Ultrahigh-throughput discovery of promiscuous enzymes by picodroplet functional metagenomics

Abstract

Unculturable bacterial communities provide a rich source of biocatalysts, but their experimental discovery by functional metagenomics is difficult, because the odds are stacked against the experimentor. Here we demonstrate functional screening of a million-membered metagenomic library in microfluidic picolitre droplet compartments. Using bait substrates, new hydrolases for sulfate monoesters and phosphotriesters were identified, mostly based on promiscuous activities presumed not to be under selection pressure. Spanning three protein superfamilies, these break new ground in sequence space: promiscuity now connects enzymes with only distantly related sequences. Most hits could not have been predicted by sequence analysis, because the desired activities have never been ascribed to similar sequences, showing how this approach complements bioinformatic harvesting of metagenomic sequencing data. Functional screening of a library of unprecedented size with excellent assay sensitivity has been instrumental in identifying rare genes constituting catalytically versatile hubs in sequence space as potential starting points for the acquisition of new functions.

Figure 1. Functional metagenomic using microfluidic droplets.

General procedure. (1) Environmental DNA (eDNA) was cloned into a high-copy plasmid and transformed into E. coli. (2) Single bacteria were encapsulated into water-in-oil droplets together with substrate (1d or 2d) and lysis agents. (3) Emulsion droplets were incubated off-chip; after single cell lysis, cytoplasmically expressed protein catalysts were able to turn over substrate. The arrow designates the droplets (∼3 × 10⁷) at the interface between fluorous oil and mineral oil (on top). (4) Emulsion droplets were re-injected (Supplementary Movie 1) into a sorting chip and strongly fluorescent droplets (‘+' channel) were separated from those with fluorescence below the threshold (‘−' channel) by dielectrophoresis (Supplementary Movie 2). (5) Selected droplets were de-emulsified and high-copy plasmid DNA was recovered following by re-transformation into E. coli. For further enrichment, iterative selections could be performed. (6) Plasmids containing eDNA coding for active catalysts were sequenced. The microfluidic devices are shown in Supplementary Fig. 5.

Figure 2. High sensitivity of the microfluidic platform allows selection and enrichment of active metagenomic variants.

(a) Minimal number of fluorescein molecules detected. (1) In a 200 μl well of a microtiter plate (conventional technology), (2) in 20 pl droplets and (3) in 2 pl droplets (according to respective calibration curves; Supplementary Fig. 2). Errors bars are set to 10% of the calculated value, corresponding to an estimation of the uncertainty of measurement. (b) Fluorescence signal distribution of 10⁷ droplets containing metagenomic cell lysate after 2 days of incubation at 22 °C. The sorting gate was set up such that droplets with two- to fivefold increased fluorescence over the population average were selected (sorting gate). All histograms corresponding to the two screening campaigns are shown in Supplementary Fig. 7. (Note: as the photomultiplier tube saturated at around 10 AU, values shown here as ∼10 AU may be higher). (c) The selection stringency was tested by monitoring the enrichment of a PTE variant PC86 over multiple rounds of microfluidic sorting. Three samples—metagenomic library before selection; DNA recovered from round 1; DNA recovered from round 2—were analysed for (i) their total plasmid content and (ii) the number of plasmids encoding PC86 by quantitative PCR. The proportion of PC86 in each DNA sample (library before selection, round 1 and round 2) was calculated by dividing the number of plasmid PC86 by the total number of plasmids. (Detailed data are shown in Supplementary Fig. 10). Error bars, s.d. from triplicates. AU, arbitrary units.

Figure 3. Unique hits isolated from metagenomic libraries.

Open reading frames (ORFs) encoding hits are highlighted in orange or green (sulfate hydrolases/transferases or PTEs, respectively). Numbers between brackets indicate the number of amino acids in the protein sequence. Grey arrows represent other ORFs isolated together with the hit sequence. As it was less obvious which of the p88.1 or p88.2 gene was most likely to code for the active triesterase, both were recloned. Selections were carried out with either sulfate ester 1d or phosphate triester 2d. (a,b) Data extracted from the Pfam database. MBL, metallo β-lactamase, also called metallo-hydrolase/oxidoreductase; AH, amidohydrolase. The other genes isolated (grey arrows) have their closest homologues in the NCBI non-redundant database predicted as: (1) transcription regulator; (2) formylglycine generating enzyme; (3) ABC transporter or membrane protein; (4) TonB-dependent receptor; (5) succinate-semialdehyde dehydrogenase; (6) carnitine dehydratase; (7) penicillin-binding protein; (8) radical SAM protein; (9) K⁺/H⁺ antiporter; (10a,b) cobalamin biosynthesis protein; (11) YKuD transpeptidase; (12) peptidase M15; (13) aminotransferase; (14a,b) permease (see Supplementary Note 1).

Figure 4. SSNs highlight the novelty of the triesterases hits.

Hits are spread over (a) the AH clan (Pfam number: CL0034) and (b) the MBL superfamily (Pfam family PF00753). Bright green nodes represent the sequences of metagenomic hits identified in this work. Annotations retrieved from the Uniprot database were used to putatively characterize sequence clusters (represented by coloured nodes). To confirm these annotations, experimentally characterized proteins (red nodes) were added into each network. Previously described OP-degrading enzymes (PTE, phosphotriesterase) are highlighted in yellow; PLL (PTE-like lactonases) are reported in the AH superfamily. 5,042 and 2,984 sequences define the AH and the MBL networks, respectively. Edges lengths represent protein sequence similarity. Only edges corresponding to similarity scores below E-values of 1 × e⁻⁹ (AH) and 1 × e⁻¹⁴ (MBL) are considered; the worst edges displayed correspond to a median 26.5% (or 27.9%) identity over an alignment length of 210 (or 218) residues for the AH (and MBL) networks, respectively. See also the position of additional hits in the α/β superfamily (Supplementary Fig. 15).

Figure 5. A triad is the catalytic feature of the α/β-hydrolase fold of the triesterase P91.

(a) The P91 structure (red) is aligned with DLH (green), a well-characterized α/β hydrolase fold (PDB ID: 1ZIC43). (b) Active site superposition with DLH reveals the catalytic triad of P91 consisting of C118, D167 and H199. E37 and H141 stabilize an alternative cysteine conformation. (c) Site-directed knockdown mutagenesis of the residues in the triad compromises P91's PTE activity (v_obs=v/[E₀]) (shown by Michaelis–Menten plot and time-course measurements using ∼1 mM of substrate (framed)). Conditions: activity towards paraoxon 2a measured at 25 °C (100 mM Tris-HCl, 150 mM NaCl pH 8.0).

Figure 6. Promiscuity is a general feature of the selected hits.

(a) Comparison of the data (catalytic efficiencies k_cat/K_M (left) and K_M values (right)) for our hits with known enzymes catalysing phosphotriester hydrolysis suggests that our screening has identified promiscuous activities. ‘OPH' summarizes PTE recovered from bacteria isolated from a pesticide-polluted environment and therefore assigned as enzymes that evolved specifically for triester hydrolysis. ‘Pro' designates enzymes for which PTE activity was shown to be a promiscuous side activity. The rates reported here are towards paraoxon 2a (for metagenomic hits) or—in the case of known PTEs—towards paraoxon 2a or methylparathion 2b. Full names (Supplementary Table 6) and reported catalytic efficiencies (Supplementary Table 7) of known PTEs are listed. (b) The catalytic promiscuity among the triesterase hits reflects bait substrate charge attributes. Substrates 1d and 2d are the bait substrates and substrates 3–8c were used for catalytic profiling. The likely bonds cleaved are represented in red. Black lines indicate that the hits share these two activities. Their width represents how often this type of enzyme promiscuity was observed. The x axis groups substrates by their charges and suggests that the bait activities select hits with consistent promiscuous activity patterns in which substrate charge is a key determinant. R₀=fluorescein; R₁=4-nitrophenyl; R₂=ethyl; R₃=methyl; R₄=glutathione (which has two additional negative charges remote from the reaction centre) R₅=C₆H₁₃. See Supplementary Fig. 12 for structures of all substrates.