To advance environmental health and hazard detection, researchers have developed whole-cell bioelectronic sensors by engineering extracellular electron transfer to be dependent on an analyte1. However, these sensors regulate a single electron transfer pathway as an electrochemical channel, limiting the sensing information to a single analyte. We have developed a multichannel bioelectronic sensor where different chemicals regulate distinct extracellular electron transfer pathways within a single Escherichia coli cell. One channel utilizes the flavin synthesis pathway from Bacillus subtilis2 and is controlled by a cadmium-responsive promoter. Another channel, the Mtr pathway from Shewanella oneidensis3, is controlled by an arsenite-responsive promoter and activates cytochrome CymA expression4, 5. We exploit the differing redox potentials of the two extracellular electron transfer pathways6 to develop a redox-potential-dependent algorithm that efficiently converts biological signals into 2-bit binary outputs. This enables our bioelectronic sensor to detect and differentiate heavy metals at EPA limits. When deployed in complex environmental water samples with lower electroactivities, our sensor effectively and accurately encodes 2-bit binary signals across various analyte conditions. Thus, our multichannel bioelectronic sensor advances the field through simultaneous detection of different chemicals by a single cell, significantly expanding information transmission and helping to safeguard human and environmental health.
Article
Multichannel bioelectronic sensing using engineered Escherichia coli
https://doi.org/10.21203/rs.3.rs-4853950/v1
This work is licensed under a CC BY 4.0 License
Version 1
posted
You are reading this latest preprint version
Environmental water contamination, stemming from both natural and synthetic chemicals, poses a significant global challenge. While sophisticated analytical techniques such as ion chromatography7 or atomic absorption spectroscopy8 can provide accurate results even in complex samples, they rely on manual collection, controlled storage, and long-distance transportation of water samples back to a laboratory. These preparation and sampling procedures are expensive and time-consuming9, thereby limiting the frequency of analysis. This lack of temporal resolution is prone to missing short-lived events, such as periodic releases of toxins or pollutants, which can have significant consequences to environmental and human health.
In contrast to lab-based analytical techniques, in situ chemical sensors allow for more frequent monitoring, even in remote locations10. However, this convenience comes at the expense of selectivity in complex conditions11, 12. To enhance the selectivity of in situ chemical sensors, preliminary sample separation steps, such as chromatography and electrophoresis are incorporated13. Nevertheless, these preparation steps require substantial power and bulkier instrumentation.
Biosensors, which rely on biomolecular recognition, are an alternative to chemical sensors. They can be highly selective, portable, and have minimal power requirements14. Enzymatic biosensors stand out due to their remarkable selectivity for a specific substrate and rapid response time15. However, enzymatic biosensors face certain drawbacks, such as the high cost associated with producing purified enzymes and the short operational lifespan16–19. In contrast, whole-cell biosensors exploit the self-maintenance and self-replication capabilities of living microorganisms, resulting in lower costs and longer lifespans20–23.
Bioelectronic sensors, a type of whole-cell biosensor, employ electroactive microorganisms to transduce chemical signals into electrical signals1, 24. One transduction process that enables microbes to produce electrical signals is extracellular electron transfer (EET), a metabolic process by which electrons from intracellular donors are transferred to extracellular acceptors. Synthetic biology allows us to construct sensing components to control the EET process, enabling bioelectronic sensing of desired targets25. Recently developed bioelectronic sensors are capable of sensing endocrine disruptors with response times1 on the minute scale. Yet, since they employ only one EET pathway within the whole cell sensor, they are limited to the detection of a single analyte. Monitoring environmental water quality requires the simultaneous measurement of multiple parameters for a complete understanding of contamination, thus necessitating the engineering of multiplex sensing capabilities into bioelectronic sensors.
To address this need for increased sensing information, we developed a bioelectronic sensor with two EET pathways in a single strain of Escherichia coli. The system allows for the sensing of two common environmental pollutants, arsenite and cadmium simultaneously. By employing two distinct heavy metal responsive promoters, the two EET pathways are independently activated in response to the presence of arsenite or cadmium, leading to increases in current values. To distinguish between the four possible states based on the presence or absence of each analyte, we developed a redox-potential-dependent algorithm that effectively encodes the signals into 2-bit binary outputs corresponding to the input signals. The algorithm enables the detection of arsenite and cadmium at EPA limits. When utilizing our bioelectronic sensor in complex environmental water samples, it effectively encodes 2-bit binary signals corresponding to four combinations of heavy metals.
To create a multichannel bioelectronic sensor, we were inspired by how light signals are sensed and processed by the color vision system26, 27 (Fig. 1 top). Visual light is composed of a spectrum of colors, each associated with a specific wavelength. These wavelengths are first sensed by three different receptors and converted into electrical signals, which then travel through separate channels to finally reach a shared bipolar ganglion cell. This cell performs an algorithmic computation on these signals through the opponent process and outputs the results to the visual cortex, which comprehends the mix of wavelengths as color28.
Similar to the color vision system, we engineered a single strain of E. coli with two EET pathways controlled by specific analytes, which transforms the biological responses into electrical signal outputs (Fig. 1 bottom). To achieve this, we constructed transcriptional regulation systems capable of sensing target analytes and activating the expression of distinct EET pathways. The EET pathways then conduct electrons from the cytoplasm to electrodes to generate electrical signals at different redox potentials, referred to as electrochemical channels. The signals from each electrochemical channel are recorded by a potentiostat-controlled electrode, processed by a voltammetric algorithm, and encoded into 2-bit binary signal outputs. Each digit corresponds to a signal or lack thereof from individual channels, and thus, the various combinations of two analytes can be represented by four different 2-bit binary signals.
Design strategy for EET pathways
As the essential signal transduction channels, the two EET pathways need to be engineerable, produce measurable current, and have distinguishable redox potentials. The CymA-limited MtrCAB pathway (notated as the CymA pathway) from Shewanella oneidensis and the flavin synthesis pathway from Bacillus subtilis were selected because both have been successfully engineered into E. coli2, 6, 29 and the activation of each individual EET pathway delivers measurably higher current signals than the basal signal associated with the wild-type E. coli. The CymA pathway uses an outer membrane protein to transfer electrons to the electrode, generating an electrical signal when the cells contact electrodes. Flavins, cycle between reduced and oxidized states to deliver electrons from the cell interior to the electrode, acting as redox mediators to facilitate this signal transmission. Additionally, the redox potentials (E1/2) associated with the CymA and flavin pathways are differentiable (0 V and − 0.4 V vs. Ag/AgCl)6, 29, 30. Thus, we chose these pathways to form the basis for our sensing channels.
Constructing Two Electron Transfer Pathways in E. coli as Sensing Channels
To enable multiplexed sensing, we employed two different types of signal receptors coupled to the EET pathways. We first selected commonly used inducible promoters to individually sense and activate the two EET pathways and refer to this strain as inducer-E. coli (Fig. 2A and 2D, second column; detailed gene circuits in Fig. S1 top). Specifically, an isopropyl β-d-1-thiogalactopyranoside (IPTG)-activated promoter regulates the expression of the cymA gene. Since CymA is the rate-limiting step along the CymA-MtrCAB pathway2, a constitutive promoter regulates the mtrCAB genes from Shewanella oneidensis29. An anhydrotetracycline (aTc)-activated promoter controls the ribADEHC genes responsible for flavin synthesis from Bacillus subtilis as the flavin synthesis pathway25. To demonstrate applicability in environmental monitoring4, 5, we incorporated heavy metal-responsive promoters to detect multiple metals simultaneously into a separate strain and refer to this as heavy metal-E. coli (Fig. 2A and 2D, third column; detailed gene circuits in Fig. S1 bottom). In the heavy metal-E. coli, a modified arsenite-regulated promoter is used to activate cymA expression in response to the presence of As5, while flavin synthesis is regulated by a cadmium-inducible promoter to sense cadmium presence4. The subsequent electrical signals delivered via the different EET pathways are thus referred to as the CymA-Channel and the Flavin-Channel.
To optimize the expression of the individual pathways, we assayed the EET activity of the CymA pathway. To assess the expression of the CymA pathway, we incubated the inducer-E. coli or the heavy metal-E. coli with cell-impermeable molybdenum-based nanoparticles, a more sensitive particle for reporting on EET than commonly used WO331. With this assay, the nanoparticles undergo an electrochromic shift from white to blue based on the level of reduction by the CymA-Channel. (The blueness of the assay represents the expression level of CymA, as shown in Fig. S2). The cells were induced at different levels with IPTG (0 µM to 2500 µM) or arsenite (0 µM to 80 µM), and the color change of the nanoparticles was recorded (Fig. S3 and S4). The EET performance of the CymA-Channel is substantially enhanced when induced at 100 µM IPTG (p = 0.016) or 2.5 µM arsenite (p = 0.002) compared to non-induced cells (Fig. 2B). Thus, for all subsequent studies (unless stated otherwise), we induced cells with 100 µM IPTG or 2.5 µM arsenite to maximize EET activity while minimizing toxicity to cells.
The activation of the flavin synthesis pathway was characterized as a function of induction level using the intrinsic fluorescence properties of flavins (Fig. S4). The production of extracellular flavins peaked with induction at 4 µM aTc and was substantially improved after induction with 0.8 µM cadmium (Fig. 2E). Thus, for all subsequence experiments on the flavin synthesis pathway, we induced cells with 4 µM aTc or 0.8 µM cadmium.
To ensure our engineered heavy metal-E. coli can specifically detect target analytes without interference from other substances in environmental monitoring, we evaluated its specificity by exposing it to various metal ions (Fig. S5). The relative expression level of the CymA pathway was only positively activated by arsenite (Fig. S5A), while the extracellular flavin concentration showed a substantial increase in the presence of cadmium (vs. Co, p = 0.004, Fig. S5B). These results showed that our engineered bioelectronic sensor specifically responds to either arsenite or cadmium.
After optimizing the two EET pathways via optical measurements, we inoculated the engineered strains into bioelectrochemical systems (BES) to assess their electrical signal production. To do so, we inoculated the engineered E. coli at OD600 of 0.5 into the working electrode chamber, polarized the working electrode at 0.2 V vs. Ag/AgCl, allowed the current to plateau, and recorded the steady-state current (jss) as the signal output. Exposure of engineered E. coli to the target analytes–either IPTG or aTc, and arsenite or cadmium–led to an increase in the steady-state current (jss) through individual channels (Fig. 2C, 2F). For the inducer-E. coli, the non-induced jss was 7.6 µA.cm− 2, serving as the control value. When cells were exposed to 100 µM IPTG, which activates CymA expression, the jss increased to 12.0 µA.cm− 2 (Fig. 2C, middle). Similarly, when cells were exposed to 4 µM aTc, which enhances the extracellular flavin concentration, the jss increased to 11.5 µA.cm− 2 (Fig. 2F, middle).
To evaluate our biosensor for heavy metal detection, we exposed heavy metal-E. coli across various combinations of arsenite and cadmium. The non-induced jss was 3.2 µA.cm− 2, serving as the control value. When cells were exposed to 2.5 µM arsenite, which activates CymA expression, resulted in a higher jss of 8.1 µA.cm− 2 (Fig. 2C right). Similarly, when cells were exposed to 0.8 µM cadmium, which enhances extracellular flavin concentrations, resulted in a higher jss of 4.3 µA.cm− 2 (Fig. 2F right). Interestingly, the inducer-E. coli synthesizes less flavins compared to the heavy metal-E. coli, while achieving a higher jss (Fig. 2E, 2F). This is likely due to the high basal activity of the IPTG promoter, leading to a higher basal current response from inducer-E. coli32. Consequently, when the Flavin-Channel was activated by aTc, the inducer-E. coli produced a greater jss. In summary, when exposed to individual target analytes, the engineered strains delivered higher signal outputs compared to the control condition without analytes, demonstrating effective analyte detection by the individual channels.
To determine whether our engineered E. coli functions as a two-channel biosensor, we inoculated the inducer-E. coli and heavy metal-E. coli under various combinations of analytes (none of the analytes, either of analytes, or both of analytes) and monitored currents at a constant potential (Fig. S6). After the current reached the plateau, we measured the steady-state current (jss) across all conditions. However, the similarity in jss presented a challenge in distinguishing between each condition. Taking the inducer-E. coli as an example (Fig. 2G), jss generated from cells induced with 100 µM IPTG was 11.0 µA.cm− 2, closely matching the 11.5 µA.cm− 2 from cells activated by 4 µM of aTc alone. We then measured expression levels of the CymA and flavin synthesis pathways for the similar jss, showing that the similar signal output was due to different individual EET pathway activities: IPTG presence enhanced CymA expression, while aTc boosted flavin concentrations (Fig. S7). Similar current values were also observed for the heavy metal-E. coli: the jss from cells activated with 0.8 µM cadmium alone was similar to the basal current (Fig. 2H). These similarities in jss present challenges for multiplex sensing, underscoring the need to develop an advanced data encoding process for improved signal differentiation.
Developing a Redox-Potential-Dependent Algorithm for Signal Processing
To distinguish between the signals from different analyte conditions, we developed an algorithm that relies on the distinct redox potentials (E1/2) associated with the two EET channels. The redox potential of the CymA-Channel is more positive than that of the Flavin-Channel. Therefore, we employed double potential step chronoamperometry (DPSC) to individually evaluate and encode the oxidation process of the CymA-Channel and the reduction process of the Flavin-Channel. After measuring jss, we applied two different potentials to create two distinct redox conditions: one close to the midpoint potential of each EET pathway as the baseline, and the other to promote either oxidation dominance in the CymA-Channel (Fig. 3A) or reduction dominance in the Flavin-Channel (Fig. 3B). The corresponding current (j) for each potential was recorded for thirty minutes (Fig. 3C and 3D).
For the CymA-Channel, we recorded the baseline current at 0 V (Fig. 3C, dashed line) and the oxidation current at 0.2 V (Fig. 3C, solid line). As expected, the baseline current from heavy metal-E. coli was similar regardless of the presence or absence of arsenite. However, the oxidation current was higher from cells exposed to arsenite, measuring 10.2 µA.cm− 2, compared to 3.9 µA.cm− 2 in the absence of arsenite. This higher current is attributed to the overexpression of the CymA pathway activated by arsenite (Eqn. S6). Consequently, the higher ratio of oxidation current to baseline current correlated with the overexpression of CymA pathway (Eq. 1), indicating the presence of arsenite. Similarly, for the Flavin-Channel, we recorded the baseline current at − 0.2 V (Fig. 3D, dashed lines) and the reduction current at − 0.4 V (Fig. 3D, solid lines). Again, as expected, the baseline current from heavy metal-E. coli was similar regardless of the presence or absence of cadmium. However, the absolute value of the reduction current was higher from cells exposed to cadmium, measuring − 2.1 µA.cm− 2, compared to − 0.5 µA.cm− 2 in the absence of cadmium. This increase is attributed to enhanced extracellular flavin production via the flavin synthesis pathway (Eqn. S6), activated by cadmium. Therefore, the higher ratio of reduction current to baseline current correlated with the overexpression of flavin synthesis pathway (Eq. 2), indicating the presence of cadmium. By normalizing current ratios to the background condition in the absence of analytes, we obtained the relative current ratios (RCymA and RFL) for each channel (Eq. 3, 4). These relative current ratios are depicted as the function of time in a heatmap, where an increase in the darkness of the colors corresponds to increased relative current ratios and expression levels of each EET pathway (Fig. 3E and 3F). The CymA-Channel is depicted in pink, while the Flavin-Channel is presented in blue.
We observed a strong correlation between the presence or absence of analytes and the color patterns displayed on the heatmaps (Fig. 3E, 3F). For both the CymA-Channel and Flavin-Channel, the presence of analytes (marked as ‘+’) resulted in higher current ratios, leading to darker shades in the heatmaps. Considering the assessment of specificity, which is shown in Fig S5, the relative overexpression via both arsenite- and cadmium-responsive regulation resulted in about a 1.2-fold increase compared to the non-induced control. We set a threshold of 1.2 for both CymA-Channel and Flavin-Channel, assigning a ‘1’ when the ratio reached the thresholds (RCymA ≥ 1.2, RFL ≥ 1.2). Conversely, in the absence of analytes (marked as ‘–’), lower current ratios were calculated, and thus lighter colors were observed on the heatmaps and assigned a ‘0’ (Fig. 3E, 3F). Hence, a 2-bit binary signal configuration was implemented wherein four states are possible. The first digit indicates the signal associated with the CymA channel, specifically responding to the presence or absence of IPTG or arsenite. The second digit encodes the signal attributed to the Flavin channel in response to the presence or absence of aTc or cadmium. This configuration allows for the encoding of information related to the activation of the two EET channels concisely and distinctly.
Moreover, this procedure can be optimized into a shorter time frame for analyte detection. For the CymA-Channel, the algorithm takes about 10 min to detect the presence of IPTG and only 1 min to recognize arsenite (Fig. 3E and 3F, pink heatmaps). For the Flavin-Channel, it takes 2 minutes to detect the presence of aTc and 3 minutes to recognize cadmium (Fig. 3E and 3F, blue heatmaps). This fast detection allows a shorter polarization time of each potential instead of 30 min. For the rest of DPSC assessments, we shortened the measurement time for each potential to 15 min.
Detection Limits and Concentration-Response Signals
With the ability to differentiate signals at single concentrations of arsenite and cadmium, we proceeded to determine the lower detection limit of our sensor. We evaluated the current ratios (RCymA and RFL) for each channel by applying DPSC across various concentrations of arsenite and cadmium. To determine the detection limit of arsenite, we assessed RCymA and RFL by exposing heavy metal-E. coli to arsenite concentrations ranging from 0.1 µM to 2.5 µM. The RFL values remained below the set threshold of 1.2, consistent with the absence of cadmium (Fig. 4A). In contrast, an increase in RCymA was observed in response to the presence of As. Notably, RCymA reached 1.2 at an arsenite concentration of 0.1 µM, which is below the EPA limit. This ratio meets the pre-determined threshold, encoding a '1' to signify the detection of arsenite, with statistical significance (p = 0.035) compared to the control without arsenite. To assess the detection limit for cadmium, we evaluated the current ratios of RCymA and RFL by exposing heavy metal-E. coli to cadmium concentrations ranging from 0.045 µM to 0.8 µM (Fig. 4B). The RCymA values stayed below the established threshold of 1.2, consistent with the absence of arsenite. Conversely, a significant increase in RFL was noted in response to cadmium presence. Notably, at a cadmium concentration of 0.045 µM—which is the EPA's maximum allowable level for cadmium in drinking water—RFL reached 1.97. This value exceeds the threshold to encode a ‘1’, indicating the detection of cadmium, with statistical significance (p = 0.0046) compared to the control without cadmium. These results show our sensor enables the separate detection of heavy metals at EPA limits.
Detection of Heavy Metals in Environmental Samples
To investigate the ability of our sensor to function in complex environmental water samples, we compared our sensor in a laboratory medium and a water sample from Brays Bayou in the Houston area. Compared to the laboratory medium, which has a neutral pH of 7.05 and a low solution resistance of 1.5 kΩ, the environmental sample exhibits a slightly higher pH of 7.73 and a significantly higher solution resistance of 5 kΩ (Fig. 5A).
To assess the signal delivery capabilities of our sensor across different water samples, we measured the steady-state currents (jss) in response to arsenite and cadmium under four conditions: none, either arsenite or cadmium, and both. The jss was collected at a constant potential of 0.2 V with glycerol as the carbon source for both samples (Fig. 5B). Overall, jss responses in environmental samples were substantially lower compared to those in the laboratory medium. This decline in jss could be attributed to the high solution resistance in environmental samples. Additionally, distinct jss were recorded across the four analyte conditions in the laboratory medium, whereas similar jss were noted in environmental samples. Therefore, measurements of jss from the heavy metal-E. coli cannot enable multiplexed sensing in environmental samples.
To test whether our redox-dependent algorithm could accurately detect arsenite and cadmium, we first exposed the heavy metal-E. coli to only cadmium in both laboratory medium and environmental samples and applied the redox-dependent algorithm to convert the resulting signals to binary signal outputs (Fig. 5C). As described earlier, we recorded the baseline (j0V) and oxidation currents (j0.2V) to calculate the RCymA value. Notably, significant variability was observed among biological replicates in both baseline and oxidation currents. The variation in oxidation currents, from maximum to minimum values, was 2.4 µA.cm− 2, accounting for about 53% of the average value. In contrast, this variability was markedly lower in the laboratory medium, amounting to only 8% of the average values (Fig. S8). Despite these variations, the calculated RCymA in environmental samples remained close to the average value at 1.01, which is below the threshold of 1.2. This indicated the absence of arsenite and reported as '0' for the CymA-Channel. For the Flavin-Channel, we polarized the electrode at − 0.2 V and − 0.4 V and recorded the baseline (j− 0.2V) and reduction currents (j− 0.4V) to calculate the RFL. A larger variation was observed in the reduction currents (j− 0.4V), while the baseline currents (j− 0.2V) remained relatively consistent. As a result, the standard deviation of the RFL values was larger, but the average RFL value, 2.06, substantially exceeded the threshold of 1.2, indicating the presence of cadmium and showing as '1' for the Flavin-Channel. Combining the individual encodings into a binary output of ‘01’ effectively corresponds to cadmium being present in the environmental samples.
With this result in hand, we applied the redox potential-dependent algorithm to detect four combinations of arsenite and cadmium using the heavy metal-E. coli in the environmental samples, as shown in Fig. 5D. As before, we distinguished signals from the CymA-Channel and the Flavin-Channel by using DSPC to determine RCymA and RFL. For the CymA-Channel, RCymA values below the 1.2 threshold (1.01 and 1) indicate the absence of arsenite, encoding as '0'; whereas values reaching or exceeding the threshold (1.24 and 1.20) indicate the presence of arsenite, showing as '1'. Similarly, for the Flavin-Channel, RFL values below 1.2 (1.11 and 1) indicate the absence of cadmium, showing as '0'; and values above the threshold (1.81 and 2.06) indicate the presence of cadmium, showing as '1'. Therefore, the combination of the heavy metal-E. coli and the redox potential-dependent algorithm effectively allows multiplexed detection of all four combinations of heavy metals in the environmental samples.
Expanding the information capacity of individual bioelectronic sensors instead of relying on multiple sensors could greatly save on cost and space in advanced sensing applications. To achieve this, we successfully engineered two different EET pathways, the CymA and the Flavin synthesis pathways, into E. coli simultaneously. Each pathway was individually activated by promoters responsive to either the inducers IPTG or aTc or the heavy metals arsenite or cadmium, resulting in strains containing two analyte-sensitive electrochemical channels. The current response generated from the two EET pathways when recording at a single potential could not reliably distinguish the analytes. To solve this, we developed a redox potential dependent algorithm that uses DPSC to extract the relative current ratios of each channel and predefined thresholds to convert the ratio to a binary readout. In this way, a simple binary readout is generated from each channel that records the presence of one analyte, the other analyte, or both analytes. This approach allows for the detection of contaminants at the EPA-prescribed detection limits. This facilitates accurate monitoring of multiple analytes, even in complex environmental samples.
In addition to expanding the sensing capability to four states in a single strain, we expect our redox-potential-based algorithm to work across different analytes or EET pathways. The strains could be re-engineered with different promoters to sense other analytes, opening up a range of sensing targets, such as small molecules33, physical changes34, and nucleic acids35, 36. Moreover, the encoded signals involve both directed and mediated electron transfer processes, highlighting their versatility across various electron transfer mechanisms. We envision that our algorithm could also be employed in systems with more than two channels as long as the redox potentials of the EET pathways are distinguishable. This expands the range of various EET-associated molecules or proteins that could serve as channels, such as OmcS37, phenazines38, and quinones39. By integrating variable sensing components and multiple channels with our algorithm, the capabilities of whole-cell bioelectronic sensors could be significantly expanded.
However, one limitation of the current approach is that the response time of our sensor is quite slow compared to other bioelectronic sensors. Arsenite-responsive promoters need 2–4 hours to activate fluorescent proteins to deliver sufficient signals5. When such a promoter was used to regulate the native Mtr pathway in Shewanella oneidensis, detectable peak currents appeared after 2 days, with a 40 µM arsenite detection limit, which is a magnitude higher than the EPA limit40. These factors may contribute to the slow response time of the CymA-Channel. For the Flavin-Channel, cadmium-responsive promoters need 8–12 hours to reach maximum fluorescent signals4. Additionally, producing and accumulating maximum extracellular flavins from engineered E. coli takes about 48 hours2. These factors may contribute to the slow response time of the Flavin Channel. Our previous bioelectronic sensors were able to detect an endocrine disruptor within 5 minutes by regulating the EET process post-translationally1. Therefore, integrating multichannel bioelectronic sensors with post-translational regulation systems could greatly shorten the response time. By leveraging allosteric conformational changes to deliver sensing signals, this approach could potentially create a reusable biosensor capable of switching on and off in response to the presence or absence of specific target analytes.
Our study expands the boundaries of bioelectronic sensing. The two transcriptionally controlled electrochemical channels impart more information than previously designed bioelectronic sensors, and the overall platform offers the flexibility to integrate diverse sensing systems and EET processes. This technology has applications for dynamic monitoring in complex conditions, such as polluted water environments and gastrointestinal tracts. By providing valuable insights into these complex systems, this technology can aid in effective environmental management strategies, disease diagnosis, and personalized healthcare interventions.
By combining synthetic gene circuit design and electron transport in bacterial cells with electrochemical signal data processing, we developed a multichannel bioelectronic sensor capable of in situ sensing of multiple contaminants in environment samples over a timescale of days. Using heavy metal responsive promoters in urban water as a proof-of-concept model system, we demonstrate a flexible strategy for gene circuit design, signal processing, and develop a universal encoding program for signal output.
ACKNOWLEDGEMENTS
The authors acknowledge the CAF Lab members for helpful discussion and Shyam Bhakta for input on the modular cloning approaches. We acknowledge the invaluable assistance of Jayashree Soman, Biki Kundu, and Matt Carpenter in proofreading the manuscript. Xu Zhang and Marimikel Charrier acknowledge support by the Cancer Prevention and Research Institute of Texas (RR190063 to C.M.A-F).
AUTHOR CONTRIBUTIONS
C.M.AF contributed to the conceptualization, funding acquisition, supervision, writing, and revising of the manuscript. X.Z contributed to the investigation, conceptualization, visualization, methodology, writing, and revising of the manuscript. M.C. contributed to the analysis, writing, and revising of the manuscript.
COMPETING INTERESTS
The authors declare no competing interests.
Plasmids assembly and transformation. A list of all plasmids with primers and strains used in this study is provided in Supplementary Table S1. E. coli 5-α was used for all plasmid construction and amplification. The promoter sequences are provided in Table S2. The key primers and overhang sequences used for cloning are shown in Table S2. The plasmids were constructed through Gibson assembly. All assembled plasmids were verified using Sanger DNA sequencing. For experiments using cells, plasmids were transformed into chemically competent E. coli BL21(DE3). E. coli was then plated on LB-agar plates containing a combination of 50 mg/L kanamycin and 34 μg/mL chloramphenicol and incubated overnight at 37 °C.
Culturing conditions. We first inoculated all the E. coli strains directly from −80 °C glycerol stocks in 5 mL LB media with antibiotics (50 μg/mL kanamycin and 34 μg/mL chloramphenicol (Sigma-Aldrich, US) and grew them at 37 °C overnight with 250 rpm shaking. To cultivate a large volume of cells for subsequent assay measurement or a bioelectrochemical characterization, cells were subcultured in 2xYT media with the same antibiotic concentrations and 1 mM δ-aminolevulinic acid (Sigma-Aldrich, US), then grown at 37 °C with 250 rpm shaking.
MoO3 Nanoparticle assay to assess metal reduction by CymA pathway. After subculturing in 2xYT media until reaching OD600 of 0.5-0.6, cells were harvest at 4000 × g at 4 °C. The cells were then washed twice with M9 minimal medium and resuspended in M9 minimal medium with 1% glycerol to an OD600 of 1. Resuspended cultures were then mixed at a 1:1 volume:volume ratio with 10 mg/mL molybdenum trioxide (MoO3) nanoparticles (Sigma-Aldrich) and transferred into a 96-well plate with a range of inducer concentrations. The plate was incubated for 2 hours at 30 °C in a 5% H2/10% CO2/85% N2 atmosphere. We recorded the color of the wells using a CanoScan LiDE 220 instrument, and the blueness of wells was determined using ImageJ with the ReadPlate3 plugin29.
Enhanced Chemiluminescence Staining (ECL) to characterize expression levels of CymA. Cells were pelleted at 4000 × g for 10 min at room temperature and washed with M9 minimal medium 3 times. Cultures with the same OD600 of 10 were mixed in a 1:1 volume:volume ratio with a gel loading solution containing mercaptoethanol (0.71 M) and 1× LDS Loading Dye Mix (NuPage), and this mixture was incubated for 15 min at 95 °C. Samples were then loaded onto a 12% Bis-Tris gel (Invitrogen) and run for 45 min at 160 mV with a Precision Plus Protein Dual Color Standards ladder. Protein samples were transferred to blotting paper using a Trans-Blot Turbo Transfer Pack (Bio-Rad) and a Trans-Blot Turbo (Bio-Rad) using the “1 midi gel” setting. Cytochromes were visualized using Clarity Western ECL Substrate (Bio-Rad) and then imaged for both visible and chemiluminescent signals using a FluorChem E imager (ProteinSimple). Chemiluminescent images were overlaid on visible images to compare protein standards and chemiluminescent cytochromes.
Fluorescence measurement of flavin concentration. Flavins are intrinsically fluorescent. Fluorescence measurements were performed with 100 µL of experimental culture in a 96-well plate. The flavins were measured with 450/530 nm as excitation/emission wavelength (Tecan Spark multimode plate reader). The concentration of [Flavin]/OD600 in each sample was calculated using a standard curve prepared with flavin concentration of 0, 1, 2, 5,10, and 20 μM, and was normalized to OD600 values.
Bioelectrochemical characterizations. We used two-chamber, 3-electrode electrochemical reactors to perform the bioelectrochemical experiments as previously described29. The reactors contained three electrodes: a working electrode (graphite felt, diameter of 3.2 cm, surface area 16 cm2 for two sides, Alfa Aesar) and a reference electrode (CHI111, Ag/AgCl with 3 M KCl, CH Instruments) in the anodic chamber, and a counter electrode (Titanium wire, Alfa Aesar) in the cathodic chamber. We also used a cation exchange membrane (CMI-7000S, Membranes International) to separate these two chambers. Each chamber contained 120 mL of M9 minimal medium (BD) and 1% glycerol (Sigma-Aldrich, US) as the electrolyte. The anodic chamber was continuously stirred with a stir bar at 300 rpm, temperature-controlled at 30 °C. Cells were cultured aerobically in foil-covered 500 mL flasks containing 300 mL of 2xYT media at 37 °C with shaking at 250 rpm. When the cell OD600 reached 0.5-0.6, they were harvested at 4000 × g at 4 °C. The cells were then washed twice with M9 minimal medium and inoculated into the reactors to a final OD600 of 0.5. A BioLogic-potentiostat was used to conduct all the electrochemical measurements (VSP-300, Bio-Logic Science Instruments). Steady-state current (jss) was measured by polarizing electrode at 0.2 V vs. Ag/AgCl.
Double potential step chronoamperometry (DPSC). To differentiate between two electrochemical channels, we developed a double potential step chronoamperometry (DPSC) measurement, which involves recording the current over a specified time decay for each potential step. The DPSC measurement was conducted at the 40-hour mark. During the measurement, the electrode was polarized at each potential for 30 minutes. For later assessments in environmental water samples, the time duration was optimized to 15 minutes.
Using DPSC to assess the CymA-Channel and Flavin-Channel, we polarized the electrode at specific potentials and recorded the corresponding current response (j). For the CymA-Channel, we polarized electrode at 0 V and 0.2 V vs. Ag/AgCl, while for the Flavin-Channel, we polarized electrode at −0.2 V and −0.4 V vs. Ag/AgCl.
For example, when exposing heavy metal-E. coli to arsenite or cadmium presence, the absolute current ratios were calculated as follows,
- Atkinson, J.T., et al., Real-time bioelectronic sensing of environmental contaminants. Nature, 2022. 611(7936): p. 548-553.
- Yang, Y., et al., Harnessing the Periplasm of Bacterial Cells To Develop Biocatalysts for the Biosynthesis of Highly Pure Chemicals. Applied and Environmental Microbiology, 2017. 84(1): p. e01693-17.
- Ross Daniel, E., et al., Characterization of Protein-Protein Interactions Involved in Iron Reduction by Shewanella oneidensis MR-1. Applied and Environmental Microbiology, 2007. 73(18): p. 5797-5808.
- Hui, C.-y., et al., Detection of Bioavailable Cadmium by Double-Color Fluorescence Based on a Dual-Sensing Bioreporter System. Frontiers in Microbiology, 2021. 12.
- Chen, S.-Y., et al., Development of a Highly Sensitive Whole-Cell Biosensor for Arsenite Detection through Engineered Promoter Modifications. ACS Synthetic Biology, 2019. 8(10): p. 2295-2302.
- Kracke, F., I. Vassilev, and J.O. Krömer, Microbial electron transport and energy conservation – the foundation for optimizing bioelectrochemical systems. Frontiers in Microbiology, 2015. 6.
- Sarzanini, C., Recent developments in ion chromatography. 14th International Ion Chromatography Symposium, 2002. 956(1): p. 3-13.
- Behbahani, M., et al., Monitoring of trace amounts of heavy metals in different food and water samples by flame atomic absorption spectrophotometer after preconcentration by amine-functionalized graphene nanosheet. Environmental Monitoring and Assessment, 2014. 186(11): p. 7245-7257.
- Ghaedi, M., et al., Cloud point extraction and flame atomic absorption spectrometric determination of cadmium(II), lead(II), palladium(II) and silver(I) in environmental samples. Journal of Hazardous Materials, 2009. 168(2): p. 1022-1027.
- Wallace, G.G., et al., Conductive electroactive polymers: intelligent polymer systems. 2008: CRC press.
- Legin, A.V., et al., Cross-sensitivity of chalcogenide glass sensors in solutions of heavy metal ions. Sensors and Actuators B: Chemical, 1996. 34(1): p. 456-461.
- Rudnitskaya, A., et al., Multisensor system on the basis of an array of non-specific chemical sensors and artificial neural networks for determination of inorganic pollutants in a model groundwater. Talanta, 2001. 55(2): p. 425-431.
- Murray, E., et al., Fully automated, low-cost ion chromatography system for in-situ analysis of nitrite and nitrate in natural waters. Talanta, 2020. 216: p. 120955.
- Scheller, F.W., et al., Biosensors: Fundamentals, applications and trends. Sensors and Actuators B: Chemical, 1991. 4(1): p. 197-206.
- Songa, E.A. and J.O. Okonkwo, Recent approaches to improving selectivity and sensitivity of enzyme-based biosensors for organophosphorus pesticides: A review. Talanta, 2016. 155: p. 289-304.
- Gibson, T.D., Biosensors: The stabilité problem. Analusis, 1999. 27(7): p. 630-638.
- Guzik, U., K. Hupert-Kocurek, and D. Wojcieszyńska, Immobilization as a Strategy for Improving Enzyme Properties-Application to Oxidoreductases. Molecules, 2014. 19(7): p. 8995-9018.
- Sassolas, A., L.J. Blum, and B.D. Leca-Bouvier, Immobilization strategies to develop enzymatic biosensors. Biotechnology Advances, 2012. 30(3): p. 489-511.
- Gouda, M.D., et al., Enhancement of operational stability of an enzyme biosensor for glucose and sucrose using protein based stabilizing agents. Biosensors and Bioelectronics, 2002. 17(6): p. 503-507.
- Prévoteau, A. and K. Rabaey, Electroactive Biofilms for Sensing: Reflections and Perspectives. ACS Sensors, 2017. 2(8): p. 1072-1085.
- Su, L., et al., Microbial biosensors: A review. Biosensors and Bioelectronics, 2011. 26(5): p. 1788-1799.
- Park, M., S.-L. Tsai, and W. Chen, Microbial Biosensors: Engineered Microorganisms as the Sensing Machinery. Sensors, 2013. 13(5): p. 5777-5795.
- Di Lorenzo, M., et al., A small-scale air-cathode microbial fuel cell for on-line monitoring of water quality. Biosensors and Bioelectronics, 2014. 62: p. 182-188.
- Terrell, J.L., et al., Bioelectronic control of a microbial community using surface-assembled electrogenetic cells to route signals. Nature Nanotechnology, 2021. 16(6): p. 688-697.
- Meyer, A.J., et al., Escherichia coli “Marionette” strains with 12 highly optimized small-molecule sensors. Nature Chemical Biology, 2019. 15(2): p. 196-204.
- Buchsbaum, G., A. Gottschalk, and H.B. Barlow, Trichromacy, opponent colours coding and optimum colour information transmission in the retina. Proceedings of the Royal Society of London. Series B. Biological Sciences, 1997. 220(1218): p. 89-113.
- Werblin, F., T. Roska, and L.O. Chua, The analogic cellular neural network as a bionic eye. International Journal of Circuit Theory and Applications, 1995. 23(6): p. 541-569.
- Horwitz, G.D., Signals Related to Color in the Early Visual Cortex. Annual Review of Vision Science, 2020. 6(1): p. 287-311.
- Su, L., T. Fukushima, and C.M. Ajo-Franklin, A hybrid cyt c maturation system enhances the bioelectrical performance of engineered Escherichia coli by improving the rate-limiting step. Biosensors and Bioelectronics, 2020. 165: p. 112312.
- Marsili, E., et al., Shewanella secretes flavins that mediate extracellular electron transfer. Proceedings of the National Academy of Sciences, 2008. 105(10): p. 3968-3973.
- Yuan, S.-J., et al., A Photometric High-Throughput Method for Identification of Electrochemically Active Bacteria Using a WO3 Nanocluster Probe. Scientific Reports, 2013. 3(1): p. 1315.
- Jensen, H.M., et al., Engineering of a synthetic electron conduit in living cells. Proceedings of the National Academy of Sciences, 2010. 107(45): p. 19213-19218.
- Ding, N., S. Zhou, and Y. Deng, Transcription-Factor-based Biosensor Engineering for Applications in Synthetic Biology. ACS Synthetic Biology, 2021. 10(5): p. 911-922.
- Jin, X. and I.H. Riedel-Kruse, Biofilm Lithography enables high-resolution cell patterning via optogenetic adhesin expression. Proceedings of the National Academy of Sciences, 2018. 115(14): p. 3698-3703.
- Chappell, J., et al., Computational design of small transcription activating RNAs for versatile and dynamic gene regulation. Nature Communications, 2017. 8(1): p. 1051.
- Su, Y. and M.C. Hammond, RNA-based fluorescent biosensors for live cell imaging of small molecules and RNAs. Nanobiotechnology · Systems Biology, 2020. 63: p. 157-166.
- Jiang, X., et al., Which Multi-Heme Protein Complex Transfers Electrons More Efficiently? Comparing MtrCAB from Shewanella with OmcS from Geobacter. The Journal of Physical Chemistry Letters, 2020. 11(21): p. 9421-9425.
- Feng, J., et al., Enhancing the performance of Escherichia coli-inoculated microbial fuel cells by introduction of the phenazine-1-carboxylic acid pathway. Journal of Biotechnology, 2018. 275: p. 1-6.
- Mevers, E., et al., An elusive electron shuttle from a facultative anaerobe. eLife, 2019. 8: p. e48054.
- Webster, D.P., et al., An arsenic-specific biosensor with genetically engineered Shewanella oneidensis in a bioelectrochemical system. Biosensors and Bioelectronics, 2014. 62: p. 320-324.
There is NO Competing Interest.