Proc. Natl. Acad. Sci. USA Vol. 92, pp. 1609-1613, February 1995 Genetics

Display of peptides and proteins on the surface of bacteriophage A (phage display/biopanning/D protein/site-specific recombination/libraries) NAT STERNBERG AND RONALD H. HoEss Dupont-Merck Pharmaceutical Company, Experimental Station E328/B33, Wilmington, DE 19880-0328 Communicated by Sankar Adhya, National Cancer Institute, Bethesda, MD, October 14, 1994 (received for review August 12, 1994)

ABSTRACT The display of peptides or proteins on the age processing has also shown that D protein is assembled as surface of viruses is an important technology for studying trimers, which appear as prominent protrusions on the capsid peptides or proteins and their interaction with other mole- surface (8). This feature suggests that peptides or proteins cules. Here we describe a display vehicle based on bacterio- fused to D protein are likely to be accessible for binding to phage A that incorporates a number of features distinct from target molecules. Recently, a display system using the V tail other currently used display systems. Fusions of peptides or protein of bacteriophage A has been reported (9). protein domains have been made to the amino terminus ofthe 11-kDa D protein ofthe A capsid. These fusions assemble onto the viral capsid and appear to be accessible to ligand inter- MATERIALS AND METHODS actions, based on the ability of a monoclonal antibody to Construction and Use of Bacterial Strains and Phage*. recognize an epitope fused to the D protein on phage heads. Escherichia coli strain YMC supF was used for titering plaque- To produce large D fusion display libraries and yet avoid the forming units (pfu). E. coli NM522 F' lacIqA(lacZ) M15 cumbersome task of cloning many fragments into A DNA, we proA+B+/supE thiA(lac-proAB) containing D fusion plasmids have used the Cre-loxP site-specific recombination system in was used as a source of extracts containing D fusion protein. vivo to incorporate plasmids encoding the D fusions into the Strains NS2973 and NS2974 are JM109 recAI supE44 lysogenic phage genome. Finally, we show that D fusion proteins can be for Aimm434 ninS and Aimm434 ninS Cre+, respectively. added in vitro to phage lacking D protein and be assembled NS3762 is E. coli strain N99 sup°recAS6 tetr(ADamlS b538 onto the viral capsid. cIts857 ninS). It was generated by infecting strain N99 recA+ with ADamlS b538 cIts857 ninS and selecting A lysogens at Surface display of peptides or proteins has become a powerful 32°C as described (10). Since the b538 deletion in this phage tool in protein engineering and for discovering ligands that removes the A integration system, the frequency of lysogeny by bind receptors. The features that make this technology valu- this phage is low, and the phage presumably insert into the able are (i) the ability to create large molecular libraries of bacterial chromosome by illegitimate recombination. A lyso- variants and (ii) in vitro selection techniques that allow effi- gen containing tandemly integrated prophage was selected by cient searches of the library without testing each member virtue of its ability to produce 10-100 phage per cell when the individually. Smith and coworkers (1, 2) first demonstrated lysogen was thermally induced (11). It should be noted that the this technology by using phage fd where amino-terminal fu- yield of A phage in this experiment is high despite the lack of sions were made to the minor coat protein gIII. Since that time the normally essential phage D protein because the phage considerable progress has been made in constructing large contains a net 21.5% which obviates the libraries of random peptides, and a variety of proteins includ- genome deletion, ing antibodies have been displayed (for recent review, see ref. need for D. Finally, the tandem lysogen was made recA- by 3). A number of other display systems have been developed P1-mediated transduction of the recA56 allele from strain that use a bacterial surface rather than the viral capsid as the NS1426 srl::TnlO recAS6. To produce phages from NS3762, an display surface (4). A feature common to both phage and exponentially growing culture was induced by shifting the cells bacterial systems is that the displayed molecule(s) must be from 32°C to 42°C for 15 min to inactivate the cIts857repressor secreted across the bacterial membrane. We describe here a and then shifted to 38°C for 40 min. The culture was com- system where secretion is not required for phage display. pletely lysed by 45 min after the temperature shift and pro- Bacteriophage A is a virus that, in its lytic mode, assembles duced 10-50 phage per cell. intracellularly prior to release of viral particles from the cell. To construct a A D- phage containing a loxP site, ADamlS Much is known about the structural proteins that make up the imm2l nin5 was grown in a dam- host, and the phage DNAwas viral shell and about the pathway(s) by which these proteins cleaved at the unique Xba I site at map coordinate 24,508. The assemble (5). The 11-kDa D protein encoded by the phage resulting two A arms were ligated to an annealed complemen- appears to play a role in stabilizing the phage prohead as it fills tary set of oligonucleotides, 5'-AGCTTATAACTTCGTAT- with DNA (6, 7). Normally D protein is essential for phage AATGTATGCTATACGAAGTTATGTTTAAAC-3' and head morphogenesis. However, when the size of the phage 5' -AGCTGTTTAAACATAACTTCGTATAG- genome is <82% of wild type, phage can be assembled in the CATACATTATACGAAGTTATA-3', that contain the loxP absence of D protein and are stable in the presence of 10 mM sequence and are flanked by 4 bp of single-stranded Xba I Mg2+, although they are exquisitely sensitive to disruption by ends. The ligated DNA was packaged into A phage using a EDTA (6). The conditional requirement for D protein in viral Gigapack II system (Stratagene). Lysates were made from assembly suggests that this protein may be the ideal choice for individual plaques, and these were tested for the presence of fusion with peptides or protein domains for display on the viral capsid. Recent work using cryo-electron microscopy and im- Abbreviations: Kmr, kanamycin resistant (resistance); Kmsi, kanamy- cin sensitive; Ap, ampicillin; Apr, Ap resistant (resistance); pfu, plaque-forming unit(s); AII, angiotensin II; Bi, Bi domain of protein The publication costs of this article were defrayed in part by page charge G from group G Streptococcus; IPTG, isopropyl f3-D-thiogalactoside. payment. This article must therefore be hereby marked "advertisement" in *The various strains used in this work can be obtained from the accordance with 18 U.S.C. §1734 solely to indicate this fact. authors. 1609 Downloaded by guest on September 25, 2021 1610 Genetics: Sternberg and Hoess Proc. Natl. Acad Sci. USA 92 (1995) the loxP site by PCR amplification of 1 Al of lysate by using two EDTA sensitivity of phage was tested. Phage were diluted into primers, 5 '-GGAATTGGTTAGCAAGTTACTACCG-3' either 10 mM Tris, pH 7.5/10 mM MgCl2 (TM) or 10 mM Tris, and 5'-ATTGAAGCAAATCTGAAACCTATTA-3', that pH 7.5/10 mM EDTA (TE) and incubated for 10 min at 37°C. flank the A Xba I site. Serial dilutions were then titered on YMC cells. Construction of D Fusion Vectors. The A D gene was am- Generation of A, D-Fusion Plasmid Cointegrates by Cre- plified by PCR using primers, 5'-GTAAGCCATGGTTAT- loxP Recombination. Strains NS2973 and NS2974 containing GACGAGCAAAG-3' and 5 '-GTTCGAATTCCTATTA- D fusion plasmids with loxP sites were grown to an OD600 of AACGATGCTGATTGCC-3', that contain the restriction -0.2 in Luria-Bertani (LB) medium containing ampicillin sites Nco I and EcoRI (underlined), respectively. After cleav- (Ap) at 100 ,ug/ml at 37°C. One milliliter of cells was then age by Nco I and EcoRI, the DNA was cloned into the Nco I centrifuged, and the pellet was resuspended in 100 ,ul of a A and EcoRI sites of pTrcHisA (Invitrogen), resulting in plasmid D-loxP lysate at a multiplicity of infection of 1. After absorp- pNS3785. This places the wild-type D gene under the control tion at 37°C for 10 min, infected cells were diluted to 1 ml in of the pTrc promoter. To create peptide or protein fusions to LB medium containing Ap (100 ,tg/ml), 10 mM MgCl2, and 1 D protein, the vector was modified by cloning an oligonucle- mM isopropyl f3-D-thiogalactoside (IPTG). Cultures were otide adaptor, 5'-CATGTCTGACCGTGTTTACATCCAC- grown for -2 hr at 37°C, and then cells were lysed. A drop of CCGTTCGGCGCGCCATCTGTTAGCGGCCGCAC-3', chloroform was added, and cell debris was removed by cen- into the unique Nco I site at the amino terminus of the D trifugation. This lysate was used to infect an exponential coding sequence to create pRH800. This oligonucleotide en- culture of YMC at a multiplicity of infection of 1, and Ap- codes the eight amino acids of the peptide hormone angio- resistant (Apr) transductants containing the desired A-plasmid tensin II (AII) in frame with D followed by unique Asc I and cointegrates were selected (17). Not I restriction sites (underlined), respectively. DNA encod- Preparation of D Fusion Extracts. Cultures (20 ml) of ing the Bi domain of protein G from group G Streptococcus NM522 containing either plasmid pNS3785, pRH804, or (Bi) was PCR amplified from pET-16BStrpG (12) using prim- pRH815 were grown at 37°C to OD600 of -0.25. IPTG was ers 5'-GTGATCGGCGCGCCAGACATCTTGGCTGC- then added to 1 mM, and the cultures were grown for an TCTGCCG-3' (forward) and 5'-TAGATCCATGCGGC- additional 2 hr at 37°C. Cells were collected by centrifugation, CGCTGCCACCTTCGGTAACCGTGAA-3' (reverse). Af- resuspended in 300 ,ul of extraction buffer [10 mM Tris, pH ter cleavage by Asc I and Not I, the DNA was cloned into 7.6/10 mM MgCl2/50 mM NaCl/1 mM dithiothreitol/1 mM pRH800, creating pRH815. Pefabloc (Boehringer Mannheim)], and then sonicated on ice Construction of D Fusion Vectors Containing loxP Sites. To two or three times in 30-sec bursts with a Kontes sonicator. recombine D fusion plasmids into the A D- loxP phage, a loxP Equal volumes of this sonicate and a ADaml5 b538 cIts857nin5 site, 5 '-AGCTTATAACTTCGTATAATGTATGCTATA- phage preparation (1 x 1011 phage per ml) lacking D protein CGAAGTTATGTTTAAAC-3', was cloned into the unique were mixed, and the mixture was incubated at 30°C for 10 min. HindIII site downstream of the D gene in pRH800 to generate The phage were then diluted and either titered on YMC or pRH804. pRH814 is similar to pRH804 except that the AII incubated with a monoclonal antibody for biopanning. coding sequence was removed and the kanamycin resistance (KMr) gene from pRH43 (13) was inserted in the HindIII site adjacent to loxP. RESULTS Construction of D Fusions Under the Control of the Arabi- D Protein Fusions Are Expressed on the Surface of A. To nose Promoter. The sequences encoding the AII-D fusion and establish that peptides or protein domains could be expressed the AII-B1-D fusions from pRH800 and pRH815, respec- on the surface of A, two model systems were tested. As a tively, were PCR amplified using the same forward primer representative peptide, an oligonucleotide encoding the eight- 5' -TCGACCCGATTTAAATGTCTGACCGTGTTT- residue peptide hormone AII was cloned as an amino-terminal ACATCC-3' and either reverse primer 5'-GTAAACA- fusion to the D gene in plasmid pRH800. As a representative GATCTCTATTAAACGATGCTGATTGCC-3' for pRH800 protein domain, a synthetic gene encoding the 65-aa Bi IgG or 5'-GTAAACGGATCCCTATTAAACGATGCTGATT- binding domain from group G Streptococcus (12, 18) was GCC3' for pRH815. The PCR-amplified fragments were di- cloned between the amino-terminal All-encoded peptide and gested with Swa I and Bgl II (pRH800) or with Swa I and the D gene in pRH815. BamHI (pRH815), respectively. The resulting fragments were These plasmids were transformed into NS3762, a strain cloned into Swa I/Bgl II-digested pKM18 (T. Patterson, Du- lysogenic for ADamlS b538 cIts857 ninS. Upon thermal induc- Pont-Merck), a derivative of pBAD18 (14). The resulting tion of the phage, it was anticipated that the D fusion protein constructs have the AII-D (pRH809) and AII-B1-D (pRH817) synthesized from the plasmid would complement the phage fusions under the control of the arabinose promoter. lacking D and that the protein would be assembled on the viral Immunoblot Analysis. Phage proteins were separated by capsid. To verify the presence of the D fusion proteins on the electrophoresis using a tricine/SDS/polyacrylamide gel (15) viral capsid, lysates were tested in two ways. First, samples of followed by electroblotting onto nitrocellulose. Fusion pro- the lysates were concentrated and phage proteins were sepa- teins were detected using KAA8 (16) as the primary antibody rated by electrophoresis and then electroblotted to a nitro- followed by biotinylated horse anti-mouse IgG (Pierce), horse- cellulose filter. The blot was then probed with a monoclonal radish peroxidase-conjugated avidin (Pierce), and Renais- antibody, KAA8, which specifically recognizes All (16). Since sance Western blot chemiluminescence reagent (DuPont/ KAA8 is an IgGl antibody, it should bind the All peptide and NEN). Bl as well. As shown in Fig. 1, lane A, phage derived from cells Biopanning. To immunologically select phage displaying a containing pRH800 show a protein band migrating at 17 kDa, surface epitope, biopanning was carried out as described (2) which is close to the calculated 13.7 kDa expected for the AII with the following modification. Rather than acid elution, peptide fused to the D protein. In lane B, a slower migrating immobilized phage were assayed by adding 1 ml of exponen- band at 24 kDa is observed, consistent with the predicted size tially growing YMC cells (3 x 108 cells per ml) in Tris-buffered of 21.1 kDa for Bi fused to the D protein. Lane C contains saline (TBS)/Mg2+ to the dish and incubating for 20 min at phage with only D protein, and, as expected, it produces no 37°C. Cells were then removed, and serial dilutions were plated signal. for pfu determination. A second test to determine if the D fusion proteins are Test for EDTA Resistance. To determine whether D protein present on the surface of A is based on the ability to physically or D fusion proteins were present on the phage capsid, the capture phage using reagents that recognize peptides or pro- Downloaded by guest on September 25, 2021 Genetics: Sternberg and Hoess Proc. NatL Acad Sci USA 92 (1995) 1611

A B C displayed. In the extreme, the phage could maximally display 405 copies (8). This polyvalency would be useful for binding 30 kDa- targets whose interaction with the D fusion protein are of low or moderate affinity. Conversely, in M13-based systems, it has been demonstrated that to select high-affinity binders from those of moderate affinity it is necessary that the peptide or protein be displayed in monovalent form (19). In the A system, 21.5 kDa- one measure of the number of D protein fusion molecules present on the capsid surface can be ascertained by the sen- sitivity of the phage to EDTA (6). Phage that are D- are rapidly inactivated in the presence of EDTA. In vitro experi- ments in which D protein is added back to phage with D- FIG. 1. Western blot detection of D fusion proteins. Phage proteins heads have shown that the phage need to be coated with all 405 from concentrated lysates were separated by electrophoresis and copies of D protein for them to become resistant to EDTA transferred to nitrocellulose by electroblotting. D fusion proteins were treatment (8). When phage are produced from lysogens con- immunodetected by KAA8 as the primary antibody. Lysates were taining pNS3875 (transcription of the D gene controlled by the made from NS3762 containing the following plasmids: lane A, strong pTrc promoter), those phage were resistant to EDTA pRH800 (AII-D); lane B, pRH815 (AII-StrpG-D); and lane C, inactivation even when the pTrc promoter was not induced by pNS3785 (D). The molecular masses indicated in kDa were estimated IPTG (data not shown). This suggests that even under re- from standards electrophoresed in the adjacent lane (Rainbow mark- pressed conditions (the pTrc plasmid contains a lacIq gene), ers; Amersham). the pTrc promoter permits sufficient amounts of D fusion teins displayed on the phage surface, a technique termed protein to be made to fully coat the progeny phage produced. biopanning (2). Phage lysates derived from cells containing In an effort to better repress the produclion of the D fusion pRH800 or pRH815 were incubated overnight with biotiny- protein, the gene was placed under the control of the more lated KAA8 and then on dishes coated with streptavidin. After effectively regulated arabinose promoter (14). This promoter can be more effectively regulated and more efficiently re- a brief incubation period, the lysate was removed and the pressed than the pTrc promoter. Two constructs, the AII- dishes were washed 10 times to remove unabsorbed phage. To tagged D fusion and the AII-B1-D fusion, were cloned into determine if phage were captured, 1 ml of exponentially grow- plasmid pKM18 containing the arabinose promoter, and the ing YMC resuspended in TBS/Mg2+ was placed on the dishes resulting constructs, pRH809 and pRH817, respectively, were and incubated for 20 min at 37°C. Assuming that phage have transformed into NS3762. The phage in these lysogens were been immobilized on the dish via their attachment to KAA8, thermally induced in the presence of glucose (promoter re- we reasoned their presence could be detected by allowing them pressed). Table 2 shows that in the presence of glucose the to infect sensitive cells. After incubation, cells were removed resulting phage are EDTA sensitive, indicating that they con- from the dish and various dilutions were plated to determine tain <405 copies of the D fusion protein. Interestingly, bio- the number of pfu. As shown in Table 1, the recovery of phage panning experiments show that they are captured efficiently by particles from the dish was at least a 1000 times higher with KAA8 (Table 2). lysates prepared from cells containing either pRH800 or Combinatorial Library Construction with A D Fusion pRH815 than with lysates prepared from cells containing Phage. The results presented thus far indicate that D fusion pNS3785. The latter plasmid produces only the wild-type D protein produced by a plasmid in a A lysogen can be displayed protein, which is not recognized by KAA8. Since the fusion on the surface of phage generated when that lysogen is in- protein encoded by pRH815 contains the AII peptide as well duced. Since the DNA encoding the D fusion is not incorpo- as Bi, one could argue that the results indicate only expression rated into the phage genome, display of the D fusion will be of the AII-Bl fusion but do not necessarily indicate function- lost upon further propagation of the virus in cells not con- ality of Bi. To address this, lysates were biopanned with taining the plasmid. For construction of combinatorial librar- biotinylated horse anti-mouse IgG, which should bind only a ies, such as in the display of random peptide sequences, it is functional Bi but not the AII peptide. As shown in Table 1, necessary to have a genetic tag for each individual in the library only the Bl-containing lysate showed interaction with the so that upon selection from the library the displayed molecule horse anti-mouse antibody. can easily be decoded by DNA sequence analysis. We have Control of D Fusion Expression. Depending on the use of devised a method by which the DNA coding for the D fusion the A display vehicle, it would be advantageous to be able to can readily be incorporated into the phage genome by site- control the number of copies of the D fusion protein that are Table 2. Expression of D fusions under repressed conditions Table 1. Biopanning lysates with KAA8 and horse anti-mouse IgG EDTA % % recovery* Plasmid Encoded protein resistant* recoveryt Horse anti-mouse pRH809 AII-D - 0.9 Plasmid Encoded protein KAA8 IgG pRH817 AII-B1-D 1.7 pNS3785 D 0.0001 0.0005 Strain NS3762 containing the designated plasmids was grown to an pRH800 AII-D 3.0 0.0015 OD600 of 0.5 in LB/Ap (100 ,ug/ml) at 30°C. Cells were then centri- fuged and resuspended in prewarmed LB/Ap with 0.2% glucose. The pRH815 AII-B1-D 5.9 0.7 cultures were incubated at 42°C for 15 min and then shifted to 37°C Strain NS3762 was transformed with each plasmid. Cells were grown for 45 min or until cell lysis occurred. Cellular debris was then removed to an OD6w0 of 0.5 and thermally induced at 42°C for 15 min followed by centrifugation. by continued growth at 37°C for -1.5 hr at which time the culture *One microliter of lysate was added to 1 ml of TM or 1 ml of TE and lysed. Cellular debris was removed by centrifugation, and phage were incubated for 10 min at 37°C. Phage were then titered on YMC. -, titered on YMC. titers from the TM were >103 times those obtained from the TE *Biopanning was done as described in Materials and Methods using samples. either 1 ,ug of biotinylated monoclonal antibody KAA8 or 1.5 &g of tBiopanning was done as described in Materials and Methods using 1 biotinylated horse anti-mouse IgG (Pierce). The percent recovery ,ug ofbiotinylated monoclonal antibody KAA8. The percent recovery was calculated by the (output titer/input titer) x 100. was calculated by the (output titer/input titer) x 100. Downloaded by guest on September 25, 2021 1612 Genetics: Sternberg and Hoess Proc. NatL Acad ScL USA 92 (1995) specific recombination (17). The strategy employs the Cre- because the phage used contains the Aatt integration system or, loxP site-specific recombination system of bacteriophage P1 alternatively, the A plasmid cointegrates can be maintained in and is outlined in Fig. 2. In a cell that contains the Cre cells as extrachromosomal elements because of their plasmid recombinase, a plasmid is introduced that contains the D replicon. In any case, the A lysogens can then be grown and fusion and a loxP recombining site. These cells are then in- induced to generate a population of phage each containing a fected with a ADaml5 imm2l ninS phage containing a loxP site. plasmid within its genome that expresses the D fusion on its During the course of the infection process, Cre recombines the surface. It should be noted that in these experiments NS2973, loxP site on the plasmid with the loxP site on the phage, NS2974, and YMC are strains containing amber suppressors. resulting in the insertion of the plasmid into the phage genome. Accordingly, the phages generated could contain both wild- Ultimately the chimeric DNA molecule is packaged into the type D protein and D fusion protein. If the desire were to viral capsid. Phage derived from such infections should con- generate phage with only D fusion protein, sup0 bacteria could tain the plasmid and also have the corresponding D fusion be used in both the loxP cointegrate and A transduction steps. protein displayed on its surface. To test this strategy, a loxP site To determine how effectively one can select a phage that was cloned into the plasmid containing the All-tagged D displays a particular D fusion from a population of noncognate fusion, generating pRH804. This plasmid was transformed into phage, the following reconstruction experiment was carried NS2974, a strain containing a cre gene and into an isogenic out. Two A D- loxP lysates were prepared in which the plasmid strain, NS2973, in which the cre gene is absent. Exponentially cointegrate contained either pRH804 with an AII-tagged D growing cells were infected at a multiplicity of 1 A D- loxP protein or pRH814 with no AII tag. In addition, the pRH814 phage per cell and grown for 2 hr at 37°C. Cell debris was plasmid carries the gene for Kmr. The two lysates were mixed removed by centrifugation, and the lysates were titered for pfu in the various ratios shown in Table 3, and then each mixture and Apr transductants. The recovery of Apr transductants per was biopanned using KAA8. After biopanning, immobilized pfu was a 1000-fold greater when the phage had been grown phage were rescued by incubating with YMC and selecting for in the host. It Apr transductants. The transductants were then subsequently cre-containing should be pointed out that not all tested for the presence or absence of the Kmr marker as a the phage that contain D fusion proteins have necessarily measure ofwhether the phage expressing the All D fusion had incorporated the plasmid DNA. To remove these phage from been selected from the population. Table 3 shows that despite the population, the lysate was then used to infect cells not an input titer comparable to the previous experiments (Table containing a cre gene, and Apr lysogens were selected. Note 1), the recovery of phage is relatively low (<0.001% in Table that the efficiency of A lysogeny under these conditions is high 3 compared to 3% in Table 1), which reflects the fact that the vast majority of phages do not display AII and hence are not A cre i / captured by KAA8. The phage that were captured, presumably those containing the AII tag, are very much enriched for the D fusion AprKms phenotype. Thus in Table 3, row 2, the initial ratio of AprKmr to AprKms (KmS = kanamycin sensitive) phage is 37,000:1, but with the biopanned population it is 1.5:1, repre- Apr senting a 25,000-fold enrichment in a single cycle of biopan- ning. This suggests that in large libraries of phage displaying XD loxP different random peptides one should readily be able to select l~~~~~oxP out rare phages with a few cycles of biopanning and amplifi- I cation. In Vitro Addition of D Fusion Protein. It has previously been ; Cre mediated recombination shown that wild-type D protein can be added in vitro to phage B proheads lacking D protein (7). This in vitro self-assembly by XD- Apr D fusion D protein could potentially be useful for adding various de- terminants to the surface of the virus without requiring the loxP loxP phage genome to encode them. It was therefore of interest to determine whether the D fusion proteins can also be assem- Infect and select Apr lysogens bled on proheads in vitro. Extracts were made from cells containing induced pNS3785 (wild-type D), pRH800 (AII-D), and pRH815 (AII-B1-D) and incubated with A D- phage. In c - tE cre all cases, the resulting phage were resistant to treatment with EDTA, indicating that the D protein had assembled on the capsid. When biopanned with KAA8, phage incubated with XD- Apr D fusion the D fusion extracts AII-D and AII-B1-D were captured loxP loxP Table 3. Selection of AII-D phage Input* % drug ApRH814 ApRH804 resistance§ (AprKmr) (AprKms) Ratiot % recoveryt AprKmr AprKms FIG. 2. Combinatorial library construction using Cre-loxP site- 5.2 x 107 0 0 4.4 x 10-4 100 0 specific recombination. (A) Cells containing D fusion plasmids were 5.2 x 107 1.4 x 103 37,000:1 7.3 x 10-4 62 38 infected with A D- phage containing a loxP site. During infection, the x x x 10-4 Cre recombinase promotes recombination between phage and plasmid 5.2 107 1.4 102 370,000:1 4.2 98 2 via recombination at loxP sites. The resulting cointegrates (B) are *Input titers are the number of Apr transducing particles. packaged as phage. Since the lysate will also contain phage that have tRatio of phage particles containing wild-type D protein and the gene not incorporated the plasmid, cointegrates are preferentially recov- for Kmr to phage containing only Apr and AII-D protein. ered by infecting cells and selecting for Apr transductants (C). Because :The number of Apr transductants recovered after biopanning with the recipient cells do not contain Cre, the cointegrates are stably KAA8 divided by the number of input Apr particles x 100%. maintained. Phage cointegrates expressing the D fusion protein can §After biopanning, 100 Apr colonies from each biopan were tested for then be produced by thermal induction of the Apr cells. the presence or absence of Kmr. Downloaded by guest on September 25, 2021 Genetics: Sternberg and Hoess Proc. Na:l Acad Sci USA 92 (1995) 1613 with an efficiency of 7.2% and 6.0% respectively. These num- immunological screens in which the entire library must first be bers are 1000-fold greater than phage incubated with a D plated as individual plaques and then tested with the antibody extract. after transfer of the plaques to nitrocellulose. A second potential use for the A display system may be as a gene delivery vehicle. Recently it has been reported that M13 DISCUSSION phage displaying a constrained RGD sequence can be specif- The A D gene display system reported here presents a useful ically taken up by Hep-2 cells (22). We envision decorating the addition to a number of biological display systems that include capsid not only with sequences required for uptake into the the A V gene (9), single-stranded bacteriophage (3), and bac- endosome but also with sequences that would allow the virus terial surface presentation systems (4). A unique feature of to disrupt the endosome. Surface decoration could be done both A systems is that the displayed peptides or proteins need either in vivo or in vitro and would not require that the coding not be secreted across the bacterial membrane because of the sequences be part of the viral genome. The A system has the intracellular assembly of the virus. In addition, the A system added advantage of being able to deliver sequences of up to 48 may have an advantage over that of single-stranded bacterio- kb in length. phage for the display of proteins or peptides that are normally We thank M. Kendall for oligonucleotide synthesis, T. Reilly for toxic to the cell. Thus, while cells must remain viable for many monoclonal antibody KAA8, J. O'Brian for the D gene clone, generations in order to continuously produce virus in the case T. Patterson for pKM18, and K. O'Neil for StrpG DNA and reading of the M13-based display system, with the A systems produc- the manuscript. tion of any toxic protein can be shut off in a repressed lysogen and then rapidly induced for only a short period prior to cell 1. Smith, G. P. (1985) Science 228, 1315-1317. lysis. Toxic proteins would be a problem only in the latter case 2. Parmley, S. E. & Smith, G. P. (1988) Gene 73, 305-318. if they interfered with phage production or cell metabolism 3. Clackson, T. & Wells, J. A. (1994) Trends Biotechnol. 12, 173-184. during the short induction period. 4. Little, M., Fuchs, P., Breitling, F. & Dubel, S. (1993) Trends BiotechnoL 11, 3-5. There are 405 copies of D protein on the capsid ofwild-type 5. Feiss, M. & Becker, A. (1983) in Lambda II, eds. Hendrix, R. W., A, and nearly all of these must be present to make the phage Roberts, J. W., Stahl, F. W. & Weisberg, R. W. (Cold Spring EDTA resistant (8). Since EDTA-resistant D fusion phage Harbor Lab. Press, Plainview, NY), pp. 279-304. displaying either the AII peptide or Bi can be isolated, those 6. Stemnberg, N. & Weisberg, R. (1977) J. Mol. Biol. 117, 733-759. phage must have -400 copies of the fusion protein on their 7. Imber, R., Tsugita, A., Wurtz, M. & Hohn, T. (1980)J. Mol. Biol. capsid. This high level ofpolyvalency that can be obtained with 139, 277-295. the A system could be useful for detecting binding of relatively 8. Dokland, T. & Murialdo, H. (1993) J. Mol. BioL 233, 682-694. low-affinity ligands or as an immunological reagent for raising 9. Maruyama, I. N., Maruyama, H. I. & Brenner, S. (1994) Proc. antibodies against the displayed molecule (20, 21). Natl. Acad. Sci. USA 91, 8273-8277. 10. Shimada, K., Weisberg, R. A. & Gottesman, M. E. (1972)J. Mol. It has been pointed out that polyvalency in the M13-based Biol. 63, 483-503. display systems is not suitable for selecting high-affinity bind- 11. Mousset, S. & Thomas, R. (1968) Cold Spring Harbor Symp. ing ligands from a background of moderate-affinity ligands. Quant. Biol. 33, 749-754. This problem has been circumvented by the development of 12. O'Neil, K T., Hoess, R. H., Raleigh, D. P. & DeGrado, W. F. monovalent display systems (19). In the case of the A system, (1995) Proteins Struct. Funct. 21, 11-21. we feel that the problem of polyvalency can most likely be 13. Abremski, K, Hoess, R. & Stemnberg, N. (1983) Cell 32, 1301- circumvented by providing a source of wild-type D protein in 1311. addition to the fusion protein. By varying the amounts of each 14. Guzman, L.-M., Barondess, J. J. & Beckwith, J. (1992) J. Bacte- of the extreme valency of the D fusion protein can riol. 174, 7717-7728. these, 15. Schagger, H. & Jagow, G. V. (1987)Anal. Biochem. 166,368-379. probably be reduced. 16. Reilly, T. M., Chiu, A. T. & Timmermans, P. B. (1987) Biochem. Perhaps one reason that A has been overlooked as a display Biophys. Res. Commun. 143, 133-139. vehicle is that cloning small random oligonucleotides for cre- 17. Hoess, R. H., Ziese, M. & Steinberg, N. (1982) Proc. Natl. Acad. ating peptide libraries is not as efficient or convenient as with Sci. USA 79, 3398-3402. the smaller single-stranded viruses or plasmid DNAs. How- 18. Frick, I.-M., Wikstom, M., Forsen, S., Drakenberg, T., Gomi, H., ever, introduction of the Cre-loxP site-specific recombination Sjorbing, U. & Bjorck, L. (1992) Proc. Natl. Acad. Sci. USA 89, alleviates this difficulty, since all the cloning can readily be 8532-8536. done in a small plasmid vector containing the D gene and then 19. Wells, J. A. & Lowman, H. B. (1992) Curr. Opin. Struct. Biol. 2, recombined onto the phage. 597-604. efficiently 20. Greenwood, J., Willis, A. E. & Perham, R. N. (1991)J. Mol. Biol. The display system described here has a number of potential 220, 821-827. uses. First, it should be possible to fuse a library of random- 21. Willis, A. E., Perham, R. N. & Wraith, D. (1993) Gene 128, primed cDNA molecules to the D gene and display the en- 79-83. coded proteins. To search the library for a given epitope using 22. Hart, S. L., Knight, A. M., Harbottle, R. P., Mistry, A., Hunger, an antibody, one could then do a biopanning experiment to H.-D., Cutler, D. F., Williamson, R. & Coutelle, C. (1994)J. Biol. select relevant phage. This has the advantage over current Chem. 269, 12468-12474. Downloaded by guest on September 25, 2021