Cell surface β‐lactamase recruitment: A facile selection to identify protein–protein interactions

Abstract

Protein–protein interactions (PPIs) are central to many cellular processes, and the identification of novel PPIs is a critical step in the discovery of protein therapeutics. Simple methods to identify naturally existing or laboratory evolved PPIs are therefore valuable research tools. We have developed a facile selection that links PPI‐dependent β‐lactamase recruitment on the surface of Escherichia coli with resistance to ampicillin. Bacteria displaying a protein that forms a complex with a specific protein‐β‐lactamase fusion are protected from ampicillin‐dependent cell death. In contrast, bacteria that do not recruit β‐lactamase to the cell surface are killed by ampicillin. Given its simplicity and tunability, we anticipate this selection will be a valuable addition to the palette of methods for illuminating and interrogating PPIs.

The identification and interrogation of protein–protein interactions (PPIs) is necessary to illuminate and better understand complex and disease‐relevant cellular processes, as well as to identify novel proteins with therapeutic potential. Facile methods to identify and interrogate PPIs are therefore of great value to a diverse set of researchers.

Popular screening‐based methods for detecting and interrogating PPIs include chemical cross‐linking (Fancy & Kodadek, 1999; Tang & Bruce, ²⁰⁰⁹), co‐immunoprecipitation (Burckhardt et al., 2021; Foltman & Sanchez‐Diaz, ²⁰¹⁶; Free et al., ²⁰⁰⁹; Lin & Lai, ²⁰¹⁷; Tan & Yammani, ²⁰²²), enzyme‐linked immunosorbant assay (Weng & Zhao, 2015), phage display (Smith, 1985), bacterial display (Kenrick & Daugherty, 2010), yeast display (Boder & Wittrup, 1997), mRNA display (Roberts & Szostak, 1997), protein‐fragment complementation (Blakeley et al., 2012; Magliery et al., ²⁰⁰⁵; Yao et al., ²⁰²⁰), and quantitative proteomic techniques (Puig et al., 2001; Rigaut et al., ¹⁹⁹⁹). These screening‐based methods can be laborious and time‐consuming, require specialized and expensive equipment and/or reagents, and often rely on complex data analysis.

In contrast to screens, selection‐based methods take advantage of Darwinian outcomes, since cells housing partners in a binding interaction survive, while those housing non‐binders die. This binary outcome dramatically simplifies the process of identifying macromolecules that interact. Popular selection‐based platforms to identify PPIs include bacterial two‐hybrid (Joung et al., 2000), yeast two‐hybrid (Fields & Song, 1989), and phage‐assisted continuous evolution (PACE) (Esvelt et al., 2011). In each of these methods, a “bait” protein or peptide is fused to a DNA binding domain and a “prey” peptide or protein is fused to an activation domain. The formation of a complex between bait and prey localizes the activation domain into proximity with the promoter of a selection gene necessary for cell survival. While incredibly powerful, each of these methods requires an interaction that orients the activation domain in a position that permits transcription of the selection gene, which can be challenging to predict a priori. While display‐based methods allow researchers to alter the concentration of exogenous targets (e.g. high prey concentrations may be used in early rounds of screening to identify early‐stage hits), tightly controlling the level of the target component using in vivo methods like two‐hybrid and PACE can be challenging.

We set out to develop a facile and broadly applicable selection‐based display platform to identify two‐component PPIs. In our approach, we drew inspiration from bacterial transformation. Selection of bacteria that have been transformed with a plasmid is typically achieved by treatment with an antibiotic since the plasmid transformed into the bacteria often encodes a gene that endows resistance to a specific antibiotic. For example, many commonly used plasmids contain a gene encoding β‐lactamase, which reacts with β‐lactam antibiotics (e.g. ampicillin), rendering them non‐lethal. We envisaged a selection in which one protein (bait) is displayed on the surface of Escherichia coli, and another protein (prey) is fused to β‐lactamase. We reasoned that an interaction between the bait and prey biopolymers would generate a highly effective concentration of β‐lactamase on the surface of the bacteria, which would provide resistance to a β‐lactam antibiotic. In contrast, E. coli lacking a displayed bait with high affinity for the prey would not recruit β‐lactamase to the cell surface and would therefore be killed in the presence of a β‐lactam antibiotic (Figure 1a).

FIGURE 1.

(a) Overview of bacterial display β‐lactamase recruitment to identify protein–protein interactions. Escherichia coli displaying a bait protein is mixed with a fusion protein consisting of prey, which binds the bait, and β‐lactamase. Bait: prey binding generates a high effective concentration of surface bound β‐lactamase, which provides resistance to β‐lactam antibiotics (e.g. ampicillin). No binding interaction between bait and prey results in cell death in the presence of ampicillin. (b) A protein–protein interaction between surface displayed GFP‐binding nanobody (GFPnb) and β‐lactamase GFP endows resistance to ampicillin (upper left quadrant). E. coli displaying GFPnb that is not incubated with β‐lactamase GFP are killed by ampicillin (upper right quadrant). E. coli not displaying GFPnb, but incubated with β‐lactamase GFP, is killed with ampicillin (lower left quadrant). E. coli not displaying GFPnb and not treated with β‐lactamase GFP is killed with ampicillin (lower right quadrant). (c) Survival of E. coli displaying GFPnb is dependent upon the concentration of exogenous β‐lactamase GFP used in the bait/prey mixing step of the selection. Selections were performed in triplicate; representative data are shown.

As a proof‐of‐concept, we transformed E. coli (DH10B‐T1^R) with pNeae2‐GFPnb, a plasmid that endows resistance to chloramphenicol and permits inducible expression of a bacterial display platform (intimin N‐terminal domain, Neae; Salema et al., ²⁰¹³) fused to a previously reported green fluorescent protein (GFP) binding nanobody (Kubala et al., 2010) (GFPnb, K _D ≈ 1.4 nM) equipped with a C‐terminal myc tag. When E. coli was induced with IPTG to display GFPnb‐myc, grown for 18 h at 25°C, then treated with either AlexaFluor‐488 labeled anti‐myc antibody or GFP‐β‐lactamase, we observed high levels of cell surface fluorescence. In contrast, E. coli treated identically, but not induced with IPTG, were not appreciably fluorescent (Supporting Information). Collectively, these data demonstrate that GFPnb display is tightly controlled, when induced with IPTG, GFPnb is displayed at appreciable levels on the cell surface, displayed GFPnb retains affinity for GFP. and GFP recruitment to the exterior of bacteria requires GFPnb display.

We next determined if β‐lactamase recruitment provided protection against treatment with β‐lactam antibiotic. A 2 mL culture of E. coli (OD₆₀₀ ≈ 0.4) containing pNeae2‐GFPnb was induced with 0.1 mM IPTG to display GFPnb, while another 2 mL culture of identical E. coli was not induced to display GFPnb. After growth for 18 h at 25°C (OD₆₀₀ ≈ 2 was reached for both +IPTG and −IPTG samples), 200 μL of each E. coli solution was transferred to Eppendorf tubes (two +IPTG samples; two −ITPG samples) and pelleted. Two hundred microliters of a PBS containing 2 μΜ β‐lactamase GFP fusion protein was added to one of the tubes containing E. coli induced to display GFPnb, as well as one of the tubes containing E. coli that was not induced to display GFPnb. The other two tubes were treated with 200 μL PBS. All four tubes were rotated at 25°C for 30 min, pelleted, and washed with 500 μL PBS for 1 min. After a final pelleting step, E. coli was resuspended in 200 μL PBS and 10 μL of each sample was plated onto LB‐agar containing 25 μg/mL chloramphenicol and 100 μg/mL ampicillin.

As shown in Figure 1b, following incubation at 30°C for 20 h, we observed robust growth of E. coli that was induced to display GFPnb and treated with a solution containing 2 μΜ β‐lactamase GFP. In contrast, no growth was observed for the sample induced to display GFPnb but not treated with β‐lactamase GFP. Similarly, bacteria not induced to display GFPnb, but treated with 2 μΜ β‐lactamase GFP, did not survive, despite treatment with a high concentration of β‐lactamase GFP before plating. As expected, bacteria not induced to display GFPnb and not treated with β‐lactamase GFP did not survive. We did not observe appreciable differences in cell culture concentration (OD₆₀₀) for induced or uninduced cells before plating. Thus, differences in the selection outcome are not attributable to different levels of bacteria concentration in induced and uninduced cultures. Collectively, these data demonstrate the viability of the selection platform: cells that recruit β‐lactamase via a cell surface PPI survive treatment with ampicillin, while bacteria lacking cell surface recruited β‐lactamase are killed by ampicillin.

It is common for bacteria transformed with a plasmid encoding IPTG‐inducible beta‐lactamase to grow into colonies, or a streak of bacteria, on agar plates containing IPTG and ampicillin. The resulting colonies survive because beta‐lactamase expression is induced by IPTG. In our case, beta‐lactamase is only present on the surface of cells directly plated onto the agar. However, we observe the growth of daughter cells from the initial parent cells containing cell surface beta‐lactamase. There are two possible mechanisms for this observation. The first is that bacteria initially plated, displaying cell‐surface beta‐lactamase, destroy neighboring ampicillin in the agar, thus providing an environment in which daughter bacteria can grow. Alternatively, parent cells displaying cell surface beta‐lactamase create a physical barrier between agar‐embedded ampicillin and daughter cells, which grow on top of the parent cells.

We next evaluated the concentration‐dependence of exogenous β‐lactamase GFP on cell survival. Samples of E. coli induced to display GFPnb were individually incubated with 2000 to 4 nM β‐lactamase GFP in PBS for 30 min at 25°C. These cells were then pelleted, washed, and resuspended as described above. Next, 5 μL of each resuspended E. coli sample was plated on LB agar containing chloramphenicol and ampicillin. As shown in Figure 1c E. coli treated with 2000 to 31 nM β‐lactamase GFP before plating survived the selection, while bacteria treated with lower concentrations of β‐lactamase GFP did not survive.

Escherichia coli displaying GFPnb are viable reagents in the selection for at least 10 days after IPTG induction. E. coli (OD ≈ 0.4) was induced to display GFPnb with 0.1 mM IPTG, grown at 25°C for 18 h, then stored in LB (with IPTG and chloramphenicol) at 4°C. Over a series of 2, 4, 6, 8, or 10 days after IPTG induction, 200 μL aliquots of E. coli induced to display GFPnb were subjected to selection conditions using 1 μM exogenous β‐lactamase GFP. After each selection, cell survival was assessed by plating cells on LB agar containing chloramphenicol and ampicillin. Satisfyingly, as shown in Figure 2, we observed similar levels of growth on selection plates, indicating that bacteria displaying GFPnb are viable reagents in the selection, even 10 days after IPTG induction.

FIGURE 2.

Escherichia coli displaying Neae‐GFPnb were treated with 1 μM β‐lactamase GFP, washed, then plated on chloramphenicol/ampicillin LB agar 2, 4, 6, or 8 days after IPTG induction. Selections were performed in triplicate; representative data are shown.

Having successfully demonstrated the β‐lactamase recruitment selection works as envisaged, we determined how quickly β‐lactamase GFP recruitment and cell survival are achieved following IPTG induction of Neae‐GFPnb expression. E. coli harboring pNeae2‐GFPnb (OD ≈ 0.4) were induced with 0.1 mM IPTG, then mixed with 1 μM GFP‐β‐lactamase 1–9 h after induction. Interestingly, appreciable growth, indicating GFPnb display and β‐lactamase GFP recruitment, was achieved after 5 h, and increased cell growth was observed for each subsequent time point (Figure 3a). In contrast, bacteria not induced with IPTG did not survive on selection plates, over the course of the entire 9 h experiment (Figure 3b). The modest time needed to achieve PPI‐dependent cell survival further simplifies the selection protocol.

FIGURE 3.

(a) Escherichia coli induced to display Neae‐GFPnb were subjected to selection conditions (1 μM β‐lactamase GFP and washing) 1–9 h after IPTG induction, then plated on chloramphenicol/ampicillin LB agar. Experiments were performed in triplicate; representative data are shown. (b) E. coli not induced to display Neae‐GFPnb were subjected to selection conditions (1 μM β‐lactamase GFP and washing) over the identical time period, then plated on chloramphenicol/ampicillin LB agar. NT, no treatment; cells were not treated with β‐lactamase GFP. Experiments were performed in triplicate; representative data are shown.

By coupling the bacterial display of a bait protein with the recruitment of an exogenous fusion protein consisting of prey and β‐lactamase, we have developed a novel selection to identify PPIs. This selection gives researchers the ability to tightly control the concentration of exogenous prey β‐lactamase fusion protein, which should allow researchers to select for high‐, medium‐, or low‐affinity PPIs. Cell survival, resulting from of a desired PPI, is achieved in as little as 5 h after IPTG induction of Neae‐bait fusion protein, and bacteria displaying bait are viable reagents at least 10 days after the initial display of GFPnb was induced. Given the simplicity and tunability of this selection, we anticipate it will be a valuable addition to the palette of experiments commonly used to detect and evaluate PPIs.

AUTHOR CONTRIBUTIONS

Brian R. McNaughton: Validation; writing – review and editing; funding acquisition; writing – original draft; supervision; project administration. Jordan A. Hinmon: Investigation; methodology; data curation; validation. Jade M. King: Investigation; methodology; data curation; validation. Latrina J. Mayo: Investigation; methodology; validation; data curation. Cierra R. Faries: Data curation; validation; methodology; investigation. Ya'hnis T. Lockett: Investigation; validation; data curation. David W. Crawford: Investigation; methodology; validation; data curation. Patrick C. Beardslee: Investigation; methodology; validation; data curation. Alexander Hendricks: Data curation; methodology; validation; investigation.

FUNDING INFORMATION

This work was supported by the National Institutes of Health/National Institute of General Medical Sciences (R15GM141783, B.R.M.) and National Institutes of Health/National Institute on Minority Health and Health Disparities (U54MD015959, B.R.M.). C.R.F. was funded by a grant from the National Institutes of Health/National Institute of General Medical Sciences (T34GM136477).

CONFLICT OF INTEREST STATEMENT

The authors declare no competing financial interest.

Supporting information

Data S1. Supporting Information.

Hinmon JA, King JM, Mayo LJ, Faries CR, Lockett YT, Crawford DW, et al. Cell surface β‐lactamase recruitment: A facile selection to identify protein–protein interactions. Protein Science. 2024;33(4):e4919. 10.1002/pro.4919