Design

Bio-computing is a promising field with potential applications in bio-security, environmental monitoring, and personalized medicine^[1]. However, there has always been a significant gap between successfully constructing logic gates and actually achieving computation. Previous designs were always limited to within a single cell, constrained by available orthogonal components and cellular metabolic pressures, making it difficult to construct complex adders^[2].

To solve this problem, we develop a novel approach to constructing a biocomputer—utilizing quorum sensing and spatial diffusion to form logic gates^[3] and build a complex adder. We also establish three orthogonal quorum sensing systems to solve the crosstalk issue between different colonies. At the same time, we design a light-controlled AHL degradation enzyme based on AiiA^[5] and a light-inducible protein degradation tag, enabling the restart and re-computation of the entire system.

Our biological computer is built on quorum-sensing (QS) systems to construct logic gates. The signaling input is the quorum-sensing molecule N-acyl homoserine lactones (AHL). Receptor proteins receive inputs and become active, and drive transcription from the cognate promoter. To prevent crosstalk between different computational modules, we selected orthogonal QS systems as the logic gate signaling system; this also allows future users to assemble complex circuits from our gates without interference.

Among these, the Rhl and Las systems are orthogonal to each other, while Rhl, Tra, and Cin are mutually orthogonal in all pairwise combinations^[4] (Figure 1).

In subsequent logic-gate construction, we detect quantitative mapping of AHL distributions after diffusion on solid media and the expression strengths of the four orthogonal QS systems. Accordingly, we construct biosensor bacteria for each system (Figure 2). Taking the Rhl system as an example: a constitutive promoter (PJ23106) drives expression of the cognate transcriptional regulator RhlR; upon binding its ligand C₄-AHL, the RhlR-AHL complex activates the Prhl promoter, initiating expression of the fluorescent protein sfGFP (Figure. 2A). Fluorescence output is proportional to local AHL concentration, enabling quantification of the signal. To generate spatial concentration profiles, a defined volume of AHL solution is dispensed onto one locus of an agar plate, and a biosensor colony is inoculated at a second locus (Figure 2B). AHL diffuses radially through the medium; the resulting fluorescence intensity at the biosensor site reports the concentration at that coordinate. Employing this assay, we build AHL biosensors and determine their diffusion-response curves (Figure 2C).

The fundamental logic gates include AND, OR, and XOR gates; their corresponding truth tables are as follows:

If we use high and low expression levels as binary outputs 1 and 0, the corresponding concentration-response curves are:

The curves above can be summarized as two basic types: high-pass curve, in which cells switch ON above a threshold concentration, and bandpass curve, in which cells are ON only within an intermediate concentration. Both response behaviors can be implemented within a single bacterial strain using simple circuits. Differences in activation thresholds—e.g., both AND and OR gates exhibit high-pass behavior, but OR responds at moderate concentrations while AND does not—can be tuned simply by adjusting the distance between the signal input site and the biosensor colony.

In the high-pass strain (Figure 5A), the circuit is constructed as follows (RhlR-CinI example). A constitutive PJ23106 promoter drives expression of the transcriptional regulator RhlR. Upon binding C₄-AHL, the RhlR-AHL complex activates the Prhl promoter, inducing expression of the synthase CinI, which produces 3-OH-C₁₄:₁ AHL as the output signal. At low input levels of C₄-AHL, no 3-OH-C₁₄:₁ AHL is generated; above a threshold concentration, the output appears, yielding a high-pass response.

In the bandpass strain (Figure 5B), the circuit is constructed as follows (RhlR-TraI-bandpass example). A constitutive PJ23106 promoter drives expression of the transcriptional regulator RhlR. Upon binding C₄-AHL, the RhlR-AHL complex activates the Prhl promoter, which drives expression of T7 RNA polymerase (T7 RNAP) and the repressor PhlF; together, these components jointly modulate expression of the downstream synthase TraI and the resulting output of 3OC₈-AHL.

The bandpass behavior is produced by a T7 promoter that is repressed by the PhlF transcription factor^[3]. T7 RNAP and PhlF are expressed from Prhl promoter, respectively. In the absence of C4-AHL, no T7 RNAP is expressed, and the output promoter is not activated. At intermediate levels of C4-AHL, T7 RNAP is produced, and TraI transcription from the output promoter occurs. At high C4-AHL levels, PhlF is expressed at a high enough concentration to inhibit TraI transcription from the output promoter.

Notably, during the single-gate verification phase, both high-pass and bandpass bacteria are engineered to express green fluorescent protein rather than downstream AHL synthases (Figure 7A). After validating the high-pass and band-pass circuits, we map their diffusion profiles on agar (Figure 7B). Feeding these profiles into our computational framework yields the optimal spatial coordinates for inputs and outputs, enabling the assembly of arbitrary logic gates (Figure 7C).

b) Addition Calculators: Half-adder and Full-adder

With single-gate modules validated, we next combine them to build half- and full-adders capable of binary addition.

The half-adder is the minimal arithmetic unit, summing two 1-bit inputs without carry-in (Figure 8A). Its designation as “half” reflects that it omits any carry propagated from a preceding stage. The full-adder overcomes this limitation by accepting three 1-bit inputs—two addends plus the carry-in—thereby enabling complete multi-bit addition (Figure 8B). Together, these adders constitute the foundational blocks of the arithmetic logic unit (ALU) in digital architectures.

Inspection of the truth table reveals that the requisite adder functions can be realised by combining distinct high-pass and band-pass colonies.

• For the half-adder, the carry-out C₁ behaves as an AND gate and is implemented with a single high-pass colony, whereas the sum bit S₁ behaves as an XOR gate and is realised with a band-pass colony (Figure 8A).
• For the full-adder, the carry-out C₁ is again an AND gate and is encoded by a high-pass colony; the sum bit Sᵢ requires a combination of high-pass and band-pass colonies to yield the correct composite response (Figure 8B).

A single full-adder handles only a 1-bit summation. To process multi-bit binary numbers we chain multiple adders in series. Each stage employs a distinct orthogonal QS system so that the output AHL of one adder becomes the diffusible input of the next. By adjusting the distance between adders and the spotting timing, we achieve serial computation; the accompanying animation depicts one complete calculation (Click the plate below to see).

D A

+ C B

-----

α β γ

Input Points

A Input

B Input

C Input

D Input

Processors

Half Adder

Amplifier

Full Adder

Output Points

α Output

β Output

γ Output

To achieve repeatable computation, we need to design a degradation enzyme capable of real-time response to blue light signals and rapid degradation of quorum sensing molecules. AiiA is an enzyme that can inactivate the acylhomoserine lactone quorum-sensing signal^[5]. We split AiiA into N-terminal (n AiiA) and C-terminal (c AiiA) fragments,, with VVD domain attached to the internal end of each fragment , whereby blue light irradiation induces VVD dimerization^[6], facilitating the reassembly of the two AiiA fragments into a functional protein. (Figure 9) This design ensures that the two enzyme fragments do not spontaneously assemble into an active enzyme in the absence of blue light. However, under blue light illumination, the interaction between the LOV domains facilitates the assembly of the two fragments into a complete enzyme ^[7], thereby reconstituting its degradation activity.

The challenge in this design lies in identifying optimal truncation sites. These sites must allow for the normal fusion of the two LOV domains, ensuring the reassembled enzyme exhibits full catalytic activity under blue light. Most critically, they also prevent spontaneous assembly of the fragments in the dark. Fortunately, through comprehensive literature research, we identified the online, web-based protein design tool SPELL to address this key issue.

SPELL is an acronym for Split Protein Reassembly by Ligands or Light. SPELL server predicts potential split sites in proteins such that two halves of a split protein are attached to a couple of other proteins (that dimerize upon ligand or light stimulation) can reassemble into a full protein upon ligand or light stimulation. The server ranges potential split sites according to the degree of prediction reliability.^[8]

Using SPELL, we select three truncation sites for the AiiA degradation enzyme: between residues 154-155, 181-182, and 208-209 (Figure 10). These truncation variants are subsequently validated through molecular dynamics simulations and protein structure analysis tools, including AlphaFold and Gromacs. We also analyze the enzymatic activity of AiiA truncation variants by plotting degradation curves.

To achieve the effect of repeatable operations, we design not only the AHL degradation module but also a light-inducible degradation module for the result display protein, enabling a refreshing for the result presentation.

We choose the modular light-inducible degradation tag LOVdeg. It is based on an LOV2 domain of Avena sativa phototropin 1 (AsLOV2). The mechanism is that the C-terminal Jα helix becomes unstructured upon blue light absorption, which could be utilized to provide light-inducible protein degradation. In the dark, the C-terminal "E-A-A" motif is structurally caged within the folded Jα helix, stabilizing the domain. Upon blue light illumination, the Jα helix unfolds, exposing "E-A-A" as an unstructured degron, which targets the protein for degradation by the host proteasome.^[9]

We fuse the light-inducible degradation tag, LOVdeg, to the C-terminus of the fluorescence protein. When exposed to blue light, the LOVdeg tag induces the degradation of the fluorescent protein, refreshing the result display. (Figure 11, 12)

Furthermore, we can also fuse the LOVdeg tag with AHL synthases such as RhlI and LasI, potentially enabling comprehensive refresh operations. (Figure 12)

To enable future teams to use our work, WHU-China plans to establish a LOGIC toolkit, aiming to contribute to the advancement of synthetic biology.

We designed and experimentally verified multiple logic gate modules, enabling modular circuitassembly.

i. Modeling Method of Spatial Computation

We use Fick's law to calculate the diffusion of AHL molecules.

ii. Software for Simulation

It integrates the simulation of molecular diffusion, facilitating the design and regulation of molecular computing circuits.

iii. Hardware for Easier Colony Localization

The hardware is precisely implemented by using a stepping motor and a gear system.

Biocomputers utilize biological molecules (e.g., DNA, RNA, proteins, or living cells) instead of silicon-based materials to perform computational tasks, employing genetic circuits and biological logic gates for information processing. Their core advantages include ultra-low energy consumption (ideal for implantable medical devices), high parallelism (DNA computing processes massive data simultaneously), and biocompatibility (enabling direct integration into biological systems).

In addition to performing binary computation, LOGIC can also be applied to detect target substances in the environment. It directly displays both the types and quantities of target substances present, providing more intuitive detection results. In addition, if our project is further developed, such as combining with microfluidic technology, there is hope for its application in the diagnosis of diseases through biomolecular detection.

LOGIC can also become an ideal educational tool, where users could input two binary numbers to obtain an answer to the addition calculation, allowing children to simultaneously experience binary computing and synthetic biology, and combining education with entertainment. Furthermore, with deep integration of advanced microfluidic technology, our project can achieve further refinement in the future.

Our project achieves the construction of a complex adder by combining multiple bacteria, thereby eliminating the need for further genetic editing of the bacteria. It constructs a lightweight and flexible adjustable computing system, providing a new method for biological computing and opening up new development directions. In the future, through precise control methods such as microfluidics, we can truly realize fully automatic biological computing chips at the sub-micron level and apply them to real-world scenarios.