Engineering of synthetic bile-salt receptors in E. coli using the EMeRALD platform.
Enteropathogenic bacteria such as Vibrio cholerae or Vibrio parahaemolyticus cause acute intestinal infections mediated by toxin secretion34. These pathogens use bile salts as an intestinal location signal to activate their virulence pathways. Bile salt sensing is mainly under the control of inner membrane sensor/cofactor couples, TcpP-TcpH for V. cholerae31 and VtrA-VtrC for V. parahaemolyticus35. As using pathogens as biosensors would involve significant host-specific regulation correction35,36 and biosafety containment issues, an alternative strategy is to rewire pathogen sensing modules of interest into a modular platform operating in a surrogate host (Fig. 1a).
The EMeRALD receptor platform, which we recently designed33 (Fig. 1b), is derived from membrane-bound one-component systems, which are bitopic proteins with a typical architecture of a cytoplasmic DNA Binding Domain (DBD), a juxtamembrane linker, a transmembrane region, and a periplasmic Ligand Binding Domain (LBD). Direct fusion between the LBD and the DBD provides a simple yet efficient solution to transduce incoming signals into a transcriptional output37,38. The EMeRALD platform operates in E. coli and uses the DBD from the CadC pH and lysine sensor which is inactive in its monomeric state39. Ligand-induced dimerization of the LBD triggers dimerisation of the cytoplasmic DBD and transcriptional activation40. This straightforward mechanism offers the potential to modularize receptor sensing and signaling by domain swapping. We previously built a synthetic receptor responding to caffeine by using a nanobody for this ligand as LBD33.
To engineer a chimeric bile-salt receptor in E. coli, we fused the V. cholerae TcpP bile-salt sensing module and its transmembrane region (TM) to the DNA binding domain of CadC (Fig. 1c). As a reporter, we placed superfolder Green Fluorescent Protein (sfGFP)41 under the control of the CadC target promoter, pCadBA. We also expressed the cofactor protein TcpH, previously described to protect TcpP from proteolysis by the V. cholerae RseP protease once dimerized in response to bile salts42. We first confirmed using inducible gene expression systems that co-expression of the TcpH cofactor was necessary for TcpP function, using the primary bile salt taurocholic acid (TCA) as a ligand (Supplementary Fig. 1–3). These data suggest that the RseP homolog present in E. coli (UniProt:P0AEH1) can degrade TcpP when not bound by TcpH. We also found that the relative expression level of CadC-TcpP and TcpH were critical parameters affecting system performance (Supplementary Fig. 3).
We then placed both proteins under the control of constitutive promoters43. We used the strong constitutive promoter P5 for TcpH and three different constitutive promoters of increasing strengths, P9, P10, and P14, for CadC-TcpP (Fig. 2a) and tested their response to TCA42 (Fig. 2b, Supplementary Fig. 4–5). We found that the P9-CadC-TcpP variant had the lowest LOD, highest dynamic range, and highest signal strength. These data confirm that the stoichiometry between CadC-TcpP and TcpH is a key parameter influencing receptor performance.
We then assessed the versatility of the EMeRALD platform by connecting the VtrA/VtrC sensor system from V. parahaemolyticus32 (Fig. 2c). We built a dual-expression system consisting of P9-CadC-VtrA and P5-VtrC, and tested its response to its canonical ligand taurodeoxycholic acid (TDCA). We found that the VtrA/VtrC EMeRALD system was functional with a slightly higher LOD, similar dynamic range and signal strength than the TcpP/TcpH EMeRALD system (Fig. 2d, Supplementary Fig. 6). These results highlight the modularity and scalability of the EMeRALD platform, which supports the connection of different sensing modules to the receptor scaffold.
Synthetic bile salt receptors exhibit different specificity profiles.
We then assessed the specificity profile of the synthetic bile salt receptors. Bile salts are classified in two categories: primary bile salts (including taurocholic acid) are produced by the liver while secondary bile salts arise from modification of primary bile salts by gut microbiome metabolism. Primary bile salts are upregulated in serum and urine of patients with liver disease27–29. Previously identified virulence activating factors for V. cholerae include taurocholic acid, glycocholate, and cholic acid42. We measured the response of the bactosensor to a panel of twelve different bile salts, including both primary and secondary types (Fig. 2e, Supplementary Fig. 7–8). Interestingly, the CadC-TcpP system was highly specific for primary conjugated bile salts (especially TCA and GCDCA), while not responding to secondary bile salts. On the other hand, the CadC-VtrA system has a larger spectrum of bile salt specificity, mainly responding to secondary conjugated bile salts GDCA and TDCA. Due to the link between primary bile salts and liver diseases, we selected the TcpP/TcpH system to develop a bile salt bactosensor for medical diagnosis.
Directed evolution of TcpP sensing module for improved LOD and higher sensitivity.
Sensor sensitivity and LOD are key parameters for biosensors applications. We aimed to identify key residues determining the sensitivity of the TcpP sensing module, and targeted those to improve synthetic receptor sensitivity and LOD. To do so, we coupled comprehensive mutagenesis with functional screening and Next-Generation Sequencing (NGS), an approach which supports the identification of functional variants together with the sequence determinants within local structural motifs44–46. This strategy has also been used to engineer orthogonal two-component systems47.
Transition from intramolecular to intermolecular disulfide bonds between TcpP monomers is a key determinant of TcpP response to bile salts and is mediated by two cysteine residues, Cys207 and Cys21842. By performing multiple sequence alignments of different TcpP bacterial homologs (Supplementary Fig. 9), we found a significant conservation of the amino-acids flanked by these two cysteines (Fig. 3a). Secondary structure prediction and ab initio 3D prediction using the Rosetta modelling suite (Fig. 3b and Supplementary Fig. 10) suggested that each cysteine was located in rigid beta-sheets separated by a flexible loop region between Asn211 and Gln214. This loop propensity to form a turn would allow the two beta sheets and the cysteines to come in close proximity and form an intramolecular disulfide bond. We hypothesized that the flexibility of the turn region between Cys207 and Cys218 was a key parameter controlling the transition rate between the two states, and that altering its amino-acid composition could change the system’s sensitivity to bile salts.
We thus built a comprehensive mutational library (NNK x 4, theoretical library complexity ≌ 1.05 x 106 variants, see methods for details) targeting the NYEK residues inside the turn, and cloned it into the plasmid constitutively expressing CadC-TcpP and producing GFP in response to bile salts (Fig. 3c). The resulting library was induced with TCA, and GFP-positive variants were isolated by fluorescence-activated cell sorting (FACS). We performed three rounds of enrichment (200\(\mu\)M of TCA as ligand in 1st and 2nd rounds of selection, and 20 µM for the 3rd round) and observed an increasing fraction of the cell population responding to different ligand concentrations (20 to 80 µM) (Fig. 3d). We collected, cultured, and sequenced single variants and tested their response to TCA (Fig. 3e). We found that comprehensive mutagenesis of residues Asn210 to Gln213 could alter the limit of detection, the sensitivity, and the fold activation of our biosensor. The 3.3-fold difference in LOD between V18 and V22 (EC50 from 28.3 to 92.5 µM, Table 1) indicated the broad range of sensitivity engineering obtained by mutating the loop region of the TcpP sensing module. Further kinetic analysis revealed that the sequence variation of this loop region changes reaction speed and system interaction of bile salts with the synthetic receptor (Supplementary Fig. 11).
To better understand the sequence features influencing the response of TcpP to bile salts, we sequenced the whole pool of enriched variants by next-generation sequencing (NGS). Surprisingly, the sequence features of functional variants were different from those expected from natural TcpP homologs (Fig. 3f and Supplementary Fig. 12). First, and in contrast with wt TcpP homologs, we observed a strong depletion of long-chain, negatively charged amino-acids (Asp and Glu) along with long-chain polar amino-acids (Asn and Gln) at position 211. Lysine at position 211 also appeared to be depleted in functional variants (despite being commonly found at this position in other TcpP homologous proteins). Second, amino acids with bulky aromatic side chain such as Phe and Tyr were highly conserved in selected functional variants, but not among TcpP homologs (with the notable exception of Tyr211 in V. cholerae TcpP), strongly indicating the important role of aromatic residues at position 211 in the function of V. cholerae TcpP. We chose the best engineered variant, termed TcpP18, for further development of a clinical bactosensor.
Development of a colorimetric version of the bactosensor.
Colorimetric assay provides a simple and intuitive method for simple and direct estimation of test results by the naked eye. In addition, colorimetric assays support straightforward development of quantitative assays using smartphone-based platforms for POC or home-based diagnosis48. We used TcpP18 coupled with the reporter beta-galactosidase LacZ (termed TcpP18-LacZ) and its substrate chlorophenol red-β-D-galactopyranoside (CPRG) to provide a colorimetric output49 (Fig. 4a, see methods for details). Similarly to the biosensor equipped with a GFP output, the bile salt specificity profile of the TcpP18-LacZ system was slightly shifted from TCA to GCDCA (Fig. 4b, Supplementary Fig. 13). We thus evaluated the LOD and signal output threshold of TcpP18-LacZ in response to increasing concentrations of GCDCA. We also explored the influence of varying cell density and incubation time (Fig. 4c and Supplementary Fig. 14). Increasing cell density or incubation time both improved the dynamic and operating ranges of TcpP18-LacZ; however, background signal also increased. After optimization, the Tcp18-LacZ was able to detect clinically relevant bile salt concentrations, such as those corresponding to early cirrhosis (~ 5.2 µM GCDCA)50 or nonalcoholic fatty liver diseases with hepatocellular carcinoma (~ 56 µM GCDCA)51.
Bactosensor-mediated detection of elevated bile salts levels in serum from patients with liver transplant.
We then prototyped our bactosensor for the detection of bile salts in clinical conditions. To do so, we tested the sensor on samples from patients having undergone liver transplantation. After liver transplant, the main complications are bile ducts stenosis and acute cellular rejection. In order to detect these complications at an early stage, liver tests are performed regularly. Serum bile salt concentration has been shown to be a good indicator for the assessment of liver dysfunction after liver transplantation52. A field-deployable method for bile salt assessment would greatly improve the monitoring of these patients, ultimately allowing fine grained monitoring performed at-home by the patients themselves.
We tested our bactosensor in clinical 21 serum samples from liver transplantation patients (Fig. 5a and Supplementary Fig. 15). The patients were followed at the Montpellier hospital after their liver transplant, most of them having been performed in the last 2 years (see Supplementary Information). These patients had received a liver transplant for end-stage liver disease as a result of alcoholic related liver disease or non-alcoholic fatty liver disease, chronic cholangitis or liver cancer. A complete hepatic check-up was performed, and serum bile salts were also measured (Supplementary Tables 3–5 for clinical data). We found that patients who had a high potential of acute cellular rejection (ACR) after liver transplantation (serum bile acid > 37 µM)53 had significant and visible colorimetric signal changes (Supplementary Fig. 16) in bactosensor assays. Three patients in particular raised our attention: patients #5, 10 and 13. These three patients had elevated serum bile salts concentration. Two of them (5 and 10) presented abnormalities in their hepatic enzymatic values (ASAT, ALAT, GGT, Pal, and bilirubin). For these patients, the bile salt bactosensor produced the strongest colorimetric change easily detectable with the naked eye (Fig. 5b). We further compared the results of the bactosensor with a commercial bile salt enzymatic assay (Fig. 5c). The high correlation between enzymatic serum bile salt assay and our bactosensor measurements (r = 0.908, p < 0.0001) indicates the reliability of our assay. In addition, we compared the results from the bactosensor with other serum liver biomarkers (Supplementary Fig. 17), and found a high correlation to a common liver biomarker, total bilirubin. These results indicate that our bactosensor is able to provide a simple, reliable, and cost-effective method for monitoring patient condition after liver transplantation.