The selection of antibody fragments via display methods for the generation of high-affinity antigen binders is highly versatile. Over the years, several methodologies such as phage, yeast, ribosomal, or bacterial display have been developed [1
]. Yet, the simultaneous identification of antigen-binding molecules and their targets has not been achieved to date. We aimed at an antibody surface display system that would (1) grant library diversity, (2) be compatible with complex mixtures, (3) permit immunoprecipitation (IP) and thus identification of targets, and (4) allow the intentional release of the binders. Thereby, we could concurrently identify the binders and their targets in complex mixtures; without prior knowledge of the antibody or the target.
Among the display systems for the surface expression of antibody-like proteins [13
] or enzymes [15
], phage display is often the system of choice to generate and maintain a highly diverse display library. However, the phages cannot be directly used for IP or fluorescence-activated cell sorting (FACS) [13
]. In addition, target identification after phage display is a tedious, multistep process [9
]. Beside this selection technique, other display systems have been developed. For example, attachment of a protein library in the bacterial periplasm [16
], on magnetosomes, and on polyhydroxyalkanoate granules have been reported. However, they require intermediate steps before panning [13
] and are more promising for the fast purification of proteins from bacterial cultures [17
]. By contrast, yeast and bacterial display are suitable for immediate cell sorting or IP, but transport across host membranes and the formation of disulfide bridges is limiting [12
]. The localization of proteins to the outer membrane of Gram-negative bacteria is usually achieved by genetic fusion to a bacterial surface protein, which increases the overall protein size and might lead to inefficient export across the inner and/or outer membrane [11
]. Nevertheless, the amount of surface-exposed proteins easily surpasses the number that is displayed on phages.
An optimal display system to mutually identify antigens and binders should fulfill certain criteria: (1) the sturdiness of the organism for IP; (2) a minimal number of membranes to be crossed to display the binder; (3) and no necessity for a genetic fusion of the binder.
The Gram-positive bacterium Staphylococcus aureus
has been used for decades for the purification of immunoglobulins [19
] and in IP of non-radiolabeled and radiolabeled materials [20
]. Its protein A presents superb binding specificity and capacity, and the thick peptidoglycan layer assures that the bacterial resin stays intact during the procedures.
The endogenous housekeeping sortase A (SrtA) covalently attaches proteins, for instance protein A, to the lipid II that is integrated into the peptidoglycan [21
]. The presence of the YSIRK/GS motif [23
] in the signal peptide initially confines a protein to the cross wall—only later, the protein localizes to the entire bacterial surface—whereas the absence of this motif restricts a protein to the cell poles or secretion sites [24
]. Accordingly, the fluorescent protein mCherry has been successfully presented at the cross or peripheral wall [26
]. Also a phage display pre-enriched library of affibodies, whose scaffold is based on the Z domain alpha helices of protein A, has been screened for binders of human tumor necrosis factor (TNF) alpha on staphylococci by FACS [27
Whereas affibodies are purely synthetic, antibody recognition domains such as single-chain variable fragments (scFv) or single-domain antigen-binding fragments from camelid heavy chain-only antibodies (VHH) can be cloned from immune cells after vaccination. Thus, they have undergone natural selection, clonal expansion, and affinity maturation against the antigen in vivo. The fact that VHH [28
] incorporate the antigen-binding loops in a single domain, and thus are not constrained by the pairing requirements of a heavy and a light chain, puts them in a favorable position for use in protein engineering compared to conventional antibodies or scFv. The proper folding of VHH is often independent of disulfide bonds [29
] and glycosylation [33
], but can be improved by the introduction of artificial disulfide bridges [34
] and glycosylation sites [33
]. The small size of VHH facilitates routine cloning, bacterial transformation, and protein expression [36
When VHH libraries with high diversity are screened against complex target antigen mixtures, affinity purification of the desired VHH and identification of the respective target is a multi-step process. It typically involves the recovery of the encoded VHH from the bacteria, sub-cloning into a suitable expression system, production and purification of the VHH, followed by IP and mass spectrometry identification of the target. In order to streamline the selection of VHH that bind to proteins in a complex mixture and the identification of their respective targets without additional sub-cloning and purification steps, we developed a surface display method that is compatible with direct IP of VHH targets using robust S. aureus cells.
Our approach aims to simultaneously identify multiple targets and their antibody binders without any knowledge about their nature. So far, most display systems have focused on the identification of binders for a single protein of interest. The abundance of displayed proteins as well as the physiology of Gram-positive bacteria with their single membrane and thick peptidoglycan layer makes them perfectly suited for display systems that need high numbers of binders on the surface and that are exposed to harsh environments e.g., during affinity chromatography. Furthermore, expressed proteins only need to be transported across one membrane to reach the extracellular milieu. Immobilization of VHH, with a protein A tail, on the staphylococcal surface was achieved by utilizing the endogenous SrtA. This avoids the need for helper plasmids—usually used in phage display in the form of a helper phage or a packaging cell line—in any form. The utilization of a signal peptide from staphylococcal enterotoxin B, lacking the YSIRK/GS motif [23
], is thought to lead to VHH hotspots proximal to staphylococcal secretion sites [24
]. This strategy might assure efficient export while it should prevent an overloading of the bacteria with the displayed VHH. Thus, we combined a strong secretion signal with the LPXTG motif of SrtA for the surface display of VHH. The reversible nature of enzymatic immobilization has the advantage of a covalent attachment without hampering further use of the displayed protein. Whereas chemical modifications of the cell surface avoid the risk of using genetically modified organisms and available click chemistry can be readily used with microorganisms [52
], the displayed proteins are externally supplied and lost during cell divisions. Furthermore, binders installed from outside will not lead to a highly diverse library.
The diversity of a library directly depends on the electroporation efficiency of the bacteria, and thus the number of colonies obtained after transformation. While Gram-positive bacteria do not match the electrocompetence of Gram-negative bacteria, efficiencies between one million [60
] and 108
transformants per microgram of plasmid DNA [61
] have been reported. These numbers are sufficient as they equal or even surpass the total number of B cells usually isolated from an immunized alpaca and used to construct a library. Furthermore, less than 50% of B cells express heavy chain-only antibodies [62
] and even fewer are specific for the protein(s) used in immunization.
Direct IP with staphylococci expressing VHH seems a valid approach, as about two-thirds of the VHH identified by phage panning bind their target’s native conformation [63
] (and M.C., unpublished observation)—a prerequisite for downstream in vivo applications. Novel VHH were found by IP with influenza immune libraries that were panned and unpanned against the virus. All new VHH recognize influenza NP, the most abundant structural protein in virions. NP likely represents the most immunogenic influenza protein—or the protein leading to the major heavy chain only antibody response—because our VHH libraries derived from peripheral blood mononuclear cells of alpacas immunized with inactivated influenza contain NP-specific VHH at large [64
In summary, we report a direct and unbiased target identification method for VHH out of libraries that obviates the need for modifying the target. Thus, any natural epitope on the target protein is preserved and available to be bound by a potential VHH. Anchoring the VHH on the staphylococcal surface by the protein A tail provides further benefits such as a stable covalent attachment and the possibility to recover VHH from the bacterial surface. Exposing the staphylococci to a triglycine nucleophile allows the attack of the SrtA-VHH acyl intermediate and the release of the VHH into the extracellular milieu. Triglycine nucleophiles equipped with biotin or small fluorescent probes can be used to site-specifically functionalize the VHH straight off the bacterial surface. VHH recovered in such fashion can be applied directly in downstream assays such as ELISA and Western blotting without the need for sub-cloning, expression and purification.
By combining our knowledge of VHH, display systems and sortase-mediated site-specific labeling we have created an unbiased and reversible staphylococcal surface display system to identify VHH and their targets. Future modifications of the reported model system can easily be achieved by endowing the single pSA-VHH-SPAXrc vector with the desired features.
4. Materials and Methods
4.1. Generation and Mass Spectrometry of the pSA-VHH-SPAXrc Shuttle Vector Construct
To achieve the surface display of VHH, we engineered a staphylococcal expression vector called pSA-VHH-SPAXrc containing the staphylococcal enterotoxin B leader sequence (amino acids 1–34), a SalI site, a VHH, a BamHI site, the staphylococcal protein A repetitive (Xr, amino acids 324–420) and constant (Xc, amino acids 421–516) regions [65
]. The C-terminal part of Xc consists of the SrtA motif LPETG (amino acids 482–486) followed by a hydrophobic domain and then a charged tail. The final vector was generated by Gibson assembly (NEB, Ipswich, MA, USA) over a sequence of intermediate steps from pRIT5-SPA [46
] and pOS1 [47
] using pSEB-SPA-490-524 [47
] as backbone. The VHH template was amplified out of the GFP-specific enh VHH [43
] carrying pHEN vector [69
Mass spectrometry was carried out as previously reported [70
]. Proteins were separated by SDS-PAGE, visualized by Coomassie or silver stain, and bands were excised. Gel pieces were subjected to dithiothreitol reduction, iodoacetamide alkylation and trypsin digestion. The digested peptides were extracted, concentrated by SpeedVac and separated using a Dionex RSLCnano high performance liquid chromatography (HPLC) system equipped with a self-packed three micron Jupiter C18 analytical column (10 cm × 75 micron, Phenomenex, Torrance, CA, USA). Peptides were eluted by standard reverse-phase gradients and analyzed in an Orbitrap Elite mass spectrometer (Thermo Fisher Scientific, Waltham, MA, USA) running nano flow configuration operated in a dependent data acquisition mode. Peptides were identified using SEQUEST algorithms, attributed to a species-specific National Center for Biotechnology Information (NCBI) database and correlated to proteins by Scaffold Q+S (Proteome Software, Portland, OR, USA) using a minimum of two peptides (peptide threshold of 95% and protein threshold of 99%).
4.2. Detection of Nitrocellulose Spotted S. aureus
S. aureus cells were grown to saturation, titrated in a twofold dilution series, and 7 μL was spotted onto nitrocellulose. After blocking in phosphate buffered saline (PBS) with 4% milk and 0.1% Tween-20 (PBSMT), the nitrocellulose was submerged in 5 μg/mL GFP containing PBSMT. Unbound GFP was washed away with PBSMT and the amount of captured GFP assessed in a Typhoon (GE, Pittsburgh, PA, USA) biomolecular imager.
4.3. Flow Cytometry and Fluorescence-Activated Cell Sorting (FACS) of S. aureus
Staphylococcal log cultures were labeled with PBS 50 μM 5-chloromethylfluorescein diacetate (CMFDA) (Life Technologies, Carlsbad, CA, USA) for 1.5 h in a 37 °C waterbath. The labeling was quenched in 2% bovine serum albumin (BSA) PBS (PBSB). The staphylococci were washed with cold PBSB and stained on ice with PBSB containing 10 μg/mL MHCII tetramer (NIH, Bethesda, MD, USA) for 45 min. Freshly isolated splenocytes were blocked with 2% ICS PBS (PBSI) and mixed with 100 μL CMFDA-labeled staphylococci on ice for 25 min. The stainings were acquired in an LSRFortessa (BD, San Jose, CA, USA). For sorting, the staphylococci were washed with cold PBSB, mixed on ice with PBSB containing GFP for 45 min, and run through a FACSAria (BD, San Jose, CA, USA).
4.4. IP with S. aureus
Frozen or fresh cultures of staphylococci were washed and blocked with ice-cold 4% PBSB. The target protein or lysate was added in ice-cold 4% PBSB or lysis buffer, and incubated at 4 °C for 60 min. After washing thrice by three minutes centrifugation at 6000 rpm and 4 °C, bound proteins were eluted with 0.2 M glycine (pH 2.2) and neutralized by addition of 1 M Tris (pH 9.1) or with 1% SDS. Radiolabeling was performed and the lysates prepared as recently reported [72
]. Briefly, cells were labeled at 37 °C using methionine- and cysteine-free medium supplemented with 10 mCi/mL [35
S]Met/Cys (PerkinElmer, Waltham, MA, USA).
4.5. Fixation of S. aureus
Saturated overnight cultures or log cultures were washed in PBS, resuspended in 4% FA PBS and fixed for one hour at 37 °C, 250 rpm. Fixed bacteria were washed in PBS and stored at 4 °C. Bacterial titers were determined before addition of FA, and complete fixation was verified by plating FA-fixed staphylococci on tryptic soy agar plates (BD Difco, San Jose, CA, USA).
4.6. Maintenance of C. elegans Strains and Lysate Preparation
Strains were grown on nematode growth medium agar plates seeded with OP50 Escherichia coli
bacteria as described previously [73
]. For lysate preparations, synchronized Day 1 adults were collected, washed vigorously with water to remove excess bacteria, snap-frozen in liquid nitrogen, and pulverized using a pre-chilled mortar. Samples were supplemented with ice-cold Nonidet P-40 lysis buffer, incubated on ice for 30 min, and further used in IP experiments.
4.7. GFP Western Blotting
Polyvinylidene difluoride (PVDF) membranes were blocked in 5% milk 0.1% Tween-20 Tris buffered saline (TBSMT), incubated with 0.2 μg/mL anti-GFP mAb (Roche, Branchburg, NJ, USA) in TBSMT followed by 1:20,000 diluted polyclonal sheep anti-mouse IgG horseradish peroxidase (HRP)-coupled antibodies (GE, Pittsburgh, PA, USA) in TBSMT and developed with SuperSignal West Femto Substrate (Pierce, Rockford, IL, USA).
4.8. VHH Release from S. aureus
Saturated overnight cultures or log cultures were washed in PBS, resuspended in sortase (50 mM Tris (pH 7.5), 150 mM NaCl, 10 mM CaCl2
] or SMM (0.5 M sucrose, 20 mM maleate, 20 mM MgCl2
] buffer. VHH was released by incubation with tryglycine nucleophile or lysostaphin at 37 °C. Detection of released VHH was performed with 1:20,000 diluted streptavidin-HRP (GGG-biotin release) or polyclonal goat anti-Llama IgG (heavy and light chain) HRP-conjugated antibodies (Bethyl, Montgomery, TX, USA). The latter has been found to cross-react with alpaca VHH [75
4.9. Specificity and Competition Enzyme-Linked Immunosorbent Assay (ELISA) and Fast Protein Liquid Chromatography (FPLC)
Plates were coated with 1–10 μg/mL His-tagged NP and 1 μg/mL biotinylated VHH was used before streptavidin-HRP (1:4000, Pierce, Rockford, IL, USA) either blocked or not with 100 μg/mL His-tagged VHH. 3,3′,5,5′-tetramethylbenzidine (Sigma, St. Louis, MO, USA) with acid stop was used as substrate.
67 μg TAMRA-labeled VHH and 200 μg influenza NP were mixed to get equimolarity. For VHH52.1-TAMRA 2-fold molar excess (114 μg) was used. The single components and the mixes were run through a S75 size exclusion column in an Äkta system (both GE, Pittsburgh, PA, USA).