Background

Nanomaterial enabled sensors are an exciting technology that provide exquisite detection, on the nanomolar to sub-picomolar level, of environmental contaminants [1,2,3,4,5]. Interest in these sensors stems from their potential for facile, in-field contaminant detection without the need for expensive lab equipment. Many past reviews in this area have grouped sensors based on the signal transduction method [2,3,4,5], nanoparticle backbone [7,8,9,10], or contaminant class [1, 11, 12], thus leaving one important paradigm virtually untouched: classifying sensors based on the analyte(s) of interest. Because environmental scientists and engineers are often interested in determining if a specific contaminant exists at a field site and if its concentration is above the regulatory limit, there was a need to organize a review based upon the detection of specific contaminants. This review has been developed to address these concerns. First, we summarize the general concepts underlying a nano-enabled sensor and then discuss recent developments in nanomaterial enabled detection of nine specific analytes: two pesticides, four metals, and three pathogens. A nearly infinite number of chemicals of environmental concern exist and although it would be impossible to outline all of them, the fundamental nanosensor designs can be seen in the examples outlined within the review. For the reader interested in nanosensors for pharmaceutical detection we direct them to the work of Nagaraj et al. [13] and the reviews of Sanvicens et al. [14] and Cristea et al. on antibiotic detection [15].

Introduction

Nanomaterial enabled sensors consist of three components: a nanomaterial(s), a recognition element that provides specificity, and a signal transduction method that provides a means of relaying the presence of the analyte (Fig. 1). These components are not necessarily distinct entities within a sensor, but every nanosensor can be characterized on the basis of these three divisions. Sensors can be designed to detect a single analyte or multiple analytes, termed multiplex detection. In addition to detecting an analyte by producing a signal, a ‘turn-on’ or ‘off/on’ sensor, some of the sensors described below are based on a ‘turn-off’ or ‘on/off’ mechanism, where-by a decrease in signal indicates the presence of an analyte.

Fig. 1
figure 1

Nanosensor design schematic. First, a class and subsequently a specific contaminant of interest is selected (i). The contaminants discussed in this review are denoted with an asterisk. Next, the number of analytes to be detected by the sensor is chosen (ii) and then the probe is designed. A nanoprobe consists of two core elements, a signal transduction method and at least one nanomaterial, and may also include a recognition element (iii). Ultimately, the sensor deployment format is selected (iv)

Full size image

Nanomaterials

Nanomaterials have enabled advances in sensor design such as miniaturization, portability, and rapid signal response times. High surface area to volume ratios and facile surface functionalization make nanomaterials highly sensitive to changes in surface chemistry thus enabling nanosensors to achieve extremely low detection limits. In some cases, the enhanced sensitivity of nano-enabled sensors is due to the fact that nanomaterials are of a similar size as the analyte of interest (e.g., metal ions, pathogens, biomolecules, antibodies, DNA) and are thus capable of interrogating previously unreachable matrices [4]. We briefly introduce three different general nanomaterial classes: quantum dots (QDs), metal nanoparticles, and carbonaceous nanomaterials.

Quantum dots

QDs are semiconductor nanocrystals with a typical composition MX where M is commonly cadmium (Cd) or zinc (Zn) and X is selenium (Se), sulfur (S), or tellurium (Te). QDs are often coated by a second MX alloy, a shell, to create core/shell QDs with highly tuned properties. Common QDs employed in sensor applications include: CdSe [16], CdSe/ZnS [17,18,19], CdTe [20,34]. The choice of cap** agent depends on the desired function and nanoparticle composition. The interested reader is referred to recent reviews by Saha et al. [7] and Wei et al. [35] for additional details on gold enabled sensors.

A range of nanostructured metal oxides (NMOs) have been explored for sensing applications. NMOs include: iron oxides, titanium oxides, zirconium oxides, cerium oxides, zinc oxides, and tin oxides. Magnetic iron oxides, such as magnetite (Fe3O4) and maghemite (γ-Fe3O4), have low toxicity, are economically friendly, and can be easily functionalized with ligands, antibodies, and other cap** agents [36]. One important allure of magnetic NPs arises from their use in facilitated separation processes and remediation applications [12]. Titanium dioxide, TiO2, has also been embraced in nano-sensor design [37, 38], but it is most typically used and studied for its photocatalytic properties.

Carbon-based nanomaterials

Carbon nanotubes (CNTs) and graphene are often employed in nano-enabled sensors because of their large surface area, excellent electrical conductivity, high thermal conductivity and mechanical strength [39]. One recent application of these nanomaterials has been their use to increase the sensitivity of glassy carbon electrodes (GCE) for electrochemical sensing [40, 9, 72]. Magnetic relaxation switches have been used to detect nucleic acids (DNA and mRNA), proteins [73] and viruses [74] among other targets.

Analytes

As defined at the outset of this review, a wide variety of different analytes can be detected by nanomaterial-based sensors. In this portion of the review, we focus explicitly on the applications of nanosensors towards detection of pesticides, metals, and pathogens.

Pesticides

There is great interest in detection of pesticides given their widespread use, their toxicity, and their proclivity for bioaccumulation. Currently, over 800 active ingredients, in 100 different substance classes are present in commercial pesticides [75]; we summarize the major pesticide classes in Table 1. Organophosphorus (OP), carbamates, neonicotinoids, and triazines are the dominant classes and to date have been the focus of nano-enabled pesticide detection. Liu et al. [75], Verma et al. [76], Aragay et al. [1], Evtugyn et al. [60] and Pang et al. [77], provide detailed reviews of pesticide detection techniques. In this section, a brief background on pesticide detection will be followed by a discussion of recent advances.

Table 1 Common pesticide classes
Full size table

Organophosphates

Pesticides are often designed to impact a specific enzyme; many forms of pesticide detection are based on observing and monitoring this enzyme either directly or indirectly. Organophosphate and carbamate pesticides inhibit the production of acetylcholinesterase (AChE) an enzyme that catalyzes the hydrolysis of acetylcholine, a neurotransmitter [78, 79]. The fundamental reaction is shown in Eq. 1.

$$acetycholine + {\text{H}}_{2} {\text{O}} \mathop \to \limits^{AChE} choline + acetate.$$
(1)

A class of rapid and sensitive electrochemical sensors has been developed around the immobilization of AChE on a solid electrode surface [11] and Ullah et al. [102].

Mercury

The negative neurological effects of mercury exposure to humans have driven extensive investigation into the geochemical cycling and detection of this element [103]. A major focus of mercury (HgII) nanosensor development has been the production of DNA-based probes [47,48,49,50, 104,105,106]. Thymine–thymine (T–T) base-mismatches in DNA are significantly stabilized in the presence of HgII [104] due to the formation of metal base pairs [107]. Two major types of oligonucleotide mercury probes have been reported in the literature: G-quadruplexes [48, 49], which unfold, and nearly complementary single strands, which hybridize [106]. A growing number of mercury sensors are being constructed using multiple nano-elements, such as the mercury sandwich assay described by Liu et al. [50]. In this assay, magnetic silica spheres encapsulated in a gold shell and Raman labeled gold nanoparticles were functionalized with complementary DNA sequences that contained five mismatched thymine sites, Fig. 5. The DNA sequences were chosen such that the binding energy between the complementary aspects of the strands was insufficient to allow them to fully hybridize. In the presence of mercury, full hybridization occurred thus decreasing the inter-probe spacing and creating a plasmonic hotspot. Owing to the magnetic particle cores, the nanoprobes could be easily recovered with an external magnet and subsequently recycled.

Fig. 5
figure 5

(Reprinted with permission from Liu et al. [50]. Copyright 2014 American Chemical Society)

Schematic of SERS-active system for HgII ion detection. Schematic illustration of the SERS-active system for HgII ion detection based on T–Hg–T bridges using DNA-Au NPs and DNA-MSS@Au NPs

Full size image

Thiol mediated assays for mercury detection have been described in the literature for a variety of nanoparticles such as gold [108,Cadmium

The body of work on nano-enabled sensors for cadmium (Cd) detection is less robust than that for mercury and lead, but detection limits on the order of nano-molar have been reported. A variety of nanomaterials have been explored including QDs [22, 118], single wall carbon nanotubes (SWCNT) [119], and antimony nanoparticles [120].

Gui et al. [22] described an off/on-sensor fluorescence sensor for CdII detection. Photoluminescent CdTe/CdS QDs were first quenched (i.e., turned-off), by ammonium pyrrolidine dithiocarbamate (APDC) due to the partial loss of the Cd–thiol surface layer and subsequent surface passivation. Introduced cadmium ions displaced the APDC from the QD surface and restored the photoluminescence (PL); thus, turning the sensor on. The sensor was highly selective for CdII, a threefold increase was seen in the PL intensity, and a limit of detection of 6 nM was determined.

Gui et al. [118] enhanced the accuracy of their CdII detection device by creating a ratiometric sensor. In this sensor, the fluorescence of two different chromophores was measured in order to minimize the error introduced by fluctuation in the photoluminescence of the QDs. To limit interactions between the QDs and the secondary dye, the CdTe QD cores were coated with a polymer, polyethylenimine (PEI), prior to conjugation with fluorescein isothiocyanate (FITC). The QDs were then quenched using sulfur (S2−) while the FITC signal was maintained. Again, upon introduction of cadmium the sensor was turned on and the photoluminescence was restored. The limit of detection was slightly higher for this sensor compared to the same groups initial report, 12 nM vs. 6 nM, but was linear across a much larger range, 0.1–15 µM compared with 0.1–2 µM.

Chromium

High chromium (Cr) absorption in vivo can result in various diseases, including fibro-proliferative diseases, airway hypersensitivity, lung cancer, nasal cancer, and other types of tumors [121]. Multiple immunoassays have been described for the detection of chromium [45, 46], but they are all based on the work of Liu et al. [46]. In pursuit of an immunochromatographic assay (ICA), Liu et al. developed novel anti-CrIII-EDTA monoclonal antibodies (McAb). Chromium ions are too small to elicit an immune response and thus they were mixed with the highly effective bifunctional chelating agent, isothiocyanobenzyl-EDTA, and conjugated to the carrier protein bovine serum albumin (BSA) before being introduced to mice from which the antibodies were ultimately extracted. The immunoassay dipstick was composed of the three main parts: (i) a conjugation pad that was dosed with the anti-Cr-EDTA antibodies; (ii) a test line that contained the analyte of interest, Cr-EDTA, and; (iii) a control line that contained goat- anti-mouse antibodies. To run a sample, liquid is introduced to the dipstick and travels into the conjugation pad where the probes are brought into solution. For a negative sample, the free antibody probes bind to the test line, whereas in a positive sample no probes will bind as all antibody sites are already occupied and thus no signal is produced at the test line. The antibodies at the control line will capture any probes in the solution even those that are bound to the target of interest and is use to verify that capillary action wicked the solution through the whole length of the dipstick. The ultimate result of Liu et al. was an assay with a visual limit of detection of 50 ng/mL and an analysis time of < 5 min.

Pathogens

Ever since John Snow’s 1854 revelation that cholera was spread through the consumption of contaminated water, waterborne pathogen detection has been a key area of research. The World Health Organization (WHO) recognizes twelve bacteria, eight viruses, seven protozoa, and two helminths as pathogens of significance in drinking water supplies, as outlined in Table 2 [6]. Pathogen detection methods typically focus on: (i) whole analyte (cell) detection or detection of a representative epitope on the cell membrane; (ii) genetic material detection; or (iii) pathogenic product (e.g., toxin) detection. For the sake of brevity, herein we confine our discussion to the detection of Vibrio cholerae and the toxin it produces, cholera toxin, Legionella pneumophila, which was responsible for greater than 50% of the waterborne disease outbreaks between 2011 and 2012 [122], and Pseudomonas aeruginosa, which the WHO recently classified as a critical pathogen in light of the proliferation of antimicrobial resistant species [123]. For expanded reviews we refer the reader to the works of Kumar et al. [124] and Mocan et al. [125].

Table 2 Waterborne pathogens and their significance in water supplies
Full size table

\(Vibrio\; cholerae\) and cholera toxin

Cholera, the infamous disease that spawned germ theory is now virtually unknown in the United States, but it continues to pose a major disease burden around the world with an estimated 1.3–4.0 million cases of cholera a year leading to between 21,000 and 143,000 deaths [126]. Cholera is an acute diarrhoeal disease caused by the ingestion of contaminated water or food containing the bacterium Vibrio cholerae. In the intestines, the bacteria colonize the mucosa and begin to secrete cholera toxin (CT), which initiates the disease symptoms [127]. Nanosensors have been fabricated to detect both Vibrio cholerae [128, 129] and CT, but the majority of the literature has focused on detection of CT subunit B (CT-B) [130,131,132,133,134] because the subunit induces cellular uptake of the toxin and not all V. cholerae isolates are toxigenic [135]. Label-based detection of CT can be achieved using antibodies, ganglioside GM1 (the binding site of CT), or β-galactose, a sugar with strong affinity towards CT. Ahn et al. [130] provide a nice summary of CT-B detection and reported a fluorescence resonance energy transfer (FRET) based method with a theoretical detection limit of 280 pM. In FRET, fluorescence from QDs is quenched, and the energy is transferred by another particle such as a gold nanoparticle. The quenching is inhibited in the presence of the target. Specifically, the cholera toxin binds to the β-galactose modified gold nanoparticles prohibiting the binding of the QDs.

\(Legionella\; pneumophila\)

Named for the famous 1976 outbreak at the American Legion, Legionnaires’ disease is a pneumonia like disease caused by the bacterium Legionella pneumophila. Under specific conditions, the bacterium can flourish in building (premise) plum** and infect people when they inhale aerosols containing the infective agent. Two approaches have been presented in the literature for nano-enabled Legionella detection: whole organism detection [136, 137] and DNA detection [138,139,140].

Martin el al. [136] developed a whole organism sensor that combined a sandwich immunoassay for bacterial capture with amperometric transduction. Magnetic nanoparticles were modified with poly(dopamine) (pDA) and ultimately functionalized with specific capture antibodies (C-Ab) to create MNPs@pDA-C-Ab probes. After incubation with the sample, a second detector antibody labeled with horseradish peroxidase was introduced and a magnetic field was used to capture the immunocomplexes on a screen-printed carbon electrode (SPCE). The authors found the assay to be specific for Legionella, but they needed a preconcentration step in order to achieve a LOD below the reference of 100 colony-forming units (CFU) L−1. However, with a runtime of < 3 h, compared to 10 days for the standard approach, and a LOD of 10 CFU mL−1, the sensor has the potential to be used as a rapid first screening method for highly contaminated water systems.

In a recent report, Melaine et al. [139] described the multiplex detection of 16S rRNA from Legionella, Pseudomonas aeruginosa (discussed below) and Salmonella typhimurium. A DNA microarray with capture DNA specific for each target was assembled on a surface plasmon resonance imaging (SPRi) substrate, e.g., gold coated nanoprisms. Upon hybridization of the DNA with isolated 16S rRNA a change in the reflectivity signal was observed, as shown in the bottom of Fig. 6. To extend the dynamic range of detection and enhance sensitivity, gold nanoparticles functionalized with a detection probe were introduced to the sample and ultimately RNA at concentrations as low as 10 pg mL−1 were detected.

Fig. 6
figure 6

(Adapted with permission from Melaine et al. [139]. Copyright 2017 American Chemical Society)

A schematic of multiplex RNA detection using surface plasmon resonance imaging (SPRi). A schematic of multiplex RNA detection using surface plasmon resonance imaging (SPRi). RNA fragments are first extracted from bacteria of interest (a). A biochip functionalized with three specific capture probes (CP) and a negative control probe (NP), each demarcated in a unique color (b (i)) is shown to exhibit no change in reflectivity (c (i)). Upon introduction to the RNA (b (ii)), there is an increase in single (c (ii)). Finally, gold nanoparticles functionalized with the detection probe (GNP-DP) are introduced and shown to enhance the change in reflectivity

Full size image

\(Pseudomonas\; aeruginosa\)

An opportunistic pathogen, Pseudomonas aeruginosa can be found in sources such as feces, soil, water, and sewage with the most important route of exposure being skin (dermal) contact with contaminated water or tools. Similar to Legionella, P. aeruginosa can colonize premise plumbing and has been associated with outbreaks of nosocomial infections in hospitals [141]. Most of the detection schemes reported for P. aeruginosa focus on whole pathogen detection [142,143,144,145,146] with the work of Melanie et al. [139], discussed above, on 16s rRNA detection being an outlier. In addition, to oligonucleotide recognition elements [139, 142,143,144], antibodies [145, 147] and bacteriophages [146] have also been used for specific detection of P. aeruginosa.

The first P. aeruginosa aptamer was discovered by Wang et al. [148] in 2011 and subsequently has been used in a range of sensors. The discussion that follows highlights two sensors that utilize optical transduction. Yoo et al. [142] and Hu et al. [144] fabricated nano-textured substrates to produce localized surface plasmon resonance (LSPR) chips (Fig. 7). Yoo et al. choose a three-step fabrication approach, first gold was deposited on a glass slide, silica nanoparticles were then deposited and then followed by the deposition of a second gold layer whereas Hu et al. opted for standard nanosphere lithography. The two groups also chose different methods to functionalize the sensor with Yoo et al. attaching the aptamers directly to the sensor surface via a gold-thiol bond. In contrast, Hu et al. used a polyethylene glycol (PEG) spacer to minimize steric hindrance for the aptamers with the goal of achieving a lower detection level. Hu et al. were successful at develo** a sensor with a linear response at low concentrations and a lower limit of detection, 10 CFU mL−1 vs. Yoo et al.’s 104 CFU mL−1. It should be noted that one of Yoo et al.’s goals was to create a low volume sensor and that their LOD was obtained in a 3 µL sample.

Fig. 7
figure 7

(Reprinted with permission from Hu et al. [144]. Copyright 2018 American Chemical Society)

Schematic of P. aeruginosa LSPR sensor chip (left). Sensor calibration curve, where error bar represents the standard deviation of all data points at a specific bacterial concentration (right).

Full size image

Conclusions

Nanosensor development for environmental contaminants is growing rapidly and, as described throughout this review, nanomaterials and recognition agents are continuously being combined in new and creative ways. The recent developments in sensor design aim to overcome the shortcomings of first-generation sensors such as nonspecific binding, particle size variation, nanoparticle aggregation, and nanoparticle stability. Questions of assay selectivity and sensitive in complex environmental matrices remains but a growing number of reports are using representative matrices to demonstrate the stability and selectivity of their sensors. The robustness of field deployable sensors is a must if individuals are going to be empowered to analyze their environment.