A hybrid transistor with transcriptionally controlled computation and plasticity

The study addresses how to programmatically couple living cellular computation to organic electrochemical transistor (OECT) outputs. OECTs are well-suited for biointerfaces due to aqueous operation, mixed ionic/electronic conduction, high transconductance, and low voltage operation, making them attractive for sensing and neuromorphic computing. Typical OECT control uses gate voltage to modulate ion-driven doping in the channel, but biological redox reactions can act as a secondary gate. Prior biofunctionalizations (lipid bilayers, ion channels, enzymes) enable specific sensing but limited on-device computation. Engineered bacteria can perform complex computations (Boolean logic, analog/digital processing, neuromorphic-like behavior), yet reliably interfacing genetic circuit outputs with OECTs has been challenging. The research question is whether extracellular electron transfer (EET) from electroactive bacteria can be mechanistically harnessed and transcriptionally controlled to modulate PEDOT:PSS OECT channel doping, thereby converting biological computation into electrical signals and enabling synaptic plasticity.

Prior work has shown OECTs as platforms for biosensing and neuromorphic functions due to volumetric ionic gating and high sensitivity. Biomembrane functionalization and enzyme coupling have enabled chemical and biological detection but with limited computational complexity. Electroactive bacteria like Shewanella oneidensis and Geobacter sulfurreducens perform EET to external acceptors (including electrodes and conducting polymers) and have been used in microbial fuel cells and emerging bioelectronic sensing. Synthetic biology has produced genetic circuits regulating EET-related genes for programmable redox outputs. A notable precedent is Méhes et al., who monitored Shewanella EET with p-type OECTs, demonstrating feasibility but lacking genetic programmability and detailed bacteria–OECT interaction mechanisms. OECTs offer advantages over traditional bioelectrochemical cells: no biofilm requirement, faster response, signal amplification from nanoampere biological signals to milliampere transistor currents, compatibility with selective channel functionalization, easy fabrication, and high-throughput formats. The current study builds on these by combining mechanistic analysis and genetic control to tie bacterial computation to OECT behavior and synaptic functions.

• Device fabrication: Planar p-type PEDOT:PSS OECTs on quartz slides with Ti/Au gate, source, and drain. PDMS formed chambers. Some devices modified for in situ UV-Vis with larger channel area; a two-electrode variant removed the gate for isolating direct channel interactions.
• Operation conditions: Typical biases VDS = −0.05 V (or 0.2 V in some viability tests), VGS varied (−0.5 to 0.5 V). Ag/AgCl pellet pseudo-reference electrodes used for accurate potential measurements (source and gate potentials vs RE). Continuous operation under anaerobic conditions in Shewanella Basal Medium with defined carbon sources (e.g., 20 mM lactate). No fumarate for most experiments to avoid alternative electron acceptors.
• Biology: Shewanella oneidensis MR-1 as wild-type. Viability and colonization assessed via LIVE/DEAD staining, CFU counts, and OD600 measurements within devices. Carbon source dependence tested (lactate, pyruvate, acetate, starvation). Controls included heat-killed and lysed S. oneidensis, culture supernatant, and non-electroactive E. coli MG1655.
• Signal analysis: Channel current decay (IDS) after inoculation fitted to a single exponential to extract a rate constant characterizing EET-driven de-doping. Transfer curves measured pre/post-incubation. Gate voltage dependence of decay measured. Gate currents recorded during voltage scans.
• Spectroscopy: In situ UV-Vis of the PEDOT:PSS channel to monitor neutral (~650 nm) and polaronic (~900 nm) absorption features as a function of bias and in the presence of cells to confirm de-doping.
• Two-electrode devices: Gate removed to isolate direct channel reduction by cells; VDS held constant to monitor IDS decay.
• Genetic perturbations: Knockouts Δbfe (reduced flavin secretion), Δlysis (impaired biofilm formation), ΔmtrC (outer membrane EET cytochromes MtrC, OmcA, MtrF deleted), and Δmtr (additional deletions of periplasmic carriers mtrA, mtrD, dmsE, SO4360, cctA). Exogenous flavin mononucleotide (FMN) titrations to probe mediated EET.
• Genetic complementation (Buffer gates): Plasmid-based expression of mtrC under Ptac symo promoter induced by IPTG in ΔmtrC; mtrCAB operon under OC6 (AHL) control in Δmtr (Δmtr + mtrCAB). Strains pre-induced 18–24 h anaerobically (1 mM IPTG or 100 nM OC6) before OECT inoculation.
• Boolean logic gates: NAND (inputs: IPTG, OC6) and NOR (inputs: OC6, aTc) transcriptional circuits controlling mtrC in ΔmtrC background. Overnight induction across input combinations, then continuous IDS monitoring and rate constant extraction; 2D heat maps of responses. Source potentials (Vs vs Ag/AgCl) and normalized transfer curves evaluated to confirm logic states.
• Synaptic behavior assays: Transfer curve hysteresis measured by sweeping VGS between −0.5 V and 0.5 V at VDS = −0.05 V. Paired pulse tests defined by pulse amplitude VP, pulse duration tP (typically 80 ms), and interval Δt. Short-term plasticity indices (A2/A1) and conductance change ΔG measured 30 s after pulses. Spike-timing dependence tested by varying Δt. Long-term modulation evaluated via repeated stimulus sessions alternating positive and negative pulses (VP = ±0.5 V, tP = 55 ms, 30 repeats per session). Time constants extracted via one-phase exponential fits. Oxygen exposure used to reset channel doping after EET-induced de-doping. EET-deficient mutants complemented with mtrC/mtrCAB used to link EET to synaptic modulation. Abiotic controls included for all assays.

• Viability and colonization: S. oneidensis maintained 67 ± 14% viability at 24 h in OECT electrolyte without fumarate; with fumarate, viability improved to 80 ± 10% and robust growth occurred, but fumarate was excluded from subsequent tests to avoid confounding electron acceptor effects.
• EET-linked de-doping: Inoculation with metabolically active S. oneidensis caused pronounced IDS decay; decay rate constant scaled with initial cell density (OD600) and depended on metabolizable carbon source (lactate ≈ pyruvate > acetate ≈ starvation ≈ abiotic).
• Specificity controls: Heat-killed and E. coli MG1655 produced minimal IDS decay; lysed S. oneidensis caused slower, linear decay, suggesting released redox components contribute but are less effective than live-cell EET.
• Electrical characterization: Transfer curves after 24 h incubation showed decreased channel current and source potential, consistent with channel reduction. Decreasing effective gate potential correlated with |IDS| reductions under fixed VGS and VDS.
• Spectroscopy: In situ UV-Vis showed increased 650 nm (neutral PEDOT) and decreased ~900 nm (polaron) absorbance in biotic devices, matching de-doped abiotic channels biased at ~0.5 V vs Ag/AgCl, confirming biological de-doping of PEDOT:PSS.
• Direct vs mediated EET: Two-electrode (gate-less) devices showed IDS decay comparable to three-terminal OECTs (VGS = 0.2 V), confirming direct cell–channel electron transfer. In three-terminal devices, more positive gate bias increased decay rate constants, indicating that electrons delivered to the gate can reduce the channel via the external circuit.
• Role of flavins and biofilms: Δbfe (reduced flavin secretion) and Δlysis (impaired biofilm) showed no significant difference from wild-type within 24 h, indicating minimal requirement for flavin secretion or biofilm in this timescale. Exogenous FMN (1 µM) increased decay rates across strains; dose–response was sigmoidal with increasing FMN, consistent with FMN-bound MtrC-mediated transfer.
• EET pathway dependence: EET-deficient ΔmtrC and Δmtr strains showed significantly reduced decay rate constants versus wild-type. Complementation with induced mtrC (ΔmtrC + mtrC) or mtrCAB (Δmtr + mtrCAB) restored higher decay rates; gate I–V scans showed increased oxidation currents only in induced complements, confirming EET activity.
• Genetic logic to electrical output: ΔmtrC strains carrying NAND (inputs IPTG, OC6) and NOR (inputs OC6, aTc) gates exhibited IDS decay rate constants matching truth tables across inducer combinations. Mean source potential shifts between logic 1 and logic 0 were 101.4 mV (NAND) and 139.6 mV (NOR). Normalized transfer curves shifted toward more negative Vs (more positive Voff) for logic 1 and wild-type controls.
• Synaptic behavior: Biotic OECTs displayed pronounced transfer curve hysteresis and asymmetric short-term plasticity: positive gate pulses (VP > 0) increased channel conductance and reduced A2/A1 with decreasing Δt; negative pulses produced negligible changes, similar to abiotic controls. Repeated stimulus sessions yielded consistent, reversible modulation with average time constants ~429.5 ± 18.6 s (positive sessions) and ~1083.1 ± 71.0 s (negative sessions). Complemented ΔmtrC + mtrC and Δmtr + mtrCAB reproduced MR-1-like ΔG and A2/A1, while empty-vector controls resembled abiotic devices, linking synaptic effects to Mtr-dependent EET.
• Reset and stability: Transition from anaerobic to aerobic conditions (3 h ambient O2) recovered ~75% of channel current and source potential toward initial values, demonstrating a pathway to reset synaptic weight after EET-induced de-doping.

The study demonstrates that living bacterial extracellular electron transfer can directly modulate the doping state and conductivity of PEDOT:PSS OECT channels, effectively coupling biological electron transport to transistor outputs. By tuning gate bias, the kinetics and extent of EET-driven de-doping can be modulated, highlighting a dual-gate paradigm where the biological redox processes act as a secondary gate. Mechanistic experiments (in situ UV-Vis, gate-less devices, reference electrode measurements) confirm both direct cell–channel electron transfer and gate-mediated pathways contribute to channel reduction. Genetic perturbations and complementation establish that the Mtr pathway is necessary and sufficient to drive the observed electrical responses; exogenous flavins further accelerate de-doping, while endogenous flavin secretion and biofilms are not required on 24 h timescales. Importantly, transcriptional logic controlling EET genes (NAND/NOR) translates reliably into electrical signatures (rate constants, source potential shifts, transfer curve shifts), enabling on-device biological computation readouts. Furthermore, EET correlates with distinct synaptic behaviors (hysteresis, paired-pulse modulation, reversible weight changes), providing a biologically programmable route to synaptic plasticity in OECTs. These findings extend OECT functionality beyond sensing ions or biomolecules to harnessing genetically controlled electron transport, opening avenues for biosensing, neuromorphic biocomputing, and hybrid devices where biological computation directly informs electronics.

This work introduces hybrid OECTs that leverage genetically controlled extracellular electron transfer from Shewanella oneidensis to modulate PEDOT:PSS channel doping, thereby converting transcriptional computations into electrical outputs and enabling synaptic plasticity. Key contributions include: (1) mechanistic validation that bacterial EET de-dopes PEDOT:PSS through direct and gate-mediated pathways; (2) genetic control (buffer gates and Boolean NAND/NOR) that maps transcriptional states to transistor responses; and (3) demonstration of EET-linked short-term synaptic behaviors and reversible weight modulation. Future directions include isolating gate- versus channel-specific EET (e.g., spatial segregation, ion-permeable membranes, multi-/floating-gate architectures), deploying reference electrodes for precise channel potential control and device reset, exploring alternative channel materials (including n-type or inorganic semiconductors) and post-processing to optimize bacteria–polymer interactions, extending to other electroactive organisms or pathways, and integrating broader genetic regulatory motifs for enhanced dynamic range and computational complexity. These developments will expand OECT-based biosensing and biocomputing platforms with modular, programmable biological interfaces.

• Incomplete separation of gate- and channel-based electron transfer: cells were present on both electrodes, preventing full isolation of direct vs gate-mediated pathways without extreme biases that risk side reactions and cellular stress.
• Use of a polarizable Au gate complicates precise potential control and onset of redox processes due to capacitive effects; Ag/AgCl pseudo-reference electrodes were required for accurate measurements.
• Timescale limitations: endogenous flavin secretion and biofilm formation may play larger roles over longer periods than the ~24 h window studied.
• Long-term stability and plasticity mechanisms remain unclear; metabolic state changes, viability over extended operation, and device/material aging need further study.
• Material specificity: findings are based on PEDOT:PSS; different processing or alternative (especially n-type) materials may exhibit different bacteria–polymer interactions and performance.
• Quantitative deconvolution of per-cell EET contributions and absolute charge transfer at the biointerface was not resolved; single-cell level measurements would refine models.