2D printed multicellular devices performing digital and analogue computation

The work addresses limitations of traditional biological computational device architectures that separate input, processing, and output layers. When applied to living systems, challenges such as the wiring problem, genetic complexity, metabolic burden, scalability, reusability, and engineering complexity hinder development and technology transfer. Prior multicellular approaches improved complexity but co-culturing different strains raises stability and reproducibility issues. The authors propose a new framework that abandons the conventional layered architecture and uses physical space as a computational element. Information is encoded in the concentration of a single carrying signal (AHL) diffusing over a 2D substrate where spatially arranged cell-based modulatory elements implement computation. The study aims to demonstrate robust, scalable, low-genetic-burden, printable cellular devices on flexible substrates (paper) capable of digital and analogue computations for potential low-cost, out-of-lab applications.

The paper situates itself within synthetic biology circuit design and multicellular computing literature. It discusses known constraints (wiring, burden, context effects) and notes EDA-inspired design tools that do not fully solve scalability and transferability. Prior distributed/multicellular computation and spatial segregation approaches showed promise but often required multiple wiring molecules or faced complex experimental setups and stability issues. The authors draw inspiration from printed electronics and paper-based devices, as well as quorum sensing-based signaling and AHL degradation (Aiia) mechanisms, to enable a single-molecule, spatially encoded computation on paper. References include foundational works on genetic circuits, multicellular NOR-based computing, quorum sensing, printed/flexible electronics, and prior paper-based biological systems.

Architecture and components: Information is encoded in a single carrying signal (AHL) produced by engineered source cells and diffusing over a 2D surface. Spatially distributed cell types modulate AHL flow. Modulators either decrease/abolish AHL (negative modulators, M−) upon input or increase AHL flow by relieving constitutive degradation (positive modulators, M+). An auto-amplifier (CA) boosts AHL upon detecting AHL to compensate diffusion losses. Reporter cells (CR) generate output (GFP) when AHL exceeds a threshold. All strains constitutively express RFP for normalization of cell population. Engineered cell library: S1 constitutively produces AHL; S2, S3, S4 produce AHL upon induction by arabinose, rhamnose, and mercury, respectively. Modulators express the AHL-degrading enzyme Aiia: M− cells induce Aiia in response to external inducers (aTc, arabinose, rhamnose), reducing AHL; M+ cells constitutively express Aiia but external inducers suppress Aiia, increasing AHL. CA cells auto-amplify AHL; CR reporter cells express GFP in response to AHL. Genetic parts and architectures are detailed in Supplementary materials. Substrate and cellular inks: Standard 75 gsm white paper is used due to compatibility with bacterial growth and AHL diffusion. Cellular inks are prepared by mixing cell cultures (OD ~0.2) with LB-agar (for ready-to-use circuits) or LB with 20% glycerol (for frozen storage), plus appropriate antibiotics. Agar confines cells upon deposition; glycerol cryoprotects for -80°C storage. Printing method (stamping): A PLA 3D-printed template with synthetic fiber absorbs distinct cellular inks and stamps them onto paper in defined topologies. Stamped paper is placed onto LB-agar plates containing specified inducers and incubated at 37°C for 24 h. Templates can be customized for different circuits. Circuits can be used immediately, stored at 4°C (short-term) or frozen at -80°C (long-term). Characterization: AHL diffusion and signal reach were assayed with S1 sources and CR reporters along paper strips; CA cells were tested for extending AHL range. Device components were combined to form transistor-like circuits, logic gates (OR, AND, NOR, XNOR), a 2-to-1 multiplexer, parity-bit logic, and an analogue band-pass filter. Inputs included arabinose (10^-3 M), aTc (10^-6 M), and rhamnose (1.5%). GFP and RFP were quantified via surface scanning using a Synergy HXT reader; GFP/RFP ratios normalized for cell population. CR response to AHL was characterized across 10^-10 to 10^-4 M and fit to a Hill function. Storage stability was evaluated after refrigeration at 4°C up to 10 days and after freezing at -80°C. An alternative substrate (nylon) was also tested. Strains and conditions: E. coli Top10 and ZN1 in LB at 37°C with chloramphenicol or kanamycin; standard inducer reagents used. Detailed volumes for ink prep: modulators (100 µl culture + 50 µl LB-agar), S1/S2/CA/CR (50 µl culture + 50 µl LB-agar).

• AHL diffuses efficiently on paper; CR cells detect source-derived AHL up to ~20 mm. CA auto-amplifier cells extend the effective AHL range.
• A stamped, transistor-like circuit was implemented: S1 (source), M−_Ara (gate), CR (drain). Without arabinose, AHL flows and CR expresses GFP; with 10^-3 M arabinose, M−_Ara expresses Aiia, degrading AHL and blocking output. Average fold change ~5.6x.
• Modulatory elements responding to aTc, arabinose, and rhamnose tune CS levels in an analogue manner; combining them enables either analogue or digital behaviors.
• Introducing multiple modulators between S1 and CR reduces CS due to basal activity; placing CA near output restores signal, enabling deeper topologies.
• General multi-branch topology maps truth tables to spatial branches; scalability analysis (Karnaugh minimization) suggests most circuits require fewer than the maximal 2^N−1 branches.
• Two-input logic gates demonstrated on paper with strong ON/OFF discrimination (average fold changes): OR 14.31x, AND 6.21x, NOR 6.58x, XNOR 5.6x.
• A 3-input 2-to-1 multiplexer (S2, S3 sources; aTc-controlled modulators) performed as specified, fold change 6.76x.
• A 3-input even parity bit circuit using four branches and CA near output matched its truth table, fold change 10.47x.
• An analogue band-pass filter using S2 and M−_Ara produced a clear band-pass GFP/RFP response across arabinose concentrations.
• Circuits printed on nylon also functioned comparably to paper.
• Storage: Devices kept at 4°C up to 10 days showed progressive fold-change reduction but remained functional; frozen at -80°C preserved performance with greater durability.

By encoding computation in the spatial modulation of a single carrying signal (AHL) across a 2D substrate, the approach circumvents key limitations of traditional genetic circuit architectures, notably the wiring problem, genetic/metabolic burden, and instability from co-culturing. Implementing each function in separate spatially segregated cell types reduces unintended interactions and competition while maintaining low genetic complexity per strain. The multi-branch mapping of truth tables to spatial patterns demonstrates scalability and reusability of a small cell library (M+ and M− per input) to build diverse logic. Experimental results validate both digital (logic gates, multiplexer, parity bit) and analogue (band-pass) computations using the same biological parts but different spatial topologies. The ability to print on inexpensive substrates (paper, nylon), store devices, and achieve robust ON/OFF fold changes supports potential out-of-lab applications such as low-cost biosensing. Compared with cell-free paper systems, living printed circuits offer richer computational versatility through growth, induction, and signal amplification, albeit with biological dynamics that influence response time and stability. Overall, findings show that space-enabled, single-signal, printable multicellular devices are a practical path toward scalable, deployable biological computation.

The study introduces a printable, spatially encoded multicellular computation framework that uses a single diffusing signal to perform digital and analogue functions on 2D substrates. Using a compact library of engineered E. coli strains (sources S1–S4, modulators M−/M+, amplifier CA, reporter CR), the authors demonstrate transistor-like behavior, standard logic gates, a 3-input multiplexer, a 3-input parity bit, and a band-pass filter with robust fold changes. Devices function on paper and nylon and can be stored for later use. The general multi-branch mapping from truth tables to spatial layouts supports scalability. This platform opens avenues for low-cost, industrially producible, single-use or field-deployable living biosensors and computational devices. Future work should focus on scaling to more inputs and functions, automating high-throughput stamping/printing, and reducing response time to broaden applicability.

• Current demonstrations involve up to three inputs; scaling to higher input counts remains to be experimentally validated.
• Response time is constrained by biological growth and diffusion; authors note the need to reduce time-to-output.
• Devices stored at 4°C show progressive reduction in fold change over ~10 days, indicating limited refrigerated shelf-life; freezing preserves function but requires cold-chain.
• Basal activities of modulators attenuate CS levels as depth increases; CA amplification mitigates but adds complexity and tuning needs.
• Circuit performance depends on inducer concentrations distributed in the agar medium and on substrate properties; environmental variations may affect reproducibility.
• Diffusion-based signaling imposes spatial-distance constraints (e.g., optimal spacing, signal decay) that must be engineered per design.