Engineered acetoacetate‐inducible whole‐cell biosensors based on the AtoSC two‐component system

Whole‐cell biosensors hold potential in a variety of industrial, medical, and environmental applications. These biosensors can be constructed through the repurposing of bacterial sensing mechanisms, including the common two‐component system (TCS). Here we report on the construction of a range of novel biosensors that are sensitive to acetoacetate, a molecule that plays a number of roles in human health and biology. These biosensors are based on the AtoSC TCS. An ordinary differential equation model to describe the action of the AtoSC TCS was developed and sensitivity analysis of this model used to help inform biosensor design. The final collection of biosensors constructed displayed a range of switching behaviours at physiologically relevant acetoacetate concentrations and can operate in several Escherichia coli host strains. It is envisaged that these biosensor strains will offer an alternative to currently available commercial strip tests and, in future, may be adopted for more complex in vivo or industrial monitoring applications.


| INTRODUCTION
Bacterial cells have evolved the ability to detect and respond to environmental cues. The repurposing of these natural abilities to create engineered biosensor strains has become a major area of research within synthetic biology. Whole-cell biosensors, which contain this sensing machinery within a single living cell, offer a number of advantages over more conventional sensing technologies including portability, self-replication, sensing of the bioavailable fraction (rather than total concentration) and low-cost of production (Courbet et al., 2015;Hicks et al., 2020;Jung et al., 2018;Wang et al., 2019). To date, whole-cell biosensors have been developed for the detection of metals in environmental samples (mercury, Mahbub et al., 2017;arsenic, Stocker et al., 2003;lead, Jia et al., 2018), for the monitoring of metabolites during bioproduction (lactate, Goers et al., 2017) and in medical applications for the monitoring of clinically relevant biomarkers (Daeffler et al., 2017;Landry et al., 2018;Lin et al., 2019;Riglar et al., 2017;Woo et al., 2020), exploiting the ability of bacteria to reach inaccessible areas of the host in vivo (Courbet et al., 2015). Whole-cell biosensors can be based on a number of bacterial sensing mechanisms. These include factors, and other environment responsive mechanisms (Hicks et al., 2020;Wen et al., 2020).
TCSs are one of the most common bacterial sensing mechanisms, with many examples found within bacterial genomes (Hoch, 2000;Stock et al., 2000). Escherichia coli alone are reported to have 32 TCSs (Stock et al., 2000). TCSs typically consist of a membrane-bound histidine kinase (HK) that detects a specific signal. This signal activates kinase activity and causes autophosphorylation of a conserved histidine residue. The phosphoryl group is then transferred to an aspartate residue on the response regulator (RR) (Hoch, 2000).
Usually RRs are transcriptional regulators capable of activating or repressing expression from a specific promoter (Mascher et al., 2006), thereby controlling gene expression in response to an external input.
The AtoSC TCS is sensitive to acetoacetate and within E. coli plays a role in short-chain fatty acid (SCFA) metabolism, motility and poly-(R)-3-hydroxybutyrate synthesis (Oshima et al., 2002;Pauli & Overath, 1972;Pilalis et al., 2011;Theodorou et al., 2013). As shown in Figure 1, acetoacetate is detected by the membrane-bound AtoS histidine kinase (AtoS HK). This then goes on to phosphorylate the AtoC response regulator (AtoC RR). The phosphorylated AtoC RR then induces transcription from the Pato promoter, which subsequently expresses the atoDAEB operon.
Acetoacetate is an extremely important metabolite in mammalian metabolism and energy regulation. It is one of three ketone bodies, alongside β-hydroxybutyrate and the less abundant acetone, which can serve as an alternative energy source in the body during periods of low glucose availability (Puchalska & Crawford, 2017). Typically ketone bodies are found at submillimolar concentrations within the blood; however, these levels can become elevated during periods of extended starvation or intense exercise (Laffel, 1999). Ketone bodies have been linked to protective effects on the neural system, which has led to the use of ketogenic diets as a method for preventing epileptic seizures (Neal et al., 2008). Another study highlighted the role ketone bodies play in stem cell differentiation and homeostasis within the intestines, through the action of β-hydroxybutyrate (Cheng et al., 2019). Human mesenchymal stem cells have been shown to have a preference for acetoacetate as an energy-yielding substrate; leading to suggestions that acetoacetate could be added to mesenchymal stem cell culture medium (Board et al., 2017). Acetoacetate has also been shown to act as a signalling molecule for muscle regeneration and can help restore muscle function in a mouse model of muscular wasting (Zou et al., 2016). However, sustained periods of increased ketone body concentrations within humans can be a sign of several pathological states. These include salicylate poisoning, alcoholic ketoacidosis and diabetic ketoacidosis (Laffel, 1999). During extreme cases of diabetic ketoacidosis ketone body concentrations can reach 20 mM and above. If left untreated, this may lead to complications such as a cerebral edema (Kanikarla-Marie & Jain, 2016;Puchalska & Crawford, 2017). As such, ketone body levels (particularly acetoacetate and β-hydroxybutyrate) are monitored regularly in patients with diabetes. Alongside their role in diabetes, ketone bodies are also measured in the blood of cows.
During pregnancy, cows become susceptible to ketosis, a condition that adversely affects the health and milk production of dairy cattle (Enjalbert et al., 2001).
It is clear therefore that methods of measuring ketone body levels may find use in a range of biomedical or agricultural applications.
Currently, strip tests that use a sodium-nitroprusside reaction, or ketone meters, can be used to monitor ketone levels in urine and blood (Misra & Oliver, 2015). Although these offer cheap monitoring systems for patients, they are only semiquantitative, and there are a F I G U R E 1 Layout of the AtoSC twocomponent system and atoDAEB operon. The AtoS histidine kinase, within the Escherichia coli inner membrane, autophosphorylates in the presence of acetoacetate. The phosphate group is then transferred to the AtoC response regulator, which triggers expression from the Pato promoter number of biomedical, scientific, and industrial applications where they are unsuitable. For example, monitoring the production of acetoacetate by engineered E. coli (Hu et al., 2010), exploring how the human microbiota utilises ketone bodies (Sasaki et al., 2020) and investigating host-microbiota interactions in model systems in vivo (Rutter et al., 2019), could all use whole-cell biosensors to report on acetotacetate concentration.
However, biosensors often display observed biosensor behaviour that is mismatched with the desired application, as such there is demand for design strategies that can be used to modify biosensor behaviour without relying on time-consuming trial and error approaches (Gonzalez-Flo et al., 2020). For transcription-factor based biosensors numerous principles have been identified that govern biosensor response, such as modulation of receptor protein expression, via approaches including design of experiments and phenomenological modelling (Berepiki et al., 2020;Chen et al., 2018;Mannan et al., 2017). Further studies have attempted to rationally design more complex TCS-based biosensors. One example used a mathematical model to show that tuning phosphatase activity was able to modify the K 1∕2 of TCS biosensors, where the HK displays both phosphatase and kinase activity (Landry et al., 2018). A further study systematically varied expression of components from a CcaSR TCS biosensor (through varying RBS strength), allowing the authors to optimise biosensor behaviour based on the plasmid library created (Sasaki et al., 2014). Within this study we apply sensitivity analysis to guide the rational design of a TCS biosensor.
Here, we report the construction of E. coli whole-cell biosensors that can be used to detect and report on the presence of acetoacetate. The whole-cell biosensors presented here are based on the AtoSC TCS, found within E. coli. We develop an ordinary differential equation (ODE) model that attempts to capture the action of the AtoSC TCS and use sensitivity analysis of this model to guide the design of an array of acetoacetate-inducible biosensors. The final biosensors display a diverse range of output responses and may be employed in the future for a variety of sensing applications.

| AtoSC plasmid construction
The plasmids within this study were constructed using the CIDAR MoClo assembly method (Iverson et al., 2016). The oligonucleotides and plasmids designed in this study are given in Tables S1 and S2, respectively, alongside details of their construction.   (

| AtoSC ODE modelling and sensitivity analysis
) Parameter sampling followed the method detailed in Campolongo et al. (2007). Specifically, we generated 1000 trajectories and sampled 50 optimal trajectories from this set. Analysis of the sensitivities to each of the Hill function variables, and additionally fold change (f f ∕ max min ), to each of the model parameters was performed using the SALib implementation of the Morris method (Campolongo et al., 2007). Plotting was performed in R (Core Team, 2021) using ggplot2 (Wickham, 2016), ggrepel (Slowikowski, 2021), and dplyr (Wickham et al., 2021). The code to reproduce and plot results can be found in a Zenodo repository: doi 10.5281/zenodo.5078479. The equations and parameter bounds used for this model are given in Supporitng Information Section 2.

| Growth curve assays
Overnight bacterial cultures were diluted to an approximate optical density 700 (OD 700 ) of 0.05, within fresh media. A volume of 120 µl of each culture was then added to the well of a 96-well clear bottom microplate (Greiner Bio-One) with a magnetic removable lid. The cultures were incubated in a Tecan Spark plate reader for 2 h at 37°C with shaking at 150 rpm (2 mm amplitude, double orbital) with an OD 700 measurement taken every 30 min. Cultures were then induced with the relevant inducer concentration and sealed with a breathe-easy permeable membrane (Diversified Biotech). The plates were then returned to the plate reader and left to grow for a further 16 h, with OD 700 measurements taken every 20 min (the same temperature and shaking conditions were maintained). Inducers were added to the 96-well plates with the aid of an I.DOT liquid handler (Dispendix). Inducers were diluted to the desired stock concentrations and then custom protocols set up within the I.DOT, using the I.DOT assay studio software, to automate addition of the correct inducer volumes. Taking the total volume in each well to 125 µl.

| Dose-response assays
Bacterial cultures, grown overnight in LB media, were diluted to an OD 700 of approximately 0.05 in fresh media, and 190 µl added to each well of a polypropylene 96-well deep-well plate (Brand, Sigma Aldrich). The plate was then sealed with an autoclaved system Duetz lid and incubated for 2 h at 37°C, with 350 rpm shaking (Duetz et al., 2000). Inducers were then added to the desired concentration, bringing the total volume in each well to 200 µl, the plate was once again sealed with an autoclaved system Duetz lid and the induced cultures incubated at 37 ∘°C , with 350 rpm shaking for 16 h. Samples were taken as required for flow cytometry analysis.
Results of the sensitivity analysis conducted on the AtoSC model. (a) Layout of the AtoSC model, illustrating the parameters varied during sensitivity analysis of the system. (b) Illustration of the different aspects of biosensor performance for which sensitivity analysis was performed, ymin and ymax represent f min and f max , respectively. K represents the K 1∕2 of the biosensor. (c) Heatmap denoting the normalised μ* results of the sensitivity analysis for the AtoSC model parameters, a higher μ* value indicates higher parameter influence on biosensor behaviour. "C" refers to AtoC RR concentration, "k_ad" to HK autodephosphorylation, "k_b1" to binding rate of the phosphorylated HK and RR, "k_b2" to the binding of HK to the phosphorylated RR, "k_b3" to the binding of the HK and RR, "k_bnd" to the binding of phosphorylated RR to the Pato promoter, "k_d1" to unbinding of phosphorylated HK and RR, "k_d2" to unbinding of HK and phosphorylated RR, "k_d3" to unbinding of HK and RR, "k_lgexp" to the rate of transcription from uniduced Pato, "k_ph" to dephosphorylation of phosphorylated RR by HK, "k_pmgexp" to the rate of transcription from induced Pato, "k_pt" to the phosphorylation of RR by phosphorylated HK, "k_unbnd" to the unbinding of phosphorylated RR to Pato, "pato" to Pato promoter availability and "S" to AtoS HK concentration (as given in Table S3) 2.6 | Flow cytometry analysis Flow cytometry was performed on an Attune NxT Acoustic Focusing Cytometer, with an Attune NxT Autosampler (Thermo Fisher Scientific), closely following a previously reported method (Rutter et al., 2019). In brief 1 µl, of the appropriate sample was diluted 1:200 in sterile phosphate-buffered saline (PBS), in a shallow polystyrene U-bottom 96-well plate. An Attune NxT Autosampler was used to record 10,000 events (for each sample) with three wash and mixing cycles between samples. GFP was excited using a blue laser (488 nm) and detected using a 530/30 nm bandpass filter. Additionally, a sample of 1:300 dilution of rainbow calibration particles in sterile PBS (Spherotech) was recorded allowing for the conversion of arbitrary units to molecules of equivalent fluorophore (MEF), using R commands from the FlopR package (Fedorec et al., 2020).
All collected flow cytometry standard (FCS) data were processed using commands from the FlopR package (Fedorec et al., 2020) and plotted using custom R scripts. Visualisation and curve fitting were performed in R, using the "ggplot2" package and "nls" fitting function (Wickham, 2016). GFP induction data was fit using a Hill function, as given in Equation (1) WCB showed no substantial change in growth when exposed to increasing concentrations of acetoacetate ( Figure S2).

| Specificity of the acetoacetate-inducible biosensor
After demonstrating that the AtoSC TCS could be used to construct acetoacetate-inducible biosensors (Figure 3b), alternative inducers were examined to check the specificity of the ASAH0 circuit. Six alternative inducers were tested, including the SCFAs acetate, propionate and butyrate, the other ketone bodies, acetone and βhydroxybutyrate, and spermidine (which has previously been reported to act as an inducer for the AtoSC TCS) (Theodorou et al., 2007). All were tested at a concentration of 20 mM, a concentration above that shown to give maximum acetoacetate induction. When characterised in LB media, acetoacetate produced the highest increase in GFP expression (Figure 3d), while the alternative inducers showed similar levels of GFP to the uninduced control. As 20 mM is a relatively high concentration, growth assays were measured to ensure the lack of GFP was not caused by toxicity of any of the inducers. All inducers produced no substantial changes in growth | 4283 compared to the uninduced control, except for spermidine ( Figure S4), which was seen to prevent cell growth. To further examine spermidine induction of BW25113 ASAH0, a full response curve was collected across a range of inducer concentrations ( Figure S5). However, no increase in GFP was seen across this range of spermidine concentrations.
Characterisation was also performed for the BW25113 ASAH0 biosensor in several different media types ( Figures S6 and S7). First, the switching of BW25113 ASAH0 in minimal media with glycerol and glucose carbon sources was compared to that in LB media ( Figure S6). The biosensor functioned across all three media types, although the fold change was greatly reduced with glucose as the carbon source. In addition, to determine whether the ASAH0 biosensor was capable of operating in more complex conditions, the biosensor was characterised in media supplemented with a range of mammalian culture media and HeLa/CHO-K1 conditioned samples ( Figure S7). BW25113 ASAH0 was found to show induction when exposed to acetoacetate in cultures supplemented with 20% of these complex samples. F I G U R E 4 A comparison of the acetoacetate biosensor response between genomic and plasmid expression of the AtoS and AtoC genes. (a) Layouts of the ASAH0 and ASAH2J06 plasmids. ASAH2J06 contains constitutive expression of the AtoS HK and AtoC RR proteins. (b) Median GFP fluorescence of the whole-cell biosensors exposed to acetoacetate (BW25113 ASAH0 and BW28878 ASAH2J06: n = 3 biological repeats, data fit with Hill function, dashed lines indicate the fitted K 1/2 values, BW28878 ASAH0: n = 2 biological repeats, line indicates lowest measured GFP median, cell symbols indicate the atoS and atoC presence status in the host chassis genome for each of the three whole-cell biosensor strains). HK, histidine kinase; RR, response regulator were kept the same. The layout of this biosensor is given in Figure 4a.

| Plasmid versus genome expression of AtoS/ AtoC components
In contrast to ASAH0, the ASAH2J06 biosensor produced an acetoacetate-inducible change in GFP expression even when host expression of both the atoS and the atoC genes were knocked-out ( Figure S8). To explore the effect of plasmid versus host gene expression in more detail the responses of ASAH0 in BW25113 (only host expression of AtoS/AtoC) and ASAH2J06 in BW28878 (only plasmid expression of AtoS/AtoC) were directly compared ( Figure 4b). BW25113 ASAH0 displayed a greater f max , dynamic range and lower f min than ASAH2J06. However, BW28878 ASAH2J06 showed a fitted K 1∕2 value an order of magnitude lower than the BW25113 ASAH0 biosensor (31.7 ± 3.78 vs. 344 ± 42.1 µM, respectively). The source of AtoS HK and AtoC RR expression were the only differences between these strains. However, these relatively simple changes produced a dramatic shift in the response curves of the BW25113 ASAH0 and BW28878 ASAH2J06 biosensors, supporting the predictions gained through sensitivity analysis of the AtoSC model.

| Engineering modified biosensor responses by varying AtoS, AtoC, and Pato concentrations
As the BW25113 ASAH0 and BW28878 ASAH2J06 biosensors displayed such different responses to acetoacetate induction we wished to explore if further changes to the concentrations of AtoS, AtoC, and the Pato promoter would result in other behaviours. To this end, a range of further AtoSC plasmids were produced. Low copy versions of both the ASAH0 and ASAH2J06 plasmids were created by replacing the origin of replications with a low copy SC101 origin. These low copy plasmids were designated ASAL0 and ASAL2J06, respectively. Furthermore, the atoC genes were removed from the ASAH2J06 and ASAL2J06 plasmids to produce the ASAH1J06 and ASAL1J06 plasmids, which have plasmid expression of the AtoS HK but rely on host expression of the AtoC RR. The response curves for all six of these biosensors are given in Figure 5.
Increasing the copy number of the plasmid containing only the Pato promoter produced a large increase in the f max of the final GFP response, while having no noticeable effect on f min (Figure 5b).
Increasing the copy number of the plasmid containing both Pato and the AtoS HK (ASAH1J06/ASAL1J06) reversed this trend, with the lower copy number displaying a higher f max . Again little difference was seen in f min (Figure 5d). Finally, increasing the copy number of the plasmid containing Pato, the AtoS HK and AtoC RR (ASAH2J06/ ASAL2J06) resulted in a vertical shift of the GFP response, increasing both f min and f max (Figure 5f). All of the fitted parameters to these response curves are given in Table S5. Of all the ato knockout (BW25113, BW28878, JW2213, and JW2214) whole-cell biosensor strains constructed, BW25113 ASAH0 displayed the largest dynamic range (114), whereas BW28878 ASAH2J06 was found to have the lowest K 1∕2 (31.7 ± 3.78 µM) followed by the low-copy BW28878 ASAL2J06 (40.1 ± 2.28 µM). From these response curves it is clear that introducing variations in the concentrations of AtoS HK, AtoC RR, and Pato promoter, through varying plasmid versus host genome expression and plasmid copy number, can produce dramatically

| Using the AtoSC TCS to create acetoacetateinducible biosensors
The ASAH0 plasmid is the simplest design for an acetoacetate sensitive biosensor that can be created from the AtoSC TCS, requiring host expression of the AtoS HK and AtoC RR. As shown in Figure S2, increasing exposure to acetoacetate did not appear to produce substantial changes in the exponential growth rate of BW25113 ASAH0. E. coli is naturally able to metabolise acetoacetate as a carbon source using genes contained within the ato operon. However, the growth curves collected here suggest that this metabolism is negligible, in LB media, as no increase in growth rate was seen.
ASAH0 was initially tested in three E. coli strains: NEB5α (a commercial cloning strain), BW25113 (the parent strain of the Keio knockout collection; Baba et al., 2006) and Nissle 1917 (a commonly used microbiome engineering strain; Ou et al., 2016). Although the ASAH0 design is simple and was found to be acetoacetate-inducible in all three strains, it provides fewer components that can be modified to tune biosensor performance. The aim of the ASAH2J06 design was to provide a self-contained plasmid that would (a) not rely on host expression of AtoS HK/AtoC RR, and (b) provide a greater potential design space from which to produce the desired biosensor behaviour. Upon construction both ASAH0 and ASAH2J06 were found to be acetoacetate-inducible ( Figure 4).
As mentioned previously, the typical concentrations of ketone bodies within a healthy adult are submillimolar. It has been reported that these may increase to approximately 1.0 mM during hyperke-  (Pauli & Overath, 1972;Theodorou et al., 2011). In addition, the two other ketone bodies, acetone and β-hydroxybutyrate were tested for induction; neither were found to induce the BW25113 ASAH0 biosensor. As discussed previously, spermidine has been reported elsewhere as an alternative inducer of the AtoSC TCS (Theodorou et al., 2007). Characterisation of BW25113 ASAH0 did not show an increase in GFP response when exposed to spermidine (Figures 3d and S5). However, it is possible that the spermidine concentrations needed to trigger GFP production from this plasmid may be above those that become toxic to the cells.
Future work could explore whether spermidine induction of the BW25113 ASAH0 biosensor can be seen in other culture conditions.
To further explore the switching response of BW25113 ASAH0, the biosensor was characterised in a range of more complex media components ( Figures S6 and S7). BW25113 ASAH0 displayed an increase in GFP upon exposure to acetoacetate in all the conditions tested; however, the fold change of the switching was found to be much lower when grown in M9 media with glucose as the carbon source ( Figure S6). Previous work has shown that the atoD gene (present at the beginning of the atoDAEB operon) is a target of the Hfq-binding small RNA, Spot 42, which suppresses numerous metabolic genes in the presence of a preferred carbon source (Beisel et al., 2012). This is a likely mechanism for the repression seen with the BW25113 ASAH0 strain. Future work could explore methods of reducing this catabolic repression within the engineered whole-cell biosensor strains.
A range of AtoSC biosensor strains, with differing expression of AtoS HK, AtoC RR, and Pato promoter were also characterised ( Figure 5). From the response curves it is clear that varying these parameters was able to produce a variety of different biosensor behaviours. This supports the results of the sensitivity analysis, which identified the concentrations of these parameters as important for governing behaviour of the biosensors. However, it should be noted that the agreement between in silico and in vivo results is qualitative and cannot be used to quantitatively predict the behaviour of the final biosensor design. Our knowledge of the regulatory interactions influencing the AtoSC TCS is incomplete and therefore our model does not capture the full biological complexity of this system.
In future, the model reported here will be adapted as our knowledge of AtoSC TCS biology improves, in turn helping to improve the predictive capability of our sensitivity analysis approach.

| CONCLUSIONS
Our study has demonstrated that the AtoSC TCS can be used to construct acetoacetate-inducible biosensors. Characterisation of the BW25113 ASAH0 whole-cell biosensor showed that the biosensor is specific to acetoacetate, with no GFP response recorded when exposed to a range of alternative inducer molecules. Sensitivity analysis performed on a model of the AtoSC TCS predicted that varying AtoS HK, AtoC RR, and Pato promoter presence would have a large effect on the final biosensor response. Subsequently, we constructed a range of AtoSC whole-cell biosensors with varying expression of AtoS HK, AtoC RR, and presence of the Pato promoter. These whole-cell biosensors were found to exhibit different response curves when exposed to acetoacetate, supporting the insights gained through sensitivity analysis. In the future, these AtoSC whole-cell biosensors may be incorporated into a range of other biomedical, bioproduction, or agricultural applications, where there is a need to monitor the levels of the acetoacetate ketone body. The design methods used here may also be applied to the construction of other TCS biosensors, where there is a mismatch between the desired and observed biosensor behaviour.