key: cord-0726428-69bovl8q
authors: Peipp, Matthias; Simon, Nicola; Loichinger, Alexandra; Baum, Wolfgang; Mahr, Kerstin; Zunino, Susan J.; Fey, Georg H.
title: An improved procedure for the generation of recombinant single-chain Fv antibody fragments reacting with human CD13 on intact cells
date: 2001-05-01
journal: Journal of Immunological Methods
DOI: 10.1016/s0022-1759(01)00298-8
sha: 41196f499004b5556ac6f96d012283e004521023
doc_id: 726428
cord_uid: 69bovl8q

Abstract A procedure was developed to generate recombinant single chain Fv (scFv) antibody fragments reacting with the extracellular domain of human cell surface antigen CD13 (hCD13; aminopeptidase N) on intact cells. Membrane fractions prepared from a stably transfected hCD13-positive murine NIH/3T3 cell line were used to immunize BALB/c mice, with the intention that hCD13 would be the major immunogenic molecule recognized by the immune system. Spleen RNA from the immunized mice served to generate a combinatorial scFv phage display library. The library was adsorbed against non-transfected NIH/3T3 or Sf21 insect cells to eliminate nonrelevant binders. The supernatant was then used for panning with either hCD13-transfected Sf21 insect cells or a hCD13-expressing human leukemia-derived cell line. Therefore, the key concepts of the procedure were the presentation of hCD13 as the sole human antigen on murine NIH/3T3 cells and a screening strategy where hCD13 was the major common antigen of the material used for immunization and panning. Two different hCD13-reactive phages were isolated and the soluble scFvs were expressed in E. coli and purified. The two scFvs, anti-hCD13-1 and anti-hCD13-3, differed at four amino acid positions in their VH regions and both had high affinities for hCD13 as determined by surface plasmon resonance (K D=7 and 33×10−10 M, respectively). Both efficiently recognized hCD13 on intact cells. Therefore, the procedure allowed the production of high affinity scFvs reacting with a desired antigen in its native conformation without requiring extensive purification of the antigen and should be useful for the preparation of scFvs against other conformation-sensitive cell-surface antigens.

1999) and functions as a membrane-bound metallop-roteinase that catalyzes the removal of N-terminal infants with leukemias associated with translocations amino acids from polypeptides and small peptides t(4;11) and t(11;19) affecting the MLL-gene on (Mizutani et al., 1993; Hooper, 1994) . CD13 is chromosome 11q23 have poor prognosis and new expressed in epithelial cells in liver, kidney and therapeutic approaches for these leukemias are intestine, in placenta, brain and spleen, and in a needed (Behm et al., 1996; Heerema et al., 1999) . variety of hematopoietic cells. Recently, roles for Immunotherapeutic approaches targeting individual CD13 in metastasis and in the degradation of ex-antigens on tumor cells have suffered in the past tracellular matrix components by tumor cells were from the fact that the same antigens were also reported (Saiki et al., 1993; Fujii et al., 1995) . In present on normal cells. By contrast, the mixedaddition, CD13 was identified as a receptor for lineage antigen spectrum is preferentially found on tumor-homing peptides in the vasculature of tumors the tumor cells and therefore offers novel possiand appears to play a role in the aberrant angio-bilities to specifically target the tumor cells. The genesis of tumor tissues. Therefore, CD13 was long-range objective of our research is to contribute proposed as a useful target for the delivery of drugs to the development of immunotherapeutic apto tumors that aim at inhibiting angiogenesis (Pas-proaches against leukemic blasts with chromosomal qualini et al., 2000) . Furthermore, CD13 is a receptor translocations t(4;11) or t(11;19) to the MLL gene. for cytomegalovirus (CMV; Soderberg et al., 1993;  For this purpose cDNAs coding for antibodies or Giugni et al., 1996) and certain coronaviruses (De-antibody fragments reactive with several of the lmas et Yeager et al., 1992) and future component antigens of the mixed lineage spectrum antiviral strategies directed against these viruses may are needed. also benefit from including CD13 as a target (Lar- The immediate purpose of the present study was to sson et al., 1998) .

generate recombinant single chain Fv fragments In the hematopoietic system, CD13 is expressed (scFv) that bind human CD13 in its native conformapredominantly in myelo-monocytic cells (Shapiro et tion on intact human leukemic blasts. To our knowlal., 1991; Shipp and Look, 1993; Olsen et al., 1997) .

edge, no reports on scFvs reactive against hCD13 on The antigen is present on early hematopoietic intact cells have been published, although a number progenitors including progenitors of the B-and T-of anti-human CD13 hybridoma antibodies have lymphoid lineages, and expression is progressively been reported (Goyert, 1997) . A standard procedure reduced with increasing maturation of these cells.

for the generation of scFvs consists in the bacterial The antigen is not detectable on mature lymphocytes expression of the antigen followed by biochemical from peripheral blood, spleen and tonsils (Spits et purification and the use of the purified antigen for the al., 1995). CD13 is often upregulated on malignant immunization of mice. However, hCD13 is heavily cells, particularly on blasts from acute and chronic glycosylated (Look et al., 1989) and an authentic B-cell leukemias (ALLs and CLLs), including infant glycosylation is not achieved by bacterial expression. pro-B cell leukemias (Makrynikola et al., 1995) . The Therefore, mice immunized with bacterially expro-B ALLs frequently express CD19, CD22 and pressed antigen may produce antibodies reacting CDw75 as lymphoid markers and CD13, CD15, only with the recombinant protein, but not with the CD33, CDw65 as myeloid markers (Stong et al., antigen in its native conformation on human cells 1985; Greil et al., 1994) . The expression of myeloid (Chowdhury et al., 1998; Huls et al., 1999) . An markers on a pro-B cell background is unusual and alternative strategy that avoids the use of bacterially highly characteristic for these particular leukemias. expressed hCD13 would involve the preparation of Therefore, special terms including 'biphenotypic cell surface membranes from CD13 positive human expression' and 'mixed-lineage phenotype' have cells, such as CD13-expressing leukemia cells been coined to designate this property. (Stong et al., 1985; Greil et al., 1994) . The dis-The mixed lineage phenotype provides a unique advantage of this approach is that the immunized opportunity to direct novel immuno-therapeutic ap-mice are likely to produce an excess of antibodies proaches against the pro-B ALLs that are often against other human cell surface antigens and antirefractory to other forms of treatment. In particular, bodies against hCD13 may be of low abundance.

Anti-hCD13 scFvs would, therefore, be difficult to containing the coding sequences for the Fc portion of isolate from combinatorial phage libraries prepared a human IgG1 heavy chain (M. Peipp, unpublished from immunized mice without further specific modi-data). The resulting chimeric protein with an Nfications of the procedure added to maximize the terminal portion consisting of the extracellular doprobability of isolating scFvs against the desired main of hCD13 and the Fc domain of human IgG1 at antigen. To avoid these problems, a novel procedure its C-terminus was designated s-hCD13. was designed here with the purpose of maximizing the yield of anti-hCD13 scFv antibodies while 2.2. Culture of eukaryotic cells avoiding the production of antibodies against other human cell surface antigens. Indeed, this procedure NIH / 3T3 murine fibroblasts and human 293T produced recombinant scFvs against human CD13 at kidney cells were cultured in DMEM-Glutamax-I high frequency. An advantage of the new procedure medium (Life Technologies, Karlsruhe, Germany) is that it does not require the extensive biochemical supplemented with 10% fetal calf serum (FCS) and purification of the antigen, which can result in the penicillin and streptomycin (Life Technologies) at partial or total loss of its native conformation. We concentrations of 100 U / ml and 100 mg / ml, respecanticipate that this procedure will be applicable to a tively. Leukemia-derived SEM cells (Greil et Germany) and transfection was performed following 1989) was digested with SpeI /EcoRV and the insert the manufacturer's protocol. Selection medium conwas ligated to SpeI/EcoRV digested pcDNA3.1(1) taining 600 mg / ml of G418 was exchanged every (Invitrogen, Groningen, Netherlands), resulting in 3-4 days. After 2 weeks of selection, single cells construct pcDNA-hCD13. For stable transfection of were subcloned. Isolated clones were analyzed for Sf21 cells, the EcoRV/NotI fragment of hCD13 surface expression of hCD13 by fluorescent-activated cDNA was ligated into pIZT-V5-His (Invitrogen), cell sorting (FACS). Sf21 cells were seeded in 60-5 resulting in pIZT-hCD13. To express soluble hCD13, mm plates at 4310 cells / well. NdeI-linearized a cDNA fragment lacking the transmembrane and pIZT-hCD13 vector (3.6 mg) and BstEII linearized intracellular domains was amplified by PCR from pIZT-V5-His vector carrying a Zeocin resistance pIZT-hCD13. The primers used for this amplification gene (0.4 mg) were cotransfected using the TransFast were: CD13-Fc-EcoRV (59-tac gat atc aga aga aca reagent (Promega, Mannheim, Germany). After 3 aga acg cca ac-39) and CD13-Fc-NotI (59-ctg ggc weeks of continuous selection with 400 mg / ml ggc cgc tcg tgt ttt ctg tga acc ac-39). The PCR Zeocin, the mixed cell population was examined by amplimer was ligated into the vector pSecTag-C-Fc FACS for surface expression of hCD13. The popula- , 1997) . PCR amplification of immunowere suspended in lysis buffer consisting of 1 mM globulin variable region cDNAs and cloning into the NaHCO , pH 8, 1 mM MgCl and 100 mM Pefab-3 2 phagemid vector pAK100 was performed as delock proteinase inhibitor (Roth, Karlsruhe, Gerscribed (Krebber et al., 1997) . The initial commany). After incubation on ice for 15-20 min, the binatorial scFv library was propagated in SBcells were disrupted in a dounce homogenizer. The medium (1% glucose; 30 mg / ml chloramphenicol) at lysate was cleared by centrifugation and sucrose was 378C under vigorous shaking. After the density of added to a final concentration of 1.59 M. Ten-ml the culture reached an OD of 0.5, 100 ml of the 600 aliquots were overlayed in SW28 rotor tubes (Beckculture were infected with helper phage M13KO7 man, Munich, Germany) with 9 ml of 1.175 M (Stratagene). The culture was incubated at 378C for sucrose in carbonate buffer (1 mM NaHCO ; pH 8). 3 another 30 min without shaking. Then, 400 ml of A second layer of 7 ml of 0.98 M sucrose and a third SB-medium were added and isopropyl b-D-thiogalaclayer of 2 ml of 0.8 M sucrose were added. Gradient topyranoside (IPTG) was supplied at a final concentrifugation was performed for 16 h at 70,0003g centration of 0.5 mM. The culture was incubated at in an SW28 rotor. Membrane fractions were col-378C for another 2 h under vigorous shaking (250 lected, diluted in 5 volumes of phosphate buffered rev. / min) and then kanamycin was added (25 mg / saline (PBS), and sedimented for 1 h at 70,0003g as ml). Finally, the culture was incubated at 308C previously described (Eylar and Hagopian, 1971) .

overnight. Phages were precipitated with poly-Membranes were resuspended in 100 ml of PBS, and ethylene glycol (20% PEG 6000, 2.5 M NaCl) and fractions containing hCD13 were identified by resuspended in 2 ml PBS. ELISA.

BALB / c mice (Charles River, Sulzfeld, Germany) Panning of phage display libraries with intact cells were maintained according to the European guide-was carried out using two different approaches. In lines for the protection of laboratory animals. Ap-both cases, the phage display library was first proximately 500 mg of gradient-purified membrane adsorbed against non-transfected cells that were protein were combined with TiterMax GoldE ad-blocked in 2% nonfat dry milk in PBS (NM-PBS) juvant (Sigma, Deisenhofen, Germany) and injected for 30 min. In the first approach, 0.5 ml of the phage subcutaneously on day 0. On days 25 and 49, mice display library was resuspended in NM-PBS and 6 were boosted with 300 mg of membrane protein. A incubated for 1 h at RT with 3310 blocked nonfinal injection of 150 mg of membrane protein transfected NIH / 3T3 cells. The pre-adsorbed library followed on day 60. Four days later, the mice were was then used for positive selection on SEM 6 sacrificed and spleens were recovered under sterile leukemic cells. For this purpose, 10 SEM cells were conditions. The serum was tested at different blocked with NM-PBS and then incubated with the timepoints for the presence of anti-hCD13 antibodies pre-adsorbed scFv library for 1 h at RT under slow by FACS analysis using 293T cells that were tran-agitation. Cells were washed 10 times with 2% NM-PBS and twice with PBS. Bound phages were nitrilotriacetic acid (Ni-NTA) magnetic agarose eluted with 1.5 ml of 50 mM HCl for 10 min. After beads (Qiagen). neutralization with 0.5 ml of 1 M Tris-HCl at pH 8, cells were pelleted and the supernatant was used to 2.9. Flow cytometric analysis infect 10 ml of exponentially growing E. coli TG1 cells. To the culture, 20 ml 23YT medium con-For monitoring of scFv binding to intact cells, 5 taining 1% glucose and 30 mg / ml chloramphenicol 5310 cells were washed with PBS containing 0.1% were added and the cells were incubated for 2 h at BSA and 7 mM Na-azide (PBA). The cells were then 378C under vigorous shaking (250 rev. / min). Cells incubated with either 50 ml of bacterial periplasmic were superinfected with helper phage. Seventy ml of extracts or 50 ml of purified scFv (10 mg/ml) for 1 h 23YT medium were then added and the IPTG on ice. Periplasmic extracts from cells producing concentration was adjusted to 0.5 mM. Two hours nonrelevant scFv fragments served as isotype conafter infection, kanamycin was added at a final trols. The cells were then washed with PBA and concentration of 25 mg / ml and the culture was incubated with 20 ml of 0.2 mg / ml Penta-His allowed to grow overnight at 308C. On the following antibody (Qiagen) for 30 min on ice. Unbound day, phages were prepared as described above. In the antibodies were removed by washing with PBA and second approach, non-transfected Sf21 insect cells 10 ml of PE conjugated rat-anti-mouse-IgG1 antiwere used to pre-adsorb the phage display library body (Becton-Dickinson, Heidelberg, Germany) and positive selection was performed on Sf21-were added to the cell pellets. The cells were hCD13 cells by the panning procedure described incubated for 30 min on ice. After a final wash, above. Individual phages were clonally purified and FACS analysis was performed on a FACSCalibur sequence analysis of the inserts was performed using using CellQuest software (Becton-Dickinson). Ten dideoxynucleotide sequencing (Sambrook et al., thousand events were collected for each sample and 1989) on an Applied Biosystems automated DNA all analyses of whole cells were performed using sequencer (ABI Prism 310 Genetic Analyzer; Perkin appropriate scatter gates to exclude cellular debris Elmer, Ueberlingen, Germany). and aggregates. For monitoring surface expression of 5 hCD13 on transfected cell lines, 5310 cells were incubated for 30 min on ice with 20 ml of anti-2.8. Bacterial expression and purification of hCD13 antibody (Clone WM-47; DAKO Diagnostica soluble scFv fragments GmbH, Hamburg, Germany) at a concentration of 5 mg / ml. Mouse IgG1 served as the isotype control. For the soluble expression of antibody fragments,

The cells were washed in PBA and 10 ml of PE plasmids were propagated in E. coli HB2151 (from conjugated rat-anti-mouse-IgG1 antibody were Dr. G. Winter, MRC Cambridge, UK). Overnight added. After a final wash, cells were analyzed as cultures were diluted in 23YT medium sup-described above. plemented with 1% glucose and 30 mg / ml chloramphenicol until the extinction at 600 nm (OD ) was 2.10. Expression of soluble hCD13 (s-hCD13) in 600 #0.1. Cultures were grown to an OD of 0.6-0.8 mammalian cells 600 at 378C and expression was induced by the addition of IPTG at a final concentration of 1 mM and by 293T cells were transiently transfected with a lowering the temperature to 308C. After 4-6 h, the plasmid containing the cDNA coding for s-hCD13 bacteria were collected by centrifugation. Periplas-using Fugene6 reagent (Roche, Mannheim, Germic extracts were prepared using established proto-many). After 24 h, medium containing 10% FCS cols (Kipriyanov, 1997) and dialyzed extensively with reduced IgG content (Life Technologies) was against buffer containing 50 mM NaH PO , pH 8, added. Beginning at 48 h after transfection, the 2 4 300 mM NaCl, and 10 mM imidazole at 48C. medium was exchanged every 24 h and culture Purification of 63His tagged scFv fragments was supernatant containing secreted s-hCD13 was colachieved by affinity chromatography with nickel-lected each day for 7 days. The purification was performed on protein A agarose columns (Biorad, tion and single cell cloning. The lines, NIH / 3T3-Munich, Germany) as described elsewhere (Harlow hCD13 clones 4 and 6, expressed elevated levels of and Lane, 1988). A commercial anti-hCD13 anti-hCD13 antigen on their surface, as judged by FACS body (Clone WM-47) was used to identify s-hCD13 staining with a commercial anti-hCD13 antibody by ELISA. (Fig. 1) . The mean fluorescence intensity of NIH / 3T3 clones 4 and 6 stained with anti-hCD13 anti-2.11. SDS-PAGE and Western blot analysis body was |20-30 times greater than non-transfected NIH / 3T3 cells and cells stained with an isotype SDS-PAGE was carried out according to Laemmli control antibody (Fig. 1) . Although the absolute (1970). Gels were stained with Coomassie brilliant surface density of hCD13 expressed in number of blue R250. Western blots (Towbin et al., 1979) were molecules per cell was not determined, the disperformed with secondary antibodies coupled to tribution of the cells with respect to their fluoreshorseradish peroxidase (HRP; Dianova, Hamburg, cence intensity was comparable to the distribution Germany) and detected using ECL reagents (Amer-obtained for the leukemia-derived SEM cell line sham, Freiburg, Germany). ScFvs were detected with (data not shown). Therefore, clones 4 and 6 exthe Penta-His antibody (Qiagen). S-hCD13 was pressed sufficient surface densities of hCD13 to be detected using a HRP conjugated anti-human-Fc useful for the preparation of membrane fractions for antibody (Sigma).

the immunization of mice.

2.12. Surface plasmon resonance and 3.2. Preparation and enrichment of plasma determination of binding constants membranes from hCD13-expressing NIH /3T3 cells and immunization of BALB /c mice S-hCD13 was immobilized on a CM5 sensor chip by standard amine coupling (Biacore AB, Freiburg,

The procedure is outlined in Fig. 2 . After generat-Germany). The amount of deposited s-hCD13 was ing an NIH / 3T3 murine line expressing hCD13, the monitored by surface plasmon resonance. Coupling membrane fraction was prepared under gentle conwas continued until the amount deposited corres-ditions to avoid denaturation of the antigen. The ponded to |800 resonance units (RU). For binding enriched plasma membrane fraction was used for the analyses, the purified scFv fragments CD13-1 and immunization of mice. With this procedure, we CD13-3 were dialyzed against running buffer (10 intended to avoid the production of antibodies mM Hepes pH 7.4, 3.4 mM EDTA, 150 mM NaCl, against human plasma membrane proteins other than and 0.005% surfactant P20). The scFvs in the fluid hCD13. The immunized BALB / c mice were exphase were allowed to interact with the immobilized pected to be tolerant against most membrane ans-hCD13 at RT and a flow rate of 10 ml / min. tigens from NIH / 3T3 cells, with the possible excep-Varying concentrations of the scFvs were analyzed tion of alloantigens. Membranes were prepared from with the same chip after regeneration. Binding NIH / 3T3 clones 4 and 6 by gentle hypotonic constants were calculated using the BIAevaluation disruption and subsequent density centrifugation in a 3.0 software (Kazemier et al., 1996; Langmuir 1:1 sucrose gradient. Several membrane fractions were model, Biacore AB).

collected from the gradient and the hCD13-positive fractions were identified by ELISA (data not shown). The peak fraction (at 0.98 M) was used to immunize 3. Results BALB / c mice. In an average preparation, 400 mg of hCD13-positive plasma membrane protein were re-8 3.1. Generation of NIH /3T3-derived cell lines covered from 5310 cells and the estimated enrichexpressing hCD13 ment of plasma membrane proteins was 20-to 50fold. After 64 days and three boosts, the immunized Two NIH / 3T3-derived cell lines stably expressing mice developed a high titer of circulating antihuman CD13 (hCD13) were established by transfec-hCD13 antibodies, as judged by sampling aliquots of their peripheral blood at various times and perform-combined for panning. The combined library con-5 ing flow-cytometric assays with hCD13-positive cells tained |8-10310 independent clones. (transiently transfected 293T cells).

3.4. Isolation of CD13-reactive phages by 3.3. Preparation of combinatorial phage libraries differential panning displaying scFv antibody fragments on their pIII tail fiber proteins Panning was performed by employing two different variants of the protocol. In variant 1, the libraries Combinatorial scFv cDNA libraries were produced were pre-adsorbed with non-transfected NIH / 3T3 from total spleen RNA of the immunized mice cells and the supernatant was panned with intact according to published procedures (Krebber et al., hCD13-expressing SEM leukemia cells. In variant 2, 1997). From 10 mg of total spleen RNA, com-the libraries were pre-adsorbed with non-transfected 5 binatorial libraries with |1-3310 independent Sf21 insect cells and the supernatant was panned chloramphenicol-resistant clones were obtained be-with hCD13-expressing Sf21 cells. Therefore, Sf21 fore amplification. Only mouse k light chain cDNAs insect cells were transfected with a cDNA construct were amplified and included in the library. The l coding for hCD13 and a mixture of stably transfected light chains were omitted because they represented clones was used for the panning procedure. This only a small proportion of mouse antibodies. Six mixture expressed hCD13 with sufficient stability as separate libraries were produced, characterized, and judged by flow cytometry and obviated the need for phages was determined and an enrichment factor was calculated relative to the first round of panning. After three rounds, reactive phages were enriched 16-fold after panning with SEM cells and 66-fold after panning with hCD13 positive Sf21 cells (Table 1) . Individually isolated phages were clonally purified and their specificity of binding was evaluated by cellular ELISA. As a result, 12 of 14 individually isolated phages from the third round of panning specifically reacted with NIH / 3T3 cells stably expressing hCD13 and showed no binding to nontransfected NIH / 3T3 cells (data not shown). Characterization of the cDNA inserts of these phages revealed that two different scFv clones, anti-hCD13-1 and anti-hCD13-3, were isolated multiple times independently, using both variants of the panning protocol. Both had identical light chain variable regions (V ), but differed in four amino acid posi-L tions of their heavy chain variable regions (V ). Two H of these changes were located in the complementarity-determining regions 1 and 2 and the other two Fig. 2 . Flow diagram of the improved procedure for the isolation were found in framework 3 (Fig. 3) .

of anti hCD13 scFvs. Key steps of the procedure are outlined. Three sequential rounds of panning were employed for the

isolation of scFvs anti hCD13-1 and -3.

single cell subcloning. The mean fluorescence in-For the purpose of large-scale expression of the tensity of Sf21 cells stained with anti-CD13 antibody recombinant antibody fragments, the cDNAs coding was |50-100 fold greater than the intensity of the for the scFvs were subcloned into the expression same cells stained with an isotype-control antibody vector pAK400. E. coli strain HB2151 was chosen or that of non-transfected Sf21 cells (Fig. 1) .

for high level expression of the scFvs. Intact recom-The phages that specifically bound to the cells binant scFvs were secreted to the periplasmic comwere eluted with 50 mM hydrochloric acid (HCl) partment, as visualized by Western blot with an and immediately neutralized. The titer of eluted antibody reactive against the hexa-histidine tag (Fig. The factor of enrichment is calculated by setting the input / elution ratio of the first round to 1 and computing the ratios measured in subsequent rounds as multiples of the first round ratio (normalization to the first round). histidine tag as a C-terminal extension, which allowed affinity purification with magnetic Ni-NTA beads. The protocol for the preparation of the resulting in the typical yield described above. The periplasmic compartment and the purification with one-step purification procedure resulted in scFvs NTA beads was reduced to a one-step procedure essentially free of contaminants detectable by stain-ing with Coomassie blue (Fig. 4B, lanes 1 and 2) .

human IgG1 antibody ( Fig. 5 ; data shown for scFv anti-hCD13-1), indicating that the two scFvs reacted with hCD13 in the The two anti-hCD13 scFvs isolated here reacted native state.

with hCD13 in its native conformation and failed to react with extracts from hCD13-positive cells on Western blots (data not shown). Therefore, to determine molecular binding constants for scFvs anti-hCD13-1 and anti-hCD13-3, it was necessary to generate the antigen in its native conformation and in a pure state. To achieve these objectives, the antigen was fused to the Fc-domain of a human IgG1 antibody, since this procedure was known to generate soluble antigens in a native conformation (M. Peipp, unpublished data). In this procedure, chimeric molecules are secreted from mammalian cells and receive the post-translational modifications (e.g. glycosylation) characteristic of mammalian cells. For the purpose of secretion, sequences coding for a leader peptide from a murine immunoglobulin k light chain were fused to the coding sequences for the extracellular domain of hCD13 (Fig. 6A) . For the purpose of high level expression, 293T cells were transiently transfected with the construct and soluble human CD13 (s-hCD13) was then produced in high quantities for several days after transfection. The chimeric protein was collected from the culture supernatant and purified by affinity chromatography on a protein A agarose column. Using this procedure, typically 50-100 mg of the recombinant protein were recovered from 1 l of culture supernatant. The chimeric molecule reacted on a Western blot with an antibody against the Fc-portion (Fig. 6B ), and no contaminants or breakdown products were visible by staining with Coomassie blue (Fig. 6C ). The material reacted in an ELISA with the commercial anti-hCD13 (WM-47) antibody, as well as with the two isolated scFvs (data not shown). Thus, the integrity of the antigenic epitope in the chimeric s-hCD13 molecule resembled that on intact cells and this The soluble chimeric ligand s-hCD13 was de- posited on a Biacore CM5 sensor chip and chemical-hCD13-3 (Fig. 7) . Each scFv was also injected over ly coupled by standard procedures. Sequential rounds a blank inactivated surface and this signal was of loading were performed until the total amount subtracted from the signal obtained on the s-hCD13 coupled corresponded to |800 relative units (RU).

surface to remove contribution from potential un-Then the scFvs were allowed to flow over the chip in specific binding. The resulting affinities were 7.273 210 210 the fluid phase, and the association and dissociation 10 M for anti-hCD13-1 and 32.9310 M for reactions were monitored by surface plasmon reso-anti-hCD13-3, suggesting that anti-hCD13-3 had a nance. This procedure allowed a reliable measure-lower affinity (Table 2) . Therefore, the sequence ment of the association (k ) and dissociation rate differences in the V regions of both scFvs may have a H constants (k ). The affinity K was determined for resulted in measurably different affinities. The nud D each scFv antibody from the quotient k /k . The merical values of the affinities were in the nanomolar d a procedure was performed with several different range and below, which is typical for high affinity concentrations of the same scFv in the fluid phase antibodies resulting from a secondary immunization and with separate chips for anti-hCD13-1 and anti- (Griffiths et al., 1994) . Thus, the novel procedure reported here was efficient in capturing scFvs with high affinities from immunized mice.

The main result of this study was the establishment of a novel procedure for the generation of recombinant scFvs reactive with pre-selected antigens on the surface of human leukocytes. The procedure delivered scFvs reacting with hCD13 in its native state on the surface of intact cells with high affinity and without requiring an extensive purification of the antigen. The principle components and distinctive features of the procedure were the following: (a) the establishment of stably transfected NIH / 3T3 mouse and Sf21 insect cell lines expressing high levels of hCD13 on their cell surface; (b) the purification of membranes from NIH / 3T3 mouse cells expressing hCD13 by gentle extraction that maintained the antigen in a native state; (c) the immunization of mice with membranes containing hCD13 as the sole human antigen; (d) the preadsorption of the phage display library with nontransfected NIH / 3T3 or Sf21 cells that allowed the removal of antibodies against alloantigens, as well as non-specific proteins; and (e) panning the pre-adsorbed library with hCD13-expressing human Other related methods producing scFvs against plasmon resonance. The soluble chimeric protein s-hCD13 was deposited as a ligand on a Biacore CM5 sensor chip and purified cell surface antigens have been previously reported soluble scFvs anti-hCD13-1 and -3 were then applied to the chip (Portolano et al., 1993; Cai and Garen, 1995 , in increasing concentrations and allowed to bind. Binding was 1997 Siegel et al., 1997; Huls et al., 1999; Winthrop monitored by surface plasmon resonance using the BIACORE XE et al., 1999; Ridgway et al., 1999; Topping et al., system. (A, B) Association and dissociation profiles for scFvs 2000). In the majority of these cases, the libraries hCD13-1 and -3, respectively. Numbers in the right margins are concentrations of purified soluble scFv used for each curve. Data were panned subtractively with tumor-and nonare representative of two separate experiments.

tumor cell lines and the resulting scFvs were analyzed without prior knowledge of their identity and without specific features of the procedure designed to augment the efficiency of isolating antibodies against Table 2 predetermined specific antigens. Only a few cases Measurement of association and dissociation rate constants and have been reported where scFvs against pre-selected affinities of scFvs anti-hCD13-1 and -3 by surface plasmon specific cell surface antigens have been isolated by resonance panning with intact cells (Portolano et al., 1993; 21 21 21 scFv Hoogenboom et al., 1999) . The key feature of the membrane fragments purified from a transfected murine cell line. When membrane preparations from procedure, mice were immunized with a specific the hCD13-transfected NIH / 3T3 cells were injected antigen (hCD13) and then given multiple boosts into recipient mice, we reasoned that the human until a secondary and possibly even a tertiary CD13 antigen would be the major immunogenic immune response was well underway. This strategy molecule recognized by the immune system. A increased the likelihood of affinity maturation of second important point was the screening strategy antibodies in vivo and subsequent isolation of high where hCD13 was the major common antigen be-affinity antibodies against hCD13. Furthermore, two tween the material used for the immunization and the different scFvs reacting specifically with hCD13 6 cells used for panning, allowing a more stringent were obtained after screening a library of only 10 positive selection on intact cells. To the best of our clones. Thus, the strategy described here allowed us knowledge, a method employing a similar combina-to isolate high affinity scFvs against our antigen of tion of steps has not been previously reported.

choice from comparatively small libraries. The pro-It is of interest to contrast the advantages and cedure should be useful for the preparation of scFvs disadvantages of our procedure with those previously against other conformation-sensitive cell surface reported. One frequently employed method is the antigens, for which cloned cDNA is available. panning of complex non-immune (naive) or synthetic

The procedure was not designed as a general libraries (Griffiths et al., 1994; Vaughan et al., 1996;  procedure to isolate novel scFvs against a large Knappik et al., 2000) . The advantage of this pro-variety of previously unrecognized cell surface ancedure is that one universal library can be used to tigens. Existing procedures involving subtractive screen for scFvs reacting with many different an-whole cell-panning without the labor-intensive retigens (Nissim et al., 1994) . Synthetic libraries do finements added here may be sufficient or superior not necessarily represent the naive repertoire, but for this purpose. In addition, the scFvs isolated here often also contain high affinity scFvs because gene-are murine scFvs, with known disadvantages when ration of these libraries includes hypermutation of used in human clinical applications. Although some the CDR3 region. Such libraries have been expanded of these disadvantages can be relieved by chimerizto considerable complexity and libraries containing ing or humanizing the murine scFvs, it may still be 9 1 1 up to 10 -10 different scFvs have been reported preferable for a majority of clinical applications to (Vaughan et al., 1996; Knappik et al., 2000) . One generate human scFvs from the beginning. Finally, disadvantage of this approach is that the libraries, in we do not wish to detract from the major advantage spite of hypermutated CDR3 regions, have not been of large combinatorial or synthetic naive human isolated from animals undergoing a secondary or scFv-libraries. Clearly, their major advantage is, that tertiary immune response against a specific antigen. from one library human scFvs against many different Although high affinity scFvs have been isolated antigens can be isolated without the need for prior using this approach, the probability of finding scFvs immunization. For a majority of general applications against a specific antigen is reduced, entailing the these existing procedures may be equal or superior to 9 1 1 need to screen a large library of 10 -10 different the procedure reported here. The strength of the clones (Perelson, 1989) . Another disadvantage is that novel procedure specifically lies in the generation of the isolation of specific binders to a defined antigen recombinant scFvs against conformation sensitive on the cell surface may be compromised using non-antigens, such as CD13, for which cDNAs are immune or synthetic libraries, because it has been available, but for which no recombinant scFvs have reported that non-immune antibody libraries contain previously been obtained, that recognize the antigen binders to dominant epitopes (Noronha et al., 1998;  in its native state on intact cells. In this case, the Hoogenboom et al., 1999) . The binders to dominant advantage of our procedure is, that it avoids the need epitopes may be preferentially amplified during for an extensive biochemical purification of the screening and decrease the probability of isolating antigen, which may destroy the conformation-sensithe desired binders to specific, predetermined an-tive epitope. This epitope may only exist when the tigens, especially when screening is performed using antigen is present in its native conformation on the complex antigens, such as whole cells. In our cell surface and may disappear when it is purified away from its nearest neighbor membrane com-active agents against HL60 myeloid leukemia cells ponents. For this narrow segment of applications, the (Xu and Scheinberg, 1995) . While it is clear that procedure reported here presents significant progress, scFvs of murine origin represent problems in human as demonstrated for the example of CD13, for which clinical applications, they may still be clinically no scFvs had previously been isolated, that recog-useful, as demonstrated by antibodies against human nized the antigen in its native state. CD20, that have recently been approved as human The two scFvs reported here reacted with the therapeutical agents and that were derived from antigen in its native state on living leukemia cells. murine monoclonal antibodies. In our perspective, Their affinities were high in comparison with pub-these scFvs and the corresponding cDNAs may lished values (Griffiths et al., 1994) . Antibodies with become valuable for the design of novel therapeutic affinities in the nanomolar range or higher are high strategies against t(4;11) and t(11;19) leukemia cells. affinity antibodies (Vaughan et al., 1996) . However, caution must be exerted when comparing the numbers obtained here with previously published figures, Acknowledgements since different measurement procedures were utilized. In our case, soluble scFvs were used that were This work was supported by research grants from monovalent and association and dissociation rate the Deutsche Forschungsgemeinschaft (DFG; SFB constants were measured by surface plasmon reso-473-B6) and grants 96.047.1 from the Wilhelm nance. In this method, one of the two binding Sander Foundation to GHF and SJZ. We thank Prof. partners was immobilized (in the solid phase); the M. Gramatzki for providing the CEM cell line; Prof. other was freely mobile in the fluid phase. ReactionsÄ . Pluckthun for the scFv vectors; Dr. L. Shapiro for with immobilized partners have different kinetic the human CD13 cDNA clone; and Dr. G. Winter for parameters compared with reactions where bothË . coli TG1. J. Dorrie, A. Birkmann and Dr. R. interacting partners remain in the fluid phase (Nieba Slany are acknowledged for valuable discussions and et al., 1996) . Furthermore, many published affinity Th. Lange for administrative assistance. constants were obtained by measurement procedures in which both the antibody and ligand were in fluid phase. Therefore, the numerical affinity values reported here are not strictly comparable to previously

References published values and are not intended to represent absolute values. They must be considered approxi-Behm, F.G., Raimondi, S.C., Frestedt, J.L., Liu, Q., Crist, W.M., mate values. With this proviso they indicate that Downing, J.R., Rivera, G.K., Kersey, J.H., Pui, C.H., 1996. scFvs of high affinity have been isolated by our Rearrangement of the MLL gene confers a poor prognosis in childhood acute lymphoblastic leukemia, regardless of presentprocedure.

ing age. Blood 87, 2870-2877.

The antibodies against hCD13 isolated here may Cai, X., Garen, A., 1995. Anti-melanoma antibodies from become useful for purposes similar to other monomelanoma patients immunized with genetically modified auclonal anti-hCD13 antibodies that have been previtologous tumor cells: selection of specific antibodies from ously described by others. These purposes include single-chain Fv fusion phage libraries. Proc. Natl. Acad. Sci. USA 92, 6537-6541.

human cell-and tissue-typing (Goyert, 1997) 

Determination of active single chain Natl. Acad. Sci. USA 95, 669-674. antibody concentrations in crude periplasmic fractions

Purification, characterization and biotinylareceptor for the entero-pathogenic coronavirus TGEV. Nature tion of single chain antibodies

Human melanoma invasion and metastasis enhanceand CDRs randomized with trinucleotides

Neutralization of human functional antibody variable domains from hybridomas and cytomegalovirus by human CD13-specific antibodies

Cleavage of structural proteins during the T

cytomegalovirus (CMV) infection exclusively in CD13-posi

The acute lymphoblastic individuals: implications for prevention of CMV transmission

Human myeloid plasma membrane glycoprotein CD13

(gp150) is identical to aminopeptidase N

Antibodies -A Laboratory Manual. blastic leukemia after in vitro stimulation

Cytogenetic studies of action of human placental aminopeptidase N in feto-placental infant acute lymphoblastic leukemia: poor prognosis of infants unit

determined from binding kinetics

nonaccessible epitopes on cell-surface receptors revealed by Midgley

Limited diversity of human scFv fragments isolated by Huls

Loyson, human melanoma cells

A recombinant, fully and expression of aminopeptidase N

Aminopeptidase N is a receptor for tumor-homing peptides and a target for inhibiting angio-Soderberg

mediates human cytomegalovirus infection

Development of thyroid-specific human autoantibodies displayed on the surface human T and natural killer cells

Human acute leukemia cell line with the t(4;11)

chromosomal rearrangement exhibits B lineage and monocytic Carter, P., 1999. Identification of a human anti-CD55 singlecharacteristics

CD13 -not just a sheets: procedure and some applications

Role of aminopeptidase N (CD13) in tumor-Isolation of human colorectal tumor reactive antibodies using cell invasion and extracellular matrix degradation

Human antibodies with sub-nanomolar Press

Developpeptidase N gene in myeloid and intestinal epithelial cells. J. ment of a hyperimmune anti-MUC-1 single chain antibody

Hematopoietic differentiation Clin

Elimination of human leukemia key!

using phage display and magnetically-activated cell sorting