Development of Aspirin-Inducible Biosensors in Escherichia coli and SimCells

An aspirin-inducible SalR/Psal regulation system, originally from Acinetobacter baylyi ADP1, has been designed for E. coli strains. SalR is a typical LysR-type transcriptional regulator (LTTR) family protein and activates the Psal promoter in the presence of aspirin or salicylate in the range of 0.05 to 10 µM. The experimental results and mathematical simulations support the competitive binding model of the SalR/Psal regulation system in which SalRr competes with SalRa to bind the Psal promoter and affect gene transcription. The competitive binding model successfully predicted that weak SalR expression would significantly improve the inducible strength of the SalR/Psal regulation system, which is confirmed by the experimental results. This provides an important mechanism model to fine-tune transcriptional regulation of the LTTR family, which is the largest family of transcriptional regulators in the prokaryotic kingdom. In addition, the SalR/Psal regulation system was also functional in probiotic strain E. coli Nissle 1917 and minicell-derived SimCells, which would be a useful biobrick for environmental and medical applications.

feedback regulation loop. E. coli DH5a organisms carrying SRS and PAR gene circuits were induced by 0, 2.5, and 50 M aspirin, and green fluorescent protein (GFP) expression was measured at the single-cell level using flow cytometry and a fluorescence microscope ( Fig. 1; see also Fig. S1 in the supplemental material). With the strong proD promoter of the salR gene, SRS has a lower background leakiness than PAR. GFP expression in both SRS and PAR can be significantly induced by 2.5 M aspirin (Fig. 1). Density plots of front scatter (FSC) reads against the GFP channel show a single population of GFP-expressing cells in SRS at each inducing concentration, ranging from 3.67 Ϯ 0.12 to 32.11 Ϯ 1.45 GFP RFU (relative fluorescent units) (Fig. 1). The baseline reading in the PAR circuit was slightly higher (7.84 Ϯ 0.82 GFP RFU) than that of the SRS circuit (3.67 Ϯ 0.12 RFU). Furthermore, the PAR circuit exhibits a slightly bistable distribution ( Fig. 1 and Fig. S1), showing two small and large populations of GFPexpressing cells, ranging from 9.69 Ϯ 0.76 (low) to 165.16 Ϯ 5.52 (high) GFP RFU (Fig.  1). The leaky and bistable expression of GFP in the PAR circuit is consistent with the classical mathematic models of positive autoregulation circuits (22), confirming that the synthetic PAR and SRS circuits behave as their original designs intended.
The effector-free form, SalR r , is able to bind P sal promoter to repress transcription. The aspirin-inducible regulatory system is controlled by the SalR regulatory protein, originally from Acinetobacter baylyi ADP1 (18). SalR belongs to an LTTR superfamily, the largest family of regulators in prokaryotes (23), which contains an N-end DNA binding domain (DBD) and a C-end chemical recognition domain (CRD) (24). Based on previous reports of characteristic features in the LTTR family (25), the sliding dimer mechanism was adopted and is illustrated in Fig. S2. This SalR/P sal regulation was validated using a salicylate biosensor, ADPWH_lux, that was previously constructed by inserting a promoterless luxCDABE cassette between salA and salR of the sal operon in A. baylyi ADP1 (18,25).
To examine the molecular mechanism, two SalR mutants, ADPWH_NΔsalR and ADPWH_CΔsalR, were constructed with N end and C end, respectively, disrupted from ADPWH_lux (Table 1). In the mutant ADPWH_CΔsalR, four bases (ATAA) in the CRD domain of the salR gene were deleted using the sacB-km counterselection method (see Section S1). We hypothesized that this deletion did not prevent the mutated SalR from binding the P sal promoter, because the N terminus was unaltered. In the case of ADPWH_NΔsalR, the DBD of the salR gene in ADPWH_lux was knocked out by inserting a kanamycin resistance gene (18) into the ClaI cutting site. Hence, this ADPWH_NΔsalR mutant has completely disrupted the DNA binding motif of SalR and is unable to bind the P sal promoter.
Although both SalR mutants ADPWH_NΔsalR and ADPWH_CΔsalR were unable to respond to salicylate acid, the background levels of bioluminescence expression were different (Fig. S3): the average background bioluminescence expression of ADPWH_NΔsalR was 5,968 Ϯ 48, which is significantly higher than that of both ADPWH_CΔsalR (3,222 Ϯ 66; P Ͻ 0.001) and ADPWH_lux (3,857 Ϯ 99; P Ͻ 0.001). ADPWH_CΔsalR with a mutated C-end chemical recognition domain (but unchanged N-end DNA binding domain) should bind and repress the P sal promoter, reducing the background expression of bioluminescence, similar to ADPWH_lux, which has an intact salR. In contrast, disruption at the N end in ADPWH_NΔsalR completely alters the SalR structure, making it unable to bind the P sal promoter and leading to a higher background expression of bioluminescence (Fig. S3). Collectively, the results suggest that uninduced SalR or the effector-free form, SalR r , should bind and repress the P sal promoter.
SalR r /SalR a competitive binding hypothesis and mathematic model construction. Accorrding to the salR gene mutagenesis results described above, we hypothesize that both SalR r and SalR a can competitively bind the P sal promoter. To assess the hypothesized mechanism and investigate how variation of different aspects of the SRS and PAR circuits affect the aspirin-regulated performance, we developed a mathematical model to simulate and predict the performance of the three gene circuits. The interaction considered in our model for the SRS and PAR systems is outlined in Fig. 2 and is implemented using deterministic differential equations and binding models which describe gene regulation and expression. This process treats the SalR regulatory mechanism as a biased competition between binding of SalR a , which activates the P sal promoter, and SalR r, which occupies the binding site and represses the P sal promoter. For the SRS circuit, the competition between overexpressed SalR r and aspirin-activated SalR a is expected to produce low GFP expression levels. This competition has little effect on the PAR circuit, as its SalR level is tuned by a self-regulated feedback loop. We created a two-state differential equation model of the system, in which the concentration of SalR is governed by for the PAR circuit and for the SRS circuit, and the circuit output (GFP) concentration is governed by (for both circuits) where the regulating function that describes the interactions between SalR and inducer concentrations and the P sal promoter is given by ͓a͔ n ͓a͔ n ϩ K a n ͪ (4) The PAR circuit closes the loop, and P sal promoter controls both gfp and salR. Initially due to transcriptional leakage, a small amount of SalR is produced, binding and repressing the P sal promoter. When the inducer is added, transcription from P sal is activated, leading to the expression of both gfp and salR. SalR binding inducers will strengthen the P sal promoter and form a loop. SalR expression is dependent on the concentrations of the inducers. (C) The simple regulatory system (SRS*) variant in which the strong promoter proD was replaced by a weak promoter, PJ109. (D) Model structure for the SRS circuit. (E) Model structure for the PAR circuit. (F) Model structure for the SRS* circuit. gfp is marked as green and salR as blue. salR is assumed to form dimer binding to P sal with high affinity. An inducer molecule (a, yellow) reversibly binds with the effector-free form of SalR dimer (SalR r ) to yield its effector-bound active form (SalR a ). SalR a dimers can cooperatively bind P sal to activate gene expression, while cooperative binding of two SalR r dimers can repress gene expression. For detailed model description and definition of parameters, see Information Section S1 in the supplemental material.
in which a is inducer concentration, L is the leakage rate from the P sal promoter, K R defines the saturation point for repressive binding with P sal , K A defines the saturation point for activating SalR binding with P sal , and K a defines the saturation point for inducer interaction with the circuit. Thus, equation 3 reflects the balance we expect between repressive SalR r binding and inducer-dependent activation of SalR a binding. Further discussion of the modeling approach, as well as the procedure used for parameter fitting, is provided in Section S2. Experimental characterization and model validation in E. coli DH5a. The performance of the SRS and PAR designs in E. coli was characterized in terms of four features: (i) expression level, (ii) response time, (iii) output dynamic range/induction fold, and (iv) input dynamic range and leakiness (uninduced expression level). These four features are motivated by a desire to quantify a biosensor's utility in drug delivery and in vivo sensing. In such an application a biosensor would ideally exhibit a high sensitivity, fast response, high induction fold change, and tightly controlled expression. To examine these criteria, we ran plate reader experiments to analyze the SRS and PAR circuits over a range of inducer concentrations and compared these data to the simulated results from the mathematical model.
The SRS and PAR circuits in E. coli DH5a were induced by different concentrations of aspirin, and the response time of GFP expression was examined in three different growth media (Luria-Bertani [LB] medium, phosphate-buffered saline [PBS] with arabinose, and PBS with glycerol) (Fig. 3 and Fig. S4 and S5). The resulting expression profiles represent a broad range of sensor behaviors that might arise from different external disturbances and internal cell state variation. We defined the time required for biosensor induction response in each condition as the time taken for GFP fluorescence to reach and exceed 10% of its final maximum fluorescence output.
The expression profile for both the SRS and PAR circuits in LB media is shown in Fig.  3A and B. The fluorescence and optical density (OD) were monitored for 16 h over a range of aspirin concentrations between 0.05 and 50 M. GFP expression in both SRS and PAR circuits was detectable within 50 min, which is close to the theoretical minimum time required for GFP maturation (26). No significant difference in growth rate was observed between the SRS and PAR circuits (Fig. S6). Experimental results from the expression of PAR and SRS circuits are in good agreement with the mathematical model simulation (Fig. 3); in both cases the PAR circuit exhibited substantially higher GFP expression than the SRS circuit. In the SRS circuit, the abundant SalR r competitively occupied and repressed the P sal promoter, reducing the possibility of SalR a binding, which repressed overall GFP expression in the presence of aspirin and lowered the baseline level. The threshold concentration of aspirin required to activate GFP expression in the SRS circuit was 0.05 M, and leaky expression was low (Fig. 3). The PAR circuit tuned the strength of SalR expression through a positive feedback loop and reduced the competitive repression, resulting in much stronger GFP expression with an activation threshold concentration of 0.05 M (Fig. 3). However, the level of leaky expression of GFP in the PAR circuit was higher than that in the SRS circuit.
Redesigned SRS* genetic circuit validates competitive binding hypothesis. To redesign the SRS circuit, the expression of salR under the control of the constitutive promoter with different strengths was examined. Promoter proD, along with three weaker promoters, J109, J115, and J106, were fused with the gfp gene separately, and the strength was evaluated by detecting GFP expression levels. Figure 4A shows that proD was the strongest promoter and J109 was the weakest promoter among the four promoters proD, J109, J115, and J106, which is consistent with previous reports (21,27). Subsequently the J109 promoter (5 times weaker than proD) was characterized and replaced proD in the construction of SRS* (Fig. 4B). Interestingly, the GFP expression level in the SRS* circuit was 3.5-fold higher than that in the original SRS circuit, and the new SRS* showed an intermediate leakiness of GFP expression compared to the SRS circuit ( Fig. 3B and 4C). The threshold concentration for the SRS* circuit remained 0.05 M (P Ͻ 0.05). This is in a good agreement with the competitive binding hypothesis, suggesting that the SRS* circuit with the weak promoter J109 should produce less SalR than the SRS circuit, ease the SalR r /SalR a competitive binding on the P sal promoter, and increase the GFP expression level (Fig. 4C). By applying the same parameters, the mathematical model fit the experimental results well and predicted that the transcriptional rate from J109 is ϳ15% of that from proD, similar to the output ratios observed in Fig. 4B. Collectively, the salR mutagenesis experiment, the behavior of the SRS* circuit, and mathematic model fitting suggest that the hypothesis of SalR r /SalR a competitive binding is reasonable.
Dynamic range and leakiness. The dynamic range of an inducer, also referred to as its dose response, denotes the difference between the induction level that results in the biosensor's maximal output and the induction level at which the biosensor's output increases significantly above the uninduced level. There are a range of metrics for quantifying this criterion, such as the fold change of inducer concentration over which the sensor goes from 10% of its maximum induction level to 90% of this level (28).  (Fig. 5). PAR had a much higher induced expression strength than SRS in all carbon condition cases ( Fig. 3 and Fig. S4); this is in line with what we expected from classical models of autoregulation, which in many established in silico and in vivo studies has demonstrated that one of the benefits of using a positive regulation system architecture is signal amplification (29)(30)(31).  These results also confirm the conceptual model of the systems (Fig. 3), indicating that the SRS circuit with strong promotion of salR has little leaky expression compared to that of the PAR circuit. In all experiments of the SRS circuit, GFP expression levels in uninduced controls and blank background had no detectable difference, and no measurable leakiness was observed over 16 h, suggesting a very tight control in this design. In contrast, although the PAR circuit has a strong induced expression strength, it showed a leaky background, particularly in LB media (Fig. 3). SRS* increased the expression level to 18,119 Ϯ 794 GFP AU from 3,790 Ϯ 405 GFP AU in SRS, at a cost of increased leakiness from 431 Ϯ 33 to 3,201 Ϯ 45 GFP AU.
Performance of SRA and PAR circuits in probiotic E. coli Nissle 1917. The SRS, SRS*, and PAR circuits were then cloned into a probiotic strain, E. coli Nissle 1917. Considering that the gut environment is semiaerobic, E. coli Nissle 1917 with SRS, SRS*, and PAR circuits was characterized under both aerobic and semiaerobic conditions. To simulate the potential growth competition with host gut microbiome (32), a nutrientlimited environment was also tested. A small amount of glucose (0.4%, wt/vol) was supplied as a sole carbon source, as it is a common carbohydrate source from the human diet. E. coli Nissle 1917 with SRS, SRS*, and PAR circuits maintained a gradient response to inducer concentration under aerobic and semiaerobic conditions, highlighting the system's modularity and robustness to a changing chassis and nutrient conditions ( Fig. 6 and Fig. S6 and S7). The SRS circuit in E. coli Nissle 1917 exhibited tight control but low response to different concentrations of aspirin, while the PAR circuit demonstrated a higher response to aspirin (Fig. 6). The performance of the SRS, SRS*, and PAR circuits in E. coli Nissle 1917 was similar to that in E. coli DH5a, consistent with the hypothesis of SalR r /SalR a competitive binding. Notably, when supplied with a limited amount of 0.4% (wt/vol) glucose, the expression profile of PAR showed less leakiness, implying that the circuit could function optimally under nutrient-limited conditions. We suspect this change is due to unknown host interference, where cells experiencing glucose starvation alter their resource allocation (33). Furthermore, we noticed there is a slight growth advantage for transformed bacteria in the case of 50 M than for controls (0 M) (Fig. S7).  (18), it is evident that the biosensor with the SalR/P sal system can be triggered by either salicylate or aspirin. However, it is unclear if aspirin directly activates SalR or whether its metabolite, salicylate, does so. Based on the fact that cells have a slight growth advantage when supplemented with 50 M aspirin (Fig. S7) and that E. coli has nonspecific esterase activity (34), it is possible that aspirin was metabolized into salicylate and acetate. Salicylate was the main product of aspirin degradation (35), while acetate is known to be an energy source of E. coli under stringent nutrient conditions (36). To investigate if aspirin could be cleaved by esterase to produce salicylate that then activates SalR, we used the iron(III) chloride assay (37) to analyze aspirin conversion into salicylate.

Aspirin is converted into salicylate in E. coli. Based on previous work
The solution of 2 mM FeCl 3 mixing with salicylate quickly changes its color, which is detectable by absorbance analysis (Fig. S8A). Absorption curves indicate that the salicylate solution produced a peak at 300 nm after interacting with FeCl 3 (Fig. S8B). This unique absorbance peak can be used to identify the presence of salicylate. Interestingly, a time course analysis shows that both the aspirin water control and aspirin with E. coli Nissle 1917 cells can produce salicylate (Fig. S9). This suggests that aspirin in water can be broken down to salicylate at 37°C after 3 h, although the level of salicylate was lower than that of aspirin mixed with E. coli Nissle 1917 cells (Fig. S9). It also indicates that nonspecific esterase activity in E. coli Nissle 1917 assisted aspirin degradation, as was reported previously (34).
Collectively, aspirin can be converted into salicylate nonbiologically and biologically: nonspecific esterase activity in E. coli can accelerate the aspirin degradation. Based on this result, it is likely that aspirin is being broken into salicylate which then activates the SalR regulator, although aspirin could not be completely ruled out as a direct activator.
Performance of SRS and PAR circuits in SimCells. The SimCells were generated from the E. coli MC1000 ⌬minD strain with plasmid PAR_Amp* ( Table 1). The PAR gene circuit was also induced and expressed in purified SimCells (Fig. 7 and Fig. S10). The lack of any growth by SimCells confirmed their lack of chromosome. Both fluorescent images (Fig. S10) and the results from the microplate reader (Fig. 7) revealed that the background leaky expression of GFP in SimCells was very low, and GFP expression in SimCells became stronger with the increase of aspirin concentrations. SimCells were usually less than 500 nm in size, and some of them were actively moving under the microscope (Fig. S10), consistent with our previous reports (20).

DISCUSSION
An aspirin-inducible SalR/P sal regulation system is functional in different E. coli strains and SimCells. Probiotic bacteria are safe for human applications, but only a few can be used in gene manipulations (38,39). E. coli Nissle 1917 is a safe probiotic bacterium frequently used for medicinal purposes (19,(40)(41)(42). Importantly, gene manipulations can be performed in E. coli Nissle 1917, making it an ideal chassis for synthetic biology. SimCells are small (400 to 600 nm), chromosome-free, and functional cells formed as a result of abnormal cell division (20). SimCells contain transcription and translation machinery, meaning they are able to faithfully express designed gene circuits without interference from background gene networks in the chromosome. Their medical utility is also aided by the fact that they are nongrowing and small. Hence, E. coli Nissle 1917 and SimCells were chosen as chassis for the expression of gene circuits in this study.
A simple and sensitive response to aspirin in SalR/P sal system. There is only a limited range of inducers that can effectively operate within the human body, as many inducers may adversely impact human health or are broadly found in common dietary materials and therefore should be avoided. For example, antibiotic inducers might alter the structure of microbiota in the gut, while sugar inducers may suffer from unwanted induction. However, aspirin is largely absent from common human diets, and its pharmacological safety and effectiveness have been demonstrated extensively through many clinical trials and in vivo studies (43,44). For the SRS and PAR circuits studied in this work, the effective range of aspirin dosage required for activation is 0.05 to 10 M, which is more than 10 times lower than the human safe dosage of aspirin (ϳ111 M). Furthermore, induced GFP expression in cells with SRS and PAR circuits was observed to start after about 50 min and reached significant strength within 16 h (Fig. 3 and 4), which is a time scale comparable to that of aspirin's biological half-life.
An aspirin/salicylate-triggered nahR regulatory system was previously developed by Royo et al. (9). A Salmonella species was engineered for in vivo expression of 5-fluorocytosine to treat mice with fibrosarcoma. This regulatory circuit is a salicylateinducible cascade expression system derived from a naphthalene degradative pathway in Pseudomonas putida (9). It demonstrated effective drug delivery to minimize tumor size in mice, proving the concept of using salicylate/aspirin as inducers to trigger synthesis of cytosine deaminase that catalyzed production of potent cytotoxic fluorouracil for cancer chemotherapy. In this case, the gene circuit configuration was a simple cascade expression without feedback, and the effective induction concentration of aspirin was 125 to 5,000 M (9). Given that aspirin shows toxic effects at concentrations greater than 832 M (14), sensitive gene circuits for aspirin have an advantage in clinical practice. In addition, the well-characterized SRS and PAR circuits are simpler gene circuits than the regulatory module nahR/P sal ::xylS2, demonstrating a dose response to a lower range of aspirin concentrations, 0.05 to 10 M. The importance of feedback loops in genetic circuits is evident in their abundance in naturally occurring gene regulatory systems (22). As a universal network motif in transcription interactions (22,45), feedback systems can provide several advantages, including robustness despite the presence of noise (7), improved temporal response (46), and stability in changing cellular contexts (47). The present study not only compared two gene circuits, SRS (no feedback) and PAR (with feedback), but also successfully applied a mathematical model to simulate the performance according to the LTTR transcriptional regulatory mechanism ( Fig. 3 and 4). This hypothesized regulatory mechanism was further supported by altering the promoter strength and creating an intermediate SRS* circuit (no feedback and with a weaker promoter).
The results from the investigation of the molecular mechanism of salR regulation suggest that aspirin/salicylate can be transported into E. coli and activate the gene circuits SRS and PAR. In A. baylyi ADP1, a putative protein encoded by salD is presumably responsible for salicylate transport across the membrane. This protein is homologous to the E. coli FadL membrane protein (16). However, it is unclear how efficiently E. coli can transport aspirin/salicylate across its membrane. In E. coli, aromatic permeant acids such as salicylate and benzoate induce a large number of low-level multidrug efflux systems, governed by the Mar operon (marRAB) as well as additional, unidentified mechanisms (48). Since E. coli does not have a specific transport protein for aspirin/ salicylate, it is likely the efflux system will remain on after the induction; hence, we hypothesize that there is a maximum equilibrium concentration of aspirin/salicylate inside the cell cytoplasm. This hypothesis fit well with the mathematical models to successfully simulate the performance of E. coli in response to different concentrations of aspirin and growth conditions.
The performance of the SalR/P sal system is robust under various conditions. In the present study, a few aspects of the SalR/P sal regulatory system were investigated: its competitive inducer binding properties, response to various inducer concentrations, and behavior in E. coli given various system architectures and growth medium conditions. To tie the diversity of experimental observations together and to test the hypotheses that they inform, we built a mathematical model and demonstrated that it was able to describe and simulate all aspects of the system's behavior. The model not only supports the mechanism that we proposed to underlie the performance of both the PAR and SRS circuits but also provides a useful tool for the development of different systems that employ the SalR/P sal (or similar) regulatory system. These may include synthetic biological devices designed for clinical applications in the future; in such cases a priori utilization of the modeling architecture developed here can inform design choices and experimental approaches, thereby optimizing designs prior to their implementation.
Since many genera of bacteria naturally coexist with tumors in clinical situations (2), probiotic bacteria such as E. coli Nissle 1917 and nongrowing SimCells should be ideal chassis for the construction of anticancer agents due to their designable features, safe nonproliferation properties, and small size. Controllable expression of gene circuits, combined with safe delivery and tissue penetration, would enable E. coli Nissle 1917 and SimCells to selectively exert cytotoxicity in tumors. This study demonstrates that the aspirin-inducible SalR/P sal system is able to function in E. coli Nissle 1917 and SimCells, which would be useful for bacterial diagnosis and therapy in medicine.

MATERIALS AND METHODS
Chemical and reagents. All reagents were purchased from Sigma-Aldrich (Dorset, UK) unless otherwise noted. Antibiotics (kanamycin and ampicillin) were obtained from Fisher Scientific (UK). Sterile 1ϫ PBS solution was made by diluting 10ϫ PBS stock solution purchased from Fisher Scientific (UK).
Strains and plasmids. The strains and plasmids used in this study are listed in Table 1. Plasmids Psal-gfp-salR (PAR circuit) and PproD-salR-Psal-gfp (SRS circuit) were synthesized by GeneArts (ThermoFisher, UK). PJ109-salR-Psal-gfp was from PproD-salR-Psal-gfp, in which the proD promoter was replaced with the J109 promoter. PAR_Amp* was from Psal-gfp-salR, in which the kanamycin gene was replaced with an ampicillin resistance gene. Other plasmids listed in Table 1 were constructed using standard molecular cloning methods. The plasmids were transformed into E. coli DH5␣ cells by heat shock. E. coli Nissle 1917 was acquired from Ardeypharm GmbH (Herdecke, Germany). E. coli MC1000 ΔminD cells were made chemically competent before transformation with PAR_Amp*. The strains were cultured in LB medium with 50 g·ml Ϫ1 kanamycin or 100 g·ml Ϫ1 ampicillin (as appropriate) in a 37°C incubator.
Gene circuit design. The designs of both the positive autoregulation and simple regulation system plasmids employed identical modular components: P sal promoter, gfp, and colE1 origin of replication ( Table 1).
The codon sequence of salR was optimized to fit E. coli. The salR binding domain and P sal promoter were determined by analyzing the sequences of the sal operon (GenBank accession no. AF150928). SRS and PRS circuits were designed, and their structures are shown in Fig. 2A and B. In the PAR circuit, the gfp gene was fused directly to the P sal promoter to replace the sequence of the original salA, codonoptimized salR replaced the original salR, and the original structure, promoter, and ribosome binding site were kept intact in the salR operon. In the SRS circuit, gene-optimized salR is under the control of a constitutive promoter, proD (21), whose structure contained promoter-TACTAGAG-B0032-TACTAG-ORF-TACTAGAG-B0015, where the ORF was the salR gene and B0032 and B0015 are standard ribosome binding sites and terminators from the Registry of Standard Biological Parts (http://parts.igem.org/Main _Page). The full sequences of SRS and PAR were synthesized and supplied by GeneArt gene synthesis (ThermoFisher Scientific, Ltd.). Further detailed information of gene circuit design can be found in Information Section S1 in the supplemental material.
Culture and purification of SimCells. Due to the selection marker on E. coli MC1000 ⌬minD::km, a plasmid of PAR (Psal-gfp-salR) with alternative ampicillin resistance (PAR_Amp*) was needed and was constructed as follows. The ampicillin resistance gene was isolated by PCR from the plasmid pGEM-T (Promega, UK) by using an In-Fusion cloning kit (CloneTech, UK) and the primers PAR_Amp R For and Rev ( Table 2). The resulting sticky-end PCR product contained an overhanging EagI site and was thereby ligated in a linearized PAR plasmid. The targeting PAR plasmid was linearized and digested with EagI-HF (New England BioLabs), which disrupted the Kan ORF. Only upon correct insertion was ampicillin resistance replaced.
Following heat shock, E. coli MC1000 ⌬minD::km transformants with plasmid PAR_Amp* were selected on LB agar selection plates containing 100 g/ml ampicillin. A single colony from each plate was cultured at 37°C with continuous shaking at 200 rpm overnight in 5 ml LB broth supplemented with 100 g/ml ampicillin (LB amp). Overnight culture was added to fresh LB-ampicillin at a ratio of 1:1,000 and cultured for 24 h. For the minicell purification, a modification of a previously described method (20) was used in order to obtain a high yield and purity while maintaining maximum ATP within the minicells. Overnight culture was centrifuged at 4°C at increasing speeds, with increments of 1,000 ϫ g, from 1,000 ϫ g to 4,000 ϫ g for 10 min at each step to remove parent cells from the suspension. The supernatant was subsequently treated with 100 g/ml ceftriaxone and incubated at 37°C for 1 h with shaking at 200 rpm and then stored at 4°C overnight. Following further centrifugation at 4,000 ϫ g for 15 min to pellet any remaining lysed and elongated parent cells as a result of the ceftriaxone treatment, the supernatant was passed through a 0.22-m nitrocellulose membrane (Sigma, UK) and the minicells resuspended from the membrane into sterile PBS solution to concentrate them 100-fold. PBS was chosen as it is a cost-effective medium in which to suspend the cells and to maintain osmotic balance. Minicell samples were maintained at 4°C prior to testing. The purity of minicell suspensions was determined by a plate count method after 24 h of incubation on LB-ampicillin agar plates at 37°C and conducted in triplicate.
Flow cytometry. Transformed E. coli DH5␣ cells were induced with an aspirin concentration gradient (final concentrations of 0, 0.05, 0.1, 0.25, 0.5, 1, 2.5, 5, 10, 20, and 50 M) in kanamycin (50 M) and LB for 16 h at 37°C in a 600-rpm shaker. Induced culture was then analyzed using a FACSCalibur (BD Bioscience), with 100,000 events being captured for each sample. Gating was performed on forward and side scatter and the GFP channel to remove noise caused by debris and background. Data were exported to Kaluza flow cytometry for subsequent analysis and visualization.
Induction and GFP measurements. A stock solution of L-arabinose, glycerol, and aspirin was made in deionized water and sterilized using a 0.2-m syringe filter to be tested with cells containing SRS and PAR circuits. A total volume of 200 l (cultured bacterial cells and inducing aspirin) was loaded in a 96-well black-sided, clear-bottomed plate (Nunclon, UK) in quadruplicate. The plate was then placed into a BioTek Synergy HT microplate reader (BioTek Corporation, UK) maintained at 37°C with reads of fluorescence (excitation, 480 nm; emission, 520 nm) and OD at 600 nm. Readings were recorded every 15 min for 16 h. Cellular intrinsic fluorescence background was subtracted from the control of E. coli without plasmid.
For experimental conditions in PBS and M9 prior to induction, transformed bacteria, which were cultured overnight in LB, were centrifuged and washed with PBS and M9 three times at 4°C to remove any residue of LB broth. For achieving microaerobic conditions, 50 l of mineral oil was added on top of the 200-l reaction volume in a 96-well plate.