra A. Nat. Protoc. (2007) 2:1528–1535.[CrossRef][Medline]
Skerra A. Nucleic Acids Res. (1992) 20:3551–3554.
Skerra A. Gene (1994) a 141:79–84.[CrossRef][Web of Science][Medline]
Skerra A. Gene (1994) b 151:131–135.[CrossRef][Web of Science][Medline]
Skerra A., Dreher M.L., Winter G. Anal. Biochem. (1991) 196:151–155.[CrossRef][Web of Science][Medline]
Sondermann P., Huber R., Oosthuizen V., Jacob U. Nature (2000) 406:267–273.[CrossRef][Medline]
Tramontano A., Chothia C., Lesk A.M. J. Mol. Biol. (1990) 215:175–182.[CrossRef][Web of Science][Medline]
Trinchieri G. Adv. Immunol. (1989) 47:187–376.[Web of Science][Medline]
Tsurushita N., Park M., Pakabunto K., Ong K., Avdalovic A., Fu H., Jia A., Vasquez M., Kumar S. J. Immunol. Methods (2004) 295:9–19.[CrossRef][Web of Science][Medline]
Tsurushita N., Hinton P.R., Kumar S. Methods (2005) 36:69–83.[CrossRef][Web of Science][Medline]
Vogt M., Skerra A. J. Mol. Recognit. (2001) 14:79–86.[CrossRef][Web of Science][Medline]
Voss S., Skerra A. Protein Eng. (1997) 10:975–982.
Winter G., Harris W.J. Immunol. Today (1993) 14:243–246.[CrossRef][Web of Science][Medline]
Wörn A., Auf der Maur A., Escher D., Honegger A., Barberis A., Plückthun A. J. Biol. Chem. (2000) 275:2795–2803.
Wu J., Edberg J.C., Redecha P.B., Bansal V., Guyre P.M., Coleman K., Salmon J.E., Kimberly R.P. J. Clin. Invest. (1997) 100:1059–1070.[Web of Science][Medline]
Yanisch-Perron C., Vieira J., Messing J. Gene (1985) 33:103–119.[CrossRef][Web of Science][Medline]
Received September 30, 2008; revised September 30, 2008; accepted October 13, 2008.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||