Proc. Nat!. Acad. Sci. USAVol. 88, pp. 7978-7982, September 1991Biochemistry

Assembly of combinatorial antibody libraries on phage surfaces:The gene III site

(combinatorial immunoglobulin repertoire/filamentous phage/surface expression/human antibodies/catalytic antibodies)

CARLOS F. BARBAS III*, ANGRAY S. KANG*, RICHARD A. LERNER*, AND STEPHEN J. BENKOVICt*Departments of Molecular Biology and Chemistry, The Scripps Research Institute, 10666 North Torrey Pines Road, La Jolla, CA 92037; and tDepartment ofChemistry, Pennsylvania State University, University Park, PA 16802

Contributed by Stephen J. Benkovic, June 21, 1991

ABSTRACT A phagemid system was developed for themonovalent display of combinatorial antibody Fab libraries onthe surface of filamentous phage M13. Fab fragments werefused to the carboxyl-terminal domain of the gene HI protein.Phage displaying Fab fragments on their surface, or Phabs,were enriched by 103- to 105-fold on antigen-coated surfacesover nonspecific phage. The method may replace currentantibody cloning techniques.

Our combinatorial approach (1, 2) provides a means ofcapturing the vast diversity of the immunological repertoire(1-5). The approach relies on the ability to clone antibodyheavy- and light-chain fragments independently and ran-domly recombine them in a system that allows the specificityof binding to be probed. Previously, we used a A phagesystem that allowed probing of plaque-lifts for the identifi-cation of desired clones. This system is limited by the size ofthe combinatorial library that may be examined (-106 mem-bers).An interesting approach to accessing larger libraries in-

volves their expression on the surface of filamentous phage.The display of libraries of small peptides on the surface offilamentous phage has proven to be a powerful approach forselecting ligands of defined specificity (6-8). Phage displaylibraries are rapidly being extended to whole proteins. Twomonomeric proteins, a single-chain antibody (9) and humangrowth hormone (10), have been successfully expressed asfusions with the gene III (gIII) product coat protein III(cpIII). We have recently reported the expression and as-sembly of considerably larger and more complex (-50-kDaheterodimers) proteins, antibody Fab fragments, as fusionswith the gene VIII (gVIII) product coat protein VIII (cpVIII)(11).The ability to rapidly sort large combinatorial Fab libraries

is particularly important for the development of catalyticantibodies (12). Here we report a strategy based on gIIIfusions, complementing our gVIII approach, for the con-struction, selection, and production of high-affinity antigen-specific Fabs from combinatorial antibody libraries. Thecombinatorial antibody technology reported here may re-place current hybridoma methods for the isolation of mono-clonal antibodies.

MATERIALS AND METHODSVector Construction. The pelB leader sequences and clon-

ing sites for the heavy-chain fragment and light chain werederived from phagemids excised from A Hc2 and A Lc2 Avectors as described (2). The sequences were modified toremove a redundant Sac I site from Hc2 phagemid and a SpeI site from the Lc2 phagemid. The combinatorial phagemid

vector pComb was constructed from these two modifiedphagemids by restricting each with Sca I and EcoRI andcombining them in a ligation reaction. Recombinants werescreened for the presence of two Not I sites yielding thecombinatorial vector pComb. The tether sequence GGGGSand gIIl fragment from Spe I to Nhe I were the product ofPCR of M13mpl8 (13) using the oligonucleotides 5'-GAGACGACTAGTGGTGGCGGTGGCTCTCCAT-TCGTTTGTGAATATCAA-3' and 5'-TTACTAGCTAG-CATAATAACGGAATACCCAAAAGAACTGG-3'.The lacZ promoter, operator, and Cap-binding site con-

trolling light-chain expression were the product ofPCR withM13mpl8 using oligonucleotides 5'-TATGCTAGCTAG-TAACACGACAGGTTTCCCGACTGG-3' and 5'-AGCT-GTTGAATTCGTGAAATTGTTATCCGCT-3'. The PCRfragments encoding the gIII fragment and lacZ promoterwere spliced by PCR overlap extension (14). The resultingproduct was digested with Spe I and EcoRI and ligated intothe corresponding sites of pComb to yield pComb 3'. Finally,pComb 3' was digested with Xho I and Spe I and ligated withthe corresponding 51-base-pair (bp) stuffer from pBluescript(15) (Stratagene) to yield pComb 3 (Fig. 1), an ampicillin-resistant phagemid. The corresponding chloramphenicol-resistant phagemid, pCBComb 3, is a derivative of pBC(Stratagene).

Electron Microscopy. Tetanus toxoid was labeled withcolloidal gold, and microscopy was done as described (11).Phage Production and Single-Pass Enrichment Experiments.

Phagemids were transformed into Escherichia coli XL1-Bluecells and grown in super broth medium (SB; 30 g of tryptone,20 g of yeast extract, 10 g of Mops per liter, pH 7) at 370Csupplemented with tetracycline at 10 ,ug/ml and carbenicillinat 50 ,ug/ml or chloramphenicol at 30 ,ug/ml. Cultures weregrown to an OD6w of 0.4 and infected with VCSM13 helperphage (phage to cell ratio, 20:1) and grown an additional hour.After 1 hr kanamycin was added (70 pug/ml), and the culturewas incubated overnight. Phage were isolated from liquidculture by polyethylene glycol 8000 and NaCl precipitation asdescribed (8). Phage pellets were resuspended in phosphate-buffered saline (50 mM phosphate, pH 7.2/150 mM NaCI)and stored at -20° C.The following procedure is a modification of that originally

described by Parmley and Smith (16) and is known aspanning. In detail, a single well of a microtiter plate (Costar3690) was coated overnight at 4°C with 25 Al oftetanus toxoidat 2 mg/ml in 0.1 M bicarbonate, pH 8.6. The well waswashed once with water and blocked by filling the well withBlotto [5% (wt/vol) nonfat dry milk in phosphate-bufferedsaline] and incubating the plate at 37°C for 1 hr. Blockingsolution was shaken out, and 50 ,ul of clonally mixed phage

Abbreviations: cpIII and cpVIII, coat proteins III and VIII, respec-tively, derived from M13 phage; NPN, p-nitrophenyl phosphonami-date; Phabs, phage displaying antibody combining sites; gIlI andgVIII, gene III and VIII, respectively; cfu, colony-forming units.

7978

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement"in accordance with 18 U.S.C. §1734 solely to indicate this fact.

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

25, 2

020

Proc. Natl. Acad. Sci. USA 88 (1991) 7979

Surface display phagemid pComb 3

| 1) Nhe 1, Spe I digest1663 bp gill fragment

2) ligate

fWZ p t le AQVKL IHeav chain a a Z pro leader AE Ught chain to

Soluble Fab expressing phagemid

FIG. 1. The pComb 3 phagemid shown in surface-display and soluble-Fab-producing forms. The vector is designed for the cloning ofcombinatorial Fab libraries. The Xho I and Spe I sites are provided for cloning PCR-amplified heavy-chain Fd sequences. The Sac I and XbaI sites are provided for cloning PCR-amplified antibody light chains. Cloning sites are compatible with previously reported mouse and humanPCR primers (2, 4). The nucleotide sequences of the peWB leader sequences were recruited from the previously reported A HC2 and A LC2constructs (2) with reading frames maintained. The sequence of the phagemid vector (pBluescript), which includes Col El and Fl origins anda ,-lactamase gene, has been reported (15). Digestion of pComb 3, encoding a selected Fab, with Spe I and Nhe I allows for the removal ofthe gIII fragment. Because Spe I and Nhe I produce compatible cohesive ends, the digested vector may be religated yielding a phagemid thatproduces soluble Fab. The following additional restriction sites appear between the Xho I and Spe I sites of the empty vector as derived fromthe 51-base-pair (bp) stuffer fragment of pBluescript (for sequence information, see ref. 15) and allow for the construction of other gIII fusions:Sal I, Acc I, HincII, Cla I, HindIII, EcoRV, Pst I, Sma I, and BamHI.

[typically 1011 colony-forming units (cfu)] were added, andthe plate was incubated for an additional 2 hr at 370C. Phagewere removed, and the well was washed once with distilledwater. The well was washed 10 times with TBS/Tweensolution (50 mM TrisHCl, pH 7.5/150 mM NaCl/0.05%Tween 20) over a period of 1 hr at room temperature. The wellwas washed once more with distilled water, and adherentphage were eluted by adding 50 A.l of elution buffer (0.1 MHCl, adjusted to pH 2.2 with glycine, containing bovineserum albumin) at 1 mg/ml and incubation at room temper-ature for 10 min. The eluate was removed and neutralizedwith 3 ,ul of 2 M Tris base. The initial phage input ratio wasdetermined by titering on selective plates. The final phageoutput ratio was determined by infecting 1 ml of logarithmicphase XLI-Blue cells with the neutralized eluate for 15 minat room temperature and plating equal aliquots on selectivecarbenicillin and chloramphenicol plates.

Library Construction and Multiple Pannings. A humancombinatorial antitetanus toxoid Fab library was constructedin pComb 3 from the previously prepared antigen-stimulatedlibrary in phage A (4). The original pairing of chains wasmaintained by ligation of the Xho I/Xba I library fragmentsinto Xho I/Xba I-prepared pComb 3. The gIII fragment andthe promoter were added by removal of the Spe I/Sac Ifragment and addition of the corresponding fragment frompComb 3. After this ligation the DNA (2 ,ug) was ethanolprecipitated, resuspended in 10 ,ul of water, and transformedby electroporation into 300 ,ul of E. coli XLI-Blue. Aftertransformation, 3 ml of SOC medium (17) was added, and theculture was shaken at 220 rpm for 1 hr at 37°C, after which10 ml of SB medium containing carbenicillin at 20 ,ug/ml andtetracycline at 10 ,g/ml was added; the culture was thenshaken at 300 rpm for an additional hour. This culture wasadded to 100 ml of SB medium containing carbenicillin at 50,ug/ml and tetracycline at 10 ,ug/ml and shaken for 1 hr, after

which helper phage VCSM13 (1012 plaque-forming units)were added, and the culture was shaken for an additional 1-2hr. After this time, kanamycin at 70 ,ug/ml was added, and theculture was incubated at 37TC overnight. Phage were isolatedand panned in two wells, 50 ,ul in each, as described above.Eluted phage were used to infect 2 ml of fresh (OD600 = 1)XLI-Blue cells for 15 min at room temperature, after which10 ml of SB medium containing carbenicillin at 20 p.g/ml andtetracycline at 10 ug/ml was added; the culture was thenshaken for 1 hr at 370C. Further growth, phage preparation,and panning were repeated as described above.Colony Screening of Panned Library. Clones from each

round of panning were streaked on LB/carbenicillin platesand grown at 370C for 4 hr. Plates were overlaid withnitrocellulose filters soaked in 5 mM isopropyl 3-D-thiogalactopyranoside and incubated overnight at 30'C. Fil-ters were removed, incubated in a chloroform chamber for 15min, then transferred to 100 ml oflysozyme buffer containing50mM Tris (pH 8), 150mM NaCl, 5 mM MgCl2, 3 g ofbovineserum albumin, 40 mg of lysozyme, and 100 units of DNase,and rocked for 1 hr. Buffer was removed, fresh lysozymebuffer was added, and filters were rocked for an additionalhour after which adhering colonies were removed by gentlerubbing (18). Filters were added to 100 ml of phosphate-buffered saline/1% (wt/vol) bovine serum albumin and in-cubated with 1 nM tetanus toxoid-labeled alkaline phospha-tase and screened as described (4).

Preparation and Analysis of Soluble Fab. Phagemid DNAfrom positive clones was isolated and digested with Spe I andNhe I. The vector DNA was gel-purified and self-ligated.Transformation of XLI-Blue afforded the isolation of recom-binants lacking the gIII fragment. Clones were grown in SBmedium containing carbenicillin at 50 pug/ml and 20 mMMgCl2 at 370C until OD6w of 0.2 was achieved. Isopropyl-3-D-thiogalactopyranoside (1 mM) was added, and the cul-

Biochemistry: Barbas et al.

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

25, 2

020

Proc. Natl. Acad. Sci. USA 88 (1991)

ture was grown at 30'C overnight. Cells were pelleted, Fabwas prepared, and competitive ELISA was done as reported(4).

RESULTS AND DISCUSSIONPhage assembly proceeds via an extrusion-like process

through the bacterial membrane (19). gIII of filamentousphage encodes a 406-residue minor phage coat protein, cpIII,which is expressed before extrusion and which accumulateson the inner membrane facing into the periplasm of E. coli.Studies by Crissman and Smith (20) have led to the assign-ment of the two functional properties of cpIII, infectivity andnormal (nonpolyphage) morphogenesis, to roughly the firstand second half of the gene. The N-terminal domain of cpIIIbinds to the F' pili, allowing for infection of E. coli, whereasthe membrane-bound C-terminal domain, P198-S406, servesthe morphogenic role of capping the trailing end of thefilament according to the vectorial polymerization model (19,20).A phagemid vector, pComb 3 was constructed to fuse the

antibody Fd chain (comprising heavy-chain variable regionand heavy-chain constant region 1 domains) with the C-ter-minal domain of cpIII. A flexible five-amino acid tether(GGGGS), which lacks an ordered secondary structure (21),was juxtaposed between the expressed Fab and cpIII do-mains to minimize interaction. The Fd-cpIII fusion andlight-chain proteins were placed under control of separate lac

promoter/operator sequences and directed to the periplasmicspace by pelB leader sequences for functional assembly onthe membrane. Inclusion of the phage F1 intergenic region inthe vector allows for packaging of single-stranded phagemidwith the aid of helper phage. The use of helper phagesuperinfection leads to expression of two forms of cpIII.Consequently, normal phage morphogenesis is perturbed bycompetition between the Fab-cpIII fusion and the nativecpIII of the helper phage for incorporation into the virion.The resulting packaged phagemid carries native cpIII, whichis necessary for infection, and the encoded Fab fusionprotein, which is displayed for selection. A similar approachhas been reported for the production of hormone phage (10).Fusion at the C-terminal domain is necessitated by thephagemid approach because fusion with the infective N-ter-minal domain would render the host cell resistant to infection(22).

Electron microscopy of phage expressing a human antitet-anus Fab revealed specific single labeling at one end of thefilamentous phage (Fig. 2). This result is consistent with theanticipated targeting ofthe Fab and can be compared with ourpreviously reported cpVIII fusion, which resulted in labelingalong the filament. Additionally, bacterial membrane frag-ments were also labeled supporting the proposed morpho-genic scheme. We suggest the acronym Phabs for phagedisplaying antibody combining sites.To demonstrate the utility of pComb 3 for the antigen-

specific sorting ofphage, constructs were prepared with Fabsof known specificity and affinity. Phabs were prepared byovernight infection of phagemid containing cells yieldingtypical titers of 1011 cfu/ml. By using phagemids encodingdifferent antibiotic resistances, ratios of clonally distinctphage were easily determined by titering on selective plates.In single-pass enrichment experiments, clonally mixed phagewere incubated with an antigen-coated plate, nonspecificphage were removed by washing, and bound phage wereeluted with acid (Table 1). Phage expressing the humanantitetanus toxoid clone lOC were enriched on a tetanustoxoid-coated surface over helper phage and phage display-ing the anti-NPN antibody by 104- and 105-fold, respectively.Clone lOC is a high-affinity clone with a Kd on the order of10-9 M, whereas clone 7E exhibits a Kd on the order of 10-

A B

j.:

FIG. 2. Electron micrographs showing antigen-specific labelingof filamentous phage displaying Fab molecules. (A) Filamentousphage as produced with the pComb 3 phagemid system displaytetanus tox~oid-specific Fab molecules on their tail. Phage werelabeled with 5- to 7-nm colloidal gold particles coated with tetanustoxoid (x140,OO0.) (B) Filamentous phage, as previously reportedusing the gVIII multivalent display system (11), shows antigen-specific labeling along the body of the phage and is provided forcomparison. (x65,000.)

M (4). This lower-afflinity clone was enriched over nonspe-cific phage by 103-fold. These enrichments are similar tothose reported for hormone phage (10).The advantage of monovalent display over multivalent

display is that it allows for the sorting of clones based onaffinity as well as specificity, as does the immune system.This is demonstrated by the 253-fold enrichment of the tightbinding clone 10C over the weaker binder clone 7E using thepComb 3 system, and this can be compared with a 5-foldenrichment using the multivalent gVIII construct. Studies ofpeptide libraries on phage that display four to five copies ofthe peptide on the phage surface (multivalent display) haveshown that multivalency prevents the separation of phagedisplaying moderate-affinity peptides (10-6 M) from thosedisplaying high-afflinity peptides (10-p M) (8). Multivalency

Table 1. Antigen-specific enrichment of phage displaying Fabfragments: Single-pass experiments

Clonal mixture Initial ratio Final ratio EnrichmentpC3-TTl0C/VCSM13 1/5714 155/47 1.88 x 104pC3-TT10C/pCC3-NPN 1/1746 89/1 1.54 x 10-1pC3-TTE/pCC3-NPN 1/350 19/5 1.33 x 103pC3-TT7E/pCC3-TT10C 600/1 1836/773 253pC8-TT7E/pCC8-TT10C 4348/1 780/1 5.5

Phagemid clone pCC3-NPN encodes the Fab designated 2b iso-lated from the original p-nitrophenyl phosphonamidate (NPN) com-binatorial library (2) in the pCBComb 3 vector. Phagemids pC3-TT10C and pC3-TT7E encode the described antitetanus toxoid Fabclones 10C and 7E (4) cloned into the pComb 3 vector. PhagemidpCC3-TIT10C contains clone 10C in pCBComb 3. Phagemids pC8-MTE and pCC8-TT10C encode clones 7E and 10C in the described

ampicillin- and chloramphenicol-resistant gVIII expression system(11).

7980 Biochemistry: Barbas et al.

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

25, 2

020

Proc. NatL. Acad. Sci. USA 88 (1991) 7981

leads to a chelation effect (for discussion, see ref. 23), whichlikely reduces the ability to discriminate between phage-bearing high-affinity Fabs from those bearing low-affinityFabs.The use of the system was further demonstrated by sorting

a previously characterized (one binder per 5000 clones)human combinatorial antitetanus toxoid Fab library (4). Thelibrary was reconstructed in pComb 3 retaining the originalpairings of heavy and light chains. The library size, 107clones, was 10-fold larger than the original A phage library.After a single round of panning, 13 of 57 clones picked weredetermined to be tetanus toxoid binders-a 103-fold enrich-ment (Fig. 3). Enrichment can be monitored without screen-ing by following the percent yield ofphage (16). Following thethird panning, the phage yield had increased 200-fold, indi-cating enrichment of specific phage (see legend to Fig. 3). Atthis point, virtually all clones were antigen-specific binders.Large combinatorial libraries of 108 members should be easilyaccessible using this system. Because 1011 phage can besorted in a single microtiter well, all members of the librarycan be selectively accessed.The pComb 3 vector was designed for rapid conversion

from a selection vector to an expression vector (Fig. 1). Thisfacilitated the preparation of soluble Fab at a level estimatedto be several milligrams per liter depending on the clone.Three positive clones selected from the antitetanus toxoidlibrary were examined by competitive ELISA (Fig. 4), andthe expected antigen specificity was confirmed. The apparentbinding affinities of the isolated clones are on the order ofthose isolated using the A phage system (4).We have demonstrated that monovalent Fab phage display

libraries can be constructed, and specific binders can berapidly selected. The reduced valency ofdisplayed moleculesallows the experimenter not only to isolate clones of defined

B4Si I4

1J

4'0V

C

tiv <it #4 %'*\*'t #

D

44% Os 4 t%%A .

"'kFIG. 3. Enrichment of antigen-specific clones from a library as

monitored by screening for antigen binding after successive roundsof panning. Positive colonies were identified on nitrocellulose filtersby using tetanus toxoid conjugated with alkaline phosphatase in acolorimetric assay. Positive colonies appear dark. A control positivecolony was included on each filter (lower right). (A) Clones pickedfrom initial library. (B) Clones picked after the first panning exper-iment. Phage input was 1.17 x 101" cfu, and 2.47 x 105 cfu wereeluted from the first panning; % yield = 2.11 x 1O-4. (C) Clonespicked after the second round of panning, where 1 x 1011 cfu wereapplied, and 1.6 x 107 cfu were eluted; % yield = 1.6 x 10-2. (D)Clones picked from the third round of panning, where 7 x 1010 cfuwere applied, and 3.1 x 107 cfu were eluted; % yield = 4.4 x 10-2.

.~60

.~40

e 20

0

-10 -9 -8 -7 -6 -5 -4

log [Competing Antigen]

FIG. 4. Specificity of antigen binding to tetanus toxoid shown bycompetitive ELISA. Experiments were done with bacterial super-natants.

specificity, but to select for higher affinity binders mimickingB-cell selection in a laboratory setting. This has implicationsfor antibody use in therapy and diagnostics. This system willcomplement our previously reported multivalent display sys-tem on cpVIII (11). A previously reported single-chain anti-body system expresses the fusion on each of the four to fivecpIII proteins and might be expected to behave as the cpVIIIsystem with respect to multivalency (9). For the selection ofhigh-affinity antibodies monovalent display is necessary,whereas selection of low-affinity clones may require multi-valent display. De novo design and selection of bindingcavities will likely require both systems.

Catalytic antibodies often suffer from substrate and prod-uct inhibition (12). Phage displaying catalytic antibody librar-ies might be selected for binding to the structural featuresunique to the transition-state analogue by substrate or prod-uct washings. Furthermore, schemes can be devised for theselection of Phabs that covalently interact with an immobi-lized substrate, thereby trapping desired amino acid sidechains that can be recruited for catalysis. The full utility ofdirectly linking a selectable protein architecture via the phagesurface to the instructions for its production, its DNA, couldbe realized by repeated rounds of clonal selection and mu-tation. This scheme would allow experimenter-controlledevolution of protein architectures similar to that reported forRNA (24) and has implications far beyond antibodies.We note that the pComb 3 system has been used to select

antigen-specific Fabs against proteins and haptens from anumber of other libraries including combinatorial librariesfrom humans, mice, and hu-PBL-SCID mice (C.F.B.,R.A.L., Dennis R. Burton, Roger Caothien, James Light, andPeter Fischer, unpublished data). In one case, the frequencyof binding clones was known to be -1 per 200,000.

We are most grateful to Cheng-Ming Chang for electron micros-copy studies, Terri Jones for DNA sequencing, Sydney Brenner(Medical Research Council, Molecular Genetics Unit, Cambridge,U.K.), Norton B. Gilula (The Scripps Research Institute), andGeorge P. Smith (University of Missouri) for helpful discussions andadvice, Dennis R. Burton and Mats A. Persson for materials, advice,and encouragement, Jack Chen for critical reading ofthe manuscript,and Diane Schloeder and Vicki Stoppenbach for technical support.C.F.B. is a National Institutes of Health Postdoctoral Fellow.

1. Sastry, L., Alting-Mees, M., Huse, W. D., Short, J. M., Sorge,

A

Biochemistry: Barbas et al.

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

25, 2

020

7982 Biochemistry: Barbas et al.

J. A., Hay, B. N., Janda, K. D., Benkovic, S. J. & Lerner,R. A. (1989) Proc. Nat!. Acad. Sci. USA 86, 5728-5732.

2. Huse, W. D., Sastry, L., Iverson, S., Kang, A. S., Alting-Mees, M., Burton, D. R., Benkovic, S. J. & Lerner, R. A.(1989) Science 246, 1275-1281.

3. Caton, A. J. & Koprowski, H. (1990) Proc. Nat!. Acad. Sci.USA 87, 6450-6454.

4. Persson, M. A. A., Caothien, R. H. & Burton, D. R. (1991)Proc. Nat!. Acad. Sci. USA 88, 2432-2436.

5. Burton, D. R. (1991) Trends Biotechnol. 9, 169-175.6. Scott, J. K. & Smith, G. P. (1990) Science 249, 386-390.7. Devlin, J. J., Panganiban, L. C. & Devlin, P. E. (1990) Science

249, 404-406.8. Cwirla, S. E., Peters, E. H., Barrett, R. W. & Dower, W. S.

(1990) Proc. Nat!. Acad. Sci. USA 87, 6378-6382.9. McCafferty, J., Griffiths, A. D., Winter, G. & Chiswell, D. J.

(1990) Nature (London) 348, 552-554.10. Bass, S., Greene, R. & Well, J. A. (1990) Protein Struct. Funct.

Genet. 8, 309-314.11. Kang, A. S., Barbas, C. F., Janda, K. D., Benkovic, S. J. &

Lerner, R. A. (1991) Proc. Nat!. Acad. Sci. USA 88, 4363-4366.

12. Lerner, R. A., Benkovic, S. J. & Schultz, P. G. (1991) Science252, 659-667.

Proc. Nati. Acad. Sci. USA 88 (1991)

13. Yanisch-Perron, C., Vieira, J. & Messing, J. (1985) Gene 33,103-119.

14. Horton, R. M., Hunt, H. D., Ho, S. N., Pullen, J. K. & Pease,L. R. (1989) Gene 77, 61-68.

15. Short, J. M., Fernandez, J. M., Sorge, J. A. & Huse, W. D.(1988) Nucleic Acids Res. 16, 7583-7600.

16. Parmley, S. F. & Smith, G. P. (1988) Gene 73, 305-318.17. Hanahan, D. (1983) J. Mol. Biol. 166, 557-580.18. Helfman, D. M., Feramisco, J. R., Fiddes, J. C., Thomas,

G. P. & Hughes, S. H. (1983) Proc. Natl. Acad. Sci. USA 80,31-35.

19. Chang, C. N., Model, P. & Blobel, G. (1979) Proc. Natl. Acad.Sci. USA 76, 1251-1255.

20. Crissman, J. W. & Smith, G. P. (1984) Virology 132, 445-455.21. Houston, J. S., Levinson, D., Mudgett-Hunter, M., Tai, M.-S.,

Novotny, J., Margolies, M. N., Ridge, R. J., Bruccoleri, R. E.,Haber, E., Crea, R. & Oppermann, H. (1988) Proc. Natl. Acad.Sci. USA 85, 5879-5883.

22. Boeke, J. D., Model, P. & Zinder, N. D. (1982) Mol. Gen.Genet. 186, 185-192.

23. Meyers, R. T. (1978) Inorg. Chem. 17, 952.24. Ellington, A. D. & Szostak, J. W. (1990) Nature (London) 346,

818-822.

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

25, 2

020