Programming bacteria for multiplexed DNA detection

The study addresses the challenge of developing specific, selective, and orthogonal biosensors that allow engineered bacteria to detect and respond to key environmental signals. While prior synthetic biology efforts have focused on sensing physical and chemical cues (e.g., light, ultrasound, quorum molecules), there remains a scarcity of well-characterized orthogonal signals to distinguish different bacterial species within communities. The authors propose extracellular DNA (eDNA) as a broadly available, programmable biomarker that can be exploited by naturally competent bacteria via homologous recombination during natural transformation. The research question is whether Bacillus subtilis can be engineered with a synthetic genetic circuit to selectively sense target DNA sequences from specific species, enabling robust, specific, and multiplexed detection, including in complex samples and without DNA extraction.

The paper builds on work showing engineered bacteria can sense environmental inputs and regulate population behaviors or antimicrobial production. Prior biosensors target physical or chemical signals, but few orthogonal signals exist for discriminating species in microbial communities. Natural competence and horizontal gene transfer via transformation are widespread and can mediate DNA uptake and recombination requiring sufficient sequence identity and length, suggesting a route to high specificity. The authors reference mechanisms of competence regulation and RecA-mediated homologous recombination in B. subtilis, prior uses of synthetic genetic circuits for pathogen sensing, and the prevalence and roles of eDNA in microbial ecosystems. They also note alternative DNA detection methods (qPCR, NGS) require instrumentation, whereas living cell-based approaches could be simpler and cost-effective. The discussion references advances in genetic circuit containment/kill switches, chassis engineering for higher competence, and potential integration with CRISPR to enhance specificity to the single-nucleotide level.

• Chassis and circuit design: The naturally competent bacterium Bacillus subtilis PY79 was engineered with a modular synthetic circuit. Competence was induced by a xylose-inducible comK (PxylA-comK) master regulator. A negative selection module consisted of an IPTG-inducible toxin–antitoxin system (txpA-ratA) and a LacI repressor governing both the toxin and a fluorescent reporter (GFP or orthogonal reporters for multiplexing). The toxin/repressor cassette was flanked by species-specific target DNA sequences ("landing pads") to enable homologous recombination upon uptake of matching extracellular DNA.
• Sensing mechanism: Upon xylose addition, a subpopulation enters competence, imports single-stranded DNA, and RecA mediates homology search. If environmental DNA contains homologous target sequences, recombination removes lacI and txpA-ratA from the genome in transformed cells. Subsequent IPTG addition induces the toxin in nontransformed cells, selectively inhibiting their growth and enriching transformed cells, which then express fluorescence.
• Target sequences and sensor strains: The authors built sensors for four species by replacing landing pads with conserved, representative sequences: E. coli (xdhABC operon), Salmonella typhimurium (sipBCDA pathogenicity island), Staphylococcus aureus (hemEH heme biosynthesis), and Clostridioides difficile (pheST phenylalanyl-tRNA synthetase). Conservation across strains and potential cross-species homology were analyzed by nucleotide BLAST (NCBI), assessing identity and coverage.
• Homology length optimization: Landing pad homology lengths from 0.5 to 2.5 kb (each flank) were tested. Transformation efficiency (ratio of transformed to total B. subtilis CFU) was measured over time with and without 100 ng/mL purified target gDNA. Effects of homology length on performance were quantified.
• Sensitivity and quantitative response: Time-series fluorescence in liquid culture after transformation with a range of gDNA concentrations (0–1500 ng/mL) was measured. A fixed fluorescence threshold defined the detection time. Linear fits of detection time versus log DNA concentration quantified the dynamic range; slopes relate to growth rate and intercepts to background mutation frequency.
• Specificity tests: Each sensor was challenged with gDNA from all four species and from closely related species harboring homologous but nonidentical sequences (e.g., S. epidermidis hemEH homolog; C. hiranonis pheST homolog). Fluorescent colony counts on selective plates provided readouts.
• Multiplexing: Orthogonal reporters (EC-G: GFP; ST-R: RFP; SA-B: BFP) enabled simultaneous detection. Mixed cultures of sensors were transformed with combinations of target gDNAs and plated to read out colored colonies.
• Complex community sensing: A six-species anaerobic community (including SA and ST plus four commensals) was assembled. Community composition over passages was assayed by 16S rRNA gene sequencing. Purified community gDNA was introduced into mixed SA-G and ST-R sensors to compare sensor readouts with sequencing-based abundance trends.
• Detection without DNA extraction: Co-culture of each sensor with live target species (initial OD600 ~0.1; ~10^7 CFU/mL) tested direct detection. Spectinomycin was used to inhibit target growth and enhance eDNA release; sensors carried resistance markers. DNase I controls verified transformation dependence on extracellular DNA. Heat treatment (90 °C, 10 min) of targets was evaluated as an antibiotic-free method to enhance eDNA and detection limit. Spike-in experiments with heat-treated E. coli and S. typhimurium in mouse cecal contents (10 mg) assessed multiplexed detection in complex matrices.
• Validation: PCR and Sanger sequencing confirmed deletion of lacI and txpA-ratA in transformants. Escape mutants were characterized by sequencing lacI/txpA. Experimental details included media, culture conditions, antibiotic concentrations, DNA extraction kits, and imaging/plate reader settings as specified in Methods.

• Circuit function and homology requirements: Transformation efficiency increased over ~10 h and scaled with landing pad homology length. A homology length ≥1 kb per flank was required to exceed background escape mutants (10^-7–10^-6 frequency), and 2.5 kb flanks yielded high performance (>10^2-fold above background). Sequencing confirmed precise recombination removing txpA-ratA and lacI. Escape mutants carried mutations in txpA or lacI that reduced toxicity.
• Sensitivity and quantitative behavior: All four sensors detected their target gDNA at 100 ng/mL over background. Detection limits (lowest concentration with significantly earlier detection time than no-DNA control) were: E. coli EC sensor 1–4 ng/mL; S. aureus SA sensor 4–16 ng/mL; C. difficile CD sensor 4–16 ng/mL; S. typhimurium ST sensor 62.5 ng/mL (approx. 10^5–10^6 genome copies/mL for EC/SA/CD; ~10^6 for ST). Detection time was linearly related to log10 gDNA concentration with high goodness-of-fit (e.g., EC: y = -0.575 log(x) + 10.9, R^2 = 0.980; ST: -0.438 log(x) + 9.97, R^2 = 0.957; SA: -0.465 log(x) + 9.39, R^2 = 0.980; CD: -0.499 log(x) + 9.97, R^2 = 0.999), indicating quantitative sensing across wide ranges.
• Specificity: For each sensor, detection times with cognate gDNA (6.1–7.1 h) were substantially earlier than with non-target gDNA or no DNA (9.1–10.7 h). The SA sensor differentiated S. aureus from closely related S. epidermidis despite 89% coverage and 77% identity of hemEH; similarly, the CD sensor distinguished C. difficile from C. hiranonis (87% coverage, 75% identity of pheST). Bioinformatic analyses showed varying degrees of cross-species conservation, consistent with observed specificity.
• Multiplexing: Mixed cultures of sensors with orthogonal fluorescent reporters accurately reported all presence/absence combinations of EC, ST, and SA gDNA in purified samples.
• Community samples: In a six-species synthetic gut community, sensor readouts from community gDNA tracked temporal abundance trends observed by 16S rRNA gene sequencing: SA decreased over time while ST remained relatively constant. SA sensor mirrored sequencing trends more closely, consistent with its higher sensitivity.
• Direct detection without DNA extraction: Co-culture with targets plus spectinomycin boosted detection for EC, ST, and SA; DNase I significantly reduced transformant counts, confirming dependence on extracellular DNA. Heat treatment (90 °C, 10 min) of E. coli improved detection limit to 5 × 10^6 cells/mL. In mouse cecal contents (10 mg), heat-treated spike-ins of EC and ST were detected multiplexedly with limits of 10^7 cells/mL (10^6 cells/gram). High target densities in single-target samples increased false positives in multiplex settings, indicating room for optimization.

Engineering B. subtilis with a competence-activated, recombination-gated kill-switch circuit enables selective sensing of long, homologous DNA sequences from the environment. The system achieves high specificity determined by stringent homology length and identity requirements, supports quantitative readouts (detection time scales with log DNA amount), and allows multiplexing via orthogonal fluorescent outputs. Importantly, it can detect targets directly from eDNA released by pretreated cells, including in complex matrices, and track relative abundance trends in communities. The approach offers a simple, low-cost alternative to instrument-heavy DNA detection methods (qPCR, NGS), with potential fieldability via robust B. subtilis spores. However, sensitivity is constrained by transformation efficiency and background escape mutations in the negative selection module. Improving chassis competence (e.g., via strain engineering or alternative naturally competent species), enhancing counter-selection stringency (e.g., more efficient kill switches), or integrating CRISPR-based discrimination could lower detection limits and accelerate detection times. The modular design suggests broad applicability to other organisms (viruses, fungi, mammalian DNA) and coupling to effector responses (e.g., antimicrobial production) for responsive sensing-control systems.

The authors devised a modular, living cell-based DNA sensor in B. subtilis that uses competence-induced homologous recombination to excise a kill switch and enable fluorescent reporting upon encountering target DNA. They demonstrated high specificity across four species (E. coli, S. typhimurium, S. aureus, C. difficile), quantitative performance across wide DNA concentration ranges, multiplexed detection with orthogonal reporters, and direct detection without DNA extraction in co-cultures and mouse cecal matrices. This establishes a foundation for programmable, multiplex DNA sensing in vitro and potentially in situ. Future directions include enhancing sensitivity by increasing transformation efficiency and reducing background mutation rates via improved counter-selectable markers; adopting alternative highly competent chassis; integrating CRISPR for nucleotide-level specificity; optimizing for complex samples to minimize false positives at high target densities; and coupling sensing to therapeutic outputs for closed-loop control in microbiomes or environmental applications.

• Sensitivity: Current detection limits (ng/mL DNA; 10^6–10^7 cells/mL in complex matrices) may be insufficient for many environmental or clinical contexts; improved transformation efficiency and reduced background are needed.
• Background escape mutants: Mutations in lacI or txpA generate false positives; more stringent or redundant counter-selection designs are necessary.
• Detection time: Reliance on growth and selection imposes multi-hour detection times; faster chassis or circuit tuning could shorten times.
• Specificity scope: The study focused on species-level discrimination; strain-level resolution was not systematically evaluated and may require unique landing pads or CRISPR augmentation.
• Multiplex in complex matrices: High target cell densities increased false positives in multiplex assays with cecal contents, indicating matrix effects and cross-talk to be optimized.
• Dependence on pretreatments: Antibiotic or heat treatment was used to enhance eDNA release; such treatments may not be universally applicable in situ and could affect sample composition.
• Limited target panel: Only four bacterial targets were tested; generalization to diverse taxa (including viruses/fungi) remains to be validated experimentally.