Protein Expression and Purification 102 (2014) 27–37

Contents lists available at ScienceDirect

Protein Expression and Purification

journal homepage: www.elsevier .com/ locate /yprep

Improved soluble expression of a single-chain antibody fragmentin E. coli for targeting CA125 in epithelial ovarian cancer

http://dx.doi.org/10.1016/j.pep.2014.07.0071046-5928/� 2014 Elsevier Inc. All rights reserved.

⇑ Corresponding author at: Department of Oncology, University of Alberta,11560-University Avenue, Edmonton, AB T6G 1Z2, Canada. Tel.: +1 780 989 8150;fax: +1 780 432 8483.

E-mail address: wuest@ualberta.ca (F.R. Wuest).1 Abbreviations used: scFv, single-chain Fragment variable; MAb, monoclonal

antibody; CA125, Cancer Antigen 125; EOC, epithelial ovarian cancer; IPTG, isopro-pyl-b-D-thiogalactopyranoside; PCR, polymerase chain reaction; WT, wild-type;IMAC, immobilized metal affinity chromatography; S, serine; G, glycine; L, leucine;K, lysine; R, arginine; CAI, codon adaptation index; I, isoleucine; P, proline.

Sai Kiran Sharma a,b, Mavanur R. Suresh a, Frank R. Wuest a,b,⇑a Faculty of Pharmacy and Pharmaceutical Sciences, University of Alberta, Edmonton, Canadab Department of Oncology, University of Alberta, Edmonton, Canada

a r t i c l e i n f o a b s t r a c t

Article history:Received 6 June 2014and in revised form 17 July 2014Available online 29 July 2014

Keywords:Single-chain Fragment variable (scFv)Codon optimizationSoluble expressionInter-domain peptide linkerCA125Epithelial ovarian cancer

Production of antibody fragments in heterologous hosts such as Escherichia coli provides a unique andcost-effective method to develop engineered vectors for tumor targeting. A single-chain Fragment vari-able (scFv) of the murine monoclonal antibody MAb-B43.13 targeting CA125 in epithelial ovarian cancerwas previously developed, expressed, purified and proposed as a functional targeting entity for biomed-ical applications. However, the yields from its soluble expression in heterologous systems were very lowfor any practical use in preclinical translational research; leave alone the defeated objective of convenientand cost-effective production. In the present work, the anti-CA125 scFv gene was re-organized and sub-cloned into pET-22b(+) vector to be in frame with the pelB leader peptide for periplasmic localization andC-terminal hexa-histidine tag to facilitate downstream purification. Six variants of the scFv were con-structed to investigate the impact of variable domain orientations, inter-domain peptide linker sequencesand codon optimization on the soluble expression of the scFv using Rosetta 2(DE3) as the E. coli host sup-plemented with tRNAs for rare codons. Expression in shake flask cultures under the control of an induc-ible T7 promoter and subsequent purification by cobalt based immobilized metal affinitychromatography yielded differential amounts of high purity scFv for all constructs. Here, we report upto 14-fold increase in the soluble expression of the scFv primarily as a result of codon optimization withminor effects from inter-domain peptide linkers and variable domain orientation in the anti-CA125 scFvmolecule. All the scFv constructs expressed and purified were found to be immunoreactive for in vitro tar-geting of CA125 antigen.

� 2014 Elsevier Inc. All rights reserved.

Introduction to as the smallest yet complete active component of the immuno-

Recombinant technology has expanded the scope for engineer-ing antibodies which have evolved from being sentinels of theimmune system to becoming key components in targeted diagnos-tic and therapeutic applications. One of the most popular format ofan engineered antibody is the single-chain Fragment variable(scFv1). Comprised of the heavy (VH) and light (VL) chain of animmunoglobulin’s variable domains that are connected by a flexiblepeptide linker, this class of recombinant molecules are often referred

globulin capable of binding to a target antigen [1,2].Some advantages of the scFv format over full-length antibodies

are (a) minimal immunogenicity due to the lack of Fc regions; (b)faster in vivo clearance and better tissue penetration owing to a rel-atively smaller molecular size; (c) cost effectiveness due to ease ofproduction in simple host systems such as Escherichia coli and (d)their amenability to recombinant engineering [3]. Furthermore,the possibilities to screen immune libraries for isolation of highaffinity scFv binders to practically any antigen by phage display[4], ribosome display [5] and yeast display [6] technologies haveexpanded scientific capabilities through ease of discovery, synthe-sis and production of engineered antibodies. Nevertheless, havingaccess to the parent hybridoma cell for a monoclonal antibody(MAb) can facilitate the isolation of its derivative scFv withretained immunoreactivity and specificity [7,8]. This gains transla-tional relevance when the scFv is sought from a clinically approvedMAb [9]. However, there are challenges with the expression ofmurine antibody sequences in heterologous prokaryotic hosts suchas E. coli in order to yield soluble and functional scFv.

28 S.K. Sharma et al. / Protein Expression and Purification 102 (2014) 27–37

Cancer Antigen 125 (CA125) is a mucinous glycoprotein thatserves as a USFDA approved tumor biomarker for the diagnosisof epithelial ovarian cancer (EOC) [10,11]. CA125 can be detectedin immunoassay techniques as an antigen shed in the serum ofpatients presenting in the clinic with pelvic masses [12]. Theimmunoassay formats and targeting applications have been greatlyfacilitated by the production of several antibodies against CA125[13–16]. Among those reported, MAb B43.13 has shown greattranslational benefit in the immunotherapy of EOC [17]. Interest-ingly, there are not as many scFv molecules developed against thistumor-associated antigen [18–22]. Prior literature indicates thatthe anti-CA125 B43.13 scFv was purified as a secreted protein fromyeast [18], as a secreted bi-specific scFv from murine myeloma NS0cells [20], and engineered to be produced as a bi-functional fusionconstruct in E. coli, albeit with low yields obtained from solubleexpression [21]. Considering its previously demonstrated potentialfor biopharmaceutical application to target CA125 [20,21,23], theproduction of this scFv in multi-milligram amounts could be astarting point for further engineering and expansion of its utilityas an EOC targeting vector for immunodiagnostics, drug deliveryand potential immunotherapy [24].

In the present work, we investigated the effects of variabledomain orientation, different inter-domain linkers and codon opti-mization to increase the yields of soluble protein from heterolo-gous expression of anti-CA125 B43.13 scFv in E. coli. The scFvwas isolated from soluble fractions of recombinant cell lysateswith high purity using a single step purification via immobilizedmetal affinity chromatography and was found to be biochemicallyactive for binding to target antigen CA125. Codon optimization wasfound to be the most important factor to positively impact solubleexpression of anti-CA125 scFv resulting in up to 14-fold higheryields than all the other murine scFv variants examined in thisstudy.

Materials

All chemicals including antibiotics used were purchased fromSigma Aldrich unless otherwise specified. Phusion high-fidelityDNA polymerase (Finnzymes, F-530S), and oligonucleotide primers(Integrated DNA Technologies) were used for PCR amplification ofthe scFv domains. Nco I (NEB, R0193S), Not I (NEB, R0189S) and T4DNA Ligase (NEB, M0202S) were used for directional cloning ofgenes encoding the scFv domains. pET-22b(+) vector (Novagen,69744) and E. coli Rosetta™ 2(DE3) (Novagen, 71400) were usedfor expression of the scFv. Bacto tryptone (BD, 212750), Yeastextract (BD, 211705) and sodium chloride (Fisher Scientific,7647-14-5) were used to prepare the 2� YT medium. Isopropyl-b-D-thiogalactopyranoside (IPTG) (Fisher Scientific, BP1755-10)was used for induction of recombinant scFv expression. BugBuster

Fig. 1. Schematic representation of the anti-CA125 MAb and its derivative scFv cloned in

Master Mix (Novagen, 71456) was used for cell lysis of harvestedE. coli cultures. TALON� Superflow resin (Clontech, 635506) wasused for immobilized metal affinity chromatography. Glycine(Bio-Rad, 161-0718) and Coomassie brilliant blue R-250 (Bio-Rad,161-0400), Pre-stained SDS–PAGE standards – Low Range (Bio-Rad, 161-0305) were used for protein analysis by gel electrophore-sis. Amicon Ultra-15, 10K MWCO filters (EMD Millipore,UFC901024) were used to concentrate the purified proteins. Fetalbovine serum (Life Technologies, 12483-020), penicillin–strepto-mycin (10,000 U/ml) (Life Technologies, 15140-122) and recombi-nant human insulin (SAFC Biosciences, 91077C) were used forculturing ovarian cancer cell lines. CelLytic™ M (Sigma, C2978)Trans-Blot nitrocellulose membrane (Bio-Rad, 162-0115), Amer-sham Hyperfilm ECL (GE Healthcare, 28906839), Amersham ECLPlus Western Blotting Detection Reagents (GE Healthcare,RPN2132), mouse anti-b actin IgG (Sigma, A1978), goat anti-mouseHRP conjugate (Sigma, A4416), 6X His MAb-HRP conjugate (Clon-tech, 631210) were used for immunoblotting. Alexa-fluor� 488goat anti-mouse antibody (Life Technologies, A-11001) and Pen-ta�His Alexa Fluor 488 conjugate (Qiagen, 35310) were used inimmunofluorescence and flow cytometry studies. Anti-CA125MAb was isolated from hybridoma B43.13 and anti-RANK scFvwas generously provided by Dr. Michael Doschak, University ofAlberta, Canada.

Methods

Cloning of anti-CA125 scFv variants

The genes encoding the variable domains of the anti-CA125scFv were amplified by polymerase chain reaction (PCR) using apreviously reported plasmid construct pWET8 [20] as the template.Engineering of inter-domain linkers and production of differentorientations of the single chain domains was performed bysplice-overlap extension PCRs to create the following constructsviz. VH-(G2S)5-VL; VL-(G2S)5-VH; VH-(G4S)3-VL; VL-(G4S)3-VH; VL-(218)-VH, (Fig. 1) using Nco I and Not I as unique 50 and 30 cloningsites respectively for ligation into pET-22b(+) expression vector.

Based on demonstrated outcomes suggested in prior reports forother scFv molecules [25–27], the VL-(218)-VH scFv gene wascodon-optimized for expression in E. coli and synthesized by Gene-Art� (Life Technologies). The codon-optimized scFv gene was sub-cloned into pET-22b(+) vector between Nco I and Not I sites. Thisconstruct is hereafter designated as GA218 in this article, whereasthe murine anti-CA125 scFv coding sequences will be referred to asthe wild-type (WT) sequence. The scFv DNA cloned in each recom-binant construct was verified by Sanger sequencing on a 3730 DNAanalyzer (Applied Biosystems, Life Technologies) prior to transfor-mation in the bacterial expression host.

to pET-22b(+) vector in different orientations for heterologous expression in E. coli.

S.K. Sharma et al. / Protein Expression and Purification 102 (2014) 27–37 29

Small scale expression analysis

Each recombinant plasmid coding for different scFv constructswas transformed into chemically competent E. coli Rosetta2(DE3) cells by heat-shock method and plated onto 1.5% 2� YTagar plates containing 100 lg/mL ampicillin and 34 lg/mL chlor-amphenicol. Single transformant colonies were picked and inocu-lated into 10 mL 2� YT medium (16 g/L tryptone, 10 g/L yeastextract and 5 g/L sodium chloride, pH 7.5) in the presence of anti-biotics and cultured overnight by incubation at 37 �C with shakingat 200 rpm. 0.2 mL of the overnight grown culture was added to20 mL of fresh 2� YT medium with the aforementioned concentra-tion of antibiotics in a 125 mL shake flask and allowed to grow at37 �C with shaking at 220 rpm until an OD600 of 0.6–0.8 wasachieved. The shake flask cultures were briefly placed in ice coldwater for 15 min prior to induction with a final concentration of0.8 mM IPTG. The induced cultures were incubated at 26 �C withshaking at 200 rpm overnight for 16 h. Spectrophotometric mea-surement of overnight grown cultures was performed to enablenormalized loading (OD600 = 1) for a comparative evaluation ofscFv expression from the different constructs. Total cell proteinsamples were prepared by adding Laemmli sample buffer(50 mM Tris–HCl pH 6.8, 2% SDS, 10% glycerol, 5 mM 2-mercap-toethanol, 0.1% bromophenol blue) to the normalized cell pelletsand heating them at 95 �C for 5 min prior to analysis on a 12%SDS–PAGE, followed by staining with Coomassie brilliant blue R-250 and immunoblotting with 6X His MAb-HRP conjugate.

Small scale batch binding

To detect soluble scFv expressed and extractable at the smallscale, pertinent volumes of the overnight grown bacterial cultures(volumes worth a total OD600 = 20) were processed further by celllysis using BugBuster� master mix reagent to perform batch bind-ing analyses with TALON� superflow metal affinity resin. 10 mMimidazole in 50 mM sodium di-hydrogen phosphate [NaH2PO4�H2O] and 100 mM sodium chloride pH 7.0 was used as wash bufferand 150 mM imidazole in 50 mM sodium di-hydrogen phosphateand 100 mM sodium chloride pH 7.0 was used as the elution buf-fer. Supernatants harvested from cell lysates were allowed to incu-bate with 100 lL of TALON resin at 4 �C for 2 h. Unbound proteinswere isolated by centrifugation of the mixture of lysate superna-tant and resin at 500g for 5 min. Two rinses were performed with1 mL of wash buffer by adding it to the resin and allowing this tomix well for 10 min prior to centrifugation at 500 g for 5 min.The resulting supernatants were collected and designated as W1

and W2 samples. Finally, 2 � 10 min rinses were performed withthe elution buffer prior to collection of eluate fractions uponcentrifugation at 500g for 5 min. The resulting fractions were des-ignated as E1 and E2 samples. All fractions were analyzed by SDS–PAGE under reducing conditions followed by Coomassie stainingand immunoblotting for hexa-histidine tagged scFv.

Medium scale expression and purification of anti-CA125 scFv variants

To compare recombinant scFv yields from 1L cultures in shakeflasks, single transformant colonies were picked for each constructand inoculated into 12.5 mL 2� YT medium with 100 lg/mL ampi-cillin, 34 lg/mL chloramphenicol and cultured overnight by incu-bation at 37 �C with shaking at 200 rpm. 5 mL of the overnightgrown cultures was added to 500 mL fresh 2� YT medium in thepresence of antibiotics in 2 L flasks and incubated at 37 �C withshaking at 220 rpm until an OD600 of 0.6–0.8 was achieved. Theshake flask cultures were briefly placed in ice cold water for15 min prior to induction with a final concentration of 0.8 mMIPTG. The induced cultures were incubated at 26 �C with shaking

at 200 rpm overnight for 16 h. The bacteria were harvested by cen-trifugation at 7000 rpm for 30 min and the pellet obtained thereofwas frozen by storage at �20 �C. Soluble scFv variants wereextracted from frozen cell pellets by lysis with BugBuster� mastermix reagent as per the manufacturer’s instructions. The cell lysatewas cleared by centrifugation at 16,000 rpm for 30 min and thesupernatant was collected for purification of soluble scFv. C-termi-nal hexa-histidine tagged recombinant scFv was purified viaimmobilized metal affinity chromatography (IMAC) by passingthe supernatant over TALON� superflow metal affinity resinpacked in a column operated under gravity using a MasterflexHV-07520-00 peristaltic pump (Cole Parmer) to provide a flow rateof 1 mL/min. Non-specifically bound protein was washed using 15column volumes of 10 mM imidazole in equilibration buffer(50 mM NaH2PO4�H2O and 100 mM NaCl pH 7.0). Soluble hexa-his-tidine tagged scFv bound to the affinity column was eluted byflowing a linear gradient of 10–150 mM imidazole in equilibrationbuffer prepared using a gradient maker GM-200 (CBS Scientific).1 mL IMAC eluted fractions were collected and analyzed by 12%SDS–PAGE under reducing conditions, followed by Coomassiestaining and immunoblotting with 6X His MAb-HRP conjugate.High purity eluate fractions were pooled together for dialysis intophosphate buffered saline (pH 7.4) and concentrated to a final vol-ume of 1 mL using Amicon Ultra-15, 10K MWCO filters. Final con-centrations of the purified scFv were quantified using a Pierce™BCA protein assay kit (Thermo Scientific, 23227) according to themanufacturer’s recommendations. Bovine serum albumin suppliedwith the kit was used to prepare a standard curve for proteinestimation.

Molecular and functional characterization of anti-CA125 scFvconstructs

Protein IDAn in-gel tryptic digest of the purified scFv samples followed by

LC–MS/MS analyses was performed to verify protein identity at theInstitute of Biomolecular Design, University of Alberta, Canada(Supplementary Data).

N-terminal protein sequencing12 lg each of scFv VL-(G4S)3-VH and GA218 were electrophore-

sed on a 4–15% Mini-PROTEAN� TGX™ precast gel (Bio-Rad) andelectroblotted onto Hybond-P 0.45 lm PVDF membrane (GEHealthcare). The blot was stained with Coomassie brilliant blueR-250, destained with methanol and washed thoroughly withmilli-Q water prior to excision of the scFv bands. N-terminal pro-tein sequencing by Edman degradation was performed at the pro-tein facility of the Iowa State University Office of Biotechnology,USA.

Cell lines and culture conditionsOvarian cancer cells NIH:OVCAR-3 (ATCC� HTB-161™) that

overexpress CA125 and SKOV3 (ATCC� HTB-77™) that do notexpress CA125 were used for in vitro functional characterizationstudies. Cells were cultured in DMEM-F12 medium supplementedwith 10% v/v fetal bovine serum, 50 IU/mL penicillin, 50 lg/mLstreptomycin. NIH:OVCAR-3 cells were additionally supplementedwith 7 lg/mL recombinant human insulin. Cells were culturedusing sterile techniques and grown in a 37 �C incubator providinghumidified atmosphere of 5% CO2 in air.

Immunoblotting7.5 � 106 NIH:OVCAR-3 and SKOV3 cells were lysed with Cel-

Lytic™ M. Cell lysates were electrophoresed on a 4–15% Mini-PRO-TEAN� TGX™ precast gel (Bio-Rad) and transferred to a Trans-Blotnitrocellulose membrane (Bio-Rad). The membranes were probed

30 S.K. Sharma et al. / Protein Expression and Purification 102 (2014) 27–37

separately to evaluate binding capabilities of the different anti-CA125 scFv constructs to target antigen in cell lysates. The blotswere blocked for 45 min with 5% non-fat dry milk (Carnation) inPBS having 0.1% Tween-20 (PBST). Anti-CA125 MAb (3 mg/mL),mouse anti-b actin IgG and anti-CA125 scFv variants (�2 mg/mL)were used as primary antibodies (1:5000 each) to probe the blotsfor 1 h at room temperature. Goat anti-mouse HRP conjugate(1:5000) was used as secondary antibody to probe the blot againstanti-CA125 MAb and mouse anti-b actin IgG for 1 h at room tem-perature. 6X His MAb-HRP conjugate (1:5000) was used as second-ary antibody to probe against antigen-bound anti-CA125 scFvvariants by incubating for 1 h at room temperature. Anti-CA125targeting monoclonal antibody MAb B43.13 [28] was used as apositive control. A hexa-histidine tagged anti-RANK (Receptor Acti-vator of Nuclear factor Kappa B) receptor binding scFv [29] wasused as an isotype control.

Flow cytometry1.5 � 106 NIH:OVCAR-3 cells were harvested by trypsinization,

rinsed twice with FACS buffer (PBS with 0.5% heat inactivated FBS,2 mM EDTA, 0.05% sodium azide) and resuspended by gentle tappingin �100 lL FACS buffer. 10 lg of anti-CA125 MAb or scFv variantswere incubated with the cell suspension for 30 min at room temper-ature. Cells were rinsed twice in FACS buffer and incubated for30 min with 1.6 lg of Alexa-fluor� 488 goat anti-mouse antibodyfor the MAb sample and 2.4 lg of Penta�His Alexa Fluor 488 conju-gate for scFv samples. Cells were rinsed twice with FACS buffer andanalyzed by flow cytometry on a BD FACS Calibur. A hexa-histidinetagged anti-RANK receptor scFv was used as an isotype control.Negative controls included unstained NIH:OVCAR-3 cells and cellsincubated with Alexa fluor 488 conjugated antibodies alone.

ImmunofluorescenceNIH:OVCAR-3 cells were plated onto glass coverslips in 35-mm

tissue culture dishes (100,000 cells/2 mL medium/dish) and incu-bated at 37 �C for 48 h. The cells were rinsed with PBS and fixedin methanol for 30 min at �20 �C. The fixed cells were incubatedin 5% non-fat dry milk (Carnation) in PBS and immunostained sep-arately for 1 h with anti-CA125 MAb and the various purified scFvconstructs. Corresponding Alexa fluor 488 labeled secondary anti-bodies were used as in the flow cytometry studies. Anti-CA125MAb and scFv variants (2 mg/mL) were used at 1:250 dilution fol-lowed by 1:500 dilutions of secondary antibodies (2 mg/mL) in PBScontaining 5% non-fat dry milk. Appropriate blank and controlsamples were included in the experiments. All antibody incuba-tions were followed by three rinses with PBST for 10 min each.Coverslips were mounted on microscopy slides (Fisherbrand, 12-544-2) using Mowiol� mounting medium (Calbiochem, 475904)supplemented with DAPI (50 lg/mL). Immunofluorescence wasobserved through a Zeiss Plan Apochromat 40X/1.3 Oil DIC M27lens on a confocal laser scanning microscope (Zeiss LSM 710).The images were analyzed using Zen 2011 software and processedfurther using Adobe Photoshop CS6.

Results

scFv cloning and codon optimization

Sanger sequencing of the recombinant plasmid constructsrevealed high fidelity DNA amplification and successful directionalcloning of all the scFv variants between Nco I and Not I restrictionsites in pET-22b(+) expression vector (Supplementary Data). Anal-ysis of the nucleotide sequence for codons in the WT anti-CA125scFv indicated that �58.3%: 136 out of 233 codons (excludingthe linker sequences) had to be modified for optimal expression

in E. coli. This was further exemplified by the fact that more than90% (32/35) of the original murine codons for serine (S) – the mostabundant residue (15%) in the anti-CA125 scFv variable domainsequence had to be optimized for E. coli expression (Fig. 2). Simi-larly, among other abundant residues, 71.4% (15/21) codons forglycine (G), 73% (11/15) codons for leucine (L), 66.6% (10/15)codons for lysine (K) and 100% (6/6) codons for arginine (R) hadto be modified (Fig. 2) to this end. The codon adaptationindex (CAI) for heterologous expression of WT anti-CA125 scFvnucleotide sequence in E. coli calculated using an online rarecodon analysis tool (http://www.genscript.com/cgi-bin/tools/rare_codon_analysis) improved from 0.64 to 0.93 after codon optimiza-tion. A CAI value between 0.8 and 1.0 is considered ideal forrecombinant expression in E. coli.

scFv expression analysis and purification yields

Small scale analysis of total cell proteins from harvestedcultures revealed successful recombinant expression of all scFvvariants constructed in this study (Fig. 3A and B).

Most of the heterologously expressed scFvs were found in cellpellets as seen by Coomassie staining and immunoblotting(Fig. 4A and B). However, a distinct band of soluble scFv was seenin the cell lysate supernatant of the GA218 construct (Fig. 4B).

Batch binding analysis revealed maximum soluble expressionand elution of scFv GA218 variant followed by scFv VL-(G4S)3-VH

(Fig. 5B and Table 1).The GA218 construct yielded�3.8-fold higher amounts (mg/g) of

soluble anti-CA125 scFv when compared to the next highestexpressing construct VH-(G4S)3-VL. In comparison with the leastexpressing construct VL-(218)-VH, there was �14-fold increase inmg/g yields of soluble anti-CA125 scFv. The difference in purifiedscFv yields between the listed constructs is represented in Fig 6,wherein equal volumes of the soluble scFv variants purified from1 L cultures and concentrated to a final 1 mL volume were electro-phoresed and analyzed by Coomassie staining and immunoblotting.

Characterization of scFv constructs

LC–MS/MS analysis of peptides from the tryptic digests of puri-fied scFv yielded high confidence score identities with mouseimmunoglobulin variable light and heavy chain proteins upon per-forming a BLAST against the NCBInr database. Furthermore, thefive peptides identified through this procedure had a 100% identitymatch with regions of the scFv coding sequence cloned into theexpression vector (Supplementary Data). N-terminal proteinsequencing of the purified scFvs revealed absence of the pelB lea-der peptide upstream of the N-terminal scFv variable domain, thusindicating that the recombinant protein was processed in the bac-terial periplasm and purified thereof (Appendices 1 and 2). Func-tional characterization of the scFv variants by immunoblotting(Fig. 7A) revealed their highly specific binding to lysates ofNIH:OVCAR-3 cells that overexpress CA125 and no binding toSKOV3 cell lysates. This experiment also demonstrated the migra-tion of CA125 in a 4–15% gradient SDS–PAGE in good agreementwith previously reported characterization studies for this antigenby Davis et al. [10]. Furthermore, the use of MAb-B43.13 as a posi-tive control and the anti-RANK scFv as a negative control strength-ens the demonstration for specificity of target binding by thevarious scFv constructs developed in this study. Results from flowcytometry (Fig. 7B) analyses indicated a shift in the forward scatterof fluorescence from samples of NIH:OVCAR-3 cells indirectlystained for CA125 using anti-CA125 MAb and scFv variants, whileno shift of fluorescence was observed in the control samples.Immunofluorescence (Fig. 8) revealed membrane-specific bindingof all scFv variants to target CA125 antigen expressed by NIH:

Fig. 2. Codon-optimized nucleotide sequences for amino acids in the variable light and heavy chains of anti-CA125 scFv. Single lettered amino acid residues are depictedsandwiched between the wild-type (WT) murine codons on top (typed in black) and the E. coli-optimized codons (Opt) for the same residue at the bottom (typed in color). Allcodons optimized for the following abundant amino acid residues in the scFv are typed in the following colors: serine (S) – red, glycine (G) – green, leucine (L) – orange, lysine(K) – purple, arginine (R) – light blue. Codons for these residues that did not need optimization for E. coli expression are typed in black and the corresponding amino acidresidue is underlined in the figure. All other optimized codons are coded in dark blue. (For interpretation of the references to color in this figure legend, the reader is referredto the web version of this article.)

S.K. Sharma et al. / Protein Expression and Purification 102 (2014) 27–37 31

Fig. 3. (A) Coomassie-stained SDS–PAGE of total cell protein samples from the six anti-CA125 scFv constructs expressed in E. coli. (B) A replica gel of the total cell proteinsamples immunoblotted for hexa-histidine tagged anti-CA125 scFv. Arrows indicate the position of migration corresponding to the molecular weight of the scFv �28 KDa andis indicative of its presence in all samples. Lane M indicates the pre-stained SDS–PAGE low range molecular weight standard.

Fig. 4. (A) Coomassie-stained SDS–PAGE of samples containing the insoluble and soluble proteins obtained after processing of induced recombinant cultures using BugBustermaster mix reagent for cell lysis. Insoluble fractions are labeled with a prefix Pell and corresponding soluble fractions for each construct are labeled with a prefix Sup. (B)Replica gel of the same samples immunoblotted for the hexa-histidine tagged anti-CA125 scFv. Arrows indicate the position for migration of the scFv corresponding to�28 KDa.

32 S.K. Sharma et al. / Protein Expression and Purification 102 (2014) 27–37

OVCAR-3 cells and virtually no binding to CA125-negative SKOV3cells. Control samples that included only Alexa fluor 488 conju-gated secondary antibodies showed no membrane bound fluores-cence, thus discounting any non-specific association observed inthese cells.

Discussion

Obtaining high yields of a functional recombinant protein is aprerequisite to its extensive characterization, molecular engineer-ing and more importantly for eventual translation as a biopharma-ceutical entity. Given the clinical potential of MAb B43.13 for

immunotherapy of epithelial ovarian cancer (EOC) [30,31] andthe previously demonstrated in vitro targeting capabilities of itsderivative scFv [20,21,23], we revisited the anti-CA125 MAbB43.13 scFv in order to extend its scope for application in molecu-lar imaging of EOC. Previous reports for this scFv by Luo et al.,had not mentioned purified yields of the scFv produced in Pichiapastoris [18], while Kriangkum et al., reported yields of � 0.2 mg/L for the scFv produced as a bi-specific scFv using NS0 myelomacells [20] and Wang et al., reported 0.75 mg/L of the same scFvas a bi-functional fusion protein produced in E. coli [21]. Since allthe previously reported yields were practically suboptimal for con-ducting extensive in vitro and in vivo studies, it was imperative to

Fig. 5. (A) Representative Coomassie-stained SDS–PAGE of batch-binding experiments performed using lysates from various anti-CA125 scFv constructs. Each construct has 5samples that have been analyzed with the following codes: UB, Unbound proteins; W1, Wash 1; W2, Wash 2; E1, Eluate 1; E2, Eluate 2. (B) Replica gel of the same samplesimmunoblotted for the hexa-histidine tagged anti-CA125 scFv. Lane M represents prestained protein molecular weight standard. Arrows indicate the position and presence ofthe anti-CA125 scFv in the samples.

Table 1IMAC purification of 1 L cultures for each scFv construct provided the following yields (averaged from n = 5).

Construct Culture vol (L) Wet cell weight (g) scFv yield (mg) scFv yield (mg/liter) scFv yield (mg/gram)

VH-(G2S)5-VL 1 6.0 0.55 0.55 0.092VL-(G2S)5-VH 1 5.5 0.35 0.35 0.064VH-(G4S)3-VL 1 5.6 1.11 1.11 0.198VL-(G4S)3-VH 1 7.5 1.05 1.05 0.140VL-(218)-VH 1 7.3 0.4 0.4 0.055GA218 1 5.2 4.0 4 0.769

Fig. 6. (A) Coomassie-stained SDS–PAGE of the IMAC purified scFv constructs expressed in E. coli. (B) Replica gel of the same samples from (A) immunoblotted for hexa-histidine tagged anti-CA125 scFv.

S.K. Sharma et al. / Protein Expression and Purification 102 (2014) 27–37 33

increase the soluble expression and yields of the anti-CA125B43.13 scFv obtained per batch of recombinant E. coli culture.

Taking cues from the reported outcomes for other scFvs as citedfurther in this article, here we investigated the effect of 3parameters on the yields of soluble anti-CA125 scFv expression

in E. coli: (a) variable domain orientation, (b) inter-domain linkersand (c) codon optimization. The (G4S)3 linker was included in thisstudy as the most canonical flexible peptide linker used betweenscFv domains [32], while the (G2S)5 inter-domain linker wasretained from the construct previously reported by Wang et al.

Fig. 7. (A) Immunoblots of SKOV3 and NIH:OVCAR-3 cell lysates probed for the presence of CA125 antigen using various anti-CA125 scFv constructs as represented at thebottom of the image. Arrows indicate the migratory positions and bands of CA125 antigen and b-actin in the cell lysates upon electrophoresis in a 4–15% gradient SDS–polyacrylamide gel. (B) A 3-dimensional representation of histograms from flow cytometry analysis for the binding of various anti-CA125 scFv constructs and controls withNIH:OVCAR-3 cells. The X-axis represents forward scatter and Y-axis represents the cell counts.

34 S.K. Sharma et al. / Protein Expression and Purification 102 (2014) 27–37

[21]. The 218 linker was chosen for its reported ability to impartproteolytic stability and decreased aggregation of the scFv [33].

Despite achieving successful expression of all the scFv con-structs examined in this study, we observed a typically challengingphenomenon with heterologous protein expression using E. coli,wherein overexpressed recombinant proteins aggregate to forminclusion bodies [34,35]. Furthermore, this feature is reported tobe more pronounced in the case of antibody fragments such asscFvs, which tend to expose hydrophobic patches on the surfacesof their in vivo folding intermediates; that ultimately favors proteinaggregation [36,37]. Though it cannot be generalized, scFv mole-cules derived from hybridomas are more likely to be victims of thisphenomenon over those selected from phage display libraries,which may have a higher propensity for soluble expression inE. coli owing to prior enrichment and selection as fusion proteinsexpressed from phagemids.

In our efforts to obviate scFv aggregation, other parameterssuch as lowering post-induction bacterial growth temperaturesbetween 16 �C and 30 �C, using various concentrations of IPTGranging from 0.01 mM to 1.0 mM for induction and the use of0.4 M sucrose for protein expression (data not shown) were tested

with no significant improvement in yields of soluble scFv. Purifica-tion of the anti-CA125 scFv VL-(G4S)3-VH from inclusion bodiesprovided highly pure scFv with yields up to 6 mg/L. Nonetheless,attempts to refold the denatured scFv in the presence of arginineand prescribed ratios of redox components repeatedly resulted inprecipitation of >95% of the protein during dialysis in phosphatebuffered saline (data not shown).

Consequently, we chose to express and purify the soluble,in vivo folded and biochemically active scFv instead of isolating itfrom inclusion bodies under denaturing conditions and performingin vitro refolding procedures. In the present work, small scale batchbinding studies were instrumental in demonstrating the purifiedsoluble scFv yields for all constructs. The intensity of chemilumi-nescent signal corresponding to anti-CA125 scFv in the E1 fractionfrom GA218 was clearly indicative of higher amounts of solublescFv produced from this construct in comparison to the others;especially since equal volumes of all eluates were used for suchan analysis starting from normalized volumes of recombinant celllysate supernatants used in the batch binding procedure.

Unlike reports for some scFv molecules [38,39], we found verylittle difference in the yields for soluble anti-CA125 scFv (WT)

Fig. 8. Representative images from immunofluorescence staining of CA125 using the various anti-CA125 scFv constructs on fixed NIH:OVCAR-3 and SKOV3 cells. GAM-A488represents Alexa fluor 488 conjugated goat anti-mouse antibody. V-His A488 represents Alexa fluor conjugated anti-penta Histidine antibody.

S.K. Sharma et al. / Protein Expression and Purification 102 (2014) 27–37 35

expression in E. coli as a result of domain orientation. Nonetheless,our results indicated that the VH–VL orientation consistently pro-vided �1.4-fold higher yields (mg per gram wet cell weight) ofthe anti-CA125 scFv, for both the inter-domain linkers: (G2S)5

and (G4S)3. We also observed that the anti-CA125 scFv constructwith (G4S)3 linker yielded more than twice the amounts of solubleWT scFv (mg/L and mg/g) over the (G2S)5 linker constructs in eitherdomain orientation. This may be partially attributed to an increaseof serine residues in the (G2S)5 linker and its corresponding non-optimized codons in the WT scFv sequence. Surprisingly, we foundthat the 218 linker-containing anti-CA125 scFv yielded muchlower soluble scFv in contrast to its expected impact for decreasedin vivo aggregation.

Going forward, we confined the scope of this study to the VL–VH

orientation and found that codon optimization of the anti-CA125scFv in this format had the maximum positive impact on its solubleexpression in E. coli. This was evidenced by the poor codon adapta-tion index of 0.64 for the WT anti-CA125 scFv sequence and thefact that 58.3% of the WT codons were modified to suit the bias

for heterologous expression in E. coli. Additionally, the challengeto E. coli’s translational machinery during heterologous overex-pression of such proteins is briefly exemplified by the fact thatwhile a native E. coli protein would comprise of �5.7% serine resi-dues [40], the anti-CA125 scFv constructs examined in this studywere comprised of �16% serine residues inclusive of those presentin the inter-domain linkers. Furthermore, �63% of these serine res-idues were encoded by rare codons such as TCA, AGT, TCG, TCC(arranged in the increasing order of their average frequency andabundance to code for serine) in E. coli [40]. Using E. coli Rosetta2(DE3) as the expression host strain supplemented with tRNAs toprovide the canonical rare codons for arginine (R), isoleucine (I),proline (P) and leucine (L) may have worked to our advantage bypreventing truncated translation of the scFv during in vivo expres-sion. However, the tendency of GA218 scFv to aggregate despitecodon optimization and expression in a suitable host can be attrib-uted to the primary protein sequence of the scFv itself and a plau-sible exposure of its hydrophobic patches in a folding intermediatethat may result in off-pathway aggregation to form periplasmic

36 S.K. Sharma et al. / Protein Expression and Purification 102 (2014) 27–37

inclusion bodies [41]. Nevertheless, in the present study, we weresuccessfully able to demonstrate and achieve highest production ofsoluble anti-CA125 B43.13 scFv primarily as a result of codon opti-mization. Further improvement of the soluble expression and cor-responding purified yields relevant to the biopharmaceuticalproduction of this scFv may be achieved via engineering of hydro-phobic amino acid residues in the framework regions of this mol-ecule without negatively affecting its immunoreactivitycombined with the use of solubility enhancing tags and fermenta-tion scale bioreactors.

In our hands, we found consistently higher yields of the scFv insoluble fractions from recombinant cell lysates upon using Bug-Buster master mix reagent when compared to the traditionalosmotic shock method (data not shown). Additionally, using thecobalt based metal affinity resin TALON�allowed single step purifi-cation of the hexa-histidine tagged scFv in high purity due toreduced non-specific binding of proteins from the cell lysate super-natant and convenient downstream processing on account of lowerconcentrations of imidazole used in elution buffers. N-terminalprotein sequencing performed for two of the purified scFv variantsVL-(G4S)3-VH and GA218 revealed the absence of pelB leader pep-tide; thus indicating that the scFv(s) were processed and success-fully purified from the soluble periplasmic fractions. Successfultranslocation of the scFv into the oxidized environment of the per-iplasm increases the possibilities for intradomain disulfide bondformation, which eventually results in the molecule attaining aproperly folded structure to be a functionally active scFv.

Finally, a set of qualitative assessments to evaluate the bio-chemical functionality by virtue of the purified scFv’s antigen bind-ing capability demonstrated that all constructs examined wereable to bind the target antigen with high specificity as observedfrom immunoblotting and flow cytometry experiments. The pat-tern of membrane bound fluorescence of CA125 observed uponimmunostaining NIH:OVCAR-3 cells was in agreement with a pre-vious report [42], which employed a FITC-labeled anti-CA125 tar-geting MAb B27.1. However, we were able to obtain higherresolution and superior quality images for immunostaining ascompared to previous reports employing anti-CA125 targetingMAb [42] and scFv [21]. Furthermore, these studies also promptthe use of NIH:OVCAR-3 and SKOV3 cells to develop tumor xeno-graft models using mice for in vivo testing of targeting by anti-CA125 B43.13 scFv.

Conclusion

In the present work, we improved the soluble expression ofanti-CA125 B43.13 scFv in E. coli to obtain significantly higheryields of the epithelial ovarian cancer targeting vector. We foundthat codon optimization of the murine anti-CA125 scFv sequenceand its expression in a tRNA supplemented host provided up to14-fold higher yields of soluble anti-CA125 scFv when all otherparameters for expression and purification were kept constant.This is a significant step forward in obtaining high yields of func-tionally active anti-CA125 scFv towards a greater scope for molec-ular characterization and engineering with extended applicationsfor biomedical research.

Acknowledgments

The authors thank Lars-Oliver Klotz, Hoon Sunwoo andDharmendra Raghuwanshi for access to molecular biology anddownstream protein purification facilities in the Faculty of Phar-macy and Pharmaceutical Sciences, University of Alberta; BonnieAndrais for providing NIH:OVCAR-3 and SKOV3 cells, Xuejun Sunand Geraldine Barron in the Cell Imaging Facility, Dept. of Experi-

mental Oncology at the University of Alberta for support with cellimaging; Jingzhou Huang for help with flow cytometry, VincentBouvet for materials procurement, Jack Moore at the Institute ofBiomolecular Design, University of Alberta, for help with LC–MS/MS analysis, Joel Nott at the Protein Facility, Iowa State Universityfor N-terminal peptide sequencing, Cheryl Nargang at the Univer-sity of Alberta for DNA sequencing.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.pep.2014.07.007.

References

[1] P. Holliger, P.J. Hudson, Engineered antibody fragments and the rise of singledomains, Nat. Biotechnol. 23 (2005) 1126–1136.

[2] R.E. Bird, K.D. Hardman, J.W. Jacobson, S. Johnson, B.M. Kaufman, S.M. Lee,et al., Single-chain antigen-binding proteins, Science 242 (1988) 423–426.

[3] S.M. Kipriyanov, High level periplasmic expression and purification of scFvs,Methods Mol. Biol. 562 (2009) 205–214.

[4] T. Clackson, H.R. Hoogenboom, A.D. Griffiths, G. Winter, Making antibodyfragments using phage display libraries, Nature 352 (1991) 624–628.

[5] C. Schaffitzel, J. Hanes, L. Jermutus, A. Plckthun, Ribosome display: an in vitromethod for selection and evolution of antibodies from libraries, J. Immunol.Methods 231 (1999) 119–135.

[6] M.J. Feldhaus, R.W. Siegel, Yeast display of antibody fragments: a discoveryand characterization platform, J. Immunol. Methods 290 (2004) 69–80.

[7] J.V. Schaefer, A. Honneger, A. Pluckthun, Construction of scFv fragments fromhybridoma or spleen cells by PCR assembly, in: R. Kontermann, S. Dubel (Eds.),Antibody Engineering, vol. 1, Springer-Verlag, Berlin, Heidelberg, 2010, pp. 21–44.

[8] L. Toleikis, A. Frenzel, Cloning single-chain antibody fragments (scFv) fromhybridoma cells, Methods Mol. Biol. 907 (2012) 59–71.

[9] C. Fields, D. O’Connell, S. Xiao, G.U. Lee, P. Billiald, J. Muzard, Creation ofrecombinant antigen-binding molecules derived from hybridomas secretingspecific antibodies, Nat. Protoc. 8 (2013) 1125–1148.

[10] H.M. Davis, V.R. Zurawski Jr., R.C. Bast Jr., T.L. Klug, Characterization of theCA125 antigen associated with human epithelial ovarian carcinomas, CancerRes. 46 (1986) 6143–6148.

[11] N. Scholler, N. Urban, CA125 in ovarian cancer, Biomarkers Med. 1 (2007)1513–1523.

[12] V. Capstick, G.D. Maclean, M.R. Suresh, D. Bodnar, S. Lloyd, L. Shepert, et al.,Clinical evaluation of a new two-site assay for CA125 antigen, Int. J. Biol.Markers 6 (1991) 129–135.

[13] K. Nustad, R.C. Bast Jr., T.J. O’Brien, O. Nilsson, P. Sequin, M.R. Suresh, et al.,Specificity and affinity of 26 monoclonal antibodies against CA 125 antigen:first report of the ISOBM TD-1 workshop, Tumor Biol. 17 (1996) 196–219.

[14] M. Nap, A. Vitali, K. Nustad, R.C. Bast Jr., T.J. O’Brien, O. Nilsson, et al., Theimmunohistochemical characterization of 22 monoclonal antibodies againstthe CA125 antigen: 2nd report from ISOBM TD-1 workshop, Tumor Biol. 17(1996) 325–331.

[15] Y. Chen, S. Clark, T. Wong, Y. Chen, Y. Chen, M.S. Dennis, et al., Armedantibodies targeting the mucin repeats of the ovarian cancer antigen, MUC16,are highly efficacious in animal tumor models, Cancer Res. 67 (2007) 4924–4932.

[16] J. Singleton, D.E. Guillen, M.S. Scully, J. Xue, J. Moffet, C. Chen, et al.,Characterization of antibodies to CA 125 that preferentially bind to the cell-associated form of the antigen, Tumor Biol. 27 (2006) 122–132.

[17] J.S. Berek, Immunotherapy of ovarian cancer with antibodies: a focus onoregovomab, Expert Opin. Biol. Ther. 4 (2004) 1159–1165.

[18] D. Luo, M. Geng, B. Schultes, J. Ma, D.Z. Xu, N. Hamza, et al., Expression of afusion protein of scFv-biotin mimetic peptide for immunoassay, J. Biotechnol.65 (1998) 225–228.

[19] Y. Hashimoto, K. Tanigawa, M. Nakashima, K. Sonoda, T. Ueda, T. Watanabe,et al., Construction of the single-chain Fv from 196–14 antibody towardovarian cancer-associated antigen CA125, Biol. Pharm. Bull. 22 (1999) 1068–1072.

[20] J. Kriangkum, B. Xu, C. Gervais, D. Paquette, F.A. Jacobs, L. Martin, et al.,Development and characterization of a bi-specific single-chain antibodydirected against T cells and ovarian carcinoma, Hybridoma 19 (2000) 33–41.

[21] W.W. Wang, D. Das, S.A. McQuarrie, M.R. Suresh, Design of a bifunctionalfusion protein for ovarian cancer drug delivery: single chain anti-CA125 corestreptavidin fusion protein, Eur. J. Pharm. Biopharm. 65 (2007) 398–405.

[22] M. Boivin, D. Lane, A. Piche, C. Rancourt, CA125 (MUC16) tumor antigenselectively modulates the sensitivity of ovarian cancer cells to genotoxic drug-induced apoptosis, Gynecol. Oncol. 115 (2009) 407–413.

[23] J. Ma, D. Luo, G.S. Kwon, J. Samuel, A.A. Noujaim, R. Madiyalakan, Use ofencapsulated single chain antibodies for induction of anti-idiotypic humoraland cellular immune responses, J. Pharm. Sci. 87 (1998) 1375–1378.

S.K. Sharma et al. / Protein Expression and Purification 102 (2014) 27–37 37

[24] L. Accardi, P. Di Bonito, Antibodies in single chain format against tumor-associated antigens: present and future applications, Curr. Med. Chem. 17(2010) 1730–1755.

[25] N.A. Burgess-Brown, S. Sharma, F. Sobott, C. Loenarz, U. Oppermann, O. Gileadi,Codon optimization can improve expression of human genes in Escherichiacoli: a multi-gene study, Protein Expr. Purif. 59 (2008) 94–102.

[26] A. Tiwari, A. Sankhyan, N. Khanna, S. Sinha, Enhanced periplasmic expressionof high affinity humanized scFv against hepatitis B surface antigen by codonoptimization, Protein Expr. Purif. 74 (2010) 272–279.

[27] C. Gustafsson, S. Govindarajan, J. Minshull, Codon bias and heterologousprotein expression, Trends Biotechnol. 22 (2004) 346–353.

[28] A.A. Noujaim, R.P. Baum, T.R. Sykes, C.J. Sykes, A. Hertel, A. Niesen, et al.,Monoclonal antibody B43.13 for immunoscintigraphy and immunotherapy ofovarian cancer, in: R. Klapdor (Ed.), Current Tumor Diagnosis: Applications,Clinical Relevance, Trends. W Zuckschwerdt Verlag, Munich, 1994, pp. 823–829.

[29] M. Newa, M. Lam, K.H. Bhandari, B. Xu, M.R. Doschak, Expression,characterization and evaluation of a RANK-binding single chain fractionvariable: an osteoclast targeting drug delivery strategy, Mol. Pharm. 11 (2014)81–89.

[30] V.J. Mobus, R.P. Baum, M. Bolle, R. Kreienberg, A.A. Noujaim, B.C. Schultes,Immune responses to murine monoclonal antibody-B43.13 correlate withprolonged survival of women with recurrent ovarian cancer, Am. J. Obstet.Gynecol. 189 (2003) 28–36.

[31] A.A. Noujaim, B.C. Schultes, R.P. Baum, R. Madiyalakan, Induction of CA125-specific B and T cell responses in patients injected with MAb-B43.13 –evidence for antibody-mediated antigen processing and presentation of CA125in vivo, Cancer Biother. Radiopharm. 16 (2001) 187–203.

[32] A.A. Kortt, O. Dolezal, B.E. Power, P.J. Hudson, Dimeric and trimeric antibodies:high avidity scFvs for cancer targeting, Biomol. Eng. 18 (2001) 95–108.

[33] M. Whitlow, B.A. Bell, S.L. Feng, D. Filpula, K.D. Hardman, S.L. Hubert, et al., Animproved linker for single-chain Fv with reduced aggregation and enhancedproteolytic stability, Protein Eng. 6 (1993) 989–995.

[34] L.A. Palomares, S. Estrada-Mondaca, O.T. Ramirez, Production of recombinantproteins: challenges and solutions, Methods Mol. Biol. 267 (2004)15–52.

[35] M.M. Carrio, A. Villaverde, Construction and deconstruction of bacterialinclusion bodies, J. Biotechnol. 96 (2002) 3–12.

[36] L. Nieba, A. Honnegger, C. Krebber, A. Pluckthun, Disrupting the hydrophobicpatches at the antibody variable/constant domain interface: improved in vivofolding and physical characterization of an engineered scFv fragment, ProteinEng. 10 (1997) 435–444.

[37] P.S. Chowdhury, G. Vasmatzis, Engineering scFvs for improved stability,Methods Mol. Biol. 207 (2003) 237–254.

[38] K. Tsumoto, Y. Nakaoki, Y. Ueda, K. Ogasahara, K. Yutani, K. Watanabe, Effect ofthe order of antibody variable regions on the expression of the single-chainHyHEL10 Fv fragment in E. coli and the thermodynamic analysis of its antigen-binding properties, Biochem. Biophys. Res. Commun. 201 (1994) 546–551.

[39] S. Hamilton, J. Odili, O. Gundogdu, G.D. Wilson, J.M. Kupsch, Improvedproduction by domain inversion of single-chain Fv antibody fragment againsthigh molecular weight proteoglycan for the radioimmunotargeting ofmelanoma, Hybrid Hybridomics 20 (2001) 351–360.

[40] J.F. Kane, Effects of rare codon clusters on high-level expression ofheterologous proteins in Escherichia coli, Curr. Opin. Biotechnol. 6 (1995)494–500.

[41] A. Knappik, A. Pluckthun, Engineered turns of a recombinant antibody improveits in vivo folding, Protein Eng. 8 (1995) 81–89.

[42] Z. Xiao, S.A. McQuarrie, M.R. Suresh, J.R. Mercer, S. Gupta, G.G. Miller, A three-step strategy for targeting drug carriers to human ovarian carcinoma cellsin vitro, J. Biotechnol. 94 (2002) 171–184.