Nanoparticle-Based Liquid–Liquid Extraction for the Determination of Metal Ions

Abstract

Traditional liquid phase extraction techniques that use optically responsive ligands provide benefits that enable cost-efficient and rapid measurements. However, these approaches have limitations in their excessive use of organic solvents and multistep procedures. Here, we developed a simple, nanoscale extraction approach by replacing the macroscopic organic phase with hydrophobic polymeric nanoparticles that are dispersed in an aqueous feed. The concentration of analytes in polymeric nanoparticle suspensions is governed by similar partition principles to liquid–liquid phase extraction techniques. By encasing optically responsive metal ligands inside polymeric nanoparticles, we introduce a one-step metal quantification assay based on traditional two-phase extraction methodologies. As an initial proof of concept, we encapsulated bathophenanthroline (BP) inside the particles to extract then quantify Fe²⁺ with colorimetry in a dissolved supplement tablet and creek water. These Fe²⁺ nanosensors are sensitive and selective and report out with fluorescence by adding a fluorophore (DiO) into the particle core. To show that this new rapid extraction assay is not exclusive to measuring Fe²⁺, we replaced BP with either 8-hydroxyquinoline or bathocuproine to measure Al³⁺ or Cu⁺, respectively, in water samples. Utilizing this nanoscale extraction approach will allow users to rapidly quantify metals of interest without the drawbacks of larger-scale phase extraction approaches while also allowing for the expansion of phase extraction methodologies into areas of biological research.

Quantifying metals is an ongoing scientific problem that has implications in a wide range of disciplines.¹⁻⁴ Traditional measurement techniques that use large-scale equipment (e.g., atomic absorption, X-ray absorbance or fluorescence, and inductively coupled plasma-mass spectrometry) have benefits such as high sensitivity, identification of oxidation states, and multiple element determination. However, there are disadvantages such as complicated operations, expensive equipment and reagents, and time-intensive sample preparation. For researchers that prefer a technique with minimal sample preparation and a simple, sensitive quantification, these techniques are impractical. Alternatively, there are small-scale analytical techniques that can be applied to a wide range of applications and are easier to integrate into routine experiments in the laboratory without the need for expertise.^5,6

A typical, simpler approach for metal quantification uses liquid–liquid extraction (LLE) procedures.⁷ These assays function on the basis of metal–ligand complex formation and its relative solubility in one of two immiscible liquids, typically an aqueous phase and an organic phase. While the metal salt resides in the aqueous phase, the organic phase consists of an optically responsive, hydrophobic ligand that binds to the metal in the hydrophobic phase, stabilizes it, and changes the optical properties to enable detection. While somewhat easy to operate, these techniques can require large amounts of organic solvents, which can be toxic and hard to dispose of. Further advances like solid phase extraction (SPE), cloud-point extraction, or dispersive liquid–liquid microextraction (among others) have been developed to quantify analytes while using significantly smaller volumes of harmful solvents;⁸ however, these techniques can require expensive equipment, extensive procedures, and a need for additional expertise.

Classic examples of LLE and subsequent metal quantification originate from assays involving probes related to 2,2′-bipyridine.^9,10 These complexes generally absorb in the visible wavelength range due to their metal-to-ligand charge transfer band and can be quantified using UV–vis spectroscopy.⁹ Building on this premise, related species that have selectivity for Fe²⁺ have gained popularity.¹¹⁻¹³ For example, 4,7-diphenyl-1,10-phenanthroline (bathophenanthroline, BP) has enabled sensitive detection of Fe²⁺ down to low micromolar concentrations even in the presence of Fe³⁺ and other potential competing analytes.^14,15 While widely accepted and seemingly straightforward, this technique has required acid digestion for sample pretreatment and Fe(BP)₃²⁺ extraction into a bulk organic phase, resulting in the same LLE issues stated above. Using the same ligand, some have performed SPE to measure Fe²⁺,¹⁶ but these methods tend to have poor response times.¹⁷

Polymeric nanosensors (PNS) are an analytical tool that gives the user the ability to measure ions or small analytes by encasing optically responsive sensing components into a nanoscale hydrophobic polymer matrix. Most of the PNS are plasticized polymer nanoparticles solubilized by a surfactant coating and dispersed in an aqueous phase. There are many sensing mechanisms for PNS discussed in fundamental reviews of the field,^17,18 and more recent mechanistic developments are discussed elsewhere.¹⁹⁻²² Conventionally, the target ion is extracted from the aqueous phase and stabilized in the hydrophobic particle core by the recognition group. Upon extraction, the luminescent output of the nanoparticle is altered. This can be a result of altering the protonation degree of a pH-sensitive dye, displacing a solvatochromic dye into the aqueous phase, or the quenching of either the fluorescence or phosphorescence of a particular transducer inside the particle.¹⁹⁻²² Polymeric nanosensors have been used to measure a wide range of analytes in both biological (in vitro and in vivo) and chemical systems and have shown advantageous character over their electrode and bulk membrane counterparts.¹⁷

One example of nanoparticle-based measurement systems is Bakker and coworkers’ exploitation of the technology to make more sensitive and selective complexometric titration reagents to measure alkali, alkaline earth metals, and the anions NO₃^– and ClO₄^–.²³⁻²⁵ The reagents had optical readouts by encapsulating solvatochromic dyes or chromoionophores, which significantly simplified the identification of the endpoint in their titrations.^24,26 The use of nanoemulsions in this regime resulted in heterogeneous complexometric reagents that are more sensitive than the homogeneous-based methods and allow for highly selective hydrophobic ligands to be utilized in titrimetry. Creating heterogeneous complexometric reagents further illuminated the potential for PNS to be used in a wide range of analytical analyses.

The extraction process of classic LLE and the sensing mechanism for PNS follow similar principles, with partitioning of the analyte between two immiscible phases and optical changes in a reporter group. Accordingly, we hypothesized that the encapsulation of optically responsive ligands into polymer nanoparticles would enable a rapid quantification of metals of interest without the drawbacks of traditional LLE. While nanoemulsions have been utilized as complexometric reagents, our nanoparticle approach has shown that ligands that are typically dissolved in bulk organic phases can be encapsulated into polymeric nanoparticles while maintaining their metal-sensing functionality. This approach enables easier analytical procedures and allows for adaptations of general LLE approaches to novel biological applications where LLE is not feasible. By utilizing the components of PNS as an extraction matrix, metals that are usually quantified through bulk liquid phase extraction methods can instead be measured with smaller volumes of harmful reagents and much higher throughput. In this manuscript, we describe the combination of optically responsive ligands traditionally used for LLE and PNS as a nanoparticle extraction method for quantifying metals in water samples (demonstrated with Fe²⁺, Al³⁺, and Cu⁺) toward a general platform for the adaptation of LLE-based methods.

Experimental Section

Materials

Polyvinyl chloride (PVC), bis ethylhexyl sebacate (DOS), bathophenanthroline (BP), 8-hydroxyquinoline (8HQ), bathocuproine (BC), tetrahydrofuran (THF), dichloromethane (DCM), sodium tetrakis-[3,5-bis(trifluoromethyl)phenyl]borate (NaBARF; Selectophore), sodium acetate, glacial acetic acid, Dulbecco’s phosphate-buffered saline (PBS), ammonium iron(II) sulfate hexahydrate, iron(III) chloride, cobalt(II) sulfate hydrate, copper(II) sulfate, nickel(II) sulfate, magnesium sulfate, and thioglycolic acid were all purchased from Sigma-Aldrich (St. Louis, MI, USA). 3,3′-Dioctadecyloxacarbocyanine perchlorate (DiO; Invitrogen) and 1,1′-dioctadecyl-3,3,3′,3′-tetramethylindodicarbocyanine, 4-chlorobenzenesulfonate salt (DiD; Invitrogen) were purchased from Thermo Fisher Scientific (Waltham, MA, USA). Microdialysis hollow fibers (13 kDa cutoff) were purchased from Spectrum Laboratories, Repligen (Waltham, MA, USA). Ferrous sulfate supplement tablets were purchased from Signature Care. Creek water was obtained from Clear Creek in Golden, CO, USA on February 21, 2021.

Fabrication

The procedure for all preparticle mixtures started by placing polyvinyl chloride (PVC) into 2 mL glass vials with Teflon-lined caps. DOS was added to the PVC and set aside. The ionic additive (NaBARF) was brought into 250 μL of THF and added to the PVC vial. This vial was vortexed until the PVC was fully dissolved. The ligand, which was dissolved in DCM, was added to the PVC vial and vortexed for 45 s. The fluorescent Fe²⁺ NS required that DiO (also dissolved in DCM) be added at the same point as the ligand. All preparticle mixtures were stored in a refrigerator at 4 °C. For the mass of each component, see Table S1.

All NS batches were made with the same protocol, following a similar procedure to Dubach et al.²⁷ Two milligrams of the PEG-lipid was put into a 4-dram scintillation vial. While the PEG-lipid is used for the solubilization of these nanoparticle sensors, other surfactants have been used in similar nanosensors (such as triblock copolymers).^28,29 Acetate buffer (pH 4.6 unless stated otherwise) was added to the scintillation vial at a volume of 5 mL and then sonicated with a probe tip sonicator (Branson Digital Sonifier 450; Branson Ultrasonics Corporation, Danbury, CT) for 30 s at 20% of its maximum power. A total of 100 μL of the preparticle mixture was added to the acetate solution and sonicated for 3 min under 20% maximum power. After sonication, the excess polymer was removed with a 0.8 μm filter (Pall Corporation, Port Washington, NY, USA), and the filtrate was stored in a 1.5-dram screw cap glass vial at room temperature in the dark.

Nanosensor Characterization

The calibration curves of the NS were obtained by loading 100 μL of analyte standard dilutions (in acetate buffer) and 100 μL of NS solution into the wells of a Nunc MicroWell 96-well optical-bottom plate (Nalgene Nunc International, Roskilde, Denmark). This yielded a final volume of 200 μL in each well and concentrations ranging from 0.5 to 150 μM Fe²⁺, 0.5 to 100 μM Cu⁺, or 1 μM to 1 mM Al³⁺ for each respective NS calibration. The selectivity of the Fe²⁺ NS was determined with the separate-solution method against Cu²⁺, Fe³⁺, Co²⁺, and Ni²⁺ dilutions. The Cu⁺ and Al³⁺ NS selectivities were obtained by two-point, separate-solution calibrations of each competing analyte at 100 μM. Each calibration was obtained in triplicate using a Synergy H1 microplate reader with absorbance readings at 533 and 478 nm or a fluorescence endpoint at 505 nm for Fe²⁺, Cu⁺, or Al³⁺, respectively.

The Fe²⁺ NS fluorescence readout was obtained by exciting each sample at 450 nm and measuring fluorescence of DiO at 501 nm. All standards were made new for each calibration to limit errors due to analyte oxidation. To obtain the Cu⁺ solution, 0.2% thioglycolic acid was added to Cu(II)SO₄ (1:10 v/v). Further standard dilutions were made from this stock.

All NS calibrations were fit to either a simple linear regression or a 4-parameter logistic equation with GraphPad Prism 9.1.2 software (San Diego, CA, USA). The statistical analyses for all calibrations were also done in GraphPad Prism 9. The limits of detection were calculated according to ref (30).

The reversibility of the Fe²⁺ NS was tested using the procedure by Ferris et al., with some variation.²² Briefly, the Fe²⁺ NS were concentrated to a 10× stock, put into a hollow fiber microdialysis tube (MWCO, 13 kDa; Spectrum Laboratories), and sealed at each end and then secured into an untreated 6-well culture plate with epoxy. After the epoxy dried, acetate buffer was put into the same well and conditioned for 3 h. Then, the initial absorbance (533 nm) and fluorescence (λ = 450/501) reading was taken on the Synergy H1 microplate reader using the area scan settings (11 × 5, 1600 × 1600 μm spacing). The acetate buffer was removed, and the well was rinsed before a dilution of 150 μM Fe²⁺ was put into the well and incubated for 30 min. After incubation, a measurement was taken with the same settings, and the Fe²⁺ dilution was removed before another rinse and addition of acetate-buffered solution. This cycle was repeated twice in triplicate.

The Fe²⁺ NS functional lifetime was measured by calibrating the fluorescence response according to the procedure on days 0, 2, 4, and 7.

Analysis of Real Samples

The reliability of the measurements was tested with the Fe²⁺ NS against a ferrous sulfate supplement tablet. One tablet (205 mg of anhydrous ferrous sulfate) was dissolved in half a liter of acetate buffer and filtered with a 0.8 μm PES membrane syringe filter. A series of two-fold dilutions were made until the theoretical concentration (calculated by the dilution factor and the 205 mg tablets) was within the NS linear range. These dilutions were mixed with NS into the wells of a 96-well plate, and an absorbance value at 533 nm was obtained with the Synergy H1 microplate reader. With the absorbance values of the supplement dilutions, the respective concentrations were back calculated with the regression line of the standard curve. Taking the calculated concentrations and dividing them by their dilution factor (stock = 1, dilution 1 = 0.5, dilution 2 = 0.25, etc.) allowed us to calculate the concentration of ferrous sulfate in the stock solution. We also attempted to measure Fe²⁺ in creek water with the standard addition and external standard method.³¹ The accuracy of the Fe²⁺ NS measurements was confirmed by a comparison to a Hanna colorimeter—Checker HC (Woonsocket, Rhode Island, USA)—that is used for selectively measuring total iron.

Results

Our Fe²⁺ nanosensors (Fe²⁺ NS) incorporate bathophenanthroline (BP, log P = 3.37³²) as the recognition agent and reporter for Fe²⁺. Instead of extracting Fe²⁺ from the aqueous to the organic liquid phase, our approach uses BP to extract Fe²⁺ into a polymer nanoparticle phase with a 3:1 BP:Fe²⁺ ratio (Scheme 1).

Scheme 1. Sensing Mechanism of the Fe²⁺ Nanosensors.

Without Fe²⁺ in the system, the sample does not absorb light at 533 nm. BP selectively extracts Fe²⁺ into the polymer phase and undergoes a large increase in absorbance at 533 nm, creating a red sample.

In the absence of Fe²⁺, BP does not absorb light from 300 to 700 nm. Increasing Fe²⁺ in the system results in a BP-mediated Fe²⁺ extraction and stabilization in the polymer phase, causing an extensive increase in absorbance with a peak at 533 nm (Figure 1). Using these data, a calibration curve was obtained using a set of standards ranging from 0 to 150 μM ferrous sulfate. The Fe²⁺ NS have a linear range from 1 to 30 μM (R² = 0.994) and a limit of detection (LOD) of 0.8 μM (Figure S2). This linear range is sufficient for immediate implementation into a variety of applications like environmental water testing^33,34 and serum-based diagnostics³⁵ (potentially requiring additional sample preparation to release transferrin-bound iron).

Figure 1.

Increasing Fe²⁺ results in a large absorbance peak at 533 nm from the nanosensors. Linear and semilog calibration curves from this data are provided in Figure S2.

The sensor mechanism and the efficiency of BP incorporation into the nanoparticles were tested by centrifugal filtration and subsequent testing on the filtrates. No optical response was seen when adding 50 μM Fe²⁺ to the nanosensor filtrate, and minimal absorbance (533 nm) was observed in the filtrate of the nanosensors that already responded to 50 μM Fe²⁺ (Figure S1).

Before the analysis of a real sample, the Fe²⁺ NS selectivity was tested against Fe³⁺ chloride and Cu²⁺, Co²⁺, and Ni²⁺ sulfates. All resulting response curves for potentially interfering metals showed negligible changes in absorbance at 533 nm (Figure 2). Phenanthrolines have also been shown to chelate other metals and absorb light at some capacity at a range of wavelengths.³⁶ Absorbance spectra for the other metals are shown in Figure S3 and confirm no significant increase in absorbance between 300 and 700 nm.

Figure 2.

Calibration curves obtained with other transition metal standards show a negligible response at 533 nm deeming the sensors highly selective for Fe²⁺ over the other analytes. n = 3, and error bars are smaller than data points.

Other BP-based assays have shown poor selectivity over Cuⁿ⁺ species,¹² as phenanthroline and its derivatives can form a 2:1 cuprous complex, which has an absorption maximum at ∼478 nm and is extractable into an organic phase.³⁷ Thus, more confirmation of the selectivity over Cu²⁺ and its impact on our Fe²⁺ calibrations was needed. According to the U.S. Agency for Toxic Substances and Disease Registry, the average amount of copper in river water is 0.16 μM.³⁸ Accordingly, we calibrated the Fe²⁺ NS to a new set of standard dilutions that were spiked with a range of relevant Cu²⁺ concentrations (0–5 μM). Figure S4 shows low Cu²⁺ interference on the Fe²⁺ response. The Fe²⁺ calibrations with spiked Cu²⁺ showed no difference compared to the standards without Cu²⁺. All calibrations held a good linear range between 1 and 30 μM.

After demonstrating Fe²⁺ NS selectivity, Fe²⁺ in a ferrous sulfate supplement tablet was measured. Dissolving the supplement tablet in 0.5 L of acetate buffer yielded a stock concentration of 2.7 mM. Measuring the two-fold dilutions that fell within the linear range of the assay resulted in a final calculated mass of 206 ± 8 mg of ferrous sulfate. A high correlation was obtained when comparing these measurements to those from a commercially available method for the determination of total iron (Figure 3). In addition to measuring Fe²⁺ in a supplement tablet, the sensors were also used to measure water collected from Clear Creek in Golden, Colorado with the standard addition method. Statistical analysis of the regressions from spiked, unknown samples and standards showed no difference between slopes, revealing negligible matrix effects in the sample (Figure S5). The method of external standards showed no measurable Fe²⁺ in the creek water, which was also confirmed with the Feⁿ⁺-selective Hanna colorimeter.

Figure 3.

Measured iron concentrations from dilutions of a supplement sample obtained with the Fe²⁺ nanosensors (y-axis) highly correlate with those obtained with a standard commercial iron selective colorimeter (x-axis). The dashed line indicates a slope of 1 between the two methods. n = 3, and error bars are smaller than the data points.

Following prior work with a fluorescence gating mechanism,²¹ we incorporated a fluorescent reporter, which is not sensitive to Fe²⁺ (DiO) into the particle core to exhibit the user-tunability of this approach to a broader set of applications. Due to the sufficient spectral overlap between the absorbance of Fe(BP₃)²⁺ and the emission of DiO, the Fe²⁺ NS show a large turn-off response to increasing Fe²⁺ (Figure 4). These calibrations resulted in a response midpoint of 2.1 μM (Figure 4C). In conjunction with the colorimetric responses shown in Figure 2, the fluorescence calibrations for the Fe²⁺ NS against the competing analytes show good selectivity (K_ij based on a 50% maximal potential signal change shows no response for off-target analytes due to minimal response) (Figure 4C). Moreover, the midpoint of the fluorescence response of these sensors is stable for at least 4 days (Figure S6). Due to the prevalence of fluorescence quenching by transition metals, including Fe²⁺,³⁹ controlling for the possibility that other targets would quench DiO was a necessity. Sensors without BP (which should not respond to Fe²⁺) had no DiO response to increasing analyte concentrations (Figure S7).

Figure 4.

(A) Introducing a lipophilic fluorescent indicator (DiO, λ_em of 501 nm) into the colorimetric Fe²⁺ NS results in a fluorescence assay for Fe²⁺ determination. (B) Exposure to Fe²⁺ results in high absorbance at 533 nm, which sufficiently gates the DiO emission, causing a decrease in measured fluorescence. (C) Sensors are selective for Fe²⁺. n = 3 for panels (B, error bars not shown) and (C).

At circumneutral pH (pH ∼6–8), Fe²⁺_(aq) is not thermodynamically stable and is rapidly oxidized to Fe³⁺ in the presence of oxygen.⁴⁰ As BP is easily protonated at acidic pH, it was necessary to determine the impact of pH on the Fe²⁺ NS response. The response at 50 μM Fe²⁺ was measured at pH 3.6, 4.6, and 5.6 (Figure S8). Indeed, the sensor response at pH 3.6 was amplified relative to the signal at pH 5.6. Prior analyses of both colorimetric and fluorescent phenanthroline-based probes show an irreversible response to Fe²⁺.⁴¹ As expected, our approach is also irreversible (Figure S9).

To test how general our nanoparticle extraction is, we used the same approach with optically responsive ligands for both Al³⁺ and Cu⁺. For sensing Al³⁺, we used the responsive probe 8-hydroxyquinoline (8HQ, log P = 1.72⁴²). 8HQ extracts and stabilizes the free Al³⁺ from the aqueous phase into the sensor core with a 3:1 stoichiometry.⁴³ Increasing Al³⁺ in the system enhances the fluorescence at 505 nm due to the formation of the Al(8HQ)₃ complex inside the particle. Using standard buffer solutions, the Al³⁺ NS show a linear range from 10 to 100 μM (Figure 5A). Al³⁺ is often leached into streams and other water bodies when the pH is below circumneutral.⁴⁴ The range of the Al³⁺ NS allows for application in the analysis of acid mine drainage or acid rain runoff where Al³⁺ can easily exceed 10 μM.⁴⁵ While 8HQ can chelate other metals, our selectivity tests (100 μM analyte) show that the Al³⁺ NS are selective (Figure 5A, inset). The spectra from the NS–analyte mixtures show no significant fluorescence peak from 300 to 800 nm (Figure S10).

Figure 5.

This approach is not exclusive to Fe²⁺ extraction and quantification. Using an optically responsive ligand (8HQ, λ_em = 505 nm) for Al³⁺ results in (A) sensitive and (A, inset) selective NS for Al³⁺ (R² = 0.9985). (B) Using BC, Cu⁺ NS show an increased absorbance at 478 nm (R² = 0.9989) and (inset) high selectivity for Cu⁺. Calibrations are in standard buffer sets with n = 3. Where not visible, the error bars are smaller than data points.

The Cu⁺ NS response is mediated by a BP derivative, bathocuproine (BC, log P = 4.47⁴⁶). Inside the particle, the Cu(BC₂)⁺ undergoes metal-to-ligand charge transfer and increases absorbance with a peak at 478 nm in response to Cu⁺ (Figure S11). The calibration of these sensors with standard buffer sets resulted in a linear range between 1 and 20 μM (Figure 5B). The Cu⁺ sensors also showed good selectivity over other metals (Figure 5B, inset).

Similar to the Fe²⁺ NS, the characterization of the extraction mechanism and confirmation of sufficient ligand partition were done through centrifugal filtration and testing of the filtrate (Figure S1). Additionally, as expected, both the Al³⁺ and Cu⁺ NS responses were impacted by pH (Figure S8).

Discussion

The quantitation of metals is a challenge that has implications in medicine, microbiology, water chemistry, and geology.¹⁻⁴ Innovations upon demanding analytical procedures that allow for quick and high-throughput analysis are a necessity. Our nanoparticle extraction assay provides a platform that enables easy quantification of metals by adopting bulk phase extraction reagents and principles. The partition principles that govern the extraction of the analyte of interest are similar to those of traditional LLE.⁴⁷ Namely, the metal ion is poorly soluble inside the hydrophobic extraction phase without the stabilization by the selective ligand; with this component, the ion is stabilized in the particle, while the other analytes remain in the aqueous phase.

Others have used similar materials (plasticized PVC) for improved extraction methods of Fe²⁺, Cu⁺, and many other analytes into a polymer membrane instead of an organic liquid.^16,48 Alternative materials have also been used for solid phase extraction (SPE); Martínez et al. investigated the partition of the phenanthroline–Fe²⁺ complex in modified acrylamide hydrogels.⁴⁹ They determined the free Fe²⁺ concentrations in milk, providing a method for spectrophotometric measurement in opaque samples. Similar to some metal-selective bulk optode membranes,^50,51 these SPE approaches have resulted in a long response time (i.e., a t₉₅ of 15–50 min^49,52) when analyzing real samples due to their dependence on both diffusion of the ligand inside the matrix as well as mass transport of the small amount of the analyte in the aqueous phase.¹⁷ For sensors with a polymeric foundation, transitioning from a planar surface to a nanoparticle dispersion has allowed for significantly lower response times due to a much larger surface area-to-volume ratio.¹⁷ In this work, the response time is too fast to measure with our equipment and approach. We know that the sensor responds faster than the time it takes between sample preparation and sample measurement (∼15 s), but the response time is likely closer to the ∼ms regime as demonstrated with similar sensors.⁵³

By adopting the structure from polymeric nanoparticle sensors (PNS), our extraction approach not only benefits from a faster response time but also adds the potential application in systems where LLE and its variations are not feasible. Nanoparticle sensors with this structure have been used to characterize complex biological systems such as clinically grown biofilms of Pseudomonas aeruginosa, eukaryotic intracellular environments, and animal models by probing for a range of diverse analytes.⁵³⁻⁵⁵ A large factor to consider when applying analytical techniques to live models is the biocompatibility of the reagents used as well as the assay’s procedure itself.⁵⁶ The reagents used in classic LLE and variations thereof are often toxic to live samples. For example, applying a large amount of an organic solvent to a live biological sample while using traditional LLE will quickly destroy the sample’s integrity. Though liquid phase microextraction (LPME) methods decrease the solvent volume, biological samples still consist of highly complex matrices that may interfere with liquid–liquid phase interfacial tension, which can be detrimental to extraction efficiency.⁵⁷ As a result, many liquid extraction methods require pretreatment of the sample, a hindrance not required with PNS.

Another benefit of the PNS methodology is the ability to substitute the sensing components inside the particles when the analyte of interest changes.¹⁷ Thus, the nanoparticles in this work can be used as an extraction phase for more than just one analyte, provided that there is a hydrophobic ligand that solubilizes inside the polymer matrix. We show in Figure 5 that this approach can be tailored to an analyte of interest depending on the ligand encased in the dispersed polymer phase while using the same structural materials with no additional steps.

Utilizing an optically responsive ligand resulted in sensors that coupled the recognition and transduction moieties inside the particle, though this is not mandatory. Other nanoparticle sensors have separated these elements by utilizing an optically inert ligand for extraction and quantification through ion exchange equilibrium with a pH-sensitive fluorescent dye^17,18 or by adding an additional optically responsive readout component into the particle.^20−22,58 Fluorescence gating mechanisms similar to our DiO approach would likely be infeasible in bulk liquid phase extractions due to the distance between the optically active groups, demonstrating an additional benefit to embracing nanoparticles for the extraction matrix.

The sensing approach that we use here is similar to the optical complexometric reagents that were introduced by Bakker and associates.²³⁻²⁶ However, in their work, the recognition and transduction elements are distinct moieties; in this work, they are combined. While mechanistically simpler than the NS with distinct recognition and transduction components, there are drawbacks here such that the sensor response is less straightforward to adjust.

The addition of a ratiometric readout to fluorescent sensors allows for the quantitation of an analyte in a highly complex sample due to the multiwavelength measurement.⁵⁹ While this readout method is preferable, especially with complex matrices, the incorporation of additional fluorophores into the nanoparticle also adds potential complexity that may be hard to predict. Our approach does not avoid this complexity. The addition of DiD as a reference signal into our Fe²⁺ NS exhibited a favorable ratiometric (DiO/DiD) response to Fe²⁺ but a large, unwanted increase in the presence of Cu²⁺ (Figure S12). Future works will focus on this off-target interaction. The ratiometric Fe²⁺ NS were still selective over metals other than Cu²⁺, which may still prove to be useful in some settings. Nevertheless, the other two readout approaches (501 nm emission and 533 nm absorption) provided high selectivity over analytes that may compete with BP’s binding moiety (Figures 2 and 4C).

Our NS are not ratiometric; thus, the sensor response is impacted by sensor concentration. Specifically, by diluting the Fe²⁺ NS sensor batch by 0.5× and 0.25× stock solutions, both the dynamic and linear range decreased in the colorimetric mode (Figure S13). As expected, in the fluorescence mode, the span of the sensors also decreased proportionally to the dilution factor with the normalized midpoint response (log EC₅₀) also decreasing (Figure S13), thus providing a slight increase in sensitivity.

This nanoscale extraction approach is beneficial for a researcher looking to rapidly quantify a metal of interest. Though this approach follows similar LLE principles, it is specified for sensing and is not all-encompassing to LLE applications. An instance in which researchers need to extract then purify their analyte of interest with chromatography for further analysis is not compatible with our approach because the metal–ligand complex stays encased in the nanoparticle. Another scenario where this method would not work is one in which the researcher needs to separate the two liquid phases after extraction to get rid of optically opaque matrices. In our method, the nanoparticle suspensions are injected directly into the sample, and the measurement is taken on the mixture. For this reason, colored or turbid matrices could significantly interfere with measuring analytes in the colorimetric mode of this approach. While there are analytical procedures to mitigate matrix effects, the fluorescence readout mode is still recommended in these cases as it may enable quantification even in some colored samples. Chemical probe synthesis is a popular area of study;⁶⁰ however, there is no probe for every application. The last setting in which this new method would not work is one in which the researcher is attempting to extract an analyte that does not have a ligand that is soluble inside a hydrophobic phase. However, water-soluble probes could be lipophilized to incorporate them into the particle core.

Conclusions

Here, we introduce a general platform for an alternative phase extraction method for quantifying metal ions. Encapsulating hydrophobic optically responsive ligands inside polymeric nanoparticles enabled successful quantitation of Fe²⁺, Al³⁺, and Cu⁺ in aqueous samples. The Fe²⁺ nanosensor response highly correlates to that of a commercial colorimeter, deeming this approach a reliable analytical option. Adopting a polymeric nanoparticle for the extraction matrix provides benefits like a one-step protocol and simple formulation procedures, reduced amounts of organic solvents, higher throughput, fast response times, and the potential for a large expansion of phase extraction applications. Future works will consist of identifying the unwanted, ratiometric response of the Fe²⁺ nanosensors to Cu²⁺, incorporating optically responsive metal probes for other analytes into the nanoparticles, and applying this nanoparticle extraction assay to samples with highly complex matrices to rapidly measure analytes of interest.

Glossary

ABBREVIATIONS

LLE

liquid–liquid extraction

NS

nanosensors

BP

bathophenanthroline

8HQ

8-hydroxyquinoline

BC

bathocuproine

PNS

polymeric nanosensors

SPE

solid phase extraction

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acssensors.1c01780.

• (Table S1) Mass components of NS formulations; (Figure S1) filtrate analysis of Fe²⁺, Al³⁺, and Cu⁺ NS; (Figure S2) linear and nonlinear calibration curves for the Fe²⁺ NS; (Figure S3) absorption spectra showing selectivity of the Fe²⁺ NS with 100 μM, competing metal solutions; (Figure S4) calibration curves for the Fe²⁺ NS with mixed-analyte (Fe²⁺:Cu²⁺) standard dilution sets; (Figure S5) comparison of Fe²⁺ standard curve calibration with known Fe²⁺ to standard addition calibration with unknown Fe²⁺; (Figure S6) fluorescence calibrations that show the functional lifetime of the Fe²⁺ NS; (Figure S7) calibration of Fe²⁺ NS that lack bathophenanthroline against all analytes; (Figure S8) calibrations of the Fe²⁺, Al³⁺, and Cu⁺ NS against a fixed analyte and varying pHs; (Figure S9) irreversibility of the Fe²⁺ NS; (Figure S10) fluorescence spectra showing selectivity of the Al³⁺ NS with 100 μM analyte solutions; (Figure S11) absorption spectra of the Cu⁺ NS with increasing concentrations of Cu⁺; (Figure S12) calibrations of the ratiometric (DiO/DiD) Fe²⁺ NS against competing metals; (Figure S13) calibrations of the diluted Fe²⁺ NS (PDF)

Author Contributions

K.J.C. and T.Z.S. conceptualized the study and performed the methodologies; T.Z.S. performed the validation and formal analysis; T.Z.S., A.A.D., A.P.C., J.N.P., G.E.G., C.M.N., and S.M.I. performed the investigation; T.Z.S. performed data curation, wrote the original draft, and performed visualization; K.J.C. and T.Z.S. reviewed and edited the manuscript, supervised the study, and performed project administration; K.J.C. performed funding acquisition.

The authors declare no competing financial interest.