Rapid, Selective, and Ultra-Sensitive Electronic Field Effect Detection of Escherichia coli

1. Introduction

Escherichia coli (E. coli) is a Gram-negative bacteria found in the human digestive tract. While most strains of E. coli are harmless, some strains can cause severe illness and be life-threatening, particularly in young children and those with compromised immune systems. Escherichia coli (E. coli) infection can be particularly life-threatening in young children and the elderly. Contaminated water, including that contaminated by E. coli, leads to the deaths of over 500,000 people annually. According to the Centers for Disease Control and Prevention (CDC), an estimated 265,000 E. coli infections occur annually in the United States(Costigan et al., 2022). The World Health Organization (WHO) estimates that up to 10% of patients infected with E. coli can develop a life-threatening disease such as hemolytic uremic syndrome (HUS)(WHO, 2018). Among those who develop a life-threatening illness, the case-fatality rate can range from 3% to 5%(Magnus et al., 2012). Although survivors of HUS may recover from the acute illness, they can suffer from serious long-term health complications, including neurological complications such as seizures, stroke, and coma, which occur in approximately 25% of HUS patients(Trachtman et al., 2012; WHO, 2011).

E. coli infection in humans can be caused by consuming contaminated foods, such as raw vegetables, as evidenced by the E. coli lettuce outbreak in 2020(Public Health Agency of Canada, 2019; Vanesse et al., 2023). E. coli has also been found in isolated bodies of water, leading to waterborne transmission through contaminated drinking water and recreational waters(Magnus et al., 2012; WHO, 2018). Effective food screening before reaching consumers and quick, cost-effective testing of suspected water contaminations can significantly reduce the incidence of E. coli outbreaks, potentially saving many lives, particularly in developing nations(Magnus et al., 2012; WHO, 2018).

Detecting E. coli before consumption is crucial for reducing E. coli-related illnesses and preventing severe health consequences, especially in vulnerable populations such as young children and the elderly. Early detection of even small concentrations of E. coli is vital for identifying fecal contamination and controlling infection sources to prevent the spread of this pathogen.

Polymerase chain reaction (PCR) is widely used for detecting E. coli due to its sensitivity to E. coli DNA and rapidity. The limit of detection (LOD) for PCR can be as low as 40 CFU per reaction, with detection times typically ranging from a few hours to a day(Ibekwe & Grieve, 2003; Li et al., 2017). However, the presence of PCR inhibitors in a sample, such as humic acids, heavy metals, and complex polysaccharides, can affect the accuracy of the results(Blumenfeld et al., 2022; Chen et al., 2012; Hariri, 2022; Park et al., 2020). Obtaining reliable results may thus require additional sample preparation steps, such as co-purification or inactivation of inhibitors. In cases where the target E. coli concentration is low, performing a pre-enrichment step may also be necessary to improve the assay’s sensitivity. These additional steps can increase assay time and complexity(Park et al., 2020).

An alternative method for detecting specific proteins or molecules is the enzyme-linked immunosorbent assay (ELISA). This method utilizes the particular interaction between antibodies and antigens, as well as the catalysis between enzyme and substrate, to detect the target molecules (Hariri, 2022; Yanan Zhao et al., 2020). ELISA is relatively simple to perform and can be used to detect multiple target molecules simultaneously. However, it may not be as sensitive as PCR or as capable of detecting specific gene sequences. The LOD for conventional ELISA (C-ELISA) for E. coli O157:H7 is typically between ${10}^5$ to ${10}^7$CFU/mL, which is inadequate for detecting low concentrations of bacteria. However, novel ELISA methods, such as FNP-ELISA, have improved sensitivity with a LOD of 68 CFU/mL and a detection time of approximately 3 hours(Shen et al., 2014).

Next-generation sequencing (NGS) is a high-throughput sequencing method that allows for the direct quantification of target bacteria by detecting specific genes(Wensel et al., 2022). The LOD for NGS in detecting antimicrobial resistance genes (ARGs) in E. coli can be achieved with a sequencing depth of ~15× coverage, requiring approximately 30 million reads for adequate sensitivity(Rooney et al., 2022). One of the significant advantages of NGS is its ability to rapidly generate large amounts of data at a relatively low cost per base. It can also detect the presence of multiple pathogens and resistance genes in a single sample by sequencing only a few kilobases of the genome. In addition, NGS kits often include software for strain typing. However, NGS has some limitations, including the inability to distinguish between live and dead bacteria and requiring highly trained personnel. The high cost and novelty of NGS methods can also be deterrents for some users.

One significant challenge in testing water for E. coli is that suspected contaminated samples may contain a wide variety of other contaminants besides the target bacteria. This can interfere with the selective and accurate detection of E. coli, making the testing process expensive, time-consuming, and difficult(Daley et al., 2018; Manzanas et al., 2023).

Recently, the use of functionalized FET for the detection of cancer markers (Kwong Hong Tsang et al., 2019; Mu et al., 2015), viruses (Seo et al., 2020), and bacteria (Moudgil et al., 2020) has shown that these types of biosensors can potentially be helpful as a platform for sensitive detection in healthcare applications. Despite this, Current field-effect transistors (FETs) used to detect bacteria often have short detection ranges for E. coli K12(Du et al., 2023; Shang et al., 2022), require long detection times(Lin et al., 2021; Shang et al., 2022), or involve cumbersome processes to prepare linkers for binding and selectivity enhancement(Wei et al., 2023). Here, we utilize a FET biosensor with B/N co-coped graphene oxide (GO) gel (BN-GO) ((Novodchuk et al., 2020)as the transducer material and functionalized with a bioreceptor derived from antibodies to demonstrate the detection of E. coli K12. Previously, a FET biosensor based on a functionalized BN-GO transducer was used to detect low-concentration solutions of ‘pure’ biomarkers, such as B-type natriuretic peptide (Novodchuk, Kayaharman, et al., 2021), which can serve as a heart-attack marker, and Covid-19 nucleocapsid protein (Novodchuk et al., 2022). This work reports the first use of functionalized BN-GO FET for bio-detection of live-culture samples.

The device was functionalized with antibodies specific for the E. coli K-12 O-antigen and exhibited a low limit of detection (LOD) and an extensive detection range over eight orders of magnitude in buffer solution. It also exhibited excellent selectivity against a live sample contaminant in the form of Bacillus thuringiensis and a detection time of less than two minutes.

2. Experimental

2.1. Materials

Hexagonal boron nitride (h-BN) ultrafine powder and GO solution were purchased from Graphene Supermarket. Phosphate buffered saline (PBS) solution, pH = 7.4, was purchased from [Thermo Scientific]. Escherichia coli and Bacillus thuringiensis were purchased from the American Type Culture Collection (ATCC 25404 and ATCC 33671, respectively). Both strains of bacteria were grown in filter-sterilized Luria broth (LB) medium (1% tryptone, 0.5% yeast extract, and 1% NaCl) for 12 hours at 37°C in sterile cell culture shaker flasks at 200 rpm. Each bacterial culture was individually centrifuged at 3,300 x g for 15 minutes; the spent LB supernatant was decanted, and then each bacterial culture was resuspended and diluted in sterile phosphate-buffered saline (PBS) to the desired concentrations.

2.2. Fabrication of BN-GO FETs

In our previous work(Novodchuk et al., 2020), we demonstrated the fabrication of the BN-GO gel FETs. Briefly, a solution containing GO and a solution containing h-BN quantum dots were mixed and irradiated using the pulses from a 1 kHz ultra-short femtosecond (fs) laser, with an average power of 1 W, focused by a 5cm lens to a spot area of 78.5×10^-5 cm² giving an intensity of 3.5 ×10¹⁵ W/cm². The focus of the laser beam was held 2 mm below the solution’s surface for 50 min. The solution was stirred throughout the ablation process using a magnetic stirrer, resulting in a BN-GO gel. The BN-GO gel was used as the channel material, with a channel width of 53 μm, between source and drain electrodes in a back-gated FET device with pre-patterned Au/Ti source and drain electrodes on top of a SiO2/p-Si substrate.

2.3. Antibody Functionalization and Device Passivation

In our previous investigation(Novodchuk et al., 2022), a methodology akin to the one employed earlier was followed to affix Anti-E. coli O-antigen antibodies onto the BN-GO functionalized channels. The choice of BN-GO gel as a functionalizing agent was predicated on its unique characteristics, which include high mobility and abundant carboxyl groups. These features are instrumental for the covalent binding of antibodies through an amine-carboxyl reaction, a process thoroughly characterized by (Novodchuk et al., 2020). The high mobility of the BN-GO gel facilitates rapid electron transport, enhancing the sensor’s response time.

In contrast, the carboxyl group of the BN-GO gel proved instrumental in facilitating the effective covalent binding of antibodies through an amine-carboxyl reaction, as explicated by (Novodchuk, Bajcsy, et al., 2021). The carboxyl groups enable strong covalent bonding with the antibodies’ amine groups, ensuring stable attachment. This chemical specificity allows antibodies to be precisely oriented and densely packed, maximizing the sensor’s ability to capture E. coli efficiently. The uniform distribution of these functional groups across the BN-GO gel surface ensures an optimal environment for antibody immobilization, significantly enhancing the sensor’s performance by improving its sensitivity and selectivity in complex biological samples. Subsequently, 5μL droplets of the antibody solution were meticulously pipetted onto the BN-GO channel surface, allowing for a 48-hour immobilization period at 4°C. Rigorous rinsing with PBS, accomplished through successive 10μL droplets of buffer solution applied and removed five times, was performed to ensure robust immobilization. To mitigate non-specific binding, the devices underwent passivation via incubation in an ethanolamine solution, followed by thorough rinsing with PBS.

2.4. Immunodetection

Electrical measurements were performed with a probe station interfaced with a KEY-SIGHT B2900A Series software-controlled source measure unit. The reference measurement was performed using a pipette to drop PBS buffer onto the channel of the FET biosensor. Solutions of different concentrations were made using a solution containing E. coli in 0.1x PBS buffer and serially diluting. 2.5μL of each E. coli solution was periodically dropped onto the BN-GO channel to verify real-time measurement capabilities. Continuous monitoring of the device’s current was performed at fixed drain and back gate voltages of −0.05 V and −0.6 V, respectively. The Dirac point change was evaluated by dropping 2.5μL of each E. coli diluted solution onto the channel of the biosensor, and a drain current vs. gate voltage plot was observed after 2 min (at a constant drain voltage of − 0.05 V). After each Dirac point measurement, the channel was rinsed with PBS solution, and then the measurement was repeated with another solution of E. coli of higher concentration. The experiments were conducted using six different BN-GO gel FET devices to ensure reproducibility and accuracy of the results.

For the selectivity experiments, the Dirac point measurement was performed using the same procedure described for E. coli, but instead of 0.1x PBS/E. coli solution, a solution of 0.1x PBS/ B. thuringiensis was pipetted.

The biosensor’s current was continuously monitored for constant gate and drain voltages of − 1.5 V and − 50 mV, respectively.

3. Results and Discussion

3.1. Device Configuration and Sensing Performance

3.1.1. Device Configuration

This study’s employed FET device configuration mirrors our prior publications (Novodchuk et al., 2022). It comprises a p-doped silicon wafer with a SiO2 dielectric layer as the substrate, source, and drain electrodes. Additionally, it incorporates a BN-GO gel channel, as depicted in Figure 1a. To enable specific detection of E. coli bacteria within a sample, the device undergoes functionalization with E. coli antibodies. Subsequently, the device is passivated with ethanolamine to eliminate any undesirable charge-related effects originating from the sample, as demonstrated in Figure 1b (Campos et al., 2019).

Figure 1. The schematics of the device configuration and biosensing mechanism of the BN-GO gel E. coli biosensor. a) The back-gated FET device configuration features the BN-GO gel channel between the source and drain electrodes. b) The circuit connections of the BN-GO gel FET device. c) The device configuration following E. coli antibody functionalization and passivation with ethanolamine. d) The biosensing response is determined by the shift in the Dirac voltage after introducing the E. coli-containing solution.

Figure 1. The schematics of the device configuration and biosensing mechanism of the BN-GO gel E. coli biosensor. a) The back-gated FET device configuration features the BN-GO gel channel between the source and drain electrodes. b) The circuit connections of the BN-GO gel FET device. c) The device configuration following E. coli antibody functionalization and passivation with ethanolamine. d) The biosensing response is determined by the shift in the Dirac voltage after introducing the E. coli-containing solution.

3.1.2. Sensing Performance

The E. coli cells were diluted in a PBS buffer solution. The drain current for BN-GO gel FET, functionalized with antibodies and passivated, was graphed against gate voltage while maintaining a constant drain voltage of -50 mV. In this measurement, we used a sterile buffer solution to represent a baseline condition in which no E. coli cells were present in the sample. Various concentrations of E. coli diluted in a buffer were deposited onto the channel for sensing experiments, and the drain current vs. gate voltage was monitored. E. coli has a net negative charge around its membrane(Wei et al., 2023); their binding to the anti-E. coli will result in an augmentation of negative charge within the BN-GO gel. This will induce hole carriers in the BN-GO channel and increase the Dirac voltage, as illustrated in Figure 1d. The performance of the BN-GO gel FET biosensor was evaluated by monitoring its response to varying concentrations of E. coli K-12. As demonstrated in Figure 2, a slight change in concentration results in a difference in the ON current and a shift in the Dirac voltage, underscoring the remarkable sensitivity of the BN-GO gel FET biosensor. This trend is caused by E. coli cells accumulating in the sample solution; the negative bound bacterial cells function as gate electrolytes, generating a negative potential. Thus, more hole carriers are induced, leading to a gradual increase in the ON current.

Figure 2. E. coli biosensing performance of the antibody functionalized BN-GO gel FET. The shift in Dirac and the increase in ON current correspond to increasing concentrations of E. coli in solution.

Figure 2. E. coli biosensing performance of the antibody functionalized BN-GO gel FET. The shift in Dirac and the increase in ON current correspond to increasing concentrations of E. coli in solution.

3.2. Sensitivity

The BN-GO gel FET biosensor’s ability to detect different concentrations of E. coli is crucial for its effective performance. Sensitivity in this context refers to the ability to detect very low concentrations and how the device responds to varying levels of the target analyte, which can be observed through changes in the Dirac voltage and fluctuations in the ON current(Novodchuk, Bajcsy, et al., 2021; Novodchuk, Kayaharman, et al., 2021).

As E. coli bacteria bind to the antibody-functionalized surface of the BN-GO gel, the sensor experiences an increase in negative charge accumulation. This accumulation influences the charge carrier density within the channel, leading to a shift in the Dirac voltage. Specifically, the relationship between the E. coli concentration and the Dirac voltage shift follows a logarithmic trend, represented by the equation ${∆V}_{Dirac}$= 3.803*ln (E. coli concentration) +2.716 (Figure 4). This logarithmic response indicates that the sensor’s sensitivity allows it to detect a wide range of E. coli concentrations with high precision. The OFF current is observed to increase with higher E. coli concentrations. This can be attributed to multiple factors, including the increased charge carrier density and surface states and trap charges due to the increased bound bacterial cells on the antibody functionalized surface. Additionally, the electric field within the FET is modulated due to the electrostatic interactions of the negatively charged bacterial cells and the channel.

The BN-GO gel FET biosensor demonstrates a low limit of detection (LOD), capable of identifying concentrations as low as one cell per sample. This high sensitivity is achieved due to the efficient charge transfer properties, the BN-GO gel’s high carrier mobility, and the specific binding interactions between the E. coli antibodies and the bacterial cells.

3.3. Selectivity and Specificity

Ensuring the biosensor’s selectivity and specificity is crucial for its reliability and applicability. The biosensor’s selectivity relies on the receptor’s ability to exclusively bind to the target biomolecules, minimizing interference from other coexisting molecules in the biological sample (Eswaran et al., 2022).

In the context of clinical diagnosis, actual biological samples are inherently complex, containing diverse biomolecules. The biosensor’s specificity establishes whether the recognition interaction will be exclusively between the analyte and the bioreceptor (Wang et al., 2020).

Different concentrations of B. thuringiensis were added to the functionalized BN-GO gel channel to examine the BN-GO gel FET biosensor selectivity and specificity. The Dirac shift in response to the different concentrations determines the sensor’s selectivity and specificity, as shown in Figure 3. Notably, the selectivity assessment of the device involves comparing the Dirac voltage response to B. thuringiensis with the response to E. coli in the LOD concentration (10CFU/mL). As seen in Figure 3, the Dirac voltage for increasing concentrations of B. thuringiensis was approximately 4mV, and the biosensor’s response was weak with 2.5 times lower response, similar to the biosensor’s response to a blank sample with only 0.1xPBS buffer. This outcome demonstrates the BN-GO gel FET biosensor’s excellent selectivity and specificity for the intended target E. coli.

3.4. Performance Compared to the State-of-Art

In addition to selectivity and specificity, the parameters of the Limit of Detection (LOD) and detection range play a critical role in assessing the biosensor’s performance(Novodchuk et al., 2022). A broad detection range allows the biosensor to detect E. coli over a wide concentration range, making it versatile and applicable to various sample types. This is especially valuable in clinical diagnostics, environmental monitoring, and food safety, where E. coli concentrations vary significantly. In addition, wide detection ranges provide a more comprehensive view of the sample’s E. coli concentration. It allows for accurate quantification across different concentration levels, ensuring reliable results regardless of whether the concentration is high or low.

The Limit of Detection (LOD) for the BN-GO gel FET biosensor is discernible in Figure 4. It indicates a detection threshold of one cell per sample, accompanied by a significant Dirac voltage shift of 5mV. This demonstrates that the sensor can detect extremely low concentrations of E. coli, as even slight increases in bacterial concentration cause noticeable shifts in the Dirac voltage. The shift in the Dirac voltage was extracted for different concentrations of E. coli in buffer solution, and the results are summarized in Figure 4. The binding of analyte molecules to the FET surface follows the Langmuir adsorption isotherm, which relates the fractional surface coverage to the analyte concentration. On a logarithmic scale, the Langmuir isotherm exhibits a linear region over a certain concentration range(Purwidyantri et al., 2021).

The BN-GO FET biosensor stretches over a wide range of concentrations from 10-${10}^8$. It is worth highlighting that when presented on a logarithmic scale, the connection between the alteration in the Dirac voltage and concentration remained consistently linear and did not exhibit saturation even at the tested concentrations, indicating that the detection range may be longer than originally tested. The results were fitted ($R^2$ = 0.98) with the following relation ${∆V}_{Dirac}$= 3.803*ln (E. coli concentration) +2.716.

In comparison to prior studies (Table 1), it is evident that none have presented a comprehensive biosensor encompassing a balanced combination of key performance parameters, including Limit of Detection (LOD), detection range sensitivity, and detection time. For instance, (Du et al., 2023)work yielded a small LOD and rapid detection time yet suffered from a limited detection range. Furthermore, (Lin et al., 2021) reported a somewhat low LOD and a broad detection range but was plagued by extended detection times. (Shang et al., 2022) developed a portable microfluidic biosensor for the specific detection of E. coli (O157:H7), utilizing a finger-actuated mechanism. While this microfluidic biosensor facilitated a “sample-in and answer-out” assay for E. coli (O157:H7) and eliminated the need for external pumps and specialized personnel, it exhibited a relatively long detection time and a limited dynamic range spanning six orders of magnitude. These characteristics may influence its practicality for real-time detection in specific applications.

Conversely, (Wei et al., 2023) achieved a low LOD and a somewhat broader detection range; however, their biosensor’s preparation process was notably complex, exceeding the complexity of our biosensor’s fabrication. Their study reported the development of an improved graphene-based field-effect transistor modified with heat-denatured casein (GFET) for the sensitive detection of E. coli O157:H7 in real water samples. The heat-denatured casein, coated on the graphene surface, served a dual purpose as a probe linker, facilitating binding with the specific antibody of E. coli O157:H7, and as a blocker to prevent interference signals stemming from non-specific adsorption between the graphene surface and impurities present in natural water samples.

Consequently, our biosensor demonstrates superiority when considering the overall sensing performance despite not achieving the absolute minimum LOD observed in certain studies. This underscores the holistic advantage of our biosensor in addressing a broader range of practical applications. In the present study, the determined Limit of Detection (LOD) was influenced by Debye screening effects. Forthcoming research endeavors must be directed toward addressing and mitigating this particular challenge. Furthermore, the focus of future investigations should extend to the practical application of our biosensor in real-world samples, such as water and food matrices, for the robust and accurate detection of E. coli.

4. Conclusion and Future Outlook

In summary, we have successfully developed a graphene-based Field-Effect Transistor (FET) biosensor capable of detecting live E. coli in buffer solutions. The underlying detection principle relies on a graphene-based semiconductor formed by a Boron Nitride-Graphene Oxide (BN-GO) gel, covalently functionalized with antibodies specific to E. coli. Our biosensor device exhibited an impressive Limit of Detection (LOD) as low as one cell within a concise time frame of 2 minutes. Moreover, the biosensor demonstrated an exceptional detection range of 8 orders of magnitude from 10 CFU/mL to ${10}^8$. Notably, the BN-GO gel FET biosensor exhibited excellent specificity and selectivity and minimal interference from other substances.

These findings underscore the significant potential of BN-GO gel FET biosensors for rapid label-free detection of E. coli. Future investigations should aim to extend the applicability of this biosensor to more complex sample matrices, such as water or food. This will test the robustness and versatility of the biosensor under varied and potentially challenging conditions. Additionally, efforts to enhance the sensitivity and detection range could involve exploring alternative functionalization materials or improving the integration of the BN-GO gel with other biorecognition elements. Furthermore, future studies should focus on analyzing potential interferences by testing the biosensor against various families of bacteria with similar surface charges to assess their effects on the biosensor’s selectivity and specificity.

Developing portable and user-friendly versions of the biosensor for on-site testing is another crucial step that would enable real-time, on-the-ground monitoring of bacterial contamination, which is particularly valuable in remote or resource-limited settings.

This study serves as a proof-of-concept demonstration of the effective functionalization of BN-GO gel with antibodies for bacteria detection, highlighting the broader potential of BN-GO gel FETs for highly sensitive biosensing applications. Exploring these possibilities promises to further the field of microbial detection and enhance public health safety measures.