PEDS Advance Access originally published online on January 25, 2006
Protein Engineering Design and Selection 2006 19(4):141-145; doi:10.1093/protein/gzj012
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Construction and optimization of a CC49-Based scFv-ß-lactamase fusion protein for ADEPT
Genencor International, a Danisco company, 925 Page Mill Road, Palo Alto, CA 94304, USA
1 To whom correspondence should be addressed at: Volker Schellenberger, PhD, Vice President of Drug Discovery, microPROTEINS, Inc., 1455 Adams Drive, #1120, Menlo Park, CA 94025, USA; E-mail: vschellenberger{at}mproteins.com
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and methodsResultsDiscussionReferences |
|---|
CC49 is a clinically validated antibody with specificity for TAG-72, a carbohydrate epitope that is over-expressed and exposed on a large fraction of solid malignancies. We constructed a single chain fragment (scFv) based on CC49 and fused it to ß-lactamase. The first generation fusion protein, TAB2.4, was expressed at low levels in Escherichia coli and significant degradation was observed during production. We optimized the scFv domain of TAB2.4 by Combinatorial Consensus Mutagenesis (CCM). An improved variant TAB2.5 was identified that resulted in an almost 4-fold improved expression and 2.5° higher thermostability relative to its parent molecule. Soluble TAB2.5 can be manufactured in low-density E.coli cultures at 120 mg/l. Our studies suggest that CCM is a rapid and efficient method to generate antibody fragments with improved stability and expression. The fusion protein TAB2.5 can be used for antibody directed enzyme prodrug therapy (ADEPT).
Keywords: ADEPT/CC49/consensus mutation/scFv/thermostability
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionReferences |
|---|
This paper describes the construction of a novel antibody–enzyme fusion protein for use in antibody dependent enzyme prodrug therapy (ADEPT). The fusion protein combines the proven tumor specificity of antibody CC49 (Muraro et al., 1988
) with the high catalytic activity of ß-lactamase (BLA). CC49 is a second-generation murine antibody with specificity for the carbohydrate epitope TAG-72 (Muraro et al., 1988). Single chain fragments (scFv) based on CC49 have been described but with a relatively low level of expression and a tendency to form aggregates (Milenic et al., 1991; Sawyer et al., 1994; Pavlinkova et al., 1999). BLA is a very versatile enzyme in its use for ADEPT. Recently, we described the construction of a BLA variant with reduced immunogenicity (Harding et al., 2005) that should further increase the therapeutic utility of this enzyme.
The literature contains several reports suggesting that consensus mutations can lead to the identification of protein variants with improved stability (Steipe et al., 1994; Lehmann et al., 2000). Consensus mutations replace an amino acid of a protein that rarely occurs in the same position in homologous proteins with another amino acid that is commonly found in the protein homologs. Unfortunately, the stabilizing effect of individual consensus mutations can be quite small and some consensus mutations can actually have destabilizing effects (Lehmann et al., 2002). Recently, we described Combinatorial Consensus Mutagenesis (CCM) as an approach to simultaneously identify and introduce multiple consensus mutations into a protein (Amin et al., 2004). Using just two consecutive rounds of CCM, we were able to identify a variant of BLA containing 11 consensus mutations that is stabilized by more than 9°. The goal of the present study was to test whether CCM could be used to identify variants of an scFv with improved expression.
For ADEPT, it is important that fusion proteins have sufficient stability and can be produced in sufficient quantities by large-scale manufacturing. In the present study, initial characterization revealed that a CC49 scFv–BLA fusion protein based on the published vH and vL sequences of CC49 had low expression and stability. Using CCM mutagenesis, we identified a CC49 scFv variant with significantly improved stability and expression. The engineered variant, TAB2.5, can be expressed in soluble form at a very high yield in Escherichia coli. The variant showed reduced degradation during production and improved thermal stability. The biological efficacy of the protein is described elsewhere (R.F. Alderson, B.E. Toki, M. Roberge, W. Geng, J. Basler, R. Chin, A. Liu, R. Ueda, D. Hodges, E. Escandon, T. Chen, T. Kanavarioti, L. Babé, P.D. Senter, J.A. Fox, V. Schellenberger, manuscript in preparation).
|
Materials and methods |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionReferences |
|---|
Construction and mutagenesis of TAB2.4
The amino acid sequence of the scFv domain of TAB2.4 is based on the antibody CC49 (Abergel et al., 1993). The vH and vL sequences were joined by a 30 amino acid linker of sequence (GGGGS)6. The corresponding nucleotide sequence was designed based on E.coli codon usage and synthesized by McLab (South San Francisco, CA). The BLA sequence was derived from Enterobacter cloacae (Siemers et al., 1996). The scFv and BLA domains were directly joined without any additional linker peptide. The TAB2.4 expression plasmid pSW239.1 was derived from the plasmid pCB04 (Amin et al., 2004). It contained a chloramphenicol resistant marker and a lac promoter. Escherichia coli TOP10F' (Invitrogen, Carlsbad, CA) cells were used for both plasmid construction and expression in shake flasks.
CCM libraries were constructed using a modified version of Multi-site QuikChange Mutagenesis (Stratagene, Torrey Pines, CA) protocol as described previously (Amin et al., 2004). Phosphorylated mutagenic primers were designed to have 17 homologous nucleotides flanking each side of the mutagenic codon. Libraries ME367 and ME368 used 21 primers at combined concentrations of 0.4 and 2 µM, respectively. After mutagenesis and DpnI digestion, 1.5 µl of PCR mix was transformed into chemically competent E.coli TOP10F' cells followed by selection on LA plates containing 20 mg/l chloramphenicol and 0.1 mg/l cefotaxime. Libraries ME374–ME377 were constructed using the same protocol and subsets of the 21 mutagenic oligonucleotides.
Screening for improved variants
Variants were cultured at 37°C in microtiter plates with 100 µl per well of LB medium containing 5 mg/l chloramphenicol and 0.1 mg/l cefotaxime for 2 days. Proteins were extracted by adding 100 µl of B-PER bacterial lysis reagent (PIERCE, Rockford, IL) to the 100 µl of culture in the wells followed by a 1 h incubation at room temperature with mixing. BLA activity was measured using nitrocefin substrate as described previously (Amin et al., 2004). To measure target binding, plates were coated overnight at 4°C with 100 µl per well of 1 µg/ml of bovine submaxillary mucin (BSM) (Sigma, St Louis, MO) in phosphate-buffered saline (PBS). The next day, plates were washed three times with PBST (PBS containing 0.01% Tween-20) and blocked with 300 µl per well of PBST for 1 h. Subsequently, 10 µl of culture extracts was added to 90 µl PBST in BSM coated wells. The plates were incubated for 1 h at room temperature and washed four times with PBST, and bound BLA activity was measured by adding nitrocefin.
Performance contributions of individual mutations were calculated using Microsoft Excel. The process is illustrated in Figure 2 and has been described in detail previously (Amin et al., 2004).
Expression, purification and biochemical analysis
TAB2.5 was expressed in TOP10F' cells (Invitrogen, Carlsbad, CA). A 5 ml seed culture was grown overnight and subsequently transferred into three shake flasks containing 1 l of Terrific Broth supplemented with 10 mg/l chloramphenicol. Cells were cultured for 48 h at 30°C in an incubator shaker. Cells were harvested and soluble proteins were extracted from the cell paste through B-PER treatment. The soluble cell lysate was collected by centrifugation and loaded onto a column packed with 9 ml of phenylboronic acid agarose resin, PBA (Sigma, A-8312). Prior to protein loading, the column was equilibrated in 20 mM TEA and 0.5 M NaCl at pH 7. Once the sample was loaded, the column was washed with equilibration buffer followed by protein elution with 0.5 M borate buffer pH 7.0 containing 0.5 M NaCl. TAB2.5 was further purified using a Superdex 75 column (Amersham Biosciences, Piscataway, NJ) for multimer and degradation product removal to achieve 99% purity by SDS gel. Protein concentrations were determined by BCA assay (Pierce, Rockford, IL). A total of 3 liters of culture produced 217 mg of crude protein, yielding 44 mg of purified protein. Protein stability was determined by Differential Scanning Calorimetry (DSC). Thermograms were collected on a VP DSC from Microcal using PBS at pH 7.4. The protein concentration varied from 0.04 to 0.7 mg/ml. Data were collected from 25 to 90°C with a scan rate of 90°C/h.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionReferences |
|---|
Construction and initial characterization of a CC49-based scFv–BLA fusion protein
The scFv–BLA fusion protein was constructed by fusing the CC49-derived scFv to the N-terminus of BLA based on the published amino acid sequence of the antibody CC49 (Abergel et al., 1993). The resultant molecule, TAB2.4, has the structure vH-linker-vL-BLA. The scFv portion was constructed by joining the vH and vL domains of CC49 with a synthetic of sequence (GGGGS)6. A long linker sequence was chosen to minimize the formation of multimers (Desplancq et al., 1994; Arndt et al., 1998). The DNA sequence encoding the scFv portion was prepared by DNA synthesis using codons optimized for expression in E.coli. The amino acid sequence of the scFv fragment is shown in Figure 1. The BLA portion of TAB2.4 is based on a lactamase sequence that has been optimized for prodrug activation (Siemers et al., 1996).
|
Although the initial evaluation of TAB2.4 showed that the protein could be secreted by E.coli, the yields were moderate and showed poor reproducibility. Gel analysis indicated that significant degradation of the construct occurred; however, the fusion protein showed binding to BSM, which contains the epitope TAG-72 (Desplancq et al., 1994
). Based on this preliminary characterization, it was concluded that TAB2.4 required optimization for improved expression and stability prior to a more thorough evaluation of its utility in ADEPT. Construction of CCM libraries
The framework amino acid sequence of TAB2.4 was compared with published multiple sequence alignments of antibody variable domains using a structural alignment, as described by J. Carlos (Almagro, Instituto de Biotecnologia, UNAM http://www.ibt.unam.mx/vir/structure/3dalinvh.html), and published amino acid frequencies for all antibody variable domain sequences (http://www.lmb.uni-muenchen.de/groups/BS/vh-mouse.html) (Steipe, 1998). It was observed that the framework of TAB2.5 contains 11 amino acids that occur in the same position in <5% of antibodies. Changing these 11 residues toward the most abundant amino acid would thus result in a relatively small library. Therefore, we decided to replace several of these 11 residues with multiple other amino acids in positions where no clear consensus was apparent. For instance, the Arg residue at position vH-62 in TAB2.4 is observed in <2% of all antibodies. Other antibodies contain Lys (40%) or Ser (37%) in the same position. Thus it was decided to evaluate both mutations. The same criteria were used to choose mutations at other framework positions. We also included two mutations in vH-32; TAB2.4 contains His in that position, which is observed in <4% of all antibodies. vH-32 is the last residue of the H1 CDR and is frequently not critical for antigen binding. Based on these considerations, we chose a total of 21 mutations in 12 positions for mutagenesis of TAB2.4 (Figure 1).
Two libraries, testing different concentrations of mutagenic primers to optimize library quality, were constructed (Amin et al., 2004). We analyzed 96 isolates from each library for expression and binding to the target BSM. In addition, the scFv portion of all isolates was sequenced. Both libraries gave similar results and the data were pooled during further analysis. The pooled library contained 55 single mutants, 26 double mutants and 8 triple mutants, and one clone contained 4 of the planned consensus mutations. With the exception of vH-H40R, we observed all consensus mutations at least once in the library. However, the absolute frequency of individual consensus mutations varied between 1 and 18. All mutagenic primers were used in equimolar concentrations during library construction and the observed bias in mutation frequencies may reflect variations in the quality of oligonucleotides. Although the mutation frequency in the pooled library was lower than anticipated, the experiment generated valuable information regarding the effect of individual consensus mutations on expression and binding.
Effect of consensus mutations on binding and expression of TAB2.4
Most isolates in the libraries showed similar levels of expression, and small improvements that were detected during initial screening could not be confirmed during subsequent analysis. This observation suggested that differences in expression were on the same order of magnitude as experimental noise. Due to the combinatorial nature of the library, most mutations were observed multiple times. This allowed us to extract valuable information by analyzing the dataset as a whole. We have described previously that the effect of individual mutations on the performance of a protein can be estimated by assuming that all mutations make additive contributions to the logarithm of performance (Amin et al., 2004). The logarithm of performance is used for the analysis to reflect the fact that the rate of reactions is typically related to the change in free energy during the reaction. The calculation of performance parameters for a library is illustrated in Figure 2. We applied the analysis to the expression and binding data for the combined library and obtained a set of parameters for each consensus mutation as shown in Figure 3. The analysis shows that five mutations at positions vH-H32 and vH-A33 lead to very significant reductions in target binding. These residues flank the H1 CDR of CC49 and our data strongly suggest that both residues are critical for target binding. The same mutations had very little effect on the expression of the fusion protein. In general, the contributions of individual mutations to protein expression were much smaller than contributions to target binding.
|
|
Construction and analysis of a secondary CCM library
New libraries were constructed that avoided mutations in proximity to the CDR H1 region. Two libraries, ME374 and ME375, used 16 mutagenic oligonucleotides by avoiding mutations in positions vH-H32 and vH-A33 from the initial set of 21 mutations. A third library, ME377, was built using only five primers, which encoded changes that showed promising results in our previous analysis (vH-D8G, vH-N40A, vH-N40P, vH-V80L and vL-A53T). Only the library ME374 contained a large number of variants with multiple consensus mutations (Table I). We combined data from all libraries to improve the sample size during statistical analysis, which was performed as described earlier. Parameters for the individual mutations are shown in Figure 4. Comparison between Figures 3 and 4 shows the importance of avoiding mutations close to H1. A significant fraction of the variants in libraries ME374–ME377 showed small improvements in expression and/or binding. However, as observed before, the effect of individual mutations on expression was small. We chose seven variants, which combined several of the performance enhancing mutations, for further validation. The best isolate, TAB2.5, produced BLA activity that was equivalent to 66 ± 7 mg/l of protein versus 18 ± 5 mg/l for the parent TAB2.4. Western blot analysis using anti-BLA antibody for detection showed that TAB2.5 samples contained significantly more intact fusion protein (91–93%) relative to the parent TAB2.4 (52–62%) (Figure 5). Variant TAB2.5 contains four consensus mutations, vH-V80L, vL-S5T, vL-A53T and vL-K79E. Data in Figure 4 suggest that further improvement in stability of TAB2.5 may be possible by introducing additional mutations with positive performance contributions into the protein. However, TAB2.5 was produced in high amounts in E.coli and most of the protein was obtained as a soluble full-length fusion protein, which suggested that the variant was very suitable for further analysis and evaluation in ADEPT.
|
|
|
TAB2.5 has increased stability relative to its parent TAB2.4
TAB2.5 was produced in shake flask cultures yielding 
120 mg/l of crude protein. The fusion protein was purified by affinity chromatography using phenylboronic acid, which efficiently captures all BLA-containing proteins. Subsequent purification steps resulted in >95% purity as examined by gel analysis and a content of >98% monomer as judged by analytical size exclusion chromatography. As expected, TAB2.4 and TAB2.5 had similar specific activities. The stability of TAB2.4 and TAB2.5 was compared by DSC as shown in Figure 6. Both proteins undergo irreversible denaturation when heated. However, the transition temperature of TAB2.5 was shifted by 2.5°C, which reflects the increased temperature stability of the variant. Interpretation of the data is complicated by the fact that the proteins are composed of two domains, which may denature independently. We have previously studied the thermal denaturation of a closely related variant of free BLA by CD spectroscopy and observed a transition temperature of 57.1°C (Amin et al., 2004). To further evaluate the stability of TAB2.5, we replaced the BLA domain in TAB2.5 with a variant of BLA that is significantly more stable toward proteolysis as well as thermal denaturation (Amin et al., 2004). Incubation of this variant with high concentrations of the non-specific protease subtilisin revealed that the fusion protein retained full binding and catalytic activity under conditions where TAB2.5 loses >90% of its lactamase activity (data not shown). This result indicates that the scFv domain in TAB2.5 is more stable than the scFv domain in TAB2.4 when exposed to a protease.
|
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionReferences |
|---|
A key challenge for ADEPT is the construction of stable and well-expressing scFv–enzyme fusion proteins (Siemers et al., 1997
). The main hindrance appears to be the scFv domain, which has a tendency to show poor expression and to form aggregates (Whitlow et al., 1993; Desplancq et al., 1994; Verma et al., 1998). Here, we show that CCM can be an efficient approach to improve the expression and stability of scFv–BLA fusion proteins. Recently, it has been demonstrated that refolding of an scFv–BLA fusion protein to form inclusion bodies could be significantly improved by selectively replacing rare framework amino acids of the scFv domain (McDonagh et al., 2003). However, protein refolding requires careful optimization of the refolding conditions and it is difficult to use this approach for the high-throughput characterization of protein variants. Therefore, we decided to utilize a soluble expression system that directly yields folded and functional fusion proteins. Using BLA as a reporter allowed the effective assessment of expression and target binding of protein variants based on crude cell lysate thus enabling the high-throughput characterization of variants. Optimization of production levels proved challenging because protein production in microtiter wells tends to have poor reproducibility compared with other biochemical parameters such as binding or stability. Protein production can be dependent on minor differences in temperature, aeration or inoculation density that cannot be sufficiently controlled in a high-throughput screen. Due to this experimental noise, we were unable to directly identify improved variants from a CCM library. However, the combinatorial nature of such libraries allowed us to estimate the effect of individual mutations on protein function by combining all screening data. This enabled the identification of a variant TAB2.5, which contained several expression-enhancing mutations. TAB2.5 showed significantly improved expression and stability compare with the parent protein TAB2.4. Our analysis suggests that further stabilization of the scFv portion of TAB2.5 may be feasible by the introduction of additional consensus mutations. However, the scFv portion of TAB2.5 appears to be at least as stable as the BLA domain and further stabilization may not result in improved in vivo performance.
Our observation that all mutations in the proximity of H1 CDR result in poor target binding is certainly of biological significance. It is likely that the residues vH-H32 and vH-A33 in CC49 resulted from somatic mutations that were introduced during the affinity maturation of CC49 and are critical for target binding.
The value of consensus mutations in improving the stability of proteins has been observed for several proteins (Steipe et al., 1994; Lehmann et al., 2000). CCM mutagenesis allows one to rapidly identify a subset of useful consensus mutations (Amin et al., 2004). Data analysis provides a means to associate performance parameters with all mutations in the library. The present work demonstrates that consensus mutations can also help to improve soluble expression in E.coli.
The fusion protein TAB2.5 can be expressed at 66 mg/l in the active form, which significantly exceeds reported yields for most scFvs (Siemers et al., 1997), and should be sufficient for pre-clinical evaluation and as well as for large-scale production of clinical material. TAB2.5 is a very promising lead for ADEPT because its parent antibody CC49 has shown very good tumor localization in patients in several trials (Murray et al., 1994; McIntosh et al., 1997; Meredith et al., 2003). The preclinical evaluation of TAB2.5 for ADEPT has been initiated.
|
References |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionReferences |
|---|
Abergel,C., Padlan,E.A., Kashmiri,S.V., Milenic,D., Calvo,B. and Schlom,J. (1993) Proteins, 17, 438–443.[Medline]
Amin,N., Liu,A.D., Ramer,S., Aehle,W., Meijer,D., Metin,M., Wong,S., Gualfetti,P. and Schellenberger,V. (2004) Protein Eng. Des. Sel., 17, 787–793.
Arndt,K.M., Muller,K.M. and Pluckthun,A. (1998) Biochemistry, 37, 12918–12926.[CrossRef][Medline]
Desplancq,D., King,D.J., Lawson,A.D. and Mountain,A. (1994) Protein Eng., 7, 1027–1033.
Harding,F.A. et al. (2005) Mol. Cancer Ther., 4, 1791–1800.
Lehmann,M., Kostrewa,D., Wyss,M., Brugger,R., D'Arcy,A., Pasamontes,L. and van Loon,A.P. (2000) Protein Eng., 13, 49–57.
Lehmann,M., Loch,C., Middendorf,A., Studer,D., Lassen,S.F., Pasamontes,L., van Loon,A.P. and Wyss,M. (2002) Protein Eng., 15, 403–411.
McDonagh,C.F., Beam,K.S., Wu,G.J., Chen,J.H., Chace,D.F., Senter,P.D. and Francisco,J.A. (2003) Bioconjug. Chem., 14, 860–869.[CrossRef][Web of Science][Medline]
McIntosh,D.G., Colcher,D., Seemayer,T. and Smith,M.L. (1997) Cancer Biother. Radiopharm., 12, 287–294.[Medline]
Meredith,R., Shen,S., Macey,D., Khazaeli,M.B., Carey,D., Robert,F. and LoBuglio,A. (2003) Cancer Biother. Radiopharm., 18, 393–404.[Medline]
Milenic,D.E., Yokota,T., Filpula,D.R., Finkelman,M.A., Dodd,S.W., Wood,J.F., Whitlow,M., Snoy,P. and Schlom,J. (1991) Cancer Res., 51, 6363–6371.
Muraro,R. et al. (1988) Cancer Res., 48, 4588–4596.
Murray,J.L., Macey,D.J., Kasi,L.P., Rieger,P., Cunningham,J., Bhadkamkar,V., Zhang,H.Z., Schlom,J., Rosenblum,M.G. and Podoloff,D.A. (1994) Cancer, 73, 1057–1066.[CrossRef][Medline]
Pavlinkova,G., Beresford,G., Booth,B.J., Batra,S.K. and Colcher,D. (1999) Nucl. Med. Biol., 26, 27–34.[CrossRef][Medline]
Sawyer,J.R., Schlom,J. and Kashmiri,S.V. (1994) Protein Eng., 7, 1401–1406.
Siemers,N.O., Yelton,D.E., Bajorath,J. and Senter,P.D. (1996) Biochemistry, 35, 2104–2111.[CrossRef][Medline]
Siemers,N.O., Kerr,D.E., Yarnold,S., Stebbins,M.R., Vrudhula,V.M., Hellstrom,I., Hellstrom,K.E. and Senter,P.D. (1997) Bioconjug. Chem., 8, 510–519.[CrossRef][Medline]
Steipe,B. (1998). In Zorbas,H. and Lottspeich,F. (eds), BioAnalytik. Spektrum Akademischer Verlag: Heidelberg, Berlin, pp. 781–813.
Steipe,B., Schiller,B., Pluckthun,A. and Steinbacher,S. (1994) J. Mol. Biol., 240, 188–192.[CrossRef][Web of Science][Medline]
Verma,R., Boleti,E. and George,A.J. (1998) J. Immunol. Methods, 216, 165–181.[CrossRef][Web of Science][Medline]
Whitlow,M. et al. (1993) Protein Eng., 6, 989–995.
Received November 8, 2005; revised December 19, 2005; accepted December 23, 2005.
Edited by Jacques Fastrez
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||