Colony assays for antibody fragments expressed in bacteria

197

Journal of Immunologwal Methods, 139 (1991) 197-205

© 1991 ElsexaerScience Pubhshers B.V. 0022-1759/91/$03.50 ADONIS 0022175900175M JIM 05929

Colony assays for antibody fragments expressed in bacteria M a r t i n L. D r e h e r 1, E r m a n n o G h e r a r d i 1, A r n e S k e r r a 2 a n d C r s a r M i l s t e i n 1 1 Laboratory of Molecular Biology, Hdls Road, Cambrtdge CB2 2QH, U.K., and 2 Max Planck 1nstttute for Btophystcs, Hemrwh-Hoffmann-Str 7, 6000 Frankfurt am Mare 71, F R G

(Received 31 December 1990, revised received 5 February 1991, accepted 5 February 1991)

This paper describes procedures for the detection and selection of bacterial colonies expressing antibody fragments of desired antigen specificity. Fab and Fv fragments are detected in a filter assay in which bacterial colonies are grown on a master filter in contact with a second, antigen-coated filter. Ab fragments diffusing onto the second filter bind antigen directly and specifically and are detected with a monoclonal antibody directed against a m y c - t a g sequence fused to the carboxy-terminal end of the hght chain or heavy chain (dtrect assay). Single-chain Fv (scFv) in which the V H and V L sequences are joined by a short linker peptide are detected by a modified procedure in which scFv are immobilized on filters coated with the a n t i - m y c - t a g sequence and subsequently detected by specific binding to radiolabelled antigen ( m d t r e c t assay). A single positive bacterial colony expressing antigen-specific Fv (or scFv) can be recovered among at least 10,000 negative colonies using the procedures described. The direct assay has been successfully used to discriminate Fv fragments which express point mutations known to increase the binding affinity of antibodies to the hapten 2-phenyl-oxazolone. The procedures described may thus prove generally useful for the selection of antigen-specific clones expressed in bacteria a n d / o r higher-affinity variants of such antibodies. Key words Colony assay; Bacterial colony; Antibody fragment

Introduction

The demonstration that functional antibody fragments can be produced in Eschertchta cob opened up new approaches in antibody research and protein engineering (Better et al., 1988; Skerra and Pliickthun, 1988). The efficient cloning and manipulation of antibody genes followed by ex-

Correspondence to. M.L. Dreher, Laboratory of Molecular Btology, Hdls Road, Cambndge CB2 2QH, U.K. Abbrevlanons: BSA, bovine serum albutmn; EDTA, ethylenedianunetetraacetlc acid, FCS, fetal calf serum, IPTG, lSOpropyl-fl-D-thlogalactopyranoslde, OxBSA, oxazolone-conjugated BSA, PBS, phosphate-buffered saline; PVDF, polyvlnyhdlenedlfluoride,

presslon in bacteria greatly enhances the potential of site-directed mutagenesls of anugen binding sites, whereas the immortalisation in vitro of large immunoglobulin repertoires from immunised animals permits the isolation of recombinant antibody fragments of desired specificities (Huse et al., 1989; Ward et al., 1989; Mullinax et a1.,1990). Our ultimate goal is the generation of high affinity antibodies, not by reproducing genetic material taken from immunized animals but by imitating in vitro the natural process of affinity maturation (Milstein, 1990; Winter and Milstein 1991). Cloning and expression of antibodies in bacteria is an excellent way to generate diversity but is constrained by the efficiency of selection of bacterial clones producing the desired antibody fragment.

198 Here we describe two methods for the rapid identification of single bacterial colonies expressing antigen specific antibody fragments. In the first procedure Fv and Fab are captured on antigencoated membranes and detected with anti-antibody reagents (direct assay); in the second procedure single chain Fv (scFv) are captured on membranes coated with a suitable antibody and antigen-specific clones are detected by binding to radiolabelled antigen. The methods are based on previously developed procedures for the identification of antibody-secreting hybridoma colonies (Gherardi et al., 1990).

Materials and methods

Plasmlds for the expresston of antibody fragments in E. coh Fv and Fab antibody fragments were produced in the TG1 strain of E. cob (Sambrook et al., 1989) harbouring expression plasmids pBQ10Fvmyc and pBQlOFab-myc, respectively. These constructs contain the light chain (V,) and heavy chain (VH) variable region genes of the mouse anti-oxazolone monoclonal antibody NQ10/12.5, the three-dimensional structure of which has been solved recently (Alzan et al., 1990). pBQlOFv-myc and pBQlOFab-myc were derived by replacing the anti-lysozyme antibody D1.3 variable region genes of pSW1/D1.3Fvmyc and pSW1/Fabmyc plasraids, respectively, with the the NQ10/12.5 variable region genes (Ward et al., 1989). The Fab constructs carry the human CH1 and C~ constant regions in which the carboxy-terminal disulphide bond forming cysteine residues have been removed (S. Ward and G. Winter, unpublished resuits). Single chain Fv (scFv) were produced in the E. colt strain N 4 8 3 0 - 1 h a r b o u r i n g plasmids pJMSNscFvNQll/7.22myc and pJMSNscFvD1.3myc which encode the single chain heavy and light chain variable regions (scFv) of the antioxazolone antibody N Q l l / 7 . 2 2 and of the antilysozyme antibody D1.3. In these constructs (kindly provided by Dr. A. Griffiths) the carboxy terminus of the heavy and amino termini of light chains are joined by a short peptide (Gly-Gly-Gly-

Gly-Ser) serving as a flexable linker (Huston et al., 1988) and are expressed as a single polypeptide. To facilitate immunological detection of expressed antibody fragments the gene segments corresponding to the carboxy termini of the light chain (in the Fv and scFv constructs) and the CH1 constant region of the heavy chain (in the Fab construct) were fused with a short D N A sequence coding for a peptide derived from the c-myc oncogene (myc-tag) (Ward et al., 1989) which is recognized by the monoclonal antibody 9El0.

Membrane coatmg wtth anttgens or monoclonal antibody 9E10 For antigen coating 82 mm diameter Hybond-N (Amersham, RPN82N) or nitrocellulose membranes (Schleicher and Schuell, cat. no. 401 116) were incubated in 10 ml PBS (0.125 M NaC1, 0.017 M Na2PO4, 0.017 M NaH2PO4, pH 7.4) containing 100 # g / m l of OxBSA, lysozyme (Sigma), or BSA (Sigroa, fraction V) for 6 h at room temperature with gentle orbital shaking. OxBSA (with an average substitution ratio of 14/1) was prepared as described by MS_kel~i et al. (1987). After coating, the membranes were washed extensively with PBS and residual binding sites blocked for 6 h at room temperature in PBS containing 3% ( w / v ) BSA and 0.5% ( v / v ) Tween 20. The coated membranes were then washed several times in PBS, soaked in a PBS solution containing 1 mM I P T G and used as described below and in Fig. 1. For coating with monoclonal antibody 9El0 (anu-myc-tag) Immobilon-P membranes (Millipore, Cat. no. P-15552) were incubated for 5-6 h at room temperature in 3 ml of a solution of purified 9El0 antibody (130 /~g/ml in PBS) and blocked for 1 h at 37 ° C in 10% FCS in PBS. Growth and mductton of bacterial colomes The procedure used for the growth and induction of E. cob transformants expressing Fab and Fv fragments is shown schematically in Fig. 1. The method is based on the use of two membranes, one for growth and induction of bacteria, the other for capture and detecUon of the antibody fragments (Skerra et al., 1990). Briefly, the bacteria were grown to small visible colonies on a Durapore filter (Millipore, GVWP09050) placed on top

199

of a T Y E agar plate supplemented with 100 # g / m l ampicillin and 1% ( w / v ) glucose (1 litre T Y E agar contains 15 g agar, 10 g bacto tryptone, 5 g yeast extract). For induction the Durapore filter was removed from the agar dish and transferred on the antigen-coated membrane. The two filters were transferred to a new T Y E agar dish containing 1 m M I P T G and 100 # g / m l ampicillin and cultured for about 16 h at room temperature. During this time the antibody fragments are released by the bacteria, diffuse through the Durapore filter and are immobilized on the antigen coated membrane. The filter with the bacteria is then removed and placed on a fresh T Y E agar plate containing 100 /xg/ml ampicillin whereas the antigen-coated m e m b r a n e is removed for immunochemical detection of the captured antibody fragments.

Bacterial colonies expressing scFv expressed under the control of the XPL promoter were cultured for 16-20 h at 30°C on ampicillin plates. Colony lifts were prepared on Durapore filters and these were transferred on I m m o b i l o n P membranes coated with antigen or antibody 9E10. The replica cultures were then incubated at 42°C for 30 min and at 37°C for 4 - 6 h before I m m o b i l o n P filters were processed as described below and in the legend to Fig. 3.

Detectton of immobthzed Fv, Fab and scFv fragments Filters contaimng Fv and Fab fragments bound to antigen-coated membranes were incubated for 60 min at room temperature with - 2 ~tg/ml of the murine anti-myc-peptide monoclonal antibody

Bacterial Colony Filter (0.2211mDurapore PVDF)

Coated Membrane a) Antigen (for "direct assay') e.g. OxBSA b) capturing Antibody (for "indirect assay) e.g. 9El0 anti-tag

TYE Agar Plate l~lmM IP'rG, 100gg/ml Ampicillin)

Filter-Sandwich [16 hours at room-temperature)

O

CSSS Detection a) via Anti-Fragment Antibody ("for "direct assay") b) via Antigen (for "indirect assay")

TYE Agar Plate [10011g/rnlAmpicillin)

Fig. 1. Schematic representaUon of the procedure used for the expression and detection of bacterial colomes expressing Fv or Fab fragmets under the control of the lacZ promotor. A shghtly different procedure was used for scFv expressed under the control of the XPL promotor (see materials and methods secUon for detatls)

200

b

C [3

Fig. 2 a mtrocellulose fdter sections that have been coated with OxBSA (A and B) and BSA (C and D). Falter-sections were covered with a Durapore filter carrying TG1 bacterta harbounng pBQlOFv-myc (A,C) or pBQlOFab-myc (B,D) b nylon membranes were coated with OxBSA (A and F), or lysozyme (C and D) or BSA (B and E) Sections A, B and C were then placed m contact with a Durapore filter which carried bacterial colomes expressing the anti-oxazolone NQ10/12.SFv and sections D, E and F were placed in contact with a second Durapore filter which camed bacterial colomes expressing the anU-lysozyme D1 3Fv. Fv and Fab fragments lmmobdlzed on the assay membranes were then vlsuahzed as described m the materials and methods section.

Fig 3. Bacterial colomes expressing antl-oxazolone ( N Q l l / 7 22) and antl-lysozyme (D1 3) scFv were incubated on Immobdon P membranes coated with either Ox~4BSA (50 pmol c a m e r / c m 2) (left panel) or antl-myc-tag antibody 9E10 (right panel). Single chain Fv bound to antigen-coated membranes were revealed by incubation with 9E10 followed by HRP-conjugated rabbit anu-mouse lmmunoglobuhn (Dako, P260) and dlarmnobenzldlne/hydrogen peroxade enzyme substrate (Gherardi et al, 1990). Single cham Fv lmmobxhzed with 9E10 were revealed with 125I-OxI4BSA. Ttus was prepared using Iodobeads (Pterce) and used at 1 × 106 c p m / m l ( - 2 # g / m l ) After incubation with 125I-Ox14BSA fdters were washed 3 × 5 nun in 1 g / l Tween 20 in PBS, dried and exposed to Fuji X ray films The film was exposed for 36 h at - 70°C

201

9E10 (Evan et al, 1985.) in 5 ml PBS. After three washings with PBS the filters were then incubated for 60 min at room temperature with alkaline phosphatase (AP) conjugated rabbit anti-mouse IgG (Dako, D314) and washed extensively with PBS. AP was visualized by incubation with 5 ml of substrate buffer (100 mM Tris/HC1 p H 8.8; 100 mM NaC1, 5 mM MgC12) containing 17 /~1 of BCIP solution (50 m g / m l 5-bromo-4-chloro-3-indolyl-phosphate 4-toluidine salt, Boehringer Mannheim, in dimethylformamide) and 33 /~1 of NBT solution (50 m g / m l nitro blue tetrazolium, Sigma, in 70% ( v / v ) dimethylformamide). The chromogenic reaction was stopped after about 5 nun by washing the filters in PBS containing 2 mM EDTA followed by air-drying. Single-chain Fv fragments xmmobilized on antigen-coated membranes were detected essentially as described above for Fv with the minor modifications detailed in the legend to Fig. 3. Single chain Fv fragments immobilized with the anti myc-tag antibody 9El0 were detected with ~25IOxBSA as detaded in the legend to Fig. 3.

Results

Dtrect assay for Fo and Fab In this procedure Fv and Fab fragments were captured and bound directly to membranes coated with antigen (OxBSA or lysozyme). BSA-coated membranes served as control. The immobilized Fv and Fab fragments were detected with the antimyc-tag mouse monoclonal antibody 9El0 followed by alkaline phosphatase conjugated antimouse IgG. Fig. 2a shows that the bacterial colonies producing anti-oxazolone Fv and Fab fragments, respectively, gave excellent s~gnals with filters coated with OxBSA (A and B), but no signals with BSA (C and D). Antibody fragments of different specificities could be distinguished by capturing them with their respective antigen. In Fig. 2b we compared the anti-oxazolone Fv fragment with the antilysozyme Fv fragment. The OxBSA (A and F) and lysozyme (C and D) coated membrane sections gave strong signals only if they were exposed to bacterm harbouring the appropriate expression

plasmids. We consistently observed higher background staining for lysozyme coated membranes. This may be due to the more hydrophoblc nature of lysozyme compared to OxBSA. The BSA coated filters (B and E) did not show any signal with either Fv fragment. However, when the same experiments were carried out with the scFv the results illustrated in Fig. 3 (left panel) were obtained. Anti-oxazolone ( N Q l l / 7 . 2 2 ) scFv readily bound to OxBSA coated Immobilon P membranes, but so did the control antl-lysozyme scFv (D1.3). Identical results were obtained with nitrocellulose membranes (data not shown). Such high non-specific binding was surprising since the binding of scFv secreted in liquid culture to antigen-coated ELISA plates was specific, namely the anti-oxazolone scFv bound to the oxazolone-coated plate while anti-lysozyme scFv bound to the lysozyme-coated plate but not vice-versa (results not shown). In order to assess how efficient the procedure was for the detection of positive bacterial colomes outnumbered by negative ones we mixed hquid cultures of bacteria harbouring pBQlOFv-myc (anti-oxazolone Fv fragment) with bacteria harbourmg the standard cloning vector pUC18 (which confers ampicilhn resistance) at a ratio of 1 : 100, respectively. Fig. 4 shows the results of a direct assay performed with OxBSA coated membranes. Out of approximately 4500 bacterial colomes (filter A), 38 were identified on the antigen-coated membrane as producing the anti-oxazolone Fv fragment (filter B). Filter C carried approximately 13,500 colonies out of which about 130 clear lmmunoreactive colonies could be detected (filter D). The results show that the positive colomes could easily be identified (and located) among a large number of non-producing bacterial colonies. However, the decreasing colony size in regions of denser growth, particularly noticeable on filter C, did result in smaller signals (compare membranes B and D). The sensitivity of the method is therefore limited by the colony size. However, we are confident that we can detect a single Fv producing bacterial clone m a 'negative' population of at least 10,000 colomes in a 9 cm agar dish. Similar results were obtained with bacteria expressing scFv's detected with the indirect procedure described below (data not shown).

202

Colonies plated:

4500

(Fv colonies 1:100)

Colonies plated- 13500

Positive Clones: 38

Positive Clones:

130

(Fv colonies: 1:100 ) Fig. 4. TG1 bacteria harbouring pBQlOFv-myc were rmxed at approximately 1 : 100 ratio with TG1 transformed with the pUC18 cloning vector (winch also confers amplcllhn resistance but does not produce Fv fragment). Filters A and C show the total n u m b e r of bacteria colonies on the Durapore filters. Membranes B and D show the bacterial colomes (from filters A and C respectively) which secrete antl-oxazolone Fv

203

Ind:rect assay for scFv As shown in Fig. 3, scFv gave a strong background in the direct assay making it useless for the detection of bacterial colonies producing antigen specific scFv fragments. On the contrary specific binding of anti-oxazolone (but not anti-lysozyme) scFv to OxBSA was observed when scFv were first captured on membranes coated with the anti myctag 9El0 and subsequently incubated with 125IOxBSA (indirect assay) (Fig. 3, right panel).

Signal intensity and antibody affimty T o test whether the method was capable of detecting differences in the affinity of the antibody fragments produced, we compared colonies transfected with plasmid constructs which specify the characteristic mutations of the light-chain which reproduce critical stages of the affinity maturation of the antibody response to oxazolone

C.

i

Fig. 5. TG1 bacteria harbounng pBQlOFv-myc (A) and pBQVHlO,V,-Oxl-myc (E) were spread on a Durapore fdter and secreted Fv were Immobilized on OxBSA-coated membranes. The colomes on sections B, C and D harbour pBQVHlO,VK-Oxl-myc plasnnd with the follovang mutations m the V,-Ox1 hght chain: Serine-31 mutated to arginine and lusudine-34 to asparagine (B); senne-31 to arginine and tyrosine-36 to phenylalanine (C); htstldme-34 mutated to asparagine (D). Sections B, C and D contain 2, 3 and 4 (respectively) pen marks wtuch appear as the most prormnent spots

(Berek and Milstein, 1987). The results, using the direct assay on OxBSA-coated membranes, are shown in Fig. 5. The germ-line gene sequence of the light chain Vk-OX~, in combination with the mature NQ10/12.5 heavy chain did not give any signal. Bacterial colonies expressing the light chain where His34 has been substituted by Asn showed a weak, but clearly positive signal and so did the introduction of a double-mutation (Ser31 for Arg and Tyr36 for Phe). A considerably stronger signal was observed when the N Q 1 0 / 1 2 . 5 V L segment, which carries all three mutations together, substituted the germ-line gene V,-Ox 1.

Discussion There is considerable interest in simple methods for production and selection of antibody fragments in bacteria. Two general strategies are under scrutiny. One aims at the expression of the antibody binding site on the surface of bacteria or phage. Success with the latter is most encouraging (McCafferty et al., 1990). The other aims at the detection of antibody fragments released by bacteria either transformed with recombinant plasmids (Better et al., 1988; Skerra and Pliickthun, 1988; Ward et al., 1989) or infected with recombinant X phages (Huse et al., 1989). The detection of antibody fragments released by bacteria is best performed in situ. A phage plaque assay for Fab fragments has been used (Huse et al., 1989; Caton et al. 1990) but previous detection of bacterial colonies secreting active fragments has not been reported. The method we report differs from the plaque assay m several other respects. In particular we use antigen- or antibody-coated filters which most likely increase the sensitivity of the assay and the signal to noise ratio. The I P T G induction and transfer of fragments to membranes also differ in several respects and the growth of bacterial colonies on filters (Fig. 1) facilitates the handling and the subsequent identification of the positive clones. In earlier experiments we transferred the filter carrying the bacterial colonies onto soaked filter paper and blotted the secreted Ab fragments to a second (antigen-coated) m e m b r a n e placed on top of the colonies. This procedure is slightly more corn-

204

plicated than the protocol illustrated in Fig. 1 but is equally effective. Both direct and ;redirect procedures have been successfully employed to capture and detect antibody fragments released by bacteria. However, m at least one case (for the detectmn of scFv) only one method (the lndlrect) was specific. The direct assay gave high non-specific binding (Fig. 3, left panel) and we were not able to solve the problem by changing the assay membrane (even though, in general, we have found testing different membranes to be cost effective). We are unable to explain the reason why the direct method gives such backgrounds, beanng in mind that they are not observed in immunoradlometric assays using the same fragments and combination of reagents. We have used the direct assay in an attempt to discriminate between stages of the affinity maturatmn of the antibody response to the hapten 2-phenyl-oxazolone and found that the assay was capable of such discrimination (Fig. 5). Obviously, to detect further increases in affimty beyond the values giving signals of maximum intensity will need modifications to the standard protocol. For instance, one may reduce the antigen density (Gherardi et al., 1990) or decrease the time allowed for the chromogenic reaction. Indeed we have observed that the strong signals appear much earher in the colour reaction. Other possibilities include competition with soluble antigen or antibody However the important point which emerges from Fig. 5 is that, in spite of presumed variations in the level of expression between individual colonies, the signal distlngmshes mutations which parallel the affinity maturation. This is important in our efforts to mimic in vitro the animal strategy for the derivation of high-affinity antibodies.

Acknowledgements The authors wtsh to thank S. Ward for providing plasmids pSW1/D1.3Fvmyc and p S W 1 / D1.3Fabmyc, A. Griffiths for providing plasmids pJMSNscFvNQ11/7.22myc and p J M S N s c FvD1.3myc and R. Pannell for large-scale 9E10 hybridoma cultures. M.L.D is a recipient of a pre-docotoral stipend from Stiftung Stipendien Fonds des Verbandes der Chenuschen Industrle

(Frankfurt, F.R.G.), A.S. was supported by a post-doctoral fellowship from the Sonderprogramm Gentechnologie (DAAD, F.R.G.).

References Alzarl, P.M, Splnelh, S, Marmzza, R A., Boulot, G , Poljak, R J., Jarvls, J.M and Mllstem, C (1990) Three-dimensional structure deterrmnatmn of an anti-2-phenyloxazolone antibody. the role of somatic mutation and heavy/hght chain pamng in the maturation of an immune response. Eur J Mol. Biol. Org. J 9, 3807 Berek, C and Mdsteln, C. (1987) Mutation drift and repertoire shaft in the maturauon of the immune response. Immunol. Rev 96, 23 Berek, C, Gnffiths, G.M and Mflstem, C. (1985) Molecular events dunng maturation of the immune response to oxazolone. Nature 316, 412 Better, M., Changh, C.P, Robinson, R R., and qorwltz, A H (1988) Eschenchaa coh secretion of an active chimeric antibody fragment Soence 240, 1041. Caton, A.J and Koprowskl, H (1990) Influenza virus hemagghaUmn-spec~fic anubodtes isolated from a combinatorial expression library are closely related to the immune response of the donor Proc. Natl Acad. SCL U S.A. 87, 6450. Evan, G I, Lewis, G.K., Ramsay, G. and Bishop, J.M (1985) Isolation of monoclonal antibodies specific for human c-myc proto-oncogene product. Mol Cell. Biol. 5, 3610. Gherardl, E., Pannell, R. and Mdstem, C. (1990) A single-step procedure for clomng and selection of anubody-secretmg hybrldomas. J. Immunol. Methods 126, 61. Huse, W D , Sastry, L, Iverson, S A , Kang, A.S., Altlng-Mees, M., Burton, D R., Benkovlc, S.J. and Lerner, R. (1989) Generation of a large combinatorial hbrary of the ~mmunoglohuhn repertoire m phage lambda Science 246, 1275 Huston, J S, Levlnson, D , Mudgett-Hunter, M., Tai, M.-S., Novotny, J., Margohes, M N., Radge, R.J., Bruccolen, R E , Haber, E, Crea, R and Oppermann, H (1988) Protein engmeenng of antibody binding sites: Recovery of specific activity m an anti-digoxm single-chain Fv analogue produced in Eschertchta colt Proc Natl. Acad Sct U.S.A 85, 5879 Makela, O , Kaartmen, M., Pelkonen, J.L T. and Karjalalnen, K. (1978) Inheritance of antibody speoficlty V. Antl-2phenyloxazolone in the mouse. J. Exp Med 148, 1644. McCafferty, J., Gnfflths, A.D., Winter, G. and Chlswell, D.J. (1990) Phage anUbodles- filamentous phage displaying antibody variable dommns. Nature 348, 552 Mllstem,, C (1990) AnUbodles: a paradigm for the biology of molecular recogmtlon (The Crooman Lecture, 1989) Proc Roy. Soc London 239, 1 Mulhnax, R.L., Gross, E.A., Amberg, J.R., Hay, B.N, Hogrefe, H.H, Kubltz, M.M., Greener, A , Altlng-Mess, M, Ardourel, D , Short, J.M, Sorge, J A. and Shopes, B. (1990) Identlficatmn of human antibody fragment clones specific

205 for tetanus toxold in a bacteriophage X lmmunoexpresslon library Prec. Natl. Acad Sci U S.A 87, 8095 Sambrook, J , Fntsch, E.F and Mamatis, T (1989) Molecular Clomng, A Laboratory Manual, 2nd edn Cold Spnng Harbour Laboratory Press, Cold Spring Harbor, NY Skerra, A and Pliickthun, A (1988) Assembly of a functional tmmunoglobuhn Fv fragment m Escherwhta colt Science 240, 1038

Skerra, A , Dreher, M.L and Winter, G (1991) submitted Ward, E S., Gussow, D., Gnffiths, A D Jones, P.T and Wmter, G. (1989) Bmdmg actlvatms of a repertotre of single lmmunoglobuhn variable domains secreted from Escherwhla colt. Nature 341, 544. Winter, G and Mdstem, C (1991) Man made antlbodtes. Nature, 349, 293