Computational design of orthogonal operators and synthetic repressors for transcriptional logic gates

Orthogonal repressor-based logic gates have enabled the predictive design of genetic circuits using automated design algorithms. However, engineering complex genetic circuits has been limited by the availability of transcriptional logic gates that are orthogonal, exhibit low cytotoxicity, and have a large dynamic range. Here, we present an algorithmic approach to design orthogonal synthetic repressors and logic gates de novo. To demonstrate the approach, we used programmable DNA-binding proteins called “transcription activator-like effectors” (TALEs) to build orthogonal transcriptional logic gates. We automated the design of synthetic repressors by developing a Python script that implements a search algorithm and genetic design rules to generate a set of operator sequences that are orthogonal from one another, the specified host genome, and an optional set of other heterologous sequences. We demonstrate this approach by creating a library of 20 characterized synthetic repressors for Escherichia coli that complements natural TetR-family repressors used to build genetic circuits. TALE repressors were built from monomer DNA modules to enable prototyping of any 19 bp DNA-binding sequence and then recoded to remove repetitive sequences for final use in logic gates. A library of 20 orthogonal NOT logic gates was tuned and characterized. Synthetic logic gates were used to construct genetic circuits for sequential logic in combination with existing natural repressor gates and genetic circuit design algorithms. This study demonstrates a computational approach to designing orthogonal operator and repressor sequences for creating logic gates and scalable genetic circuit construction.

Complex behavior and dynamic responses can be encoded in bacteria using synthetic genetic circuits and have been employed for a range of applications, including in situ health monitoring, therapeutics, environmental monitoring, bioremediation, and biomanufacturing (Kotula et al., 2014; Xu et al., 2014; Danino et al., 2015; Gupta et al., 2017; Tay et al., 2017; Doong et al., 2018; Isabella et al., 2018; Raman et al., 2021; Triassi et al., 2023; Chemla et al., 2026). These genetic regulatory networks can be implemented in bacteria using various regulator classes, including CRISPRi (Nielsen and Voigt, 2014; Didovyk et al., 2016; Santos-Moreno et al., 2020; Liu et al., 2022; Tickman et al., 2022; Barbier et al., 2023; Kusumawardhani et al., 2025), riboregulators (Green et al., 2017; Kim et al., 2019; Gao et al., 2025; Hoynes-O’Connor and Moon, 2016), recombinases (Bonnet et al., 2012; Siuti et al., 2013; Huang et al., 2024; Greco et al., 2025), and DNA-binding transcriptional repressors (e.g., TetR family and bacteriophage repressors) (Gardner et al., 2000; Nielsen et al., 2016; Andrews et al., 2018; Bartoli et al., 2020; Shin et al., 2020; Lebovich et al., 2023; 2024; Padmakumar et al., 2025). However, it is necessary that the constituent regulators be highly specific and regulate expression without cross-talk between the gates and components of the genetic circuit’s design.

Transcriptional repressors and their corresponding regulated promoters have been widely used to build logic gates and have been layered to construct genetic circuits in bacteria (Gardner et al., 2000; Yokobayashi et al., 2002; Ellis et al., 2009; Tamsir et al., 2011; Nielsen et al., 2016). A primary source of these regulators has been the identification of natural repressors from various genomes. Genomic mining of bacterial genomes yielded 16 natural TetR-family repressors and cognate promoters that are suitable for use in genetic circuits (Stanton et al., 2014). However, some of the TetR-family repressors are highly toxic to the host cell and exhibit compounded toxicity when combined with one another (Nielsen et al., 2016; Andrews et al., 2018). More recently, the utilization of natural repressors from various bacteriophages yielded 12 orthogonal gates suitable for the construction of genetic circuits (Padmakumar et al., 2025). The availability of the orthogonal gates determines the size and complexity of the available circuit design space. Another necessary requirement is that a chosen repressor for a genetic circuit does not have off-target binding and unintended regulation of the host’s genome. Therefore, the use of natural repressors is limited in the organism from which they originated. The generation of synthetic repressors could circumvent these limitations and expand the components available for generating genetic circuits.

Transcription activator-like effectors (TALEs) could serve as a source of synthetic repressors for use in genetic circuits. TALEs are a class of programmable DNA-binding proteins originally discovered in bacterial plant pathogens belonging to the genus Xanthomonas (Boch et al., 2009; Moscou and Bogdanove, 2009). Their DNA-binding domains are composed of 34 amino acid repeat motifs, with each repeat recognizing a single nucleotide. Within each repeat are two variable amino acids at positions 12 and 13, known as “repeat variable diresidues” (RVDs), that determine the DNA nucleotide base that the repeat binds to (NI = adenine, NG = thymine, NN = guanine, and HD = cytosine preferentially) (Boch et al., 2009; Moscou and Bogdanove, 2009; Becker et al., 2022). The DNA-binding sequence recognized by a TALE is determined by the identity of the repeats as they are strung together. Previous research has shown that synthetic TALE repressors targeted to the lac operator could repress the lac promoter more tightly than LacI in Escherichia coli (Politz et al., 2013; Copeland et al., 2016). The ability of custom-built TALEs to target specified DNA sequences (Garg et al., 2012) and strongly regulate transcription in prokaryotes makes them a potential strategy for expanding the availability of characterized logic gates for use in the automated design of bacterial genetic circuits.

In this study, we show that highly orthogonal synthetic repressors can be designed computationally and prototyped using modular TALE repressors to create libraries of transcriptional gates for use in genetic circuits (Figure 1A). We report a Python script to automate the constraint-based design of large sets of orthogonal operator sequences. We demonstrate this approach by constructing a library of 20 synthetic repressors that are recoded and used in 84 characterized NOT gates. Functional genetic circuits that use these TALEs in combination with TetR-family repressors are shown. We also demonstrate custom TALE repressors to regulate endogenous gene expression of the host bacterium.

Strains and media

Constructs were cloned and tested using E. coli NEB10-beta (New England Biolabs), which is a derivative strain of DH10B. The strains used for cloning purposes were grown in Miller Luria-Broth (LB) medium or on LB agar supplemented with ampicillin (100 μg/mL), kanamycin (50 μg/mL), or spectinomycin (50 μg/mL) (Gold Biotechnology), as appropriate. Assays used M9 medium (6.78 g/L Na 2 HPO 4, 3 g/L KH 2 PO 4, 1 g/L NH 4 Cl, 0.5 g/L NaCl, 0.34 g/L thiamine HCl, 0.4% w/v D-glucose, 0.2% w/v casamino acids, 2 mM MgSO 4, and 0.1 mM CaCl 2) supplemented with antibiotics as appropriate.

DNA assembly and cloning of constructs

PCR was performed using Q5 DNA polymerase (New England Biolabs). Primers and DNA oligonucleotides were ordered from Integrated DNA Technologies (IDT). Type IIS DNA assembly was performed using high-concentration T4 DNA ligase and the IIS restriction enzymes BbsI, BsaI, or SapI (New England Biolabs). Plasmid purifications were performed with the Zyppy Plasmid Miniprep Kit (Zymo Research). DNA purifications were performed using the DNA Clean & Concentrator 5 µg kit (Zymo Research). Lists of the plasmids and DNA sequences used in this study can be found in the Supplementary Information (Supplementary Tables S1, S2).

Synthetic operator design

A custom Python script was created to design synthetic TALE operator sequences based on GC content and homology between the designs and the host organism’s genome, synthetic operators in the library, and other specified input sequences. The BLAST+ application, optimized for short sequences, was used to align the synthetic operator sequences and determine homology (Camacho et al., 2009). Detailed information about the script, including its usage, inputs, and outputs, can be found in the Supplementary Information (Supplementary Note 1). The synthetic operator sequence library used in this study was created by constraining the first and last nucleotide to adenosine and thymine, respectively, and randomly selecting the middle 17 nucleotides (nt) to create a 19-nt operator sequence. Each operator sequence was restricted to GC content between 30% and 60%, less than 13 bp homology to the E. coli DH10B genome (accession number CP000948.1), and less than 9 bp homology between each synthetic operator and the operator sequences for a set of TetR homologs (Stanton et al., 2014). The Python script is available on GitHub (https://github.com/AndrewsLabSynBio).

Synthetic TALE repressor hierarchical DNA assembly

TALE monomers coding for each of the four RVDs (NN, NI, NG, and HD) were ordered as gene fragments (IDT) and assembled using hierarchical Type IIS DNA assembly. Gene fragments were PCR-amplified to add IIS restriction enzyme cut sites and orthogonal linker sequences for hierarchical DNA assembly (Supplementary Table S3 and Supplementary Note 2). Orthogonal linkers were identified using Linker Design, a published Python script (Woodruff et al., 2017). PCR products of monomer RVDs were inserted into a ColE1 plasmid backbone carrying a kanamycin resistance marker (Kan R) via IIS assembly with BbsI. Six monomer plasmids were used to assemble 6-mers into a ColE1 plasmid backbone carrying an ampicillin resistance marker (Amp R) via IIS assembly with BsaI. Assembly of the 6-mers was confirmed via Sanger sequencing (Genewiz). To construct a full TALE, the plasmids for three 6-mers (groups A, B, and C) and a PCR product for the final half-repeat RVD (0.5-mer) were assembled using BbsI into a plasmid backbone (KanR, ColE1) with the TALE N- and C-terminal domain sequences flanking the insertion site (Supplementary Figure S1).

TALE recoding

Recoding of the TALE gene library was performed using a combination of Genome Calligrapher and the EFM Calculator (Christen et al., 2015; Jack et al., 2015). First, a sequence file containing the original coding of all 20 TALE genes was uploaded to Genome Calligrapher. The program was set to remove repeated sequences >20 bp and avoid the recognition sequences for BbsI, BsaI, and SapI. The output from Genome Calligrapher was then checked using the EFM Calculator, which identified repeats that Genome Calligrapher failed to remove. Any remaining repeats were removed through a process of manual recoding and evaluation with the EFM Calculator. The originally used linker sequences for the 6-mers and between the third and fourth monomers were restored in the recoded sequences to allow for DNA assembly (Supplementary Table S3).

Recoded TALE assembly

Recoded TALE sequences were divided into fragments to facilitate gene synthesis. The divided sequences included the N-terminus, the C-terminus with the half-repeat, and the repeat domains divided into 3-mers. DNA fragments were ordered (Twist Bioscience), cloned into plasmids, and sequenced. Pairs of 3-mers were assembled into the original 6-mer plasmid backbones. The constructed 6-mers were then assembled with the N- and C-termini to form the full recoded TALEs in a DNA assembly reaction with BbsI.

DNA assembly of genetic circuit and gate plasmids

DNA constructs were assembled using Type IIS assembly as described by Woodruff et al. (2017). In brief, repressor genes were placed on a plasmid backbone (Kan R, ColE1) with the ribozyme, RBS, and terminator for the gate to create a repressor composite part flanked by BsaI cut sites and linkers for the subsequent assembly reaction. The repressor composite part was then assembled to the desired promoter on a plasmid (Amp R, ColE1) to form a transcription unit (TU) construct. Depending on the position of each TU in the final construct, the TU plasmid included flanking BbsI sites with orthogonal linkers for the final assembly. TUs were cloned into a three-input circuit backbone (pAN871) via BbsI assembly (Andrews et al., 2018).

Ribosome binding site tuning

Ribosome binding site (RBS) tuning of TALE expression was performed by designing RBSs for E. coli using the RBS Calculator (Salis et al., 2009; Borujeni et al., 2014; Espah Borujeni and Salis, 2016; Espah Borujeni et al., 2017). For the original TALEs, RBSs with 500, 1,000, and 3500 TIR were designed. RBSs were introduced by inverse PCR onto the TALE plasmid backbone prior to the full TALE assembly. RBSs for each recoded TALE were designed to have 500, 1,000, and 3,000 TIR. Additional RBSs were designed if needed and were cloned into the TALE plasmid backbones by IIS DNA assembly using a pair of annealed DNA oligonucleotides to simultaneously assemble the TALE with the RBS (Supplementary Figure S2).

Repressor toxicity assay

E. coli NEB10-beta carrying a TALE or TetR repressor gate characterization plasmid was grown overnight (16 h) in V-bottom 96-well plates from a fresh colony in 200 µL of M9 medium with kanamycin, shaking at 1,000 RPM and 37 °C in an ELMI DTS-4 microtiter plate shaker incubator. Overnight cultures were diluted to an OD 600 of 0.01 in 200 µL M9 medium supplemented with kanamycin and 1 mM isopropyl β-d-1-thiogalactopyranoside (IPTG) and grown in flat-bottom 96-well plates at 1,000 RPM and 37 °C. OD 600 measurements were taken every hour using an Epoch microplate spectrophotometer (BioTek). Toxicity was quantified by comparing the specific growth rate during the log phase with that of NEB10-beta harboring the empty circuit plasmid (pAN871). Specific growth rate was determined by linear regression of the exponential phase data points.

Gate characterization assay

Characterization of NOT gates and other genetic parts was performed as per Lebovich et al. (2023). In brief, strains were grown overnight from a single colony for 16 h in V-bottom 96-well plates sealed with a breathable membrane in 200 µL M9 medium supplemented with the appropriate antibiotic and incubated at 1,000 RPM and 37 °C in an ELMI DTS-4 incubator. Overnight cultures were diluted 178-fold (two serial dilutions of 15 µL into 185 µL) and grown for 3 h. Cultures were then diluted twice with 15 µL into 185 µL, and then 3 µL was added to 145 µL of M9 with varying concentrations of inducer and the appropriate antibiotic. Cultures were grown for 5 h as mentioned above and then diluted into 200 µL of phosphate-buffered saline (PBS, pH 7.4) containing 2 mg/mL kanamycin to arrest protein expression. After incubation for 30 min at room temperature, flow cytometry was run on a 3-laser BD DUAL LSRFortessa (BD) to measure cell fluorescence for approximately 10,000 cells. Unless otherwise noted, output measurement was eYFP fluorescence converted to relative promoter units (RPUs) using the standard reference plasmid (pAN1717) in NEB10-beta and wild-type NEB10-beta, which normalizes the J23101 promoter strength to 1 RPU (Andrews et al., 2018).

SR latch circuit assay

Genetic circuits containing SR latches integrated with synthetic NOT gates were characterized as previously described by Lebovich et al. (2023). In brief, strains carrying a circuit construct plasmid and one of the two appropriate output plasmids were grown overnight for 16 h in V-bottom 96-well plates sealed with a breathable membrane in 200 µL M9 medium supplemented with kanamycin and spectinomycin at 1,000 RPM and 37 °C in an ELMI DTS-4 incubator. Overnight cultures were diluted 178-fold (two serial dilutions of 15 µL into 185 µL) and grown for 3 h. Cultures were then diluted by transferring 15 µL into 185 μL of M9 medium, and 3 µL of that dilution was added to 145 µL of M9 medium containing either both inputs (1 mM IPTG and 2 ng/mL anhydrotetracycline, aTc) or no inducers, along with the appropriate antibiotics. Cultures were incubated for 8 h. Afterward, cultures were again diluted by transferring 15 µL into 185 μL of M9 medium, and 3 µL of that dilution was added to 145 µL of M9 medium containing the input to the synthetic NOT gate, either 1 mM IPTG or 2 ng/mL aTc, and grown for an additional 8 h. After each 8-h growth period, a sample from the culture was collected for flow cytometry and diluted into 200 µL PBS (pH 7.4) containing 2 mg/mL kanamycin. After a 30-min incubation at room temperature, flow cytometry was run on a 3-laser BD DUAL LSRFortessa (RRID: SCR_027100) to measure cell fluorescence. Output of the circuit was converted to relative promoter units (RPU) using the pAN1717 RPU standard plasmid (Andrews et al., 2018). Outputs of the circuits were predicted by applying the signal-matching algorithm using the NOT gate response functions as described by Nielsen et al. (2016) and Andrews et al. (2018).

Motility assay

E. coli cultures were grown from a single colony in M9 medium supplemented with kanamycin (50 μg/mL), with shaking in culture tubes at 250 RPM and 37 °C for 16 h. The overnight cultures were diluted to OD 600 of 0.01 in M9 medium with kanamycin and cultured at 37 °C without (–IPTG) or with 1 mM IPTG (+IPTG) and. When the culture reached an OD 600 of 0.5, 100 µL of the culture was spot plated onto M9 plates containing 0.2% agar, kanamycin (50 μg/mL), and +/− 1 mM IPTG. The plates were put in a sealed plastic bag with a wet paper towel and incubated for 24 h at 37 °C. Motility was quantified by measuring the diameter of the halo surrounding the initial inoculation point. When no motility was observed, the diameter of the initial inoculation patch was measured.

mreB repression assay

E. coli cultures were grown overnight in M9 medium supplemented with kanamycin (50 μg/mL), with shaking in culture tubes at 37 °C and 250 RPM for 16 h. Overnight cultures were diluted to an OD 600 of 0.01 in M9 with kanamycin and either with or without 1 mM IPTG. These cultures were incubated at 37 °C and 250 RPM for 5 h before being imaged. Cells were pipetted onto a 1% agarose pad and imaged with a Nikon Eclipse Ti N-SIM E microscope equipped with a Hamamatsu Orca Flash 4.0 camera with a numerical aperture of 1.49 (RRID:SCR_021148 core facility).

Algorithmic design of orthogonal operators and synthetic promoters

To design orthogonal DNA-binding sites for the synthetic repressors, we developed a Python script that performs a random search algorithm and applies user-specified design constraints (Figure 1B). The script allows a set of orthogonal DNA-binding sites to be designed either to complement an existing set of repressors or from scratch. The script can design a set of 20 synthetic operators in approximately 1 minute on a basic laptop (8 GB RAM, Intel i5 CPU). There are 14 user-defined input parameters for the script, which allow specification of the operator sequence length, desired library size, upstream or downstream flanking sequences, GC content, maximum homology between synthetic operator sequences, and maximum homology to the genome and other DNA sequences (Supplementary Note 1). The script can be used to design orthogonal operators and synthetic repressors for any sequenced organism and strain. A FASTA file of the host organism’s genome is a required input file for the script. There is an optional input FASTA file that can contain other genetic part sequences to avoid, including existing regulators, so that the sequences are designed to be orthogonal from all genetic parts that may be used in combination with the synthetic repressors. This script could be used to design orthogonal DNA-binding sequences for different classes of regulators.

Using the script, we generated 19 bp DNA sequences for each synthetic operator. The first (A) and last (T) nucleotides within the −10 and −35 regions, respectively, were held as constant, and a 17 bp variable sequence was designed between them. The TALE synthetic repressor was designed to bind in the reverse orientation of the promoter to maintain a 5′-T for strong DNA binding (Meckler et al., 2013). Here, our aim was to design a set of synthetic repressors that could be used in combination with TetR-family repressors commonly used in genetic circuits (Stanton et al., 2014; Nielsen et al., 2016; Andrews et al., 2018) and augment the set of transcriptional gates available. Therefore, we input the genetic circuit part sequences to the operator design script. The following design constraints were applied to design the library of synthetic operators for this study: GC content (30%–60%), homology to the E. coli K12 DH10B genome (<13 bp), and homology between synthetic operator sequences in the library and specified genetic circuit parts (<9 bp). The BLAST+ application (Camacho et al., 2009) was used to analyze sequence homology (“Methods” and Supplementary Note 1). Using these constraints, a set of up to 4,052 synthetic operator sequences was created after analyzing approximately 27 million random designs. Twenty of the generated operator sequences were chosen at random to design a library of synthetic NOT/NOR gates and corresponding P TALE promoters.

We next built the corresponding set of 20 synthetic P TALE promoters containing the designed operator sequences to serve as the gate output promoters. We used the strong constitutive promoter J23119 (BBa_J23119) as the core promoter sequence and replaced the 17 bp between the −10 and −35 sequences (Figure 1C) with each of the variable operator sequences designed by the algorithm (Supplementary Tables S4, S5). A strong core promoter was chosen with the aim of making the gates inter-compatible with existing TetR-homolog gates (Nielsen et al., 2016) and able to achieve highly activated expression of the genetic circuit’s output. The activities of the synthetic promoters were then measured in relative promoter units (relative to the pAN1717 (Nielsen et al., 2016) RPU standard plasmid) by substituting the promoter in the insulated transcriptional unit for eYFP expression (Figure 1D). The strengths of the P TALE promoters were 1.2–4.8 RPU, while the original J23119 was measured to be 4.16 ± 0.22 RPU (n = 3) (Supplementary Table S6). This confirmed that the synthetic promoters remained strong following the substitution of the variable sequences with the designed synthetic operators.

Hierarchically assembled synthetic repressors strongly repress

A hierarchical cloning scheme was devised to build TALE repressors using Type IIS DNA assembly (Weber et al., 2011) from monomer plasmids that could be used to construct TALEs targeting any 19 bp operator sequence. The TALE monomer part plasmids were cloned into 6-mer intermediate constructs, which were subsequently assembled into full TALE constructs (Supplementary Figure S1; Supplementary Note S2). This approach allows initial prototyping of the synthetic repressors using reusable DNA part plasmids and the construction of a full TALE repressor gene in approximately 6 days. For these initial prototype designs, we utilized the identical ribozyme insulator (RiboJ54) (Nielsen et al., 2016) and TALE N-terminal domain sequence across TALE constructs, enabling reuse of ribosome binding sites (RBSs).

To assess the ability of TALEs to repress the P TALE promoters, four TALEs were tested with the corresponding cognate promoter (P TALE1, P TALE2, P TALE9, or P TALE19). TALEs were constructed to bind the cognate 19-bp operator sequence. Repression was tested by construction of TALE NOT gate characterization plasmids (Supplementary Figure S3). The NOT gates were tested by inducing TALE expression from the P Tac promoter with varying inducer (isopropyl β-D-1-thiogalactopyranoside, IPTG) and measuring eYFP output regulated by the P TALE promoter (Figure 2A). Initial tests showed strong repression of the TALE promoters ranging from 20- to 30-fold repression (Figure 2B). Notably, the response functions exhibit a linear increase in repression with increasing input prior to reaching maximal repression, which is consistent with the known non-cooperative DNA binding of TALEs (Mak et al., 2012).

However, we observed that the ON state for the NOT gates (0 mM IPTG) was substantially less than the unrepressed output of the P TALE promoters (Figure 1D). Therefore, we hypothesized that we could improve the attainable dynamic range of the NOT gates by reducing the RBS strength to limit the effects of leaky basal expression. Weaker RBSs were designed to be fractions of the original RBS strength (approximately 1/2, 1/8, and 1/16) using the RBS Calculator (Salis et al., 2009; Borujeni et al., 2014; Reis and Salis, 2020) and subsequently assayed (Supplementary Table S7; Figure 2C). The strength for each RBS was experimentally assayed using fluorescent fusion proteins comprised of 90 bp of the N-terminus of the TALE gene connected to superfolder GFP (sfGFP) by a flexible linker, and a strong correlation between the measured and predicted RBS strength was observed (r = 0.999) (Supplementary Figure S4). The RBS tuning improved the dynamic range of the NOT gates to ≥50-fold repression and increased the ON state output for all four NOT gates (Figure 2D). The greatest dynamic range observed for each gate was achieved with the lowest-strength RBS (predicted translation initiation rate, TIR, of 500 au).

NOT gates with recoded synthetic repressors achieved similar dynamic range

To limit genetic instability and spurious homologous recombination, we recoded the TALE repressor genes to eliminate any repeated sequence longer than 20 bp across the set of 20 synthetic repressors. The recoding efforts improved the predicted genetic stability of the TALE genes by a median value of 704-fold as measured by the evolutionary failure mode (EFM) calculator (Jack et al., 2015) (Supplementary Figure S5). The average DNA sequence identity between the TALEs was reduced from 98% to an average of 76% (Supplementary Figure S6). The recoding also improved the codon adaptation index (CAI) score of the TALEs by 15% on average (Supplementary Figure S7) (Sharp and Li, 1987). Each TALE transcriptional gate was assigned a unique terminator (Chen et al., 2013) and one of 15 ribozyme insulators (Lou et al., 2012; Nielsen et al., 2016) (Supplementary Table S8). We performed the recoding as a precautionary measure for long-term stability. Interestingly, when we assayed the genetic stability of a cascade circuit containing four TALE NOT gates, we did not detect any large deletions in either the original or recoded versions of the circuit over 12 days of serially passaged cultures in E. coli NEB10-beta (Supplementary Figure S8).

For each of the recoded TALE genes, we designed three unique RBSs predicted to have weak to moderate strength (500, 1,000, and 3,000 TIR). For TALE NOT gates that did not achieve >50-fold dynamic range with these initial RBSs, additional RBSs were designed and tested (up to 7 total). More than half of the recoded TALE NOT gates (out of 84 in total) achieved a dynamic range on par with or better than the original unrecoded TALEs (Figure 3A; Supplementary Table S9). The dynamic range achieved by the selected NOT gates (20 recoded TALE repressors and 26 RBS variants) was >30-fold and averaged 51.9-fold repression, with 11 exhibiting <50-fold repression and only three exhibiting <40-fold repression. The response function for each NOT gate in the library was determined (Supplementary Table S10).

TALE-based synthetic repressors are orthogonal

To measure the orthogonality of the synthetic repressors, all 20 recoded synthetic repressors were experimentally assayed for their ability to repress all 20 P TALE promoters in the library (400 combinations). Plasmids with the synthetic repressors under the control of the P Tac promoter were co-transformed with output plasmids containing one of the P TALE promoters driving expression of the eYFP output cassette and using the initial RBS (3000 TIR) for each (Supplementary Figure S9). The synthetic repressors are highly orthogonal and repress each of their cognate promoters with specificity (Figure 3B). The average fold repression for noncognate P TALE promoters was 1.04 ± 0.1, showing negligible repression. Given that this experiment did not use the final optimized RBS variant for each, we expect the higher dynamic range variants of the synthetic repressors to have equivalent or better orthogonality. This level of orthogonality surpasses that of other naturally derived repressor libraries of equal size, such as TetR-family repressors, likely due to the natural repressors’ common ancestry (Stanton et al., 2014). Smaller libraries of natural repressors have achieved comparable orthogonality, such as a set of 12 phage repressors (Padmakumar et al., 2025), yet this was demonstrated for far fewer combinations tested (144 combinations). The specificity of the TALE repressors was so strong that, in one case, we were able to detect a recoding error that created a single RVD–nucleotide mismatch, which eliminated its cognate P TALE repression activity.

Synthetic repressors have a broad range of toxicity

To assess the growth toxicity of the synthetic repressors on E. coli, we performed growth curves with induced repressor expression and measured the exponential specific growth rate. Strains for the experiment harbored the NOT gate characterization plasmids for the recoded TALEs, and we included six repressors from the TetR family for comparison (Supplementary Figure S10; Supplementary Table S11). Consistent with published results, expression of the TetR-family repressor LitR caused high observed toxicity, with a 23.4% decrease in specific growth rate compared to wildtype NEB10-beta, while others such as BetI (native to E. coli) and BM3RI had low toxicity (0.1% and 5.3% reduction, respectively) (Nielsen et al., 2016). Surprisingly, the synthetic TALE repressors displayed a wide range of toxicity. Some gates exhibited low toxicity, including Syn_rT01 (RBS T1_500, 4% reduction) and Syn_rT19 (RBS T19_1K, 1.5% reduction), while others exhibited higher toxicity, up to 22.6% slower growth rate for Syn_rT02 (RBS 20K) and Syn_rT03 (RBS T3_1K). For the TALE-based gates, a growth rate reduction of 14.1% ± 5.9% was observed for the 26 gates, while 11 gates showed toxicity on par or better than the TetR-family repressors tested. No significant difference was found in toxicity between the expression of the TALE and TetR-family repressors (Welch’s two-sided t-test, p = 0.130). We investigated possible explanations for the observed toxicity of the synthetic repressors. However, the observed toxicity did not correlate to the RBS strength, gate dynamic range, or basal OFF state output (r = 0.05, r = 0.08, and r = 0.2, respectively) (Supplementary Figure S11). Given the variation in TALE repressor toxicity, an additional screening step in building future synthetic repressor libraries could be applied to screen out the most toxic repressors and may be able to be applied prior to recoding.

Synthetic gates can be used with TetR-family repressors in genetic circuits

We chose to integrate the new synthetic gates with previously reported set-reset (SR) latches (Andrews et al., 2018) constructed from TetR-family NOR gates to test their composability (Supplementary Figure S12). We chose five SR latches to combine with synthetic TALE signal inverters, generating ten different sequential logic circuits. For each SR latch, two circuit designs were made, with either signal feeding into the latch being inverted by a different chosen synthetic NOT gate (Figure 4A). Except for two circuit designs (SR3 and SR4), which contained the toxic repressor QacR, the other eight sequential logic circuits performed as expected, having close agreement with the predicted outputs from signal matching for each condition and holding their states for 8 h after being transferred to input-neutral media (Figure 4B). The toxicity of SR3 and SR4 illustrates the importance of accounting for compounded toxicity of multiple repressors, as has been reported using toxicity score criteria (Nielsen et al., 2016). Syn_rT06 and Syn_rT15 were not the most toxic synthetic gates used (18% and 11% reduced growth rate, respectively); however, in combination with QacR, they were sufficiently toxic to hinder the experiments because of low cell density. Overall, these results indicate the utility of synthetic repressors in expanding the number of repressors available for use in genetic circuits in conjunction with natural repressors.

Synthetic repressors can be designed to regulate native gene expression

Outside of genetic circuits, synthetic repressors can be used to regulate the expression of endogenous genes to control phenotypic changes. Using our library of TALE RVD monomers for the hierarchical TALE assembly, we next built synthetic repressors to test the repression of endogenous gene targets. The first target, mreB, is a cytoskeletal gene responsible for coordinating cell-wall synthesis. When MreB is depleted in E. coli, its morphology changes from rod-shaped to spherical (Kruse et al., 2005). The second target, cheZ, is a chemotaxis regulator located at the end of an operon. CheZ-deficient mutants have a tumble-only phenotype that eliminates their ability to chemotax (Parkinson et al., 1983).

We built three TALEs, each targeting one of the three promoters upstream of mreB (P mreB1, P mreB2, and P mreB3) (Wachi et al., 2006) (Supplementary Figure S13A). To quantify their repression, we cloned the whole promoter region of mreB into an output plasmid upstream of eYFP. We expressed each TALE individually and all three simultaneously (Supplementary Figure S14). Only the TALE targeting P mreB3, located closest to mreB, had a significant effect on promoter output, eliciting five-fold repression (Supplementary Figure S13B). We also observed a shift to spherical cell morphology when mreB was repressed (Supplementary Figure S13C). To repress cheZ expression, we built two TALEs that bind within the first 53 bp of the cheZ coding sequence to utilize RNA polymerase roadblocking for repression (Supplementary Figures S13D, S15). To test for cheZ repression and the associated chemotactic defect, we performed a soft agar motility plate assay. Consistent with successful cheZ repression, the TALE-expressing strain had reduced motility after 24 h (Supplementary Figures S13E-F). However, there were signs of spotty chemotaxis after 72 h, and subsequent streak purification and diagnostic restriction digests revealed deletion mutations of the cheZ -targeting TALEs in the isolated escapee colonies. This observation supports the conclusion that repetitive, hierarchically constructed TALEs are prone to deletion mutations in vivo, yet these mutations did not arise in our 12-day liquid culture experiment (Supplementary Figure S8).

This study demonstrates the use of TALE-based synthetic repressors to expand the library of available transcriptional logic gates for genetic circuits in E. coli. While TALEs are large genes, we used reusable monomer plasmids and hierarchical DNA assembly to facilitate the prototyping of synthetic repressors prior to recoding and chemical DNA synthesis. For future studies, the results presented here suggest that the repressors could be recoded from the outset and skip the prototyping step. We developed a systematic approach for designing, building, and testing synthetic repressor logic gates that can guide additional library expansion or library customization for a different chassis. This includes a script for synthetic operator design that allows a user to generate orthogonal DNA-binding sequences while avoiding off-target genomic binding for the user-specified organism and genetic parts. This strategy was used to develop 26 transcriptional inverter gates using 20 recoded TALE repressors and their cognate synthetic P TALE promoters. The synthetic NOT gates showed a suitable dynamic range (30–70-fold) and were functional in genetic circuits when used in combination with TetR-family repressor transcriptional gates and a signal-matching algorithm. The synthetic repressors exhibited high specificity for their cognate promoter using the designed synthetic operators. The growth toxicity of the synthetic TALE repressors was determined to be comparable to, though sometimes higher than, TetR-family repressors, yet with variable TALE toxicity observed. This observation suggests that future uses may benefit from an expanded starting pool of candidate synthetic gates and an early screening step to remove highly toxic repressors. We also demonstrated the potential uses of synthetic TALE repressors to regulate native gene expression by targeting an endogenous promoter or coding sequence in E. coli. The design of orthogonal synthetic operators and the expansion of available repressor proteins for a specified bacterial strain are expected to improve the utility of genetic circuits in diverse prokaryotes for future applications.