Modular Scalable Synthetic Gene Circuits for Complex Functions Within Minimal Computational Layers in Human Cells Authors: Keren Roas, Ilanit Kovalski, Odelia Mouhadeb, Tamar Aminov, Hadas Weinstein-Marom, Lior Nissim Published: 2026 ABSTRACT Engineering mammalian cells to execute complex genetic programs remains a significant challenge in synthetic biology. Synthetic gene circuits typically implement sophisticated programs through cascaded computational layers. However, these architectures require numerous orthogonal parts, increase genetic payload, and deplete cellular resources, thereby limiting functionality and scalability. ‏ ‏We present a modular design framework for engineering scalable gene circuits that execute complex functions within fewer computational layers. The platform integrates orthogonal trans-splicing-based AND gates, native-synthetic hybrid promoters for tunable regulation, and synthetic microRNAs that implement inhibitory logic. Using this approach, we engineer complex circuits, including a three-input combinatorial logic gate, a half adder, a full adder, and a dynamic 3-to-1 multiplexer with a dedicated Selector Overload Status output, generated only when both selector inputs are activated. By minimizing the number of computational layers while maintaining functionality, this strategy addresses scalability barriers in gene circuit engineering and expands applicability for biomedicine, biotechnology, and fundamental biology. Introduction Synthetic biology is focused on engineering artificial genetic tools with novel functions. Synthetic gene circuits that integrate biomolecular inputs using Boolean logic can be layered to execute increasingly complex functions, analogous to electronic devices such as full adders and multiplexers. These genetic devices facilitate the rational programming of cellular activities, applicable to fundamental research, biotechnology, and biomedicine. This strategy enables precise therapeutic interventions, including targeted cancer therapies, immunotherapies, gene therapies, and the management of metabolic and immunological disorders1–6. Despite these advances, engineering complex synthetic gene circuits in mammalian cells is often hindered by their dependency on numerous orthogonal genetic components and multistage computational cascades. For example, mammalian full adders are restricted to a single irreversible recombinase-based computation or necessitate cascaded communication between multiple reprogrammed cell populations7–9. Therefore, a critical objective in synthetic biology is to develop orthogonal genetic components and strategies to execute complex programs within minimal computational layers. RNA-based designs enable the straightforward generation of orthogonal synthetic genetic elements through sequence modifications10–24, such as CRISPR/Cas-based orthogonal synthetic transcription factors12,16. Additionally, RNA-based architectures enable sophisticated wiring of circuit components that minimizes the number of computational layers10–14. RNA-based circuits were previously constructed using mechanisms such as trans-splicing riboregulators in E. coli15, alternative splicing17, microRNAs11,14,18,19, and CRISPR-Cas systems12,13,20–24. Therefore, leveraging additional RNA-based mechanisms can expand the functionality of synthetic gene circuits. RNA trans-splicing is a naturally occurring process in which exons from distinct pre-RNA molecules are joined into a single chimeric transcript (Fig. 1A)25,26. Trans-splicing has been applied for biomedical applications, for example, by replacing a mutated exon with a wild-type sequence27, or targeting disease-specific mRNA28. Yet, trans-splicing has not been implemented for engineering mammalian synthetic gene circuits.Fig. 1Design, optimization, and modularity of trans-splicing-based AND gates.A Schematics of cis- and trans-splicing. In cis-splicing (top), a precursor RNA (pre-mRNA) encodes Exon 1, a 5′ splice-donor (5′D), an intron containing 3′ splicing elements (3′SE), and Exon 2. The spliceosome excises the intron and ligates Exon 1 to Exon 2. In trans-splicing (bottom), two separate pre-mRNAs are joined. Pre-mRNA 1 encodes Exon 1, 5’D, and an intronic binding domain in a sense orientation (BD-S). Pre-mRNA 2 encodes an intronic complementary anti-sense sequence of BD-S (BD-AS), 3’SE, and Exon 3. BD-S and BD-AS hybridize, enabling the spliceosome to excise the introns and ligate Exon 1 to Exon 3, thereby producing a mature mRNA. B Trans-splicing-based AND gate design. The circuit produces mKate2 only when both input promoters (Input 1 and Input 2) are simultaneously active (mK2-Ex1 and mK2-Ex2 denote exon 1 and exon 2 of mKate2, respectively). C AFP-based trans-splicing design. Input states are defined as 0 (module absent) or 1 (module present) (n = 3 biological replicates; p = 0.000029). D BD6-based trans-splicing design. Input states are defined as 1 (input promoter is CMVp) or 0 (input promoter is a transcriptionally inactive sequence) (n = 3 biological replicates; p = 0.000017). E Output mRNA levels of the BD6-based AND gate, quantified by junction-specific qPCR (n = 3 biological replicates; p = 0.000197). F BD6-based AND gate generating an ECFP output (n = 3 biological replicates; p = 0.000112). G Schematics of a universal multi-output AND gate design. The circuit output is a synthetic transcription factor (GAD). GAD activates the synthetic promoter GAL4p, which regulates the output gene. Auxiliary vectors encoding GAL4p-regulated outputs enable flexible multi-output expression. H Output levels generated by the multi-output AND gate (n = 3 biological replicates; p = 0.000086). I IL-15 secretion generated by the multi-output AND gate (n = 3 biological replicates; p = 0.0457). J Orthogonality assessment of six distinct BD pairs generating mKate2 output. Data presented as mean ± SEM; n = 3 biological replicates. Data were statistically analyzed by one-way ANOVA with Tukey’s multiple comparisons test; ∗ indicates p < 0.05, ∗∗∗ indicates p < 0.001 (95% confidence interval). Source data are provided as a Source Data file. Panels A, B, and G were created in BioRender.com Nissim, L. (2026) https://BioRender.com/5gjbb8a. Design, optimization, and modularity of trans-splicing-based AND gates. A Schematics of cis- and trans-splicing. In cis-splicing (top), a precursor RNA (pre-mRNA) encodes Exon 1, a 5′ splice-donor (5′D), an intron containing 3′ splicing elements (3′SE), and Exon 2. The spliceosome excises the intron and ligates Exon 1 to Exon 2. In trans-splicing (bottom), two separate pre-mRNAs are joined. Pre-mRNA 1 encodes Exon 1, 5’D, and an intronic binding domain in a sense orientation (BD-S). Pre-mRNA 2 encodes an intronic complementary anti-sense sequence of BD-S (BD-AS), 3’SE, and Exon 3. BD-S and BD-AS hybridize, enabling the spliceosome to excise the introns and ligate Exon 1 to Exon 3, thereby producing a mature mRNA. B Trans-splicing-based AND gate design. The circuit produces mKate2 only when both input promoters (Input 1 and Input 2) are simultaneously active (mK2-Ex1 and mK2-Ex2 denote exon 1 and exon 2 of mKate2, respectively). C AFP-based trans-splicing design. Input states are defined as 0 (module absent) or 1 (module present) (n = 3 biological replicates; p = 0.000029). D BD6-based trans-splicing design. Input states are defined as 1 (input promoter is CMVp) or 0 (input promoter is a transcriptionally inactive sequence) (n = 3 biological replicates; p = 0.000017). E Output mRNA levels of the BD6-based AND gate, quantified by junction-specific qPCR (n = 3 biological replicates; p = 0.000197). F BD6-based AND gate generating an ECFP output (n = 3 biological replicates; p = 0.000112). G Schematics of a universal multi-output AND gate design. The circuit output is a synthetic transcription factor (GAD). GAD activates the synthetic promoter GAL4p, which regulates the output gene. Auxiliary vectors encoding GAL4p-regulated outputs enable flexible multi-output expression. H Output levels generated by the multi-output AND gate (n = 3 biological replicates; p = 0.000086). I IL-15 secretion generated by the multi-output AND gate (n = 3 biological replicates; p = 0.0457). J Orthogonality assessment of six distinct BD pairs generating mKate2 output. Data presented as mean ± SEM; n = 3 biological replicates. Data were statistically analyzed by one-way ANOVA with Tukey’s multiple comparisons test; ∗ indicates p < 0.05, ∗∗∗ indicates p < 0.001 (95% confidence interval). Source data are provided as a Source Data file. Panels A, B, and G were created in BioRender.com Nissim, L. (2026) https://BioRender.com/5gjbb8a. Here, we present a toolkit to generate orthogonal trans-splicing-based AND gates, tunable hybrid promoters comprising native and synthetic enhancers, and synthetic microRNAs for inhibitory logic. We first systematically optimized AND gates that produce an output gene only when two input promoters are simultaneously active. Scalability and modularity were demonstrated by multiple orthogonal AND gates, subsequently adapted to concurrently generate multiple outputs. Synthetic miRNAs were incorporated to enable additional functions, including the integration of three input promoters (Input 1, Input 2, and Input 3) by a [(I1 AND I2) AND (NOT I3)] logic, producing an output only when both I1 and I2 are active, and I3 is inactive. Finally, we constructed circuits that execute functions analogous to electronic devices, including a half adder, a full adder, and the first dynamic mammalian 3-to-1 multiplexer incorporating a Selector Overload Status (SOS) output, which is activated when both selector inputs are active. In summary, this work removes limitations in mammalian gene circuit engineering by expanding the genetic parts repertoire and enabling the construction of synthetic gene circuits that execute complex functions within fewer computational layers. Such functions are essential for detecting disease onset or treating medical conditions with combinatorial or heterogeneous characteristics, such as cancer, metabolic disorders, and autoimmunity. Therefore, this platform establishes practical tools with broad applicability in fundamental and applied research12,26–34. Results Modular trans-splicing-based AND gate design As a core circuit component, we constructed and optimized an RNA-based AND gate, which produces an output protein only upon trans-splicing of two distinct pre-mRNAs, each regulated by a separate input promoter. The initial design comprised two modules that produce the red fluorescent protein mKate2 as output. Specifically, Module 1 consists of the SSX1 promoter (SSX1p)35 (activity levels shown in Fig. S1), regulating a pre-mRNA that encodes the first exon of mKate2 (mK2-Ex1), followed by a 5’ Donor sequence (5’D) that defines the intron start, and a 50 bp intronic binding domain in a sense orientation (BD-S), that mediates hybridization between pre-mRNAs required for efficient trans-splicing. Module 2 consists of the H2A1 promoter (H2A1p)35 (activity levels shown in Fig. S1), regulating a pre-mRNA that encodes the anti-sense sequence of BD-S (BD-AS), followed by 3’ splicing elements (3’SE) comprising a branch point, a polypyrimidine tract, and a splice acceptor that promote spliceosome recruitment and define the intron end, and the second exon of mKate2 (mK2-Ex2). Consequently, when both pre-mRNAs are expressed, the complementary intronic BD-S and BD-AS hybridize and facilitate spliceosome recruitment. The spliceosome then identifies the 5’D in the first pre-mRNA and the 3’SE in the second, ligates the two pre-mRNAs, and removes the intronic sequences, thereby producing a functional mKate2 mRNA. Trans-splicing efficiency is inherently lower than that of cis-splicing. Since efficient trans-splicing and orthogonal BDs are essential to establishing scalable AND gate designs, we systematically optimised circuit components, including the 3’SE, BDs, and linker sequences, aimed to minimise cross-interference between circuit elements. As a starting point, we utilized splicing elements from a trans-splicing system developed to target tumor cells expressing the hepatocellular carcinoma-related alpha-fetoprotein (AFP) gene by specifically replacing the AFP exon with an exon encoding a killer gene28. To evaluate the trans-splicing efficiency of these AFP-based elements (i.e., 5’D, AFP BD-S, AFP BD-AS, and AFP-3’SE), we transfected HEK293T cells with either both modules together, only Module 1, only Module 2, or none, and measured the mKate2 output levels using flow cytometry (Fig. 1C, Fig. S2). The AFP-based configuration provided robust trans-splicing efficiency, with mKate2 produced by both modules together approximately 72- to 83-fold higher than that generated by either a single or no modules. Deletion of the 5’D, 3’SE, or both resulted in an approximately 176- to 231-fold decrease in output levels (Fig. S3B), whereas removal of BD-S, BD-AS, or both led to an approximately 18- to 28-fold reduction (Fig. S3C). Together, these data highlight the importance of splicing motifs and BD sequences to circuit performance. Therefore, we focused on these elements for subsequent circuit optimization. Optimization of splicing sequences and BDs In the first optimization step, we compared the trans-splicing efficiencies of the AFP-based 3’SE to three different canonical splicing motifs (Supplementary Data 1 for 3’SE sequences)12,36,37. In parallel, we examined the effect of modifications to the original AFP-BD (Fig. S4), which comprises 50 bp with approximately 30% GC content, includes an eight-adenine stretch in BD-S, and contains a two-nucleotide mismatch between BD-S and BD-AS at positions 18–19 (Supplementary Data 2 for BD sequences). To this end, we transfected HEK293T cells with modified versions of Modules 1 and 2 and measured mKate2 fluorescence using flow cytometry. The canonical and AFP-based 3’SE provided comparable trans-splicing efficiencies. Therefore, 3’SE v1 was utilized in all our subsequent designs. Doubling the AFP-based BD-S (AFP-S x2) did not improve trans-splicing efficiency (Supplementary Data 2 for BD sequences). Trans-splicing efficiency of BDs comprising only the first 30 bp from AFP-BD and its exact antisense (AFP 1-30) was comparable to that of the full-length AFP-BD, whereas BDs comprising only the last 30 bp of the AFP-BD and its exact antisense (AFP 21-50) yielded approximately 31% of this efficiency. Restricting the overlap of BD-S and BD-AS to only 10 nucleotides, including BD pairs AFP(1-30)S + AFP(21-50)AS and AFP(21-50)S + AFP(1-30)AS, significantly reduced trans-splicing efficiency, indicating the inferior performance of short BDs. Finally, BDs comprising random sequences with varying lengths and GC content (BD1-BD4). BD1 (50 bp, 50% GC), BD3 (30 bp, 50% GC), and BD4 (30 bp, 30% GC) mediated approximately 44%, 30%, and 19% trans-splicing efficiency, respectively, whereas BD2 (50 bp, 30% GC) mediated only 2% efficiency (Supplementary Data 3 for BD sequences). Therefore, BD optimization is critical for enhanced circuit performance. We further examined additional BD modifications, including removal or alteration of the two mismatches originally introduced in the AFP-BD to prevent editing triggered by long double-stranded RNAs, GC content adjustments, and randomised BD sequences (Figure S5; see Supplementary Data 4 for BD sequences). Mismatch alterations did not significantly affect trans-splicing efficiency compared to AFP-BD. AFP-BD variation in which the GC content was increased to 50% (BD5) and a randomized BD with 30% GC content (BD6) demonstrated approximately two-fold higher efficiency. Finally, we tested whether AFP-BD, BD5, and BD6 are orthogonal by measuring the outputs generated from all pairwise BD combinations (Fig. S6A). Consistent with orthogonal parts, complementary BD pairs produced significantly higher output than non-complementary counterparts, highlighting the scalability of this design strategy. Collectively, these results suggest basic BD design guidelines (Fig. S6B): (i) 30–50 bp with around thirty contiguous complementary nucleotides to ensure stable hybridization; (ii) approximately 30-50% GC content, evenly distributed to prevent secondary structures that could hinder splicing; (iii) exclude extended homopolymeric runs; and (iv) minimize internal mismatches in short BDs ( ≤ 30 bp), although two mismatches are tolerable. Nevertheless, additional considerations must be taken into account. 50 bp dsRNA can activate antiviral innate immunity, although innate immune sensing depends on multiple factors, including RNA length, sequence, secondary structures, subcellular localization, post-transcriptional processing, and modifications38–40. Therefore, it is recommended to limit BD length to 30 bp, and introduce mismatches into longer BDs to disrupt dsRNA continuity and thereby hinder recognition by cellular innate immunity sensors38–40. Furthermore, since long intronic dsRNA could be processed by the miRNA production machinery, it is critical to exclude basal UG, Drosha, and Dicer processing motifs14,41. Finally, it is essential to avoid known regulatory sequences, such as binding sites for native miRNAs. Optimization of intronic sequences and linkers As a last optimization step, we tested modifications to intronic sequences and linkers (Figure S7B). Module 1 contained an optimized 50 bp synthetic intronic domain (ID) known to enhance splicing efficiency, located between the 5’D and the BD-S. An additional 34-bp random linker (Linker L1) was added downstream of BD-S to prevent interference of the poly-A tail with BD binding and splicing. Module 2 contained a 40 bp linker upstream of BD-AS (Linker L2), a 34 bp linker downstream of the BD-AS (Linker L3), and a 56 bp linker upstream of the poly-A tail (Linker L4). Since linkers within exons are most likely to affect trans-splicing, we focused on optimizing L1-L3. Removing the ID sequence in the AFP-based design impaired trans-splicing, whereas concurrently removing Linkers L1, L2, and L3 provided a marked increase in efficiency (Fig. S7B). Similar results were obtained for BD6 and BD5, further supporting these findings (Fig. S7C). Therefore, the configuration lacking Linkers 1-3 was used in all subsequent designs. Modular trans-splicing-based AND gate design We next evaluated the optimized circuit in executing an AND gate function (Fig. 1D–H, Fig. S2). To this end, we measured the circuit output across all four possible input states: State [0,0], in which both modules are inactive; State [0,1], in which only Module 2 is active; State [1,0], in which only Module 1 is active; and State [1,1], in which both modules are simultaneously active. Unless otherwise indicated, active modules in all experiments were regulated by an active promoter, whereas inactive modules were regulated by a non-specific DNA sequence. This sequence was selected for its residual activity, which simulates the leaky basal activity typical of native mammalian promoters that must be taken into account in circuit design (basal activity levels, enhanced in the presence of intronic miRNA, are shown in Fig. S1). The mKate2 output in State [1,1] was at least 45-fold higher compared to all other states, demonstrating a robust AND gate function (Fig. 1D, Fig. S2). Representative fluorescence microscopy images of the AND gate were acquired to qualitatively validate the spatial distribution of output expression and cellular morphology across all input states, demonstrating high transfection efficiency with minimal cytotoxicity (Fig. S7D). To quantify output mRNA levels, junction-specific qPCR was performed across all input states of the BD6-based AND gate. As expected, significant levels of trans-spliced mKate2 mRNA were detected exclusively in the [1,1] state (Fig. 1E). To provide direct molecular evidence for the formation of appropriate mKate2 mRNA, we reverse-transcribed from the trans-spliced product generated by this AND gate and sequenced the resulting cDNA (Fig. S7E). Sanger sequencing results validated the formation of the expected mKate2 sequence, thereby confirming the fidelity of the trans-splicing process. Similar results were obtained in a design adapted to produce an ECFP output by replacing the mKate2 exons with ECFP exons (Fig. 1F), highlighting the modularity of this design strategy for output substitution. However, such designs are limited to a single output and could require separate re-optimization for each distinct output. Nevertheless, trans-splicing efficiency can be affected by multiple factors in the output pre-mRNA, such as RNA length and secondary structures. Consequently, comparable quantities of pre-mRNAs encoding distinct output genes may produce different protein concentrations. Therefore, further optimization of splicing motifs, linkers, and BDs may be required to yield defined output concentrations. A universal design in which output expression is regulated downstream of the trans-splicing reaction would enable flexible output substitution without requiring circuit re-optimization. Furthermore, many applications require combinatorial multi-output expression, including immunotherapies, gene therapies, and regenerative medicine14,42–45. Such applications will require multiple orthogonal single-output circuits, each encoding a distinct output, thereby increasing circuit complexity, genetic payload, and metabolic burden. To enable a universal multi-output AND gate design, we constructed a circuit that produces a synthetic transcription factor (GAD), previously generated by fusing the yeast GAL4 DNA-binding domain with the viral VP16 transcriptional activation domain35,46. GAD activates the expression of output genes regulated by its cognate synthetic promoter GAL4p (Fig. 1G). Therefore, utilizing GAL4p for output regulation enables the expression of multiple genes through two mechanisms: either using a single output vector encoding a polycistronic mRNA, in which genes are separated by self-cleaving peptides (2 A) or internal ribosomal entry sites (IRES), or co-delivery of multiple output vectors, each encoding a distinct GAL4p-regulated output gene14. The multi-output design executed a robust AND gate function (Fig. 1H), highlighting its potential for applications requiring the expression of combinatorial outputs. To demonstrate the generalizability and potential applicability of this system using secreted effector protein outputs, we replaced the mKate2 output in the GAD-based AND gate with human IL-15, a key cytokine that regulates the activation and proliferation of T and natural killer (NK) cells47. Consistent with AND gate logic, significant IL-15 secretion was detected with ELISA exclusively in the [1,1] state (Fig. 1I). Scalable and orthogonal AND gates To further evaluate the scalability of this approach, we designed four additional BD sequences (BD7, BD8, BD9, and BD10) using the guidelines derived from our previous findings. We then examined the trans-splicing efficiency of each BD with its complementary counterpart and with all other BDs by measuring output levels for each pairwise BD combination by flow cytometry. Significantly higher outputs were generated exclusively from complementary BDs compared to non-compatible pairs (Fig. 1J), further demonstrating the relatively straightforward generation of orthogonal circuit components (Supplementary Data 5 and 6). Finally, to assess potential off-target intermolecular RNA hybridization, we performed paired-end RNA-Seq (median depth ~66 million reads per sample) and STAR alignment with chimeric junction detection enabled. Across all BD constructs and mKate2 controls, off-target junctions (chrMKATE2) constituted only 0.06–0.75% of the total chimeric events and ≤0.013% of all mapped reads, comparable to the expected technical chimeric background of RNA seq libraries (Supplementary Data 8). A combinatorial circuit executing a three-input [(I1 AND I2) AND (NOT I3)] function To demonstrate the processing of more than two inputs using additional Boolean logic, we designed a combinatorial circuit that integrates the activity of three input promoters using a [(I1 AND I2) AND (NOT I3)] logic function. For this purpose, we integrated an intronic synthetic miRNA-based regulation mechanism into the circuit design10,14. The circuit, comprised three modules (Fig. 2A, Fig. S8): Module 1 that consists of SSX1p (I1), regulating the expression of mK2-Ex1 exon, followed by a 5’D and a downstream BD6-S; Module 2 that consists of H2A1p (I2), regulating the expression of BD6-AS, followed by 3’SE, a subsequent mK2-Ex2 exon and downstream three tandem binding sites to a synthetic miRNA mirS1 (mirS1-BSs); and Module 3 that consists of CMVp (I3), regulating the production of mirS1, encoded as an intron within two exons of the ECFP fluorescent gene. Consequently, when only one or none of the input promoters are active, no mKate2 output is generated because no trans-splicing occurs, and the mRNA encoding mKate2 is not produced. Similarly, when either I1 and I3 or I2 and I3 are active, no mKate2 mRNA is produced. When all three promoters are active, the RNAs produced from Modules 1 and 2 are trans-spliced to produce the mKate2 mRNA. However, at the same time, mirS1 is produced by Module 3 and binds its cognate mirS1-BSs encoded in Module 2, thus inhibiting mKate2 production. Only when I1 and I2 are simultaneously active and I3 is not, the mKate2 mRNA is produced by trans-splicing but not inhibited by mirS1, enabling the production of the mKate2 protein. As expected from a circuit executing this function, the mKate2 output in State [1, 1, 0] was approximately 14- to 144-fold higher compared to all other states (Fig. 2B), further demonstrating the modularity of this design approach.Fig. 2A three-input synthetic gene circuit executing a [(Input 1 AND Input 2) AND (NOT Input 3)] logic function.A Schematics of the circuit, incorporating trans-splicing and a synthetic intronic microRNA. Circuit modules are regulated by SSX1p (Input 1), H2A1p (Input 2), and CMVp (Input 3). Modules 1 and 2 generate mKate2 following trans-splicing only when Input 1 and Input 2 are simultaneously active, while Module 3 generates mirS1 that inhibits output expression when Input 3 is active B Output levels under all eight input states. High mKate2 was observed exclusively in the [1,1,0] State, confirming correct circuit function. Input states are defined as 1 (input promoter is active) or 0 (input promoter is a transcriptionally inactive sequence). Data presented as mean ± SEM; (n = 3 biological replicates; p = 0.000109). Data were statistically analyzed by one-way ANOVA with Tukey’s multiple comparisons test; ∗∗∗ indicates p < 0.001 (95% confidence interval). Source data are provided as a Source Data file. Input 1 (I1), Input 2 (I2), Input 3 (I3), mKate2 exon 1 (mK2-Ex1), mKate2 exon 2 (mK2-Ex2), ECFP exon 1 (EC1), ECFP exon 2 (EC2),5’ donor (5’D), 3’ splicing elements (3’SE), binding domain 6 in a sense orientation (BD6-S), antisense sequence of BD6-S (BD6-AS), synthetic microRNA 1 (miRS1), miRS1 binding sites (miRS1-BSs), enhanced cyan fluorescent protein (ECFP). A dashed arrow indicates trans-splicing. Purple line: inhibition by miRS1. Panel A was created in BioRender.com Nissim, L. (2026) https://BioRender.com/5gjbb8a. A three-input synthetic gene circuit executing a [(Input 1 AND Input 2) AND (NOT Input 3)] logic function. A Schematics of the circuit, incorporating trans-splicing and a synthetic intronic microRNA. Circuit modules are regulated by SSX1p (Input 1), H2A1p (Input 2), and CMVp (Input 3). Modules 1 and 2 generate mKate2 following trans-splicing only when Input 1 and Input 2 are simultaneously active, while Module 3 generates mirS1 that inhibits output expression when Input 3 is active B Output levels under all eight input states. High mKate2 was observed exclusively in the [1,1,0] State, confirming correct circuit function. Input states are defined as 1 (input promoter is active) or 0 (input promoter is a transcriptionally inactive sequence). Data presented as mean ± SEM; (n = 3 biological replicates; p = 0.000109). Data were statistically analyzed by one-way ANOVA with Tukey’s multiple comparisons test; ∗∗∗ indicates p < 0.001 (95% confidence interval). Source data are provided as a Source Data file. Input 1 (I1), Input 2 (I2), Input 3 (I3), mKate2 exon 1 (mK2-Ex1), mKate2 exon 2 (mK2-Ex2), ECFP exon 1 (EC1), ECFP exon 2 (EC2),5’ donor (5’D), 3’ splicing elements (3’SE), binding domain 6 in a sense orientation (BD6-S), antisense sequence of BD6-S (BD6-AS), synthetic microRNA 1 (miRS1), miRS1 binding sites (miRS1-BSs), enhanced cyan fluorescent protein (ECFP). A dashed arrow indicates trans-splicing. Purple line: inhibition by miRS1. Panel A was created in BioRender.com Nissim, L. (2026) https://BioRender.com/5gjbb8a. Half-adder circuit Digital adders are core computational components that perform binary addition by integrating multiple input bits to generate SUM and CARRY outputs7–9,48–51. Half adders, which integrate two input bits, are the fundamental building blocks of complex adder circuits, such as full adders. Biological adders are suited for biomedical applications in which the sum of active disease markers is critical for diagnosis and therapeutic intervention. Since the number of active symptoms is often a marker of disease progression, an adder circuit could support treatment tailored to the corresponding disease stage. Therefore, adder circuits would be utilized to treat complex diseases, including cancer, metabolic disorders, and autoimmunity, or to program cellular functions such as combinatorial antigen sensing of CAR-T cells52. To facilitate the construction of a genetic full adder, we engineered a genetic half adder comprising two modules, each including two plasmids: Module M1, which processes Input 1, and Module M2, which processes Input 2 (Fig. 3A, Fig. S9). This circuit simultaneously integrates the activity of two input promoters via an XOR gate to generate the SUM output (e.g., mKate2) and via an AND gate to generate the CARRY output (e.g., ECFP). Additionally, mKate2 production by the XOR gate is mediated by cis-splicing, which is substantially more efficient than the trans-splicing mechanism that generates ECFP in the AND gate. We therefore implemented two alternative modular mechanisms to reduce mKate2 output to levels comparable to ECFP, including synthetic protein degradation tags and promoter engineering.Fig. 3A genetic two-input half-adder.A Schematics of a circuit, incorporating synthetic intronic microRNAs, promoter engineering, and trans-splicing. The circuit executes a half-adder function within a single computational layer to generate appropriate SUM (mKate2) and CARRY (ECFP) outputs. B SUM (mKate2) and CARRY (ECFP) outputs under all four input states. Each input state generated the expected logical output. Input states are defined as 1 (input promoter is CMVp) or 0 (input promoter is a transcriptionally inactive sequence). Data presented as mean ± SEM; n = 3 biological replicates. Thresholds were calculated as \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$(\frac{{{\rm{Max}}}{WM}-{{\rm{Min}}}{WM}}{2})+{{\rm{Min}}}{WM}$$\end{document}(MaxWM−MinWM2)+MinWM are presented by dashed lines for Sum (red) and Carry (blue). Source data are provided as a Source Data file. Input 1 (I1), Input 2 (I2), mKate2 exon 1 (mK2-Ex1), mKate2 exon 2 (mK2-Ex2), ECFP exon 1 (EC1), ECFP exon 2 (EC2), 5’ donor (5’D), 3’ splicing elements (3’SE), binding domain 9 in a sense orientation (BD9-S), antisense sequence of BD9-S (BD9-AS), synthetic microRNA 1 (miRS1), synthetic microRNA 2 (miRS2), miRS1 binding sites (miRS1-BSs), miRS2 binding sites (miRS2-BSs), synthetic protein degradation tag (PEST), TATA-box elements (TATA), enhanced cyan fluorescent protein (ECFP). Panels A and B were created in BioRender.com Nissim, L. (2026) https://BioRender.com/5gjbb8a. A genetic two-input half-adder. A Schematics of a circuit, incorporating synthetic intronic microRNAs, promoter engineering, and trans-splicing. The circuit executes a half-adder function within a single computational layer to generate appropriate SUM (mKate2) and CARRY (ECFP) outputs. B SUM (mKate2) and CARRY (ECFP) outputs under all four input states. Each input state generated the expected logical output. Input states are defined as 1 (input promoter is CMVp) or 0 (input promoter is a transcriptionally inactive sequence). Data presented as mean ± SEM; n = 3 biological replicates. Thresholds were calculated as \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$(\frac{{{\rm{Max}}}{WM}-{{\rm{Min}}}{WM}}{2})+{{\rm{Min}}}{WM}$$\end{document}(MaxWM−MinWM2)+MinWM are presented by dashed lines for Sum (red) and Carry (blue). Source data are provided as a Source Data file. Input 1 (I1), Input 2 (I2), mKate2 exon 1 (mK2-Ex1), mKate2 exon 2 (mK2-Ex2), ECFP exon 1 (EC1), ECFP exon 2 (EC2), 5’ donor (5’D), 3’ splicing elements (3’SE), binding domain 9 in a sense orientation (BD9-S), antisense sequence of BD9-S (BD9-AS), synthetic microRNA 1 (miRS1), synthetic microRNA 2 (miRS2), miRS1 binding sites (miRS1-BSs), miRS2 binding sites (miRS2-BSs), synthetic protein degradation tag (PEST), TATA-box elements (TATA), enhanced cyan fluorescent protein (ECFP). Panels A and B were created in BioRender.com Nissim, L. (2026) https://BioRender.com/5gjbb8a. The XOR gate is executed by plasmids M1.1 (Module 1, Plasmid 1) and M2.1 (Module 2, Plasmid 1). When active, each plasmid induces mKate2 expression while simultaneously repressing the other plasmid using synthetic miRNAs. In the XOR-gate plasmid M1.1, a CMVp input promoter regulates the expression of an RNA that encodes two exons of mKate2 (mK2-Ex1, mK2-Ex2), separated by the synthetic intronic miRNA mirS1. To attenuate mKate2 levels, we located a known tunable synthetic protein degradation tag (PEST) downstream of mK2-Ex253,54. Finally, we encoded multiple perfect match binding sites to the synthetic miRNA mirS2 (mirS2-BSs) downstream of the PEST sequence. Upon expression, this transcript is spliced and produces an mRNA encoding mKate2 fused to PEST with downstream 3’ mirS2-BSs, while the intron produces mirS1. In the XOR-gate plasmid M2.1, we attenuated mKate2 production by engineering the CMVp input promoter. For this purpose, we separated the enhancer region (CMVe) from the TATA-box elements (TATA) using a transcriptionally inactive 80 bp spacer. Consequently, mKate2 production from the modified CMVp (eCMV-80bp-TATA) was decreased approximately 6.5-fold compared to the native CMVp (Fig. S10B). The modified CMVp regulates the expression of an RNA encoding mK2-Ex1 and mK2-Ex2, separated by the synthetic intronic miRNA mirS2, followed by multiple binding sites for the synthetic miRNA mirS1 (mirS1-BSs) located downstream of mK2-Ex2. Upon expression, this transcript is spliced and produces an mKate2 mRNA containing 3’ mirS1-BSs, while the intron is processed by the cell to produce mirS2. The AND gate is executed by plasmids M1.2 (Module 1, Plasmid 2) and M2.2 (Module 2, Plasmid 2). When active, these plasmids express distinct transcripts that undergo trans-splicing to produce ECFP when simultaneously expressed. In the AND gate plasmid M1.2, a CMVp input promoter regulates the expression of an RNA that encodes BD9-AS and a subsequent second ECFP exon (EC-Ex2). In the AND gate plasmid M2.2, a CMVp input promoter regulates the expression of an RNA that encodes the first exon of ECFP (EC-Ex1), followed by a downstream BD9-S. Thus, when both input promoters are active, both transcripts are expressed, trans-splicing occurs, and the CARRY output ECFP is produced. The circuit design ensures that each input state yields the appropriate combination of SUM and CARRY outputs as defined by a half adder logic. When no input promoter is active (State [0, 0]), no transcript is expressed and therefore no output is produced (SUM = 0, CARRY = 0). When only input 1 is active (State [1,0]), mKate2 is produced by Module M1.1, but only the second ECFP exon is expressed from Module M1.2, and no ECFP output is produced (SUM = 1, CARRY = 0). When only input 2 is active (State [0,1]), mKate2 is produced by Module M2.1, while the first ECFP exon is expressed, and no ECFP output is produced (SUM = 1, CARRY = 0). When both inputs are active (State [1, 1]), mirS1 produced by Module M1.1 inhibits the mKate2 transcript produced by Module M2.1 by binding to its corresponding mirS1-BSs, while mirS2 produced by Module M2.1 inhibits the mKate2 transcript produced by Module M1.1 by binding to its corresponding mirS2-BSs. Therefore, no mKate2 is produced (SUM = 0). Simultaneously, the transcripts expressed by M1.2 and M2.2 undergo trans-splicing, and ECFP is produced (CARRY = 1). We examined the circuit output under all possible input states to verify correct SUM and CARRY logic (Fig. 3B). To evaluate this design, activation thresholds were determined by measuring the minimum (OFF) and maximum (ON) reporter WM values from experimental data8. Specifically, the OFF/ON threshold was calculated as \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\left(\frac{{\mbox{Max}}{WM}-{\mbox{Min}}{WM}}{2}\right)+{\mbox{Min}}\,{WM}$$\end{document}MaxWM−MinWM2+MinWM55, yielding a threshold of ~464,666 WM for mKate2 and ~532,168 WM for ECFP. In State [0,0], no significant output was generated by the circuit, corresponding to SUM = 0, CARRY = 0. In States [0, 1] and [1, 0], high mKate2 expression was observed, while ECFP levels remained negligible, corresponding to SUM = 1, CARRY = 0. In State [1, 1], mKate2 expression was suppressed, while ECFP was robustly expressed, corresponding to SUM = 0, CARRY = 1 outputs. These results confirm that the circuit produced the correct output combinations for each input state, consistent with half-adder logic. Full-adder circuit We next demonstrated the applicability of our design approach for the engineering of complex circuits, such as full adders. A digital full adder requires at least two computational layers to integrate three input bits (A, B, and C) into SUM and CARRY outputs. The SUM output is calculated using two sequential XOR operations. Inputs A and B are first integrated by an XOR gate to generate an intermediate signal (S1 = A ⊕ B). This signal is then integrated with input C by a second XOR gate to yield the final SUM output: (SUM = S1 ⊕ C). The CARRY output is computed using two AND gates and one OR gate. The first AND gate integrates inputs A and B (C1 = A ∧ B), and the second AND gate integrates the intermediate signal S1 with input C (C2 = S1 ∧ C). C1 and C2 are then integrated by an OR gate to generate the final CARRY output: (CARRY = C1 ∨ C2). To this end, we constructed a circuit that executes a full adder function to integrate the activity of three input promoters, in which the SUM output is mKate2, and the CARRY output is ECFP. Unlike the multi-layered computational architecture of a digital adder, the full adder circuit comprises three modules, each processing a corresponding input, with all reactions simultaneously occurring within a single computational layer (Fig. 4, Fig. S11).Fig. 4A genetic three-input full-adder.A Schematics of a full-adder circuit, incorporating synthetic intronic microRNAs, native-synthetic hybrid promoters, and trans-splicing. The circuit executes a full-adder function within a single computational layer to generate appropriate SUM (mKate2) and CARRY (ECFP) outputs. B SUM (mKate2) and CARRY (ECFP) under all eight input states. Each input state generated the expected logical output. Data presented as mean ± SEM; n = 3 biological replicates. Thresholds were calculated as \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\left(\frac{{{\rm{Max}}}{WM}-{{\rm{Min}}}{WM}}{2}\right)+{{\rm{Min}}}{WM}$$\end{document}MaxWM−MinWM2+MinWM are presented by dashed lines for Sum (red) and Carry (blue). Source data are provided as a Source Data file. Input 1 (I1), Input 2 (I2), Input 3 (I3), Module 1, Plasmid 1 (M1.1), Module 1, Plasmid 2 (M1.2),Module 1, Plasmid 3 (M1.3), Module 2, Plasmid 1 (M2.1), Module 2, Plasmid 2 (M2.2),Module 2, Plasmid 3 (M2.3),Module 3, Plasmid 1 (M3.1), Module 3, Plasmid 2 (M3.2),Module 3, Plasmid 3 (M3.3),Module 3, Plasmid 3 (M3.4),TATA-box elements (TATA), mKate2 exon 1 (mK2-Ex1), mKate2 exon 2 (mK2-Ex2), ECFP exon 1 (EC1), ECFP exon 2 (EC2), GAD exon 1 (GAD-Ex1), GAD exon 2 (GAD-Ex2), synthetic transcription factor (GAD), 5’ donor (5’D), 3’ splicing elements (3’SE), binding domain 5 in a sense orientation (BD5-S), antisense sequence of BD5-S (BD5-AS), binding domain 6 in a sense orientation (BD6-S), antisense sequence of BD6-S (BD6-AS), binding domain 9 in a sense orientation (BD9-S), antisense sequence of BD9-S (BD9-AS), synthetic microRNA 1 (miRS1), synthetic microRNA 2 (miRS2), synthetic microRNA 3 (miRS3), miRS1 binding sites (miRS1-BSs), miRS2 binding sites (miRS2-BSs), miRS3 binding sites (miRS3-BSs), GAL4 binding sites (G3BSs), GAL4 promoter (G3p), synthetic protein degradation tag (PEST), enhanced cyan fluorescent protein (ECFP). Panel A was created in BioRender.com Nissim, L. (2026) https://BioRender.com/5gjbb8a. A genetic three-input full-adder. A Schematics of a full-adder circuit, incorporating synthetic intronic microRNAs, native-synthetic hybrid promoters, and trans-splicing. The circuit executes a full-adder function within a single computational layer to generate appropriate SUM (mKate2) and CARRY (ECFP) outputs. B SUM (mKate2) and CARRY (ECFP) under all eight input states. Each input state generated the expected logical output. Data presented as mean ± SEM; n = 3 biological replicates. Thresholds were calculated as \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\left(\frac{{{\rm{Max}}}{WM}-{{\rm{Min}}}{WM}}{2}\right)+{{\rm{Min}}}{WM}$$\end{document}MaxWM−MinWM2+MinWM are presented by dashed lines for Sum (red) and Carry (blue). Source data are provided as a Source Data file. Input 1 (I1), Input 2 (I2), Input 3 (I3), Module 1, Plasmid 1 (M1.1), Module 1, Plasmid 2 (M1.2),Module 1, Plasmid 3 (M1.3), Module 2, Plasmid 1 (M2.1), Module 2, Plasmid 2 (M2.2),Module 2, Plasmid 3 (M2.3),Module 3, Plasmid 1 (M3.1), Module 3, Plasmid 2 (M3.2),Module 3, Plasmid 3 (M3.3),Module 3, Plasmid 3 (M3.4),TATA-box elements (TATA), mKate2 exon 1 (mK2-Ex1), mKate2 exon 2 (mK2-Ex2), ECFP exon 1 (EC1), ECFP exon 2 (EC2), GAD exon 1 (GAD-Ex1), GAD exon 2 (GAD-Ex2), synthetic transcription factor (GAD), 5’ donor (5’D), 3’ splicing elements (3’SE), binding domain 5 in a sense orientation (BD5-S), antisense sequence of BD5-S (BD5-AS), binding domain 6 in a sense orientation (BD6-S), antisense sequence of BD6-S (BD6-AS), binding domain 9 in a sense orientation (BD9-S), antisense sequence of BD9-S (BD9-AS), synthetic microRNA 1 (miRS1), synthetic microRNA 2 (miRS2), synthetic microRNA 3 (miRS3), miRS1 binding sites (miRS1-BSs), miRS2 binding sites (miRS2-BSs), miRS3 binding sites (miRS3-BSs), GAL4 binding sites (G3BSs), GAL4 promoter (G3p), synthetic protein degradation tag (PEST), enhanced cyan fluorescent protein (ECFP). Panel A was created in BioRender.com Nissim, L. (2026) https://BioRender.com/5gjbb8a. In this design, Module 1, which processes Input 1, includes three plasmids. Plasmid 1 of Module 1 (M1.1) is regulated by a native-synthetic hybrid promoter, constructed by replacing the transcriptionally inactive 80 bp spacer in the modified CMVp with three tandem GAL-4 binding sites (G3BSs). Consequently, the activity of the hybrid promoter is attenuated in the absence of GAD and restored upon GAD expression (Fig. S10B). The hybrid promoter regulates the expression of an RNA encoding two exons of mKate2 (mK2-Ex1 and mK2-Ex2) separated by the synthetic intronic mirS2, followed by sequential mirS1-BSs and mirS3-BSs. Plasmid 2 of Module 1 (M1.2) is regulated by a bidirectional hybrid promoter. The bidirectional hybrid promoter consists of a native CMVp encoded in sense orientation, fused to a GAL4p comprising three tandem GAL4 binding sites (G3p) encoded in an anti-sense orientation. The CMVp regulates the expression of an RNA encoding the first exon of ECFP (EC-Ex1), followed by 5’D, BD5-S, and sequential BD9-S. G3p regulates the expression of a separate RNA that is identical to the RNA encoded on plasmid M1.1, but encoded in anti-sense orientation. Therefore, this transcript is expressed only when GAD is produced (Fig. S10B). Finally, in Plasmid 3 of Module 1 (M1.3), a CMVp regulates the expression of an RNA encoding EC-Ex1, followed by a downstream 5’D and BD5-S. Module 2, which processes Input 2, includes three plasmids. In plasmid 1 of Module 2 (M2.1), a CMVp regulates the expression of an RNA encoding mK2-Ex1 and mK2-Ex2, separated by the synthetic intronic mirS1, followed by a PEST tag and sequential mirS2-BSs and mirS3-BSs. In plasmid 2 of Module 2 (M2.2), a CMVp regulates the expression of an RNA encoding the first GAD exon (GAD-Ex1), followed by 5’D and sequential BD6-S. In plasmid 3 of Module 2 (M2.3), a CMVp regulates the expression of an RNA encoding BD9-AS, followed by 3’SE and a sequential second exon of ECFP (EC-Ex2). Module 3, which processes Input 3, consists of four plasmids. In plasmid 1 of Module 3 (M3.1), a CMVp regulates the expression of an RNA encoding mK2-Ex1 and mK2-Ex2, separated by the synthetic intronic mirS3, followed by a PEST tag and sequential mirS1-BSs and mirS2-BSs. In plasmid 2 of Module 3 (M3.2), a CMVp regulates the expression of an RNA encoding BD6-AS, followed by 3’SE and a sequential second GAD exon (GAD-Ex2). In plasmid 3 of Module 3 (M3.3), a CMVp regulates the expression of an RNA encoding BD5-AS, followed by 3’SE and sequential EC-Ex2. In plasmid 4 of Module 3 (M3.4), a CMVp regulates the expression of an RNA encoding EC-Ex1, followed by 5’D and a sequential BD9-S. The circuit design ensures that each input state yields the appropriate combination of SUM and CARRY outputs as defined by a full adder logic. In State [0,0,0] (no input promoter is active), no circuit components are expressed, and no output is produced (SUM = 0, CARRY = 0). In State [1,0,0] (only Input 1 is active), the following reactions simultaneously occur: (1) M1.1 produces mKate2 together with mirS2 that lacks any target mRNA; while (2) M1.2 and M1.3 express only the EC-Ex1, therefore no ECFP is produced. Consequently, only the SUM output mKate2 is produced (SUM = 1, CARRY = 0). In State [0,1,0] (only Input 2 is active), the following reactions simultaneously occur: (1) M2.1 produces mKate2 together with mirS1 that lacks any target mRNA; (2) M2.2 expresses only GAD-Ex1, therefore no GAD is produced; and (3) M2.3 expresses only EC-Ex2, therefore no ECFP is produced. Consequently, only mKate2 is produced (SUM = 1, CARRY = 0). In State [0,0,1] (only Input 3 is active), the following reactions simultaneously occur: (1) M3.1 produces mKate2 together with mirS3 that lacks any target mRNA; (2) M3.2 expresses only GAD-Ex2, therefore no GAD is produced; (3) M3.3 expresses EC-Ex2 fused to BD5-AS; and (4) M3.4 expresses only EC-Ex1 fused to BD9-S that is not compatible with BD5-AS, therefore no trans-splicing occurs and ECFP is not produced. Consequently, only mKate2 is produced (SUM = 1, CARRY = 0). In State [1,1,0] (only Inputs 1 and 2 are active), the following reactions simultaneously occur: (1) M2.1 produces mirS1 that binds the mirS1-BSs in the mKate2 transcript expressed by M1.1, thereby suppressing mKate2 production by Module 1; (2) M1.1 produces mirS2 that binds the mirS2-BSs in the mKate2 transcript expressed by M2.1, thereby suppressing mKate2 production by Module 2; (3) EC-Ex1 expressed by M1.2 and EC-Ex2 expressed by M2.3 trans-splice following the binding of BD9-S and BD9-AS, therefore ECFP is produced; and (4) EC-Ex1 expressed by M1.3 and GAD-Ex1 expressed by M2.2 do not produce any protein. Consequently, ECFP is produced, but not mKate2 (SUM = 0, CARRY = 1). In State [1,0,1] (only Inputs 1 and 3 are active), the following reactions simultaneously occur: mKate2 production by M1.1 and M3.1 is mutually inhibited since (1) M3.1 produces mirS3 that binds the mirS3-BSs in the mKate2 transcript expressed by the moderately active native-synthetic hybrid promoter regulating M1.1, thereby suppressing mKate2 production by Module 1; and (2) M1.1 produces mirS2 that binds the mirS2-BSs in the mKate2 transcript expressed by M3.1, thereby suppressing mKate2 production by Module 3; while (3) EC-Ex1 expressed by M1.3 and EC-Ex2 expressed by M3.3 trans-splice following the binding of BD5-S and BD5-AS, therefore ECFP is produced; and (4) EC-Ex1 expressed from both M1.2 and M3.4, and GAD-Ex2 expressed from M3.2 do not produce any protein. Consequently, ECFP is produced, but not mKate2 (SUM = 0, CARRY = 1). In State [0,1,1] (only Inputs 2 and 3 are active), the following reactions simultaneously occur : (1) M3.1 produces mirS3 that binds the mirS3-BSs in the mKate2 transcript expressed by M2.1, thereby suppressing mKate2 production by Module 2; (2) M2.1 produces mirS1 that binds the mirS1-BSs in the mKate2 transcript expressed by M3.1, thereby suppressing mKate2 production by Module 3; (3) EC-Ex1 expressed by M3.4 and EC-Ex2 expressed by M2.3 trans-splice following the binding of BD9-S and BD9-AS, therefore ECFP is produced; (4) EC-Ex2 expressed from M3.3 do not produce any protein; (5) GAD-Ex1 produced by M2.2 and GAD-Ex2 produced by M3.2 trans-splice following the binding of BD6-S and BD6-AS, therefore GAD is produced; (6) GAD binds G3BSs in M1.1, and moderately activates mKate2 expression from the modified native-synthetic hybrid promoter since the upstream enhancer region is not active, yet mirS1 produced by M2.1 and mirS3 produced by M3.1 bind their BSs in this transcript and suppress mKate2 production; and (7) GAD binds G3p in M1.2 and activates mKate2 expression from the bidirectional hybrid promoter, yet mirS1 produced by M2.1 and mirS3 produced by M3.1 bind their BSs in this transcript and suppress mKate2 production. Consequently, ECFP is produced, but not mKate2 (SUM = 0, CARRY = 1). In State [1,1,1] (all Inputs are active), the following reactions simultaneously occur : (1) EC-Ex1 expressed by M1.2 and EC-Ex2 expressed by M2.3 trans-splice following the binding of BD9-S and BD9-AS, therefore ECFP is produced; (2) EC-Ex1 expressed by M1.2 and EC-Ex2 expressed by M3.3 trans-splice following the binding of BD5-S and BD5-AS, therefore ECFP is produced; (3) EC-Ex1 expressed by M1.3 and EC-Ex2 expressed by M3.3 trans-splice following the binding of BD5-S and BD5-AS, therefore ECFP is produced; (4) EC-Ex1 expressed by M3.4 and EC-Ex2 expressed by M2.3 trans-splice following the binding of BD9-S and BD9-AS, therefore ECFP is produced; (5) mirS1, mirS2, and mirS3 are produced by M1.1, M1.2, M2.1, and M3.1 bind to their cognate BSs, and inhibit mKate2 production; yet, (6) GAD-Ex1 expressed by M2.2 and GAD-Ex2 expressed by M3.2 trans-splice, therefore the GAD protein is produced; (7) GAD binds G3BSs in M1.1, and together with the upstream active enhancer sequence restores a full activation of the native-synthetic hybrid promoter, further enhancing mKate2 expression to overcome inhibition by mirS1 and mirS3, and therefore mKate2 is produced; and, in addition, (8) GAD binds G3p in the M1.2 bidirectional hybrid promoter and activates it, inducing mKate2 expression that complements the output from M1.1, thereby further overcoming inhibition by mirS1 and mirS3 and enabling mKate2 production. Consequently, both ECFP and mKate2 outputs are produced (SUM = 1, CARRY = 1). Circuit output was examined under all possible input states to verify circuit performance (Fig. 4B). Inactive modules in previous designs, comprising ≤4 plasmids, were regulated by inactive promoters with minimal basal activity. To avoid transfection-mediated cytotoxicity in the 10-plasmid full adder circuit, we kept the total delivered DNA similar to that used in previous experimental setups. However, although this minimized cytotoxicity, partial plasmid delivery into cells during transfection became significantly more common (Fig. S12). Since proper circuit function requires the simultaneous presence of all 10 plasmids, the absence of essential modules prevents correct processing of the input state, and amplifies the impact of basal activity from leaky promoters in OFF-state modules, leading to misinterpreted outputs56,57. In contrast, empty plasmid backbones do not produce any transcript, thereby eliminating unintended expression from missing modules. Therefore, inactive modules were replaced with empty backbones to minimize incorrect outputs generated in cells that received only a subset of the circuit modules. Consequently, the ability to precisely estimate the contribution of input promoter leakiness to the generation of improper outputs is limited in this experimental setup. Activation thresholds were determined as described above, yielding a threshold of ~335,571 WM for mKate2 and ~293,953 WM for ECFP. In State [0,0,0], both reporters remained below threshold (mKate2~18,000WM, ECFP~15,000WM), corresponding to SUM = 0, CARRY = 0. In all states with only a single active input, mKate2 expression was ON and ECFP expression was OFF (mKate2~371,000–653,000 WM, ECFP~12,000–85,000WM), corresponding to SUM = 1, CARRY = 0. In states with only two active inputs, mKate2 expression was OFF, and ECFP expression was ON (mKate2~83,000–154,000 WM, ECFP~353,000-506,000 WM), corresponding to SUM = 0, CARRY = 1. In State [1,1,1], both reporters were ON (mKate2~472,000 WM, ECFP~575,000 WM), corresponding to SUM = 1 and CARRY = 1. While State [1,1,1] enables more productive EC‑Ex1/EC‑Ex2 trans‑splicing reactions compared to States [1,1,0], [1,0,1], and [0,1,1], ECFP levels were comparable across all four states. These results are consistent with constraints on ECFP production in State [1,1,1], including a stoichiometric bottleneck in the availability of EC‑Ex2 pre‑mRNAs expressed from plasmids M2.3 and M3.3. In summary, the circuit generated correct output states across all input combinations, consistent with full-adder logic. An extended 3-to-1 multiplexer with embedded selector overload status (SOS) In electronics, a multiplexer (MUX) selects one signal from multiple input channels and transmits it to a single output line, based on the State of designated selector inputs. It functions as a digital switch, enabling multiple data sources to share a single transmission channel or computational resources. This architecture reduces hardware complexity, minimizes wiring, and facilitates optimized signal routing in digital systems, including data buses, communication circuits, and memory address decoders9. Genetic multiplexers integrate multiple regulatory elements with selector modules to enable conditional gene expression based on defined regulatory patterns, supporting complex decision-making processes within living cells58,59. In standard multiplexers, the number of input channels equals 2ⁿ, where n denotes the number of selectors. 4-to-1 genetic multiplexers that follow this design principle were previously demonstrated in bacteria and yeast, using recombinases-based architectures, CRISPRi/CRISPRa-based transcriptional circuit gates, synthetic chemoreceptors/ synthetic TFs58–64. In 4-to-1 multiplexers, the selector State [1,1] propagates an output corresponding to the State of a fourth input. By contrast, 3-to-1 multiplexers omit the fourth input, leaving the [1,1] selector State undefined or treated as an error. Alternatively, this State can be assigned to activate a predefined system-level function, such as a monitoring signal, self-test mode, or a debugging process. These ‘selector-overload’ aware designs could facilitate programmable system-level behavior within a minimal encoding space, which is particularly useful for synthetic biology applications. For example, selector-overload states could be engineered to report circuit functionality, detect abnormal cell-states such as cancer and trigger a safety mechanism (e.g., expression of a killer gene), sense extreme cellular stress to modulate circuit activity (analogous to low-power states in electronic circuits), or irreversibly lock the MUX state using a recombinase output. Such configurations could therefore extend the functionality and applicability of genetic multiplexers. To this end, we constructed a 3-to-1 multiplexer with an embedded Selector Overload Status (SOS) that processes the activity of two selector promoters (S0 and S1) and three input promoters (I1, I2, and I3) (Fig. 5, Fig. S13). As in standard multiplexers, selector States [0,0], [0,1], and [1,0] generate outputs corresponding to inputs I1, I2, and I3, respectively. However, under the SOS State [1,1], the circuit produces a dedicated system-level output. As in our previous designs, all reactions occur in parallel within a single computational layer. While standard multiplexers generate a single output corresponding to one of the input lines, we assigned a distinct fluorescent reporter to each selector-defined output, including Sirius for I1, ECFP for I2, mKate2 for I3, and EYFP for the SOS state (Fig. 5A, Fig. S13). This architecture facilitates design optimization by enabling independent verification of each selector-input output and expands circuit functionality by allowing selection of output combinations for downstream iterations. For example, this design can be converted into a combinatorial gate by assigning specific outputs to predefined input combinations. Such functions are essential to address challenges in diseases of a combinatorial nature. For instance, treating autoimmune disorders requires selective inhibition of specific immune pathways only in response to disease-specific input combinations, which often overlap with normal inflammatory signals. This challenge can be addressed by assigning distinct therapeutic outputs to predefined combinations of inflammatory markers. Nonetheless, the modular architecture allows straightforward reconfiguration of the distinct outputs into a single output to execute MUX functionality.Fig. 5A 3-to-1 genetic multiplexer with a Selector Overload Status (SOS).A Schematics of the multiplexer circuit integrating three input promoters (Input 1, Input 2, and Input 3), two selector promoters (S0, S1), and four output reporters: Sirius (reporting Input 1 status), ECFP (reporting Input 2 status), mKate2 (reporting Input 3 status), and EYFP (reporting SOS). The design incorporates trans-splicing and synthetic microRNAs to execute selector-dependent logic. B Normalized fluorescence output for each fluorophore under all selector and input states. Threshold is set to 0.5 RF for all data. Each selector state activated exclusively the corresponding reporter or the SOS output, confirming precise 3-to-1 multiplexer logic and robust implementation of the SOS state. Data presented as mean ± SEM; n = 3 biological replicates. Thresholds were calculated as \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\left(\frac{{{\rm{Max}}}{RF}-{{\rm{Min}}}{RF}}{2}\right)+{{\rm{Min}}}{RF}$$\end{document}MaxRF−MinRF2+MinRF), and is set to 0.5 RF for all data. Thresholds are presented by dashed lines for I1 (purple) I3 (red), I2 (blue) and SOS(yellow). Source data are provided as a Source Data file. Selector 0 (S0), Selector 1 (S1), Input 1 (I1), Input 2 (I2), Input 3 (I3), Selector Overload Status (SOS),TATA-box elements (TATA), mKate2 exon 1 (mK2-Ex1), mKate2 exon 2 (mK2-Ex2), ECFP exon 1 (EC1), ECFP exon 2 (EC2), GAD exon 1 (GAD-Ex1), GAD exon 2 (GAD-Ex2), synthetic transcription factor (GAD), 5’ donor (5’D), binding domain 5 in a sense orientation (BD5-S), antisense sequence of BD5-S (BD5-AS), binding domain 6 in a sense orientation (BD6-S), antisense sequence of BD6-S (BD6-AS), binding domain 9 in a sense orientation (BD9-S), antisense sequence of BD9-S (BD9-AS), synthetic microRNA 1 (miRS1), synthetic microRNA 2 (miRS2), miRS1 binding sites (miRS1-BSs), miRS2 binding sites (miRS2-BSs), GAL4 binding sites (G8p), enhanced cyan fluorescent protein (ECFP), and enhanced yellow fluorescent protein (EYFP). Panel A was created in BioRender.com Nissim, L. (2026) https://BioRender.com/5gjbb8a. A 3-to-1 genetic multiplexer with a Selector Overload Status (SOS). A Schematics of the multiplexer circuit integrating three input promoters (Input 1, Input 2, and Input 3), two selector promoters (S0, S1), and four output reporters: Sirius (reporting Input 1 status), ECFP (reporting Input 2 status), mKate2 (reporting Input 3 status), and EYFP (reporting SOS). The design incorporates trans-splicing and synthetic microRNAs to execute selector-dependent logic. B Normalized fluorescence output for each fluorophore under all selector and input states. Threshold is set to 0.5 RF for all data. Each selector state activated exclusively the corresponding reporter or the SOS output, confirming precise 3-to-1 multiplexer logic and robust implementation of the SOS state. Data presented as mean ± SEM; n = 3 biological replicates. Thresholds were calculated as \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\left(\frac{{{\rm{Max}}}{RF}-{{\rm{Min}}}{RF}}{2}\right)+{{\rm{Min}}}{RF}$$\end{document}MaxRF−MinRF2+MinRF), and is set to 0.5 RF for all data. Thresholds are presented by dashed lines for I1 (purple) I3 (red), I2 (blue) and SOS(yellow). Source data are provided as a Source Data file. Selector 0 (S0), Selector 1 (S1), Input 1 (I1), Input 2 (I2), Input 3 (I3), Selector Overload Status (SOS),TATA-box elements (TATA), mKate2 exon 1 (mK2-Ex1), mKate2 exon 2 (mK2-Ex2), ECFP exon 1 (EC1), ECFP exon 2 (EC2), GAD exon 1 (GAD-Ex1), GAD exon 2 (GAD-Ex2), synthetic transcription factor (GAD), 5’ donor (5’D), binding domain 5 in a sense orientation (BD5-S), antisense sequence of BD5-S (BD5-AS), binding domain 6 in a sense orientation (BD6-S), antisense sequence of BD6-S (BD6-AS), binding domain 9 in a sense orientation (BD9-S), antisense sequence of BD9-S (BD9-AS), synthetic microRNA 1 (miRS1), synthetic microRNA 2 (miRS2), miRS1 binding sites (miRS1-BSs), miRS2 binding sites (miRS2-BSs), GAL4 binding sites (G8p), enhanced cyan fluorescent protein (ECFP), and enhanced yellow fluorescent protein (EYFP). Panel A was created in BioRender.com Nissim, L. (2026) https://BioRender.com/5gjbb8a. In the Selector Module, two plasmids process the State of the first selector, S0 (plasmids S0.1 and S0.2), and two additional plasmids process the State of the second selector, S1 (plasmids S1.1 and S1.2). Plasmid S0.1 consists of the selector input promoter S0p (specifically, CMVp), regulating the expression of an RNA encoding mK2-Ex1 and BD5-S, separated by an intronic mirS2. Importantly, this configuration preserves the 5’ Donor sequence following splicing and production of mirS2, enabling subsequent trans-splicing reactions with suitable target RNAs. Plasmid S0.2 consists of the selector input promoter S0p, which regulates the expression of an RNA encoding BD9-AS upstream of GAD-Ex2. Plasmid S1.1 consists of the selector input promoter, S1p (specifically, CMVp), which regulates the expression of an RNA encoding EC-Ex1 and BD6-S, separated by an intronic mirS1. As in S0.1, this configuration preserves the 5’ Donor sequence following splicing and production of mirS1, enabling subsequent trans-splicing reactions with suitable target RNAs. Finally, plasmid S1.2 consists of the selector input promoter S1p, which regulates the expression of an RNA encoding GAD-Ex1 upstream of BD9-S. In the Inputs Module, plasmids I1, I2, and I3 process the State of each of the three input promoters I1p, I2p, and I3p, respectively. An additional plasmid (plasmid SOS) processes the SOS State. Plasmid I1, which processes Input 1, consists of an input promoter I1p (specifically, CMVp), regulating the expression of an RNA encoding the Sirius gene, followed by sequential mirS1-BSs and mirS2-BSs. Plasmid I2, which processes Input 2, consists of an input promoter I2p (specifically, CMVp) that regulates the expression of an RNA encoding BD6-AS upstream of EC-Ex2, followed by sequential mirS2-BSs. Plasmid I3, which processes Input 3, consists of an input promoter I3p (specifically, CMVp), regulating the expression of an RNA encoding BD5-AS upstream of mK2-Ex2, followed by sequential mirS1-BSs. Finally, plasmid SOS, which processes the SOS State, encodes a GAL4p comprising eight tandem GAL4 binding sites (G8p) that regulate the expression of EYFP. In the selector State [0,0] (S0p and S1p are inactive), the circuit is designed to report only the activity of I1. In this case: (1) if I1p is inactive, the Selectors Module does not expresses any circuit component; (2) if I1p is active, Sirius is produced; (3) if I2p is active, it expresses only the second ECFP exon (EC-Ex2), therefore no ECFP is produced; (4) if I3p is active, it expresses only the second mKate2 exon (mK2-Ex2), therefore no mKate2 is produced; and (5) G8p remains inactive since no GAD is produced, and thus no EYFP is produced. Consequently, if I1p is inactive, no output is produced, corresponding to an I1 OFF output. If I1p is active, Sirius is produced, corresponding to an I1 ON output. I2, I3, and SOS do not produce any output regardless of their activity, reflecting their exclusion from the output line. In selector State [0,1] (S0p inactive, S1p active), the circuit is designed to report only the activity of I2. In this case: (1) plasmids S0.1 and S0.2 expresses no circuit components; (2) plasmid S1.1 expresses EC-Ex1 and produces mirS1; (3) plasmid S1.2 expresses only GAD-Ex1, and therefore no functional GAD protein is produced; (4) if I1p is active, mirS1 produced from S1.1 binds the mirS1-BSs in the Sirius transcript, thereby suppressing Sirius production by I1; (5) if I2p is active, EC-Ex2 expressed by plasmid I2 trans-splices with EC-Ex1 expressed by S1.1 following BD6-AS and BD6-S binding, therefore ECFP is produced; (6) if I3p is active, plasmid I3 expresses only the second mKate2 exon (mK2-Ex2), therefore no mKate2 is produced; and (7) the SOS plasmid remains inactive since no functional GAD is produced, and therefore no EYFP is produced. Consequently, if I2p is inactive, no output is produced, corresponding to an I2 OFF output, while if I2p is active, ECFP is produced, corresponding to an I2 ON output. I1, I3, and SOS do not produce any output regardless of their activity, corresponding to an exclusion from the output line. In selector State [1,0] (S0p active, S1p inactive), the circuit is designed to report only the activity of I3. In this case: (1) plasmid S0.1 expresses mK2-Ex1 and produces mirS2; (2) plasmid S0.2 expresses only GAD-Ex2, and therefore no functional GAD protein is produced; (3) plasmids S1.1 and S1.2 do not express any circuit components; (4) if I1p is active, mirS2 produced from S0.1 binds the mirS2-BSs in the Sirius transcript, thereby suppressing Sirius production by I1; (5) if I2p is active, plasmid I2 expresses only the second ECFP exon (EC-Ex2), and therefore no ECFP is produced; (6) if I3p is active, mK2-Ex2 expressed by plasmid I3 trans-splices with mK2-Ex1 expressed by S0.1 BD5-AS and BD5-S binding, therefore mKate2 is produced; and (7) the SOS plasmid remains inactive since no functional GAD is produced, therefore no EYFP is produced. Consequently, if I3p is inactive, no output is produced, corresponding to an I3 OFF output, while if I3p is active, mKate2 is produced, corresponding to an I3 ON output. I1, I2, and SOS do not produce any output regardless of their activity, corresponding to an exclusion from the output line. In selector State [1,1] (both S0p and S1p active), the circuit is designed to report only the SOS State. In this case: (1) plasmid S0.1 expresses mK2-Ex1 and produces mirS2; (2) plasmid S0.2 expresses GAD-Ex2; (3) plasmid S1.1 expresses EC-Ex1 and produces mirS1; (4) plasmid S1.2 expresses GAD-Ex1 that trans-splices with GAD-Ex2 expressed by S0.2, thereby functional GAD protein is produced; (5) if I1p is active, mirS2 produced by S0.1 and mirS1 produced by S1.1 bind the corresponding mirS2-BSs and mirS1-BSs in the Sirius transcript, thereby suppressing Sirius production by I1; (6) if I2p is active, EC-Ex2 expressed by plasmid I2 can trans-splice with EC-Ex1 expressed by S1.1 following the binding of BD6-AS and BD6-S, but mirS2 produced by S0.1 binds mirS2-BSs in this transcript, thereby suppressing ECFP production; (7) if I3p is active, mK2-Ex2 expressed by plasmid I3 can trans-splice with mK2-Ex1 expressed by S0.1 following the binding of BD5-AS and BD5-S, but mirS1 produced by S1.1 binds mirS1-BSs in this transcript, thereby suppressing mKate2 production; and (8) the functional GAD protein binds and activates the G8p promoter encoded on plasmid SOS, thereby EYFP is produced. Consequently, EYFP is exclusively produced in the selector State [1,1], corresponding to the SOS State ON output. I1, I2, and I3 do not produce any output regardless of their activity, corresponding to an exclusion from the output line. Circuit performance was systematically evaluated under all selector-input combinations (Fig. 5B). In this experimental setup, active modules were regulated by the CMV promoter (CMVp), while inactive modules were regulated by a non-specific DNA sequence. To enable direct comparison, fluorescence outputs for each fluorophore were normalized to their maximum intensity within each replicate and defined on a relative fluorescence (RF) scale ranging from 0 to 1. The OFF/ON threshold was set at approximately 0.5, consistent with the threshold calculations. In the selector state [0,0], designed to report only the activity of input I1, relative Sirius fluorescence corresponding to input I1 was high exclusively when promoter I1p was active (RF~0.68–0.99) and remained low (RF~0.02–0.03) when inactive. The RF of ECFP, mKate2, and EYFP consistently remained below the threshold, regardless of the input promoter states (~0.09–0.13, ~0.04–0.09, and ~0.25–0.33, respectively), confirming the exclusive reporting of input I1. In the selector state [0,1], designed to report only the activity of input I2, relative ECFP fluorescence corresponding to input I2 was high exclusively when promoter I2p was active (RF~0.73–0.99), and low (RF~0.08–0.16) when inactive. The RF of Sirius, mKate2, and EYFP consistently remained below threshold levels regardless of their input promoter states (~0.02–0.06, ~0.06–0.09, and ~0.15–0.25, respectively), confirming the exclusive reporting of input I2. In selector state [1,0], designed to report only the activity of input I3, mKate2 fluorescence corresponding to input I3 was high exclusively when promoter I3p was active (RF~0.50–0.93) only upon activation of I3p, and remained low (RF~0.05–0.06) otherwise. Sirius, ECFP, and EYFP fluorescence were consistently below threshold levels (~0.02–0.07, ~0.11–0.16, and ~0.17–0.20, respectively), confirming exclusive reporting of input I3. In the SOS state [1,1], EYFP RF was always high (~0.52–0.84), indicating successful activation of the dedicated SOS state output. RF from Sirius, ECFP, and mKate2 was consistently below threshold (~0.01–0.04, ~0.14–0.22, and ~0.07–0.10, respectively), verifying that this State exclusively reported the SOS state. Collectively, these data demonstrate the first implementation of robust and dynamic 3-to-1 selector-dependent multiplexing with an SOS state, further highlighting the modularity and scalability of our circuit design strategy. Discussion In this study, we present a modular and scalable strategy for constructing synthetic gene circuits that execute complex functions within mammalian cells, which combines RNA trans-splicing, synthetic microRNAs, and native-synthetic hybrid promoters. This circuit design strategy effectively addresses challenges associated with the limited availability of orthogonal genetic elements while reducing the number of genetic components and computational layers required to execute complex functions in mammalian cells (Supplementary Data 7). By lowering genetic complexity, this approach reduces the encoding capacity and cellular resource demands, thereby supporting the practical implementation of synthetic gene circuits65,66. As proof of concept for this approach, we established versatile circuits executing Boolean gates, combinational logic, and functions analogous to electronic devices, including a half adder, a full adder, and an extended 3-to-1 multiplexer with embedded SOS state. Such functions are essential to enable diagnosis of disease onset, therapeutic intervention tailored to disease progression, precise targeting, or addressing complex diseases of a combinatorial nature by assigning specific outputs to predefined input combinations6,14,42–45,52,62,67–69. Future studies focusing on the comprehensive optimization of circuit components are expected to enhance performance and orthogonality70,71. The generation of BD sequences would benefit from a systematic evaluation of key parameters, including length, GC content, internal mismatches, nucleotide composition, secondary structures, and sequence complexity. In addition, the design of synthetic miRNAs must account for sequences present in native RNAs to avoid unintended off-target interactions with the cellular machinery. In parallel, circuit designs should avoid components that could become potential targets for endogenous regulators, such as sequences targeted by native miRNAs. Moreover, integrating complex multi-vector circuits into mammalian genomes presents an additional layer of complexity. To address this challenge, future work should explore methods such as landing pad technologies or CRISPR/Cas-mediated integration to facilitate precise, stable, and efficient genomic insertion of these circuits3,6,12,20,26,31,59,72,73. To maintain circuit functionality, appropriate insulator sequences could be incorporated to avoid cross-regulation with flanking regulatory elements or silencing by DNA methylation74–76. Consequently, further optimization of circuit design that accounts for vector copy number and delivery platform may be required. Finally, this approach could be expanded by incorporating additional synthetic components. Cell-state-specific promoters that are selectively activated in defined cellular contexts can be used as circuit inputs. Such promoters would provide precise regulation, reduce off-target effects, and enable rational programming of cellular activities essential for biomedical research and therapeutic interventions77–80. Artificial promoters, including Synthetic Promoters with Enhanced Cell-state Specificity (SPECS), typically exhibit minimal basal activity, enhanced specificity, and potent ON-state activity, which outperform native promoters77. Therefore, SPECS can be particularly useful for enhancing circuit performance and mitigating leaky promoter activity in the OFF-states. Furthermore, orthogonal synthetic transcription factors based on CRISPR-Cas9 systems, TALEs, or zinc-fingers would provide additional scalability6,23,42,81,82. Such genetic components could be easily integrated into our circuits, given the modular nature of these designs. In conclusion, this modular synthetic gene circuit engineering platform constitutes a significant advancement in synthetic biology, offering a robust, tunable, scalable, and orthogonal toolkit for executing complex computational operations in mammalian cells. This approach provides essential building blocks for precision diagnostics, targeted therapies, sophisticated biosensors, and cellular computing devices. For example, multi-input circuits enable the programming of immune cells for cell therapies83,84, enhance the precision of gene therapies, and facilitate in vivo diagnostics of complex diseases3. Furthermore, combinatorial and MUX circuits are well-suited to address inherently heterogeneous diseases such as cancer85 or immunological disorders that require targeting of specific inflammatory patterns86. Therefore, this work supports next-generation biomedical and biotechnological applications. Methods Cell culture Low-passage HEK-293T cells (CRL-3216, ATCC) were cultured in Dulbecco’s modified Eagle medium (DMEM; 01-055-1 A, Biological Industries) supplemented with 10% fetal bovine serum (FBS; 04-007-1 A, Biological Industries), 1% sodium pyruvate (03-042-1B, Biological Industries), 1% MEM non-essential amino acids (MEM/NEAA; 01-340-1B, Biological Industries), and 100 units/ml penicillin with 0.1 mg/ml streptomycin (03-031-1B, Biological Industries). Cells were cultured at 37 °C with 5% CO2. Plasmid construction Plasmids in this study were procured from Addgene or constructed using conventional restriction cloning or Gibson assembly. Principal plasmids and associated sequences are available from Addgene. Addgene catalogue numbers for these plasmids are detailed in the ‘Supplementary Data 9’ file. Sequences of plasmids not deposited to Addgene are provided in GenBank format in the ‘Supplementary Data 10’ file. All plasmids can be obtained from the corresponding author upon request. Plasmid compositions Plasmid compositions for each circuit design are detailed in the Supplementary Data 9 file. Transfections HEK293T cells were transiently co-transfected in 6-well plates with a mixture of 6 μL of the transfection reagent SusFexin (Fisher Scientific, 50-204-5795), with 100 μL of OptiMEM (31985-047, Thermo Fisher Scientific), and combined with a DNA mixture containing up to 10 plasmids expressing the appropriate circuit parts. The total DNA amount per sample was up to 4 μg, adjusted according to the number of plasmids in each circuit. Each plasmid was added in equal amounts within a given circuit, resulting in the same total DNA in each sample of the transfected circuit. HEK293T suspension cells were prepared and diluted to 1.2 × 10⁶ cells/mL in culture medium. 0.5 mL of the cell suspension (0.6× 10⁶ cells) was added to each SusFexin/DNA complex tube and mixed. The mixture was then transferred to the designated well of a 6-well plate to a final volume of 1.6 mL and incubated at 37 °C with 5% CO₂. 24 h post-transfection medium was replaced with 2 mL of fresh culture medium. In some experiments, mentioned in figure legend, HEK293T cells were co-transfected in 24-well plates in the same manner with a mix of 2 μL of the transfection reagent FuGENE-HD (Promega, E2312), 42.5 μL OptiMEM, and 0.25 μg, of each plasmid. 0.5 mL of 10⁵ suspension cells was added to a mixture tube. 24 h post-transfection, 100μLfresh media was carefully added. Flow cytometry To characterize fluorescent protein expression, 48 h post-transfection, samples were prepared for flow cytometry analysis. Cell medium was removed, cells were washed with PBS (02-023-1 A, Biological Industries), detached with 0.5 µL trypsin-EDTA (25200056, Rhenium), neutralized with an equal volume of medium, and centrifuged at 1500 g for 3 minutes to remove residual trypsin. Cell pellet was resuspended in PBS and analyzed using a CytoFLEX S flow cytometer (Beckman Coulter) with the appropriate laser setting. Gating strategy is depicted in Fig. S14. A minimum of 20,000 cells was recorded per sample. Data analysis was performed using Kaluza software 2.1 (Beckman Coulter). The weighted median (WM) was calculated by multiplying the percentage of fluorescence-positive cells (% positive) by the median fluorescence of the positive cell population, yielding a value that correlates fluorescence intensity with the number of fluorescent-positive cells (Nissim et al., 2014). Microscopy All images were created using the EVOS M5000 Imaging System (ThermoFisher Scientific). All images included in the study were generated at a resolution of 300 DPI. RNA extracion 48 h post-transfection cell medium was removed, and cells were washed with PBS, detached with 0.5 µL trypsin-EDTA, neutralized with an equal volume of medium, and centrifuged at 1500 g for 3 minutes to remove trypsin residues. The cell pellet was twice resuspended in PBS. After cells were washed, PBS was removed, and the cell pellets were frozen at −80 °C for 20–24 hours. RNA was isolated using RNeasy plus mini kit (74134, QIAGEN), according to the manufacturer’s protocol. RNA concentrations were measured by Tape Station. cDNA and Real-Time PCR (RT PCR) cDNA synthesis was performed with 600 ng of RNA template using qScript reverse transcriptase (95047-100, Quantabio). The reaction conditions were 22 °C for 5 min, followed by 42 °C for 30 min, and inactivation at 85 °C for 5 min. 20 ng of cDNA template was used for the real-time RT-PCR with Luna qPCR master mix (NEB-M3003E, NEB) and specific primers set for the detection of the trans-spliced mKate2, our output gene of interest. Primer sequences: prKRQ1: TGTACGGCAGCAAAACCTTC, prKRQ2: CGTATGTGGTGACTCTCTCCC. Quantification of PCR signals of each sample was performed by the ΔCT methods normalized to Gapdh housekeeping gene. cDNA and reverse-transcriptase PCR cDNA was performed using 1ug RNA template with a specific mkate2 primer(R: GGAATTAAGTTTGTGCCCCAGT) by using SuperScript RTIII (18-080-085, Invitrogen) kit, with 10 μM of dinucleotide triphosphates (dNTP) mix. The reaction conditions were 55 °C for 60 min, followed by inactivation at 70 °C for 15 min. PCR on cDNA template was performed using specific mKate2 primers for full-length mKate2 gene sequencing. Primers sequencing: F: TGTCTAAGGGCGAAGAGCTGAT, R: GGAATTAAGTTTGTGCCCCAGT. PCR products were sent for sequencing analysis. ELISA assay IL-15 secreted protein was identified in the cell culture supernatant of each sample using the Human IL-15 ELISA kit (435104, BioLegend) according to the manufacturer’s instructions. The measurements were performed with a plate reader. Absorbance was measured at 450 nm in all samples. RNA-Seq data analysis RNA-seq libraries were prepared using poly-A selection with Watchmaker mRNA Library Prep Kits and sequenced on an Illumina NovaSeqX platform. Paired-end reads were trimmed for poly-A/T stretches and Illumina adapters using Cutadapt87, discarding those shorter than 30 bp, then aligned to the human GRCh38 reference genome augmented with the mKate2 sequence (with Ensembl annotations) (with Ensembl annotations) using STAR88 in two-pass Basic mode with --alignEndsType EndToEnd, --outFilterMismatchNoverLmax 0.04, and chimeric detection enabled (-- chimSegmentMin 20, --chimJunctionOverhangMin 15, --chimOutType WithinBAM HardClip Junctions) to identify potential trans-splicing events. PCR duplicates were removed with Picard MarkDuplicates. The pipeline was executed via Snakemake89, with custom Python scripts used to parse Chimeric.out.junction files, extracting chrMKATE2-involving events and distinguishing off-target fusions (MKATE2 paired with other genes) from intra-chrMKATE2 junctions. Statistics & reproducibility All graphs and statistical analyses were performed using GraphPad Prism 10 software (GraphPad Software Inc., San Diego, CA, USA). Data are presented as the mean of at least n = 3 biological replicates, with individual data points depicted as dots and error bars denoting the standard error of the mean (SEM). Data were statistically analyzed using one-way ANOVA followed by Tukey’s, Dunnett’s, or Bonferroni’s multiple comparisons tests, as specified in the figure legends. Statistical significance was defined as P-values < 0.05 with a 95% confidence interval. Thresholds were calculated as \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\left(\frac{{\mbox{Max}}{WM}-{\mbox{Min}}{WM}}{2}\right)+{\mbox{Min}}\,{WM}$$\end{document}MaxWM−MinWM2+MinWM. No statistical method was used to predetermine sample size. No data were excluded from the analyses. The investigators were not blinded to allocation during the experiments and outcome assessment. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Supplementary Information Description of Additional Supplementary Files Supplementary Data 1-9 Supplementary Data 10 Reporting Summary Transparent Peer Review file Supplementary Information Description of Additional Supplementary Files Supplementary Data 1-9 Supplementary Data 10 Reporting Summary Transparent Peer Review file Source data Source Data Source Data