The association between food consumption and human disease was recognized very early. Hippocrates (460 B.C.) was the first to report a strong link between consumed food and human illness [1]. In fact, every year, around 600 million people fall ill as a result of ingesting contaminated food, a number estimated to increase year after year [2]. Exposure to pathogens can cause symptoms as mild as nausea or vomiting, but serious cases can ultimately lead to organ failure or death [3].

Water and food products such as unpasteurized milk and dairy, meat, green leafy vegetables, and water are known to be at high risk for E. coli contamination, which is also often used as an indicator of fecal pollution in water [4,5]. Shellfish such as bivalves that feed through the filtration of large volumes of water are exceedingly prone to E. coli bioaccumulation. This is especially concerning considering that they are often consumed raw or only slightly cooked [6].

E. coli strains are important causes of wound infection, urinary tract infection, peritonitis, pneumonia, meningitis, and septicemia [7]. Overall, E. coli is thought to kill more than 2 million humans per year through both intestinal and extraintestinal diseases. Indeed, the incidence of extraintestinal infections in humans is increasing, and major uraemic syndrome epidemics such as the 2011 epidemic in Europe [7] regularly occur.

Reducing the number of outbreaks is essential, and to achieve this it is necessary to ensure food safety through close monitoring at critical processing points, which safeguards both product quality and safety. Because of this, there are guidelines for specific issues that need to be addressed [8].

EU regulation classifies bivalve production into 4 classes of harvesting safety: A, B, C, and Forbidden. Bivalves may only be caught and readily sold for direct human consumption in class A areas, which corresponds to a limit of E. coli of 230 most probable number (MPN) per 100 g of bivalve tissue and intravalvular liquid. In class B or C area, bivalves are either destined for depuration, where they are placed in clean high-quality waters to purge any stored contaminants, or for processing in an industrial facility, where they undergo appropriate heat treatments before being placed on the market. Harvesting is completely prohibited if any result exceeds 46,000 MPN/100 g of tissue [9].

Reference methods for pathogen detection are standardized at the international level by the International Organization of Standardization (ISO). ISO standards for the detection of E. coli are based on conventional microbiology, namely culturing methods, which include steps like sample preparation, enrichment, dilution, plating, enumeration, and isolation of individual colonies [3].

However, while they are extremely dependable and effective, these traditional methods are laborious and time-consuming, making them difficult to implement. Considering that the samples have to be transported to accredited laboratories, in addition to the time required for the enrichment step, the incubation time for bacterial growth, and the processing of biochemical tests, the process of identification can typically take between 3 and 7 days to complete [10]. Timely decision-making is of most importance in food handling facilities, in an effort to identify contaminated food products before they are made available to the public, so it is desirable to employ a system to monitor E. coli that is not only reliable and fast but also minimizes complicated sample handling and associated transportation time and costs to a diagnosis facility [11].

In light of the previously mentioned requirements for E. coli detection, new methods have been studied, particularly using biosensors, which are analytical devices that comprise a biorecognition element, a transducer, and a reading system [12]. A wide range of optical and electrochemical biosensors, many of which are used in label-free designs, have been proposed as a cost-effective and highly sensitive way to detect E. coli at low limits of detection. Nevertheless, such devices, especially those based on nanomaterials, still need to be improved in terms of reproducibility, stability, and selectivity/specificity in complex media for reliable, on-site operation in real-world applications [13].

An interesting type of technology uses liquid crystals (LCs), whose properties are intermediate between those of an isotropic liquid and of a solid crystal [14]. LCs possess the fluidity, droplet formation, and coalescence typically associated with liquids, while also maintaining optical, electrical, and magnetic anisotropy, as well as a periodic arrangement of molecules, often seen in crystals [15,16]. Since their discovery in 1888 by F. Reinitzer, LCs have been extensively studied across various fields of knowledge, which has resulted in significant advancements in science and technology [17]. In fact, LCs became the quintessential self-assembling molecular materials for LC displays [18], but numerous studies have evidenced their importance in fields such as optics, energy, and especially sensing applications [19].

In 1998, the Abbott research group initiated their work on the use of LCs associated with antigen-AB immunoreactions, and by 2002, they discovered that sodium dodecyl sulfonate could cause LCs to adopt an ordered vertical arrangement, a major breakthrough that led to significant advancements in LC biosensors [20]. The alignment of the LC molecules, known as the director orientation, is sensitive to external stimuli and can thus be used to detect and quantify changes in the environment. This orientation can be affected by changes in temperature, topography, or chemical structure [20], allowing, for example, label-free and fast amplification and transduction of biologically relevant binding events into optical outputs that may even be visible to the naked eye [18].

This extraordinary sensitivity can lead to switching between dark (black) and bright images [16]. The birefringence of LCs, a property of materials that present different refraction indexes (RIs) for different light polarizations, is behind these striking optical effects. When light passes through a birefringent material, it splits into two beams, corresponding to the ordinary and extraordinary RIs no and ne, and the difference between the two (no−ne) can be defined as birefringence.

Phase modification occurs when light emerges from the LC, i.e., the optical retardation of the incident radiation δ, which can be quantified by [21] δ=2πdλ(ne−no),where λ is the incident light’s wavelength and d is the path length of the light within the LC compound. Thus, LC-based devices depend on the ability to modulate the polarization of light, which can easily be visualized with two crossed polarizers or using a polarized optical microscope (POM). The intensity of the light emerging between crossed polarizers can be quantified by [21] I=I0sin2(2φ)sin2(δ2),where I0 is the original light intensity and φ is the angle between the polarizer and the LC molecules’ orientation. If the LC molecules are aligned perpendicularly to the confinement surface (i.e., homeotropic alignment, δ≈0), between crossed polarizers, the intensity of transmitted light I is negligible; thus, a dark image is observed. If the orientation of LC molecules is distorted (δ≠0), there is an alteration of the polarization of light along the LC medium so the intensity of the transmitted light I becomes significant, and a bright and colorful image is observed. This transition can be caused by minute traces of target molecules, as a strong but very local interaction on the surface triggers the disorder of many LC molecules resulting in signal amplification [22].

In sensing applications, the LC material is typically contained in a cell with transparent substrates on both sides, and the alignment of the LC molecules in a specific direction is induced by physical or chemical treatment of the confinement substrates. Currently, the most prominent methods of constructing LC sensing substrates include constructing a gold film, a mechanically rubbed film, and a silylating reagent self-assembly film [20].

The gold film requires a very high skill level, so its application is limited [23]. The rubbed film method is simple but can generate dust particles or detach matter during friction, and it is difficult to control the uniformity of the friction [24]. The silylation reagent method acts by controlling the length of the alkyl chain, which is a simple, effective, and widely used technique. Dimethyloctadecyl[3-(trimethoxysilyl)propyl]ammonium chloride (DMOAP), for example, has a long alkyl chain that, after being adsorbed onto the substrate surface, can induce LCs to arrange in an orderly manner along the alkyl chain. In general, functional molecules such as (3-aminopropyl)triethoxysilane (APTES) are introduced onto the substrate surface to immobilize the target molecule [20].

There has been a lot of research towards the biosensing application of LC-based systems, namely towards target molecules such as cholesterol or glucose. However, their use for the detection of pathogens, namely E. coli, has been very limited [25]. Most systems proposed in the literature can only differentiate between Gram-negative (such as E. coli) and Gram-positive bacteria [26], are associated with high costs and a significant number of false positives [27], or exhibit sensitivity towards a very specific strain [28].

Herein, we developed an LC sensor based on the distortion of the director field that selectively detects the presence of E. coli in a culture medium or contaminated water samples by measuring the quantity of light that emerges when the sensor is observed between cross-polarizers. This sensor was assembled using a sandwich approach with two glass coverslips and was functionalized with APTES, anti-E. coli antibodies, and DMOAP to ensure the vertical alignment of the LC molecules. In addition, the development of a simple prototype represents another significant advancement since it allows an even faster testing time, avoiding bulky equipment and measurements in-situ.

Deionized (DI) water, obtained from a Milli-Q water purification system, was used throughout the work. Phosphate buffer saline (PBS) tablets (pH=7.4, 10 mM; 1M = 1 mol/L) were obtained from Fisher Bioreagents, USA. Sulfuric acid (H2SO4, 95%–97%), (3-aminopropyl)triethoxysilane (APTES, 98%), and cysteamine hydrochloride (≥98%) were purchased from Sigma-Aldrich, Germany. Bovine serum albumin (BSA) solution (10 μg/mL in PBS) was obtained from Alfa Aesar, USA. N-hydroxysuccinimide (NHS, 0.5 M), N-(3-dimethylaminopropyl)-N′-ethylcarbodiimide hydrochloride (EDC, 0.2 M), and hydrogen peroxide (H2O2, 30% in volume fraction) were obtained from Merck, Germany. The nematic LC [4-cyano-4′-pentylbiphenyl (5CB)] was also acquired from Merck, under the designation K15, and used without any further purification. Anti-E. coli serotype O/K polyclonal antibody (AB) (5 mg/mL) and N-dimethyl-N-octadecyl(3-aminopropyl)trimethoxysilyl chloride (DMOAP, 60% in methanol) were acquired from ThermoFisher Scientific, Switzerland. Lysogeny broth (LB) growth medium was acquired from NZYTech, Portugal. Epoxy glue (Araldite) was purchased from Ceys, Spain.

To evaluate the developed sensors, a collaborator from the NOVA School of Science and Technology supplied four different bacterial strains, propagated in lysogeny agar (LA) or LB medium at 37°C with aeration. The main strain used was E. coli ATCC 35218, a quality control testing strain, which will be referred to as just “E. coli” in this paper for convenience. Other strains were used to assess selectivity, namely Vibrio atlanticus, Pseudomonas gallaeciensis, and Acinetobacter sp., which were isolated from an aquaculture plant in Portugal.

To estimate bacterial concentration, the optical density (OD) of the suspensions was measured using ultraviolet-visible (UV-Vis) spectroscopy. The UV-Vis spectra were obtained using a Shimadzu UV-2501 PC spectrophotometer, equipped with a halogen lamp Shimadzu (062-65004-06) and deuterium lamp Shimadzu (062-65055-05) at 600 nm. This OD600nm method is commonly used in microbiology to determine the concentration of cells in a liquid, while causing little damage or growth hindrance to cells. OD is directly related to concentration, and, for E. coli, 1.0 OD equates to 8×108cells/mL. The desired concentrations for each strain were obtained by serially diluting the initial bacteria suspensions with the corresponding growth medium.

Considering the long-term goal of applying these sensors to aquaculture settings, tests with contaminated water samples were also performed, through a collaboration with a researcher from the Centre for Environmental and Marine Studies (CESAM) and Department of Biology of the University of Aveiro. These samples were collected at various times during a depuration process in which the circulating water was treated with a UV lamp. Although no other method could be used to estimate the E. coli concentration in each sample at the time, it is known that bacteria concentration decreases over time during depuration [29]. Thus, it was still possible to assess if the developed sensors could detect this decrease in concentration, which is of the utmost importance for its intended end use in aquaculture, where samples similar to those collected need to be analyzed.

Each LC sensor was built from two 26mm×10mm glass slides, obtained by cutting 26mm×76mm microscope slides from ThermoFisher Scientific (Switzerland). Each glass substrate had only one functionalized surface, obtained by following the procedure schematically represented in Fig. 1. Because of this, it was important to ensure that throughout the process, the reagents were applied to the same side of each glass substrate.

The first step in the functionalization process aimed to yield free amine groups on each glass substrate, to enable the subsequent covalent immobilization of anti-E. coli ABs. First, each glass was carefully immersed in piranha solution (3:1 volume ratio H2SO4:H2O2) for 10 min to allow surface hydroxylation. After being thoroughly rinsed with DI water to ensure that all remaining residues were washed off, the hydroxylated glasses were silanized and chemically treated to allow the homeotropic alignment of the LCs. To achieve this, each glass was immersed for 1 h in a mixture of ethanol and DI water (70:30 volume ratio) containing 5% (volume fraction) APTES and 2% (volume fraction) DMOAP, followed by three wash cycles with DI water to remove unbound reagents. Finally, they were cured at 120°C for 20 min and left to cool. Once at room temperature, the amine-terminated glasses were ready to be functionalized with the anti-E. coli ABs. A fresh mixture of 100 μL AB solution (100 μg/mL), 50 μL EDC (0.2 M), and 50 μL NHS (0.5 M) was prepared and carefully pipetted onto each glass substrate, and it was left to react for 2 h, ensuring that each surface was completely covered. Next, the glasses were rinsed three times with PBS and put in contact with a BSA solution (10 μg/mL) for 2 h to passivate the surface [30]. Finally, the glasses were washed with PBS three times and left in this solution overnight.

After the functionalization, E. coli suspensions with concentrations between 2.8 and 2.8×109cells/mL were prepared for the detection tests. Then, a pair of functionalized glasses was immersed for 30 s in each of the testing samples, including one pair immersed in LB medium (sterile growth medium, the control). An additional assay was performed using contaminated water samples, following the same procedure as described for the E. coli suspensions. After being immersed in the suspensions, each corresponding pair of glass substrates was glued together by putting Araldite epoxy glue on the edge of each glass. Finally, by capillary rise, a drop of LC was placed between the two glass substrates, filling the entire cell.

To evaluate the response of the sensors, each assembled sample was observed under a polarized optical microscope (Olympus BX51) coupled with a CCD camera (Olympus DP73) and the Stream Basic v.1.9 Olympus software. Images were obtained in transmission mode between crossed linear polarizers with an objective of 10× (Olympus, MPlanFL N) and automatically scaled by the software. To quantify the light transmission of each sample, the transmitted light spectra were obtained using the incorporated monochromator from Sense+ from Sarspec and the software LightScan. The integrated spectral light flux was calculated for each spectrum by integrating the curve over the wavelengths between 300 and 1000 nm. Figure 2 depicts the working principle and assembly of these sensors.

To expedite the analysis of the transmitted light, a prototype, as shown in Fig. 3, was developed. It includes a laser, which acts as a light source to observe the samples that are placed into a holder, between two crossed polarizers. On the other side, there is a detector, incorporated in a Wio terminal (D51R), which measures the intensity of light that reaches it. The higher the E. coli concentration present in a sample, the greater the disruption of the LC molecules, which results in more light being able to emerge from the polarizers. Here, the light intensity is quantified as a percentage.

The LC sensors were also evaluated regarding their selectivity towards E. coli. To achieve this, Vibrio atlanticus, Pseudomonas gallaeciensis, Acinetobacter sp., and E. coli suspensions with the same OD600nm were used (equivalent to 2.8×109E. coli cells/mL). Each pair of functionalized glass substrates was immersed in the respective bacteria suspension and assembled as previously explained. The sensors were then observed under POM, and the integrated spectral light flux for each sample was calculated and compared.

Since it provides information on surface topography and morphology, scanning electron microscopy (SEM) is one of the most commonly used techniques for sample characterization [31]. In this work, the developed LC sensors were observed under a Hitachi Regulus 8220 (Hitachi Corporation, Japan). To observe the E. coli cells, they were fixed in 2.5% (volume fraction) glutaraldehyde (Sigma-Aldrich), for 20 min. After two washes with DI water, the fixed bacteria were dehydrated using a graded series of ethanol (50%, 70%, 90%, and 100% in volume fraction). Before visualization, the samples were dried in adhesive carbon substrates and sputter-coated with a 15 nm iridium film, using a Quorum Q150TES (Quorum Technologies, UK).

After being immersed in the different E. coli suspensions and assembled, the LC-based sensors were observed under a POM, which made it possible to evaluate their performance by assessing the integrated spectral light flux for the ten tested E. coli concentrations, compared to the control sample. Figure 4 shows the optical appearance of the developed sensors under a POM.

Here, it can be seen that the images corresponding to sensors exposed to a higher concentration of E. coli are brighter and more colorful. In fact, the control sample [Fig. 4(a)] appears completely dark, perfectly displaying what happens when there are no bacteria present to bind to the ABs and disrupt the homeotropic alignment of the LC molecules. As the E. coli concentration increases, the disruption imposed on the LC molecules also increases, which results in a higher amount of light being able to emerge from the sensor, which is maximum for the sample corresponding to the highest concentration [Fig. 4(k)].

Furthermore, analysis of the acquired spectra of light intensity for each sample also generally supports this visual tendency. The average obtained spectra for each sample of the three independent tests are shown in Fig. 5(a). Here it is clear that the spectrum regarding the control sample has an associated area much smaller than that corresponding to the sample of 2.8×109cells/mL. By integrating the area below each curve, the corresponding average integrated spectral light flux for each sample can be obtained, and these values can be presented as a function of E. coli concentration, as displayed in Fig. 5(b). This graph evidences a positive linear variation of light flux with E. coli concentration. Regarding the control sample, an average integrated light flux of (1.90±0.5)×105a.u. was obtained, contrasting with the much larger value of (3.7±1.0)×106a.u. obtained for the 2.8×109cells/mL.

The results suggest that this type of immunosensor is suitable for differentiating between negative and positive samples for E. coli contamination, being capable of detecting concentrations as low as 2.8 cells/mL. This sample presented an average integrated light flux of (1.1±0.3)×106a.u., which is a clear difference from the value of the control, so this represents the estimated limit of detection found in this work. This LOD value is lower than others reported in the literature for E. coli detection, which was 100 CFUs/mL for a label-free impedimetric aptasensor [32], or in the range of 100–300 CFUs/mL depending on the bacteria suspension medium for an electrochemical biosensor based on target-induced aptamer displacement [33].

It is imperative to ensure that the developed sensors are selective towards E. coli, in order to avoid false positives that arise from contact with other types of bacteria. Thus, a selectivity test was performed, in which the glass slips were immersed in four different bacterial cultures, at the same growth phase and the same OD600nm, frequently found in marine water ecosystems, namely, Vibrio atlanticus, Pseudomonas gallaeciensis, Acinetobacter sp., and E. coli were used. Spectral light flux was analyzed by observing the developed sensors under POM, and the obtained average spectrum for each bacteria strain is shown in Fig. 6(a), where it is possible to observe that the light intensity associated with the E. coli samples is much higher than the intensity for the samples with the other three bacteria. This is also evidenced by the images taken under POM, where it is possible to note that Figs. 6(b)–6(d) all appear completely dark, which means that there was no binding of bacteria to the AB on the sensors’ surface. Only in Fig. 6(e) it is possible to witness the effects of the disruption of the LC molecules by binding to the specific AB, resulting in a bright and colorful image.

Consequently, there is a great difference between the spectral light flux for E. coli and all other bacterial species, as evidenced by Fig. 7, which represents the histogram comparison between the spectral light flux for all tested samples. The sensors immersed in E. coli presented an integrated spectral light flux of (3.73±1.04)×106a.u., contrasting with the values in the order of 103a.u. found in every other sample. These results clearly show that the developed sensors have a high selectivity towards E. coli, thanks to the specific binding between the anti-E. coli AB and the outer membrane of these bacteria.

To evaluate if the binding between E. coli and the functionalized surfaces of the sensor occurred, SEM images of the glasses were acquired.

The functionalized glass slides were immersed in 2.8×109cells/mLE. coli solution and washed three times. Several images were collected with different magnifications, and Fig. 8 shows representative images of this analysis. Here, it is possible to see the typical rod shape of E. coli attached to the surface of the glasses. Given its dimensions of 1.59±0.05μm and 0.47±0.02μm, these are easily differentiated from the rod-like molecules of the 5CB LC, which are around 20 Å long. These results confirm that the desired binding correctly occurred.

The LC sensors were also used to assess the concentration of E. coli in water throughout a depuration process. Samples were taken at the beginning of this process (0 h), after 30 min, 3 h, 4.5 h, and 22.5 h, and finally after 24 h. As the depuration process progresses, the concentration of E. coli decreases, and the sample taken after 24 h should show a much lower concentration than the samples collected first. This is proven by the obtained results, where it is possible to observe that Figs. 9(a) and 9(b) appear much more colorful than Fig. 9(f), and this can also be noted in the optical spectra corresponding to each sample [Fig. 9(g)].

By calculating the integrated spectral light flux associated with each sample, it is possible to see that there is a general tendency for this value to decrease over time [Fig. 9(h)]. The maximum value of 4.70×106a.u. decreased to 1.72×105a.u. for the 24 h sample. It is important to note that the maximum was not detected for the 0 h sample. However, this can be explained by the fact that the bacteria may not have been well distributed throughout the tank, so when the water began circulating it increased the value of the sample collected. Anyway, as only 30 min had passed since the start, it was unlikely that a considerable reduction in E. coli would happen.

Again, it was not possible to validate the concentration of E. coli present in the water using a standard method, which limits the type of analysis that can be done. Still, it is clear that this type of sensor methodology has great potential for the intended application, and further testing should be conducted to fully explore its limits of application and detection.

Given the promising results obtained using the LC-based sensors, a prototype for an even faster analysis was developed. While the previously explained method is considerably easier when compared to traditional E. coli detection (such as culture-based methods), it still does not provide immediate results. Also, the fact that the detection of E. coli is made by integrating the correspondent spectrum, meaning that there is no measurement unit associated, is a downside that should be addressed.

Thus, a preliminary test was performed in order to analyze if the developed system can measure the intensity of the light passing through the sensor properly and provide comparable results with those obtained from the observation of the samples in the POM. Five sensors with different concentrations of E. coli were tested, and Fig. 10 shows the comparison between the results obtained between the analysis in the POM (integrated spectral flux) and the percentage of light measured by the prototype. These results show that a decrease in the integrated spectral flux corresponds to a decrease in the percentage of light read by the prototype. In fact, it is possible to notice that the maximum value of integrated spectral flux, 1.04×107a.u., corresponded to a maximum of 94% measured by the prototype, meaning that the amount of E. coli present in the analyzed sample was large enough to distort the LC molecules in such a way that the light was able to pass almost entirely through the polarizers. Similarly, for the sample with the lowest concentration of E. coli, having obtained a spectral flux value of 1.70×106a.u., the prototype measured light intensity of only 2%. These are very promising results, which show that the developed system can provide results comparable to those obtained by the POM method. This brings great advantages since it facilitates the analysis process and allows immediate results to be obtained.

However, in the conducted tests, it was not possible to adequately measure the amount of E. coli present in each sample, which made it impossible to carry out a more thorough analysis.

Nonetheless, considering the promising preliminary results obtained, and the fact that the application of this prototype would allow much faster measurements, more tests should be performed in the future. In addition to verifying that there is a decrease in the percentage of light with a reduction in the number of bacteria, research should be done to try to establish a relation between the percentage of transmitted light and E. coli concentration.

Maintaining appropriate public health conditions is and always will be a global concern, and this includes having access to food that is safe for consumption. To ensure the safety of the food supply, it is crucial to test samples for E. coli contamination. However, the currently available methods of detection still face a lot of problems particularly related to waiting times and the need for specialized institutions and personnel. Considering these setbacks, the primary objective of this work was to develop a highly selective, fast, and sensitive LC-based E. coli immunosensor. The detection was made possible by observing the optical response of the LC molecules under POM and quantifying the amount of light flux that could emerge from each sample when observed between cross-polarizers. The sensors worked with an exposure time of only 30 s, and with concentrations as low as 2.8 cells/mL and up to 2.8×109cells/mL. An increase in light flux (resulting in brighter images) was found with increasing E. coli concentration, as expected. SEM visualization of the samples evidenced the binding between E. coli and the glass substrates, proving the functionalization procedure and the detection tests were a success. Results using other bacteria species showed a very high selectivity towards E. coli, and it was possible to detect the decrease in E. coli concentration during depuration, from the assays with contaminated water samples. The development of a working prototype represents another significant advancement since it allows an even shorter testing time. Considering the very limited application of LC molecules for the detection of E. coli, the results shown in this work evidence very positive progress. As future work, we intend to improve our portable platform to be able to use sunlight as a light source and include a camera and an optical objective to obtain the images, and also use artificial intelligence tools to analyze them. We are also interested in analyzing samples in the presence of a mixture of the cultures as well.

Acknowledgment. Maria Simone Soares acknowledges FCT/MCTES for the PhD fellowship. This work was developed within the scope of the projects CICECO and DigiAqua, financed by national funds through the Portuguese Science and Technology Foundation/MCTES (FCT I.P.). We also acknowledge the financial support of CESAM by FCT/MCTES through national funds. This work is part of the project DepurD, supported by Portugal and the European Union through MAR2020, Portugal2020, and FEAMP. The research was co-funded by the financial support of the European Union under the REFRESH—Research Excellence For REgion Sustainability and High-tech Industries, project number CZ.10.03.01/00/22_003/0000048 via the Operational Programme Just Transition. This work was also supported by the Ministry of Education, Youth, and Sports of the Czech Republic conducted by the VSB-Technical University of Ostrava, under grant no. SP2024/081.