Anchored Periplasmic Expression, a Versatile Technology for the Isolation of High-Affinity Antibodies from Escherichia Coli-Expressed Libraries

Home , Phage display

Anchored periplasmic expression, a versatile technology for the isolation of high-affinity antibodies from Escherichia coli-expressed libraries

Barrett R. Harvey*, George Georgiou*†‡§, Andrew Hayhurst*†, Ki Jun Jeong*†, Brent L. Iverson*§¶, and Geoffrey K. Rogers†

*Institute for Cellular and Molecular Biology and Departments of †Chemical Engineering, ¶Chemistry and Biochemistry, and ‡Biomedical Engineering, University of Texas, Austin, TX 78712

Edited by James A. Wells, Sunesis Pharmaceuticals, Inc., South San Francisco, CA, and approved May 10, 2004 (received for review January 9, 2004) Anchored periplasmic expression (APEx) is a technology for the fluorescently labeled antigen, and cells exhibiting a desired level isolation of ligand-binding proteins from combinatorial libraries of fluorescence are isolated. With sorting rates of Ͼ400 million anchored on the periplasmic face of the inner membrane of cells per hour, commercial FC machines can be used to screen Escherichia coli. After disruption of the outer membrane by Tris- libraries of the size accessible within the constraints of microbial EDTA-lysozyme, the inner-membrane-anchored proteins readily transformation efficiencies. Furthermore, multiparameter FC bind fluorescently labeled ligands as large as 240 kDa. Fluores- can provide valuable information regarding the function of each cently labeled cells are isolated by flow cytometry, and the DNA of and every library clone in real time, thus helping to guide the isolated clones is rescued by PCR. By using two rounds of APEx, the library construction process and optimize sorting conditions affinity of a neutralizing antibody to the Bacillus anthracis protec- (11, 12). tive antigen was improved >200-fold, exhibiting a final KD of 21 E. coli offers facile expression of recombinant protein and high pM. This approach has several technical advantages compared with DNA transformation efficiencies that allow for efficient large previous library screening technologies, including the unique abil- library production and increased coverage of protein library ity to screen for ligand-binding proteins that bind endogenously sequence space. Previously, we have shown that the outer expressed ligands fused to a short-lived GFP. Further, APEx is able membrane of E. coli can be selectively permeabilized, allowing to display proteins either as an N-terminal fusion to a six-residue the diffusion of fluorescently conjugated antigens into the cell sequence derived from the native E. coli lipoprotein NlpA, or as a where they can bind soluble proteins localized within the C-terminal fusion to the phage gene three minor coat protein of periplasmic space (13). In addition, others have demonstrated M13. The latter fusions allow hybrid phage display͞APEx strategies that association of single-chain variable fragment (scFv) with the without the need for further subcloning. peptidoglycan layer can allow selection of fluorescently labeled antigen via FC (14). However, these approaches are limited to ecombinant antibodies are increasingly being used as ther- small molecule and peptide antigens. Large antigens such as Rapeutic and diagnostic tools over a broad spectrum of proteins cannot be used because conditions that allow the applications, ranging from cancer treatment to microbial infec- accessibility of high molecular weight species to the recombinant tions. Currently, 13 antibodies are FDA approved, and 30 more scFv also result in the destruction of the scFv linkage to the cell. are in late-stage clinical trials (1). At the heart of the new Here, we report a protein library-screening technology, based generation of antibody therapeutics is protein engineering to on anchored periplasmic expression (APEx). In APEx, proteins reduce immunogenicity and increase antigen affinity (2). For are expressed in the periplasm, tethered to the inner membrane example, we have recently reported studies with a series of of E. coli via lipidation of a small N-terminal 6-aa fusion or as antibody fragments produced by directed evolution in which aC-terminal fusion to the N terminus of the M13 phage gene 3 toxin neutralization efficacy correlated with antitoxin antibody minor coat protein (g3p), for FC analysis of clones selected by affinity in an animal model of anthrax intoxication (3). phage display without further subcloning. After chemical͞ Directed evolution involves first the generation of a recom- enzymatic permeabilization of the bacterial outer membrane, E. binant library of protein-expressing clones with randomized coli cells expressing anchored scFv antibodies can be specifically sequences using molecular biology techniques, and second, the labeled with fluorescent antigens, ranging in size up to at least use of screening technologies for the isolation of the protein 240 kDa, and analyzed by FC (Fig. 1). By using APEx, we have variants that exhibit the most enhanced activity. The screening demonstrated the efficient isolation of well expressed antibodies SCIENCES

of large libraries requires a physical link among a gene, the with markedly improved ligand affinities, including an antibody APPLIED BIOLOGICAL protein it encodes, and the desired function. Such a link can be fragment to the protective antigen (PA) of Bacillus anthracis with established by using a variety of in vivo display technologies that an affinity that was increased Ͼ200-fold. Further, we show that have proven to be invaluable mechanistic studies, biotechnolog- fusions between GFP and antigen can be expressed endoge- ical purposes, and proteomics research (4–6). nously and captured by perplasmically anchored scFv. Thus, Display on M13 bacteriophage represents the oldest and after a washing step, cells that express both the fluorescent currently most widely used protein library-screening method (7, antigen and an APEx-anchored scFv are highly fluorescent and 8). Phage display has been used successfully for the isolation of antibodies from human and animal repertoire libraries. An alternative approach utilizes the anchoring of protein libraries on This paper was submitted directly (Track II) to the PNAS office. the surface of bacteria or yeast cells, most commonly Escherichia Abbreviations: APEx, anchored periplasmic expression; FC, flow cytometry; g3p, M13 phage gene 3 minor coat protein; PA, protective antigen; Dig, digoxigenin; SPR, surface plasmon coli and Saccharomyces cerevisiae, respectively. Unlike phage, the resonance; PI, propidium iodide; Meth, methamphetamine; BODIPY, 4,4-difluoro-4-bora- relatively large size of bacteria and yeast allows screening by flow 3{␣},4{␣}-diaza-s-indacene; scFv, single-chain variable fragment; scAb, single-chain Ab cytometry (FC) (9, 10). FC combines high-throughput with fragment. real-time quantitative multiparameter analysis of each library §To whom correspondence may be addressed. E-mail: [email protected] or biverson@ member. For FC screening of antibody libraries, microorganisms mail.utexas.edu. displaying the library are incubated with a limiting amount of a © 2004 by The National Academy of Sciences of the USA

www.pnas.org͞cgi͞doi͞10.1073͞pnas.0400187101 PNAS ͉ June 22, 2004 ͉ vol. 101 ͉ no. 25 ͉ 9193–9198 Downloaded by guest on September 26, 2021 (5Ј-CTGGGCCATGGCCGGCTGGGCCTCGCTGCTAC- TCTGGTCGCAACC-3Ј). The resulting NlpA fragment was used to replace the pelB leader sequence of pMoPac1 (15) via NdeI and SfiI to generate pAPEx1. scFv specific for digoxin (16), B. anthracis PA (PA) (3), and methamphetamine (Meth) (B.R.H., A. Shanafelt, B.L.I., and G.G., unpublished work) were inserted downstream of the NlpA fragment in pAPEx1 via the noncompatible SfiI sites. Corresponding g3p fusions of the scFv were made by cloning the same genes into phage display vector pAK200 (17).

Growth Conditions. E. coli ABLE C (Stratagene) was the host strain used throughout. E. coli transformed with the pAPEx1 or pAK200 derivatives were inoculated in terrific broth (12 g of pancreatic digest of casein͞24 g of yeast extract͞9.4 g of dipo- tassium phosphate͞2.2 g of monopotassium phosphate, pH 7.2) supplemented with 2% glucose and chloramphenicol at 30 ␮ ͞ g ml to an OD600 of 0.1. Cell growth and induction were performed as described (13). After induction, the cellular outer membrane was permeabilized as described (18). Briefly, cells Ϸ (equivalent to 1mlof20OD600) were pelleted and resuspended in 350 ␮l of ice-cold solution of 0.75 M sucrose͞0.1 M Tris⅐HCl, pH 8.0͞100 ␮g/ml hen egg lysozyme. We gently added 700 ␮l of ice-cold 1 mM EDTA, and the suspension was left on ␮ ice for 10 min; 50 l of 0.5 M MgCl2 was added, and the mix was left on ice for a further 10 min. The resulting cells were gently pelleted and resuspended in 1ϫ PBS with 200 nM probe at room temperature for 45 min before evaluation by FC.

Fluorescent Probe. The synthesis of digoxigenin (Dig)–4,4- difluoro-5,7-dimethyl-4-bora-3a,4a-diaza-s-indacene-3-propio- nyl ehtylenediamine (BODIPY) has been described (19). Meth– fluorescein conjugate was a gift from Roche Diagnostics. Purified PA protein, kindly provided by S. Leppla (National Institutes of Health, Bethesda), was conjugated to BODIPY plus FL SE D-2184 at a 1:7 molar ratio according to the manufacturer’s instructions. Unconjugated BODIPY was removed by dialysis. To synthesize Dig-phycoerythrin, R-phycoerythrin, 3-amino- 3-dioxy-Dig hemisuccinamide, and succinimidyl ester (Molecu- lar Probes) were conjugated at a 1:5 molar ratio according to the manufacturer’s instructions. Free Dig was removed by dialysis in excess PBS.

Affinity Maturation of scFv Libraries with FC. Libraries were made from the 14B7 parental scFv by using error-prone PCR with standard techniques (20) and cloned into the pAPEx1 expression vector. Upon transformation, induction, and labeling, the cells were then stained with propidium iodide (PI) (emission 617 nm) to monitor inner membrane integrity. Cells were analyzed on a MoFlo (Cytomation, Fort Collins, CO) droplet deflection flow cytometer by using a 488-nm Argon laser for excitation. Cells Fig. 1. A schematic diagram showing the principle of APEx for the FC-based were selected based on improved fluorescence in the fluores- isolation of high-afﬁnity antibody fragments. cein͞BODIPY fluorescein emission spectrum detecting through a 530͞40 band-pass filter and for the absence of labeling in PI emission detecting through a 630͞40 band-pass filter. can be readily sorted from cells that express either only an scFv E. coli captured after the first sort were immediately resorted or GFP-antigen fusion alone. This feature should be particularly through the flow cytometer. Subsequently, the scFv genes in the useful for high-throughput antibody selections in proteomics sorted cell suspension were amplified by PCR. Once amplified, applications, epitope mapping, or when searching genomes for the mutant scFv genes were then recloned into pAPEx1 vector, interacting pairs of proteins. retransformed into cells, and grown overnight on agar plates at 30°C. The resulting clones were subjected to a second round of Materials and Methods sorting plus resorting as described above, before scFv genes were Recombinant DNA Techniques. The leader peptide and first six subcloned into pMoPac16 (15) for expression of single-chain amino acids of the mature NlpA protein flanked by NdeI and SfiI antibody fragment (scAb) protein. sites were amplified by whole-cell PCR of XL1-Blue (Strat- agene) by using primers BRH#08 (5Ј-GAAGGAGATATAC- Surface Plasmon Resonance (SPR) Analysis. Monomeric scAb pro- ATATGAAACTGACAACACATCATCTA-3Ј) and BRH#09 teins were purified by immobilized metal affinity chromatogra-

9194 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0400187101 Harvey et al. Downloaded by guest on September 26, 2021 phy size-exclusion FPLC as described (15). Affinity measure- ments were obtained via SPR by using a BIACORE 3000 (Biacore, Uppsala) instrument. We coupled Ϸ500 response units of PA to a CM5 chip by using 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide͞N-hydroxy succinimide chemistry. BSA was similarly coupled and used for in-line subtraction. Kinetic analysis was performed at 25°C in Hepes-buffered saline-EP buffer (Biacore) at a flow rate of 100 ␮l͞min. Five 2-fold dilutions of each antibody beginning at 20 nM were analyzed in triplicate. Methods and procedures for the endogenously expressed antigen–GFP fusions can be found in Supporting Text, which is published as supporting information on the PNAS web site. Results APEx Detection of Ligand Binding. For screening applications, an ideal expression system should minimize cell toxicity or growth abnormalities that can arise from the synthesis of heterologous polypeptides. We have devised an expression system that enables the anchoring of proteins on the periplasmic face of the E. coli inner membrane via a lipoprotein targeting motif, without any of the complications that are associated with transmembrane protein fusions (21, 22) (Fig. 1). Unlike membrane proteins, bac- Fig. 2. Examples of targets visualized by APEx. (A) Fluorescence distribution terial lipoproteins are not known to require the signal recogni- of ABLEC cells expressing PA-specific (14B7), scFv (white), and Dig-specific tion particle or YidC pathways for membrane anchoring (23). (Dig) scFv (green), and labeled with 200 nM BODIPY-conjugated fluorescent antigens. Histograms represent the mean fluorescence intensity (M) of 10,000 Lipoproteins are secreted across the membrane via the Sec E. coli events. (B) Histograms of cells expressing 14B7 scFv (white) or Dig scFv pathway and, once in the periplasm, a diacylglyceride group is (green) labeled with 200 nM of the 240-kDa Dig-phycoerythrin conjugate. attached through a thioether bond to a cysteine residue on the C-terminal side of the signal sequence. The signal peptide is then cleaved by signal peptidase II, the protein is fatty acylated at the Library Screening by APEx. The gene encoding the anti-PA 14B7 modified cysteine residue, and finally the lipophilic fatty acid scFv was mutagenized by error-prone PCR and fused to the inserts into the membrane, thereby anchoring the protein (24). NlpA membrane anchoring sequence in an appropriate expres- NlpA is a nonessential E. coli lipoprotein that exclusively sion vector, and the resulting library was transformed into E. coli, localizes to the inner membrane (25, 26). A sequence encoding giving rise to 1 ϫ 107 independent clones. DNA sequencing of the leader peptide and first six amino acids of the mature NlpA 12 library clones selected at random revealed an average of 2% (containing the putative fatty acylation and inner membrane nucleotide substitutions per gene. After induction of NlpA–scFv ␤ targeting sites) was used for anchoring scFv antibodies to the synthesis with isopropyl -D-thiogalactoside, the cells were periplasmic face of the inner membrane (see Materials and treated with Tris–EDTA–lysozyme, washed, and labeled with Methods). Of particular note is the aspartate residue adjacent to 200 nM PA–BODIPY. Inner membrane integrity was monitored ϫ 8 the fatty acylated cysteine residue that is thought to be a by staining with PI. A total of 2 10 bacteria were sorted by consensus residue for inner membrane targeting (25, 27, 28). using an ultrahigh-throughput Cytomation. MoFlo droplet de- NlpA fusions to the 26–10 antidigoxin͞Dig scFv and to the flection flow cytometer selectively gating for low PI fluorescence anti-B. anthracis PA 14B7 scFv were constructed and expressed (630-nm emission) and high BODIPY fluorescence. Approxi- from a lac promoter in E. coli. Note that the presence of only a mately 5% of the cells sorted with the highest 530-nm fluores- 6-aa fusion partner should have minimal influence on scFv cence were collected, immediately restained with PI alone, and folding, so that well expressed antibodies in this fusion format resorted as above. Because no antigen was added during the resort, only cells expressing antibodies with slow dissociation are expected to be well expressed in soluble form in E. coli. kinetics remained fluorescent. The plating efficiency of this After induction of NlpA-[scFv] synthesis using isopropyl ␤-D- population was low, presumably due to a combination of poten- thiogalactoside, the cells were incubated with EDTA and ly- tial scFv expression toxicity (29, 30) and the Tris–EDTA– sozyme to disrupt the outer membrane and the cell wall. The SCIENCES lysozyme treatment. Therefore, to avoid loss of potentially permeabilized cells were mixed with their respective antigens

high-affinity clones, genes encoding scFvs were rescued by PCR APPLIED BIOLOGICAL conjugated to the fluorescent dye BODIPY (200 nM), and the amplification of the DNA collected from Ϸ1 ϫ 104 fluorescent cell fluorescence was determined by FC. Treated cells expressing Ϸ events resorted. It should be noted that the conditions used for the NlpA-[14B7 scFv] and the NlpA-[Dig scFv] exhibited 9- PCR amplification result in the quantitative release of cellular and 16-fold higher mean fluorescence intensity, respectively, DNA from the cells that have partially hydrolyzed cell walls due compared to controls (Fig. 2A). to the Tris–EDTA–lysozyme treatment during labeling. After 30 To evaluate further the ability of antibody fragments anchored rounds of PCR amplification, the DNA was ligated into pAPEx1 on the cytoplasmic membrane to bind bulky antigens, we exam- and transformed into fresh E. coli. A second round of sorting was ined the ability of the NlpA-[Dig scFv] to recognize Dig conju- performed exactly as above, except that in this case only the most gated to the 240-kDa fluorescent protein phycoerythrin (PE). fluorescent 2% of the population was collected and then imme- The conjugate was mixed with cells expressing NlpA-[Dig scFv] diately resorted without relabeling. and treated with EDTA-lysozyme. A high cellular fluorescence The scFv DNA from the second round was amplified by PCR was observed, indicating binding of Dig-PE conjugate by the and ligated into pMoPac16 (15) for expression of the antibody membrane anchored antibody (Fig. 2B). Labeling with Dig-PE fragments in soluble form in the scAb format. The scAb antibody followed by one round of FC resulted in a Ͼ500-fold enrichment fragment is comprised of an scFv in which the light chain is fused of bacteria expressing NlpA-[Dig scFv] from cells expressing a to a human ␬ constant region (30, 31). We picked 20 clones in similar fusion with an scFv having unrelated antigen specificity. the scAb format at random, and they were grown in small-scale

Harvey et al. PNAS ͉ June 22, 2004 ͉ vol. 101 ͉ no. 25 ͉ 9195 Downloaded by guest on September 26, 2021 Fig. 3. Analysis of anti-PA antibody fragments selected using APEx. (A) SPR analysis of anti-PA scAb binding to PA. (B) Table of afﬁnity data acquired by SPR. (C) FC histograms depicting the mean ﬂuorescence (FL) intensity (M) of E. coli expressing anti-PA scFv clones in pAPEx1 and labeled with 200 nM PA-BODIPY conjugate as compared with those expressing the anti-Meth scFv as a negative control.

shake-flask cultures. After induction with isopropyl ␤-D- a further increase in affinity, as well as significantly improved thiogalactoside, the periplasmic fractions were isolated, and the expression levels. It is noteworthy that the M5, M6, and M18 scAb proteins were rank-ordered with respect to their relative were isolated after a single round of asexual PCR, yet they all had antigen dissociation kinetics using SPR analysis. Of the 20 higher affinity relative to the best antibody that could be isolated clones, 11 exhibited slower antigen dissociation kinetics com- by phage display, even after multiple rounds of sexual mutagen- pared to the 14B7 parental antibody. The three scAbs with the esis and selection. slowest antigen dissociation kinetics were produced in large scale M18, the highest-affinity clone isolated by APEx, contained and purified as monomers through a combination of Ni chro- the S56P mutation but lacked the Q55L substitution found in 1H, matography followed by gel-filtration FPLC. All of the library- M5, and M6. When the Q55L substitution was introduced into selected clones exhibited excellent expression characteristics and M18 by site-specific mutagenesis, the resultant scAb exhibited a ϭ resulted in yields of 4–8 mg of purified protein per liter in further improvement in antigen binding (KD 21 pM) with a kon ϫ 6 Ϫ1⅐ Ϫ1 ϫ Ϫ5 Ϫ1 shake-flask culture. Detailed SPR analysis indicated that all of 1.1 10 M sec and a koff of 2.4 10 sec , corre- three clones exhibit a substantially lower KD value for PA sponding to a complex half life of almost 12 h. However, the compared with the parental 14B7 antibody (Fig. 3 A and B). The introduction of this mutation reduced the yield of purified Ͼ ͞ improved KD values result primarily from slower antigen disso- protein 5-fold to 1.2 mg liter in shake-flask culture. ciation (i.e., slower koff). The highest-affinity clone, M18, exhib- ϫ Ϫ5⅐ Ϫ1⅐ Ϫ1 ited a KD of 35 pM, with a koff of 4.2 10 M sec , APEx of Phage Displayed scFv Antibodies. Numerous antibody corresponding to an M18-PA half life of 6.6 h. This represents fragments to important therapeutic and diagnostic targets have a Ͼ120-fold affinity improvement compared to the parental been isolated from repertoire libraries screened by phage dis- ϭ antibody 14B7 (KD 4.3 nM, as determined by BIACORE play. Such repertoire libraries are an important resource in 3000). antibody discovery efforts. Antibodies are most commonly dis- The fluorescence intensity of Tris–EDTA–lysozyme perme- played on filamentous phage via C-terminal fusion to the abilized cells expressing NlpA fusions to the isolated mutant N-terminus of the g3p (33). During phage morphogenesis, g3p antibodies varied in proportion to their respective antigen- becomes transiently attached to the inner membrane via its binding affinity in solution (Fig. 3C). After incubation in excess extreme C terminus, before it can be incorporated onto the buffer for 1.5 h after labeling with the fluorescent probe, cells growing virion (34). We evaluated whether g3p fusion proteins expressing the NlpA-[M18 scFv] protein displayed a mean can be exploited for antibody library screening purposes using fluorescence of 220, and antibodies with intermediate affinities the APEx format. The high-affinity anti-PA M18 scFv discussed displayed intermediate fluorescence intensities. above, the antidigoxin͞Dig 26–10 scFv, and an anti-Meth scFv The three clones analyzed in detail, M5, M6, and M18, (Meth) were cloned in-frame to the N terminus of g3p down- contained 7-, 12-, and 11-aa substitutions, respectively. In earlier stream from a lac promoter in phagemid pAK200 (see Support- studies using phage display (3), we had isolated a variant of the ing Text), which is widely used for phage display purposes and 14B7 scFv by three cycles, each consisting of (i) mutagenic error utilizes a short variant of gene III for g3p display (17). After prone PCR, (ii) five rounds of phage panning, and (iii) DNA induction with isopropyl ␤-D-thiogalactoside, cells expressing shuffling of the postpanning clones. The best clone isolated in scFv-g3p fusions were permeabilized by Tris–EDTA–lysozyme that study, 1H, contained Q55L and S56P substitutions and and labeled with the respective fluorescent antigens (Fig. 4). exhibited a KD of 150 pM. These two mutations likely increase High fluorescence was obtained for all three scFvs only when the hydrophobicity of the binding pocket adding to the mounting incubated with their respective antigens. Significantly, the mean evidence that an increase in hydrophobic interactions are often fluorescence intensities of the scFvs fused to the N terminus of a dominant effect in antibody affinity maturation (32). The same g3p were comparable to those obtained by fusion to the C amino acid substitutions are also found in the M5 and M6 clones terminus of the NlpA anchor. The results in Fig. 4 demonstrate isolated by APEx. However, the presence of the additional that (i) antibody fragments cloned into phagemids for display on mutations beyond Q55L and S56P in these two clones conferred filamentous phage can be readily analyzed by FC using the APEx

9196 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0400187101 Harvey et al. Downloaded by guest on September 26, 2021 Fig. 5. Fluorescent cell labeling using endogenously expressed antigen–GFP fusions. Histograms depict 10,000 E. coli events expressing (i) GFP–peptide fusion alone (mean ϭ 13), (ii) GFP–peptide coexpressed with Dig scFv (mean ϭ 15), (iii) GFP without peptide fusion coexpressed with 7C2 antipeptide scFv (mean ϭ 14), and (iv) GFP–peptide coexpressed with 7C2 antipeptide scFv (mean ϭ 46).

Thus, in this case, upon treatment with Tris–EDTA–lysozyme, the cells retain their fluorescence. This concept was demonstrated by using the 7C2 scFv that recognizes the peptide antigen (CFTFKEFQNNPNPRSLVK) from the MacI protein with a KD of 142 nM. After Tris–EDTA–lysozyme treatment, cells expressing NlpA-[7C2 scFv] together with the ssTorA peptide antigen– GFP–SsrA fusion exhibited a 3-fold higher fluorescence compared with control cells expressing (i) the ssTorA peptide antigen–GFP–SsrA fusion alone, (ii) the ssTorA peptide antigen–GFP–SsrA fusion coexpressed with an NlpA-fused irrele- vant scFv, or (iii) ssTorA–GFP–SsrA without peptide antigen coexpressed with an NlpA-[7C2 scFv] (Fig. 5). Even though the fluorescence signal obtained was not as high compared with that Fig. 4. N- vs. C-terminal anchoring strategy comparison. Data represent the mean fluorescence intensity (M) of 10,000 E. coli events. (A) Anti-Dig scfv, obtained with exogenous fluorescent antigens, it is sufficient for anti-PA M18 scFv, and anti-Meth scFv expressed as N-terminal fusions in the library screening purposes. A higher signal may be obtained in pAPEx1 vector in E. coli specifically label with 200 nM of their respective strains that afford more efficient degradation of GFP–SsrA in antigen. (B) C-terminal fusions of same scFv in pAK200 vector specifically the cytoplasm or by using leader peptides optimized for GFP labeled with 200 nM of their respective antigen. export (36). Discussion format, and (ii) scFv antibodies can be anchored on the cyto- We have developed a FC-based method using bacterial expres- plasmic membrane either as N- or C-terminal fusions without sion for the efficient selection of high-affinity ligand-binding loss of antigen binding. proteins, and specifically scFv antibodies, from combinatorial libraries. APEx is based on the anchoring of proteins to the Fluorescent Cell Labeling Using Endogenously Expressed Antigen–GFP periplasmic side of the inner membrane, followed by disruption Fusions. We have developed a strategy that capitalizes on APEx of the outer membrane before incubation with fluorescently for labeling by antigens expressed endogenously as fusions to labeled antigen and FC sorting. This strategy offers several short-lived GFP. GFP, tagged on the C terminus with the SsrA advantages over previous bacterial periplasmic and surface peptide, is targeted for rapid degradation by the powerful ClpXP display approaches. (i) APEx is an E. coli-based system and proteolytic machinery of the bacterial cytoplasm (35). Fusion of therefore provides an easy route to the creation of large libraries GFP–SsrA to the ssTorA leader that directs protein export via by transformation and preparative protein expression of isolated SCIENCES twin arginine transporter pathway enables export of ssTorA- antibodies. (ii) By using a fatty acylated anchor to retain the APPLIED BIOLOGICAL GFP–SsrA protein into the periplasm where it sequestered away protein in the inner membrane, a fusion as short as 6 aa is all that from the ClpXP protease and rescued from proteolytic degra- is required for display. The short fusion is unlikely to influence dation (36). As a result, cells expressing ssTorA–GFP–SsrA the affinity or expression characteristics of the isolated proteins. exhibit high fluorescence. However, the fluorescence is lost (iii) The inner membrane lacks molecules such as LPS or other when the outer membrane is permeabilized and the ssTorA– complex carbohydrates that can sterically interfere with large GFP–SsrA fusion can escape into the extracellular fluid. antigen binding to displayed polypeptides. (iv) The fusion must We constructed fusions consisting of the sequence of a peptide only traverse one membrane before it is displayed, and therefore antigen recognized by a scFv antibody sandwiched between the biosynthetic limitations that might restrict the export of certain ssTorA leader peptide and GFP–SsrA (see Supporting Text). The sequences to the yeast or bacterial surface may be circumvented. ssTorA peptide antigen–GFP–SsrA protein chimera is exported (v) Display is accomplished by using either N- or C-terminal into the periplasm rendering the cells fluorescent, but after outer fusion. (vi) APEx can be used directly for proteins expressed membrane permeabilization, the fluorescence is lost. However, from widely used phage display vectors. This latter point is in cells that also express an APEx-displayed antibody specific to particularly important because it enables the use of the many the peptide antigen, the ssTorA peptide antigen–GFP–SsrA available phage display antibody fragment libraries, along with protein chimera becomes associated with the cell via its inter- hybrid library screening strategies, in which clones from a phage action with the membrane-tethered scFv antibody fragment. panning experiment can be quantitatively analyzed or sorted

Harvey et al. PNAS ͉ June 22, 2004 ͉ vol. 101 ͉ no. 25 ͉ 9197 Downloaded by guest on September 26, 2021 further by FC without the need for any subcloning steps. Finally, contrast, the short (6-aa) sequence required for N-terminal (vii) APEx provides a means for the simultaneous expression of tethering of proteins onto the cytoplasmic membrane in APEx fluorescent antigen and antibodies within the same cell. This is unlikely to affect the expression characteristics of the fusion. latter approach, which is likely to be particularly important for Consistent with this hypothesis, all three affinity-enhanced peptide antigens, circumvents time-consuming processes for clones to the anthrax PA toxin isolated by APEx exhibited synthesis, purification, and conjugation of preparative amounts excellent soluble expression characteristics despite having nu- of probe, as is required when the fluorescent antigen is incubated merous amino acid substitutions. Similarly, well expressing with the library. clones have been obtained in the affinity maturation of a Meth APEx can be used for the detection of antigens ranging from antibody (B.R.H., A. Shanafelt, B.L.I., and G.G., unpublished small molecules (e.g., Dig and Meth, Ͻ1 kDa) to phycoerythrin work), suggesting that the isolation of clones that can readily be conjugates (240 kDa). In fact, the phycoerythrin conjugate used produced in soluble form in bacteria on a large scale may be an in Fig. 2B is not meant to define an upper limit for antigen intrinsic feature of APEx selections. detection, because larger proteins have not yet been tested. In this study, we used APEx for affinity maturation purposes In the present study, genes encoding scFvs that bind the and engineered scFvs to the B. anthracis PA exhibiting KD values fluorescently labeled antigen were rescued from the sorted cells as low as 21 pM. The scFv-binding site exhibiting the highest by PCR. An advantage of this approach is that it enables the affinity for PA has been humanized and converted to full-length isolation of clones that may not be viable because of the IgG, and its neutralizing potential to anthrax intoxication is combination of potential scFv toxicity and Tris–EDTA– being evaluated in preclinical studies. In addition to affinity lysozyme disruption. Yet another advantage of PCR rescue is maturation, APEx can be exploited for several other protein that the amplification of DNA from pooled cells can be carried engineering applications, including the selection of enzyme out under mutagenic conditions before subcloning. Thus, after variants with enhanced function, because the cell envelope each round of selection, random mutations can be introduced provides sites for retention of enzymatic catalytic products, into the isolated genes, simplifying further rounds of directed thereby enabling selection based directly on catalytic turnover evolution (37). Further, PCR conditions that favor template (38), or for the analysis of membrane protein topology, whereby switching among the protein-encoding genes in the pool may be a scFv antibody anchored in a periplasmic loop is able to bind used during the amplification step to allow recombination fluorescent antigen and serves as a fluorescent reporter. among the selected clones. An important issue with any library screening technology is We thank Dr. Stephen Leppla for providing the purified PA protein and the ability to express isolated clones at a high level. Existing Dr. Armen Shanafelt for providing the anti-Meth hybridoma and Meth–fluorescein conjugate. This work was supported by the U.S. Army display formats involve fusion to large anchoring sequences, Army Research Office͞Multidisciplinary University Research Initiative which can influence the expression characteristics of the dis- program and, in connection with contract no. DADD17-01-D-0001, by played proteins. For this reason, scFvs that display well as fusions the U.S. Army Research Laboratory. K.J.J. was supported by a post- in phage, yeast, or bacteria may not necessarily be amenable to doctoral fellowship from the Korea Science and Engineering Foundation high expression in soluble form as nonfusion proteins (15). In (KOSEF).

1. Hudson, P. J. & Souriau, C. (2003) Nat. Med. 9, 129–134. 20. Fromant, M., Blanquet, S. & Plateau, P. (1995) Anal. Biochem. 224, 347–353. 2. Brekke, O. H. & Loset, G. A. (2003) Curr. Opin. Pharmacol. 3, 544–550. 21. Mingarro, I., von Heijne, G. & Whitley, P. (1997) Trends Biotechnol. 15, 3. Maynard, J. A., Maassen, C. B., Leppla, S. H., Brasky, K., Patterson, J. L., 432–437. Iverson, B. L. & Georgiou, G. (2002) Nat. Biotechnol. 20, 597–601. 22. Miroux, B. & Walker, J. E. (1996) J. Mol. Biol. 260, 289–298. 4. Hoess, R. H. (2001) Chem. Rev. 101, 3205–3218. 23. Samuelson, J. C., Chen, M., Jiang, F., Moller, I., Wiedmann, M., Kuhn, A., 5. Hayhurst, A. & Georgiou, G. (2001) Curr. Opin. Chem. Biol. 5, 683–689. Phillips, G. J. & Dalbey, R. E. (2000) Nature 406, 637–641. 6. Wittrup, K. D. (2000) Nat. Biotechnol. 18, 1039–1040. 24. Pugsley, A. P. (1993) Microbiol. Rev. 57, 50–108. 7. Rodi, D. J. & Makowski, L. (1999) Curr. Opin. Biotechnol. 10, 87–93. 25. Yamaguchi, K., Yu, F. & Inouye, M. (1988) Cell 53, 423–432. 8. Smith, G. P. (1985) Science 228, 1315–1317. 26. Yu, F., Inouye, S. & Inouye, M. (1986) J. Biol. Chem. 261, 2284–2288. 9. Feldhaus, M. J., Siegel, R. W., Opresko, L. K., Coleman, J. R., Feldhaus, J. M., 27. Seydel, A., Gounon, P. & Pugsley, A. P. (1999) Mol. Microbiol. 34, 810–821. Yeung, Y. A., Cochran, J. R., Heinzelman, P., Colby, D., Swers, J., et al. (2003) 28. Yakushi, T., Masuda, K., Narita, S., Matsuyama, S. & Tokuda, H. (2000) Nat. Nat. Biotechnol. 21, 163–170. Cell Biol. 2, 212–218. 10. Georgiou, G. (2000) Adv. Protein Chem. 55, 293–315. 29. Somerville, J. E., Jr., Goshorn, S. C., Fell, H. P. & Darveau, R. P. (1994) Appl. 11. Boder, E. T. & Wittrup, K. D. (2000) Methods Enzymol. 328, 430–444. Microbiol. Biotechnol. 42, 595–603. 12. Daugherty, P. S., Chen, G., Iverson, B. L. & Georgiou, G. (2000) Proc. Natl. 30. Hayhurst, A. (2000) Protein Expr. Purif. 18, 1–10. Acad. Sci. USA 97, 2029–2034. 31. McGregor, D. P., Molloy, P. E., Cunningham, C. & Harris, W. J. (1994) Mol. 13. Chen, G., Hayhurst, A., Thomas, J. G., Harvey, B. R., Iverson, B. L. & Immunol. 31, 219–226. Georgiou, G. (2001) Nat. Biotechnol. 19, 537–542. 32. Li, Y., Li, H., Yang, F., Smith-Gill, S. J. & Mariuzza, R. A. (2003) Nat. Struct. 14. Fuchs, P., Weichel, W., Dubel, S., Breitling, F. & Little, M. (1996) Immuno- Biol. 10, 482–488. technology 2, 97–102. 33. Barbas, C. F., 3rd, Kang, A. S., Lerner, R. A. & Benkovic, S. J. (1991) Proc. 15. Hayhurst, A., Happe, S., Mabry, R., Koch, Z., Iverson, B. L. & Georgiou, G. Natl. Acad. Sci. USA 88, 7978–7982. (2003) J. Immunol. Methods 276, 185–196. 34. Boeke, J. D. & Model, P. (1982) Proc. Natl. Acad. Sci. USA 79, 5200–5204. 16. Chen, G., Dubrawsky, I., Mendez, P., Georgiou, G. & Iverson, B. L. (1999) 35. Gottesman, S., Roche, E., Zhou, Y. & Sauer, R. T. (1998) Genes Dev. 12, Protein Eng. 12, 349–356. 1338–1347. 17. Krebber, A., Bornhauser, S., Burmester, J., Honegger, A., Willuda, J., 36. DeLisa, M. P., Samuelson, P., Palmer, T. & Georgiou, G. (2002) J. Biol. Chem. Bosshard, H. R. & Pluckthun, A. (1997) J. Immunol. Methods 201, 35–55. 277, 29825–29831. 18. Hayhurst, A. & Harris, W. J. (1999) Protein Expr. Purif. 15, 336–343. 37. Hanes, J. & Pluckthun, A. (1997) Proc. Natl. Acad. Sci. USA 94, 4937–4942. 19. Daugherty, P. S., Olsen, M. J., Iverson, B. L. & Georgiou, G. (1999) Protein Eng. 38. Olsen, M. J., Stephens, D., Griffiths, D., Daugherty, P., Georgiou, G. & 12, 613–621. Iverson, B. L. (2000) Nat. Biotechnol. 18, 1071–1074.

9198 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0400187101 Harvey et al. Downloaded by guest on September 26, 2021