Redesign of ultrasensitive and robust RecA gene circuit to sense DNA damage.

SOS box of the recA promoter, PVRecA from Vibrio natriegens was characterized, cloned and expressed in a probiotic strain E. coli Nissle 1917. This promoter was then rationally engineered according to predicted interactions between LexA repressor and PVRecA . The redesigned PVRecA-AT promoter showed a sensitive and robust response to DNA damage induced by UV and genotoxic compounds. Rational design of PVRecA coupled to an amplification gene circuit increased circuit output amplitude 4.3-fold in response to a DNA damaging compound mitomycin C. A TetR-based negative feedback loop was added to the PVRecA-AT amplifier to achieve a robust SOS system, resistant to environmental fluctuations in parameters including pH, temperature, oxygen and nutrient conditions. We found that E. coli Nissle 1917 with optimized PVRecA-AT adapted to UV exposure and increased SOS response 128-fold over 40 h cultivation in turbidostat mini-reactor. We also showed the potential of this PVRecA-AT system as an optogenetic actuator, which can be controlled spatially through UV radiation. We demonstrated that the optimized SOS responding gene circuits were able to detect carcinogenic biomarker molecules with clinically relevant concentrations. The ultrasensitive SOS gene circuits in probiotic E. coli Nissle 1917 would be potentially useful for bacterial diagnosis.

Introduction

Naturally occurring microorganisms living on and in the human body present an opportunity for in situ bacterial diagnostics and therapy (Kurtz et al., 2019; Forbes, 2010). Despite advancements in treating gut inflammation (Forbes, 2010; Archer et al., 2012; Kurtz et al., 2019), hyperammonemia (Kurtz et al., 2019), obesity (Chen et al., 2014) and various cancer treatments (Ganai et al., 2009; He et al., 2017; Yu et al., 2019), bacterial therapy still lacks accurate control systems that can act as ‘input modules’. Synthetic biology could contribute to the development of inexpensive, rapidly deployable bacterial diagnostics and therapy (Slomovic et al., 2015). An ideal module to bacterial diagnostics and therapy should be highly sensitive to low dosage of inducer; low baseline with minimal background expression; strongly responsive to produce readily detectable outputs; stable to resist nutrient and environmental fluctuations, spatially controllable to enable a localized activation. Finally, the engineered host bacteria containing the module should be clinically safe.

Some carcinogenic or DNA damaging compounds generated in the human body are responsible for cancers (Seitz and Becker, 2007; Ridlon et al., 2016; Nguyen et al., 2018). The SOS response system is a global response to such DNA damage and represents a universal regulation mechanism in bacteria (Cohen et al., 2008; Simons et al., 2010), yeast (Fu et al., 2008) and humans (Michel, 2005). The response mechanism of the SOS system showed significant overlap in different prokaryotic organisms (Erill et al., 2007; Kreuzer, 2013). However, the wild‐type E. coli SOS system is not sufficiently sensitive or rapid for diagnosis of cancer biomarker molecules. Therefore, a sensitive, low background, strong responsive, stable and spatially controllable SOS system is required for the application of bacterial diagnostics and therapy.

The Gram‐negative and non‐pathogenic Vibrio natriegens are the fastest growing bacterium that has been isolated (Lee et al., 2016). First reported in the early 1960s, V. natriegens was found to have a doubling time of fewer than 10 min under optimal conditions, which is less than half of the shortest reported doubling time of E. coli recorded in the literature (Erill et al., 2007). Considering its fast growth rate (Lee et al., 2016), it is expected that V. natriegens may be equipped with a robust and sensitive SOS system for DNA repairing in short time due to natural errors of DNA replication. The expression of the universal SOS response regulator, RecA, in response to UV radiation by V. natriegens was previously reported (Booth et al., 2001). However, the response time and sensitivity of the recA promoter of V. natriegens to DNA damage remain unknown.

In this study, we first identified LexA binding sites of V. natriegens recA gene promoter (PVRecA) according to the predicted promoter‐LexA interactions. The transcriptional sigma factor, LexA binding affinity and interactions with the PVRecA promoter were then characterized by mutagenesis. We then rationally designed a spacer between LexA binding sites in PVRecA and tested its performance to sense DNA damage, such as carcinogenic chemicals and UV radiation. To demonstrate the promoter modularity and to further strengthen the tightness of the control, we created a series of chimeric promoters and examined the performance of the redesigned recA promoter.

Natural heterogeneity, noise and variability of gene expression at a constantly changing environment are challenges for bacteria therapy. The physiology condition in human tissue (pH, oxygen and nutrient availability) (Courbet et al., 2015) can alter the dynamic range of synthetic circuits, thus undermining an engineered system’s performance. Inspired by recent work on disturbance‐resistant control systems (Aoki et al., 2019), we further improved the stability and sensitivity of the SOS system through a designed negative feedback loop, which is an integrated HrpRS‐based transcriptional amplifier and TetR‐based repression system (Wang et al., 2014). This operational feedback design was demonstrated to improve the robustness of the sensing circuit when subjected to different environmental changes.

We further demonstrated that the optimized system could be used as an optogenetic tool, which uses light to control gene expression and have broad applications in synthetic biology. Optogenetic control is particularly attractive because of its ability to enable remote (i.e. non‐contact), precise (predictable input‐output response) and rapid actuation of gene expression (Olson et al., 2014; Zhao et al., 2018; Aoki et al., 2019). This optimized SOS system can be can be precisely controlled by UV light.

All circuits were tested and validated in E.coil Nissle 1917 due to its established safety and efficacy in clinical applications (Sarate et al., 2019; Crook et al., 2019) and genomic similarities with other E. coli variants (Grozdanov et al., 2004). Overall, we developed a toolbox for SOS gene circuit for the detection of DNA damage, which can detect a range of disease biomarker molecules and carcinogenic antibiotics in a sensitive and robust manner.

Result

Cloning recA promoter from V. natriegens into E. coli Nissle 1917 for SOS response

A schematic diagram of the SOS DNA repair mechanism is shown in Fig. S1A. A LexA dimer binds the SOS box and represses the promoter of genes related to DNA repair (Fig. 1A), including the recA gene. When DNA damage occurs, the single‐stranded (ssDNA) from mismatched DNA and RecA protein form a complex (ssDNA‐RecA), which induces the self‐cleavage of LexA (Butala et al., 2009), resulting in de‐repressing of gene expression in the SOS machinery (Fig. S1B). Alignment of recA promoters in various bacteria showed that the binding sites of LexA dimer are highly conserved CTGT‐N8‐ACAG which is flanked by −35 and −10 sigma binding site (Fig. 1A and B). Sequence alignment of V. natriegens and E. coli indicated a high similarity in both conserved LexA and sigma factor binding sites (Table S1). Sequence alignment of other Vibrio fischeri (˜280 min) exhibited similar conserved region but different upper/lower flanking sequences (Table S2). It implicates that the flanking region of LexA binding motifs could affect the response time of SOS. V. natriegens and E. coli Nissle 1917 share 72% homology in LexA protein sequence and the peptide sequence of putative binding site is similar (Fig. S1C) (Little et al., 1981; Brent and Ptashne, 1981). It suggests that LexA in E. coli Nissle 1917 could recognize a similar SOS box in Vibrio natriegens recA (vrecA) gene promoter.

Fig. 1

Binding of LexA and recA/lexA promoter.

A. WebLogo diagram of consensus LexA binding site compiled between six bacteria species, the consensus sequence for SOS box is (CTGTN8ACAG). In the 3D structure interaction, DNA base pair interaction with LexA dimer was highlighted in orange box where each of the motif base pair interact with the α helix domain directly. 3D crystal structure was acquired from (UniProt ID P0A7C2).

B. Sequence alignment of the highly conserved region of LexA binding site (Orange), Sigma factor binding site (Yellow) and 8‐bp spacer (Red) among nine promoters from different species.

C. Circuit performance with different SOS binding box sequence. Sequence alignment of SOS box of E. coli recA promoter and three variants of V. natriegens recA promoter. GFP induction of all three v. RecA variants of 3 h under induction of UV. Sigma factor binding site −35 (TAGACA) and −10 (TATAAT) was highlighted in yellow and LexA binding site highlighted in orange.

D. Dosage response and induction fold of change profile for each construct under the exposure of UV. Each experiment has three biological and three measurement replicates.

All strains and plasmids in this study are listed in Table 1. We constructed a plasmid carrying PVRecA (V. natriegens recA promoter, NCBI ref: NZ_CP016347.1) promoter to control the expression of a superfolder GFP (sfGFP) gene (Pédelacq et al., 2006) in E. coli Nissle 1917 (Fig. S2B). To make the gene circuit comparable, we constructed a basic PERecA promoter with native E. coli Nissle 1917 promoter of recA gene using the same ribosome binding site (RBS, AAAGAGGAGAAA, BBa_B0030, parts.igem.org) and the gap sequence GGTACC between RBS and ATG of sfGFP gene (Fig. S2). Fig. 1C shows that the PVRecA promoter was more sensitive and had a higher expression amplitude than the E. coli native promoter PERecA. The PVRecA promoter can be triggered by UV with exposure time as short as 2 s, while the E. coli native recA promoter PERecA required at least 4 s (Fig. 1C). The fold change of PVRecA promoter was significantly higher than PERecA (Fig. 1D). UV exposure durations up to 10 s did not affect cell growth compared to a control without UV exposure in the testing conditions (Fig. S3). However, PVRecA also demonstrated a higher leakiness than the PERecA (Fig. 1C).

Table 1

The strains and plasmids used in this study.

Strains	Genotype
E.coli Nissle 1917 (EcN)	Ardeypharm GmbH (Herdecke, Germany)
E.coli DH5a	fhuA2 Δ(argF‐lacZ)U169 phoA glnV44 Φ80 Δ(lacZ)M15 gyrA96 recA1 relA1 endA1 thi‐1 hsdR17
E.coli BL21(DE3)	fhuA2 [lon] ompT gal (λ DE3) [dcm] ∆hsdS λ DE3 = λ sBamHIo ∆EcoRI‐B int::(lacI::PlacUV5::T7 gene1) i21 ∆nin5
E.coli JM109	endA1, recA1, gyrA96, thi, hsdR17 (rk –, mk +), relA1, supE44, Δ( lac‐proAB), [F´ traD36, proAB, laqIqZΔM15].

Genes of recA and lexA with their native promoters of E. coli Nissle 1917 were cloned into a plasmid to make PVRecA‐AT ‐sfGFP. The results demonstrated that the sensitivity to mitomycin C has been increased (Fig. 2A and B), and the baseline has been significantly repressed due to the expression of extra lexA on PVRecA‐AT ‐sfGFP compared to the cases without (Fig. 2D). This confirms that LexA from E.coli Nissle 1917 can indeed bind with PVRecA‐AT promoter from V. natriegens and repress baseline expression.

Fig. 2

Gene circuit performance with addition of LexA and RecA module cloned from E. coli Nissle 1917.

A. Genes of lexA and recA with original promoters were cloned into PvrecA_AT system and tested in E. coli Nissle 1917 B) PvrecA_AT system and C) PE‐recA_PE‐lexA with no sfGFP.

D. The baseline leakiness for each system (A, B and C) was compared using pair t tests. Each experiment has three biological and three measurement replicates.

Characterizing the SOS box of V. natriegens recA promoter

To further analyse the PVRecA promoter, we used the LexA crystal structure and LexA‐promoter interaction model (Zhang et al., 2010) to annotate PVRecA promoter. We identified the SOS box in PVRecA promoter, which is a 16‐bp sequence (CTGTATGAGCATACAG) flanked by sigma factor binding sites, −35 (TAGACA) and −10 (TATAAT) (Supplementary information). Table 1 shows that the structure of the SOS box in V. natriegens (PVRecA) is similar to that in E. coli Nissle 1917 (PERecA), and their LexA binding sites are identical, recognizing the reverse complementary CTGT/ACAG with 8 bp spacer in between (CTGT‐N8‐ACAG).

We then created a series of mutations and truncations of the PVRecA promoter of wild‐type V. natriegens, including a full −35, SOS box and −10 deletion; partial SOS box deletion; full SOS box deletion with intact −35; and −35 and SOS box deletion. It showed that deletion, extension of SOS box, or removal of the −35 sequence completely silenced and disrupted the PVRecA promoter in response to UV‐induced DNA damage (Fig. S4), whilst leaving only −35 site intact increased the background expression level. These results validated our annotation of the SOS box and −35 and −10 sites in the PVRecA promoter, although previous reports proposed different annotations (Bernstein et al., 2005; Jiang et al., 2016; Simons et al., 2010).

Rational redesign SOS box of PVRecA promoter

The LexA binding motifs of CTGT and ACAG are reverse complementary (Table 1). The 8‐bp  spacer sequence between the binding sites can twist, allowing LexA dimer to recognize and bind two identical positions of CTGT on recA promoter (Zhang et al., 2010). Notably, the baseline expression level in the wild‐type promoter of V. natriegens was higher than the native recA promoter in E. coli Nissle 1917 (Fig. 1). Based on the predicted structure of LexA‐DNA complex (Zhang et al., 2010), the AT‐rich spacer sequence between CTGT and ACAG in the SOS box would be prone to twist, due to the weaker hydrogen bond in AT pairs, which should enhance LexA‐CTGT binding and repression. Hence, such an AT‐rich spacer may be able to reduce the baseline expression of PVRecA promoter Tables 2 and 3.

Table 2

SOS box design in the promoter of E. coli and V. natriegens.

	SOS sequence 5’ → 3’
PERecA	TTGATACTGTATGAGCATACAGTATAAT
PVRecA	TAGACACTGTATGAATCAACAGTATAAT
PvRecA‐AT‐Repeats	TAGACACTGTATATATATACAGTATAAT
PvRecA‐A‐Tandem	TAGACACTGTAAAAAAAAACAGTATAAT

Table 3

The performance of gene circuits in E. coli Nissle 1917.

Circuits	Dynamic range (rfu)	LOC (UV/MMC)	Inducer range$ / f.c (fold change)
PERecA‐sfGFP	2292 ± 40–6738 ± 47	1 s	1–10 s
PVRecA_sfGFP	10 941 ± 767–42 957 ± 1075	1 s	1–10 s
PVRecA‐AT_sfGFP	3588 ± 127–18 844 ± 501/2193 ± 75–13 284 ± 307	1 s/0.1 µM	1–10 s/0.1 µM–10 µM
PVRecA_AA _sfGFP	53 527 ± 2214–10 068 ± 1965	1 s	1–10 s
PVRecA_Jungle_Down_sfGFP	2089 ± 74–7088 ± 121	1 s	1–10 s
PVRecA_Jungle_Up _sfGFP	1464 ± 132–6130 ± 162	1 s	1–10 s
PVRecA_Jungle_Both _sfGFP	1271 ± 39–3428 ± 281	1 s	1–10 s
PVRecA_AT‐GFP	2765 ± 31–5969 ± 882/3682 ± 55–22771 ± 919	1 s/0.5 µM	1–10 s/0.5–10 µM f.c = 20
PVRecA_sfGFP + PE. LexA‐PE. RecA	1268 ± 47 – 7579 ± 124	16 nM	16–750 nM/f.c = 46.87
PVRecA‐Amplifier*	1361 ± 186–12 032 ± 257	6.125 nM	6.125 nM–200 nM/f.c = 32.65
PVRecA‐Amplifier + Feedback+	12 864 ± 261–26 100 ± 1105	6.125 nM	6.215–100 nM/f.c = 16.32

PVRecA‐Amplifier*: PVRecA_AT connected with HrpRS transcriptional amplifiers (PVRecA_AT_HrpRS_PhrpL_GFP). PVRecA‐Amplifier + Feedback+: PVRecA_AT connected with HrpRS transcriptional amplifiers linked with PtetO‐tdtomato‐HrpV repressor. (PVRecA_AT_HrpRS_PhrpL_sfGFP_TetR + PTetO_tdtomato_HrpV). $: An inducer was UV light (s) or MMC.

Two spacer variants of SOS box with different dynamic responses are shown in Fig. 1 and Table 2. It was found that a spacer region of four AT repeats (PVRecA‐AT) (ATATATAT) created a lower baseline expression level, and a more significant fold change than wild‐type promoter and AAAAAAAA (PVRecA‐AA) mutant (Fig. 1). Despite a slight decrease in the absolute intensity, compared to the wild‐type PVRecA SOS (TGAATCAA), the baseline expression of PVRecA‐AT was ˜ 2.4 fold lower than that of PVRecA. This indicates that PVRecA‐AT allows tight binding of LexA dimer, enhancing repression (Fig. 1). Interestingly, alternation of the spacer to A‐tandem resulted in higher leakiness and less fold change (Fig. 1) despite having similar AT ratio. The spacer with AT repeat might have a high affinity to LexA and lower dissociation constant Kd. These results suggest that PVRecA‐AA and PVRecA‐AT should have different twist angles and minor grooves in the spacer, affecting LexA binding (Zhang et al., 2010). Alternatively, other factors other than LexA binding efficiency may be at play, which is worth further investigation. The performance of variants with different spacers is shown in Table 3.

After obtaining an optimized SOS box (spacer AT repeats), we created three chimeric promoters with multiple LexA binding sites to further suppress the leakiness of PVRecA‐AT promoter. Inspired by the EilR repressor promoter structure in a Jungle Express system (Ruegg et al., 2018), an additional SOS box with 5’ complement sequence was inserted downstream of the original SOS sequence to enhance repression. As shown in Fig. 3 and Table 3, this additional SOS box reduced the system background level (10 941 GFP rfu) down ˜ 5.2 fold to 2089   GFP rfu PVRecA‐AT Jungle_Down, which could be potentially due to the formation of a hairpin loop between two SOS boxes or recruitment of additional LexA protein at the promoter region (Fig. S5). Upstream insertion of an additional SOS box (PVRecA‐AT Jungle_Up) displayed a further reduction of baseline leakiness (to 1464  GFP rfu) with lower noise (Fig. 3A and Table 3). The insertions of both upstream and downstream SOS boxes (PVRecA‐AT Jungle_Both) displayed the lowest leakiness (1271  GFP rfu) despite lowering the amplitude of system output (Fig 3A and C and Table 3). All strains harbouring system variants demonstrate similar cell growth (Fig. S6). Overall, these results demonstrated that the PVRecA‐AT promoter could be redesigned to generate different dynamic responses via simple addition of SOS boxes. This modularity could be useful in creating other chimeric promoters with dual‐chemical/spatial control input.

Fig. 3

Characterization of chimeric promoters with SOS boxes. Design of each jungle promoter is shown within the graph. Green box indicates the sigma factor binding site −35 and −10. Yellow box indicates the LexA binding site, and red box indicates the SOS box.

A. Induction profile of different chimeric promoters under UV exposure.

B. A comparison of baseline levels of different chimeric promoters. Pair t tests were carried out. P < 0.002 **P < 0.02 *P < 0.001 **** (C) Dosage response and induction fold of change profile for each construct under the exposure of UV light. Each experiment has three biological and three measurement replicates.

Signal amplification through transcriptional amplifiers

To enhance the sensitivity of the DNA damage sensing system, we added an HrpRS‐based transcriptional amplifier (Wang et al., 2014) to amplify the output signal of the DNA damage sensing system (PVRecA‐AT). As shown in Fig. 4 and S8, the sensitivity can be improved, allowing detection down to 6.125 nM of mitomycin C (MMC), which is by far the most sensitive genotoxic sensor reported (Lemoine et al., 2018). The maximum amplification factor for PVRecA‐AT promoter can reach 4.3‐fold when coupled with HrpRS‐based amplifier (Fig. 4) without showing metabolic burden (Fig. S7). However, the introduction of this amplification system led to a significant delay of activation, which the response time extended to ˜ 2.5 h. Compared to the original arsenic amplifiers construct which also displayed a similar time of delay (Wang et al., 2014), the delay is likely to be independent of the biosensor target. This delay could be due to the global sigma‐54 regulation of HrpRS system and additional time required for the translation of HrpRS complex in E. coli Nissle 1917. We have tested a number of nutrient depleting conditions which have previously been shown to reduce response delay in sigma‐54 dependent promoter (Huang et al., 2008; Reitzer and Schneider, 2001). We noticed there is a significant difference in response time when different carbon sources were used. Glycerol as carbon source had a shorter delay (2.5 h) compared to glucose as carbon source (4 h) (Fig. S8). Since little difference was observed in the OD, the delay might be related to the global catabolite repression in bacteria (Görke and Stülke, 2008).

Fig. 4

The amplification effect of PVRecA promoter coupled with HrpRS transcriptional amplifier.

A. Original PVRecA_AT promoter fused with GFP as an output.

B. PVRecA_AT promoter with amplifier. The expression of HrpR and HrpS subunits subsequently activate the PhrpL promoter with a GFP output. Induction kinetics were tested with a concentration gradient of mitomycin C (0;6.125;12.5; 25; 50;100; 200 nM) over 16 h.

C. A comparison of PVRecA promoter with and without amplifier in response to different concentrations of mitomycin c (MMC). The maximum amplification can be 4.3‐fold. Each experiment has three biological and three measurement replicates.

A negative feedback gene circuit to resist fluctuations of environmental conditions

Although the amplifier constructs demonstrated high sensitivity with minimal leakiness (Fig. 4), it lacks stability and robustness against environmental fluctuation, such as changes in the nutrient richness, pH and temperature. To improve the stability and robustness of our SOS sensing system, we coupled tuneable TetR repression with the transcription amplifier to make negative feedback loop (Fig. 5). This was achieved by adding TetR regulator protein under the control of PhrpL and the counter repressor unit HrpV for HrpRS complex under the control of TetR regulating promoter (PtetO). Upon the activation of HrpRS amplifier, the flux of TetR repressor can be tuned by anhydrotetracycline (aTc), thus forming a closed‐loop negative feedback system. To visualize the performance of the negative feedback system, we incorporated a red fluorescence tdTomato downstream of HrpV. When PtetO was activated, we observed an increased expression of tdTomatoes with a decreased expression of sfGFP (Fig. 5C). Upon over induction of PVRecA‐AT, the promoter PhrpL was induced to a high level, which also leads to a high expression of TetR. In the absence of aTc, expression of HrpV was repressed, and the overall expression of sfGFP was high. In the presence of aTc, the expression of HrpV caused negative regulation through its interaction with HrpS (Jovanovic et al., 2011), and the overall expression of sfGFP decreased. Once aTc reached between 6.25 and 12.5 nM, HrpV was likely to reach saturation, leading to significant repression of overall sfGPP level (Fig. 5). Hence, this negative regulation system can be controlled by the concentration of aTc (<12.5 nM), stabilizing the overall gene expression.

Fig. 5

Performance of PVRecA‐AT promoter with amplifier and negative feedback loop.

A. Schematic diagram of the genetic operational amplifier with a negative feedback loop. A plasmid consists of the optimized PVRecA‐AT promoter fused with hrpRS, sfGFP gene and tetR; another plasmids consist of a negative regulator PhrpL and Tdtomatoes reporter.

B. Expression profile under the induction of both aTc gradient (0 to 200 nM) and MMC gradient (0 to 200 nM).

C. Heatmap of sfGFP and Tdtomatoes expression corresponding to different concentration of inducers.

D. Comparison of fold change of PVRecA‐AT amplifier gene circuits with and without the negative feedback loop. Steady state of aTc = 3 nM and MMC = 50 nM at the end of 16 h time course were used to calculate, optimal condition indicates as pH7, oxygen available, 37°C, M9 supplemented with 2%w/v glucose. pH6 was calibrated using HCl titration, micro‐oxygen environment was created using oil seal over each well. 50% nutrient indicated 1% w/v glucose supplemented instead of 2% w/v used in optimal condition. Each experiment has three biological and three measurement replicates.

To verify the performance of this negative feedback gene circuit, we subjected it to several conditional changes including temperature, nutrient level, oxygen availability and pH (Fig. S9 and S10). We showed that in the presence of the negative feedback module (under the induction of 3 nM of aTc – thus allowing the expression of HrpV) the PVRecA‐AT sensing circuit exhibited system stability and robustness against environment fluctuations by maintaining the same level of induction fold at steady state (16 h post‐induction) (Fig. 5D, S9, and S10). Despite the decrease in the relative fold of change and narrow operational range caused by HrpV, which could be solved through mining lower affinity mutant. The increased robustness indicates the potential of our PVRecA‐AT sensing system with additional feedback module for in situ applications, where exposure to the real‐world environment or gut environments can make it challenging for synthetic circuits to perform as predicted.

Overall, we described an ultrasensitive DNA damage sensor equipped with a functional genetic operational amplifier that combines a chimeric promoter as the noise canceller, a HrpRS‐based construct as the signal amplifier and a PtetO‐negative feedback system.

Spatial control and adaptation of SOS gene circuit in response to UV light

We next demonstrated the robust and sensitive response of the PVRecA promoter, demonstrating UV inducible gene circuits through a spatial control. As a proof‐of‐concept, E. coli Nissle 1917 with PVRecA‐AT system formed colonies on LB plates, which were activated by a UV light to express sfGFP (Fig. S11). The fluorescence of sfGFP‐expressing colonies was imaged after UV exposure for 10 and 20 s, respectively, following 4 h incubation at 37°C. The sfGFP expression under chimeric PVRecA‐AT Jungle_Down promoter displayed a lower baseline but less sensitivity, compared to that under PVRecA‐AT (Fig. S11A), the results from solid biofilm are consistent with the experiments done in bacterial liquid culture condition shown in Figs 1 and 2. Interestingly, a ring‐like structure was observed (Fig. S11), and this could suggest SOS system being induced to express sfGFP in newly grown cells more readily than the growth stalling old cell in the centre.

To investigate temporal dynamics and adaptation of the DNA damage sensor, we tested it using a Chi. Bio turbidostat platform(Steel et al., 2020), which can maintain a constant culture OD while measuring fluorescence and inducing DNA damage with 280 nm UV radiation. Two variants were tested (PVRecA‐AT with either sfGFP or sfGFP attached with ssrA fast degradation tag), with each subjected to a UV dosing regime in which radiation dose intensity was doubled every 5 h (Fig. 6). For the sfGFP variant, we observed a log‐linear relationship between UV dose and sfGFP expression, with the system demonstrating a reliable response over a 128‐fold increase in induction level (Fig. 6). In each cycle, a return‐to‐baseline took approximately 5 h which we hypothesized was due to the time necessary for GFP dilution during growth. The sfGFP‐ssrA (LVA‐ BBa_K1399001) construct (in which the fluorescent marker is degraded on a faster timescale than cellular growth) supported this hypothesis, demonstrating that our optimized PVRecA‐AT reporter system can be used to generate short (˜1 h) pulses of gene expression in response to UV induction.

Fig. 6

Adaptation of E. coli Nissle 1917 with PVRecA‐AT –sfGFP and PVRecA‐AT – sfGFP‐ssrA in turbidostat over 40 h with interval UV exposure.

A. Dynamics of sfGFP expression under periodic UV exposure. Every 5 h is a cycle. The first UV dose was 10 mW 280 nm for 2.4 s and the UV doses doubled in the next cycle. Small fluorescence expression was observed for PVRecA‐AT – sfGFP‐ssrA as a rapid degradation of fluorescent protein with ssrA tag.

B. Relationship between applied UV dose and fluorescence peak height. Error bars are the standard deviation of the ten reads in fluorescence taken in the 10 min window entered on each peak. Each experiment has four replicates.

C. OD was maintained constant in the turbidostat during the experiment.

We next analysed the sfGFP kinetics using a microplate reader to assess circuit behaviour in a different physiological context. In Fig. S12, using 2 s of UV exposure, sfGFP‐ssrA is able to generate a short GFP pulse behaviour with fast degradation, 2.5 h are required for the emerging peak to return to ground states. In comparison, the original sfGFP showed similar pulse behaviour but did not return to zero within 16 h. Collectively, the PVRecA‐AT promoter induction in E. coli Nissle 1917 can be triggered in < 2h and the duration of the downstream expression is between 2.5 and 5 h and the strength of SOS expression can be adapted to reach significantly higher levels after intermittent stimulation and continuous cultivation.

Application of SOS biosensors to gut toxins and antibiotics

We applied the optimized SOS sensing system PVRecA‐AT to different DNA damaging compounds. One application is the detection of genotoxic biomarkers that are associated with colon cancer, which includes bile salt derivatives: deoxycholate, taurocholic acid (Nguyen et al., 2018; Sánchez, 2018) and other DNA damaging compounds related to diet such as acetaldehyde (Seitz and Becker, 2007). As shown in Fig. 7, the DNA damaging sensing circuit of PVRecA‐AT can detect genotoxic effects of 0.5 mM deoxycholate, which is a common secondary bile salt derivate metabolized by gut microbiome (Bernstein et al., 2011). The normal gut concentration of deoxycholate ranges from 450 ˜ 700 µM (Hamilton et al., 2007), high concentrations of deoxycholate have been correlated with colon cancer and could induce DNA breakage (Bernstein et al., 2011). Furthermore, the sensing system demonstrated a detection threshold of 9 mM taurocholic acid, which is within the range of estimated local concentration in the gut (Bernstein et al., 2005; Ridlon et al., 2016). This is the first report to directly quantify the genotoxicity of taurine conjugated bile salt. We proved that acetaldehyde (Na and Lee, 2017) (> 450 µM) is also detectable using PVRecA‐AT sensing system (Fig. 7C). It has been shown that bile salts are able to damage membrane and require bacterial membrane transport system to get into cells (Merritt and Donaldson, 2009; Begley et al., 2005). It might explain the delay of response in deoxycholate, taurocholic acid (Fig. 7A and B). Overall, by using the redesigned gene circuit of SOS sensing system carried by a probiotic‐based E. coli Nissle 1917, we were able to detect carcinogenic or genotoxic compounds. Our approach has clear merits compared to existing methods, which often rely on animal testing (Gad, 2014) and costly mammalian cell‐based assays such as AME test (Guy, 2014) or UMU assay (Hamer et al., 2000).

Fig. 7

Application of PVRecA‐AT –sfGFP for the detection of bile salt derived carcinogens and other DNA damaging compounds: (A) deoxycholate, (B) taurocholic acid, (C) acetaldehyde and (D) mitomycin C. Each experiment has three biological and three measurement replicates.

To compare the strength of genotoxic effect of deoxycholate, taurocholic acid and acetaldehyde, we also examined the induction of mitomycin C, which is a widely used chemotherapeutic drug that causes cancer‐specific DNA cross‐link during cell replication (DNA Repair and Human Disease, 2006; Paz et al., 2012; Händel et al., 2016). A dosage guideline for mitomycin C was provided in Table S3, where the concentration can be varied between 12 µM and 24 µM depending on the target tumours. Within the testing concentrations of these carcinogenic compounds, there is no impact on the cell growth of E. coli Nissle 1917 with sensing plasmid PVRecA‐AT‐sfGFP (Fig. S13) except the case when deoxycholate is used as an inducer, where deoxycholate could serve as a carbon source for E. coli Nissle 1917 (Nzakizwanayo et al., 2015).

To further extend the application spectrum in food security, we have also tested several fluoroquinolones derived antibiotics (levofloxacin, enrofloxacin and ciprofloxacin) which represent some of the most commonly prescribed antibiotics for bacterial infections (Marchant, 2018). Despite their efficacy and high penetrability, these antibiotics show severe carcinogenic effects on mammalian cells, including irreversible nerve damages and tendon rupture (Herbold et al., 2001; Anchordoquy et al., 2019). Although the prescription of fluoroquinolones to human patients has been warned against by the FDA (WHO, 1998), fluoroquinolones are still commonly used in animal feedstock which poses an enormous risk in food security.

Existing methods to detect the presence of fluoroquinolone commonly involved micellar electrokinetic chromatography (MEKC) (Prutthiwanasan and Suntornsuk, 2018), high‐performance liquid chromatography (HPLC) (He and Blaney, 2015) and mass spectrometry (MS) (Zhang et al., 2018). These instruments lack portability and are costly in use. In contrast, our optimized PVRecA‐AT sensing system could detect fluoroquinolone down to 0.97 µg l‐1, which is far below the usage limit of fluoroquinolones for animal feedstock (1.5–15 mg l‐1) (Subbiah et al., 2016) (Fig. 8 and Fig. S14). Similarly, metronidazole is another type of antibiotic that is also heavily used in animal feedstock and veterinary. However, the PVRecA‐AT sensing system showed no response, which might be due to the fact that the cytotoxicity of metronidazole only arises following reduction by the liver (Dingsdag and Hunter, 2018). The OD graph (Fig. S14) confirms our suggestion that metronidazole does not affect cells growth, but all three other antibiotics were found to stall cell growth at a concentration higher than 125 µg l‐1.

Fig. 8

Application of PVRecA‐AT –sfGFP for the detection of the genotoxicity effect of some antibiotics: (A) ciprofloxacin, (B) enrofloxacin, (C) levofloxacin and (D) metronidazole. (E) Dosage responses of different antibiotics. Each experiment has three biological and three measurement replicates.

Discussion

In this paper, we analysed the recA promoter in V. natriegens, identifying the sigma binding sites (−35 and −10) and SOS box in order to develop an ultrasensitive, rapid and robust DNA damage sensor. Although there is no direct evidence to support a correlation between SOS response time and bacteria growth rate, several studies reported links between SOS and DNA replication (Burby and Simmons, 2020), modulating growth pattern (Kreuzer, 2013) and growth fitness (Erill et al., 2007). We found that recA promoter PVRecA from the fast‐growing V. natriegens has higher sensitivity and broader dynamic range than that of E. coli PERecA. We then engineered the PVRecA promoter through a rational and systematic approach to reduce its background noise and enhance its sensitivity. Compared to conventional ad hoc method, our approach provided an example of bottom‐up engineering. Based on the structure of the LexA protein and its quantified interaction with the promoter (Kozuch et al., 2020; Culyba et al., 2018), we identified the precise location of the SOS box and tuned and optimized the SOS box to create a PVRecA‐AT promoter with low leakiness and broader dynamic response. This engineering process also showed that the SOS box is a modifiable biological part that could be fitted into the synthetic gene circuit design process (Ceroni et al., 2010).

An ideal synthetic gene circuit should have the capacity to withstand stochastic and environmental perturbations while maintaining a high signal‐to‐noise ratio. To improve our circuit’s robustness, we employed a previously reported Jungle Express system (Ruegg et al., 2018) which could incorporate our optimal SOS sequence. Meanwhile, by using a HrpRS‐based genetic amplifier system, we managed to increase the detection sensitivity and signal amplitude significantly with a minor increase of the system response time. To improve the stability of our optimized PVRecA sensing circuit, we incorporated a negative feedback system to achieve a more predictable operation upon changes in the environment. Together, we proved that our DNA damage sensing circuit confers similar robustness in response to changes in temperature, nutrient level, oxygen availability and pH. In essence, we engineered a functional SOS responsive operational amplifier by combining a noise canceller, a signal amplifier and a negative feedback system within a synthetic gene circuit. The merit of this DNA damage sensing system is not merely the potential to perform under real‐world stochastic environments such as the human gut, but also the engineering room for creating user defined dynamics and potential chimeric promoter, these features allow such circuits to be used in a plug‐and‐play fashion.

We demonstrated that our DNA damage sensing system could be used as an optogenetic actuator through the induction of UV light. In the experiment where we exposed the macro‐colonies transformed with our DNA damage sensing system with UV light, we showed that a 10‐s temporal difference could alter the phenotype of the macro‐colonies. As a proof‐of‐concept, our sensing system can be extended as a valuable tool to study signalling dynamics, generation of persisters and acquisition of antibiotic resistance within a biofilm (Goldman and Travisano, 2011; Yin et al., 2019). For instance, using our system to monitor the UV damage and its corresponding biofilm formation. Despite the system activation being different to canonical optogenetic systems such as the Cph8/OmpR system and BphP1(Liu et al., 2018) which commonly involve a light‐sensing domain, we proved that our ultrasensitive DNA damage sensing system is adaptable and functional upon UV induction without altering the growth of the bacterial host. Although the UV dose of PVRecA_AT‐sfGFP gene circuit was increased 128‐fold in 40 h, no mutation on the plasmid PVRecA‐AT‐sfGFP and PVRecA‐AT‐sfGFP‐ssrA was found during the entire course of the experiment (Fig. S15), according to DNA sequencing. It has been reported that bacterial gene regulation systems can be adapted in response to the environment without DNA mutation (Ibarra et al., 2002). This further verifies the potential of our sensing system as a programmable medicinal tool, strains with preloaded PVRecA‐AT‐sfGFP can be trained to extend its operational dynamic as ‘field‐ready’ strains with different characteristics. Another merit of using irradiation over molecular triggers as it is not restricted by diffusion limitations. Nevertheless, since our sensing system uses transcriptional machinery to detect UV light, the operational timescale could be longer than translational sensors such as a riboswitch (Chang et al., 2018; Wang et al., 2020) or two components system with membrane receptors (Chang et al., 2018).

Finally, we demonstrated the potential of this optimized PVRecA‐AT promoter for its real‐world application. The key application of this sensing system is to detect carcinogens, such as genotoxic compounds remaining in agricultural or dairy products, and secondary metabolites generated by the human microbiome. Our DNA damage sensing system can potentially be further developed into a Point of Care detection device or used as a programmable vehicle for living medicine. Compared to currently existing tools for carcinogen detection such as the Comet Assay (Fairbairn et al., 1995) and the Ames (Bacterial reverse mutation) tests (Hamel et al., 2016), our sensing system offers a cost‐effective, fast and easy‐to‐operate solution for such measurements. (Fairbairn et al., 1995).

Experimental procedures

Each experiment for the performance of gene circuits has at least three biological replicates and three measurement replicates.

Bacterial strains, media and chemicals

Escherichia coli DH5α was routinely used for molecular cloning and plasmid maintenance. E. coli BL21(DE3) was used for the initial screening and characterization of the genotoxic biosensor. Final optimized constructs were tested in probiotics strain EcN (E. coli Nissle 1917), obtained from ardeypharm GmbH (Herdecke, Germany). Enrofloxacin, metronidazole and levofloxacin were acquired from Adamas‐beta, Shanghai, China. Ciprofloxacin was acquired from TCI, Shanghai, China. Unless otherwise stated, all chemicals used were from Sigma‐Aldrich (Dorset, UK).

For routine cell growth, bacteria were inoculated from a colony and grow in Luria–Bertani (LB) broth with corresponding antibiotics kanamycin (60 μg ml‐1); ampicillin (100 μg ml‐1) and incubated at 37°C, 250 rpm for 16 h. For double‐plasmid transformation, equal molar of the two plasmids was electroporated into competent cells at the same time, followed by using an LB cocktail with kanamycin (50 μg ml‐1) and ampicillin (100 μg ml‐1) for inoculation and bacteria culturing over 16hrs.

For the expression experiments, M9 medium was made from M9 5X minimal salts (Sigma‐Aldrich) supplemented with 0.4% w/v glucose and corresponding antibiotics. SOC medium was made according to CSH protocol (cshprotocols.cshlp.org).

Gene circuits design and plasmid construction

All molecular procedures were carried out with enzymes obtained from New England Biolabs (NEB), and all primers were synthesized by Sigma‐Aldrich. For initial screening and annotation of the promoter region, the randomized promoter library was constructed using randomized oligos (NNNNNNNNACAGTATAATAACTTTCATTGCTGAGCG&NNNNNNNNACAGTGTCTATACCTGTATAGAAAAACTTTAGC) using the template (PV. RecA_sfGFP), followed by Q5 SDM protocol. A mixture of sfGFP plasmids with the randomized promoter (SOS box) was then transformed in E. coli BL21 competent cells. Colonies with different degree of background leakiness were screened using VersaDoc® imaging system by Bio‐Rad. Kinetics reads were followed by an overnight bacterial culture of the colony of interests. For each transformed mutant, plasmids were purified from QIAprep Spin Miniprep Kit and the sequence of promoter and SOS box region was determined via ‘TubeSeq Service’ provided by Eurofins Genomics. The quality of DNA was routinely checked with agarose gel electrophoresis (1% w/v) and NanoDrop (260/280 and 260/230 ratio ). Plasmids used in this study are shown in Table S1, and detailed plasmids map are provided in Table S4. For standard cloning procedure, gene fragments were synthesized via GeneArt with the backbone of pMK( KanR ColE1 ori), RBS (B0032) and sfGFP used in this study are standard parts obtained from Registry of Standard Biological Parts. (www.partsregistry.org). The v.recA binding site and promoter sequence were determined by analysing the sequences of v.recA operon (the GenBank accession No. CP016351.1). Similarly, control E.coli K12 recA operon was retrieved from GenBank: No. CP025268.1and for E. coli Nissle 1917 GenBank No. CP007799.1.

Bacterial transformation

To make chemically competent E. coli Nissle 1917 cells, bacteria were first grown to exponential phase by inoculating overnight culture in fresh LB broth at 1:100 dilution and incubated at 37°C, 200 rpm for 2 h or until OD600 of 0.3 or 0.4 was reached. Cell culture at exponential phase was chilled on ice for 30 min followed by centrifugation at 3500 rpm, 4°C for 10 min. The supernatant was removed, and the cells were washed twice with ice‐cold CaCl2 (0.1 M) and resuspended in CaCl2 (0.1 M) and glycerol (20% w/v).

For transformation, competent cells (100 μl) were added with a plasmid (100 ng) and mixed by flicking. The cell suspension was chilled on ice for 30 min and transferred to a 42°C water bath for 45 s, followed by incubation on ice for 2 min. SOC media (900 μl) was added and mixed, and the cell suspension was incubated at 37°C, 200 rpm for an hour before plating on kanamycin (60 μg ml‐1) agar plate.

DNA damage induction by mitomycin C, UV, cancer biomarkers, genotoxic antibiotics

Evaluation of the genotoxicity biosensor was done in Corning Black 96‐well plate (flat, clear bottom): Overnight bacteria culture of 5 μl was added into 190 μl medium containing M9 minimal media (1x), glucose (0.4 %w/v) and kanamycin (60 μg ml‐1), and 5 μl mitomycin C at concentrations ranging from 0.05 to 250 μM. For UV induction, 5 μl overnight bacteria culture was added into 195 μl medium containing M9 minimal media (1x), glucose (0.4 %w/v) and kanamycin (60 μg ml‐1), followed by 6W, 302 nm UV exposure from 0 to 10 s using Benchtop UVP Transilluminators (Thermo Scientific UK). The plate was incubated in a BioTek plate reader at 37°C for 16 h with OD600 and luminescence or GFP reading being measured every 15 min. All experiments were carried out in three biological repeats with three technical replicates each time. Error bars represent the mean and standard deviation, respectively. The processed data were plotted with OriginPro Version 9.1 and analysed using in the built statistical function of prism 8. Unpaired t‐test was carried out for comparing baseline within each variant. Significantly different P < 0.05 was used.

UV induction in turbidostats

For turbidostat experiments, we employed the Chi. Bio experimental platform57 with culture volumes of 20ml. M9 minimal media (1x), 5 mM MgSO4•7H2O, 0.01%w/v thiamine, glucose (0.4 %w/v) and kanamycin (60 μg ml‐1) were used, and experiments were run until substantial biofilms were observed to form in the culture chamber. Fluorescence measurements of sfGFP were performed each minute using excitation/measurements of 457/510 nm, respectively. Cultures were grown to an OD of 0.3 and maintained via Turbidostat functionality for ˜ 5 h prior to the first UV dose in each experiment. This first UV dose (i.e. 1X) was equal to activating a 10 mW (optical power) 280 nm LED for 2.4 s. In each subsequent dose, the duration of LED activation was doubled. Peak height in Fig. 6 is defined as the difference between the maximum GFP reading in the 5 h following a UV dose and the minimum UV reading in the preceding 5 h (e.g. since the previous UV dose). For the sfGFP‐ssrA experiment, fluorescence peaks were indistinguishable from measurement noise below UV doses of 16X.

Macro‐colony growth and UV induction

For biofilm growth, 100 μl of transformed overnight culture was aliquoted onto LB agar with 50 μg ml‐1 kanamycin supplemented. This was allowed to grow into a circular macro‐colony (˜2cm diameter) upon 16 h incubation at 37°C. Subsequently, the fully developed biofilm was exposed to 302nm UV for 10 and 20 s. Fluorescent reads were taken 4 h after the exposure using Molecular Imager VersaDoc imaging system by Bio‐Rad (Kidlington, UK). Light micrographs were subsequently analysed using ImageJ version 1.50b, with fluorescent maxima automatically counted after consistent thresholding.

Funding Information

WEH acknowledges support from EPSRC (EP/M002403/1 and EP/N009746/1).

Conflict of interest

The authors claim no conflict of interest.

Click here for additional data file.

Click here for additional data file.