RATEX: A Scalable RNA‐Based Platform for Logical and Multi‐Layered Cellular Programming

ABSTRACT

Scalable genetic circuits are essential for implementing complex functions in living cells. Toward this goal, RNA regulators can provide a much‐needed parts library with added benefits of low metabolic load, design flexibility, and logic capacity. However, despite the great potential of synthetic RNA circuits, constructing such circuits with wide dynamic ranges and multiplexed regulatory cascades remains a challenge. To address this, we introduce RATEX (Ribosome‐Assisted Transcriptional EXpression controller) by integrating a translation‐to‐transcription converter with synthetic RNA regulators, enabling a compact and scalable RNA‐programmed circuit architecture. The RATEX platform repurposes a large library of well‐characterized translation regulators with up to 1,492‐fold gene regulation, while leveraging natural ribosome‐mediated sensing of diverse environmental inputs, such as metabolites. We demonstrated multi‐input logic processing with up to a 6‐input OR logic gate for RNA inputs and hybrid 3‐input logic gates to sense diverse metabolite and small‐molecule inputs alongside RNA signals. Signal amplification with multiplexed combinatorial control of RNA outputs was achieved through multiplexed signaling cascades. Finally, the RNA‐ and metabolite‐sensing 3‐input AND gates were used to control cellular morphology and intracellular spatial organization. Together, the RATEX platform, with its scalable and modular architecture, offers a broad potential design space for synthetic biology and biotechnology.

RATEX integrated ribosome‐mediated transcription control with synthetic RNA regulators, enabling simultaneous processing of RNA, metabolite, and small‐molecule inputs through complex logic operations within a single transcript. The platform achieves up to 1,492‐fold gene regulation, multi‐input hybrid logic gates, RNA‐programmed signaling cascades, and logic‐gated control of cell morphology and intracellular spatial organization.

1. Introduction

Synthetic biology enables the design and creation of customized genetic circuits to execute programmed cellular functions. Advances of synthetic biology approaches have revolutionized a broad range of applications, including biosensing, biotherapeutics, biomaterials, and bioremediation, to address critical current and emerging challenges [1, 2, 3, 4, 5, 6]. By rewiring multiple regulatory modules to interact in a predictable and pre‐programmed manner, synthetic genetic circuits can be implemented to achieve desired biological functions, such as oscillators, logic gates, and feedback controllers [7, 8]. The rapid pace of development of well‐characterized libraries of DNA‐, RNA‐, and protein‐based regulatory elements has considerably enhanced our ability to precisely control gene expression, optimize multi‐gene pathways, and program sophisticated cellular behaviors [9, 10].

Recently, de novo‐designed RNA‐based regulators have emerged as a versatile platform, enabling the efficient assembly of sophisticated synthetic biological systems through seamless integration with diverse regulatory elements in a compact encoding space [11, 12]. Notably, by taking advantage of natural and engineered sensing mechanisms, the repertoire of RNA‐based sensors covers a wide range of input stimuli, including small molecules, proteins, temperature, pH, and nucleic acids [12, 13, 14, 15, 16, 17, 18]. These riboregulators typically function immediately upon transcription and have short half‐lives, enabling faster signal propagation and rapid kinetics in multi‐layered signaling cascades [12, 19, 20, 21]. Further, these synthetic RNA elements provide highly compact genetic tools with a minimal genomic footprint and the added benefits of a low metabolic burden on the cell due to the lack of translation steps [22]. Together with advanced in silico design tools for RNA‐based circuits built on the predictable base pairing and well‐characterized sequence‐structure‐function relationships, these advantageous features position synthetic RNA regulators as a powerful platform for engineering complex cellular behaviors [23, 24, 25, 26].

Despite the great potential for scaling synthetic genetic circuits via the strategic positioning of RNA regulators, current approaches for RNA‐based multiplexed and multi‐layered circuitry often face limitations in achieving desired fine‐tuned dynamic control of circuit outputs. The challenges in precisely programmed dynamic control of RNA‐based circuits arise in part due to coupling and interference among regulatory parts, especially when colocalizing multiple parts with overlapping functional domains. Moreover, while cells in natural environments must integrate heterogeneous signals—including nucleic acids, metabolites, and small molecules—synthetic RNA circuits for sensing small molecules and for RNA signals have remained largely separated. As such, the simultaneous processing of chemically distinct input classes – e.g., RNA and metabolite signals – through complex logic operations within a single transcript remains underexplored [27, 28]. Addressing this challenge requires a novel design framework for highly modular and programmable synthetic RNA‐based circuits.

Synthetic RNA transcription regulators allow for the direct control of functional noncoding RNAs and polycistronic operons, thereby facilitating resource‐efficient RNA networks [29, 30, 31, 32, 33] and coordinated multi‐gene expression [34, 35]. Yet, synthetic RNA transcriptional regulators such as small transcription activating RNAs (STARs), which utilize natural and synthetic terminator elements, often underperform in multiplexed signaling cascades due to design constraints; these regulators require strong secondary structures, which lead to difficulties in controlling folding kinetics and avoiding interference with adjoining functional elements [12, 21, 36, 37, 38]. Similarly, conditional guide RNAs in the CRISPR/Cas system that embed a sensor domain within the effector guide RNA sequence have been utilized to sense and respond to a wide range of molecules for transcriptional control in cells [39, 40]. Despite their successful demonstration as a synthetic RNA‐based controller, conditional guide RNAs often face analogous challenges due to overlapping functional domains with sequence constraints, which in turn restrict both modularity and performance [41, 42]. On the other hand, synthetic translational RNA regulators were shown to be less constrained by RNA folding and kinetic control processes, as they operate post‐transcriptionally, allowing for design flexibility and programmability with multiplexed signal processing via the colocalization of multiple regulatory elements [43].

To take advantage of RNA‐programmed signaling cascades as well as the multiplexed input processing and design flexibility of translational controllers, we turned our attention to ribosome‐mediated transcription regulation as a potential solution to provide a translation‐to‐transcription converter in a seamless manner [44]. This natural mechanism of transcription attenuation leverages ribosome stalling to control transcription termination events, thereby offering a unique opportunity and a rich design space to control transcription in response to translational inputs [45]. Since such transcription attenuation elements typically lack fixed structural motifs or sequence constraints, their integration with synthetic translational RNA regulators, exhibiting exceptional programmability, could provide a novel framework for constructing large orthogonal transcriptional switch libraries and implementing complex signal processing circuits at the transcription level [12, 43, 46, 47]. Moreover, as ribosome‐mediated transcription regulation is intrinsically coupled to the translational state of the cell, co‐opting ribosome‐mediated sensing offers a natural interface to diverse metabolite and small‐molecule inputs via leader‐peptide‐based metabolite sensors and riboswitches [48, 49, 50, 51, 52]. This unique multi‐modal input sensing capability within a single regulatory framework could enable synthetic circuit architectures that leverage the natural systems for enhanced cellular programming. Taken together, leveraging ribosome‐mediated transcriptional regulation could enhance the design flexibility and functional diversity of RNA‐based genetic circuits, providing a unified design strategy for scalable genetic circuit architecture.

In this work, we present a novel RNA‐based design framework, the Ribosome‐Assisted Transcriptional EXpression controller (RATEX), that integrates de novo‐designed RNA translation regulators with ribosome‐mediated transcriptional regulation to construct scalable and multiplexed genetic circuits (Scheme 1). Through the systematic optimization of regulatory components sourced from a large orthogonal RNA switch library, together with natural and engineered ribosome‐stalling sequences, we established design strategies that yield high‐performance RATEX devices with up to 1,492‐fold regulation and tunable dynamic ranges. RATEX devices were then deployed to implement multi‐input logic circuits, including AND, OR, and NIMPLY gates, as well as multi‐layered RNA‐programmed cascades for signal amplification and multiplexed control of outputs, demonstrating a robust information processing capability.

SCHEME 1.

Schematic overview of the RATEX system that converts RNA‐, metabolite‐ and small molecule‐mediated translational logic into functional RNA outputs, enabling RNA‐based layered networks and logic‐gated systems for cellular reprogramming.

By interfacing RNA‐responsive translational regulators with ribosome‐mediated metabolite sensing, including L‐tryptophan, and with riboswitch‐based small molecule sensing, such as theophylline and vitamin B₁₂, RATEX enables hybrid multi‐input logic gates that simultaneously process heterogeneous input signals within a single circuit layer. Furthermore, RATEX devices could be repurposed to implement logic‐gated CRISPR interference (CRISPRi) circuits including hybrid three‐input logic gates integrating RNA and metabolite inputs for the regulation of endogenous targets, thereby facilitating cellular morphology control. The same logic‐gated RATEX device was also repurposed for phase‐separating RNA scaffold outputs to achieve programmable intracellular spatial organization, establishing the platform as a tool for engineering cellular architecture. Together, by providing a general framework for programmable transcriptional control, multi‐modal logic computation, and multilayered RNA signal propagation, the RATEX platform can unlock a novel toolkit for constructing scalable synthetic genetic circuits to program complex cellular behaviors.

2. Results and Discussion

2.1. Design and Optimization of Ribosome‐Mediated RNA Transcriptional Regulators

To implement an RNA‐based regulatory architecture that capitalizes on ribosome‐mediated transcription regulation, we adapted the translation‐to‐transcription converter (TTC) module from the Escherichia coli (E. coli) tryptophanase operon [44], where the synthesis of a leader peptide, TnaC, within the operon, is crucial for the transcription regulation of downstream genes. In this scheme, an exposed Rho‐utilization (rut) site located just downstream of the tnaC facilitates the recruitment of the Rho transcription termination factor; the interaction of the Rho factor with the RNA polymerase (RNAP) leads to the promotion of transcription termination [53, 54]. However, tryptophan‐dependent ribosome stalling during TnaC translation masks the rut site, thereby inhibiting Rho factor access and allowing transcriptional elongation (Figure 1a). Consequently, the translation of TnaC in the presence of high tryptophan concentrations allows continued transcription of downstream genes.

FIGURE 1.

Integration of ribosome‐mediated transcription controller with the toehold switch. (a) Mechanism of translation to transcription conversion in E. coli tryptophanase operon. The nascent TnaC peptide interacts with the ribosome exit tunnel to form a complex that recognizes L‐tryptophan within the tunnel and induces ribosome stalling. In the absence of tryptophan, the lack of ribosome stalling during tnaC translation leads to Rho‐dependent transcription termination. In contrast, tryptophan sensing induces ribosome stalling by TnaC, preventing Rho access to the rut site to allow transcriptional readthrough. (b) Mechanism of toehold switch. Switch RNA consists of a hairpin structure that sequesters the RBS, preventing gene expression. When a cognate input RNA (Trigger RNA) binds to the toehold switch, the hairpin structure unfolds via toehold‐mediated strand displacement, exposing the RBS and allowing translation of downstream RNA to proceed. (c) Schematic of the RATEX mechanism. RATEX integrates the two regulatory modules, TTC and toehold switch, to enable RNA‐triggered control of transcriptional elongation. In the absence of the trigger RNA, the translation of TnaC is blocked such that the Rho factor‐mediated transcription termination occurs. Conversely, in the presence of the trigger RNA, the translation of TnaC leader peptide proceeds, stalling ribosome at rut site, leading to transcription continuation. (d) Characterization of RATEX performance. (Top) Green fluorescent protein (GFP) fluorescence measurements for RATEX and an open RBS control. (Bottom) Relative RNA expression levels measured by RT‐qPCR. Cells were induced with 0.1 mM isopropyl β‐D‐1‐thiogalactopyranoside (IPTG) for 3 h 30 min. Dots represent individual data points, bars represent the mean ± s.d. of n = 3 biological replicates. (e) Flow cytometry histograms for the best‐performing RATEX. Autofluorescence was measured from induced cells containing a control plasmid without the GFP gene. Decoy represents an input RNA that does not interact with the switch RNA. TTC: translation to transcription converter; THS: toehold switch.

To repurpose this natural mechanism for synthetic RNA circuits, we employed a fusion construct where a synthetic RNA translational switch is connected to downstream genes via the TnaC leader peptide. For the precise regulation of the translation status of TnaC leader peptide, we utilized the toehold switch library, which provides wide dynamic ranges, modularity, programmability, and the capacity for scalable logical computation [43]. This synthetic RNA regulator contains the ribosome binding site (RBS) and the start codon within a strong hairpin that prevents the ribosome from initiating translation of the downstream coding sequence (Figure 1b). In the presence of a trigger RNA input, the programmed hybridization between the switch RNA and trigger RNA reconfigures the structure of the switch RNA through a toehold‐mediated strand displacement. This allows the ribosome to bind to the RBS and start codon, thereby initiating the translation of the downstream coding region. In this architecture, the presence of a trigger RNA input can control the activation of the toehold switch and translation of TnaC leader peptide, which in turn modulates the transcription of downstream genes.

Leveraging the combined benefits of ribosome‐mediated transcriptional regulation and RNA‐responsive translational regulation, we engineered a synthetic RNA device, named RATEX, to achieve precise transcriptional control via RNA inputs (Figures 1c and S1). In the absence of a trigger RNA input, the translation of the TnaC leader peptide is repressed, resulting in transcriptional termination via the Rho factor. Upon trigger RNA binding, however, a conformational change in the toehold switch allows the translation of the TnaC leader peptide, thereby inducing ribosome stalling that masks the rut site to enable continued transcription of downstream genes. Experimental characterization of RATEX with a GFP reporter showed that the integration of a toehold switch within RATEX enables precise regulation of TnaC translation and subsequent reporter gene expression in response to its cognate trigger RNA, thereby controlling transcription states using specific RNA inputs (Figure S1).

To further enhance the performance of the RATEX system, we explored the structural elements governing ribosome‐mediated transcription regulation. First, the coding region of the TnaC leader peptide was shortened to assess its impact on transcription control. The truncated TnaC peptide with only 11 amino acids (aa) exhibited markedly reduced fold changes, possibly due to the loss of critical TnaC residues essential for ribosomal arrest [55]. In contrast, a tnaC variant encoding 13 aa displayed enhanced GFP fluorescence and higher fold changes compared to the full‐length tnaC (Figure S2). Quantitative reverse transcription polymerase chain reaction (RT‐qPCR) analysis corroborated these GFP fluorescence trends, verifying that the observed differences arise at the transcriptional level (Figure S3; Table S1). Thus, we chose a 13 aa TnaC coding region for all the subsequent experiments. Next, the impact of RNAP on transcription regulation was tested (Figure S4). While the RATEX system with bacteriophage T7 RNAP exhibited at most a 2‐fold ON/OFF ratio, the E. coli RNAP‐based construct showed substantially lower levels of leaky expression, resulting in a 561‐fold change. This difference may result from the approximately 6‐fold higher elongation rate of the phage‐derived RNAP compared to that of E. coli RNAP [56], which could impair Rho‐dependent termination [57, 58, 59, 60].

The optimized RATEX device showed minimal leaky expression, comparable to autofluorescence levels, as well as strong reporter expression with cognate RNA inputs, achieving up to 1,492‐fold activation, with changes in GFP fluorescence reflecting underlying changes in GFP mRNA abundance (Figure 1d,e). Thus, the optimized RATEX device exhibited GFP output changes comparable to those of high‐performance toehold switches when tested alone [43]. RT‐qPCR analysis of GFP mRNA levels revealed that a standalone toehold switch showed negligible change in the transcript levels upon activation, whereas the RATEX device demonstrated a 70‐fold change in the output transcript levels. This transcription regulatory performance is comparable to the STAR AD1 variant [61] (Figure S5; Table S2), positioning RATEX as a high‐performance controller that effectively bridges translational inputs to robust transcriptional outputs.

2.2. Expanding the RATEX Repertoire via Modular Integration of Diverse Genetic Elements

A diverse genetic part library is a prerequisite for expanding the capacity to engineer cells with customized traits. To evaluate the design flexibility of RATEX, we first aimed to characterize its compatibility with various genetic elements, such as RNA devices, other TTC modules, a library of promoter and RBS variants.

Due to their minimal sequence constraints and well‐characterized design rules, toehold switches provide a large library of highly orthogonal variants with wide dynamic ranges [43]. Initially, we aimed to expand the orthogonal repertoire of RATEX by co‐opting the strong orthogonality of toehold switches. Specifically, we incorporated 14 different toehold switch RNA variants at the 5’ end of RATEX devices and assessed their performance via GFP expression (Figure 2a; Table S3). All 14 RATEX devices showed a functional response to their cognate RNA inputs, indicating the modularity and flexibility of the RATEX design. Among these variants, 9 RATEX devices exhibited especially low leaky expression without cognate RNA inputs, reaching signal levels comparable to autofluorescence. Furthermore, 11 RATEX devices showed strong activation exceeding 100‐fold changes, with three devices surpassing 1,000‐fold activation (Figure 2a). Another important aspect of the RATEX device library is the orthogonality of designed elements. To assess this, we systematically assessed crosstalk across all 196 pairwise switch‐trigger combinations (14 x 14) and observed that most non‐cognate pairs showed little to no crosstalk (Figures 2b and S6). A subset of 10 RATEX devices showed crosstalk levels below 10% for all pairwise interactions tested.

FIGURE 2.

Integration of diverse genetic control elements into RATEX. (a) GFP fluorescence outputs and fold changes of RATEX devices with toehold switch RNA variants. Flow cytometry measurements were taken 3 h 30 min after induction with 0.1 mM IPTG. (b) Crosstalk matrix for 196 pairwise combinations of switch and trigger RNAs. Each value indicates the percentage of non‐cognate activation, calculated as the GFP output of a trigger‐switch pair normalized to that of the cognate pair. The red box highlights the 100 pairwise switch‐trigger combinations (10 x 10) with less than 10% overall crosstalk levels. (c)‐(e) Schematics and experimental validation for (c) a toehold repressor, (d) a theophylline riboswitch, and (e) a vitamin B₁₂ riboswitch integrated into the RATEX framework. (f) Schematics of tryptophan‐independent (eTrpL, SecM) or tryptophan‐sensing (TnaC) TTC modules within the RATEX framework. (g) GFP fluorescence of RATEX devices incorporating various TTC modules, measured in LB medium. (h) Schematic of the experimental setup for cells cultured in M9 minimal medium with or without tryptophan. (i)–(k) Characterization of various TTC modules in RATEX, showing GFP fluorescence of cells harboring (i) TnaC, (j) eTrpL, or (k) SecM in M9 medium in the absence (grey background) or presence of 10 mM tryptophan (yellow background). Dark grey and green bars represent the addition of decoy and cognate trigger RNAs, respectively. Dots represent individual data points, bars represent the mean ± s.d. of n = 3 biological replicates. Detailed methods are described in the Supplementary Methods.

To evaluate the functional versatility of RATEX devices, we first implemented a NOT logic gate by incorporating a toehold repressor, which permits translation in the absence of a trigger RNA but blocks the translation of the downstream gene upon cognate RNA binding (Figures 2c and S7; Table S4). By combining this toehold repressor module with the tnaC TTC, an inverted RATEX device was created, exhibiting an 83‐fold reduction in GFP expression upon the addition of the cognate input (Figures 2c and S7). Beyond RNA‐sensing devices, we investigated whether RATEX could interface with ligand‐responsive riboswitches. First, a theophylline riboswitch [62, 63], which natively activates translation by exposing the RBS upon ligand binding, was seamlessly integrated upstream of the TTC module, yielding a robust 46‐fold activation of GFP output in the presence of theophylline (Figures 2d and S8). Similarly, a vitamin B₁₂‐responsive riboswitch, which represses translation by sequestering the RBS in the presence of vitamin B₁₂, was concatenated to form a vitamin B₁₂‐responsive RATEX device, exhibiting an 80‐fold reduction in GFP output upon treatment with its cognate ligand (Figures 2e and S9) [64].

A distinctive feature of the RATEX device is the functional coupling of ribosomal translation activity to transcriptional termination. Therefore, the ability to harness the wide repertoire of ribosome‐mediated regulation could open an untapped design space for complex synthetic circuits (Table S5). To this end, we explored an array of leader peptides known to sense tryptophan, similar to TnaC, as well as others that should be largely independent of tryptophan (Figure 2f). We first tested a trpL‐mediated TTC derived from the tryptophan operon, where tryptophan‐deficiency‐induced ribosome stalling disrupts the formation of an intrinsic terminator, leading to continued transcription [44, 48]. By replacing the tnaC TTC with the trpL‐mediated TTC, the input RNA for the upstream toehold switch could modulate the Rho‐independent transcriptional termination of the downstream reporter analogously, achieving a 23‐fold ON/OFF ratio (Figure S10). To further optimize this, we developed an engineered trpL (eTrpL) variant by substituting the tryptophan codon (UGG) with a stop codon (UGA). This modification led to a 71‐fold activation and higher GFP expression (Figure 2g), likely due to prolonged ribosome stalling at the stop codon that may enhance transcriptional readthrough [65].

Encouraged by this, we next constructed a new RATEX device by substituting the Rho‐dependent TnaC coding sequence with that of SecM. Unlike the TnaC polypeptide, the SecM arrest peptide inhibits the progression of translation in a tryptophan‐independent‐manner [66] (Figure S11). The SecM‐RATEX device showed slightly increased GFP fluorescence levels compared to other devices with a 265‐fold change upon activation (Figure 2g). In addition, we tested whether the presence of tryptophan affected the operation of eTrpL‐, SecM‐, and TnaC‐RATEX devices (Figure 2h). As expected, the TnaC‐RATEX device exhibited strong dependence on the presence of tryptophan (Figure 2i), whereas the SecM‐ and eTrpL‐RATEX were not affected by tryptophan (Figure 2j,k). The tryptophan‐independent eTrpL and SecM TTC variants showed more consistent performance across diverse media conditions (Figure S12) and throughout early to mid‐log growth phases (Figure S13).

To finely tune gene expression levels for synthetic circuits, a large library of promoters and RBSs can be tapped [67, 68]. RATEX devices can analogously be fine‐tuned by the combinatorial use of constitutive promoters and RBS variants (Figure S14a,b; Tables S6 and S7). Characterization of the engineered RATEX devices revealed that the use of a strong RBS in combination with an intermediate‐strength promoter results in large dynamic ranges (Figure S14c).

2.3. RATEX‐Based Multi‐Input Logic Circuits

Cells utilize exquisite signaling networks to process and integrate diverse signals from their environment and internal states, enabling programmed and adaptive responses for increased fitness. Synthetic analogs of such signal processing circuits including digital logic circuits within cells could be constructed by the combination of modular RNA regulators [46, 47]. Building on the diverse regulatory modules characterized above, we sought to extend the capability of RATEX devices to perform transcription‐level logic computation. Here, we aimed to construct multi‐input Boolean logic gates by combining these modules within a single RATEX architecture (Table S8).

First, to construct an AND logic gate, the RATEX device was reconfigured to contain two orthogonal switch RNAs arranged in tandem, each coupled with its own tnaC TTC (Figure S15). In this design, the input RNA for each switch RNA is required for the transcription elongation at each TTC checkpoint, such that both input RNAs are required for the synthesis of full‐length downstream mRNA. Experimental characterization of the AND‐gate RATEX demonstrated the expected high GFP signal only in the presence of both input RNAs, with low GFP outputs for all three logical FALSE cases. When both cognate input RNAs were present, we observed a 135‐fold increase in GFP fluorescence compared to the control with decoy RNAs (Figure S15).

Next, we constructed a B NIMPLY A (B AND NOT A) gate by combining a toehold switch with a toehold repressor within a single RATEX device (Figure S16). GFP expression required the presence of trigger RNA B to activate the toehold switch module and the simultaneous absence of trigger RNA A to avoid repression by the toehold repressor module, achieving an approximately 70‐ to 80‐fold ON/OFF ratio (Figure S16). This indicates that various toehold‐based riboregulators, including repressors and activators, can be seamlessly integrated within the RATEX framework to support a functionally complete set of Boolean operations.

Then, we extended these gates to 3‐input logic configurations by leveraging the tryptophan‐sensing capability of the tnaC TTC module. To achieve this, cells were cultured in M9 minimal medium to ensure stringent control over the availability of tryptophan, allowing it to serve as a regulatory input rather than a constant permissive factor. First, a 3‐input AND gate (RNA A AND RNA B AND tryptophan) was demonstrated by employing the same 2‐input AND gate with tryptophan acting as an additional regulatory input (Figure 3a). The two RNA inputs, Trigger RNA A and Trigger RNA B, and tryptophan, were treated as three independent variables to control GFP output, successfully demonstrating 3‐input AND behavior. Next, an inhibitory 3‐input AND gate (RNA B AND NOT RNA A AND tryptophan)  was analogously implemented by using the previous B NIMPLY A gate and controlling tryptophan availability (Figure 3b). Furthermore, by incorporating the theophylline and vitamin B₁₂ riboswitches alongside TnaC RATEX, we demonstrated two 3‐input hybrid circuits: a theophylline‐responsive 3‐input AND gate (theophylline AND RNA input AND tryptophan) and an inhibitory 3‐input AND gate (RNA input AND NOT vitamin B₁₂ AND tryptophan) (Figure 3c,d). These constructs were able to process RNA, metabolite, and small‐molecule inputs simultaneously within a single RNA transcript architecture, illustrating the potential for multi‐modal signal integration.

FIGURE 3.

Implementation of multi‐input hybrid logic gates using RATEX. (a) 3‐input AND gate. Schematic representation and GFP fluorescence measurement for a 3‐input AND circuit requiring RNA A, RNA B, and tryptophan for activation. (b) Inhibitory 3‐input AND gate. In this configuration, RNA A acts as a negative regulator via a toehold repressor, while RNA B and tryptophan act as activators. (c) 3‐input AND gate with a theophylline riboswitch. Design and characterization of a 3‐input gate combining a theophylline riboswitch, a toehold switch, and a tryptophan‐sensing TTC. (d) Inhibitory 3‐input AND gate with a vitamin B₁₂ riboswitch. In this circuit, a concatenated vitamin B₁₂ riboswitch serves as a negative regulator. (e) 2‐input OR gate. Schematic and GFP fluorescence levels for two switch RNA modules (A and B) arranged in tandem with a single TTC. Activation is achieved by the presence of either RNA input A or B. (f) Schematic and GFP fluorescence levels for a 6‐input OR gate that processes six orthogonal RNA signals (A‐F). The ‘ALL’ bar represents the co‐expression of all six input triggers. Dots represent individual data points, bars represent the mean ± s.d. of n = 3 biological replicates. Trp: tryptophan; Theo: theophylline; B₁₂: vitamin B₁₂. Detailed methods are described in the Supplementary Methods.

To achieve an OR logic function, we chose to utilized a concatenated toehold switch RNA design analogous to the OR‐gate ribocomputing devices [46]. For this design, mutually orthogonal switch RNAs are concatenated in‐frame within the same RNA transcript, such that the activation of any one of the switch RNAs by its cognate input RNA can result in the translation of the downstream mRNA (Figure S17). Using this architecture, a 2‐input OR RATEX device demonstrated expected OR logic behavior; the presence of either one or both input RNAs increased the reporter output compared to the basal level observed in the absence of both inputs (Figure 3e). The ON/OFF GFP fluorescence levels ranged from 15‐ to 53‐fold for the three logical TRUE cases compared to the sole logical FALSE case (Figure S17). To test the scalability of this design, we then implemented a 6‐input OR RATEX device. This design enables any single input RNA among the six cognate inputs for switch RNAs to activate the expression of the downstream reporter (Figure 3f) [46]. Characterization of 6‐input OR RATEX showed increased GFP signals for each cognate input RNA with low leaky expression in the presence of non‐cognate RNAs, as expected. The GFP fold changes for logical TRUE cases ranged from 9‐ to 79‐fold (Figure 3f, S18). Notably, the circuit with all six inputs exhibited a GFP signal less than some of the other input combinations indicating that there could be additional effects such as balancing metabolic load and input RNA stoichiometry (Figures 3f and S18; Table S9).

2.4. Harnessing RATEX for Synthetic RNA‐Based Layered Networks

RNA‐programmed regulatory circuits offer a streamlined approach to genetic network designs, reducing metabolic burden, and minimizing genetic footprints by bypassing the need for protein synthesis [11, 36, 37, 69]. Additionally, a broad repertoire of regulatory RNAs could facilitate the realization of complex regulatory functions by exploiting the sequence specificity of RNA interactions. Leveraging the capability of RATEX designs to drive RNA synthesis in a precisely programmed manner, we aimed to construct multi‐layered RNA circuits for signal amplification and multiplexed output control (Table S10).

As a basic architecture for an RNA‐mediated signaling cascade, the cognate input RNA is produced in the first stage to activate the RATEX device in the second stage, where the output from RATEX is in turn used to drive the expression of a reporter in the final stage (Figure 4a). Upon activation of RATEX by input from the first stage (Trigger A), the continued transcription of downstream RNA would result in the second stage output (Trigger B). Note that the Trigger B RNA is still connected to the toehold switch and TnaC TTC at its 5’ end. This elongated transcript may interfere with the proper functionality of Trigger B via unintended secondary structure formation, and thus, mechanisms to insulate the output RNA from the regulatory domains were explored [70, 71]. RNA processing by a cis‐cleaving ribozyme or a trans‐cleaving endoribonuclease would result in the physical separation of the RNA signal from the upstream element, enabling facile interaction with switch RNA in the final stage to drive reporter gene expression (Figure 4a). Specifically, we compared cis‐cleaving ribozymes and Csy4 endoribonuclease recognition hairpins for insulation strategies (Figure S19). The Csy4‐based RNA processing yielded a 93‐fold activation of GFP signal, a substantial improvement over the control architecture without signal cleavage (35‐fold) and the ribozyme‐mediated RNA processing design (12‐fold). Computational predictions suggest that the Csy4 recognition hairpin forms a stable secondary structure that facilitates efficient cleavage, whereas ribozyme misfolding may hinder activation [72]. Therefore, we adopted the Csy4‐based RNA processing module for all subsequent experiments due to its superior performance. This cascade produced strong GFP expression only in the presence of Trigger RNA A in its first stage together with the intermediate stage RATEX device that converts this signal to Trigger RNA B (Figure 4b), indicating that RATEX designs can be used to control regulatory RNA expression for constructing RNA‐based signaling cascades.

FIGURE 4.

Construction of multiplexed signaling cascade circuits with RATEX. (a) Schematic of a three‐stage cascade. Three‐stage RNA cascade: trigger RNA A activates the intermediate RATEX module to produce trigger RNA B, which is cleaved and activates the final‐layer toehold switch driving GFP expression. Bracketed Tr region denotes tandem trigger copies for (c). (b) GFP fluorescence outputs for the RATEX‐driven RNA cascade for different combinations of intermediate layer and inputs. (c) Signal amplification by multiple trigger RNA B copies in the cascade. The increased number of tandem trigger RNA B copies led to enhanced GFP fluorescence. (d) Multiplexed RNA cascades via RATEX. The GFP and mCherry outputs under the control of orthogonal switch RNAs were simultaneously regulated for all combinations of inputs and intermediate layers. Insets in (b) and (c) show the GFP fluorescence data on a logarithmic scale. Blue circle in the top layer denotes trigger RNA A, while gold and orange circles in the intermediate layer denote the trigger RNA B and C, respectively. Green and magenta circles in the final layer indicate GFP and mCherry outputs. Gray nodes represent inactive signaling layers where decoy RNA inputs or inactive switch RNAs were used. (e), (f) Integration of Boolean logic into signaling cascades. (e) An AND logic gate within the signaling cascade. The intermediate layer processes two inputs to generate tandem trigger B copies (x2) only upon dual activation. (f) An OR logic gate cascade. The circuit design allows signal propagation and subsequent GFP expression when either or both inputs are present. Cells were induced with 1 mM IPTG and 12.5 ng/mL aTc for 5 h. Dots represent individual data points, bars represent the mean ± s.d. of n = 3 biological replicates. Tr: Trigger RNA.

An important consideration for signaling cascades is whether the signal could be maintained across multiple cascade layers, for which a signal amplification strategy may be required to enhance circuit robustness. Unlike transcription factor‐based regulatory cascades, where the translation steps for intermediate protein regulators provide a certain form of signal amplification, RNA‐based synthetic cascades lack this amplification stage due to the bypassing of protein expression steps. Therefore, an alternative strategy for signal amplification is needed to boost signal strength and ensure robust signal propagation. Taking inspiration from the RNA arrays of the CRISPR system and previous approaches [69, 73], we constructed RATEX devices encoding multiple tandem copies of regulatory RNA signals (Figure S20). This modified design resulted in increased GFP fluorescence with an increased number of regulatory RNA outputs, with 4.3‐ and 8.9‐fold increases for two and three RNA outputs over the single RNA output, corresponding to GFP ON/OFF ratios of up to 380‐fold activation (Figure 4c).

To further explore the capacity of the RATEX platform to coordinate multiple regulatory pathways, we designed RATEX‐based cascades encoding orthogonal regulatory RNAs that target independent switch RNAs controlling GFP and mCherry reporters (Figure S21). In this design, the first stage input RNA (Trigger A), which drives the expression of second stage RATEX‐mediated outputs (Trigger B and C), ultimately produces the GFP reporter downstream of switch B and the mCherry reporter downstream of switch C. Importantly, while multiple signal RNAs are produced and processed for interaction with different stages of regulators simultaneously, the signaling cascades correctly processed the multiplexed regulation of downstream GFP and mCherry reporters for all combinations of inputs and cascade components tested (Figure 4d). To test the capacity for multiplexed signal integration within layered architectures, we implemented 2‐input AND and OR circuits where RATEX serves as the logical signal processor within a signaling cascade. The 2‐input AND cascade yielded a 30‐fold induction only when both inputs were present (Figure 4e), while the 2‐input OR cascade achieved 10‐ to 40‐fold activation with at least one input present (Figure 4f), confirming logical signal integration across cascade layers.

2.5. Logic‐Gated Control of Cell Morphology and Subcellular Compartmentalization via RATEX

Building on the successful application of RATEX for sRNA synthesis and RNA regulatory cascades, we next aimed to interface RATEX with diverse RNA effectors to program higher‐order cellular phenotypes: the modulation of cell morphology via CRISPRi‐mediated gene repression [36, 74, 75] and the de novo construction of subcellular compartments via RNA‐driven phase separation [76].

We first aimed to regulate cell morphology by controlling the biogenesis of single guide RNAs (sgRNAs) targeting the endogenous ftsZ gene via CRISPRi (Figure 5a). Repression of ftsZ, which is essential for bacterial cell division, induces a filamentous phenotype characterized by cell elongation [77]. As a first step, a gfp‐targeting sgRNA was designed and integrated into a RATEX device to regulate expression levels of a gfp‐coding gene on a plasmid, demonstrating that RATEX‐mediated sgRNA expression could effectively modulate target gene transcription (Figure S22; Table S11). Next, activation of the RATEX with ftsZ‐targeting sgRNA output via cognate input RNA resulted in elongated cell morphology as confirmed by fluorescence microscopy and flow cytometry (Figure S23; Tables S12 and S13). Integrating ftsZ‐targeting sgRNA output with previously demonstrated 2‐input AND and OR RATEX logic gates showed the expected logical regulation of cellular morphology as confirmed by fluorescence microscopy and flow cytometry (Figures S24 and S25).

FIGURE 5.

Logic‐gated control of high‐order cellular phenotypes via RATEX. (a), (b) Design of a hybrid 3‐input AND gate for morphological control. (a) Schematic of endogenous ftsZ regulation. RATEX‐gated biogenesis of ftsZ‐targeting sgRNA enables CRISPRi‐mediated repression, inducing a filamentous phenotype. (b) Schematic of the 3‐input hybrid AND gate RATEX device. (c), (d) Functional characterization of morphology control. (c) Representative fluorescence microscopy images (top) and corresponding flow cytometry FSC‐A histograms (bottom) across all eight possible input combinations. Scale bars = 20 µm. (d) Quantification of morphological changes based on FSC‐A measurements. (e), (f) Design of a hybrid 3‐input AND gate for RNA‐condensate formation in cells. (e) Schematic of the RNA output module for RNA‐condensate formation. The transcript consists of 47x rCAG repeats for liquid‐liquid phase separation (LLPS) and 12x MS2 aptamers for visualization via tdMCP::GFP recruitment. (f) Schematic of the 3‐input hybrid AND‐gate configured to control the LLPS‐RNA output. (g), (h) Logic‐gated control of subcellular compartments. (g) Representative microscopy images (top) and corresponding demography profiles (bottom) across all eight possible input combinations. Scale bars = 1 µm. (h) Quantitative analysis of condensate formation. Cells were grown in M9 minimal medium in the presence or absence of 10 mM L‐tryptophan. For morphology control (c), (d), cells were induced with 1 mM IPTG and 10 ng/mL aTc and collected 4 h 30 min post‐induction. For spatial programming (g), (h), cells were induced with 0.125 mM IPTG and 5 ng/mL aTc and collected 3 h 30 min post‐induction. Statistical significance in d) and h) was determined by Welch's t‐test (**p<0.01); Dots represent individual data points, bars represent the mean ± s.d. of n = 3 biological replicates.

To capitalize on the ability of TnaC TTC as a natural L‐tryptophan sensor, we sought to test the higher‐order sensing capability of RATEX for cellular morphology control via a 3‐input hybrid AND gate (Figures 5b and S26). In this configuration, the production of the ftsZ‐targeting sgRNA is gated by the simultaneous presence of two RNA triggers (Trigger A and B) and L‐tryptophan as previously illustrated in Figure 3a. As before, cells were cultured in M9 minimal medium, allowing L‐tryptophan to function as a control knob (Figure 5b). Fluorescence microscopy and flow cytometry analysis revealed that the elongated cell phenotype was exclusively observed under the logical TRUE (1,1,1) condition (Figure 5c) with a 7‐fold increase in forward scatter (FSC‐A) values only when all three inputs were present (Figure 5d). These results confirm the high‐fidelity integration of heterogeneous molecular and metabolic signals via RATEX for orchestrating morphological control.

To demonstrate that the RATEX device can be used for regulation of architectural change within the intracellular environment, we repurposed the device to gate the formation of RNA‐based membraneless organelles (Figure 5e; Table S14). We utilized an RNA output consisting of tandem rCAG repeats and MS2 aptamers to induce liquid‐liquid phase separation (LLPS), which could be easily confirmed by the recruitment of the tandem‐dimeric MS2 coat protein (tdMCP)::GFP fusion proteins [76, 78, 79]. Building on preliminary results with a single‐input control (Figure S27) and quantification methods for condensate formation (Figure S28 and Supplementary Methods), we scaled up this spatial programming application with the 3‐input hybrid AND logic device (Figure 5f). As before, this circuit requires the concurrent presence of Trigger A, Trigger B, and L‐tryptophan to activate the downstream expression of the LLPS‐inducing RNA. Quantitative imaging analysis via demography profiles revealed logical control over subcellular compartmentalization, where distinct intracellular foci were predominantly observed only under the logical TRUE condition (Figure 5g). Notably, the proportion of condensate‐forming cells reached 53.5% in the presence of all three inputs, about 30‐fold higher than the other logical FALSE conditions (Figure 5h).

The identical logical RATEX device could be used for ftsZ‐mediated morphological changes as well as LLPS‐driven spatial organization, highlighting the modularity and versatility of the RATEX platform. Collectively, these results suggest that RATEX can potentially function as a universal logic processor to orchestrate diverse and complex cellular phenotypes, thereby providing a novel pathway for cellular engineering and reprogramming.

3. Conclusion

Here, we present RATEX, a compact and scalable platform for synthetic RNA‐based circuit design by coupling programmable translational RNA regulators to ribosome‐mediated transcriptional control, achieving up to 1,492‐fold activation of target gene expression in living cells. The RATEX platform can be flexibly reconfigured in a plug‐and‐play manner with diverse promoters, RNA switches, and RBS variants, while simultaneously allowing multiplexed control of RNA and protein output signals. A notable feature of RATEX design is the ability to process multiple RNA inputs as well as metabolites and small molecules to provide hybrid multi‐input processors capable of complex logical operations such as a 6‐input OR gate for RNA inputs, and a 3‐input AND gate for RNA with small molecule inputs. Engineering programmable RNA‐based signaling cascades with an RNA signal amplifier enabled multiplexed RNA circuits including 2‐input AND and OR cascade architectures, underscoring the scalability of RATEX for multi‐layered circuit design. Finally, integration with the CRISPR/Cas system demonstrated logic‐gated control of endogenous gene regulation via RATEX, with clear phenotypic consequences for a hybrid 3‐input AND gate that processes RNA and metabolite inputs simultaneously. Moreover, by gating the production of RNA condensate scaffolds for membraneless organelle assembly, we demonstrated that RATEX could orchestrate intracellular spatial programming [78].

The RATEX design presented here can offer advantageous features that compare favorably to prior transcriptional regulators [36, 37, 61]. First, RATEX, while used as a synthetic transcriptional regulator, can leverage de  novo designed translational regulators with large dynamic ranges and orthogonal libraries [43, 46, 47] as demonstrated by a maximal fold change greater than 1,000 and an orthogonal library of 10 x 10 with less than 10% crosstalk. Second, by utilizing concatenated translation regulators, much like previous ribocomputing strategies [46, 47], the RATEX enables complex intracellular signal computations to be carried out in a single circuit layer as demonstrated by multi‐input logical evaluations including a 6‐input OR gate. Third, the RATEX platform provides advantageous features of RNA‐programmed devices, in which both inputs and outputs are RNA species. This feature enables a compact architecture and flexible control of diverse output targets, including RNA‐based regulators, CRISPR/Cas systems, and RNA‐binding proteins [29, 43, 80, 81], as well as signal multiplexing and layered signaling cascades via RNA signal processing and amplification [77, 82, 83]. Finally, the incorporation of ribosome‐mediated transcription regulation allows RATEX to interface with a wide spectrum of natural and synthetic ribosomal sensing mechanisms, enabling spatiotemporal regulation and environmental signaling [84, 85, 86, 87, 88]. Taken together, these properties position RATEX as a scalable and programmable RNA‐based regulatory architecture that leverages ribosome‐mediated transcription regulation to enable context‐sensitive genetic circuit designs.

The versatility and modularity of RATEX platform can be further underscored by the design flexibility and consistent performance across different sensing modalities. Beyond TnaC‐based tryptophan sensing, the TTC architecture can accommodate a diverse array of metabolite‐responsive leader peptides; by simply interchanging the leader peptide module without further engineering, we successfully implemented RATEX devices responsive to L‐ornithine (SpeFL) [89], L‐threonine/isoleucine (ThrL) [90, 91], and L‐phenylalanine (PheL/PheM) [92, 93] (Figure S29, Table S15). Further, RATEX can be used to process full‐length protein‐coding mRNAs as inputs, such as mCherry mRNA transcripts (Figure S30, Table S16), potentially extending its application scope to endogenous transcript sensing [94, 95]. Additionally, other dynamic regulatory circuits, including a NOR gate (Figure S31, Table S17), a negative feedback loop (Figure S32, Table S17), and a multi‐level controller with reduced transcriptional leakage (Figure S33, Table S17) were implemented, highlighting the platform's capacity for constructing RNA circuits with programmed dynamic regulatory features [96, 97].

Despite the advantages, the RATEX devices present areas that would benefit from further optimization. First, while RATEX demonstrated robust portability across laboratory strains E. coli BL21 (DE3) and MG1655, and probiotic strain E. coli Nissle 1917 (Figure S34, Table S18), a reduced performance was observed for RATEX devices in Pseudomonas putida KT2440 (Figure S34). This divergence likely stems from host‐specific variations in Rho‐dependent termination pathways [98] and ribosome‐mediated arrest peptide recognition [49]. Thus, consideration of these endogenous regulatory components would likely be required to ensure consistent performance when porting RATEX to other microbial chassis. Second, the multi‐input logic gates that concatenate multiple sensing elements can exhibit context‐dependence, in line with previous reports [46]. To address the potential RNA structure interference or incomplete integration of transcription–translation coupling, incorporating standard sensing and coupling modules across circuit designs together with high‐throughput characterization and machine learning may enhance the overall robustness and predictability of RNA‐based computational devices [99, 100]. Third, while RATEX could be used to detect full‐length mCherry mRNA transcripts, achieving robust endogenous mRNA sensing would require additional optimization to overcome the challenges, including low intracellular target concentrations and ribosome‐mediated displacement of bound sensors [99, 100]. Combining RATEX with signal amplification strategies or engineered high‐affinity binding modules may provide a path toward this goal.

In conclusion, by seamlessly integrating the highly programmable de  novo‐designed riboregulators with versatile ribosome sensing mechanisms to allow multiplexed and multi‐layered RNA cascades, CRISPR/Cas regulation, and programmable spatial organization via RNA condensates, the RATEX system is poised to provide a step forward for next‐generation synthetic biology applications across microbial platforms.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

The authors have cited additional references within the Supporting Information [101, 102, 103, 104, 105].

Data Availability Statement

The data that support the findings of this study are available in the Supporting Information of this article.