Codon optimization, expression, and characterization of an internalizing anti-ErbB2 single-chain antibody in Pichia pastoris

Protein Expression and PuriWcation 47 (2006) 249–257 www.elsevier.com/locate/yprep

Codon optimization, expression, and characterization of an internalizing anti-ErbB2 single-chain antibody in Pichia pastoris Siyi Hu, Liangwei Li, Jingjuan Qiao, Yujie Guo, Liansheng Cheng, Jing Liu ¤ Laboratory of Molecular and Cellular Immunology, School of Life Sciences, University of Science and Technology of China, Hefei, Anhui 230027, PR China Received 30 August 2005, and in revised form 10 November 2005 Available online 13 December 2005

Abstract Anti-ErbB2 antibodies are used as convenient tools in exploration of ErbB2 functional mechanisms and in treatment of ErbB2-overexpressing tumors. When we employed the yeast Pichia pastoris to express an anti-ErbB2 single-chain antibody (scFv) derived from the tumor-inhibitory monoclonal antibody A21, the yield did not exceed 1–2 mg/L in shake Xask cultures. As we considered that the poor codon usage bias may be one limiting factor leading to the ineYcient translation and scFv production, we designed and synthesized the full-length scFv gene by choosing the P. pastoris preferred codons while keeping the G + C content at relatively low level. Codon optimization increased the scFv expression level 3- to 5-fold and up to 6–10 mg/L. Northern blotting further conWrmed that the increase of scFv expression was mainly due to the enhancement of translation eYciency. Investigation of culture conditions revealed that the maximal cell growth and scFv expression were achieved at pH 6.5–7.0 with 2% casamino acids after 72 h methanol induction. Secreted scFv was easily puriWed (>95% homogeneous product) from culture supernatants in one step by using Ni2+ chelating aYnity chromatography. The yield was approximately 10–15 mg/L. Functional studies showed that the A21 scFv could be internalized with high eYciency after binding to the ErbB2-overexpressing cells, suggesting this regent may prove especially useful for ErbB2-targeted immunotherapy. © 2005 Elsevier Inc. All rights reserved. Keywords: ErbB2; scFv; Pichia pastoris; Codon optimization; Internalization

The ErbB2 receptor tyrosine kinase (Her2/P185) plays an important role in normal cell growth and diVerentiation as well as in human tumorigenesis [1]. A number of monoclonal antibodies (mAbs) have been developed against the extracellular domain (ECD)1 of ErbB2, and they are used as convenient tools in exploration of the mechanisms of ErbB2-mediated signal transduction and in treatment of ErbB2-overexpressing tumors [2,3]. In contrast to the intact antibodies, single-chain antibody (scFv) fragments consisting of the variable domains of heavy chain (VH) and light chain (VL) are more suitable for antibody-based immunotherapy [4]. Recently, anti-ErbB2 scFvs which are selected *

Corresponding author. Fax: +86 551 3601443. E-mail address: [email protected] (J. Liu). 1 Abbreviations used: scFv, single-chain antibody; mAbs, monoclonal antibodies; ECD, extracellular domain; SDS–PAGE, sodium dodecyl sulphate–polyacrylamide gel electrophoresis. 1046-5928/$ - see front matter © 2005 Elsevier Inc. All rights reserved. doi:10.1016/j.pep.2005.11.014

from phage libraries and can be rapidly internalized have been developed as newer tumor targeting molecules [5,6]. In the previous study, we generated an anti-ErbB2 ECD mAb A21, and it was shown to speciWcally inhibit the growth of ErbB2-overexpressing cancer cells both in vitro and in vivo [7,8]. Subsequently, we constructed a single-chain chimeric antibody by fusing the VL and VH domains of A21 mAb to human IgG1 Fc. The engineered antibody was expressed in CHO-GS system to evaluate its potential applications in therapy of ErbB2-overexpressing tumors [8,9]. We also cloned the A21 scFv gene into the pCANTAB5E vector and expressed it in Escherichia coli, but the soluble periplasmic expression was found to be very low (<0.1 mg/L). This did not meet the requirement of large amounts of scFv for studying its biological eVects and therapeutic applications. The yeast Pichia pastoris has been explored to successfully express recombinant antibody fragments with yields typically above several milligram per liter using the

250

S. Hu et al. / Protein Expression and PuriWcation 47 (2006) 249–257

Saccharomyces cerevisiae -factor secretion signal [10–15]. Therefore, we cloned the A21 scFv gene into the pPIC9K vector and employed P. pastoris to express the functional scFv, but the yield did not exceed 1–2 mg/L in the standard shake Xask cultures despite the observed high transcription level. Although the expressing capacity of an heterologous protein in P. pastoris is greatly inXuenced by its inherent properties, the yield can still be signiWcantly enhanced by genetic manipulation of the rate-limiting factors involving gene transcription, mRNA translation, protein folding and secreting [16,17]. As the native A21 scFv gene exhibited a poor codon usage bias for P. pastoris, we considered that it should be one of the limiting factors leading to the ineYcient translation and scFv production. In this paper, we designed and constructed a full-length synthetic gene by changing the codon usage to that preferred by P. pastoris in order to improve the scFv expression. We also investigated the eVects of gene copies and culture conditions on scFv accumulation. Under optimal conditions, the scFv expression was increased up to 15 mg/L in shake Xask cultures. Further studies showed that the scFv could be rapidly internalized into the cytoplasm after binding to the ErbB2-overexpressing cells, making it an attractive targeting vehicle for antibody-based immunotherapy. Materials and methods Yeast strain and culture medium The P. pastoris strain GS115, the expression vectors pPIC9 and pPIC9K were purchased from Invitrogen. MD/ MM medium (1.34% yeast nitrogen base, 400 g/L biotin, 2% agar, 2% dextrose or 1% methanol) was used for selecting transformants with Mut+ or Muts methanol utilization phenotype. YPD-G418 medium (1% yeast extract, 2% peptone, 2% dextrose, 2% agar, and 0.5–3.0 mg/mL G418) was used for selecting multicopy transformants. The P. pastoris cells were cultured in BMGY medium (1% yeast extract, 2% peptone, 1% glycol, 400 g/L biotin, and 0.1 M potassium phosphate, pH 6.0) for growth and in BMMY medium (1% yeast extract, 2% peptone, 400 g/L biotin, 1% methanol, and 0.1 M potassium phosphate, pH 6.0) for induction. Antibody and cell line A21 mAb was puriWed from ascites as described previously [7]. Human breast cancer SKBR3 cells were cultured in Dullbecco’s modiWed Eagle’s medium (Gibco-BRL) supplemented with 10% fetal bovine serum at 37 °C with 5% CO2. Design of the synA21 gene The synthetic gene synA21 was based on the amino acid sequence of A21 mAb (GenBank Accession No. AY077781 for VL and AY077783 for VH) and constructed in VL–(G4S)4–VH mode. A polyhistidine sequence (his6-tag) was introduced to C-terminus of the synthetic gene for

facilitating antibody detection and puriWcation. The tables of P. pastoris preferred codons were according to the results reported by several groups [18–20]. Synthesis of the synA21 gene The synA21 gene was synthesized by recursive PCR (rPCR) strategy [21]. A total of 17 oligonucleotides were designed to have lengths between 55 and 83 nucleotides and overlaps of 19–23 bp with theoretical melting temperature of 55–60 °C using the Oligo 6.0 software. The sequences encoding the VL-linker and VH-His6 were synthesized by Wrst PCR with oligonucleotides P1–P8 and P9–P16, respectively. AmpliWcation was performed with 10 nM of each internal oligonucleotides and 1 M of each 5⬘-Xanking oligonucleotides under optimal conditions: 95 °C for 2 min; 30 cycles of 95, 56, and 72 °C each for 1 min; 72 °C for 10 min. The two products were gel-puriWed, and equal molar of them were mixed for 10 cycles of ampliWcation of 95, 56, and 72 °C each for 1 min. Then, the Xanking oligonucleotides P0 and P16 were added for additional 20 cycles of ampliWcation to generate the full-length gene. Construction of expression vectors The synA21 gene was cloned into the XhoI–EcoRI sites of the pPIC9 vector to generate the pPIC9/synA21 construct. The 1100 bp BamHI–EcoRI digestion products containing the -factor sequence and synA21 gene were cloned into the BamHI–EcoRI sites of the pPIC9K vector to generate the pPIC9K/synA21 construct. Similarly, the native scFv gene (scA21) was fused with a C-terminal His6-tag by PCR ampliWcation from the pCANTAB5E/scFv vector [8] and cloned into the pPIC9K vector to generate the pPIC9K/scA21 construct. Transformation and screening for multicopy transformants Five to ten micrograms of plasmid DNA was linearized with SalI and electrotransformed into P. pastoris strain GS115 (1.5 kV, 200 , 25 F; Bio-Rad Gene Pulser). His+ transformants were selected on histidine-deWcient MD plates. In vivo screening of multiple inserts was according to the Invitrogen pPIC9K Expression Manual and antibiotic G418 was used in three concentrations: 0.5, 1.5, and 3.0 mg/mL. The Mut+ transformants were screened out by replica-plating the G418-resistant transformants on MD/ MM plates. For small-scale culture experiments, single clone was cultured in 2 mL BMGY medium at 30 °C overnight with shaking at 250 rpm. The cells were then resuspended in 2 mL BMMY medium to OD600 nm 1.0 to induce expression by addition of 1% methanol every 24 h. Sodium dodecyl sulphate–polyacrylamide gel electrophoresis Two hundred microliters of culture supernatant was mixed with 1/9 volume of 100% TCA and incubated on ice for 30 min. After centrifugation (15,000g, 15 min, 4 °C), the pellet

S. Hu et al. / Protein Expression and PuriWcation 47 (2006) 249–257

was washed once with ice-cold acetone, air-dried, resuspended in 20 L non-reducing loading buVer, and separated on 12% polyacrylamide gels. The gels were stained with Coomassie brilliant blue R250 and the band densities were evaluated by densitometric scanning. The scFv was quantiWed by comparing the signal densities present in lanes with that present in lanes containing known quantity of puriWed scFv.

251

resuspended in 10-mL BMMY medium to OD600 nm 1.0 in 100-mL baZed Xasks. The phosphate buVer was adjusted to pH 5.0–8.0 with 0.5 pH intervals. The induction was maintained for 6 days by addition of 1% methanol every 24 h. If required, 2% casamino acids or 2.5 mM EDTA was added to the BMMY medium. At the desired time points, 0.2 mL cell aliquots were withdrawn and then replaced with equal amount of fresh medium.

Quantitative ELISA Large-scale expression and puriWcation of scFv Supernatant samples were diluted 100-fold with 50 mM sodium carbonate buVer (pH 9.6) and coated on 96-well microtiter plate (Nunc) overnight at 4 °C. To construct a standard reference curve, a series of dilutions containing 0–10 ng puriWed scFv were included in each assay. The plate was blocked with 1% non-fat milk in TPBS (PBS with 0.05% Tween 20) for 1 h at room temperature. The plate was then incubated with mouse anti-histag antibody (1:2000; Qiagen) followed by HRP-conjugated goat anti-mouse IgG (1:1000; Pierce) each for 1 h at room temperature. The color reaction was developed by addition of the substrate solution (1 mg/ mL O-phenylenediamine and 0.1% H2O2 in 0.1 M citrate buVer, pH 5.5) and stopped by 2 M H2SO4. The absorbance at 490 nm was measured in ELX800 Microplate Reader (BioTek Instruments Inc.). It was observed that there was a close agreement between the values obtained by sodium dodecyl sulphate (SDS)–polyacrylamide gel electrophoresis (PAGE) and quantitative ELISA.

The best clone was cultured in 200 mL BMGY at 30 °C for 2 days. The cells were resuspended in 1 L BMMY medium at pH 7.0 with 2% casamino acids and induced for 3 days. The supernatant was harvested by centrifugation (10,000g, 30 min, 4 °C), Wltered through 0.22-rmum Wlter, dialyzed against binding buVer (20 mM sodium phosphate, 500 mM NaCl, pH 7.2) and loaded onto 3 mL Ni2+ Chelating Sephorase Fast Flow resins (Pharmacia Biotech) at a Xow rate of 3 mL/min. The contaminant materials were washed away with binding buVer containing 0.01 M imidazole. Recombinant scFv was eluted with binding buVer containing 0.1 M imidazole. The appropriate fractions were pooled, buVer exchanged to PBS and condensed by ultraWltration using 10 kDa cut-oV membrane (Amicon, Millipore). Protein concentrations were determined by the Lowry Protein Assay Kit after DOC and TCA precipitation using BSA as standard (Sigma).

Southern and Northern blotting analysis

Determination of scFv binding aYnity by ELISA

The DNA fragment containing the 5⬘ untranslated mRNA leader sequence of AOX1 gene and the -factor signal sequence was PCR-ampliWed from the pPIC9K vector with 5⬘-primer: ACA GAA GGA AGC TGC CCT GTC T and 3⬘-primer: AGC TTC AGC CTC TCT TTT CTC. This fragment was labeled with [-32P]dCTP using the Random Primer Labeling Kit (Takara) and used as the hybridization probe. Genomic DNA or total RNA was extracted from the P. pastoris transformants after 24 h methanol induction using the Yeast Genomic DNA or Total RNA Isolation Kit (Watson’s Biotech). Southern blotting was according to the procedures described by Hohenblum et al. [22] with some modiWcations. Both pPIC9K control DNA and genomic DNA were digested with BglII and XhoI, electrophoresed in 0.8% agarose, transferred to Hibond-N+ nylon membrane, and hybridized with 10-ng labeled probes overnight at 65 °C. For Northern blotting, RNA samples (20 g each) were electrophoresed in 1.2% Mops/formaldehyde agarose and the hybridization was performed at 42 °C. The blots were visualized by autoradiography and the band intensities were quantiWed by densitometric scanning.

For direct binding assay, microtiter plate was coated with puriWed ErbB2 at 0.5, 0.25, 0.125, and 0.63 g/mL, respectively. Serially, threefold diluted scFv from 2.5 £ 10¡4 to 15 g/mL concentration was added for 1 h incubation. The bound scFv was detected with mouse anti-histag antibody and then with HRP-conjugated goat anti-mouse IgG as described before. The aYnity constant was calculated using the formula KaV (M¡1) D 1/2(2[Ab⬘]t ¡ [Ab]t) [23]. For competitive ELISA, microtiter plate was coated with 0.1 g/mL ErbB2. The mixture of A21 at 1.5 g/mL and scFv at various concentrations were added for 1 h incubation. The residual binding of A21 to ErbB2 was detected with HRP-conjugated goat anti-mouse IgG.

Conditions for scFv expression in shake Xask culture A single clone was cultured in 50 mL BMGY medium at 30 °C overnight. To induce expression, cell aliquots were

Internalization of scFv in SKBR3 cells SKBR3 cells were grown on coverslips in six-well plates to 50% conXuency and treated with 20 g/mL scFv or 10 g/mL A21 mAb at 4 or 37 °C each for 2 h. The cells were washed with ice-cold PBS and then with stripping buVer (500 mM NaCl, 0.1 M glycine, pH 2.5) if necessary, Wxed in 4% paraformaldehyde for 10 min, and permeabilized with 0.2% Triton X-100 for 10 min. The cells were saturated with 1% TPBS/BSA for 1 h and then incubated with mouse antihistag antibody (1:50) followed by FITC-conjugated goat anti-mouse IgG (1:200). The location of the antibody in cell

252

S. Hu et al. / Protein Expression and PuriWcation 47 (2006) 249–257

membrane and cytoplasm was detected by Xuorescence microscope with 400£magniWcation scale (Olympus). Antibody internalization assay The eYciency of antibody internalization was quantiWed according to the method described by Neve et al. [6]. Antibodies were labeled with 125I by the Chloramine-T method. SKBR3 cells (5 £ 105) were seeded in six-well plates and incubated with 2 g radioiodinated antibodies (1.3– 1.6 Ci/g) at 37 °C for 2 h. The cells were rinsed four times with Hanks’ balanced salt solution, and two times with icecold stripping buVer. The cells were then lysed with 2 mL of 100 mM TEA for 4 min at 4 °C. Radioactivities of the rinses and the acid-washed cell lysate were determined. Results Design of the synA21 gene As we suspected that the poor codon bias in the native A21 scFv gene may be a possible cause of the ineYcient translation and scFv production in P. pastoris, we designed a codon optimized gene to overcome this problem. Several groups have proposed the tables of P. pastoris preferred codons on the basis of diVerent calculation methods [18–20]. Analysis of their results revealed that a number of amino acids have two preferred codons. Therefore, although the most-preferred codons were used during codon optimization in principle, the second-preferred codons were alternatively used to meet the following considerations: (1) elimination of stable secondary structures such as hairpins exceeding 6 bp especially in the vicinity of the 5⬘ and 3⬘ encoding regions of the corresponding mRNA sequence [27]; (2) prevention of probable depletion and congestion of tRNAs due to consecutively choosing the same codon for the frequently used amino acids; (3) removal of undesirable repeats and false priming events for facilitating subsequent gene synthesis. Finally, a total of 114 non-preferred codons in the native gene were replaced by the P. pastoris preferred codons in the synthetic gene, which represents 44% of the total amino acid sequence (Fig. 1). As a result, codon optimization signiWcantly reduced the overall G + C content of the full-length gene from 56 to 43%. Construct of the synA21 gene The synA21 gene was synthesized by rPCR strategy [21] using a series of overlapping oligonucleotides, which are also shown in Fig. 1. To reduce the possibly introduced errors during PCR ampliWcation, the designed gene was subdivided into two fragments: the VL-linker fragment (480 bp) assembled with oligonucleotides P1–P8 and the VH-His6-tag fragment (370 bp) with oligonucleotides P9– P16. The two rPCR products were then assembled to the full-length synA21 gene (830 bp) by overlap extension PCR with Xanking primers P0 and P16. The synA21 gene was

then cloned into the XhoI–EcoRI sites of the pPIC9K vector as in frame fusions with the -factor signal sequence under the control of AOX promoter. Ten randomly picked clones were submitted to DNA sequencing analysis. Two of them had correct coding sequences and the others had at least one error, typically with single-point mutants or short deletions. EVect of codon optimization on scFv expression To test the eVect of codon optimization on scFv expression, both the native and the synthetic constructs were screened for multicopy clones by increasing the antibiotic G418 concentrations. The synthetic gene typically yielded 6–10 mg/L scFv in BMMY medium after methanol induction for 3 days (Fig. 2). The mean expression levels of the transformants resistant to 0.5, 1.5, and 3.0 mg/mL G418 were 7.9, 8.5, and 8.4 mg/L, respectively. Statistics analysis revealed that there was no signiWcant diVerence in the scFv levels among three groups of the transformants. This yield was approximately 3- to 5-fold increase in comparison with the 1–2 mg/L scFv level of the native gene. The higher copy number in the more G418-resistant clones was further conWrmed by Southern blotting analysis. Typically, the clones resistant to 1.5 mg/mL G418 acquired 3–5 copies and the clones resistant to 3.0 mg/mL G418 acquired more than Wve copies, basically in accordance with the results proposed in the Pichia pPIC9K Expression Manual (Invitrogen). These data suggested that copy number had no obvious eVect on the scFv expression. To precisely evaluate the steady-state mRNA levels of the native and synthetic constructs, total RNA was extracted after 24 methanol induction and analyzed by Northern blotting using the DNA probe containing the 0.4kb fragment of the 5⬘ untranslated AOX1 mRNA leader sequence and the -factor signal sequence. The use of this probe ensured the cross-hybridization of the mRNAs of both constructs strictly with the same eYciency. Fig. 3 shows that all of the transformants had the full-length transcripts of the expected size (1.4 kb) without obvious degradation or premature polyadenylation. The mRNA levels of the synthetic construct were only 1.2- to 1.4-fold higher than that of the native construct. However, at the protein level, a signiWcant improvement of approximately Wvefold was observed for the synthetic gene (data not shown). In addition, both constructs of two gene copies showed similar mRNA levels to that with one copy. Therefore, codon optimization did not appear to signiWcantly inXuence the gene transcription. EVect of culture pH and medium composition on scFv expression The eVect of culture pH on the cell growth and scFv expression was investigated by a time-course experiment in BMMY medium to reduce protease activities. The results showed a basically positive correlation between the cell

S. Hu et al. / Protein Expression and PuriWcation 47 (2006) 249–257

253

Fig. 1. Design and construction of the synA21 gene. The synthetic gene is based on the primary amino acid sequence of the A21 scFv and Xanked upstream by the -factor signal sequence and downstream by the His6-tag sequence. The DNA sequences were aligned and the changed codons were labeled out. For gene synthesis by rPCR, 17 overlapping oligonucleotides (P0–P16) were designed, which are shown in dotted lines for forward primers and solid lines for backward primers. Restriction enzyme sites introduced for vector construction are underlined.

growth and the scFv accumulation. At all of the tested pH conditions from 5.0 to 8.0, the viable cell densities displayed a rapid increase during the Wrst 72 h induction and almost kept a platform of approximately 30–40 OD600 nm afterward until 144 h induction. The scFv accumulation reached the maximal level during 72–120 h induction and began to decline afterward. It was also observed that the best scFv production was achieved at pH 6.5–7.0 with 10–12 mg/L level, which was 1.2to- 1.3-fold higher than that at pH 6.0, although the cell growth rates were very similar at these pH conditions. Extremely at pH above 8.0, the cell growth retained 70% level but the scFv production reduced to only 40% level. The eVect of casamino acids and EDTA on the scFv expression was shown in Fig. 4. The addition of 2% casamino

acids in BMMY medium increased the scFv yield approximately 1.5-fold. This improvement may partly result from the increased cell growth rate, as the cell density also increased 1.2-fold. However, the addition of 2.5 mM EDTA exhibited no eVect on cell growth, while the scFv production reduced to 80% level. The best clone produced up to 15 mg/L scFv after 96 h methanol induction in BMMY medium at pH 7.0 supplied with 2% casamino acids. PuriWcation of scFv As P. pastoris cells only secreted small amounts of extracellular proteins into the culture supernatant, the secreted his6-tagged scFv was easily puriWed by one-step Ni2+

254

S. Hu et al. / Protein Expression and PuriWcation 47 (2006) 249–257

Fig. 2. Expression of scFv in P. pastoris. Transformants of the native or synthetic constructs resistant to diVerent concentrations of G418 (0.5, 1.5, and 3.0 mg/mL) were submitted to methanol induction for 3 days. The amount of secreted scFv in culture supernatant was quantiWed by quantitative ELISA using mouse anti-histag antibody. The number of clones (n) analyzed is indicated.

Fig. 4. EVects of casamino acids and EDTA on scFv expression. Two clones were used for methanol induction for 3 days in BMMY or BMMY with 2% casamino acids or 2.5 mM EDTA. The scfv level was determined by quantitative ELISA. Data were calculated from three independent experiments.

Fig. 5. SDS–PAGE analysis of scFv puriWcation. The secreted scFv was puriWed from the induced culture supernatant by one-step Ni2+ chelating aYnity chromatography after dialysis. Lane 1, 200 L supernatant; lane 2, 200 L supernatant in the Xow-through fractions; lane 3, protein weight marker. Lanes 4–8, fractions eluted with 0.01 M imidazole. Lanes 9–14, fractions eluted with 0.1 M imidazole. Fig. 3. Northern blotting analysis. Total RNA was extracted from the P. pastoris transformants after 24 h methanol induction. Twenty micrograms of RNA was used for each sample. 32P-labeled DNA probe representing the AOX1 mRNA 5⬘ leader sequence and the -factor signal sequence were used for hybridization. The relative amount of mRNA was quantiWed by densitometric scanning. Lane 1, GS115 negative control; lane 2, native gene, one copy; lane 3, native gene, tow copies; lane 4, synthetic gene, one copy; lane 5, synthetic gene, two copies.

chelating chromatography. It was found that recombinant scFv bound to the resins with high eYciency after the dialyzed supernatant was loaded onto the column, because the scFv band almost disappeared in the Xow-through fractions (Fig. 5, lanes 1 and 2). After the non-speciWcally bound contaminants were washed away with binding buVer containing 0.01 mM imidazole (Fig. 5, lanes 4–8), the scFv was eluted with 0.1 M imidazole (Fig. 5, lanes 9–14). The puriWed scFv showed an homogeneous band with apparent molecular weight of approximately 30 kDa as examined by SDS–PAGE, corresponding to the predicted size of 29 kDa. The amount, purity and recovery yield of scFv for each puriWcation step are summarized in Table 1. A typical puriWcation process allowed for the puriWcation of 14.1 mg scFv with the purity above 95% and the recovery yield of about 90% from 1-L induced supernatant.

Table 1 PuriWcation of recombinant scFv from Pichia pastoris culture supernatant PuriWcation stepa

Total protein (mg)b

scFv (mg)

Purity (%)

Recovery yield (%)

Culture supernatant Dialyzed supernatant Ni2+ column elution UntraWltration and condensation

175.6 85.0 15.1 14.7

15.5 15.1 14.4 14.2

8.8 17.8 95.3 96.6

100 97.4 92.9 91.0

a The puriWcation experiment was carried out from 1-L induced culture supernatant. b The amount of total protein was determined using the Lowry Protein Assay Kit (Sigma).

Binding aYnity of scFv for ErbB2 The binding aYnity of scFv for puriWed ErbB2 antigen was determined and compared with that of the parent A21 mAb by two ELISA methods. In the direct binding assay, the scFv reacted to ErbB2 in a concentration-dependent manner (Fig. 6A). The aYnity constant KaV was calculated to be 0.45 £ 108, 1.16 £ 108, 0.67 £ 108 L/mol, respectively. The average KaV was 0.76 £ 108 L/mol and the dissociation

S. Hu et al. / Protein Expression and PuriWcation 47 (2006) 249–257

255

Fig. 8. Antibody internalization assay. SKBR3 cells were incubated with 125 I-labeled antibodies (4D5 control, A21 mAb, scFv) at 37 °C each for 2 h. The proportion of internalized antibody was calculated as the ratio of cell lysate radioactivity to the sum of radioactivities of cell lysate and acid rinses.

Internalization of scFv by SKBR3 cells

Fig. 6. Binding aYnity of puriWed scFv. (A) Direct binding assay. PuriWed ErbB2 antigen was used in 0.5, 0.25, 0.125, and 0.63 g/mL. The scFv was threefold diluted from 2.5 £ 10¡4 to 15 g/mL. (B) Competitive ELISA. PuriWed ErbB2 was used in 0.1 g/mL concentration. A21 at 10 nM and scFv at twofold serial dilutions from 0 to 1.28 M were competed for binding to ErbB2.

constant KD was 13.1 nM. In the competitive ELISA, the binding of A21 for ErbB2 at 10 nM concentration was halfinhibited by the scFv at 61 nM concentration (Fig. 6B). As the KD of A21 had been determined to be 1.8 nM [7], the KD of scFv was calculated to be 11.0 nM, slightly lower than the result of direct ELISA.

The puriWed scFv was tested for its ability to undergo receptor-mediated internalization after binding to the ErbB2-overexpressing SKBR3 cells. A strong membrane staining was observed in the cells treated with the scFv at 4 °C as visualized by immunoXuorescence microscope, indicating that the scFv could speciWcally bind to the cell-surface ErbB2 receptor (Fig. 7A). In contrast, an obviously intracellular staining was found after the cells were treated with the scFv at 37 °C for 2 h, indicating that the bound scFv could undergo eYcient internalization (Fig. 7B). Similarly, the A21 mAb could also be internalized eYciently into the cytoplasm after binding to the SKBR3 cells (Fig. 7C). As a negative control, an anti-GPIb mAb showed no speciWc immunoXuorescent staining (data not shown). Further 125I-labeled antibody internalization assay showed that the distribution of the scFv between the cell surface and the intracellular compartment was slightly higher than that of A21 mAb after treatment at 37 °C for 2 h, with 46.6 and 43.4% being intracellular, respectively (Fig. 8). As a control, only 21.6% of 4D5 mAb was internalized, in accordance to the results of humanized 4D5—Herceptin [6]. Therefore, the internalizing ability of A21 mAb

Fig. 7. Internalization of scFv in SKBR3 cells. SKBR3 cells were incubated with scFv at 4 °C (A), scFv at 37 °C (B) or A21 at 37 °C (C) each for 2 h. After washed with PBS and then with stripping buVer (37 °C only), the cells were Wxed and permeabilized. The cell-surface bound and internalized antibodies were detected with mouse anti-histag antibody and FITC-conjugated goat anti-mouse IgG by Xuorescence microscope.

256

S. Hu et al. / Protein Expression and PuriWcation 47 (2006) 249–257

and its scFv was approximately twofold higher than that of 4D5. Discussion In this study, we evaluated the utility of codon optimization to improve the expression of an anti-ErbB2 scFv in P. pastoris. We designed the full-length scFv gene by choosing the most- or second-preferred codons, while avoiding the formation of stable secondary structures in the corresponding mRNA sequence. Resultingly, codon optimization moderately increased the scFv expression level 3- to 5-fold and up to 10 mg/L in the standard shaker Xask cultures. Although this increase in magnitude is somewhat lower than those of usually above 5- to 10-fold published for codon optimization studies in P. pastoris [20,24,25], it is still higher than the results of some other reports [26,27]. The translational eYciency hypothesis related to translation initiation and elongation rates has been well accepted for explaining the codon usage bias in the prokaryotes and unicellular eukaryotes [28]. Northern blotting revealed that the mRNA levels were similar between the native and the synthetic scFv constructs, suggesting that the increased expression eVect was most attributable to the enhancement of post-transcriptional processing. As both the native and the synthetic genes were placed after the -factor secretion peptide and not in the vicinity of translation start codon context, we speculated that the increased scFv expression by codon optimization should be mainly due to the enhanced eYciency of translation elongation instead of translation initiation. This is experimentally supported by the Wndings that the amount of rare codons positively correlated with the decreased mRNA stability and translation elongation rate in the yeast S. cerevisiae [29,30]. In addition to codon choice, mRNA structure can also inXuence the translation eYciency in yeasts. The speciWc secondary structures formed or lost near the mRNA untranslated region and the start codon were shown to signiWcantly inXuence the mRNA degradation rate and translation initiation eYciency [31]. However, less data are available regarding the presence or the absence of secondary structures in the coding region aVecting the mRNA stability and translation. It was reported that optimization of the G + C content alone, despite the actual codon bias, could signiWcantly improve the recombinant protein expression in P. pastoris as well as in mammalian cells [20,32]. Actually, as P. pastoris prefers A/T-ended codons while mammals tend toward G/C-ended codons, codon optimization of a mammalian-derived gene into the P. pastoris preferred one usually leads to the relatively low G + C content accompanied with the mRNA of less stable secondary structures. Regardless of the mechanisms to be further studied, our data suggest that codon optimization and mRNA structure reduction can have positive eVects on heterologous protein expression in P. pastoris. Our attempt to improve the scFv expression by increasing the gene copies of the synthetic gene was found to be

less eVective. Woo et al. also reported that there was no diVerence in the expression level of a scFv-fused immunotoxin between the clones of single and double copies [25]. It seems that other rate-limiting factors including protein folding within the endoplasmic reticulum and secretion signal processing may also determine the scFv secretion ability. Recently, the unfolded protein response concerning a series of proteins was identiWed to prevent the secretion enhancement of heterologous proteins by increasing gene copies in the yeasts [22,33]. Consistently, several groups have reported that co-expression of chaperone proteins such as BiP, PDI and SEC4 to assist folding and secreting remarkably improved the expression of recombinant proteins including scFv fragments [34–36]. Further studies are needed to investigate whether this co-expressing strategy can help to express the A21 scFv in P. pastoris. As cell density and protease activity may have impactful inXuence on the yield of secreted proteins in P. pastoris, we investigated the eVect of culture pH and medium composition on scFv production. It was found that scFv accumulation positively correlated with the cell growth in BMMY medium and achieved the highest level at pH 6.5–7.0. In contrast, Shi et al. [14] did not observe a correlation between the cell growth, protease activity and scFv accumulation, as the anti-serpin scFv was sensitive to extracellular proteolysis. Although P. pastoris can secrete several types of extracellular proteases into the culture medium, it seems that the inclusion of yeast extract and peptone provides eYcient prevention of our scFv from protease degradations. This is supported by our observations that the addition of casamino acids only limitedly increased the scFv production and even the addition of EDTA led to a negative eVect. These data thus provide useful parameters for optimizing fermentation conditions to achieve higher scFv production. The ability of anti-ErbB2 antibodies to inhibit tumor cell growth diVers much with their speciWcally recognized epitopes and associate commonly with their abilities to induce ErbB2 receptor endocytosis [37,38]. Although the aYnity of the A21 scFv expressed in P. pastoris was sixfold lower than that of A21 mAb and approximately 100-fold lower than that of 4D5, the internalizing ability of the scFv was slightly higher than A21 mAb and twofold better than 4D5. Our data conWrmed the Wndings that the internalizing eYciency of anti-ErbB2 antibodies did not necessarily depend on either aYnity or bivalency [6]. The A21 scFv should be especially useful for construction of fusion molecules for delivery of drugs, toxins or DNA into the cytoplasm for ErbB2-based immunotherapy. Acknowledgments This work is supported by Hi-Tech Research and Development Program (“863” Program) of the Ministry of Science and Technology of China (No. 2001AA215381) and Specialized Research Fund for the Doctoral Program of Higher Education (No. 20020358048).

S. Hu et al. / Protein Expression and PuriWcation 47 (2006) 249–257

References [1] R.M. Neve, H.A. Lane, N.E. Hynes, The role of over-expressed HER2 in transformation, Ann. Oncol. 12 (Suppl. 1) (2002) 9–13. [2] J. Baselga, J. Albanell, Mechanism of action of anti-HER2 monoclonal antibodies, Ann. Oncol. 2 (Suppl. 1) (2001) 35–41. [3] L.Y. Yum, L.W. Robyn, Anti-ErbB2 monoclonal antibodies and ErbB2-directed vaccines, Cancer Immunol. Immunother. 50 (2002) 569–587. [4] G.P. Adams, R. Schier, Generating improved single-chain Fv molecules for tumor targeting, J. Immunol. Methods 231 (1999) 249–260. [5] M.A. Poul, B. Becerril, U.B. Nielsen, P. Morisson, J.D. Marks, Selection of tumor-speciWc internalizing human antibodies from phage libraries, J. Mol. Biol. 301 (2000) 1149–1161. [6] R.M. Neve, U.B. Nielsen, D.B. Kirpotin, M.A. Poul, J.D. Marks, C.C. Benz, Biological eVects of anti-ErbB2 single chain antibodies selected for internalizing function, Biochem. Biophys. Res. Commun. 280 (2001) 274–279. [7] P. Li, Y. Li, C. Wang, J. Liu, Investigation on anti-cancer activities of anti-P185 monoclonal antibodies in vitro, Chin. J. Immunol. 18 (2002) 33–35. [8] L.S. Chen, A.P. Liu, J. Liu, Construction, expression and characterization of the engineered antibody against tumor surface antigen P185cerbb-2 , Cell Res. 13 (2003) 35–48. [9] J. Wang, Y. Shi, Y. Liu, S. Hu, J. Ma, J. Liu, L. Chen, PuriWcation and characterization of a single-chain chimeric anti-p185 antibody expressed by CHO-GS system, Protein Expr. Purif. 41 (2005) 68–76. [10] R. Fischer, J. Drossard, N. Emans, U. Commandeur, S. Hellwig, Towards molecular farming in the future: Pichia pastoris-based production of single-chain antibody fragments, Biotechnol. Appl. Biochem. 30 (1999) 117–120. [11] A. Goel, G.W. Beresford, D. Colcher, G. Pavlinkova, B.J. Booth, K.J. Baranowska, S.K. Batra, Divalent forms of CC49 single-chain antibody constructs in Pichia pastoris: expression, puriWcation, and characterization, J. Biochem. 127 (2000) 829–836. [12] C. Marty, P. Scheidegger, H.K. Ballmer, Production of functionalized single-chain Fv antibody fragments binding to the ED-B domain of the B-isoform of Wbronectin in Pichia pastoris, Protein Expr. Purif. 21 (2001) 156–164. [13] Y. Wang, K. Wang, D.C. Jette, D.S. Wishart, Production of an antiprostate-speciWc antigen single-chain antibody fragment from Pichia pastoris, Protein Expr. Purif. 23 (2001) 419–425. [14] X. Shi, T. Karkut, M. Chamankahh, M.M. Alting, S.M. Hemmingsen, D. Hegedus, Optimal conditions for the expression of a single-chain antibody (scFv) gene in Pichia pastoris, Protein Expr. Purif. 28 (2003) 321–330. [15] C. Gurkan, S.N. Symeonides, D.J. Ellar, High-level production in Pichia pastoris of an anti-p185HER-2 single-chain antibody fragment using an alternative secretion expression vector, Biotechnol. Appl. Biochem. 39 (2004) 115–122. [16] K. Sreekrishna, R.G. Brankamp, K.E. Kropp, D.T. Blankenship, J.T. Tsay, P.L. Smith, J.D. Wierschke, A. Subramaniam, L.A. Birkenberger, Strategies for optimal synthesis and secretion of heterologous proteins in the methylotrophic yeast Pichia pastoris, Gene 190 (1997) 55–62. [17] J.M. Cregg, J.L. Cereghino, J. Shi, D.R. Higgins, Recombinant protein expression in Pichia pastoris, Mol. Biotechnol. 16 (2000) 23–52. [18] K. Sreekrishna, Strategies for optimizing protein expression and secretion in the methylotrophic yeast Pichia pastoris, in: R.H. Baltz, G.D. Hegeman, P.L. Skatrud (Eds.), Industrial Microorganism: Basic and Applied Molecular Genetics, American Society of Microbiology, Washington, DC, 1993, pp. 119–126. [19] X. Zhao, K.K. Huo, Y.Y. Li, Synonymous codon usage in Pichia pastoris, Chin. J. Biotechnol. 16 (2000) 308–311.

257

[20] G. Sinclair, F.Y.M. Choy, Synonymous codon usage bias and the expression of human glucocerebrosidase in the methylotrophic yeast Pichia pastoris, Protein Expr. Purif. 26 (2002) 96–105. [21] C. Prodromou, L.H. Pearl, Recursive PCR: a novel technique for total gene synthesis, Protein Eng. 8 (1992) 827–829. [22] H. Hohenblum, B. Gasser, M. Maurer, N. Borth, Diethard Mattanovich eVects of gene dosage, promoters, and substrates on unfolded protein stress of recombinant Pichia pastoris, Biotechnol. Bioeng. 85 (2004) 367–375. [23] J.D. Beatty, B.G. Beatty, W.G. Vlahos, Measurement of monoclonal antibody aYnity by non-competitive enzyme immunoassay, J. Immunol. Methods 26 (1987) 173–179. [24] S. Brocca, C. Schmidt-Dannert, M. Lotti, L. Alberghinaand, R.D. Schmid, Design, total synthesis, functional overexpression of the Candida rugosa lip1 gene coding for a major industrial lipase, Protein Sci. 7 (1998) 1415–1422. [25] J.H. Woo, Y.Y. Liu, A. Mathias, S. Stavrou, Z. Wang, J. Thompson, D.M. Neville, Gene optimization is necessary to express a bivalent anti-human anti-T cell immunotoxin in Pichia pastoris, Protein Expr. Purif. 25 (2002) 270–282. [26] A.M. WolV, O.C. Hansen, U. Poulsen, S. Madrid, P. Stougaard, Optimization of the production of chondrus crispus hexose oxidase in P. pastoris, Protein Expr. Purif. 22 (2001) 189–199. [27] N.S. Outchkourov, W.J. Stiekema, M.A. Jongsma, Optimization of the expression of equistatin in Pichia pastoris, Protein Expr. Purif. 24 (2002) 18–24. [28] X. Xia, How optimized is the translational machinery in Escherichia coli, Salmonella typhimurium and Saccharomyces cerevisiae? Genetics 149 (1998) 37–44. [29] A. Hoekema, R.A. Kastelein, M. Vasser, H.A. Boer, Codon replacement in the PGK1 gene of Saccharomyces cerevisiae: experimental approach to study the role of biased codon usage in gene expression, Mol. Cell. Biol. 7 (1987) 2914–2924. [30] D. Herrick, R. Parker, A. Jacobson, IdentiWcation and comparison of stable and unstable mRNAs in Saccharomyces cerevisiae, Mol. Cell. Biol. 10 (1990) 2269–2284. [31] C.C. Oliveira, J.J. van den Heuvel, J.E.G. McCarthy, Inhibition of translational initiation in Saccharomyces cerevisiae by secondary structure: the roles of the stability and position of stem-loops in the mRNA leader, Mol. Microbiol. 9 (1993) 521–532. [32] C.H. Kim, O. Younghoon, T.H. Lee, Codon optimization for highlevel expression of human erythropoietin (EPO) in mammalian cells, Gene 199 (1997) 293–301. [33] K.J. KauVman, E.M. Pridgen, F.J. Doyle, P.S. Dhurjati, A.S. Robinson, Decreased protein expression and intermittent recoveries in BiP levels result from cellular stress during heterologous protein expression in Saccharomyces cerevisiae, Biotechnol. Prog. 18 (2002) 942–950. [34] E.V. Shusta, R.T. Raines, A. Pluckthun, K.D. Wittrup, Increasing the secretory capacity of Saccharomyces cerevisiae for production of single-chain antibody fragments, Nat. Biotechnol. 16 (1998) 773–778. [35] S.H. Liu, W.I. Choua, C.C. Sheub, M.D. Changa, Improved secretory production of glucoamylase in Pichia pastoris by combination of genetic manipulations, Biochem. Biophys. Res. Commun. 326 (2005) 817–824. [36] R. Vad, E. Nafstad, L.A. Dahl, O.S. Gabrielsen, Engineering of a Pichia pastoris expression system for secretion of high amounts of intact human parathyroid hormone, J. Biotechnol. 30 (2005) 251–260. [37] E. Hurwitz, I. Stancovski, M. Sela, Y. Yarden, Suppression and promotion of tumor growth by monoclonal antibodies to ErbB-2 diVerentially correlate with cellular uptake, Proc. Natl. Acad. Sci. USA 92 (1995) 3353–3357. [38] L.N. Klapper, N. Vaisman, E. Hurwitz, P.K. Ronit, Y. Yarden, M. Sel, L.N. Clapper, A subclass of tumor-inhibitory monoclonal antibodies to ErbB-2/HER2 blocks crosstalk with growth factor receptors, Oncogene 14 (1997) 2099–2109.