Construction of a Large Synthetic Human Fab Antibody Library on Yeast Cell Surface by Optimized Yeast Mating

Abstract

Yeast surface-displayed antibody libraries provide an efficient and quantitative screening resource for given antigens, but suffer from typically modest library sizes owing to low yeast transformation efficiency. Yeast mating is an attractive method for overcoming the limit of yeast transformation to construct a large, combinatorial antibody library, but the optimal conditions have not been reported. Here, we report a large synthetic human Fab (antigen binding fragment) yeast surface-displayed library generated by stepwise optimization of yeast mating conditions. We first constructed HC (heavy chain) and LC (light chain) libraries, where all of the six CDRs (complementarity-determining regions) of the variable domains were diversified mimicking the human germline antibody repertoires by degenerate codons, onto single frameworks of VH3-23 and $V{\kappa}1$-16 germline sequences, in two haploid cells of opposite mating types. Yeast mating conditions were optimized in the order of cell density, media pH, and cell growth phase, yielding a mating efficiency of ~58% between the two haploid cells carrying HC and LC libraries. We constructed two combinatorial Fab libraries with CDR-H3 of 9 or 11 residues in length with colony diversities of more than $10^9$ by one round of yeast mating between the two haploid HC and LC libraries, with modest diversity sizes of ${\sim}10^7$. The synthetic human Fab yeast-displayed libraries exhibited relative amino acid compositions in each position of the six CDRs that were very similar to those of the designed repertoires, suggesting that they are a promising source for human Fab antibody screening.

Introduction

Synthetic antibody libraries are very useful for bypassing several limitations of immunization and hybridoma technology to isolate highly selective, high affinity antibodies to given antigens [19]. Various in vitro display technologies, such as phage display, yeast surface display, and ribosome display, have been developed for generation of human antibody libraries in various formats, including single domains of variable heavy chain (VH) or variable light chain (VL), single chain variable fragment (scFv), antigen binding fragment (Fab), and even full-length immunoglobulin G (IgG) [21]. Each display technology has pros and cons in terms of library size, screening speed and capacity, and displayable antibody formats.

Yeast surface display technology takes advantage of the disulfide bond-mediated linkage of secreted Aga2p with Agalp covalently attached to the cell wall via a phosphatidylinositol glycan tail, leading to anchoring of the Aga2p-fused passenger proteins on the cell surface [3, 5]. The technology has been extensively used in antibody library construction in the scFv [8], Fab [29], and even IgG [21] formats. The advantages of yeast surface-displayed antibody libraries over other display technologies, such as phage and ribosome display, include no loss of clonal diversity during amplification and expansion; quantitative screening of expression; stability; antigen binding properties of the antibody; and quality control of antibody expression in eukaryotic systems [5]. However, yeast surface display technology has a critical limit in the realizable library size (typically 106 ~ 108) compared with phage (109 to 1011) [30] and ribosome (1011 to 1012) display technologies [21] owing to limitations in yeast transformation efficiencies [20]. Multiple transformations and improved yeast transformation methods could overcome this limitation [1]. However, it is a time-consuming and labor-intensive process. In this respect, yeast mating can be used as a powerful tool for generating a large antibody library. The yeast mating was achieved by cellular fusion between two haploid cells of opposite mating types through interaction within sexual adhesion molecules such as a-agglutinin of MATa cells and α-agglutinin of MATα cells [17]. After mating, two distinct plasmids in each haploid cell are combined into one diploid cell, expressing simultaneously the encoded proteins from each plasmid in the subsequent diploid cells.

Fab antibody fragments comprise two chains; a heavy chain (HC) with VH and CH1 (the first domain of heavy chain constant regions) and a light chain (LC) with VL and CL (light chain constant domain). Thus, yeast mating is suitable for the construction of a combinatorial Fab library from two haploid cells of opposite mating types containing HC and LC libraries. The heterodimerization of secreted LC to yeast surface-anchored HC by formation of a disulfide bond between the two C-terminal Cys residues of CH1 and CL facilitates the construction of the display Fab without using any artificial physical linkers on the yeast cell surface. This concept was successfully verified by the combinatorial Fab library generation on yeast cell surfaces [29] and affinity maturation of a Fab antibody through light chain shuffling [2, 18]. Despite the usefulness of yeast mating for the combinatorial antibody library construction and antibody engineering, there have been few studies describing optimal mating procedures for generation of a large antibody library. Furthermore, only a human-naïve immune Fab library has been reported via yeast mating [29], highlighting the necessity of exploring yeast mating for different types of Fab libraries, such as a synthetic antibody library. In this study, we report optimized yeast mating conditions and yeast surface-displayed human synthetic Fab libraries with a large diversity (>109), generated via one round of yeast mating between two HC and LC library-containing haploid cells of modest library size (~107).

 

Materials and Methods

Yeast Strains and Media

Yeast strains Saccharomyces cerevisiae JAR200 (MATa, GAL1-AGA1::KanM×4ura3D45, ura3–52 trp1 leu2D1 his3D200 pep4::HIS3 prb1D1.6R can1) and YVH10 (MATα, PDI::ADHII-PDI-Leu2 ura3-52 trp1 leu2D1 his3D200 pep4::HIS3 prb1D1.6p can1 GAL) were provided by Prof. Dane Wittrup (Massachusetts Institute of Technology, USA) [5]. JAR200 was derived from the Aga1p overexpressing strain EBY100 with mating type a (MATa) by inserting the KanM×4 gene conferring resistance to G418 into the chromosomal URA3 gene, giving three auxotrophic markers (Leu−, Trp−, and Ura−), and was used as a host for the cell surface display of the HC library [29]. YVH10, a BJ5464 haploid derivative with mating type α (MATα) and two auxotrophic markers (Trp− and Ura−), was used as a host for secretion of the LC library [18]. The standard yeast media of YPD (20 g/l dextrose, 20 g/l peptone, and 10 g/l yeast extract in deionized H2O) and SDCAA (-Ura, -Trp) (6.7 g/l yeast nitrogen base, 5 g/l casamino acids, 5.4 g/l Na2HPO4, 8.56 g/l NaH2PO4·H2O, and 20 g/l dextrose in deionized water) were used [5, 12]. The expression of the HC and LC libraries was induced in the selective medium of SGCAA, which contained the same composition as SDCAA, except glucose was replaced with galactose [12, 15]. For the selective media, 0.0004 g/l tryptophan and/or 0.0002 g/l uracil was added to SDCAA and SGCAA. To adjust media to pH 4.5, 14.7 g/l sodium citrate and 4.29 g/l citric acid monohydrate were added. For the plate, 15 g/l agar was added to the media. The media were sterilized by autoclave. The number of yeast cells was estimated by measurement of optical density at 600 nm (OD600) (an absorbance of 1 at 600 nm is about 1 × 107 cells/ml) [5].

Construction of HC and LC Expression Vectors

For the construction of the yeast surface display plasmid of HC, we modified the CEN-based shuttle vector pRS414 with TRP1 marker (Invitrogen) to express HC with its C-terminal c-myc tag and sequentially Aga2p under the GAL1 promoter, generating the pYDS-H vector (Supplementary Fig. S1). HC composed of VH, CH1 with the Cys residues intact at the last position, and upper hinge region with two Cys substituted with Ser (SPPSP) of antitumor necrosis factor α (TNFα) monoclonal antibody (mAb) adalimumab (Humira) was cloned using the unique restriction sites of NheI/BamHI [13]. We introduced an ApaI site between VH and CH1 by silent mutation so that the VH region could be digested by NheI and ApaI to insert the VH gene library. For the construction of the secretion plasmid of LC, we modified the CEN-based shuttle vector pRS416 with URA3 marker (Invitrogen) to express the LC of adalimumab with its C-terminal Flag tag under the GAL1 promoter, generating a pYDS-K vector (Supplementary Fig. S1). We introduced a BsiWI site between VL and CL by silent mutation so that the VL region could be digested by NheI and BsiWI to insert the VL gene library.

Preparation of Synthetic VH and VL Gene Libraries

VH and VL gene libraries were constructed by serial overlapping polymerase chain reaction (PCR) with degenerative primers designed to introduce mutations on the three CDRs of the respective single framework, VH3-23 and Vκ1-16, respectively. The primers used are l isted in the Supplementary Table S1. The designed single frameworks of VH3-23 and Vκ1-16 were prepared by DNA synthesis (Bioneer, Korea). PCRs were performed with compositions of 25 ng template DNA, 200 μM deoxyribonucleotide triphosphate (dNTP), 0.25 μM forward primer, 0.25 μM reverse primer, 1× PCR buffer, and 1.25 units of speed pfu polymerase (Nanohelix, Korea) in a 50 μl total reaction volume. All of the PCRs were carried out in less than 20 cycles to avoid possible redundancy of the library. Each fragment amplified by proper primer sets was purified after electrophoresis on 2% agarose gel. The purified fragments were mixed at equal molar ratio (10 nM) for the subsequent overlapping PCR to prepare full-length VH and VL gene library inserts. The VH/VL inserts were purified from 1.5% agarose gels and concentrated using Pellet Paint (Novagen) [12, 15]. The DNA concentration was measured by absorbance at 260 nm in a spectrophotometer.

Transformation of HC and LC Libraries into Haploid Strains

Yeast cells that were competent for transformation by electroporation were prepared by the protocol described by Chao et al. [5]. The amplified VH gene library (5 μg) and pYDS-H vector linearized by NheI/ApaI digestion (1 μg) were co-transformed ten times in parallel into the JAR200 strain by homologous recombination technique using a Bio-Rad Gene Pulser electroporation apparatus [5, 12, 15]. Likewise, the VL gene library (5 μg) and NheI/BsiWI digestion-linearized pYDS-K vector (1 μg) were co-transformed ten times in parallel into the YVH10 strain. The transformants were pooled and propagated directly in liquid selective media (i.e., HC library-transformed JAR200 cells in SDCAA+Ura and LC library-transformed YVH10 cells in SDCAA+Trp). Library size was determined by plating serial 10-fold dilutions of the transformed cells on the selective agar plates. Plasmid DNA was recovered by using the Zymoprep yeast plasmid miniprep kit (Zymo Research Co., CA, USA) and sequenced using standard techniques [12].

Fab Library Construction on Diploid Cells by Yeast Mating

We first tested three yeast mating parameters sequentially to determine optimal yeast mating conditions for the HC- and LC-containing haploid cells: cell density, media pH, and cell growth phase. We then generated a combinatorial Fab library on diploid cells under the optimized mating conditions. The HC-containing JAR200 (MATa) and the LC-containing YVH10 (MATα) haploid cells were grown overnight at 30℃ and 250 rpm in 500 ml of the selective medium. Then the haploid cells with 10-fold excess of library size were inoculated at the initial cell OD600 ≈ 0.1 to freshly prepared selective media (adjusted to pH 6.4 or 4.5) and grown until they reached certain optical densities (OD600 ≈ 0.25 for early-exponential, ≈ 1.2 for mid-exponential, ≈ 5.0 for late-exponential, and ≈ 10 for stationary phase). Mating of the two grown haploid cells was performed by mixing equal numbers of cells by vortexing, spreading them on YPD agar plates (adjusted to pH 6.4 or 4.5), with a cell density ranging from 1 × 107 to 9 × 107 cells/cm2, and incubating at 30℃ for 6 h. During plating, the plates were gently shaken to spread the cells homogeneously. Cells were collected by gentle scraping with 50 ml of the double-selective SDCAA medium and pelleted by centrifugation (2,500 ×g for 3 min). The cells were washed three times by resuspension with 40 ml of sterilized cold deionized water and centrifugation to remove remaining media components. The cells were then inoculated to SDCAA media at the very low cell density of OD600 = 0.1 and grown at 30℃ for 24 h to enrich the diploid cells by outgrowing over unmated haploid cells. To estimate the mating efficiency, the washed cells were resuspended in a total volume of 10 ml of SDCAA liquid media, serially diluted, and spread out onto double-selective SDCAA and single-selective SDCAA+Trp agar plates. Plates were incubated at 30℃ for 3 days and the number of colonies was counted. Percentage mating efficiency was calculated as the number of diploid colonies in the double-selective SDCAA plates divided by the number of total colonies in the single-selective SDCAA+Trp plates [2].

Yeast Colony PCR for Sequencing VH and VL

A single colony randomly chosen from diploid cells grown on the SDCAA plate was transferred to PCR tubes containing 20 μl of 0.1% sodium dodecyl sulfate (SDS), and heated at 95℃ for 5 min. The sample was then transferred to a 1.5 ml Eppendorf tube and pelleted using a table centrifuge set as 13,000 rpm for 10 sec. Then 2 μl of the supernatant was mixed with 200 μM dNTP, 0.25 μM forward primer (scFv-H-F or scFv-K-F), 0.25 μM reverse primer (scFv-H-R or scFv-K-R), 1× PCR buffer, and 1.25 units of speed pfu polymerase (Nanohelix, Korea) in a 50 μl total reaction volume. The primer sequences are listed in Supplementary Table S2. PCR was carried out (95℃, 5 min; 30 cycles of 30 sec at 95℃, 30 sec at 55℃, and 20 sec at 72℃; 7 min at 72℃ and kept at 4℃) and the VH and VL fragments were purified after agarose gel electrophoresis. The purified VH and VL fragments were mixed at equivalent molar ratio, and overlapping PCR was performed under the above conditions to generate scFv. The scFv fragments were sequenced with internal primers (for VH: 5’-GAG AGC TGA AGA CAC AGC TG-3’; for VL: 5’-CAA CAA TTT AGG AGC CTT TCC AG-3’).

Fab Library Analysis

Diploid yeast cells carrying the Fab library grown at 30℃ overnight in SDCAA were transferred at the initial OD600 = 0.5 to the induction media of SGCAA and grown at 18℃, 20℃, or 25℃ for 48 h in a shaking incubator set as 250 rpm. The cell-surface expression levels of the Fab libraries were determined using flow cytometry by immunofluorescence labeling of the LC C-terminal Flag tag with anti-Flag mAb (F-3165; Sigma) at 25℃ for 30 min. Cells were washed with PBSB (phosphate-buffered saline, pH 7.4, plus 0.1% bovine serum albumin) and secondarily labeled with fluorescein isothiocyanate (FITC)-labeled anti-mouse mAb (F-0257; Sigma) at 4℃ for 20 min. After washing quickly with PBSB, the cells were subjected to flow cytometric analysis. For the double labeling of Fab expression and antigen binding, cells were primarily labeled with anti-Flag mAb and biotinylated antigen and then secondarily labeled with FITC-labeled anti-mouse mAb and SA-PE (S-860; Invitrogen) prior to the flow cytometric analysis [12, 15].

Statistical Analysis

Data are reported as the mean ± standard deviation (SD) of at least three independent experiments performed in triplicate, unless otherwise specified. Statistical significance was analyzed by a 2-tailed unpaired Student’s t-test using Excel (Microsoft Inc.) [6]. A P value of less than 0.05 was considered statistically significant.

 

Results and Discussion

Construction of HC and LC Expression Vectors for Yeast Surface Display of Fab

For yeast surface display of Fab by assembly of independently expressed HC (VH-CH1) with LC (VL-CL), we first generated two vectors pYDS-H and pYDS-K expressing HC and LC of anti-TNFα mAb adalimumab [13], respectively. The plasmid pYDS-H was designed based on pRS414 to express HC in the C-terminally fused c-myc tag and Aga2p cell surface anchor protein under the GAL1 promoter (Supplementary Fig. S1). The C-terminal Aga2p facilitates cell surface display of HC with the antigen binding sites (VH) away from the cell wall surface [29]. The plasmid pYDS-K was designed based on pRS416 to secrete LC with its C-terminal Flag tag under GAL1 promoter (Supplementary Fig. S1). The two vectors commonly have a CEN6/ARS4 replication origin, which can helps to stabilize one or two copies of each vector in one haploid or diploid cell [22].

The newly generated pYDS-H and pYDS-K vectors were transformed into JAR200 (MATa) and YVH10 (MATα), respectively. The two transformed haploid cells were mated to selected diploid cells for Fab expression. Fluorescence-activated cell sorting (FACS) analysis revealed that the diploid cells exhibited simultaneously two positive detections of the Flag tag fused at the C-terminal LC, and antigen binding to TNFα in a concentration-dependent manner (Supplementary Fig. S2). These results demonstrated that the secreted LC correctly assembled with the Aga2p-fused HC to form Fab with endogenous antigen binding activity on the yeast cell surface.

Design of Fully Human Synthetic VH and VL Libraries with Six Diversified CDRs

We generated fully human synthetic HC and LC libraries by diversifying all of the six CDRs based on single stable framework scaffolds. We first chose VH3 and Vκ1 family genes as a library VH/VL framework pair, because each family gene is most frequently used in a natural repertoire and the VH3/Vκ1 pair has shown favorable biophysical characteristics, such as stability and expression levels [7, 9, 27]. We analyzed human V gene germline sequences of the VH3 and Vκ1 families from the international ImMunoGeneTics information system (IMGT) database (http://www.imgt.org/) and obtained their consensus sequences with the most predominant amino acids at each position (Fig. 1). Even though the consensus sequences of VH3 and Vκ1 could be immediately used as a library scaffold, some artificial sequences deviating from the context of germline locus might raise several problems, such as immunogenicity, stability, and/or poor pairing of VH/VL [10]. Accordingly, instead of the consensus sequence, we searched the nearest germline gene segments to the consensus sequence and chose the IGVH3-23*04 and IGVK1-16*01 genes as single master frameworks for VH and VL, respectively (Fig. 1A). The J gene was chosen by a similar approach, and JH4 for VH, and Jk1 for Vκ were selected. The VH3-23*04 framework was directly used because its framework sequence was identical to the consensus sequence of the VH3 family except for K94, the residue of which Arg is in the consensus sequence. The chosen Vκ1-16*01 framework was used after a point substitution of S46 for Leu, since L46 is largely conserved in the Vκ1 family and plays a critical role in VH/VL interactions [14].

Design of six CDR regions was based on the analysis of most prevalent amino acids in each position of germline VH3 and Vκ1 family genes to reflect the human pre-immune diversity (Fig. 1 and Table 1). This strategy incorporating natural CDR repertoires into a stable VH/VL framework may provide improved functional diversity of the library [31] as well as contribute to the stability and folding efficiency of displayed antibody fragments [16, 30]. The CDR diversities in CDR-H1 (positions 31-35 in Kabat numbering used throughout this study [11]), -H2 (positions50?65), -L1 (positions24?34), and -L2 (positions50?56) were introduced using partially degenerate codons, designed to cover more than 70% of the natural diversity in each position (Fig. 1B and Table 1). Nonetheless, some positions, such as L32, H32, H33, H35, and H58, had a relatively low coverage owing to the limit of codon degeneracy. For CDR-L3, we designed a single VL library of nine amino acids (positions89?97) reflecting the most common length frequency [9, 16]. Residues 89?95 of CDR-L3 are encoded by VL gene segments, residue 96 is mostly diversified by VJ recombination, and the rest of the CDR-L3 and framework 4 are encoded by JL gene segments [28]. Accordingly, positions 89, 90, 93, 95, and 97 were fixed with the most prevalent amino acid in each position, but positions 91, 92, 94, and 96 were diversified with degenerate codons to cover the natural germline frequency (Fig. 1B and Table 1). Among the six CDRs, CDR-H3 showed the most diversity in natural antibodies in terms of amino acid composition and loop length (typically 7?16 in length) [16, 30]. The residues of CDR-H3, defined as the sequence from positions 95 to 102, were not germline encoded; instead their diversity was generated by recombination of the VDJ gene segments [23]. Thus we designed two VH libraries with different CDR-H3 lengths of 9 and 11 amino acids, which belong to the most common frequency in the recovered binders [9, 16]. The first six positions (95?100) of CDR-H3 with 9 amino acids and eight positions (95?100b) of CDR-H3 with 11 amino acids were diversified by a degenerate codon of NNK, which encodes 20 amino acids and 1 stop codon, to maximize diversity. However, the C-terminal last three positions 100a(c), 101, and 102 are not highly variable, but usually contain the consensus sequence of Phe/Met/Ile?Asp?Tyr owing to the encoding by JH gene segments. Thus we used a degenerate codon of WTK to introduce hydrophobic amino acids at 100a(c), but fixed 101 and 102 residues with Asp and Tyr, respectively (Fig. 1B and Table 1).

Fig. 1.Amino acid sequence alignment of VH (A) and VL (B) and library construction strategy by diversifying six CDRs according to the germline frequency. The consensus sequences of VH and VL were derived from VH3 and Vκ1 family amino acid sequences in the IMGT database, respectively. Consensus CDR residues were determined by selecting the most prevalent amino acids observed in natural germline VH3 and Vκ1 family genes. The germline VH3-23*04 and Vκ1-16*01 showing the closest sequences to the consensus sequences of VH3 and Vκ1 family genes were chosen, but slightly modified as described in the text to be used as the single master framework sequences (“templates”) for the library construction. The amino acid prevalence (%) at each position of CDRs is shown by the indicated color code. Diversity of CDR-L1/L2/L3 and CDR-H1/H2 was designed by the indicated degenerated codons to cover the natural germline diversity (theoretical “coverage”) as close to 70% or higher as possible. CDR-H3, designed to be 9 or 11 amino acids long, was maximally diversified by the indicated codons. Diversified positions are marked by X in the template sequence. Residues are numbered according to the Kabat numbering scheme and CDRs were marked according to Kabat et al. [11].

Table 1.aSingle-letter codes for amino acids and nucleotides are used according to the IUPAC-IUB. Residue numbering and CDR definition are according to the Kabat numbering scheme. bPrevalence (%) in natural Vκ1 and VH3 germline gene family was analyzed in the IMGT database and represented as a percentage.

Construction of HC and LC Libraries on Haploid Yeast

To generate a yeast surface-displayed human combinatorial Fab library by assembly of HC and LC libraries, we first constructed independent repertoires of HC and LC in haploid yeast strains of opposite mating types (Fig. 3). We amplified VH and VL gene libraries on the chosen template by performing overlapping PCRs using partially overlapping oligonucleotides designed to introduce random mutations with degenerate codons (Fig. 2). The amplified VH library insert with gap repair tailed sequence (40 bps) was mixed with a linearized pYDS-H vector and transformed into the JAR200 (MATa) strain by the homologous recombination method [12, 13]. The library size determined on the selective agar plates was about 4 × 107 for the HC-1 library with 11 residues in CDR-H3, and 1 × 107 for the HC-2 library with 9 residues in CDR-H3 (Table 2). Likewise, transformation of the mixture of VL library insert with linearized pYDS-K vector into the YVH10 (MATα) strain and selection in the selective media generated a LC library with a diversity of 1 × 107 (Table 2). When individual plasmids were rescued from 20 randomly chosen clones in each unselected library, more than 70% of the VH or VL genes exhibited authentic sequences, indicative of fidelity of the library diversity.

Fig. 2.VH and VL gene library preparation. (A) Overall scheme of VH and VL gene library preparation by fragment PCRs and then overlapping PCRs. The three fragments were respectively amplified from template DNA with the indicated primers. Then equimolar fragments were mixed and amplified to generate full-length VH and VL gene libraries by two external primers (H1-F and H7-R) with two 40 bp extra sequences at both 5’ and 3’ termini for the homologous recombination during the transformation step. Detailed sequences of the primers are shown in Supplementary Table S1. (B) Representative agarose gel electrophoresis showing the sizes of amplified fragment #1 (F1), #2 (F2), and #3 (F3) and full-length insert VH library (VH).

Table 2.aMating efficiency was calculated in percentage as the number of diploid colonies grown on the double-selective plates divided by diploid/haploid colonies grown in the single-selective plates. bDiversity determined by number of colonies grown on the selective plates. cPercentage of library clones with intact open reading frame (ORF) without deletion/insertion mutations, determined by sequence analysis of more than randomly chosen 50 clones. dFunctional diversity obtained by multiplying colony diversity by functional ORF (%).

Fig. 3.Schematic process for generating combinatorial Fab antibody library on yeast cell surface via yeast mating between two haploid cells carrying HC and LC libraries. The haploid HC library was constructed by co-transformation of the VH gene library and linearized pYDS-H vector to JAR200 (MATa) to express the HC library on the cell surface by Aga2p anchoring to the cell wall. The haploid LC library was generated by co-transformation of the VL gene library and linearized pYDS-K vector to YVH10 (MATα) to secrete the LC library. The combinatorial Fab library on the diploid cell surface was generated by mating the two haploid library cells to achieve random pairing of the secreting LC library to the cell wall-anchored HC library by disulfide linkage between Cys residues at the C-terminus of CL and CH1. The Fab assembly of LC and HC could be detected by labeling the LC C-terminal Flag tag on flow cytometry.

Optimization of Yeast Mating Conditions

Yeast mating has been used to construct a human-naïve combinatorial Fab library on the yeast surface by combination of HC and LC diversity [29] and to isolate a high-affinity Fab binder by chain shuffling of HC and LC [2]. Blaise et al. [2] reported that co-incubation of MATa and MATα strains for 4–6 h at a cell density of between 107 and 108 cells/cm2 on YPD agar plates resulted in higher mating efficiency of up to 65%. However, the mating efficiency between two haploid cells carrying HC and LC repertoires was relatively low, showing 29% [2], highlighting an urgent need to optimize yeast mating protocols for two haploid cells carrying HC and LC libraries. We investigated the effects of three major mating parameters on the mating efficiency and numbers of diploid cells for the two HC and LC haploid cells sequentially in the order of cell density, media pH, and cell growth phase.

Cell density. We examined the mating efficiency by homogeneously mixing two haploid cells comprising HC-containing JAR200 (MATa, TRP) and LC-containing YVH10 (MATα, URA) in a 1:1 cell ratio at five different cell densities, 1, 3, 5, 7, and 9 × 107 cells/cm2, on YPD/pH 6.4 agar plates incubated at 30℃ for 6 h (Fig. 4). Yeast cells at the late-log phase (5 × 107 cells/ml) were used. Selection of diploid cells on the double-selective plates (-Trp, -Ura) demonstrated that yeast mating efficacy is very sensitive to cell density, with cell densities of 1~3 × 107 cells/cm2 showing the highest mating efficiency of ~51% (Fig. 4A). In addition to the mating efficiency, the number of diploid cells selected by the mating of two haploid libraries indicates the actual library size. In this context, the cell density of 3 × 107 cells/cm2 was most effective in terms of both the mating efficiency and practically generated diversity size of diploid cell repertoires (Fig. 4B).

Media pH. Y east m ating at l ow p H media h as been shown to increase the mating efficiency [26]. HC and LC haploid library-containing cells were grown at 30℃ until the late-log phase in the selective medium adjusted to pH 6.4 or 4.5. The haploid cells were mated at 30℃ on the same pH-adjusted YPD agar plate at the optimized cell density of 3 × 107 cells/cm2 for 6 h. Mating on the YPD/pH 4.5 plate exhibited slightly improved efficiency (~56%) compared with the YPD/pH 6.4 plate (~51%) (Figs. 4C and 4D). The exact molecular mechanism by which yeast mating efficiency at the low pH was increased is not clear. We speculate that, since mating is initiated via the protein-protein interaction between MATa and MATα agglutinin and sexual adhesion molecules expressed on the cell surface [4, 25], growth and mating under acidic conditions might increase the expression of the relevant proteins and/ or provide more intensive interactions between them.

Cell growth phase. Since yeast mating is a cell fusion event, we hypothesized that the physiological condition of yeast cells may influence the efficacy [24]. The two haploid library cells grown at 30℃ up to different growth phases in the selective media were mated at the above optimized conditions. Mating of the cells grown up to the early-exponential phase exhibited the highest mating efficiency of ~72% compared with the mid- (~61%) and late-exponential (~61%) phase cells (Figs. 4E and 4F). However, the stationary phase cells were poor at mating (~40%). It seems that younger yeast cells have relatively high cell viability and high expression levels of sexual adhesion molecules. Moreover, the thinner and more flexible cell walls of the exponential growth phase cells, compared with the stationary phase cells, may contribute to more efficient cell fusion.

In summary, we found that yeast mating with midexponential growth-phased cells at the cell density of 3 × 107 cells/cm2 on YPD/pH 4.5 plates at 30℃ for 6 h is optimal, with mating efficiency ranging from 56% to 61% for the generation of a large Fab yeast library from two HC- and LC-haploid yeast libraries.

Fig. 4.Mating efficiency (A, C, E) and number of diploid cells (B, D, F) determined by mating between two haploid cells carrying HC and LC libraries, performed at 30℃ for 6 h on YDP agar plates under the indicated conditions. (A, B) Cells grown to late-exponential phase were mated at the indicated cell densities on YPD/pH 6.4 plates. (C, D) Cells grown to late-exponential phase were mated at the density of 3 × 107 cells/cm2 on YPD/pH 6.4 or YPD/pH 4.5 plates. (E, F) Cells grown to the indicated growth phases were mated at the density of 3 × 107 cells/cm2 on YPD/pH 4.5 plates. Error bars, ± SD. *p < 0.05.

Generation of Synthetic Fab Libraries Under the Optimized Yeast Mating Conditions

Using the above-established optimized conditions of yeast mating, either HC-1- or HC-2-containing haploid cells at mid-log growth phase were mated with LC-containing haploid cells at mid-log growth phase on the YPD/pH 4.5 plates at 30℃ for 6 h. The cells on the plates were then pooled and seeded in liquid SDCAA (-Ura, -Trp) medium at the very dilute cell density of OD600 = 0.1. The cells were allowed to grow for 24 h, then transferred into fresh SDCAA media at OD600 = 0.1, and grown for another 24 h before collection. The above procedures successfully enriched the diploid cells to ~90% of the grown cells by out-growth over unmated haploid cells in the double-selective medium.

Fig. 5.Analysis of the scFv fragments generated from the isolated VH and VL inserts. (A) Yeast colony PCR to isolate the cognate VH-VL pair from a single Fab colony and convert it into scFv format for sequencing and further characterization. VH and VL could be amplified by the respective unique primer sets and then converted into scFv format by overlapping PCR. (B) Representative agarose gel electrophoresis of amplified VH (left panel), VL (middle panel), and scFv (right panel) fragments from the indicated single Fab colonies.

Through the above mating and selection procedures, we constructed two yeast surface-displayed combinatorial Fab libraries: Fab-1 with 11 CDR-H3 residues and Fab-2 with 9 CDR-H3 residues at diverse sizes of ~1.2 × 109 and ~1.1 × 109, respectively, by one round of yeast mating (Table 2). To estimate the fidelity of the constructed library, we sequenced 50 VH and 50 VL inserts amplified by yeast colony PCR from two respective Fab library-containing diploid cells (Fig. 5). All of the sequenced VH and VL genes exhibited unique sequences showing ~74% and ~78% functional open-reading frame without stop codons and insertion/deletion mutations (Table 2). Thus the functional diversity sizes of Fab-1 and Fab-2 libraries were estimated as ~7.0 × 108 and ~6.4 × 108, respectively. We also confirmed the amino acid compositions at each position of six CDRs. From the Fab-1 library, 44 VH sequences and 46 VL sequences, which were all functional in-frame, were analyzed to compare amino acids compositions in each position of six CDRs between the library design and constructed library with degenerated codons. The relative amino acid composition in each position of six CDRs in the constructed VH/VL library closely matched the theoretical design (Fig. 6). The Fab-2 library also showed close similarity to the amino acid composition profile of the design. These results suggest that the two Fab libraries were successfully constructed according to the design.

Fig. 6.Relative distribution of amino acid representation in each of six CDRs (A-F), analyzed in the Fab-1 library (11 residue CDR-H3). From the Fab-1 library, a total of 44 VH and 46 VL unique clones were analyzed as described in Fig. 5. The designed (indicated by D) frequency in each residue was closely matched to that of the constructed Fab-1 library (indicated by L).

Although yeast cells grow well at 30℃, induction of Fab antibody for the cell surface display has been accomplished below 30℃, but the results were inconsistent [2, 18, 29]. We determined the optimal growth media, temperature, and induction duration for the cell surface expression of the Fab library by labeling the Flag tag fused at the C-terminal LC using flow cytometry. As shown in Fig. 7, the Fab library was most highly expressed on the cell surface of ~21% after the cells were induced in 2× SGCAA medium at 20℃ for 48 h.

In this study, we generated yeast surface-displayed, combinatorial synthetic Fab libraries, designed to have similar amino acid compositions at each of six CDRs to those in the germline repertoires. To overcome the small library size that is characteristic of yeast surface display, owing to low transformation efficiency, we took advantage of yeast mating between two haploid cells containing HC and LC libraries and optimized the conditions of cell density, media pH, and cell growth phase. Under the established optimal conditions, the mating efficiency was more than 60%, which is much higher than the reported 29% between antibody libraries containing two haploid cells [2]. Our established optimal conditions could also be useful for other yeast mating experiments, such as techniques involving two hybrids. Using the optimal mating conditions, we rapidly constructed yeast surface-displayed Fab libraries with a colony diversity size of more than 109 through one round of mating between two haploid cells containing HC and LC libraries with modest diversity sizes of ~107. The constructed Fab libraries exhibited relative amino acid compositions in each position of the six CDRs that were very similar to those of the designed germline repertoires, suggesting that they would be extremely useful in isolating human Fab antibodies with improved biophysical properties and reduced immunogenicity for given antigens.

Fig. 7.The surface expression levels of the Fab library on diploid cells. (A-C) Fab-1 diploid cells induced by growth under the indicated conditions (induction medium, temperature, and induction periods). (D) Representative flow cytometry dot plots for Fab-1 diploid cells un-induced (left panel) and induced (right panel) in 2× SGCAA for 48 h. The medium of 2× SGCAA is composed of 2-fold higher amount of yeast nitrogen base and casamino acids, but the same amount of galactose, compared with SGCAA. Fab expression levels were determined using flow cytometry by immunofluorescence labeling of the LC C-terminal Flag tag with anti-Flag mAb and represented as the percentage of colonies detected in the gate indicated in (B). Error bars, ± SD. *p < 0.05.