1. Introduction

Rapid and timely detection of pathogens is essential for maintaining proper health conditions, not only in clinical settings to prevent and stop the spread of infections, but also in environmental monitoring, biomedical research, and, particularly, in the food industry to ensure proper food safety [1]. The latest global estimates by the World Health Organization on foodborne diseases, published in 2015, suggest that around 600 million people fall ill each year as a result of ingesting contaminated food. Of these, 420,000 people die and this number is estimated to increase year after year [2].

Exposure to pathogens can result in symptoms with varying degrees of severity, ranging from mild short-term symptoms, such as nausea or vomiting, to long-term illnesses, such as cancer, kidney or liver failure, neural disorders, and ultimately death [3].

Thus, early detection of pathogenic microorganisms in contaminated food and water is critical to ensure proper food safety, reduce the number of outbreaks, and ensure that ill people are given appropriate treatment after properly identifying the pathogen they have been exposed to [4]. Escherichia coli (E. coli), Listeria monocytogenes, Salmonella Typhimurium, and Campylobacter are the pathogenic strains that cause most foodborne diseases [5]. In particular, E. coli is a very common bacterium that can be found in the gastrointestinal tracts of humans and other warm-blooded animals and is a part of the normal bacterial flora. Most strains are considered harmless, however, several intestinal diseases, urinary tract infections, neonatal meningitis, and gastroenteritis have been reported to be caused by E. coli [6,7].

People primarily become infected with pathogenic E. coli by eating or handling contaminated food or water, or through contact with infected animals. Products such as unpasteurized milk and dairy products, meat products, green leafy vegetables, and water are known to be at high risk for E. coli [8]. Additionally, shellfish are also highly susceptible to bioaccumulation of E. coli which, in addition to its pathogenic potential, is also a good indicator of fecal contamination that may be associated with other pathogens. This is of particular concern in Portugal, where bivalve harvesting is an ancient and traditional activity, that is of great social, economic, and cultural value [9].

According to Regulation (CE) No. 854/2004, Regulation (CE) No. 1021/2008, and Regulation (EU) 2015/2285, bivalve production areas can be classified 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 correspond to a limit of E. coli of 230 MPN per 100 g of bivalve tissue. In Zones B or C, 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 46000 MPN/100 g of tissue and intravalvular liquid [10].

The most conventional method to inspect food for E. coli contamination relies on culture-based-methods, that use specific microbiological media to isolate and count E. coli in food samples. These methods are effective, but they are time-consuming and require proper laboratory conditions to be carried out [11]. Timely decision-making is of most importance in food handling facilities, so it is desirable to employ a system to monitor E. coli that is not only reliable and fast but minimizes complicated sample handling and associated transportation time and costs to a diagnosis facility [4].

Improvements in pathogen detection came in the form of enzyme-linked immunosorbent assay (ELISA) and polymerase chain reaction (PCR) based assays, providing accurate and comparatively quicker results. However, these assays are expensive and require trained personnel and complex instrumentation, both of which are also associated with great costs, and make them impossible to use for rapid on-site monitoring [8]. Because of these setbacks, a lot of work has been done in the biosensors field, in order to develop systems that are reliable and fast but also economically viable and that allow in situ detection [12]. In general, biosensors are devices that comprise three main components: a bioreceptor, that provides specific biorecognition for a target analyte, a transducer, that transforms this biochemical reaction into a measurable signal, and a signal readout processor [13]. Biosensors can be classified by their biorecognition element, which can be based on catalytic (enzymes, cells, or tissues) or affinity biosensors (antibodies (AB) or nucleic acids) [14]. But they can also be classified according to the transducing mechanism, which can be electrochemical, piezoelectric, optical, and thermometric, among others [15]. Particularly, optical biosensors have been studied extensively since they can offer real-time and in situ detection compared to conventional analytical techniques, and they can be based on multiple biosensing principles [16].

Regarding the biomedical field, surface plasmon resonance (SPR) sensors are especially useful in biosensing applications [17]. Surface plasmons (SP), a concept first introduced by Ritchie in 1957 [18], can be defined as collective oscillations of free electrons at the interface between two materials with dielectric constants of opposite signs. This typically occurs between a thin metal film and a dielectric medium, which is usually a liquid and represents the analyzing environment [19–21]. When light traveling through the dielectric medium hits the metal surface under the resonance condition of the SP, i.e., when the energy and momentum of both the incident light and SP match, SPR is generated [22,23]. This phenomenon results in a partial omission of reflected light intensity, also referred to as a sharp spectral dip. SPR sensors are a very good approach for biosensing since the resonance condition is strongly dependent on the dielectric medium RI so any small changes, such as surface modifications, variations in composition or concentration of the analyzing sample, or even biomolecular binding events can cause a measurable shift in resonance [17].

The standard components of an SPR sensor are a light source, an optical coupling component (either a prism, grating, waveguide, or optical fiber), an imaging optical system, and a detector [24]. The first most relevant SP configurations involved a prism: the Otto configuration and the Kretschmann configuration. In the Otto configuration, the analyte separates the prism and the metal layer. When the prism is illuminated with an incidence angle greater than the critical angle, light suffers total internal reflection (TIR) and an evanescent wave (EW) is created at the prism/analyte interface. The EW excites the SP in this interface, causing SPR. However, EW decays with distance, so to excite the SP, the metal layer had to be placed very close to the prism. Kretschmann updated this configuration, placing the metal layer between the prism and the analyte layer. This new configuration enhances the propagation of the EW, so SPR is more easily achieved [23].

Although these configurations are already mature for SPR sensing, their bulk, complexity, and troublesome manufacturing cause their application in remote, narrow, and long-distance sensing to be very limited [23,24]. Thus, there has been a focus on applying this technology to optical fibers, which offer great advantages over traditional prism-based SPR systems, such as simplicity, cost-effectiveness, and more importantly, miniaturization [25]. Their compact structure enables their application in narrow spaces, real-time detection, and in situ measurements [23].

A great advantage of optical fibers in telecommunications poses a challenge when it comes to biosensing. Since the core RI is higher than that of the surrounding cladding, incident light is confined within the core region, as a result of TIR [26]. This provides minimal losses to the surrounding environment, which in turn prevents the interaction of the light with the fiber coating and analytes [22]. In fact, in a standard optical fiber, the evanescent field is almost zero in the cladding, which makes it impossible to achieve SPR, so the solution is to modify the fiber. Two main types of optical fiber biosensors based on SPR can be considered: grating-assisted and geometry-modified [23]. Regarding geometry-modified sensors, there is a considerable number of configurations, including uncladded, U-shaped, D-shaped, tapered, and end-face reflected [23]. All these structures are based on light transmission except the end-face reflected configuration, in which a portion of the fiber’s tip is uncladded, and a reflective layer is applied. The light is reflected backwards once it reaches this reflective layer, doubling the optical path, which enhances the response of the sensing region [27]. The unique design of end-face reflected fibers, also called Tips, makes them especially attractive when the intended application is to use the sensors as probes.

Most SPR sensors, both commercial and presented in literature, use gold (Au) and/or silver (Ag). Au presents better SPR performance and much higher chemical stability, whereas Ag tends to oxidize when exposed to air, resulting in short lifespans [22]. On top of this, to further increase sensitivity, the addition of different materials, such as graphene [28] or MoS₂, to the typical metal layer has been tested [29]. The direct band gap and hydrophobic nature of MoS₂ enable a highly sensitive detection of bio targets since hydrophobic surfaces exhibit a high affinity for protein-surface adsorption. Its hydrophobicity is found to be comparable to that of an Au surface and, in addition, MoS₂ being a 2D material has a large surface density; these suggest that the immobilization process of proteins and AB on a MoS₂ surface is similar to that on an Au surface [30].

ABs are Y-shaped glycoproteins that belong to the immunoglobulin superfamily and are formed by a fragment crystallization region and two fragment antigen binding regions in which the specific antigen binding site is located [31]. A prominent class of affinity biosensors, called immunosensors, take advantage of the highly specific and selective immunoreaction that takes place once an AB recognizes and binds to a specific antigen [32]. After the AB are immobilized on the sensing surface, when it is brought into contact with a solution containing analyte molecules, the binding between AB and analyte occurs, changing the surrounding RI on the sensor’s surface. This change gives rise to a change in the propagation constant of the SP, which can be measured by a change in one of the characteristics of the incident light, such as intensity, wavelength, or phase [33].

Many different configurations have been developed in the last few years regarding biosensors for the detection of pathogens, namely E. coli [5,34–37]. However, even though the developed biosensors had good performances, there are other important factors, such as the possibility for miniaturization, a short response time, and simple instrumentation and analysis procedures that are still lacking. In fact, to achieve high performance and specificity, many proposed biosensors rely on very complex instrumentation, which makes them difficult to implement in real environments due to the financial and logistical burden they pose to the industry [38].

Thus, we aimed to develop and characterize an Au-coated optical fiber immunosensor with an end-face reflected configuration for E. coli detection based on the SPR effect. First, we characterized the fiber to the surrounding RI using glucose solutions. Afterward, Au functionalization with anti-cortisol antibodies was performed, using cysteamine (Cys) as the intermediate linker. The functionalized immunosensors were then tested for different E. coli concentrations ranging from 1 to 10⁵ cells/mL, contaminated water samples, and different bacterial strains to assess specificity and selectivity.

2. Materials and methods

2.1 Reagents

All reagents used in this work are listed below. Deionized (DI) water, obtained from a Milli-Q water purification system, was used throughout the work. Phosphate buffer saline (PBS) (pH = 7.4, 10 mM) was obtained from Fisher Bioreagents, USA. Glucose (D-(+)-Glucose ≥99.5%), sulfuric acid (H2SO4, 95-97%), cysteamine hydrochloride (≥ 98%), MoS₂ and 1-Methyl-2-pyrrolidinone (NMP) 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% v/v) were obtained from Merck, Germany. Anti-E. coli serotype O/K Polyclonal AB (5 mg/mL) was acquired from ThermoFisher Scientific, Switzerland. Lysogeny broth (LB) growth medium was acquired from NZYTech, Portugal.

2.2 Bacterial strains and depuration scenario testing

A partner from NOVA School of Science and Technology provided the bacterial samples, which were propagated in LB medium at 37 °C with aeration. The main strain used was E. coli ATCC 35218, a quality control testing strain, also referred to in this work as only E. coli for simplicity. However, a different type of E. coli (ATCC 25922) and Morganella morganii sibonii, Vibrio cyclistrophicus, and Pseudoalteromonas marina, isolated from an aquaculture plant in Portugal, were used to assess selectivity and specificity. The strains were selected because of their large presence in water and genetical proximity to E. coli. Morganella was selected as it is the closest available bacteria to E. coli isolated from aquaculture (they belong to the same order). Vibrio e Pseudoalteromonas are both gram-negative rods, like E. coli, and they are very abundant in water.

To estimate concentration, the optical density (OD) of the suspensions was measured using ultraviolet-visible (UV-Vis) spectroscopy. The UV-Vis spectra were obtained on a Shimadzu UV-2501 PC spectrophotometer, 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 estimate the concentration of cells in a liquid, exhibiting little damage or growth hindrance to cells [39].

The OD, obtained from the value of absorbance measured by the spectrometer is directly related to concentration and, for E. coli, the correspondence between 1.0 OD and 8 × 10⁸ cells/mL was considered. The desired concentrations for each strain were obtained by serially diluting the initial bacteria suspensions with the corresponding growth medium.

Since the long-term objective of developing these sensors is to use them in aquaculture settings, it was considered useful to conduct some tests making use of samples of contaminated water rather than relying solely on bacterial cultures. To accomplish this, a collaboration with researchers from CESAM and Department of Biology was formed as part of DigiAqua and DepurD projects. Water samples were gathered at various phases of the depuration process and subsequently utilized as the sensors’ testing solutions. However, at the time of these tests, it was not possible to estimate the E. coli concentration in each sample, but it is known that, during depuration, the bacteria concentration decreases over time. Thus, it was still possible to evaluate if the developed sensors could detect this decrease, which is of the utmost importance, as these types of samples resemble more closely the intended end use of the sensors.

2.3 End-face reflected gold-coated silica optical fiber production

To develop the desired sensors, T400 Step-Index multimode fibers with a core diameter of 400 μm (Thorlabs, USA) were used. These are made up of a silica core, surrounded by a 25 μm polymeric cladding and further protected by an external hard cladding.

The process to achieve the end-face reflected geometry (Tips) is summarized in Fig. 1. It starts with pieces of optical fibers (15 cm), whose cladding is removed at one end with a clamp, creating unclad regions of around 4 cm in each fiber. To remove the hard polymer cladding, this portion was soaked in acetone and then wiped using an acetone-soaked cloth. After that, to obtain an unclad sensing region of 1 cm, each optical fiber was cleaved at a 90° angle. The unclad region was cleaned using piranha solution (H₂SO₄ and H₂O₂ 3:1) and then silanized in methanol using 1% (trimethylsilyl)-3-propanethiol solution for 15 minutes, after which the fibers were left to dry at room temperature overnight.

 

Fig. 1. Schematic representation of the production of Au-coated end-face reflected fibers (Au- Tips).

Download Full Size | PDF

The prepared fibers were then coated with an Au film by the sputter deposition process, which is a physical vapor deposition method that involves extracting the material (in this case Au) from the cathode by bombardment with accelerated ions [40]. The Au coating was applied to both the sides and the end of the fibers to ensure proper reflection of the light. To achieve this, the fibers were mounted on a support with the uncladded portion placed vertically and pointing towards the cathode and a single Au sputter deposition of 50 nm thickness was performed, which has been reported in the literature to maximize sensitivity [41]. The deposition was performed under vacuum (10⁻⁴ mbar) with argon, into a LEICA EM SCD500 sputter-coater with integrated quartz microbalance to measure the Au thickness.

2.4 Modification with MoS₂

Three Tips were further modified with MoS₂, a material that given its layered structure, can be exfoliated into individually separated thin layers using the ultrasound-assisted liquid exfoliation technique, for example, which is a method that makes it possible to exfoliate bulk materials by overcoming the weak van der Waals force between their layers [42].

Hence, 30 mg of MoS₂ powder was added to 10 mL of NMP, in a 20 mL beaker and then placed in an ultrasonic bath sonicator (Ultrasons-HD) for 30 minutes, followed by 30 minutes using the probe sonic tip (UP100 H). Finally, the dispersion was centrifuged at 4000 rpm for one hour. To coat each Tip, it was dipped in 1 mL of the solution containing the exfoliated MoS₂ nanosheets for 8 cycles, each comprising 20 seconds of immersion, followed by a drying stage of 2 minutes. To ensure proper robustness, the optical fibers went through an annealing process for 2 hours at 50°C [5].

2.5 Experimental setup

The same experimental setup was used throughout this work, which takes advantage of the probe-like nature of the Tips (Fig. 2). Using a vertical support, the fiber was suspended over an eppendorf, held in place by a metallic support placed on an elevating platform. This allowed the eppendorf to be lifted, bringing the solution in contact with the Au Tip, or lowered, so it could be swapped for an eppendorf with a different solution. Connected to the other end of the fiber, through a bifurcating optical cable, were a light source (tungsten halogen light source, HL-2000 from Ocean Insight, USA), with an emission range from 380 nm to 2500 nm, and a spectrometer (FLAME-T-UV-vis manufactured by Ocean Optics, USA) with a detection range of 180-890 nm and resolution of 0.19 nm. The acquired spectra were displayed on a laptop, connected to the spectrometer, using OceanView software provided by Ocean Insight.

 

Fig. 2. Schematic representation of the experimental setup for the tests performed throughout this work using the developed Tips.

Download Full Size | PDF

2.6 Optical characterization

SPR-based optical fiber sensors, such as those developed here, are highly sensitive to changes in RI, so it is important to evaluate their performance before further testing with E. coli. This RI characterization was carried out by immersing the fibers in glucose solutions with different concentrations and, consequently, different RIs. Eight different solutions were prepared with 0, 1, 5, 10, 20, 30, 40, and 50% (w/v) glucose, which correspond to 1.333, 1.334, 1.340, 1.346, 1.358, 1.368, 1.377, and 1.386 RIU, respectively. These values were obtained by analyzing the solutions using a refractometer (Abbemat 200, Anton Paar) at 21°C.

After the acquisition of reference spectra with the sensors only exposed to air, they were immersed in DI water as a cleaning step. Then, in increasing order of concentration, they were consecutively immersed in each glucose solution for 2 minutes. After this waiting period, a spectrum was recorded for each solution, on a laptop using OceanView software, applying an integration time of 1000 ms, averaging of 10 scans, and a boxcar width of 10.

The reference spectrum of air was subtracted from the acquired spectra, and the resulting data was smoothed with an FFT filter of 100 points, using OriginLab software. It was then possible to analyze the shift of the SPR signature by analyzing the resonance wavelengths, which correspond to the minimums of the spectra. To calculate the sensitivity, the changes in the resonance wavelength were plotted as a function of the RI of the different solutions. The sensor’s sensitivity was then given by the slope of a linear adjustment of the experimental points:

2.7 Functionalization

The functionalization step aims to immobilize ABs on the surface of the sensors, to achieve E. coli responsiveness (Fig. 3). From the many approaches available for this effect, given the strong physicochemical interaction between Au and sulfur, the Au Tips were first exposed to a thiol derivative, cysteamine (Cys). Then, using EDC/NHS, the ABs were covalently immobilized through the carboxylic acid (-COOH) functional groups to the -NH2 functional groups of Cys in the fiber surface. This is possible because EDC activates carboxylic groups, promoting the formation of amide bonds when in the presence of amine groups, a reaction that can be stabilized by using NHS [22].

 

Fig. 3. Schematic representation of the functionalization procedure for the developed Tips, highlighting the immobilization of anti-E. coli AB on the Au surface.

Download Full Size | PDF

Firstly, to clean any debris off the fiber, it was immersed in DI water for two minutes, followed by two minutes in PBS and another two minutes in DI water. To obtain an amine-terminated fiber, the Tip was left to incubate for eight hours in an aqueous solution of Cys (20 mM, 250 μL). After this time, the Tip was washed three times in DI water to remove unbound Cys, before being placed in PBS. Afterward, it was immersed for two hours in a mixture (pH = 7.4) of 100 μL of AB solution (500 μg/mL), 50 μL of EDC (0.2 M, in PBS), and 50 μL of NHS (0.5 M, in PBS). Then, it was washed with PBS, before the surface passivation with BSA solution (10 μg/mL, prepared in PBS) for 30 minutes, to avoid non-specific interactions. Finally, the Tip was cleaned with and left immersed in PBS overnight, to ensure that any unbound component was subsequently eliminated.

All the steps were performed at a controlled room temperature of 21 °C, and the optical response of the fibers was monitored at each step by acquiring the corresponding reflection spectra.

2.8 E. coli detection tests

Finally, to assess the sensors’ performance, different E. coli suspensions with concentrations of 1, 10, 10², 10³, 10⁴, and 10⁵ cells/mL were prepared. Before the functionalized fiber could be used, it was taken off the PBS and left to air dry, and a new reference spectrum was acquired in the air. The Tip was then immersed in PBS one last time for 30 minutes, before being exposed to each bacteria suspension.

The fiber was first immersed in the control sample (sterile LB medium) and after ten minutes, it was cleaned three times with PBS, and the correspondent reflection spectrum was recorded with the fiber still completely immersed in this solution. This process was then repeated for all suspensions in order of increasing concentration. Alternatively, when using the depuration water samples, the procedure was the same, except DI water was used instead of PBS for the washings in between samples and to acquire the spectra after contact with the contaminated samples. Additionally, DI water was also used as the control sample, and the test started with the sample collected after 5 hours of depuration and ended with the sample collected right at the start, to ensure that it was done in increasing order of concentration.

2.9 Selectivity and specificity tests

To evaluate the sensors regarding selectivity towards E. coli, similar detection tests were performed using other strains of bacteria, namely Morganella morganii sibonii, Vibrio cyclistrophicus, and Pseudoalteromonas marina. This time, the prepared suspensions all had the same OD₆₀₀nm, equivalent to 10⁵ E. coli cells/mL. The essay started by using just the LB medium as a control, followed by each of the different species of bacteria and ended with E. coli. This was done in the hopes that the sensors only reacted to E. coli, so it should be the last species to be tested. Two different strains of E. coli were also used, namely E. coli 25922 and the E. coli (35218), to assess specificity.

2.10 Scanning electron microscopy and energy-dispersive x-ray spectroscopy

Scanning electron microscopy (SEM) is one of the most commonly used techniques for sample characterization, as it provides information on surface topography and morphology. Alongside SEM, energy-dispersive X-ray spectroscopy (EDX) is a technique that provides important information regarding the identification of elements on the sample surface. Through this procedure, it is possible to identify different elements and estimate their distribution, providing a chemical mapping of a sample’s small area [43].

SEM images of the optical fiber sensors were acquired using TESCAN Vega3 SB equipment in secondary electron mode with a high voltage of 15.0 kV. The fibers were attached to an aluminum sample holder using double-sided carbon tape. EDX was performed using a Bruker Xflash 410 M Silicon Drift Detector incorporated into the SEM equipment.

3. Results and discussion

3.1 Optical characterization

To evaluate the response of the sensors to RI changes, three different fibers were immersed in solutions with increasing glucose concentrations (0-50%w/v). Figure 4(a) evidences the changes in optical spectra attained for one of the sensors, where it is possible to note that there is a noticeable shift to longer resonance wavelengths (a redshift) with increasing glucose concentration. This behavior was verified for all three fibers tested. The obtained resonance wavelengths, which correspond to the dips in the spectra, were then used to obtain each sensor’s sensitivity to RI changes.

 

Fig. 4. Results of the optical characterization using three different fibers: (a) Example of the reflection spectra recorded for glucose solutions in a concentration range from 0 to 50%, for one fiber; (b) Average resonance wavelength variation as a function of RI (results for triplicates are shown), and respective linear fit.

Download Full Size | PDF

As represented in Fig. 4, across the three fibers the resonance wavelength changed, on average, from 605.33 ± 6.14 nm to 708.02 ± 2.29 nm, corresponding to an average sensitivity of 1849.9 ± 63.0 nm/RIU with R²= 0.991.

The error bars evidence minor differences in sensitivity between the three fibers, which is expected, as it is impossible to ensure that all fibers are exactly equal and that the thickness of the Au coating is exactly the same in all of them. Nonetheless, the results show that the three fibers can accurately measure changes in external RI, which is an essential requisite for them to be applied as E. coli sensors.

3.2 Functionalization

After the RI characterization, the fibers went through a very important functionalization procedure, aiming to immobilize the anti-E. coli AB on the surface of the Au Tips. To study spectral changes caused by each step, the fibers were immersed in PBS and a reflection spectrum was acquired, to establish a comparison with the initial spectrum obtained in PBS, before functionalization.

The spectral evolution for one of the fibers is shown in Fig. 5(a), evidencing a redshift, a behavior common to all sensors. The average wavelength shifts associated with each step across the three fibers were thus obtained (Fig. 5(b)). The immersion in Cys allowed the formation of a self-organized monolayer on the fibers’ surface, which led to an average redshift of 18.01 ± 7.37 nm. Next, after the AB immobilization onto the fiber’s surface, an average redshift of 24.83 ± 9.32 nm was recorded. The last step, which was the surface passivation with BSA, resulted in a total average redshift of 25.79 ± 8.96 nm, which represents a very small change from the wavelengths recorded after AB immobilization. Considering that every measurement was carried out in the same surrounding RI and under identical ambient conditions, the spectral changes had to be a result of RI changes on the surface of the fibers. These changes were thus caused by the formation of a Cys monolayer, immobilization of the anti-E. coli AB onto the fibers’ surface, and the passivation with BSA. However, it is worth noting that the observed wavelength shifts have a considerable level of variability between tests, evidenced by the error bars in Fig. 5(b). Since the fibers were subjected to the same reagents, this variability may be attributed to differences in the fibers’ morphology, namely the uniformity of the Au coating. Nonetheless, the obtained results suggest that the surface functionalization was successful.

 

Fig. 5. Functionalization results: (a) Spectra collected in PBS after each functionalization step for one fiber; (b) Average resonance wavelength shifts obtained in PBS after each biofunctionalization step (results for triplicates are presented).

Download Full Size | PDF

3.3 E. coli suspension tests

The three functionalized were ready to be tested as E. coli immunosensors. Their performance was evaluated by monitoring shifts in resonance wavelength, as a result of immunorecognition between the immobilized AB on the fibers’ surface and the bacteria present in the different suspensions.

The sensors were tested with concentrations ranging from 1 to 10⁵ cells/mL, a broad range that should cover the concentrations of legal interest in bivalve aquaculture. The optical spectra obtained throughout a detection test with one of the fibers are highlighted in Fig. 6(a), where it is possible to note that there was a redshift with increasing concentration, a behavior that was common to the three fibers. From these spectra, it is possible to obtain the resonance wavelength for each sample, which corresponds to the dip in the spectra. For each concentration, the corresponding wavelength shift was obtained in reference to the resonance wavelength for the control sample (LB medium).

 

Fig. 6. (a) Spectral response of one functionalized Au-Tip, acquired in PBS, as a response for E. coli suspensions in a concentration range from 1 to 10⁵ cells/mL, after incubation for 10 min. (b) SPR signature wavelength shift as a function of the logarithm of E. coli concentration (results for triplicates are shown).

Download Full Size | PDF

This analysis was performed for the assays with the three different fibers, so an average of the wavelength shift in relation to the control samples could be plotted (Fig. 6(b)), considering the error in each measurement to be the standard deviation between the three independent tests. The obtained results indicate that there was an average total redshift of 1.49 ± 0.05 nm and that the sensors present a logarithmic response to he increase in E. coli concentration.

A logarithmic sensitivity (LS), which can be defined as the shift of the SPR wavelength per unit change in the logarithm of E. coli concentration, of 0.21 ± 0.01 nm/log(cells/mL) and an R²= 0.979 were obtained.

Considering that all spectra were taken in fresh PBS, the observed wavelength shifts that occurred must be due to the detection of E. coli. As the concentration of E. coli increases, more bacteria bind to the AB, thus increasing the surrounding RI of the fibers’ surface. This increase in RI is what causes the observed redshift.

It can be concluded that it is possible to detect concentrations as low as 1 cell/mL, a concentration that resulted in an average wavelength shift of 0.42 ± 0.06 nm in the three tests. This marks a great improvement to what was previously reported in the literature for similar approaches [5], where more complex instrumentation was used, namely an etched fiber and a flow cell, and a more limited range of E. coli concentration was tested. Results suggest that the developed sensors can be successfully applied in aquaculture since their operating range should be large enough to make the distinction between zone classes.

3.4 E. coli in water from depuration tests

In addition to the tests with E. coli suspensions, one test was also performed using contaminated water samples, to evaluate the performance of this type of sensor in a more complex matrix, such as water samples that contain other substances besides just E. coli. This is important, as the presence of other types of materials could interfere with the binding of E. coli to the AB, thus leading to inaccurate results.

The water samples were directly collected from a contaminated tank of circulating water, which was subjected to a UV lamp to eliminate the existing bacteria, using a pasteur pipette, before they were properly identified and stored for further use. Samples were taken at the start of this process (0 hours), then after 30 minutes, 3 hours, and 5 hours. As the depuration process progresses, there is a decrease in E. coli concentration, so the sample collected at 5 hours is much less contaminated than the sample taken at the start. The results from this test are present in Fig. 7.

 

Fig. 7. (a) Spectral response of a functionalized Au-tip, acquired in DI water, as a response for contaminated water samples, after incubation for 10 min. (b) SPR signature wavelength shift as a function of the depuration time.

Download Full Size | PDF

A longer wavelength was detected in the samples containing more bacteria, i.e., the samples taken earlier. These results suggest that the developed sensors can detect E. coli in more complex water samples, which is a promising step toward their implementation in aquaculture. Nonetheless, further investigation is still needed, as the collected data was very limited and it was not possible to evaluate the concentration of E. coli in the collected water samples by traditional methods, which prevented a more thorough analysis from being performed.

3.5 Modification with MoS₂

To evaluate the impact of the MoS₂ modification, three fibers were modified as explained previously and tested in the same range of E. coli concentration, from 1 to 10⁵ cells/mL. From the obtained data, the average resonance wavelength shifts for each sample could be plotted as a function of the logarithm of E. coli concentration (Fig. 8). It is possible to observe that there was a larger overall shift in the resonance wavelength when compared to the Au-Tips. While the previous value for the 10⁵ concentration was found to be 1.49 ± 0.05 nm, the addition of MoS₂ resulted in a recorded shift of 2.17 ± 0.05 nm.

 

Fig. 8. Resonance wavelength shift as a function of the logarithm of E. coli concentration for the Au-Tips (black squares) and an Au/MoS2-Tip (red circles) and respective linear fits (results for triplicates are shown).

Download Full Size | PDF

Additionally, the sensitivity was also calculated, once again in the form of LS, revealing a result of 0.33 ± 0.03 nm/log(cells/mL), and a R²= 0.970. The addition of MoS₂ had no influence on the linear response the Tips had already exhibited with the logarithm of E. coli concentration, nor did it influence the detection range. However, a lower variability between the three different tests was found, evidenced by the smaller error bars in Fig. 8, when compared to the Au-Tips, suggesting better reproducibility.

Overall, the addition of MoS₂ is a good option to increase the sensitivity of SPR-based optical fiber sensors, and in this case, an increase of approximately 50% was achieved.

3.6 Selectivity and specificity tests

To evaluate the sensors’ selectivity towards E. coli, an essay with different bacterial species often found in aquatic environments was performed. Namely, Morganella morganii sibonii, Vibrio cyclistrophicus, and Pseudoalteromonas marina were used. Two different strains of E. coli were also used, namely E. coli 25922 and the already previously tested E. coli (35218), to assess specificity. The AB used was polyclonal, so the developed sensors should respond to different strains of E. coli. All bacterial suspensions had the same OD₆₀₀nm.

Figure 9 shows the comparison in histogram form of the resonance wavelength shift for each bacterial sample, in the order that they were tested. To establish a comparison with the previous E. coli detection test described in section 3.3, the value of the wavelength shift corresponding to the same OD₆₀₀nm is also presented (E. coli 35218 (previous test)). It is possible to note that after the fiber was in contact with the Morganella bacteria, a small redshift of 0.13 nm compared to the control sample occurred, which can be justified by the fact that these bacteria are part of the same family as E. coli and some cross-sensitivity can occur. After this, the fiber was immersed in Vibrio bacteria, which had no influence on the resonance wavelength. This is the expected result, as no bond of these types of bacteria to the fiber’s surface, as intended. Thirdly, the fiber was immersed in Pseudoalteromonas, and a blueshift occurred, making it so that the total wavelength shift dropped to 0.06 nm. This may be because the weak bonds that had previously occurred with other bacteria were now washed away. Nonetheless, it is a positive result that no redshift was monitored. Finally, when the fiber was put in contact with the new E. coli strain, a large redshift of 0.32 nm was achieved. This suggests that the developed sensors can efficiently and selectively detect more than one strain of E. coli. Lastly, when using the same E. coli as before, a redshift of 0.51 nm was found. This further confirms that the sensors are much more sensitive to E. coli than other bacteria.

 

Fig. 9. Histogram comparison of the shift in resonance wavelength caused by immersion for 10 minutes in suspensions of different strains of bacteria for one Au-Tip.

Download Full Size | PDF

Nonetheless, if these results are compared with the one obtained during the E. coli detection tests, it is possible to see that this is much lower. In fact, previously, the recorded shift was 1.49 ± 0.05 nm. This difference could be because the fiber was now subject to higher optical density bacteria from the start, some of the AB may have been degraded and washed off, and by the time the sensor was immersed in E. coli, there were much fewer free AB for E. coli to bond to.

3.7 SEM and EDX characterization

To evaluate surface topography and morphology, alongside the fiber’s elemental composition, an Au-Tip was observed by SEM and EDX. Figure 10(a) shows an image of the fiber, which reveals a clean and uniform surface. However, other portions of the Tip appeared less uniform, and with increased magnification, it was possible to notice some fissures on the surface (Fig. 10(b)). To further analyze this, another image was taken, under higher magnification, to perform EDX analysis. From this, it was possible to obtain the elemental characterization of the fiber. The obtained EDX spectrum, present in Fig. 10(c), reveals clear peaks of oxygen (O), silicon (Si), and Au, which correspond to the peaks of the highest intensity. This result was expected, as it is a silica (SiO₂) fiber with an Au coating.

 

Fig. 10. SEM images of an Au-Tip with (a) 400x and (b) 15kx magnification; (c) EDX spectra and elemental mappings of (d) Au, (e) Si and (f) O of an Au-Tip.

Download Full Size | PDF

Additionally, in Fig. 10(d)–(f), it is possible to note the individual Au, Si, and O maps, respectively. These results reveal that there is indeed Au on the surface of the fiber, although, there is not a uniform distribution. There are some areas in which there is barely any Au detected. Coincidentally, when comparing with the Si map, it is possible to conclude that the areas that lack Au indicate a strong presence of Si. From this, it is safe to conclude that some portions of the fiber are not fully coated in Au, leaving the SiO₂ exposed.

This analysis was performed after the fiber was used in an E. coli detection test, however, the present defects most likely were not due to this process. A much more reasonable explanation is that the sputtering procedure was unable to produce a perfectly even coating. It is desirable, in the future, to adapt the method employed to ensure that all produced fibers have the same uniform Au coating.

4. Conclusion

Considering the current paradigm of E. coli detection, it is evident that there is significant potential for improvement. Thus, the primary aim of this work was to develop end-face reflected Au-coated optical fiber sensors based on the SPR effect, functionalized with anti-E. coli AB. Three independent tests with three identical fibers were performed and a linear relation between resonance wavelength shift and the logarithm of E. coli concentration was obtained, which translated into an average logarithmic sensitivity, LS of 0.22 ± 0.01 nm/log(cells/mL) and a working range of 1-10⁵ cells/mL. This sensitivity value was increased to 0.33 ± 0.03 nm/log(cells/mL) by coating the fibers with MoS₂ while maintaining the same detection range. In the future, more work should be done regarding the use of Au nanoparticles instead of the Au film, as it may lead to an increase in sensitivity (because SPR is localized).

Selectivity and specificity tests carried out with other types of bacteria showed that the developed sensors have a highly specific response to E. coli compared to other bacteria. Preliminary tests conducted with contaminated water samples showed that these sensors can be employed for the detection of E. coli in more complex matrices. Both these results are essential for their intended purpose of being implemented in a real-world scenario, such as aquaculture, as the presence of other compounds or species of bacteria could lead to false positive results. The results shown in this work represent a positive progress compared to the traditional methods and the systems described in the literature.