Programming gene expression in multicellular organisms for physiology modulation through engineered bacteria

The study addresses the challenge of programming animal physiology using synthetic biology. While logic circuits and sophisticated control have been achieved in bacteria and mammalian cells, extending such programmable control to higher organisms is hindered by system complexity and delivery barriers. The authors leverage the lifelong association between animals and microbes to transfer decision-making logic from engineered bacteria to a host animal. Using C. elegans as a model, they propose to control worm gene expression and physiology via ingestion of engineered E. coli that produce RNA interference triggers under the control of bacterial genetic logic gates. This approach aims to create a generalizable platform for modulating animal traits (gene expression, behavior, metabolism) with external signals processed by bacterial circuits.

Prior work established programmable bacteria with logic functions, memory, oscillations, and sensing of diverse signals (light, nutrients), and engineered mammalian cells (e.g., CAR-T) with logic circuits. However, organism-level programming remains challenging due to delivery of gene editing cascades and complexity of whole-organism engineering. Animals’ microbiota influence nutrition, immunity, behavior, and metabolism, and engineered bacteria have been explored for pest control, plant growth promotion, diagnostics, and therapeutics. RNA interference (RNAi) provides a conserved, tractable mechanism for cross-kingdom gene regulation; in C. elegans, feeding bacteria expressing dsRNA effectively induces gene knockdown. SID-1-mediated transport favors longer dsRNA, affecting RNAi efficiency and enabling tunability via RNA length. These foundations motivate the current strategy of using engineered bacteria to compute and transmit RNAi signals to program animal physiology.

Model system: C. elegans (wild-type N2 and SD1084, which expresses nuclear SUR-5::GFP) fed with engineered E. coli BL21(DE3). Bacterial engineering: - Initial RNAi vector placed gfp dsRNA under a bidirectional lac promoter (IPTG-inducible). Full-length dsRNA caused silencing even without IPTG due to promoter leak. To obtain a dynamic ON/OFF range, the authors modulated the intermediary RNA length and strandedness to limit SID-1 transport and RNAi efficacy. They constructed RNA fragments from 100–400 bp in double- and single-stranded formats on pET24a derivatives. Single-stranded 200, 300, and 400 bp gfp RNAs produced clear IPTG-dependent ON/OFF in worms; ss 200 bp was selected for subsequent experiments. - Dose-response: Using the ss 200 bp gfp RNA, IPTG concentration was varied; RT-qPCR quantified bacterial RNA levels, correlating with worm GFP silencing. Genetic circuits: - AND gate: Split T7 RNA polymerase (split between aa 179–180) used to implement logical AND. PBAD (L-arabinose inducible) drives one fragment (e.g., C-terminus), and pTet (anhydrotetracycline inducible) drives the other (N-terminus). Both inducers are required to reconstitute T7 RNAP and transcribe target RNA from PT7 (e.g., ss gfp). - OR gate: Tandem PBAD and pTet promoters control transcription of full-length T7 RNAP; activation of either promoter yields T7 RNAP expression to drive target RNA from PT7. Target genes and phenotypes: - gfp (gene expression readout in SD1084). - unc-22 (twitching phenotype; RNAi induces levamisole-sensitive twitching due to locomotion defects). - sbp-1 (SREBP homolog; RNAi reduces fat storage). Constructs and controls: - For dsRNA gfp/unc-22: RNA sequence inserted between opposing T7/LacO sites on pET24a; for ssRNA: sequence downstream of a single T7/LacO. For sbp-1, T7 lysozyme (lysS) and LacI were constitutively expressed from proD promoters to reduce basal expression; proD also used to express repressors (TetR, AraC) to tighten pTet/pBAD control. - Logic gate plasmids: pTet-Ara-Split-T7-AND-Gate-proD and pTet-Ara-OR-Gate-proD co-transformed with target RNA plasmids. Culture and feeding: Engineered E. coli grown in LB with antibiotics, seeded on NGM plates supplemented with inducers. Synchronized L1 larvae placed on lawns; day-1 adults used for phenotyping after 2 days at room temperature. Induction conditions: - GFP logic experiments: L-arabinose 0.2 mg/mL and aTc 0.1 µg/mL; IPTG up to 15 µM for dose-response with lac system. - Twitching: AND: Ara 2 µg/mL and/or aTc 0.1 µg/mL with ds unc-22 400 bp; OR: Ara 0.2 mg/mL and/or aTc 0.1 µg/mL with ds unc-22 200 bp. - Fat storage: AND: Ara 2 µg/mL + aTc 0.1 µg/mL with ds sbp-1; OR: Ara 0.2 mg/mL + aTc 0.1 µg/mL with ds sbp-1. Measurements: - GFP imaging on Axiovert 200 M (FITC filter); quantification by relative fluorescence in ImageJ. - Twitching phenotype recorded immediately after levamisole (3 mM) anesthesia. - Fat storage stained by Nile Red after fixation (40% isopropanol), imaging, and quantification. - RT-qPCR of bacterial RNA (gfp) normalized to cysG to validate inducer-dependent RNA levels. Statistics and sample sizes: - GFP and fat assays: n = 15 worms per condition across three independent experiments; two-sided t-test (Bonferroni correction where applicable), p < 0.001. - Twitching: >200 worms per condition; binary outcome (all twitch or none per condition).

- Tunable RNAi via RNA length and inducer: Single-stranded gfp RNAs of 200–400 bp yielded clear ON/OFF GFP silencing dependent on IPTG; ss 200 bp selected. Dose-response showed up to 81% GFP silencing at 15 µM IPTG; bacterial RNA levels by RT-qPCR tracked worm silencing. Residual ~20% GFP in the OFF state attributed to neuronal expression resistant to RNAi. - AND gate for GFP: Only with both L-arabinose and aTc did worms show strong GFP silencing; >70% reduction in GFP intensity with both inducers; with zero or one inducer, GFP remained high (>95% of control). Bacterial RNA levels matched AND logic profiles. - OR gate for GFP: Any single or both inducers caused >90% reduction in GFP; the no-inducer control showed only a 24% reduction. RT-qPCR confirmed OR logic in bacterial RNA output. - Twitching phenotype (unc-22 RNAi): AND gate—100% of worms twitched only when both inducers were present; 0% twitching in the other three conditions. OR gate—100% twitching with one or both inducers; no twitching without inducers. - Fat storage (sbp-1 RNAi with Nile Red): AND gate—significant reductions in body size and fat storage only with both inducers. OR gate—groups with one or both inducers showed >60% reduction in fat storage; no-inducer group resembled normal controls. - Inducers themselves (L-arabinose, aTc) did not alter worm phenotypes in the absence of engineered bacterial RNA outputs, supporting that phenotypic changes were programmed by bacterial circuits. - The approach generalizes across three outputs in C. elegans: gene expression (GFP), behavior (twitching), and metabolism (fat storage), using AND/OR logic implemented in bacteria and transmitted via RNAi.

The work demonstrates that engineered bacteria can process external signals through genetic logic and transmit the resulting program to a host animal, thereby modulating gene expression and physiology. By tuning the intermediary (RNA) length and inducer levels, the authors achieved a practical dynamic range for RNAi-mediated control without extensive promoter reengineering. The AND and OR gates controlled three distinct worm outputs, indicating that bacterial logic can be faithfully translated into animal phenotypes through ingestion. The lack of inducer effects on worm phenotypes underscores that programming arises from bacterial circuit outputs. The strategy leverages the relative ease of bacterial circuit design versus animal engineering and suggests scalability: more complex logic, memory, and multi-signal sensing present in bacteria could be ported to control animal traits. For organisms where RNAi alone may be insufficient, bacterial metabolites provide alternative mediators, offering broader avenues to program host physiology and study microbe–host communication in situ.

Engineered E. coli can program C. elegans gene expression and physiology by producing RNAi triggers under the control of synthetic logic gates. The platform enables externally signaled AND/OR control over GFP expression, locomotor behavior, and fat metabolism, with dose-dependent and gate-consistent outputs validated by RT-qPCR and phenotypic assays. This bacterium-to-animal programming strategy can be extended to additional hosts (e.g., Drosophila, zebrafish, plants, mammals), more complex logic and memory circuits, and alternative mediators (e.g., microbial metabolites) for applications in agriculture, therapeutics, and basic research. Future directions include multi-input (>2) gate integration, environmental signal sensing, reducing basal leakage (e.g., T7 lysozyme strategies), expanding target gene sets, and translating the approach to higher organisms.

- The study is demonstrated in C. elegans; generalization to higher organisms may require different mediators since RNAi alone may not robustly alter physiology. - Initial promoter leakage necessitated tuning via RNA length; basal expression remains a consideration, though mitigations (e.g., T7 lysozyme, tighter repression) are discussed. - Neuronal gene knockdown is less efficient in worms, leading to residual GFP signal (~20%) in OFF states. - Dependence on synthetic inducers (L-arabinose, aTc, IPTG) limits immediate environmental applicability; future work could leverage endogenous signals or biosensors. - Phenotypic assays are limited to specific outputs (GFP, twitching, fat storage) and conditions; long-term stability, reversibility, and off-target effects were not extensively characterized.