Journal of Chromatography A, 1359 (2014) 140–146

Contents lists available at ScienceDirect

Journal of Chromatography A journal homepage: www.elsevier.com/locate/chroma

Determination of prostaglandin analogs in cosmetic products by high performance liquid chromatography with tandem mass spectrometry James B. Wittenberg ∗ , Wanlong Zhou, Perry G. Wang, Alexander J. Krynitsky U.S. Food and Drug Administration, Center for Food Safety and Applied Nutrition, 5100 Paint Branch Parkway, HFS-717, College Park, MD 20740-3835, USA

a r t i c l e

i n f o

Article history: Received 24 March 2014 Received in revised form 1 July 2014 Accepted 13 July 2014 Available online 19 July 2014 Keywords: Prostaglandins Cosmetics QuEChERS LC-MS/MS

a b s t r a c t A method was developed and validated for the determination of 16 prostaglandin analogs in cosmetic products. The QuEChERS (Quick, Easy, Cheap, Efficient, Rugged, Safe) liquid–liquid extraction method, typically used for pesticide residue analysis, was utilized as the sample preparation technique. The prostaglandin analogs were chromatographically separated and quantified using high performance liquid chromatography with tandem mass spectrometry (HPLC-MS/MS). Thirty-one cosmetic products were surveyed, and 13 products were determined to contain a prostaglandin analog with amounts ranging from 27.4 to 297 ␮g/g. The calculated concentrations for the cosmetic products were in a similar range when compared to the concentrations of three different prostaglandin analog-containing prescription products. © 2014 Elsevier B.V. All rights reserved.

1. Introduction Eyelash and eyebrow enhancing cosmetic products have been increasing in demand and popularity in recent years. Some of the chemicals responsible for enhancements such as lengthening and darkening eyelashes and eyebrows are a class of compounds called prostaglandins [1]. Prostaglandins are used in many biological processes and have a wide variety of effects. A few prostaglandin analogs (bimatoprost, latanoprost, and travoprost) are used to control intraocular pressure, or treat glaucoma, through the administration of prescription eye drops [2–9]. Side effects of this treatment include conjunctival hyperemia, excessive tearing, inflammation, increased coloring of the iris, periocular skin pigmentation, as well as an increase in eyelash thickness and length [10–12]. These side effects that increase and enhance eyelashes gave reason to implement prostaglandin analogs for use in cosmetic products. The prostaglandin analogs studied, including the three used as drugs for glaucoma treatment, are structurally similar (Fig. 1). They all are comprised of a cyclopentanediol backbone with two arms: one arm containing an aromatic moiety, and the other containing a carbonyl functional group. Fortunately, they are dissimilar enough to be chromatographically separated and quantified using high performance liquid chromatography with

∗ Corresponding author. Tel.: +1 2404022206. E-mail address: [email protected] (J.B. Wittenberg). http://dx.doi.org/10.1016/j.chroma.2014.07.032 0021-9673/© 2014 Elsevier B.V. All rights reserved.

tandem mass spectrometry (HPLC-MS/MS). A few of the noted prostaglandins, along with many others, have previously been analyzed in various matrices. Many techniques, including high performance liquid chromatography paired with ultra-violet absorption (HPLC-UV) [13–16] or fluorescence detection [17,18], gas chromatography–mass spectrometry (GC/MS) [19], high performance liquid chromatography-electrospray ionization–tandem mass spectrometry (HPLC-ESI-MS/MS) [20–29], and high performance liquid chromatography–atmospheric pressure chemical ionization–tandem mass spectrometry (HPLC-ACPI–MS/MS) [30,31], have been used to determine prostaglandins contained in a wide variety of matrices ranging from in vitro enzyme incubations to pharmaceuticals and natural products. In general, cosmetic product matrices are complex and can be completely different in composition for two products used for the same purpose. Some of these matrices include serums, oils, lotions, creams, mascaras, powders, lipsticks, etc. Sample preparation, therefore, can be challenging. Traditionally, cosmetic samples are prepared using conventional liquid–liquid extraction (LLE) and solid–liquid extraction (SLE) techniques [32]. While these conventional sample preparation techniques are useful when dealing with a variety of matrices, there are disadvantages such as use of solvent (large quantities and potentially carcinogenic) and time consuming multi-step processes. For this project, the QuEChERS (Quick, Easy, Cheap, Efficient, Rugged, and Safe) extraction method was implemented for sample preparation. The QuEChERS method was originally developed and utilized for pesticide analysis since these samples come from a large variety of matrices [33–36].

J.B. Wittenberg et al. / J. Chromatogr. A 1359 (2014) 140–146

141

Fig. 1. Structures of the prostaglandin analogs studied.

QuEChERS was chosen for this study because it was envisioned to be a quick, easy, and cheap way to extract the relatively non-polar prostaglandin analogs from the aqueous serums and mascaras. To date, no method has been reported for the detection and quantitation of prostaglandin analogs in cosmetics. In this study an LC/MS/MS method was developed to determine 16 prostaglandin analogs contained in a survey of 31 eyelash and eyebrow enhancing cosmetic products using four stable isotopically labeled internal standards. The QuEChERS extraction method was adopted and utilized for sample preparation of the cosmetic samples without any need for further sample cleanup. To our knowledge, this would be the first account of the QuEChERS method being used for cosmetic sample preparation. This method may be used for the detection and quantitation of the noted prostaglandin analogs in a variety of matrices.

2. Material and methods 2.1. Chemicals LC-MS grade water (H2 O), methanol (MeOH), and acetonitrile (MeCN) were purchased from Fisher Scientific (Pittsburgh, PA). All prostaglandin analogs (bimatoprost (bima), bimatoprost isopropyl ester (bima IE), bimatoprost serinol amide (bima SA), bimatoprost free acid (bima FA), latanoprost (lat), latanoprost free acid (lat FA), tafluprost (taf), tafluprost ethyl amide, (taf EA) tafluprost ethyl ester (taf EE), travoprost (trav), (+)-cloprostenol (clo), (+)-cloprostenol isopropyl ester (clo IE), 17-phenyl trinor prostaglandin E2

serinol amide (17-PTPE2 SA), 17-phenyl trinor prostaglandin F2␣ methyl amide (17-PTPF2␣ MA), 17-trifluoromethylphenyl trinor prostaglandin F2␣ ethyl amide (17-CF3 PTPF2␣ MA), 16-phenoxy prostaglandin F2␣ ethyl amide (16-PPF2␣ EA)) and internal standards (bimatoprost-d4 (bima-d4 ), bimatoprost free acid-d4 (bima FA-d4 ), latanoprost-d4 (lat-d4 ), and latanoprost free acid-d4 (lat FAd4 )) were purchased from Cayman Chemical (Ann Arbor, MI). All chemicals were sold as ≥95% pure and were used without further purification. Thirty personal care products were purchased via the internet. 2.2. Instrumentation The liquid chromatography was carried out using an Acquity UPLC® (Waters, Milford, MA) consisting of a binary solvent manager and a sample manager. The sample manager was set to 4 ◦ C during operation. The separations were carried out using a 2.6 ␮m Kinetex XB-C18 column (100 mm × 2.1 mm i.d., Phenomenex, Torrance, CA) coupled to a 0.5 ␮m KrudKatcher ultra HPLC in-line filter (0.004 in i.d., Phenomenex, Torrance, CA). Elution was completed using a 20-min gradient program operating with a 0.50 mL/min flow rate with a 10 ␮L injection volume. Mobile phase A was composed of 0.1% formic acid in 95:5 H2 O:MeOH (v/v) and mobile phase B was composed of 0.1% formic acid in 5:95 H2 O:MeOH (v/v). The gradient parameters were: 0–1 min 20% B, 1–15 min 20% B to 100% B, 15–18 min 100% B, 18–19 min 100% to 20% B, 19–20 min 20% B. The LC elute was introduced to the ion source only between 7 and 12.5 min during the run using a Valco valve switch to prevent contamination.

142

J.B. Wittenberg et al. / J. Chromatogr. A 1359 (2014) 140–146

Table 1 Scheduled MRM parameters for quantitation and confirmation transitions for prostaglandin analogs studied. Analyte

Average mass

Precursor ion

Product ion

17-PTPE2 SA

RT (min) 7.7

459.3

Bima SA

8.2

461.3

16-PPF2␣ EA

8.4

417.3

17-PTPF2␣ MA

8.7

401.3

Bima

9.2

415.3

Bima-d4

9.2

419.3

Bima FA

9.4

388.2

Bima FA-d4

9.4

392.2

Taf EA

9.7

437.2

17-CF3 PTPF2␣ EA

10.3

483.3

Lat FA

10.7

390.2

Lat FA-d4

10.7

394.3

Clo FA-Na+

10.8

446.2

Taf EE

11.4

438.2

Bima IE

11.5

430.3

Lat

11.7

432.3

Lat-d4

11.7

436.3

Clo IE

11.8

466.2

Taf

11.9

452.2

Trav

11.9

500.2

442 442 444 444 418 418 384 384 398 398 402 402 387 387 391 391 438 438 430 430 389 389 393 393 423 423 439 439 413 413 433 433 437 437 489 467 453 453 501 501

406 92 408 317 234 270 348 317 362 317 366 321 343 193 347 197 306 232 343 385 345 147 349 147 295 219 335 289 377 131 337 379 341 383 447 321 335 261 321 249

CE (V) 23 27 19 23 15 19 15 21 19 21 19 21 −28 −34 −28 −34 15 21 31 25 −36 −44 −36 −44 −24 −24 17 23 17 41 23 15 21 19 35 17 19 25 11 19

DP (V) 81 101 56 61 96 96 76 76 76 56 76 56 −90 −45 −90 −45 86 56 161 181 −140 −60 −140 −60 −60 −85 81 81 66 76 91 91 81 86 141 91 81 81 96 106

CXP (V) 10 10 10 12 18 16 12 20 10 22 10 22 −9 −5 −9 −5 14 10 12 20 −15 −7 −15 −7 −41 −15 16 10 22 22 26 18 14 10 18 16 22 16 46 6

Note: RT, retention time; CE, collision energy; DP, declustering potential; CXP, cell exit potential. The first product ion for each analyte was used for quantitation, and the second product ion was used for confirmation.

The LC was interfaced to an AB Sciex QTrap 5500 (AB Sciex, Foster City, CA) equipped with a Turbo Ion Spray® source (electrospray). LC-MS/MS operation and data acquisition were controlled by the AB Sciex Analyst software version 1.6. Quantitation was completed using the AB Sciex MultiQuant software version 2.1.1. The ion spray voltage was set to 5000 V. The turbo heater was maintained at 400 ◦ C. Nitrogen was used as the curtain gas and collision gas. The curtain gas, ion source gas 1, and ion source gas 2 were set to pressures of 30, 40, and 50 psi, respectively. The entrance potential was set to 10 V for all transitions. The collision gas was set to “Medium”. The mass spectrometer was operated in scheduled selected reaction monitoring (SRM) mode. All the parameters listed above were used for both positive and negative mode acquisitions. Other parameters (declustering potential, collision energy, and cell exit potential) were optimized for each compound and are listed in Table 1. 2.3. Preparation of standard solutions A stock solution containing 5 ␮g/mL of bima, taf EA, and 17-PTPF2␣ MA; 10 ␮g/mL of bima IE, bima SA, clo IE, lat, trav, and 16-PPF2␣ EA; and 20 ␮g/mL of taf, taf EE, 17-PTPE2 SA, 17CF3 PTPF2␣ EA, bima FA, lat FA, and clo FA-Na+ in 50:50 H2 O:MeOH (v/v) was prepared in a volumetric flask and stored at −20 ◦ C. A stock solution of internal standard mixture containing 2 ␮g/mL of bima-d4 , lat-d4 , bima FA-d4 , and lat FA-d4 in 50:50 H2 O:MeOH

(v/v) was prepared separately in a volumetric flask and stored at −20 ◦ C. Eight calibration solutions were prepared using the stock solution. The concentration ranges for the standard solutions were: 0.25–50 ng/mL for bima, taf EA, and 17-PTPF2␣ MA; 0.5–100 ng/mL for bima IE, bima SA, clo IE, lat, trav, and 16-PPF2␣ EA; and 1–200 ng/mL for taf, taf EE, 17-PTPE2 SA, 17-CF3 PTPF2␣ EA, bima FA, lat FA, and clo FA-Na+ . A constant concentration of 10 ng/mL of the internal standard mixture was added to each standard solution. 2.4. Sample preparation The majority of the cosmetic samples (including all samples containing prostaglandins) were water-based solutions and sample preparation was relatively straightforward. Stock solutions of each cosmetic sample were prepared by weighing a known amount sample (approx. 50–100 mg) and dissolving it in 50:50 H2 O:MeCN (v/v) in a volumetric flask. Any cosmetic sample that did not completely dissolve (e.g., mascaras) was sonicated for 10 min before sample preparation. The Bond Elut QuEChERS extraction kit (Agilent Technologies, Santa Clara, CA) was utilized for this project and samples were prepared in triplicate. A 100 ␮L aliquot of a stock solution of each sample was placed into a 50 mL centrifuge tube with 10 mL H2 O, 10 mL MeCN, and 100 ␮L of the internal standard stock solution and shaken for 1 min. A salt mixture (4 g NaCl, 1 g MgSO4 ) was added to the tube to induce phase separation and then mixed on

J.B. Wittenberg et al. / J. Chromatogr. A 1359 (2014) 140–146

a Vortex Maxi Mixi mixer (Thermo Scientific, Waltham, MA) for 3 min. The sample was then centrifuged at 6000 rpm for 5 min. A 500 ␮L aliquot from the organic phase was diluted with 500 ␮L of H2 O, vortexed, and filtered through a 0.2 ␮m PTFE Acrodisc CR 13 filter (Pall Life Sciences, Port Washington, NY) directly into an autosampler vial for LC-MS/MS injection. 3. Results and discussion 3.1. Mass spectrometry optimization All parameters were optimized by infusing a 1 ␮g/mL solution of each prostaglandin analog at a flow rate of 10 ␮L/min. Positive ionization mode was used for all but the four free acids (bima FA, bima FA-d4 , lat FA, lat FA-d4 , and clo FA-Na+ ), which were optimized using negative ionization mode. Both electrospray ionization (ESI) and atmospheric pressure chemical ionization (APCI) were tested as ionization sources. APCI was initially chosen as the ionization source because a significant number of the prostaglandin analog parent ions observed for ESI were sodium adducts [M + Na]+ . As an attempt to break up these sodium adducts, ammonium formate (NH4 HCO2 ) was added to the mobile phase with no positive results. While APCI resulted in less adducts forming, ESI provided greater sensitivity to the majority of compounds (e.g., 4-fold for Bima IE and >70-fold for Bima FA), and thus was ultimately used for this study. To avoid using the [M + Na]+ adducts as the precursor ions, the ions indicating loss of H2 O were instead used. As seen in Fig. 1, all the compounds studied contain multiple –OH groups and readily lose H2 O during ionization. 3.2. Liquid chromatography optimization A number of C18 columns were tested. Initially, APCI was chosen as the ionization source. When using APCI it is important to use a high flow-rate (≥0.5 mL/min) since sensitivity is flowdependent. Therefore, a long, large-bore column (Kinetex 5 ␮m XB-C18 250 mm × 4.6 mm, Phenomenex, Torrance, CA) was initially chosen. Unfortunately, peak shape and separation were not ideal. The switch to ESI allowed for lower flow rates and the possibility to use a sub-2 ␮m UPLC column (Acquity UPLC BEH Shield RP18 1.7 ␮m 100 mm × 2.1 mm, Waters, Milford, MA) which did produce acceptable peak shape and separation. However, the ability to utilize a single LC method for both ESI and APCI was intriguing. A Kinetex 2.6 ␮m XB-C18 100 mm × 2.1 mm column was chosen for the final method since it permits higher flow rates with lower back pressures and produces similar results to the chromatography observed with sub-2 ␮m UPLC columns. All 16 prostaglandin analogs eluted between 7.5 and 12 min during the 20 min run (Fig. 2a). A mobile phase composed of H2 O:MeOH:1%FA was used. The addition of 1% formic acid was beneficial to the positive ionization while not significantly decreasing the negative ionization, thus the mobile phase was used in both positive and negative ionization modes to allow for a single chromatographic run. MeOH was chosen initially over MeCN because it is more conducive to the ionization process in APCI. The mobile phase was not changed once ESI was implemented as there was no significant change in ionization due to mobile phase change. Also, using the H2 O:MeOH:1%FA gradient allows for either ESI or APCI to be utilized in separate runs with the same LC method. 3.3. Method validation An eight-point calibration curve was created for each prostaglandin analog in the concentration ranges from 0.25 to 50 ng/mL for bima, taf EA, and 17-PTPF2␣ MA; 0.5 to 100 ng/mL for bima IE, bima SA, clo IE, lat, trav, and 16-PPF2␣ EA; and 1 to

143

200 ng/mL for taf, taf EE, 17-PTPE2 SA, 17-CF3 PTPF2␣ EA, bima FA, lat FA, and clo FA-Na+ . The lower limit of quantitation (LLOQ) was established to be the lowest point on the validated calibration curve for each analyte based on a signal-to-noise ratio minimum of 10:1 for the primary transition (quantitative transition), a precision of 20%, and an accuracy of 80–120% [37,38]. The accuracy for all analytes ranged from 91 to 108%. The LLOQ for each analyte were determined (along with their precision in %RSD) to be: 0.25 ng/mL for bima (3.5%), taf EA (2.3%), and 17-PTPF2␣ MA (3.1%); 0.5 ng/mL for bima IE (3.7%), bima SA (3.6%), clo IE (3.3%), lat (1.9%), trav (2.7%), and 16-PPF2␣ EA (2.4%); and 1 ng/mL for taf (1.9%), taf EE (3.4%), 17-PTPE2 SA (4.6%), 17-CF3 PTPF2␣ EA (1.8%), bima FA (3.0%), lat FA (1.1%), and clo FA-Na+ (2.8%). The limit of detection (LOD) was also established to be the lowest point on the validated calibration curve for each analyte based on a signal-to-noise ratio minimum of 3:1 for the secondary transition (identification transition). A weighting factor of 1/x2 , where x is the analyte concentration, was applied and all r values (correlation coefficient) were greater than 0.99. Both intraday (Table 2) and interday (Table 3) validations were performed. The standards were stable in the autosampler set at 4 ◦ C over the course of the stability studies (8 days). All stock and working solutions were stored at −20 ◦ C and were used over the course of 6 months. 3.4. Sample preparation The majority of the cosmetic products studied were water-based serums, and the remainder of the products were mascaras. Since the matrix of each product differed from the next, the QuEChERS liquid–liquid extraction (LLE) was implemented as the sample preparation technique. Although salt-out organic solvent (acetonitrile) extraction in QuEChERS has been applied to the analysis of pesticides, natural toxins, veterinary drugs, environmental pollutants, and other chemical contaminants in primarily food matrices, this universal approach has been applied for the first time, to the analysis prostaglandin analogs in cosmetic products. The modular approach of QuEChERS and small sample sizes allow for optimization to be easily achieved by the selection of the appropriate extraction solvent and rapid evaluation of a variety of clean-up solvents. For example, we were able to evaluate two different 2 mL dispersive solid-phase extraction (dSPE) tubes: one containing 50 mg Carbon X, 50 mg C18 , and 150 mg MgSO4 , and the other containing 50 mg C18 and 150 mg MgSO4 (United Science); were used as a clean-up step after the LLE, but were not implemented in the method due to a significant loss of the analyte into the sorbent (