Rapid construction of metabolite biosensors using domain-insertion profiling

Abstract

Single-fluorescent protein biosensors (SFPBs) are an important class of probes that enable the single-cell quantification of analytes in vivo. Despite advantages over other detection technologies, their use has been limited by the inherent challenges of their construction. Specifically, the rational design of green fluorescent protein (GFP) insertion into a ligand-binding domain, generating the requisite allosteric coupling, remains a rate-limiting step. Here, we describe an unbiased approach, termed domain-insertion profiling with DNA sequencing (DIP-seq), that combines the rapid creation of diverse libraries of potential SFPBs and high-throughput activity assays to identify functional biosensors. As a proof of concept, we construct an SFPB for the important regulatory sugar trehalose. DIP-seq analysis of a trehalose-binding-protein reveals allosteric hotspots for GFP insertion and results in high-dynamic range biosensors that function robustly in vivo. Taken together, DIP-seq simultaneously accelerates metabolite biosensor construction and provides a novel tool for interrogating protein allostery.

Figure 1. High-throughput biosensor construction using DIP-seq.

(a) Illustration of a single-FP metabolite protein biosensor. cpGFP is fused to a LBD in a manner such that metabolite binding by the LBD causes a change in fluorescence of the attached cpGFP. (b) Overview of the domain-insertion profiling method used to create and identify functional biosensors. A diverse library of fusions, with cpGFP inserted into a LBD, is created and screened with FACS. Initial and sorted libraries undergo NGS analysis and these data are used to identify insertion sites within the LBD that are enriched during screening. Clones of interest are individually tested to validate biosensor functionality. (c) Method of domain-insertion library creation. An engineered transposon containing a selectable marker is inserted into a staging plasmid carrying the LBD ORF using an in vitro transposase reaction. Staging plasmids with an insertion are selected for, purified and digested with an enzyme that releases the LBD ORF from the staging backbone (grey). LBD ORFs with an inserted transposon are size-selected and cloned into an expression plasmid (black). Finally, a domain of interest (in this paper, cpGFP) is inserted into the cloning site created by the modified transposon.

Figure 2. Maltose-binding protein as proof-of-concept target of random domain insertion.

(a) Fluorescence histograms of representative MBP-cpGFP libraries. Induced libraries were sorted via FACS with gating set to collect cells above a threshold GFP fluorescence level shown with a dashed line. (b) Enrichment of cpGFP-insertion sites (log₂ of fold change) from FACS mapped onto MBP crystal structure (PDB 1ANF). Enrichment calculated from NGS, comparing pre- and post-sort libraries. Insertion sites that went from undetectable to detectable were set at +6, while sites that were cleared from the library (not sequenced in post-sort library) were set at −6. Amino acid 170 is indicated with an arrow. (c) Functional biosensor insertion sites mapped onto the MBP structure. Sites highlighted with sphere representation of side chains, with colours showing level of activity (ΔF/F=(F_ligand−F₀)/F₀). (d) Activity assay of individual constructs. Fluorescence change on addition of saturating maltose (1 mM) or PBS was measured, shown by mean±s.d. of ΔF/F (three biological replicates). Sample 165-PPYF is the previously published construct EcMBP165-cpGFP.PPYF.T203V (ref. 19).

Figure 3. Domain-insertion profiling enriches for fluorescent in-frame proteins.

(a) Fluorescence histograms of a representative TMBP-cpGFP library during initial sort. Histograms compare uninduced with induced libraries. The induced library was treated and sorted with 1 mM trehalose. (b) Fluorescence histograms of sorted library shown in a during sort two (that is, sort for low fluorescence in the absence of trehalose). (c) Fluorescence histograms of the sorted library from b during sort three (that is, high fluorescence in 1 mM trehalose). For a–c, 25,000 events are shown. The dashed line indicates events at a non-fluorescent sample threshold and green shaded regions indicate gates for sorted cells. (d) Histogram of enrichment values, comparing productive (in-frame) insertion constructs and non-productive (out-of-frame and reverse) insertion constructs. Enrichment, shown as log₂ of the fold change, was calculated with DESeq from two biological replicates comparing the final library NGS read counts to the initial library. Calculated enrichments with P<0.1 shown. (e) Profile of enrichment values (initial versus final) for in-frame insertions along TMBP primary sequence. Calculated enrichments with P<0.1 shown. Insert sites cleared from the library are represented with gray dashed lines set to −5. Colours are binned for enrichment values, matching Fig. 4c. Asterisks mark isolated and sequenced constructs with activity ΔF/F⩾0.5 (Supplementary Fig. 9a).

Figure 4. DIP-seq with successive rounds of positive and negative selection enriches for functional biosensors.

(a) Activity versus final NGS read count. Activity is reported as mean±s.d. of ΔF/F from at least three biological replicates (asterisk denotes one sample with two replicates). Read count is combined from two biological replicates from the final sort 3 library. Activity is not correlated with final read count (Pearson's R²=8.6 × 10⁻⁵, P=0.96). (b) Activity versus enrichment. Activity is as shown in a. Enrichment, shown as fold change, is calculated from two biological replicates comparing the initial and final library NGS read counts. Calculated fold change with P<0.1 shown. Activity is linearly correlated with enrichment (Pearson's R²=0.73, P=5.3 × 10⁻¹¹). Colours of points in a,b are binned based on fold-change values after all sorting rounds. (c) Enrichment of cpGFP-insertion sites (log₂ of fold change) mapped onto TMBP crystal structure (PDB 1EU8). Sites cleared from the library are set at a value of −5. Colours are binned for enrichment values, with white representing sites of nonsignificant enrichment (P⩾0.1). Sites in the top bin, having the highest enrichment, are highlighted by sphere representation of their side chains.

Figure 5. The optimized trehalose biosensor reports trehalose concentrations in vivo.

(a) The fluorescence response of the parental and optimized trehalose biosensors, Tre-334 and Tre-C04, respectively, to 1 mM trehalose addition (indicated by arrow). Negative controls of media and TMBP are also shown. Data are mean±s.d. for three replicates. (b) Response of optimized trehalose biosensor, Tre-C04, to 1, 10, 100, or 1000 μM trehalose during exponential growth. Arrows indicate the timing of trehalose addition. Fluorescence was background corrected against a culture treated with water instead of trehalose. Data are mean±s.d. for 12 replicates. (c) OD₆₀₀ (upper panel) and fluorescence response (lower panel) of Tre-C04 to the addition of water, PBS or 300 mM NaCl (once or twice). Dashed lines indicate additions. Data are mean±s.d. for four biological replicates.