Highly multiplexed bioactivity screening reveals human and microbiota metabolome-GPCRome interactions

Abstract

The human body contains thousands of metabolites derived from mammalian cells, the microbiota, food, and medical drugs. Many bioactive metabolites act through the engagement of G-protein-coupled receptors (GPCRs); however, technological limitations constrain current explorations of metabolite-GPCR interactions. Here, we developed a highly multiplexed screening technology called PRESTO-Salsa that enables simultaneous assessment of nearly all conventional GPCRs (>300 receptors) in a single well of a 96-well plate. Using PRESTO-Salsa, we screened 1,041 human-associated metabolites against the GPCRome and uncovered previously unreported endogenous, exogenous, and microbial GPCR agonists. Next, we leveraged PRESTO-Salsa to generate an atlas of microbiome-GPCR interactions across 435 human microbiome strains from multiple body sites, revealing conserved patterns of cross-tissue GPCR engagement and activation of CD97/ADGRE5 by the Porphyromonas gingivalis protease gingipain K. These studies thus establish a highly multiplexed bioactivity screening technology and expose a diverse landscape of human, diet, drug, and microbiota metabolome-GPCRome interactions.

Keywords: G protein-coupled receptor (GPCR); high-throughput screening; human metabolome; microbiome; microbiota metabolites.

Figure 1.

Design and assessment of PRESTO-Salsa. (A) Modules of PRESTO-Salsa. PRESTO-Salsa consists of three modules: a GPCR plasmid, which contains a GPCR ORF, V2 tail, TEV cleavage site (TEVcs) and tTA; an adaptor module, which encodes a fusion protein of β-arrestin 2 and the catalytic domain of TEV protease; and a PRESTO-Salsa reporter plasmid, including a TRE promoter, mini promoter, EGFP, and a unique 10 bp barcode. (B) Principle of the PRESTO-Salsa system. Co-expression of the three PRESTO-Salsa modules transforms HEK293T cells into unique GPCR reporter cell lines. ① Once stimulated with a ligand, ② the GPCR will recruit β-arrestin 2/TEV fusion protein and trigger cleavage of the engineered GPCR at the TEVcs. ③ The released transcription factor tTA will translocate into the nucleus, ④ inducing the transcription of the EGFP/barcode mRNA. The fold induction of a single GPCR or a GPCR panel can be quantified based on EGFP detection or barcode quantification by next-generation sequencing (NGS), respectively. (C) The PRESTO-Salsa pipeline. PRESTO-Salsa reporter cells, a HEK293 T-derived cell line that stably expresses β-arrestin 2/TEV fusion protein, are seeded into four 96-well plates. When the cell density reaches 90%, ① the cells in each individual well are co-transfected with one GPCR plasmid (314 GPCRs total) and one of 314 unique PRESTO-Salsa reporter plasmids, transforming the PRESTO-Salsa reporter cells into GPCR reporter cells. Each well thus contains one of 314 unique GPCR reporter cell lines. ② 6 hours after co-transfection, the cells from all 314 wells are lifted and pooled to construct a GPCR reporter cell line library consisting of 314 GPCR reporter cell lines. This library is then reseeded into Poly-D-lysine pre-treated 96-well plates followed by ③ ligand stimulation, ④ mRNA extraction, ⑤ cDNA synthesis, and ⑥ sequencing library construction via 1^st PCR and 2^nd PCR. ⑦ The amplicon library containing 2,304 sequencing samples is sequenced by NovaSeq. ⑧ Each GPCR-paired barcode is quantified, and the activation of all 314 GPCRs in each well is calculated as the fold induction of each unique barcode over unstimulated controls.

Figure 2.

Detection of bioactive small molecule metabolites with PRESTO-Salsa. (A) Three titrating doses of eleven small molecule ligands were screened against the near-complete conventional GPCRome using PRESTO-Salsa. (B) Four human microbiome species can produce phenethylamine, histamine, dopamine, and tryptamine, respectively. (C-D) Detection of microbiome-derived bioactive metabolites from bacterial cultures. Four species were cultured in a chemically defined minimal medium (MM) (C) or gut microbiota medium (GMM) (D) for 24 hours. Cleared bacterial supernatants were screened against the human GPCRome using PRESTO-Salsa. n=3 technical replicates per group (A, C-D).

Figure 3.

Detection of bioactive proteins with PRESTO-Salsa. (A) The activities of diverse chemokines against the near-complete conventional GPCRome. Three titrating doses of seven chemokines were screened via PRESTO-Salsa. (B) C3AR1 is activated by human serum but not other mammalian sera. Six mammalian sera (5% serum concentration) were screened against the human GPCRome via PRESTO-Salsa. (C) Human serum activates C3AR1. Serial dilutions of mammalian sera were tested against human C3AR1 with GPCR activity measured by C3AR1 Tango assay. (D) Complement protein C3-deficient human serum shows no activity on C3AR1, as measured with C3AR1 Tango assay. (E) Human serum eliminates acetylcholine (ACh)-mediated activation of muscarinic acetylcholine receptors (CHRMs). The GPCRome-wide activity of 1 or 10 μM of ACh was measured via PRESTO-Salsa with or without 5% human serum. (F) Mammalian sera abolish ACh activity on CHRM3. Activation of CHRM3 by titrating doses of ACh was measured by CHRM3 Tango assay with or without 5% mammalian sera. (G) The acetylcholinesterase (AChE) inhibitor (AChEi) reverses the inhibitory effect of human serum on CHRM3. 5% human serum was pretreated with 0.1, 1 or 10 μM of AChEi (Neostigmine) before assessment of CHRM3 activation via Tango. Data in all panels other than A, B, E are representative of at least three independent experiments. n=3 technical replicates per group.

Figure 4.

PRESTO-Salsa screening reveals human metabolome-GPCRome interactions. (A) PRESTO-Salsa screening of 1,041 human metabolites (100 μM) against 314 GPCRs. PRESTO-Salsa screening and data analysis revealed 561 putative hits. 138 of these hits were validated via arrestin-based Tango assays and 92 of these validated hits were subsequently validated via G protein assays. (B) Summary of validated GPCR activation by human metabolites. Data are displayed on a hierarchical tree of GPCRs organized by class, ligand type, and receptor family. The intensity of the red circles represents the normalized maximum magnitude of activation relative to the maximal activation by endogenous ligands across the entire data set—i.e., the maximum activation of a given GPCR by any human metabolite in our collection. Purple circles represent GPCRs without known maximal activation (including orphan GPCRs) that were activated by at least two-fold by at least one metabolite. Radii of the circles at each tip are scaled based on the number of human metabolites that activated a given receptor or receptor family (i.e., the number of hits across the complete data set). Validated hits were defined as activation of a given receptor more than two-fold over background as measured by individual Tango assays. The detailed information for all human metabolome-GPCRome interactions is presented at: http://palmlab.shinyapps.io/presto-salsa/ (C) Select previously undescribed metabolite-GPCR interactions. Previously unreported human metabolite-GPCR interactions were validated via Tango- and G protein-based assays; the dose responses are merged results from three independent experiments. (D) Off-target effects of medical drugs on GPCRs. Off-target effects of select small molecule drugs on GPCRs. The dose-response curves are merged from three independent experiments. n=3 technical replicates per group (B-D).

Figure 5.

PRESTO-Salsa reveals a cross-tissue human microbiome-GPCRome interaction atlas. (A) PRESTO-Salsa screening of human microbiome strains. 435 human microbiome strains were cultivated in GMM or Gifu medium for 2 days. Sterilized supernatants were screened against 314 GPCRs via PRESTO-Salsa. 418 positive hits were observed, and 299 positive hits were validated by PRESTO-Tango. (B) Human microbiota strains from multiple host tissues whose cultures can activate any GPCRs are displayed. Known endogenous ligands are also displayed for the purpose of comparison. Comprehensive information about the human microbiota-GPCRome interaction atlas can be explored at: http://palmlab.shinyapps.io/presto-salsa/ (C) Four strains of P. gingivalis activate CD97/ADGRE5. The activities of 435 bacterial cultures on CD97/ADGRE5 were measured with a CD97 Tango assay and G protein assay. n=3 technical replicates per group (B-C). (D) Schematic of biologically relevant human microbiome-GPCRome interactions. Diverse human microbiome strains from multiple host tissues can secrete known or unknown bioactive molecules that can activate host GPCRs and potentially exert diverse regulatory roles on host (patho)physiology.

Figure 6.

Gingipain K from P. gingivalis activates CD97/ADGRE5. (A) P. gingivalis selectively activates CD97/ADGRE5. The impacts of P. gingivalis culture supernatant on 17 human adhesion GPCRs were tested with PRESTO-Tango. CD97/ADGRE5 and GPR56/ADGRG1 were activated by P. gingivalis. (B) Size fractionation (<100 kDa) and heat inactivation abolished P. gingivalis supernatant-mediated activation of CD97/ADGRE5. (C) Molecular weight considerations and the autoproteolysis model of adhesion GPCR activation implicate three P. gingivalis endopeptidases in CD97/ADGRE5 activation. (D) A gingipain K (Kgp) inhibitor abolishes P. gingivalis-mediated CD97/ADGRE5 activation. CD97/ADGRE5-transfected HTLA or CD97/ADGRE5 and SRF-Luc reporter plasmid co-transfected HEK293T were pretreated with 1 μM KYT-1 (RgpA/RgpB selective inhibitor) or 1 μM KYT-36 (Kgp selective inhibitor) before assessment of CD97/ADGRE5 activation by P. gingivalis culture. (E) Kgp-deficient P. gingivalis fails to activate CD97/ADGRE5. WT, Kgp-deficient, and RgpA/RgpB double-deficient, and RgpA/RgpB/Kgp triple-deficient P. gingivalis strains were cultivated in TSB media and had similar OD.600 values. The activities of sterilized cultures on CD97/ADGRE5 were measured via Tango assay. (F) Dose response of CD97/ADGRE5 to purified Kgp. Kgp was purified from RgpA/RgpB double-deficient P. gingivalis cultures, and titrating doses of Kgp against CD97/ADGRE5 were measured via Tango and G protein assays. Data in all panels other than C are representative of at least three independent experiments. n=3 technical replicates per group.

Figure 7.

Catalytically Active Kgp Activates CD97/ADGRE5 Via Cleavage at K290. (A) The catalytic activity of Kgp is required for the activation of CD97/ADGRE5. The activation of CD97/ADGRE5 by purified Kgp was measured in the presence of 1 μM KYT-1 or KYT-36 via Tango or G protein assays. (B) CD97/ADGRE5 harbors 22 lysines in its extracellular domain in addition to its five classic EGF-like domains and a GAIN domain that includes a GPS (GPCR Proteolysis Site). (C) K290 in CD97/ADGRE5 is critical for its activation by purified Kgp. The activities of 5 nM purified Kgp (~ 820 ng/ml) on 22 lysine to alanine mutants of CD97/ADGRE5 were measured via Tango and G protein assays. (D) K290A CD97/ADGRE5 is unresponsive to Kgp-mediated activation. (E) K290A and WT CD97/ADGRE5 exhibit similar cell-surface expression. (G) Comparison of PRESTO-Salsa and PRESTO-Tango. Data in all panels other than B and F are representative of at least three independent experiments. n=3 technical replicates per group.