Abstract
Whole cell biosensors have great potential to be used as diagnostic and treatment tools in biomedical applications. Bacterial biosensors that respond to specific metabolites can help analyze spatio-temporal gradients in the body or cue expression of therapeutic agents in diseased sites. In this study, we developed an acetoacetate bacterial sensor by inserting the promoter of the atoSC two-component system (TCS) into a promoterless GFP expression plasmid and transformed it into E. coli DH5α and E. coli Nissle 1917, which both contain the atoSC TCS. This bacterial biosensor has a dynamic range of 0.01-1mM of acetoacetate, which is within the range of physiological concentrations in the blood, and exhibits up to a 100-fold change of induced GFP expression as measured by relative fluorescence. Comparing combinations among the two host strains and high/low-copy plasmid variations of the biosensor, we observed the fastest and highest response to acetoacetate with the E. coli Nissle 1917 low copy plasmid biosensor. Induction experiments on the biosensor using 50mM of the short chain fatty acids (SCFA) acetate, proprionate, and butyrate showed no response, indicating its specificity between acetoacetate and SCFAs. This acetoacetate bacterial sensor is a valuable contribution to the library of biosensors that may eventually be used for in vivo clinical applications.
1 Introduction
Natural sensing systems in bacteria detect external signals in the environment and relay these into the cell for appropriate responses. As an integral part of a complex biological system, these have evolved to be sensitive, specific, and robust, making them attractive to harness for technological applications. In the field of synthetic biology, bacteria are engineered into whole cell biosensors by combining three parts: (i) a membrane-bound or cytoplasmic sensor to detect the signal, (ii) an internal circuit to process signals, and (iii) an actuator to transduce the processed signal into an output [1]. Sensors can come in the form of two-component systems (TCS), cytoplasmic regulatory proteins, RNA riboswitches, and other environment-responsive sensors [2]. Whole cell biosensors can be used to detect harmful substances in the environment or monitor key metabolites in bioprocessing units. When the sensor is linked to fluorescence expression, real-time information of the system can be obtained quickly and inexpensively through flow cytometry [3]. In addition, biosensors can be integrated directly to metabolic pathways to control and regulate enzyme expression to improve titer yields [4]. For medical applications, whole cell biosensors can provide non-invasive and targeted therapeutic strategies when incorporated with sensors that respond to metabolites or conditions in diseased sites of the body [5, 6]. Although many challenges and limitations still remain for whole cell biosensors [7], the growing number, and improved methods in assembling internal genetic circuits, will help their translation into clinical applications.
In the present study, we developed and characterized an acetoacetate biosensor in E. coli. Acetoacetate is one of the three ketone bodies (acetone, acetoacetate, and β-hydroxybutyrate) produced in the liver. It serves as an alternative source of energy that is distributed through the bloodstream at concentrations between 0.02-3.0mM [8, 9, 10] when glucose levels are low. Ketone bodies also confer protective effects to the brain, which is why the ketogenic diet (high fat-low carbohydrate) is used to induce ketosis as a method of treatment against epilepsy [11]. However, excess ketone bodies accumulated in the blood from prolonged starvation, diabetes (lack of insulin), or excessive alcohol consumption can result in ketoacidosis [12], a harmful and extreme form of ketosis. Ketone body levels, in the form of acetoacetate or/-hydroxybutyrate, are monitored in urine and blood in diabetic patients to help control blood glucose and reduce incidence of complications. Ketone bodies are also monitored in the blood and milk of cows, who are susceptible to ketosis during pregnancy and can adversely affect health and milk production [13]. While strip tests reacting sodium nitroprusside with acetoacetate will likely be easier and cheaper in terms of blood and urine diagnostics, biosensors can be applied into more complex, fine scale, and real-time applications such as monitoring acetoacetate levels in the liver and gut to provide important information on the dynamics and spatial patterns of acetoacetate metabolism. For example, bacterial biosensors that can colonize tissues and microenvironments in vivo have already been demonstrated in the rat liver to detect tumors [5] and in the rhizosphere of plants to detect nitrate availability [14]. Bacterial biosensors that contain trigger and memory elements have also been developed in murine gut derived E. coli to show how genetic circuits can be used to monitor the complex gut environment [15].
We developed the whole cell biosensor using the atoSC two-component system, which is triggered by acetoacetate and is responsible for activating the genes necessary for short chain fatty acid (SCFA) and acetoacetate metabolism, poly-(R)-3-hydroxybutyrate biosynthesis, and chemotaxis in E. coli [16, 17, 18, 19]. The atoSC two-component system is composed of a membrane-bound homodimeric histidine kinase sensor, atoS, which detects acetoacetate in the environment and auto-phosphorylates. AtoS then passes on the phosphoryl group to atoC, which dimerizes and binds onto the two 20bp palindromic DNA stretches (Pato) upstream the atoDAEB operon to promote σ-54-dependent transcription [16, 20, 21, 22, 23] (Fig. 1). To construct the acetoacetate biosensor, the Pato promoter region was inserted into a high-copy promoterless GFP expression plasmid and transformed into E. coli DH5α, which contains the atoSC two-component system within its genome. Characterisation experiments were performed to determine the dynamic range of acetoacetate concentration and fold-change expression of GFP in the biosensors. These experiments were monitored through the plate reader to obtain growth curves and normalized aggregate population fluorescence time courses, and through flow cytometry to obtain fluorescence intensity population distributions of the induced cultures. The whole cell biosensor was shown to be able to detect acetoacetate in a range of 0.01-1mM and exhibit up to a 100-fold change in induced GFP expression. Additionally, we demonstrated the sensor to work in a low-copy plasmid and in E. coli Nissle 1917, a human probiotic strain which has been widely used in clinical settings and microbiome engineering studies [24, 25, 26, 27].
2 Materials and Methods
2.1 Bacterial strains and media
The bacterial strains and plasmids used in this study are listed in Table 1. Both E. coli DH5α and E. coli Nissle 1917 were confirmed to carry the atoSC two-component system by PCR (Fig. S1). Primers used were developed from a whole genome assembly of E. coli DH5α [28]. All bacterial cultures were carried out in 5ml liquid Luria-Bertani (LB) broth and incubated at 37°C and 180rpm in a shaking incubator. Antibiotics were added at a final concentration of 50μg/ml kanamycin when indicated. Media was supplemented with acetoacetate, in the form of lithium acetoacetate solution (Sigma-Aldrich), during acetoacetate induction experiments. Chemically competent cells for all bacterial strains were prepared by CaCl2 method and then stored at 80°C until used for chemical transformation.
2.2 Construction of plasmids
OG241 (Oxford Genetics), a pUC high-copy (500-700 copies/cell) promoterless GFP expression plasmid, was used as a backbone plasmid for the Pato promoters. Different variations of promoters were PCR amplified and extended with restriction enzyme sites from the genomic DNA of E. coli DH5α and inserted upstream the GFP gene to the multiple cloning site (MCS) of OG241 by restriction enzyme digest and ligation (Figs. S2-4). PatoG-OG241 contains the Pato promoter region up until the start codon of atoD to continue in-frame with GFP. PatoK-OG241 is the same as PatoG-OG241 except that the GFP is out of frame with the start codon. PatoL-OG241 con-tains the Pato promoter region up until before the probable native ribosomal binding site (RBS) and continues with Bba_B0034, a strong RBS from the iGEM registry (www.parts.igem.org), and then the start codon of GFP. Oxb19-OG241 contains a strong constitutive promoter, Oxb19 (Oxford Genetics), upstream of GFP. All promoter regions of the plasmids were confirmed by capillary sequencing. Low copy versions of plasmids OG241, Oxb19-OG241, PatoL-OG241, and PatoM-OG241 were made by switching the pUC origin of replication region with SC101, a low-copy (2-5 copies/cell) origin of replication (Figs. S5-6). These plasmids were labeled as LOG241, for low copy OG241. The plasmids were then transformed into E. coli DH5α and E. coli Nissle 1917 for growth and induction experiments. Further details are documented in the Supplementary Material. Plasmids are available at Addgene (www.addgene.org, Plasmid IDs 72838-72842 for the high-copy plasmids and 72844-72848 for the low-copy plasmids).
2.3 Growth curves and flow cytometry induction assays
The plasmid transformed bacteria were cultured in 5ml LB broth with 50μg/ml kanamycin overnight (12 hours) at 37°C and then diluted 1:100(v/v) into fresh 5ml LB broth with 50μg/ml kanamycin. The diluted cultures were incubated for 1 hour and then induced with varying concentrations of acetoacetate (0mM, 1e-3mM, 1e-2mM, 1e-1.5mM, 1e-1mM, 1e-0.5mM, 1mM, and 10mM). After induction, 200μl of the cultures are aliquoted into a clear bottom, black walled 96-well plate (Grenier), sealed with a gas permeable membrane, and placed into a plate reader (BMG FLUOstar OPTIMA). The plate reader was set at 37° C with shaking (150rpm, 4mm double orbital), absorbance (540nm), and fluorescence (480nm excitation, 520nm emission) readings every 10 mins. The gain used for fluorescence readings was 2,000 and all plate reader readings were taken by 3mm diameter orbital averaging. The remainder of the induced cultures were incubated further for 15 hours for flow cytometry analysis. flow cytometry samples were prepared by pelleting 50μl of the cultures by centrifugation (1 min, 13,000xg), aspirating out the supernatant, washing with 500μl phosphate-buffered saline solution (PBS), and then resus-pending in fresh 500μl PBS. Samples were analyzed on a flow cytometer (BD Acuri C6, 22μm core) using the slow mode (14μl/min) to record FL1 filter readings (533/30nm for GFP) for 10,000 events. For time series experiments, flow cytometry samples were aliquoted, prepared, and analyzed from the incubating cultures every 1 hour. All experiments were done in triplicate using separate bacterial colonies.
2.4 Data analysis
Growth curve results were processed by taking the natural logarithm of the absorbance values and plotting ln (OD540) against time. Relative fluorescence units (RFU) values were normalized by dividing with OD540. Ratios of normalized RFU values of induced and uninduced cultures were used to compare the induction responses of the different strains and plasmid copies over time in the plate reader. Growth curves were modelled using a modified version of the Gompertz model [30, 34] where N0 is the initial optical density, K is the carrying capacity, μ is the maximal growth rate and λ is the lag phase time. Flow cytometry characterisation curves were created by plotting the density of log10(no. of events) vs. uorescence intensity, log10 (relative uorescent units, RFU). Characterisation curves were modelled using a Hill function where I is the median RFU, [X] is the concentration of acetoacetate and Pmin, Pmax, Kd and n are the minimum fluorescence, maximum fluorescence, dissociation constant, and Hill coefficient, respectively. Both models were initially fitted using maximum likelihood followed by a Bayesian analysis using Markov chain Monte Carlo (MCMC) [35] in the R statistical software environment [36]. Further details of the model fitting are provided in the Supplementary Material.
3 Results
3.1 Induction of high-copy plasmids in E. coli DH5α
Initially, the biosensors were tested for inducibility at 0mM and 10mM acetoacetate. E. coli DH5α and E. coli DH5α harboring OG241 and PatoK-OG241 showed no GFP expression by acetoacetate induction. E. coli DH5α harboring Oxb19-OG241 constitutively expressed GFP. E. coli DH5α harboring the plasmids PatoG-OG241, PatoL-OG241, and PatoM-OG241 were observed to be acetoacetate inducible (Fig. S8). Growth curves and specific growth rate (μ) distributions showed no apparent difference that could be caused by the additional carbon source from acetoacetate. E. coli DH5α harboring PatoL-OG241 was also used to test induction by short chain fatty acids acetate, proprionate, and butyrate, but showed no induction at 50mM for each SCFA (Fig. S10).
3.2 Dynamic range of high-copy plasmids in E. coli DH5α
The dynamic ranges of the E. coli DH5α with high copy PatoG-OG241, PatoL-OG241, and PatoM-OG241 were determined by acetoacetate induction at concentrations between 0mM and 10mM. GFP expression was induced starting after 1e-2mM acetoacetate for all three biosensors and was saturated at 1e-0.5mM for PatoG-OG241 (Fig. 3A) and 1mM acetoacetate for both PatoL-OG241 (Fig. 3C) and PatoM-OG241. Within their dynamic ranges, GFP expression, as measured by RFU intensity, increased >100-fold for PatoG-OG241 and >10-fold for PatoL-OG241 and PatoM-OG241. Hill function fits (Figs. 3B, 3D, and S27-29) showed that the medians of the parameters Pmin, Pmax, Kd and n were 708.90, 13,749.18, 0.159, and 2.832 for PatoG-OG241, 557.42, 4,598.02, 0.221, and 1.684 for PatoL-OG241, and 589.67, 2,855.60, 0.228, and 1.639 for PatoM-OG241 in E. coli DH5α, respectively.
3.3 Induction and temporal response of low-copy plasmids in E. coli DH5α
Low-copy versions of the OG241, Oxb19-OG241, PatoL-OG241, and PatoM-OG241 biosensors were constructed using LOG241 (SC101 origin of replication) in place of OG241 as the backbone plasmid. The PatoL-LOG241 and PatoM-LOG241 in E. coli DH5α biosensors also showed inducibility with acetoacetate, but was observed to have a heterogeneous population with respect to fluorescence intensity (Figs. 4B and S8K-L) as compared to the high-copy version (Figs. 4A and S8E-F). The time series of induced cultures showed that the high-copy PatoL-OG241 in E. coli DH5α had one homogeneous population shifting in fluorescence intensity from 3-16 hours (Fig. 4A). On the other hand, the low-copy PatoL-LOG241 in E. coli DH5α showed two separate populations at low and high fluorescence intensity. Along the succeeding hours after induction, the low fluorescence population decreases while the higher fluorescence population increases. Additionally, similar to the high-copy PatoL-OG241, there is an observed general shifting towards higher fluorescence in the low-copy PatoL-LOG241 (Fig. 4B).
3.4 E. coli Nissle 1917 acetoacetate sensors
PatoL-OG241 and PatoL-LOG241 plasmids transformed into E. coli Nissle 1917 also exhibited acetoacetate inducibility. Growth curves for both plasmids in E. coli Nissle 1917 (Figs. 5A & 5C) reached their stationary phases in a shorter amount of time and specific growth rates were higher as compared to E. coli DH5α strains (Figs. 2A & 2C). Specific growth rate (μ) medians of PatoL-OG241 in E. coli Nissle 1917 for uninduced and induced cultures were 0.00527 and 0.00438, respectively. Specific growth rate (μ) medians of PatoL-LOG241 in E. coli Nissle 1917 for uninduced and induced cultures were 0.00424 and 0.00341, respectively (Figs. 5A, 5C, S23-26). While the E. coli DH5α with PatoL-OG241 biosensors had a specific growth rate (μ) of 0.00164 and 0.00184 and E. coli DH5α with PatoL-LOG241 biosensors had a specific growth rate (μ) of 0.000972 and 0.000990 for uninduced and 10mM acetoacetate induced treatments, respectively (Figs. 2C, S15-16 and S19-20). The low-copy PatoL-LOG241 plasmid in E. coli Nissle 1917 also resulted in a heterogeneous population (Fig. 5D and S9F) similar to the low-copy biosensors in E. coli DH5α. In both cases of host strains, the high-copy plasmid resulted in a higher fold change of induction.
Induced/uninduced ratios of normalized RFU vs. time plots describe the response of the bacterial biosensors in culture to the acetoacetate inducer. Comparing the high-copy PatoL-OG241 (Fig. 6A) and low-copy PatoL-LOG241 (Fig. 6B) implementations in E. coli DH5α, the low-copy biosensor results in larger standard deviations for the replicates. In E. coli Nissle 1917, the low-copy PatoL-LOG241 biosensor (Fig. 6D) results in a faster time to a maximum RFU induction ratio as compared to the high-copy biosensor PatoL-OG241 (Fig. 6C).
4 Discussion
We constructed a whole bacterial cell biosensor for acetoacetate and demonstrated it to specifically respond to concentrations of acetoacetate within its physiological range in human blood. The biosensor had a dynamic range between 0.01mM to 0.10mM for PatoG-OG241 and 0.01mM to 1.0mM for PatoL-OG241 and PatoM-OG241 in E. coli DH5α. Physiological acetoacetate concentrations in human blood ranges in between 0.02-3mM [8, 9, 10]. The use of the entire native Pato region until the start codon of the ato operon in PatoG-OG241 showed a greater response in fluorescence output as compared to the PatoL-OG241 and PatoM-OG241 biosensors by approximately 10-fold (Figs. 3A, 3C, and S27-29). This may be due to transcriptional or translational regulatory mechanisms because PatoL-OG241 and PatoM-OG241 differ with PatoG-OG241 only between their transcriptional start site and the start codon, including the RBS. The median magnitudes of the Hill coefficient (n), representing the cooperativity of the system, are lower for PatoL-OG241 (1.684) and PatoM-OG241 (1.639) compared to PatoG-OG241 (2.832) in E. coli DH5α. The lower Hill coefficients result in a wider dynamic range against acetoacetate induction concentrations. The different behaviors resulting from biosensor variations show how sequence elements can alter the properties of the biosensor. These variations can be used to adjust the dynamic range or response output of the sensor appropriate for the application.
Although the atoSC two-component system induces the atoDAEB operon necessary for SCFA metabolism, the acetoacetate bacterial biosensor was specific towards acetoacetate only and not acetate, proprionate, or butyrate at concentrations of 50mM (Fig. S10). This is in agreement to previous studies that linked acetoacetate as the inducer of the atoDAEB operon. Wild type E. coli was observed to be able to grow on acetoacetate as a sole carbon source, but not with SCFAs. However, the mutant E. coli strain containing atoCc, which caused constitutive expression of the atoDAEB operon, was able to grow on butyrate as a sole carbon source [17, 18] - indicating that the atoSC two-component system is one of the regulatory mechanisms for SCFA metabolism. Spermidine was also observed to increase poly-(R)-3-hydroxybutyrate synthesis, which is regulated by the atoSC two-component system and the atoDAEB operon. However, whether spermidine directly induces atoS autophosphorylation in a similar manner to acetoacetate, or one of its metabolic products somehow influences the atoSC two-component system, is unknown [31]. The acetoacetate biosensor developed in this study could be used to further investigate the mechanism of spermidine on the the atoSC two component system.
Induction experiments of high and low copy biosensors showed a difference of population fluorescence heterogeneity. In E. coli DH5α, the high copy PatoL-OG241 biosensor had one homogeneous population throughout induction (Figs. 4A & 5B) and the low copy PatoL-LOG241 plasmid had two distinct subpopulations with either low or high fluorescence (Figs. 4B & 5D). In the low copy biosensor, the low fluorescence population decreased and the high fluorescence population increased as time after induction progressed. These subpopulations could be a result of the competition for the transcription factor atoC between the two Pato promoters present in the GFP expression plasmid and the genome of E. coli. In high copy plasmids, the probability of a phosphorylated atoC binding onto the Pato promoter on PatoL-OG241, as compared to the genomic Pato, is much higher. A similar heterogeneous behavior in a low copy reporter plasmid was also observed in an arabinose-inducible bacterial sensor where a copy of the inducible system was also present in the chromosome of the host E. coli. This resulted in two distinct subpopulations of induced cells, as opposed to the homogeneous population observed in the arabinose transport-deficient strains harboring the reporter plasmid [32].
The acetoacetate sensor worked in both E. coli DH5α and E. coli Nissle 1917 but performed differently on each host and plasmid vector. Growth rates of E. coli DH5α were slow due to mutations in its genome that cause inferior growth phenotypes [33]. Median specific growth rates (//) ranged between 0.0014-0.0022 for E. coli DH5α while 0.0035-0.0053 for E. coli Nissle 1917. The slow growth rate of E. coli DH5α affects the response time of the biosensor on acetoacetate, which begins to have significant induced GFP expression at the exponential phase up until the stationary phase. E. coli Nissle 1917 grows faster and reaches its stationary phase more quickly. The high copy plasmid in E. coli DH5α had a more apparent rise in normalized RFU ratios between induced and uninduced cultures as compared to its low copy strain (Figs. 6A & 6B), while the low copy plasmid in E. coli Nissle 1917 exhibited a faster increase of normalized RFU ratio compared to the high copy strain (Figs. 6C & 6D).
This acetoacetate bacterial sensor is a valuable addition to the steadily growing library of biosensors. Whole cell biosensors will eventually be able to help open up research opportunities to monitor biological systems, whether in an industrial microbiology setting or within in vitro and in vivo biomedical applications. The gut environment is of special interest for whole cell biosensors to monitor gut health. Stable populations of bacteria can be established in the gut microbiota and be used for long term diagnostics of specific metabolites in the gastrointestinal tract [6, 15].