Identification of 15 Phthalate Esters in Commercial Cheese Powder via Cyclodextrin-Promoted Fluorescence Detection.

A challenge for detecting phthalates in commercial products such as cheese powders is that the composition of the products is highly complex, and current methods for detection rely on gas chromatography-mass spectrometry, which is not portable and cannot be used by individual consumers at a time and place of their choosing. Herein, we report the development of a new method for phthalate detection in cheese powder using cyclodextrin-promoted fluorescence detection, in which the presence of the phthalate analytes leads to highly analyte-specific changes in the fluorescence emission signal of a fluorophore bound in a cyclodextrin cavity. This method relies on subtle changes in the analyte affinity for the fluorophore and the cyclodextrin cavity and provides for markedly more straightforward sample preparation procedures and an extremely rapid read-out signal, with potential for the development of portable fluorescence sensors. Using this method, we were able to detect 15 phthalate esters with highly analyte-specific responses and at concentrations as low as 0.12 μM, which is well below regulatory levels of concern. Computational investigations strongly support the observed experimental trends.

Introduction

Toxic chemicals have been found in a wide variety of commercial products,[1] including processed foods[2] and beverages.[3] Examples of such chemicals include phthalates,[4] phenol and phenol derivatives,[5] and a variety of other organic and inorganic analytes.[6] Exposure to these chemicals is concerning because of their known and suspected toxicity,[7] which can cause deleterious health effects to humans[8] and other species exposed to the chemicals.[9] Moreover, the long-term environmental persistence of such chemicals means that potential exposure can occur long after the usage of the chemicals and the initial release into the environment.[10] Current methods to detect these compounds include gas chromatography–mass spectrometry (GC–MS) and liquid chromatography–mass spectrometry,[11] although newer methods such as electrochemically based methods[12] and spectroscopic (including Raman-based) methods[13] have also been reported. Although these known methods operate with high sensitivity, concerns about the ability of the average consumer to access such methods persist.

Our group has recently developed a fundamentally new method for toxicant detection in complex environments, using cyclodextrin-promoted energy transfer from the toxicant of interest to a high-quantum-yield fluorophore, for photophysically active analytes (Figure A),[14] and cyclodextrin-promoted fluorescence modulation, for nonphotophysically active analytes (Figure B).[15] These proximity-induced, noncovalent interactions among the cyclodextrin host, toxicant analyte, and fluorophore reporter result in highly analyte-specific fluorescence read-out signals, which have been used for toxicant detection in a broad variety of complex environments. Such environments include human plasma,[16] breast milk,[17] urine,[18] extracts collected from oil[19] and fuel spills,[20] and contaminated marine environments.[21] Compared to most currently utilized methods, fluorescence-based toxicant detection has the potential to lead to more rapid read-out signals, highly portable devices, and improved detection sensitivity and selectivity.[22]

Figure 1

Schematic illustration of (A) cyclodextrin-promoted analyte-to-fluorophore fluorescence energy transfer and (B) cyclodextrin-promoted analyte-specific fluorescence modulation.

The detection of toxicants in complex food matrices such as macaroni and cheese products using cyclodextrin-promoted fluorescence detection has not been reported to date, although such detection is particularly relevant in light of recent news reports that phthalates have been found in commercial cheese powder.[23] Although these initial reports were of chemical contamination in only a single cheese powder brand, there are a broad variety of commercial macaroni and cheese products available, and the composition of the cheese powder in these products is somewhat variable. Moreover, the concentration of available phthalates in the final cooked product will depend on the cooking method, the chemical composition of the cookware, and other highly sample-specific parameters, which means that developing detection methods that can be operated by the end user is of significant interest. Current methods for toxicant detection in macaroni and cheese products are time consuming and require specialized laboratory equipment, which precludes individual consumers from analyzing their own macaroni and cheese products at a time and place of their choosing, i.e., after the food has been prepared.[24]

Reported herein is a detailed GC–MS analysis of the samples that confirms the existence of phthalate esters in all of the powders investigated. In parallel, we report the development of a new fluorescence-based detection method for phthalate esters in commercial cheese powder, which operates with high sensitivity (as low as 0.12 μM limits of detection) and general applicability (for operating in complex environments and for correctly identifying individual analytes and analyte mixtures). In support of the experimental results, computational investigations provide additional insight into the mechanism that underlies the highly analyte-specific cyclodextrin-promoted detection. Together, these two detection methods and computational supporting investigations provide a detailed picture of the samples in question and an important proof-of-concept of a new sensor for phthalate detection in complex food matrices.

Results and Discussion

Four brands of macaroni and cheese, including both organic and nonorganic varieties, were bought from a highly popular United States-based retail store, with the goal of sampling products that are reaching large numbers of consumers and have a significant potential health effect if found to be contaminated. Phthalate esters chosen for analysis (compounds 1–15, Figure ) include those that have been found in macaroni and cheese powder[25] as well as a variety of other analogues, with the goal of determining the generality of the fluorescence-based method and GC–MS-based analysis across a broad swath of structures with a relatively wide range of potential toxicities.[26] For the fluorescence-based detection method, we selected γ-cyclodextrin as the supramolecular host due to the known ability of cyclodextrin to bind hydrophobic small molecules in its interior[27] and the specific ability of γ-cyclodextrin to facilitate the binding of two small-molecule guests to form a ternary complex.[28] A commercially available boron-dipyrromethene (BODIPY) fluorophore was selected as the source of the read-out signal, based on the well-documented ability of BODIPY fluorophores to display high quantum yield and robust performance under a variety of experimental conditions.[29]

Figure 2

Structures of all phthalate ester analytes (1–15), control analyte (16), and high quantum yield fluorophore (17) used in these investigations.

GC–MS analysis of the four selected cheese powders indicated that all powders contained multiple phthalate esters (Table ) as well as significant amounts of fatty acids, including hexanoic acid and octanoic acid. Total concentrations of the phthalate esters in the cheese powder extract were relatively modest, ranging from 0.46% (for organic name-brand cheese powder) to 9.05% (for regular name-brand cheese powder). The identity of all peaks was confirmed using the NIST14 library as well as by matching the retention times and mass spectra to known standards.[30] Of note, many of these analytes have the potential to affect the fluorescence of fluorophore 17 when added to the complex matrix, through the formation of micellar structures[31] and/or through general environmental changes.[32] Such effects partially underlie the ability of the fluorescence-based detection to distinguish between different cheese powders investigated herein.

Table 1

Phthalates Detected in Cheese Powder Samples via GC–MS Analysisa

	regular cheese powder		organic cheese powder
analyte	name-brand	store-brand	name-brand	store-brand
1				X
2		X
3
4			X	X
5	X
6	X		X	X
7		X	X
8	X
9
10		X
11
12
13
14
15

The presence of an “X” indicates that the analyte has been found in the particular cheese sample investigated.

Because phthalates were detected in all of the cheese powder samples investigated, we then decided to dope higher amounts of a broader variety of phthalates into the cheese powders, with the goal of understanding how the presence of the phthalates affects the emission signal of fluorophore 17 and how such analyte-induced changes can be used for phthalate detection. Of note, the actual amount of phthalate esters in each sample investigated is comprised of the amount doped into the sample as well as the amount that is already found therein. Analyte-induced changes in the fluorescence emission, which were quantified using eq , are summarized in Table and indicate that in every case, the introduction of the analyte resulted in increased fluorescence emission of fluorophore 17. This phenomenon can be attributed to the large cavity size of γ-cyclodextrin, which facilitates association between the analyte and the fluorophore, further restricting the free rotation of the fluorophore and increasing the observed fluorescence emission. Moreover, the continued interaction between the analyte and the fluorophore afforded by the ternary complex facilitates low limits of detection for all phthalate ester analytes (vide infra). Ample literature precedent supports the notion that increased steric hindrance and/or rigidification around fluorophores can lead to enhanced fluorescence emission due to decreases in nonradiative decay pathways available.[33,34] Notable outliers include analytes 10 and 11, with markedly higher modulation values compared to the other analytes, and analytes 12–14, which displayed relatively low modulation values. These analytes were subjected to detailed computational analysis (vide infra) to explain these anomalous trends.

Table 2

Selected Fluorescence Modulation Ratios Obtained with Fluorophore 17 and Analytes 1–16a

	regular cheese powder		organic cheese powder
analyte	name-brand	store-brand	name-brand	store-brand
1	1.53	1.25	1.32	1.42
2	1.69	1.30	1.27	1.37
4	1.56	1.70	1.48	2.32
5	1.45	3.27	1.97	1.99
6	1.32	1.32	1.87	1.34
8	1.91	1.24	2.37	1.73
10	4.02	4.80	4.59	4.00
11	3.61	2.56	2.34	4.63
12	1.35	1.18	1.28	1.38
13	1.40	1.24	1.32	1.27
14	1.06	1.20	1.25	1.40

Fluorescence modulation values were calculated according to eq , and each value represents the average of at least six trials. All modulation experiments were subjected to 1 h equilibration time prior to analysis, in accordance with the results of equilibration studies (see Supporting Information for more details).

Analyte-specific changes in the fluorescence emission of fluorophore 17 in the presence of cyclodextrin in the cheese powder matrices were further analyzed to determine limits of detection (LODs) for each analyte. These LODs ranged from 11.89 to 0.12 μM, depending on the identity of the analyte, cheese powder, and fluorophore selected (Table ). For comparison, the Consumer Product Safety Committee reports that the maximum acceptable intake of phthalate esters ranges from 152 mg/kg-day for diisononyl phthalate to as low as 0.5 mg/kg-day in the case of dibutyl phthalate. Using approximations of a one-cup serving size of macaroni and cheese and an approximate adult human weight of 63 kg,[35] our highest LOD (i.e., least sensitive detection capability) is still an order of magnitude lower than concentrations that can be safely consumed. For comparison, detection limits of phthalates obtained via gas chromatography–mass spectrometry methods were reported to be as low as 0.1 μM with excellent separation observed between analyte signals,[36] although such methods required additional preparation procedures prior to analysis. Electrochemical methods reported even lower limits of detection (4.5 nM) for a single phthalate ester (dibutyl phthalate),[37] although without the broad substrate scope reported herein. Moreover, the concentration of phthalates found in commercial macaroni and cheese products ranged from approximately 1 to 160 μg/kg,[38] which encompasses the detection limit range reported herein.

Table 3

Selected Limits of Detection (Reported in μM) for Analytes 1–15 Using Fluorophore 17 and Cyclodextrin-Promoted Fluorescence Modulationa

	organic cheese powder		regular cheese powder
analyte	name-brand	store-brand	store-brand	name-brand
1	5.72	11.5	6.25	3.88
2	4.16	3.38	3.88	4.04
4	0.25	1.58	0.99	2.97
5	0.99	1.49	0.31	1.27
6	1.02	1.71	4.09	4.39
8	1.17	0.49	4.16	0.85
10	0.24	0.25	0.24	1.49
11	0.12	0.46	0.53	0.80
12	2.38	1.93	10.3	11.9
13	2.18	3.13	7.63	5.32
14	4.50	2.95	7.73	8.12
15	5.75	2.43	3.28	8.03

All limits of detection were calculated using eq , and results reported represent an average of at least six trials.

Of note, the strong fluorescence modulation values for analytes 10 and 11 also led to lower LOD values, with compound 11 having the lowest limit of detection observed (0.12 μM). In contrast, the analytes with higher LOD values tended to be the smaller structures, with dimethyl phthalate and monomethyl phthalate having much higher average LOD across all samples. Moreover, the monosubstituted phthalates (compounds 12–15) were shown to have larger LODs compared to their disubstituted analogues. This trend is likely due to weaker binding of the monosubstituted analogues in the cyclodextrin cavity, due to increased hydrophilicity and decreased steric complementarity, which, in turn, weakens this hydrophobically promoted complexation. The relatively polar nature of analytes 12–15, as visualized through their electrostatic potential surface maps, is shown in Figure , where the red and blue colors indicate areas of high polarity. These results are in line with previously reported work from our group, which highlighted hydrophobic association as a particularly relevant mechanism for the inclusion and detection of highly hydrophobic analytes such as the analytes and fluorophores reported herein.[39]

Figure 3

Electrostatic potential mapping of analytes 12–15 highlighting polar areas in blue and red. Analyte 2 is shown as a less polar comparison.

Further computational investigations of analytes 10 and 11 were conducted using Molecular Operating Environment (MOE), in which the complex of cyclodextrin and fluorophore 17 was minimized in an aqueous environment (Figure A). The introduction of analytes 10 and 11 led to close association between the analytes and the fluorophore (Figure B,C), which ensures the high modulation values and low limits of detection that are observed experimentally. In comparison, the introduction of analyte 8 to the BODIPY–cyclodextrin complex led to the displacement of the fluorophore from the cavity (Figure D); the resulting weaker interaction between the analyte and the fluorophore leads, in turn, to a decreased responsiveness to the presence of the analyte.

Figure 4

Computed energy-minimized structures using the Molecular Operating Environment (MOE) software, with energies minimized in a fully aqueous environment: (A) Fluorophore 17 in γ-cyclodextrin; (B) Fluorophore 17, analyte 10, and γ-cyclodextrin; (C) Fluorophore 17, analyte 11, and γ-cyclodextrin; (D) Fluorophore 17, analyte 8, and γ-cyclodextrin.

Such differences in the behavior of analytes 10 and 11 were further seen in the analyte-induced fluorescence modulation (Figure ), wherein the introduction of analytes 10 and 11 to a solution of fluorophore 17 in γ-cyclodextrin resulted in a substantial increase in the emission intensity (reflected in high fluorescence modulation values, Table ) as well as a red-shift in the position of the fluorescence emission maxima for these two analytes. For comparison, analyte 8 induced a lower fluorescence intensity increase as well as no noticeable change in the position of the emission maximum of fluorophore 17, highlighting the unique properties of analytes 10 and 11 in cyclodextrin-based complexes.

Figure 5

Fluorescence modulation of fluorophore 17 in γ-cyclodextrin (black line) in the presence of analyte 8 (red line), analyte 10 (blue line), and analyte 11 (green line).

Conclusions

In conclusion, reported herein is the detection of phthalate ester analytes in four commercial cheese powder samples using initial characterization by GC–MS and a newly developed method of cyclodextrin-promoted, fluorescence-based detection. Compared to our previously reported fluorescence detection methods that operated predominantly in aqueous solutions, the work reported herein indicates that the cyclodextrin-based detection strategy is sufficiently robust to operate in highly complex environments. Compared to GC–MS-based methods, this fluorescence modulation strategy allows for a broad substrate scope and rapid result generation, while maintaining high sensitivity and simple sample preparation procedures. GC–MS analysis, by contrast, required significant time for method development and individual sample analysis. In parallel with these experimental methods, computational experiments provided unique insight about the nature of the cyclodextrin-promoted interaction, especially for the analytes that provided the greatest modulation values and lowest limits of detection. Although the fluorescence-based method reported herein is not yet quantitative, efforts toward achieving that goal are currently underway in our laboratory; even in the absence of quantitative detection, such a method provides notable operational advantages. Current efforts in our laboratory are dedicated toward decreasing detection limits further and exploring additional cyclodextrin–fluorophore combinations, and results of these and other investigations will be reported in due course.

Experimental Section

Materials and Methods

Fluorescence measurements were recorded on a Shimadzu RF-5301PC spectrophotofluorimeter with 1.5 nm excitation and 1.5 nm emission slit widths. All analytes and fluorophores (compounds 1–17, Figure ) were purchased from Sigma-Aldrich Chemical Company and used as received. All cyclodextrins were purchased from Tokyo Chemical Industry and used as received. Computational experiments were performed using the Spartan 18 software for electrostatic potential maps and Molecular Operating Environment for system modeling. All GC–MS measurements were performed using a Shimadzu GC–MS QP-2020 gas chromatograph–mass spectrometer.

Extraction Procedure for GC–MS Analysis

Each cheese powder sample (5 g) was placed in an 8 mL vial with 3 mL of deionized water and vortexed for 5 min. Sodium chloride (3 g) was added to the solution along with 3 mL of analytical-grade acetonitrile. In a glass vial, 5 g of sample was dissolved in 3 mL of deionized water followed by vortexing for 3 min and sonicating for 5 min. Acetonitrile (3 mL) was then added to the solution along with 5 g of sodium chloride. Samples were exposed to 5 min of further sonication and vortexing, followed by refrigeration for 10 min. Centrifugation at 5500 rpm yielded a two-layer system with the top layer collected into a round-bottom flask. The process was repeated two times, and subsequent extractions were collected into the same round-bottom flask. The combined extracts were evaporated to dryness using a rotary evaporator and brought to a final volume of 1 mL, which was transferred into an analysis vial. A standard sample doped with 15 nM of analytes 1–15 in tetrahydrofuran (THF) was also put through the same extraction procedure as a positive control for retention time, spectra, and procedural viability.

Instrumental Procedure for GC–MS Analysis

The oven temperature was set at 150 °C for 1.00 min followed by a 5.00 °C/min ramp to 270 °C with a 5.00 min hold. The injection temperature was set to 100 °C, with a splitless injection, with a 150 °C ion source temperature and 230 °C interface. A 4.00 min solvent cut time was used with the m/z ratio range set to 35.00 m/z (low mass cutoff) to 500 m/z (high mass cutoff).

Procedure for Equilibration Experiments

In a quartz cuvette, 1.25 mL of a 10 mM cyclodextrin solution dissolved in phosphate-buffered saline (PBS, buffered to pH 7.4) and 1.25 mL of a cheese powder sample (5 g/L in water) were combined with 100 μL of a 0.1 mg/mL solution of fluorophore 17 and set on a shaker table for single minute increments prior to sampling. Spectra were obtained by fluorescence analysis at a 460 nm excitation wavelength. The sampling continued until the change in the fluorescence signal was less than 5% over 5 min, as calculated by eq where Fl minutes represents the integrated fluorescence emission 5 min after the initial time point and Flinitial represents the integrated fluorescence emission at the initial time point.

General Procedure For Fluroscence Modulation Experiments

In a quartz cuvette, 1.25 mL of a 10 mM cyclodextrin solution dissolved in PBS and 1.25 mL of a cheese powder sample (5 g/L in water) were combined and mixed thoroughly by shaking on a shaker station for 1 h. Next, 100 μL of a 0.1 mg/mL of fluorophore 17 solution in tetrahydrofuran (THF) was added, and the solution was excited four times using a 460 nm excitation wavelength. Then, 50 μL of analytes 1–16 (1.0 mg/mL in THF) was added to the mixture, and the solution was again excited four times at 460 nm. The fluorescence emission spectra were integrated vs wavenumber on the X-axis using the OriginPro software, and the degree of fluorescence modulation was determined using eq where Flanalyte is the integrated emission of the fluorophore in the presence of the analyte and Flblank is the integrated emission of the fluorophore in the absence of the analyte. Fluorescence modulation ratios greater than 1 indicate an enhancement of fluorescence emission of the fluorophore in the presence of the analyte, fluorescence modulation ratios less than 1 indicate a decrease in fluorescence emission of the fluorophore in the presence of the analyte, and fluorescence modulation ratios close to 1 indicate minimal change in the fluorescence emission of the fluorophore in the presence of the analyte.

General Procedure for Limit of Detection Experiments

Limit of detection experiments were conducted following literature-reported procedures.[40] In brief, 1.25 mL of γ-cyclodextrin and 1.25 mL of 5 g/L cheese powder in water were added to a quartz cuvette and mixed thoroughly. Next, 100 μL of fluorophore 15 was added to the cuvette and the solution was excited six times at 460 nm. Next, 10 μL of the analyte was added, and again the solution was excited six times at 460 nm. This step was repeated for 20 μL of the analyte, 30 μL of the analyte, 40 μL of the analyte, 50 μL of the analyte, 60 μL of the analyte, 70 μL of the analyte, 80 μL of the analyte, 90 μL of the analyte, and 100 μL of the analyte.

All fluorescence emission spectra were integrated vs wavenumber on the X-axis, and calibration curves were generated. The curves plotted the analyte concentration, measured in μM, on the X-axis and the fluorescence modulation ratio on the Y-axis. The curve was fitted to a straight line, and the equation of the line was determined. The limit of detection was calculated according to eq where SDblank is the standard deviation of the blank sample and m is the slope of the calibration curve. In all cases, the limit of detection was calculated as a concentration in micromoles.

General Procedure For Computational Experiments

Molecular models of the analytes, fluorophores, and cyclodextrin host were generated using the build function in the Molecular Operating Environment (MOE) software. Analytes were defined as ligands, and the γ-cyclodextrin host was defined as the receptor. The ligand–receptor system underwent a “quick preparation”, which is an initial energy minimization function, followed by a more in-depth dynamics calculation using a Berendsen velocity/position scaling algorithm and Amber 14: EHT forcefield. The calculations done in the solvent were prepared using a 90 angstrom cubic cell populated with roughly 2500 water molecules to display the most likely conformation in aqueous environments. Computational models of the molecules were built in Spartan 18 and underwent a ground-state geometry calculation using semiempirical PM3 methods in the gas phase followed by an electrostatic potential map, ESP, surface calculation. The values for the ESP maximum and minimum energies were normalized across all samples to better compare polarities and electrostatic characters.