Overall Design
Our aim is to design an in vivo, long-term effective, and multi-potent probiotic platform, which functions to detect, treat, and record feline inflammatory bowel disease (IBD). Guided by synthetic biology principles, our overall design integrates sensing, processing, and output modules to achieve a therapeutic platform to promote feline gut health.
We consider food input as an overall switch to probiotic functioning, which is achieved by implementing a bile acid biosensor, in order to lower the metabolic burden of our probiotics. Targeting feline IBD, we make the IBD biomarker another key input of our platform through a calprotectin biosensor. Those two inputs are integrated by an AND logic gate in the information processing module. Once both inputs are detected, our platform will start generating the therapeutic and visual outputs. Those include producing butyrate as an anti-inflammatory molecule, degrading hairballs by expressing and anchoring keratinase, and expressing visual color signals to indicate disease conditions to cat owners. Further, to trace disease progression for more individualized treatment in the future, we incorporated the Retro-Cascorder system to record the activities of promoters-of-interest over time, and to store the information in the bacterial genome. (Figure 1)
Figure 1. Overview of our project design.
Information Input Module
Food Input Detection
Certain bacteria sense bile acids as signals for environmental adaptation. To build a probiotic platform responsive to TCA in the feline gut, we explored two different TCA sensors: one based on the TetR-family repressor CmeR from Campylobacter jejuni, and the other based on the transmembrane sensor TcpP from Vibrio cholerae.
The CmeR TCA biosensor system contains CmeR repressor and the regulated promoter pCmeR. In the context of C. jejuni genome, CmeR is constitutively expressed, binds to the repeated motifs in pCmeABC in the absence of inducers (Figure 2A), and regulates the transcription of downstream cmeABC genes, which collectively express the efflux pump that transports antibiotic substances out of the cell. In the presence of cytoplasmic antimicrobial bile salts, CmeR dissociates with pCmeABC and activates the expression of CmeABC (Figure 2B, C); TCA, especially, increases CmeABC expression by about 16-fold. (Lin et al., 2005) TCA's capability to bind to the allosteric site of CmeR was also verified in vitro, in which this binding was characterized with a dissociation constant of 1.5 μM and an enthalpic contribution of -59 kcal/mol (Lei et al., 2011). In previous characterization of CmeR in E. coli, it has shown inducibility by salicylate (8-fold) and other 21 molecules, but the response to TCA has never been tested in E. coli (Nasr et al., 2022).
Figure 2. The CmeR structure predicted by Alphafold 3. A, the homodimer structure (chain A: cyan; chain B: green) of CmeR binds to the cmeO DNA region. B, the ligand-bound structure of CmeR homodimer (chain A: cyan; chain B: green) when bound with TCA (pointed by red arrows). C, structural alignment of CmeR DNA-bound structure (green) and ligand-bound structure (cyan); a structural shift was observed at the helix-turn-helix DNA binding domain (red circled).
In our plasmid design, we placed CmeR downstream of a constitutive promoter J23119(mut), and placed sfGFP as a reporter gene downstream of pCmeR. The design of pCmeR upstream to sfGFP has undergone additional rounds of engineering. First, a prototype of pCmeR was designed by using pQacR (BBa_J428050) as the template and then replacing qacO with cmeO (Nasr et al., 2022). Since the relative positioning of cmeO with -35 and -10 hexamers can greatly affect the interaction between transcription factor, CmeR repressor, and promoter region, we designed and tested multiple variants with different cmeO positioning, including downstream to -10 hexamer, overlapping with -35 hexamer, and within the spacer between the two hexamers (more details are described in our Engineering Success section). (Figure 3) Although we eventually concluded that placing cmeO within the spacer between -35 and -10 hexamers was optimal compared to the other two designs, we also identified a fatal flaw of this design, that TCA has low permeability to the plasma membrane of E. coli (Elkins & Savage, 2003) and therefore is unable to serve as a cytoplasmic ligand of CmeR.
Figure 3. Plasmid designs for CmeR-based TCA sensor. Over the course of engineering and optimization, we tried different promoters upstream to CmeR [J23119 and J23119(mut)], pCmeR promoter with different cmeO placing, and different RBS upstream to GFP [B0031 and RBS(Bgl)]. For constitutive promoters and RBS, darker red color represents higher transcriptional or translational activity. Created by biorender.com.
This flaw drove our design of the second TCA sensor. The transmembrane sensor TcpP is responsible for bile salt sensing in V. cholerae; in the presence of ligands, TcpP dimerizes and acts as a transcriptional factor, therefore activating downstream gene expression (Xue et al., 2016). Its subcellular localization (on the membrane with its ligand-binding domain oriented towards the extracellular side) makes it compatible with our need for the TCA sensor and addresses the problem of low membrane permeability of TCA. However, TcpP is rapidly degraded through proteolysis by YaeL in V. cholerae without the protection of TcpH (Yang et al., 2013), and similar degradation takes place in E. coli through proteolysis by a RseP homolog (Chang et al, 2021). Therefore, it is also important to incorporate TcpH into our design.
Above TcpP-TcpH TCA sensor has shown high compatibility to E. coli after being integrated into a previously well-characterized, widely-applicable synthetic receptor platform called EMeRALD (short for Engineered Modularized Receptors Activated via Ligand-induced Dimerization). In the design of this TcpPH-EMeRALD system, since the transcriptional factor in V. cholerae might not be compatible with promoters and RNA polymerases in E. coli, the DNA-binding domain of the native TcpP was replaced by the DNA-binding domain of CadC, a membrane-bound pH sensor in E. coli that regulates the promoter pCadBA (Lindner & White, 2014). The fusion protein CadC-TcpP, composed of the ligand-binding and transmembrane domains of TcpP and the DNA-binding domain of CadC, keeps the function of both as a TCA sensor and as a transcriptional factor endogenous to E. coli. This TcpPH-EMeRALD system demonstrated its ability to sense TCA with high sensitivity (with the EC50 of 28.344 μM) and decent dynamic range (84.92-fold) in E. coli. (Chang et al, 2021)
In our plasmid design, we put CadC-TcpP and TcpH downstream to two different constitutive promoters, respectively, P9 and apFAB339, because the relative expression levels of CadC-TcpP and TcpH proteins under this combination of promoters were tested to show the optimal sensing performance (Chang et al, 2021). We placed sfGFP as a reporter protein downstream to pCadBA, which will be expressed upon binding with CadC DNA-binding domain in the presence of TCA. (Figure 4)
Figure 4. Plasmid design of TcpPH-based TCA sensor. Created by biorender.com.
IBD Biomarker Detection
Calprotectin is a heterodimeric protein that shows antimicrobial activity by inducing metal starvation in bacteria. It is mainly secreted by neutrophils and monocytes and has been widely recognized as a clinical biomarker for inflammatory bowel disease (IBD) (Xia et al., 2023). Elevated concentrations of calprotectin have also been observed in the feline gastrointestinal tract for cats with IBD conditions (Riggers et al., 2023), making it a valuable indicator of feline IBD.
Building on this property, we constructed a calprotectin-responsive biosensor that utilizes its zinc-chelation mechanism. Our design incorporates the endogenous zinc uptake regulator (Zur) of Escherichia coli Nissle 1917 (EcN) together with its corresponding metal-sensitive promoter. Among the four candidate promoters tested by Xia et al. (2023), ykgMO consistently outperformed the others. It exhibited the strongest GFP signal with 21.2-fold induction in minimal medium (M9), and shows the strongest dynamic range of 16.9-fold in complex medium (LB), whereas other promoters showed markedly reduced zinc sensitivity in LB medium. (Xia et al., 2023) Such robustness across different environments is particularly important, as the intestinal tract is highly complex. In addition to the ykgMO promoter's performance, the incorporation of Zur provides further advantages. This sensing regulator is native to EcN, making it more likely to remain stable and functional in our design. Together, these features make the ykgMO-Zur module promising for reliable calprotectin detection in the feline gut.
Mechanistically, free zinc ions normally bind to Zur (Figure 5A), enabling it to repress transcription by occupying the intergenic space (IGS) upstream of the ykgMO promoter. However, when calprotectin is present, it competes for zinc (Figure 5B) and sequesters it away from Zur. This competitive interaction disables Zur-mediated repression, thereby activating downstream gene expression from the ykgMO promoter.
Figure 5. The zinc-bound structures of Zur and calprotectin predicted by Alphafold 3. A, Zur. B, calprotectin.
The ykgMO promoter and its upstream IGS were previously uploaded by the iGEM team Princeton 2024 (BBa_K5180004). However, they only modeled the promoter’s performance and did not conduct experimental characterization. Building on their design, we cloned the ykgMO promoter together with its IGS into the pUC57-Kan backbone. To enable experimental characterization, sfGFP was placed downstream of the ykgMO promoter as a reporter. (Figure 6A) However, the endogenous expression level of Zur might be insufficient to fully repress the ykgMO promoter, which potentially leads to high leaky expression. To address this, previous studies have shown that supplementing Zur expression in plasmids can reduce basal expression by around 100-fold, while maintaining similar expression upon induction. Among various constitutive promoters tested, expressing Zur downstream to J23109 (BBa_J23109) achieved the highest dynamic range. (Zhu et al., 2025) Therefore, we designed and constructed another plasmid placing Zur downstream to J23109, in addition to the ykgMO-regulated sfGFP expression. (Figure 6B) Those calprotectin-responsive sensors were further validated using kinetic assays.
Figure 6. Plasmid designs of ykgMO-based calprotectin sensor. A, ykgMO-GFP without supplementing Zur expression in plasmid. B, ykgMO-GFP that supplements Zur expression by placing it downstream to J23109 promoter. Created by biorender.com.
However, calprotectin is expensive and requires a complicated production process. Therefore, we employed TPEN (N,N,N′,N′-tetrakis-2-pyridylmethyl-ethylenediamine) as a functional substitute, following previous studies (Zhu et al., 2025). With comparable zinc-binding properties, TPEN serves as a high-affinity chelator, offering six potential coordination atoms for zinc ions. This makes it a practical replacement for calprotectin in characterizing our system.
Information processing module
The AND gate constitutes a fundamental component of our platform's logic design. Specifically, we employed an AND gate to collectively monitor food intake, represented by the bile acid taurocholic acid (TCA), and intestinal inflammation, indicated by elevated calprotectin levels. This dual-input system ensures that therapeutic responses and visual outputs are triggered only under specific pathological conditions following food consumption (Figure 7), while preventing false activation in healthy states and avoiding unnecessary metabolic burden on the probiotic chassis.
Figure 7. Overall design of AND gate. Created by biorender.com.
To apply this logic gate, we explored two distinct architectures acting at different regulatory layers. The first is a promoter-repressor transcriptional system, which combines transcriptional control elements to enforce dual-input dependence (Stanton et al., 2014). The second is a split T7 RNA polymerase (RNAP)-based protein-level system, in which functional RNAP activity is restored only when both inputs are present (Schaerli et al., 2014).
Investigating both strategies provides alternative pathways toward robust circuit behavior: while transcriptional systems offer modularity and direct coupling to promoters, post-translational systems have the potential to expand dynamic range and minimize leaky expression. By comparing these approaches, we aim to identify a design that balances sensitivity, dynamic range, and metabolic efficiency, thereby ensuring reliable logic execution in a complex gut environment.
pAND system
The first AND gate we examined was the pAND plasmid originally developed by Stanton et al. (2014). This design relies on the modular combination of multiple prokaryotic transcriptional repressor-promoter pairs, enabling the construction of composite logic gates in E. coli. For pAND, orthogonal transcriptional NOT gates (built from repressors PhIF and QacR) are used to invert the two input promoters, and the inverted outputs are combined into a NOR gate (built from the repressor BetI) that controls the downstream reporter. This combined logic implements the AND logic gate. Thus, only when two input signals are both present, will the reporter be expressed. (Figure 8)
Figure 8. Schematic and the truth table of pAND. Created by biorender.com.
For the circuit design, the constitutively expressed LacI repressor inhibits the Tac promoter, and the TetR repressor inhibits the Tet promoter, with the PhIF and QacR coding sequences placed downstream of the pTac and pTet promoters, respectively. Their cognate promoters, pPhlF and pQacR, are arranged in series to drive BetI expression, which in turn represses pBetI and thereby controls the expression of reporter gene downstream of pBetI (Figure 9). Upon addition of IPTG and aTc, LacI and TetR are inactivated, releasing pTac and pTet and enabling production of PhlF and QacR. These repressors then bind to their targets upstream of betI, suppressing its transcription. As a result, under double induction, BetI is not expressed, pBetI repression is elevated, and fluorescent protein is expressed.
Figure 9. Plasmid design of pAND. Created by biorender.com.
In the original plasmid design, YFP was used as the reporter, yielding a 4.4-fold response between the ON state (+/+) and the strongest OFF state (Stanton et al., 2014). In our construct, YFP was replaced with sfGFP to standardize measurement with other components of our project intended for kinetic assays, and to reduce background interference caused by the spectral overlap between YFP and the intrinsic coloration of bacterial cultures. The performance of the pAND circuit was subsequently evaluated through both endpoint qualitative characterization and kinetic analysis.
However, this design has potential limitations: the process of transcriptional layering can result in time delays, and the relatively modest dynamic range may hinder signal discrimination in noisy environments such as the gut. Therefore, an alternative AND gate system was also incorporated into our design.
Split T7 RNAP system
The second AND gate we evaluated was the split T7 RNA polymerase (RNAP) system originally developed by Schaerli et al. (2014). In this design, the T7 RNAP is divided into N- and C-terminal fragments, each fused to an intein domain. When both fragments are expressed, the inteins mediate protein splicing, reconstituting a functional T7 RNAP. Because enzymatic activity arises only upon the co-expression and assembly of both protein fragments, this is considered a post-translational AND gate strategy. The reconstituted polymerase then drives transcription from a T7 promoter, thereby enabling a protein-interaction-based control of transcriptional activity. Compared with pAND, the split T7 RNAP system offers an expanded dynamic range and lower leakage. Most importantly, the system was implemented in probiotic E. coli strain Nissle 1917, which lacks endogenous T7 RNAP. This ensures orthogonal transcription, with output occurring exclusively through the split T7 RNAP system, further minimizing basal expression and making it a promising AND gate candidate.
Figure 10. Plasmid designs of split T7 system with different RBS. Created by biorender.com.
For plasmid construction, two compatible vectors with p15A and ColE1 origins of replication were employed. On the p15A plasmid, LacI is constitutively expressed to repress the pTac promoter, while AraC functions as a dual regulator that activates the pBAD promoter in the presence of arabinose. Downstream of pTac, the N-terminal fragment of T7 RNAP (amino acid residues 1-514) is fused to a UmuD degradation tag to reduce metabolic burden. This fragment also carries the Npu intein (NpuN, 102 aa) from Nostoc punctiforme dnaE, containing the L25S mutation to enhance splicing efficiency at 37 °C in E. Coli. Downstream of pBAD, the C-terminal fragment of T7 RNAP (amino acid residues 515-884) is fused to the Ssp intein (SspC, 36 aa) from Synechocystis sp. PCC6803 dnaE, likewise optimized with a P21R mutation. (Schaerli et al., 2014); Figure 10)
When both inducers are present, IPTG binds LacI to derepress pTac, while arabinose binds AraC to activate pBAD, enabling simultaneous expression of the two T7 RNAP fragments. Through Npu–Ssp intein-mediated trans-splicing, the fragments are covalently ligated to generate a full-length T7 RNAP, restoring enzymatic activity. The reconstituted enzyme then activates the T7 promoter located on the ColE1 plasmid and drives sfGFP expression (Figure 10). Previous study reported a 57-fold increase in fluorescence upon co-induction with IPTG and arabinose, using a system with a mutant pT7(−3G) promoter that retains only ~20% of wildtype activity (Schaerli et al., 2014). Given that our design employed the consensus T7 promoter, an even stronger induction was expected. The functionality of our split T7 RNAP system was evaluated by both qualitative and quantitative assays (more details are described in our Results section).
To minimize leaky expression, we constructed two sets of plasmids for the RNAP fragments and two sets for the sfGFP reporter, each carrying ribosome binding sites (RBS) of different strengths (Figure 10). We then predicted the translation initiation rates of different RBS using the RBS Calculator algorithm of De Novo DNA (LaFleur et al., 2022; Salis et al., 2009). For the RNAP fragments, we compared a strong RBS (g10-L) with a weaker one (B0032) (Figure 11A-D). For the sfGFP reporter, the same RBS (B0034) was used but positioned at varying distances from the promoter and start codon, resulting in different strengths, with the shorter spacing producing a higher translation rate (Figure 11E-F). We then cross-combined these variants into a matrix and identified the optimal combination.
Figure 11. Predicted translation rates for six RBS variants. Red circles highlight the correct translation peaks corresponding to the N-terminal RNAP fragment, C-terminal RNAP fragment, or full-length sfGFP. A-B, the predicted translation rates of two T7 RNAP fragments under the RBS g10-L. C-D, the predicted translation rates of two T7 RNAP fragments under the RBS B0032. E, the predicted translation rate of sfGFP under the RBS B0034(strong). F, the predicted translation rate of sfGFP under the RBS B0034(weak).
Further, after identifying the optimal RBS combination as B0032 + B0034(strong), we cloned the two RBS into a single plasmid (Figure 12). This design increases genetic stability and, by placing all regulatory elements on one vector, enables more precise control over the spacing of promoters, RBSs, and ORFs. Together, these strategies ensure a stringent AND gate with low basal expression and high dynamic range in probiotic E. coli Nissle 1917, providing a platform for signal processing.
Figure 12. Plasmid design of split T7 system with optimal RBS combination. Created by biorender.com.
Functional Output Module
Anti-inflammatory Production
Short-chain fatty acids (SCFAs), such as butyrate, are critical for maintaining a healthy gut environment. Butyrate is of particular interest due to its anti-inflammatory properties, which have shown promise in treating inflammatory bowel disease (IBD). While research has primarily focused on human applications, we believe that butyrate can also be effective in the feline gut, given the similarities in gut physiology. A study by Zhang et al. (2025) demonstrated that supplementing a cat's diet with sodium butyrate improved feline IBD, suggesting its potential for promoting gut health and reducing inflammation in cats.
Various bacteria are able to produce butyrate. Referring to the design of BNDS-China 2024, we utilize the Tes4 enzyme (BBa_K3838613), which is a bacterial acyl-ACP thioesterase derived from Bacteroides fragilis, for butyrate synthesis. Previous iGEM teams (including SZU-China 2021, HZAU-China 2022, and BNDS-China 2024) have characterized this enzyme, revealing that expression of Tes4 in E. coli strain BL21 leads to the production of butyrate. The metabolic pathway for butyrate production initiates with the formation of butyryl-ACP from glucose through the native fatty acid biosynthesis pathway (FASII) in E. coli (Figure 13). This is subsequently followed by the conversion of butyryl-ACP into butyrate via the action of Tes4 (Kallio P. et al., 2014).
Figure 13. Tes4 butyrate production pathway.
Keratin Degradation
Keratinases are proteases that can degrade keratin. They are found across diverse organisms, and their subtle structural variations can result in significant functional differences. For our proposed application in the feline gastrointestinal tract, the keratinase in our final implementation should exhibit high keratinolytic activity specific to α-keratin (the type of keratin that composes most of the cat hairball), operate within pH and temperature ranges compatible with the feline gut environment, and avoid eliciting immune responses or damaging the intestinal epithelium (Sato et al., 2019). To achieve those, we selected four keratinases based on their stability, keratinolytic activity, substrate specificity, and safety profiles. We plan to further optimize the most promising candidate in the future, aiming to generate a single enzyme well-suited for our project. Furthermore, the keratinase will be displayed on the cell surface via fusion with a truncated ice nucleation protein (INPNC) to ensure its extracellular degradation.
The first enzyme, KrMKU3 (BBa_K3895005), originally isolated from Bacillus licheniformis MKU3, was previously characterized in E. coli by Radha et al. and by the iGEM team SZ-SHD 2021, in which both qualitative and quantitative assays confirmed its keratin-degrading activity. However, the article noted that the presence of a signal peptide may interfere with the active site and reduce activity. Specifically, expression of KrMKU3 in E. coli with its native signal peptide (pETK3) showed no measurable activity, whereas removal of the signal peptide (pETproK3) restored and even enhanced keratinolytic activity to 74.32 ± 1.13 U/mL, corresponding to 135% of the activity of the native B. licheniformis MKU3 enzyme (55.0 ± 1.01 U/mL). (Radha et al., 2007) Thus, we predicted the signal peptide cleavage site of KrMKU3 using SignalP and constructed a signal peptide-removed version. Since similar issues may occur in other keratinases, we predicted the signal peptide for all keratinases (Figure 14A, D, G, J). For KrLCB12 and KrOcL9, which had not yet been expressed in E. coli, we constructed both native and signal peptide-removed versions. In contrast, although KrUS575 is also predicted to contain a signal peptide, previous studies have successfully expressed it in E. coli with its native signal peptide, so we only expressed its native variant. For all constructs, structural predictions were performed using AlphaFold3 to aid in subsequent analysis (Figure 14B-C, E-F, H-I, K-L). This design further facilitates the subsequent fusion with the INPNC cell surface display tag. We were then able to compare the enzyme activities of these variants across different keratinases, as well as between the native and signal peptide-removed versions.
Figure 14. Signal peptide prediction and structural modeling of keratinases. (A, D, G, J), Signal peptide prediction results obtained from SignalP 6.0. The green vertical dotted line indicates the predicted signal peptide cleavage site. (B, E, H, K), AlphaFold3-predicted protein structures generated using the full-length amino acid sequences, including the predicted signal peptides. (C, F, I, L), AlphaFold3-predicted protein structures generated using sequences with the signal peptide regions removed. Four keratinases are shown: KrMKU3 (A-C), KrLCB12 (D-F), KrOcL9 (G-I), and KrUS575 (J-L). Created by biorender.com.
We next considered the second enzyme, KrLCB12, which was originally derived from Bacillus sp. LCB12. This keratinase is specific to α-keratin, the major keratin type in cat fur, and its degradation ability was qualitatively confirmed in the original study using goatskins, which are also mainly composed of α-keratin and have a similar composition to cat hairballs. The same study also completes quantitative assays, which further demonstrated its high keratin-degrading activity: the fermentation supernatant exhibited a keratinase activity of 1332 U/mL (with a caseinase activity of 3379 U/mL), while the purified enzyme reached a specific activity of 9813.2 U/mg. (Tian et al., 2019) These results highlight its strong potential for achieving efficient α-keratin degradation in our platform. Although KrLCB12 has not yet been expressed in E. coli, its successful expression and characterization in other prokaryotic hosts suggest its compatibility with the prokaryotic expression system, leading to a high likelihood of successful expression in E. coli. Based on this, we selected KrLCB12 for further characterization to evaluate its efficacy in hairball degradation.
The third keratinase, KrOcL9, was identified from Bacillus sp. OcL9. This enzyme was chosen because it showed no detectable activity against collagen types I and II, indicating that it would not damage the feline gut interior, thereby ensuring efficient hairball degradation with minimal side effects. Additionally, it was shown to effectively degrade α-keratin, as also confirmed through assays on goatskins. The purified enzyme exhibited a specific activity of 85565.21 U/mg at its optimal conditions (75 °C, pH 8). Although its relative activity toward keratin was only 12.03% and decreased further at 40 °C, the enzyme still maintained a considerable activity level. (Li et al., 2022) While KrOcL9 has not yet been expressed in E. coli, it has been successfully expressed and characterized in prokaryotic Bacillus species, suggesting its feasibility for prokaryotic expression.
We selected KrUS575 as the fourth keratinase. This enzyme, sourced from Brevibacillus brevis US575, has been successfully expressed in E. coli and exhibits optimal activity at pH 8 and 40 °C, conditions that closely match the environment of the feline gut. Under these conditions, KrUS575 also demonstrates high substrate specificity and catalytic efficiency. When rabbit fur was used as the substrate, the enzyme showed a keratinolytic activity of 5500 U/mL, corresponding to a relative degradation efficiency of 86%; with goat hair as the substrate, its activity reached 3135 U/mL, with a relative degradation efficiency of 77%. (Jaouadi et al., 2013) Both substrates are rich in α-keratin and structurally similar to cat fur. Based on these optimal conditions and relative efficiencies on α-keratin-rich substrates, we expected KrUS575 to perform effectively in our system.
The coding sequences of four keratinases after codon optimization, KrMKU3 (BBa_25J6XNWA), KrLCB12 (BBa_25JQNEX5), KrOcL9 (BBa_25VB4UDR), and KrUS575 (BBa_25EY32UM), were separately synthesized by Tsingke in the pET-28a(+) vector, with both N- and C-terminal 6×His tags flanking each keratinase. The vector carries a kanamycin resistance gene for colony selection and the lacI repressor for IPTG induction. We also constructed signal peptide-removed versions of KrMKU3 (BBa_25WHQJ10), KrLCB12 (BBa_2597O3NS), and KrOcL9 (BBa_25C6M1U8) via cloning. All constructs were assembled in E. coli Trelief 5α and subsequently expressed and characterized in E. coli BL21(DE3). (Figure 15) The recombinant keratinases were purified by Ni-NTA affinity chromatography, and the purified proteins were used for subsequent quantitative and qualitative characterization.
Figure 15. Plasmid design of 7 plasmids that express different keratinases with or without the native signal peptide. Created by biorender.com.
Since E. coli cannot uptake hairballs into the cytosol, the keratinase can only degrade cat hairballs when exposed to the feline gut. We therefore incorporate the INPNC surface display system to anchor the expressed keratinase onto the extracellular side of the cell membrane. The INPNC tag (BBa_K811003), previously characterized by the iGEM team Penn 2012, is a truncated version of the iced nucleation protein (INP), where the N- and C-termini are fused to facilitate its localization at the cell surface. (Figure 16) To test the functionality of this surface display system, we fused the INPNC tag with sfGFP via a flexible GS linker uploaded by the iGEM team ELTE 2022 (BBa_K4375010), which was then cloned into the pET28a(+) vector. The 13 aa-long GS linker attaches sfGFP to the outward-extending C-terminus of INPNC, directing the movement of sfGFP toward the cell surface. Comparing the localization of green fluorescence between this plasmid and another plasmid that expresses sfGFP without INPNC fusion, we could qualitatively test whether the INPNC tag successfully transports the linked protein to the extracellular side and functions as expected.
Figure 16. INPNC prediction and structural modeling. A, AlphaFold3-predicted INPNC structure. B, AlphaFold3-predicted INPNC structure anchored within the phospholipid bilayer.
Color Visual Display
We want to alert the owner when their pet is in a diseased condition. Thus, we designed our probiotic to express fluorescent proteins and chromoproteins upon sensing IBD biomarkers, creating a visual cue in cat feces.
To ensure the color report has the best visibility, we tested fluorescent proteins and chromoproteins of a wide range of colors. We employed the constitutive promoter J23100 to initiate their expression, ensuring continuous production of visible signals. Through cloning of the proteins and by introducing Golden Gate overhangs, 9 chromoproteins and 25 fluorescent proteins cloned from the iGEM distribution kit were converted into switchable sequences, providing flexibility for downstream applications. These constructs were transformed into E. coli Trelief 5α for further characterization. (Figure 17)
Figure 17. Plasmid designs of the color visual display module and the proteins involved. A, plasmid design of color signal module. B, proteins involved. Created by biorender.com.
In order to evaluate their practical performance, E. coli with color signals were then mixed with curry powder and soy sauce to simulate the appearance of feline feces. Their color performance was assessed under both ultraviolet light and natural sunlight to determine the visibility of the signals. Through this comparative screening process, we selected the most distinguishable and reliable color outputs, which formed the foundation of our diagnostic reporting module.
Data Storage Module
The Basic Design
To record and use time-based data of feline gut health, we incorporated the transcriptional recording system called Retro-Cascorder into our platform. It is able to chronologically record IBD-related transcriptional events of promoters-of-interest.
Specifically, Retro-Cascorder harnesses the Cas1-Cas2 adaptation machinery together with a retron module to capture transcriptional events in real time (Figure 18A). When a promoter of interest is activated, it drives transcription of a non-coding RNA (retron-ncRNA) containing specific barcodes, which is then reverse-transcribed by a constitutively expressed retron reverse transcriptase (retron-RT) into complementary DNA. These DNA fragments are subsequently integrated into the CRISPR array as new spacers. Besides retron-derived spacers, Cas1-Cas2 also incorporates spacers derived from other double-stranded DNA sources such as the host genome or plasmids. These non-retron-derived spacers, collectively referred to "N," accumulate alongside retron-derived spacers during the course of recording. Thus, each array can contain three types of spacers: retron-derived spacers corresponding to signal A or B, and background spacers denoted as N. Over time, the sequential acquisition of A, B, and N spacers forms a chronological record within the CRISPR array, which can be decoded by sequencing to reconstruct the order of transcriptional events.
This recording design was implemented by two plasmids that together enable the generation of retron-derived DNA barcodes and their integration into the CRISPR array (Figure 18B-C). The first plasmid (BBa_25X8951T) follows the design of Bhattarai-Kline et al. (2022) and serves as the reverse transcription-integration system. It carries the Cas1-Cas2 adaptation machinery along with the Eco1 retron reverse transcriptase (RT), where Cas1 and Cas2 expression is controlled by a LacI-regulated T7 promoter and the Eco1 RT is constitutively expressed by the J23115 promoter. The T7 RNA Polymerase (RNAP) is expressed in E. coli BL21(AI) genome, which is induced by arabinose. Therefore, both arabinose and IPTG are required to activate the integration machinery by turning on Cas1 and Cas2 expression. (Figure 18B)
The second plasmid (BBa_253V08M1) provides the signal-responsive retron ncRNAs, in which barcode A and barcode B are placed under the control of inducible promoters responsive to aTc and choline, respectively. The two expression cassettes are placed in adverse directions to reduce the disturbance at the intersection. Their expression is regulated by TetR and BetI, enabling external stimuli to selectively trigger transcription of the corresponding retron ncRNAs. Together, these plasmids couple input signals to retron DNA barcode production and subsequent Cas1-Cas2 integration, thereby realizing Retro-Cascorder’s ability to encode transcriptional histories directly into CRISPR arrays. (Figure 18C)
Figure 18. Principles of Retro-Cascorder transcriptional recording. A, Upon activation of a promoter of interest by signal A or B, the downstream retron non-coding RNA (RT-ncRNA) would be transcribed, which is reverse-transcribed by retron reverse transcriptase (RT, shown in blue) into complementary DNA. The Cas1-Cas2 integrase complex (yellow and brown) then integrates DNA sequentially into the CRISPR array as spacer A (green) or B (purple). Meanwhile, double-stranded DNA from other sources like genome or plasmids could also be incorporated by Cas1-Cas2 as spacer N (yellow). Over time, different signals are captured as new spacers, creating a chronological transcriptional history that can be read out by sequencing. B, The plasmid expressing Cas1, Cas2, and Eco1 retron-specific RT. The LacI-regulated T7 promoter drives the expression of Cas1 and Cas2, while Eco1 RT is constitutively expressed by the J23115 promoter. C, Retron ncRNA encoding barcode A and B were expressed by aTc- or Cho-inducible promoters regulated by TetR and BetI. Created by biorender.com.
The overall workflow of our design proceeds as follows: (1) activation of an inducible promoter on the first plasmid drives transcription of a barcode-tagged retron-ncRNA; (2) the Eco1 retron RT on the second plasmid reverse-transcribes this ncRNA into DNA; (3) the Cas1-Cas2 complex captures the resulting DNA fragments and integrates them into the CRISPR array as new spacers. Then, the genome is extracted and sequenced, from which the sequencing reads are analyzed to identify retron-derived barcodes, and reconstruct the temporal order of promoter activities. See more details about the experimental workflow and procedures in our Measurement section.
Cas1 Enzyme Optimization
To improve the efficiency of our recorder system, we tested and compared three additional Cas1 variants identified by Yosef et al. (2023): Cas1(E296G), Cas1(V76L), and Cas1(V76L E269G). E296G represents the substitution of glutamic acid with glycine at position 296, while V76L refers to the replacement of valine with leucine at position 76. The double mutant Cas1(V76L E269G) carries mutations at both positions.
Structurally, the E269 residue in one of the two Cas1 monomers protrudes toward the integration-site DNA. Since the E residue is negatively charged, substituting it with the charge-neutral glycine residue can enhance DNA binding. This substitution might also destabilize the helix, which might further contribute to the integration efficiency. On the other hand, V76 is located within the β domain at the interface between the α and β domains of Cas1, and can form van der Waals interaction with G189 and A193 residues in the α domain. Replacing V76 with leucine, a larger hydrophobic residue, may thereby allosterically alter the inner dynamics of Cas1 dimer. (Yosef et al., 2023)
To test the integration efficiencies of the three Cas1 variants, we cloned them onto three different plasmids [Cas1(E296G): BBa_25VCQRET; Cas1(V76L): BBa_25WYL7QM; Cas1(V76L E269G): BBa_25OKVR2V]. Their construction followed the design in Figure 18B (BBa_25X8951T), which carries the wild-type Cas1. This plasmid contains the constitutive promoter J23115 for Eco1 RT expression, as well as the LacI-regulated T7 promoter for Cas1-Cas2 expression, all built on the pRSFDuet backbone. (Figure 19)
Figure 19. Plasmid designs for expression of Cas1 variants. Created by biorender.com.
Genome Integration
To enable our Retro-Cascorder recorder system in E. coli Nissle 1917, we first addressed the absence of an endogenous CRISPR array in its genome by integrating a native CRISPR array from BL21(AI) genome. We utilized the CRISPR Cas9 and Lambda Red recombination system to integrate the CRISPR array (Li et al., 2021). By aligning the genomes of BL21(AI) and EcN, we identified two loci in EcN whose flanking sequences are homologous to the termini of the BL21(AI) CRISPR array. We PCR-amplified these homologous regions from BL21(AI) and used them as homology arms to direct targeted integration in EcN, thereby minimizing potential context effects on the inserted array.
Genome integration was performed using a two-plasmid editing system (Figure 20). The first plasmid, pEcgRNA (BBa_25PAAV39), is a high-copy ColE1 plasmid that constitutively expresses a single guide RNA (sgRNA) under a strong constitutive promoter; the sgRNA contains a 20-nt spacer complementary to the intended EcN genomic target, which is fused to the conserved gRNA scaffold required for Cas9 recognition. The second plasmid, pEcCas (BBa_25T8TRM7), is a low-copy pSC101-based vector carrying both the Cas9 nuclease and the λ-Red recombination system (Gam, Bet, Exo). In this design, Cas9 is expressed constitutively to ensure efficient cleavage, whereas λ-Red recombinases are placed under the arabinose-inducible araBAD promoter to provide temporal control of recombination activity. To facilitate subsequent plasmid curing, pEcCas also encludes a rhamnose-inducible gRNA (gRNA-pBM1 under the rhaBAD promoter, activated by RhaS/RhaR in the presence of rhamnose) that directs Cas9 cleavage of pEcgRNA, and a sacB counter-selectable marker for sucrose-based elimination of pEcCas itself.
Figure 20. Plasmid design for a CRISPR-Cas9/Lambda Red based genome integration system. Some constitutive promoters and terminators are emitted for simplicity. Created by biorender.com.
For editing, the donor cassette, comprising the BL21(AI) CRISPR array flanked by the amplified homologous arms, was supplied together with the two editing plasmids. Upon induction of the editing program, Cas9 guided by the constitutively expressed sgRNA generates a double-strand break at the chromosomal target. Concurrent expression of the λ-Red recombination proteins promotes homology-directed repair using the provided donor DNA, resulting in precise integration of the BL21(AI) CRISPR array into the EcN chromosome.
Correct integration of the CRISPR array in engineered EcN was verified by colony PCR and further validated by nanopore sequencing. To eliminate the editing plasmids and prevent recurrent cleavage, we applied a two-step curing strategy. First, pEcgRNA was selectively removed by inducing gRNA-pBM1-directed Cas9 cleavage in cultures grown with rhamnose and kanamycin. Subsequently, pEcCas was cured via sucrose counter-selection, as expression of sacB is lethal to E. coli on sucrose-containing media. Colonies dying in this two-step process represent EcN strains carrying the integrated CRISPR array but free of editing plasmids, suitable for further characterization.
References
↑ Lin, J., Cagliero, C., Guo, B., Barton, Y. W., Maurel, M. C., Payot, S., & Zhang, Q. (2005). Bile salts modulate expression of the CmeABC multidrug efflux pump in Campylobacter jejuni. Journal of bacteriology, 187(21), 7417–7424. https://doi.org/10.1128/JB.187.21.7417-7424.2005
↑ Lei, H. T., Shen, Z., Surana, P., Routh, M. D., Su, C. C., Zhang, Q., & Yu, E. W. (2011). Crystal structures of CmeR-bile acid complexes from Campylobacter jejuni. Protein Science: A Publication of the Protein Society, 20(4), 712–723. https://doi.org/10.1002/pro.602
↑ Nasr, M. A., Timmins, L. R., Martin, V. J. J., & Kwan, D. H. (2022). A Versatile Transcription Factor Biosensor System Responsive to Multiple Aromatic and Indole Inducers. ACS Synthetic Biology, 11(4), 1692–1698. https://doi.org/10.1021/acssynbio.2c00063
↑ Elkins, C. A., & Savage, D. C. (2003). CbsT2 from Lactobacillus johnsonii 100-100 is a transport protein of the major facilitator superfamily that facilitates bile acid antiport. Journal of molecular microbiology and biotechnology, 6(2), 76–87. https://doi.org/10.1159/000076738
↑ Xue, Y., Tu, F., Shi, M., Wu, C. Q., Ren, G., Wang, X., Fang, W., Song, H., & Yang, M. (2016). Redox pathway sensing bile salts activates virulence gene expression in Vibrio cholerae. Molecular microbiology, 102(5), 909–924. https://doi.org/10.1111/mmi.13497
↑ Yang, M., Liu, Z., Hughes, C., Stern, A. M., Wang, H., Zhong, Z., Kan, B., Fenical, W., & Zhu, J. (2013). Bile salt-induced intermolecular disulfide bond formation activates Vibrio cholerae virulence. Proceedings of the National Academy of Sciences of the United States of America, 110(6), 2348–2353. https://doi.org/10.1073/pnas.1218039110
↑ Chang, H.-J., Zúñiga, A., Conejero, I., et al. (2021). Programmable receptors enable bacterial biosensors to detect pathological biomarkers in clinical samples. Nature Communications, 12, 5216. Programmable receptors enable bacterial biosensors to detect pathological biomarkers in clinical samples | Nature Communications
↑ Lindner, E., & White, S. H. (2014). Topology, dimerization, and stability of the single-span membrane protein CadC. Journal of molecular biology, 426(16), 2942–2957. https://doi.org/10.1016/j.jmb.2014.06.006
↑ Xia, J. Y., Hepler, C., Tran, P., Waldeck, N. J., Bass, J., & Prindle, A. (2023). Engineered calprotectin-sensing probiotics for IBD surveillance in humans. Proceedings of the National Academy of Sciences, 120(32). https://doi.org/10.1073/pnas.2221121120
↑ Riggers, D. S., Xenoulis, P. G., Karra, D. A., Enderle, L. L., Köller, G., Böttcher, D., Steiner, J. M., & Heilmann, R. M. (2023). Fecal Calprotectin Concentrations in Cats with Chronic Enteropathies. Veterinary sciences, 10(7), 419. https://doi.org/10.3390/vetsci10070419
↑ Zhu, D., Galley, J., Pizzini, J., Musteata, E., Douglas, M. V., Chazin, W. J., Skaar, E. P., Tabor, J. J., & Britton, R. A. (2025). Microbial Biosensor for Sensing and Treatment of Intestinal Inflammation. Advanced science (Weinheim, Baden-Wurttemberg, Germany), 12(27), e2504364. https://doi.org/10.1002/advs.202504364
↑ Stanton, B. C., Nielsen, A. A., Tamsir, A., Clancy, K., Peterson, T., & Voigt, C. A. (2014). Genomic mining of prokaryotic repressors for orthogonal logic gates. Nature Chemical Biology, 10(2), 99–105. https://doi.org/10.1038/nchembio.1411
↑ Yolanda Schaerli, Magüi Gili, Mark Isalan, A split intein T7 RNA polymerase for transcriptional AND-logic, Nucleic Acids Research, Volume 42, Issue 19, 29 October 2014, Pages 12322–12328, https://doi.org/10.1093/nar/gku884
↑ LaFleur, T. L., Hossain, A., & Salis, H. M. (2022). Automated model-predictive design of synthetic promoters to control transcriptional profiles in bacteria. Nature Communications, 13(1). https://doi.org/10.1038/s41467-022-32829-5
↑ Salis, H. M., Mirsky, E. A. & Voigt, C. A. (2009). Automated design of synthetic ribosome binding sites to control protein expression. Nature Biotechnology, 27(10), 946–950. https://doi.org/10.1038/nbt.1568
↑ Zhang, A., Li, D., Yu, T., Zhang, M., Cui, Y., Wang, H., Dong, T., & Wu, Y. (2025). Multi-omics approach to evaluate effects of dietary sodium butyrate on antioxidant capacity, immune function and intestinal microbiota composition in adult Ragdoll cats. Metabolites, 15(2), 120. https://doi.org/10.3390/metabol15020120
↑ Kallio, P., Pásztor, A., Thiel, K., Akhtar, M. K., & Jones, P. R. (2014). An engineered pathway for the biosynthesis of renewable propane. Nature Communications, 5(1). https://doi.org/10.1038/ncomms5731
↑ Sato, J., Ogawa, C. Y. L., Sandrini, M., Medina, A. N., Sato, F., & Vasconcellos, R. S. (2019). Study of keratin hair of domestic cat under methionine and cystine experimental diet using FT-Raman spectroscopy. Vibrational Spectroscopy, 100, 1–5. https://doi.org/10.1016/j.vibspec.2018.11.001
↑ Radha, S., & Gunasekaran, P. (2007). Cloning and expression of keratinase gene in Bacillus megaterium and optimization of fermentation conditions for the production of keratinase by recombinant strain. Journal of Applied Microbiology, 103(4), 1301–1310. https://doi.org/10.1111/j.1365-2672.2007.03372.x
↑ Tian, J., Xu, Z., Long, X., Tian, Y., & Shi, B. (2019). High-expression keratinase by Bacillus subtilis SCK6 for enzymatic dehairing of goatskins. International Journal of Biological Macromolecules, 135, 119–126. https://doi.org/10.1016/j.ijbiomac.2019.05.131
↑ Li, Xiaoguang & Zhang, Shihao & Zhang, Qian & Gan, Longzhan & Jiang, Guangyang & Tian, Yongqiang & Shi, Bi. (2022). Characterization and application of a novel halotolerant protease with no collagenase activity for cleaner dehairing of goatskin. Process Biochemistry, 113. https://doi.org/10.1016/j.procbio.2022.01.006
↑ Jaouadi, N. Z., Rekik, H., Badis, A., Trabelsi, S., Belhoul, M., Yahiaoui, A. B., Ben Aicha, H., Toumi, A., Bejar, S., & Jaouadi, B. (2013). Biochemical and molecular characterization of a serine keratinase from Brevibacillus brevis US575 with promising keratin-biodegradation and hide-dehairing activities. PloS one, 8(10), e76722. https://doi.org/10.1371/journal.pone.0076722
↑ Bhattarai-Kline, S., Lear, S. K., Fishman, C. B., Lopez, S. C., Lockshin, E. R., Schubert, M. G., Nivala, J., Church, G. M., & Shipman, S. L. (2022). Recording gene expression order in DNA by CRISPR addition of retron barcodes. Nature, 608(7921), 217–225. https://doi.org/10.1038/s41586-022-04994-6
↑ Yosef, I., Tridib Mahata, Goren, M. G., Degany, O. J., Ben-Shem, A., & Udi Qimron. (2023). Highly active CRISPR-adaptation proteins revealed by a robust enrichment technology. Nucleic Acids Research, 51(14), 7552–7562. https://doi.org/10.1093/nar/gkad510
↑ Li, Q., Sun, B., Chen, J., Zhang, Y., Jiang, Y., & Yang, S. (2021). A modified pCas/pTargetF system for CRISPR-Cas9-assisted genome editing in Escherichia coli. Acta biochimica et biophysica Sinica, 53(5), 620–627. https://doi.org/10.1093/abbs/gmab036