LC-MS/MS signal enhancement for estrogenic hormones in water samples using experimental design

Abstract

This paper describes optimization of LC-MS/MS conditions to develop a method for selective and sensitive determination of estrogens, namely: estradiol (E2), estrone (E1), estriol (E3), and synthetic estrogen as ethinyl estradiol (EE2) by using statistical experimental design methods. The optimization studies were conducted in three stages: (i) determination of minimum alkaline volume ratio (0–17% NH₄OH at 0.2 M) in mobile phase to maximize peak area by single factor experimental design; (ii) optimization of LC elution conditions including flow rate, acetonitrile concentration in standard and in mobile phase by Box–Behnken response surface method (RSM), and (iii) optimization of LC-MS/MS conditions for seven factors to maximize peak areas by Box–Behnken RSM. NH₄OH volume ratio significantly affected the peak area and it was maximized at 3–5% volume ratio. Predicted optimal LC elution conditions were % ACN_standard: 28, % ACN_mobile:44 and flow rate of 137 μL min⁻¹. The optimum instrumental conditions were determined as sheath gas pressure:33 arbitrary unit (Arb), ion sweep gas pressure: 0.4 Arb, aux gas pressure: 17 Arb, capillary temperature: 254 °C, vaporizer temperature: 352 °C, collision gas pressure: 1.9 mTorr, and spray voltage: 2740 V. Optimization provided substantial improvements in peak symmetry and resolution factor and a 20–25 times peak signal gain with respect to the instrumental self-optimized condition at detection limits of ng L⁻¹ levels were achieved. The lower detection limits were obtained by coupling the method with a SPE procedure for attaining high pre-concentration factors. The signal enhancement was about three orders of magnitude, which constitutes a remarkable sensitivity of the method.

---

1. Introduction

Endocrine disrupting compounds (EDCs) are known as steroids, pharmaceuticals, personal care products, industrial chemicals, and pesticides. It was reported that they cause feminization of males, sexually immature fish, hermaphroditism, and decreased fertility in wildlife.^1,2 Their adverse effects on aquatic life has led to studies to identify these chemicals in natural water samples.^3–5 Natural steroid estrogenic hormones, estrone (E1), 17β-estradiol (E2), estriol (E3) secreted by humans and animals, and the synthetic 17α-ethinyl estradiol (EE2) used by women were identified as having the highest endocrine disrupting potential. Among these, natural and synthetic estrogenic hormones excreted by humans and animals were shown to have the highest degree of estrogenic activity in aquatic environments.^6–8 The main source of steroid estrogens in aquatic environment is the discharge of municipal wastewater to receiving water. Since water treatment plants were not designed for the removal of estrogenic hormones, they may be either partially removed or biotransformed from one form to another during treatment.^9,10 Due to their adverse effects on aquatic life, and also their presence in municipal wastewaters, steroid hormones are considered to be future candidates in the list of EC Water Framework Directives. Hence, there is an urgent need to develop well-established analytical methods to measure ng L⁻¹ or even pg L⁻¹ levels of them in water samples.¹¹ Detection of such low level concentrations requires very sensitive analytical method development in terms of both sample preparation and instrumental conditions. Several analytical methods have been developed to identify and quantify these substances in water samples. Gas chromatographic techniques coupled to a mass spectrometric detector (GC-MS)^12,13 and tandem mass spectrometry (GC-MS/MS)^14,15 were used after derivatization of estrogenic hormones. Alternatively, liquid chromatographic techniques do not require a derivatization procedure and therefore, LC methods coupled with a number of detectors are becoming methods of choice. Different types of detectors including diode array (DAD),^16,17 electrochemical,¹⁸ fluorescence,¹⁹ mass spectrometry (MS),^20–22 and tandem mass spectrometry (MS/MS)^23–27 have been used. Data collection has been usually made with multiple reactions monitoring (MRM)^28–30 or selected reaction monitoring (SRM)^31–34 with triple quadrupole or quadrupole-ion trap systems to improve sensitivity and selectivity.

A literature survey revealed that optimization procedures for improving the lowest limit of detection (LOD) were usually achieved by studying each factor separately in a traditional trial and error method. Although sophisticated instruments provide an ease of self-optimization in many applications, a statistical method based on an experimental design will certainly give more insight about any interaction between the factors, particularly for methods that require handling a number of parameters. The number of experiments required can also be reduced by this means. However, there are only a few studies in the literature dealing with optimization of LC instrumental parameters by using a chemometric approach and comparing results with those obtained with self-optimization of the LC system.^35–37

By considering the needs for estrogenic hormone analysis in water samples at ng L⁻¹ concentrations, with a large number of factors and their interactions affecting the optimal conditions in tandem MS, our study was designed to use a response surface experimental design approach to develop an analytical method for estrogenic hormones measurement in water. By conducting only 62 experiments, LC-MS/MS conditions for seven factors (sheath gas pressure, ion sweep gas pressure, aux gas pressure, capillary temperature, vaporizer temperature, collision gas pressure, and spray voltage) were optimized to maximize peak areas in the chromatograms using the Box–Behnken response surface method (RSM). Using RSM for this purpose provided fast progress in method development for measuring estrogenic hormones (namely estradiol (E2), estrone (E1), estriol (E3), and synthetic estrogen as ethinyl estradiol (EE2)) and also for selective and sensitive determination of hormones in tap water. The chromatograms recorded under optimized conditions revealed that peak symmetry, resolution factor, and peak area of the hormones were substantially improved with regard to those obtained by instrument self optimized conditions.

2. Experimental

2.1. Chemicals

All the reagents used were analytical grade. Estrogenic hormones (E1, E2, E3, and EE2) were bought from Sigma-Aldrich (St Quentin Fallavier, France). The purity of hormones was at least 98%. Stock solutions of hormones were 10 mg L⁻¹ in methanol. Working solutions to inject onto LC-MS/MS were prepared by diluting each stock solution of hormones to the desired concentration in acetonitrile/water (28/72, v/v). HPLC MS grade acetonitrile, methanol, formic acid, and ammonia solution (25% in water) were supplied from Merck (Darmstadt, Germany). The C18 bonded silica SPE cartridge (500 mg) was obtained from Supelco (Oakville, Canada). Ultra pure water (18.2 MΩ cm) was used to prepare solutions. A mechanical vacuum pump under a 20 in Hg vacuum (Edwards E2M2, Crawley, UK) was used during sample preparation.

2.2. Instrumentation

LC-MS/MS measurements were performed using a Thermo TSQ Quantum Access Max Triple Quadrupole MS (Les Ulis, France) equipped with an Electrospray Ionization (ESI) probe, a degasser, a binary pump, an autosampler, and a column oven. Xcalibur software from Thermo Finnigan was used for data processing. The column was a Hypersil gold C18 (2.1 mm × 50 mm/1.9 μm particle size). The column oven temperature was set to 25 °C and injection volume was set to 25 μL in full loop mode. The cone position was “D” throughout the study. Samples were analyzed in negative ionization mode by using SRM mode. Precursor and product ion masses of the studied hormones are shown in Table 1.

Table 1 Precursor and product masses of studied estrogenic hormones

Analyte	E1^a	E2^a	E3^a	EE2^a
a E1: estrone, E2: estradiol, E3: estriol, EE1: ethinyl estradiol.
Precursor ion mass	269.1	271.1	287.0	295.1
Product ion mass	143.2	145.1	145.0	145.0
	145.3	183.2	171.1	159.1

---

2.3. Experimental conditions

This study comprises four parts including: (i) effect of alkaline solution in mobile phase on peak area of hormones, (ii) optimization of LC elution conditions using the Box–Behnken response surface method, (iii) optimization of LC-MS/MS operation conditions using the Box–Behnken response surface method, and (iv) development of a calibration curve for the optimized instrumental and elution conditions. Design Expert software (Trial Version 7.0.0 Stat-Ease Inc., Minneapolis, MN, USA) was used to estimate coefficients of the response functions developed in a response surface experimental design method and to determine the optimal conditions for LC separation and MS/MS acquisition.

2.3.1. Effect of alkaline solution in mobile phase. The aqueous mobile phase containing 40% ACN was pumped through the column at a flow rate of 300 μL min⁻¹. The concentration of standard hormone solutions injected into the LC-MS/MS system was fixed at 50 μg L⁻¹. Standard NH₄OH solution (0.2 M) was added into the main stream from a second line to be 3–17% in volume ratio. In order to evaluate the effect of volume ratio of NH₄OH in the mobile phase, a single factor experimental design method was used. The replicate number at each point was 10 and control experiments were also conducted in the absence of NH₄OH. Variance analysis was carried out to determine any significant effect of the volume ratio on peak area. Then, Least Significant Difference test (LSD) was applied to select the most significantly different volume ratio for the highest peak area. The autosampler conditions were as follows, injection volume: 25 μL, tray temperature: 25 °C, column oven temperature: 25 °C, and injection mode was full loop. Interface conditions were capillary temperature (CT): 280 °C, vaporizer temperature (VT): 120 °C, sheath gas pressure (SGP): 25 Arb, aux gas pressure (AGP): 20 Arb, ion sweep gas pressure (ISGP): 2.0 Arb, collision gas pressure (CGP): 1.5 mTorr, and spray voltage (SV) (positive/negative polarity): 2800/3500 V.

2.3.2. Optimization of LC elution conditions. The Box–Behnken experimental design method was used for optimization of LC elution conditions. The three independent variables were the volume percentage of ACN in standard solution (ACN_s) (X₁ = 22–28%) and in mobile phase (ACN_m) (X₂ = 44–50%), and flow rate (X₃ = 100–200 μL min⁻¹). The mobile phase was modified by adding 0.2 M NH₄OH solution to be 3% in flow conditions. The dependent variables were peak symmetry (PS) and resolution factor (RF). The coded and actual experimental points of the Box–Behnken experimental design are given (S-1).† Center point (C) was repeated 5 times.

The quadratic polynomial equation is given in eqn (1) for three significant independent variables, where y is estimate response, β₀ is constant; β₁, β₂ and β₃ are linear coefficients; β₁₂, β₁₃ and β₂₃ are two level interaction coefficients; and β₁₁, β₂₂ and β₃₃ are quadratic coefficients of the factors.^38,39

y = b₀ + b₁X₁ + b₂X₂ + b₃X₃ + b₁₂X₁X₂ + b₁₃X₁X₃ + b₂₃X₂X₃ + b₁₂₃X₁X₂X₃ + b₁₁X₁² + b₂₂X₂² + b₃₃X₃² (1)

The interface conditions were as follows; CT: 300 °C, VT: 375 °C, CGP: 1.5 Arb, SGP: 35 Arb, AGP: 20 Arb, ISGP: 0 Arb, and SV (positive/negative polarity): 3000/2750 V. Autosampler conditions were the same as the conditions given in Section 2.3.1.

2.3.3. Optimization of MS/MS conditions. The Box–Behnken experimental design was used to reveal optimal MS/MS conditions for analysis of target hormones. The independent variables were sheath gas pressure (35–38 Arb), ion sweep gas pressure (0–2 Arb), aux gas pressure (10–20 Arb), capillary temperature (250–350 °C), vaporizer temperature (325–425 °C), collision gas pressure (1.5–2.5 mTorr), and spray voltage (2500–3000 V). The dependent variable was peak area of each studied hormone at 1.0 ng mL⁻¹ concentration. The low, high, and center points of the factors are presented in S-2† where 62 experiments with 6 replicate at center points were randomly conducted. The actual design points are given in S-3.†

2.4. Sample analysis procedure

Prior to the extraction, aliquots of tap water samples were vacuum filtered through a PTFE membrane filter with a pore size of 0.45 μm. The pH was adjusted to 3.0 with addition of formic acid solution in appropriate amounts. Then, a 500 mL water sample was passed through the C18 cartridge conditioned with 3 mL of MeOH and 3 mL water. After the sample was loaded into the cartridge, the retained compounds were eluted with 10 mL of MeOH. In a thermostatic bath set at 25 °C, the extracts were evaporated until dry under a gentle nitrogen stream. Upon addition of 500 μL of ACN/water (28 : 72, v/v) solution, it was injected to LC/MS/MS system.

3. Results and discussions

Self-optimization of LC-MS/MS for the targeted hormones resulted in asymmetrical and not well resolved peaks. In order to improve the peak symmetry and resolution along with the sensitivity, preliminary studies were dedicated to a statistical experimental design to optimize selected parameters, primarily the chemical composition of the mobile phase and elution and operation conditions of the LC-MS/MS system. Then, calibration and validation studies were conducted under optimized conditions.

3.1. The effect of NH₄OH volume ratio on the peak area

It is known that the ionization ratio of the hormones can be substantially improved by the addition of alkaline solutions.²⁸ Considering their acidity constant (pK_a ∼ 10), it can be expected that increasing alkalinity will have a positive effect on ionization. Therefore, a single factor experimental design method was used to investigate the effect of NH₄OH volume ratio. For this purpose, 0.2 M standard NH₄OH solution was pumped through the mobile phase in different volume ratios. Fig. 1 collectively shows the change in the mean peak area obtained for E1, E2, E3, and EE2 at different NH₄OH ratios. ANOVA test of the data depicted that volume ratio significantly affects the peak areas of all hormones (p < 0.05). The maximum peak area was obtained at 3–5% range of NH₄OH volume ratio for each studied hormone. A LSD test showed that there is no significant difference between 3% and 5% in terms of the peak area (p > 0.05). Therefore, NH₄OH volume ratio was selected as 3% for further studies.

Fig. 1 Variation of the mean (n = 10) peak area of E1, E2, E3, and EE2 with NH₄OH volume ratio in the mobile phase.

3.2. Optimization of LC elution conditions

Gaussian peak shape and well separated peaks are desired in chromatographic analysis. The main factors that influence peak symmetry depend upon retention factors, solvent effects, incompatibility of the solute with the mobile phase, or development of an excessive void at the inlet of the column. Resolution can be affected significantly by peak sizes, peak shapes, etc. In order to achieve a Gaussian peak shape and well separated peaks, LC elution conditions were optimized by using a Box–Behnken RSM experimental design method. The selected responses were peak symmetry and resolution factor in the design. Three independent variables were ACN_s% (X₁), ACN_m% (X₂), and flow rate (X₃) in three levels as given in S-1.† Response equation coefficients for different hormones were determined by regression analysis and quadratic equations (eqn (2)–(5)) and developed as given in S-4.† Observed and predicted values for the resolution factors and peak symmetry for studied hormones based on the Box–Behnken experimental points are given in S-5.† The linear regression coefficient between observed and predicted values were R² > 0.90 for E2, E3, and EE2 revealing a good agreement. However, this value was found less than 0.90 for E1. Another statistical parameter used to determine if coefficients can be used to predict the response is Adequate Precision (AP). It measures the signal to noise ratio in the experiments and gives a factor by which a model can be judged to see if it is “adequate” to navigate through the design space and be able to predict the response. A ratio greater than 4 is desirable and states that a model equation can be used to predict the response for any value of factors within the range of experimental design.^40,41 Although R² value for E1 was less than 0.90, AP was found larger than 4. Consequently, it can be concluded that model coefficients are reliable to predict response at any experimental point within the studied ranges of the factors.

Two-level interactions of the factors for different hormones at constant % ACN_s (25%) were evaluated. Fig. 2A depicts the variation of PS values of all the targeted hormones at different ACN_m% and flow rates. Increasing flow rate adversely affects the peak symmetry of E1. The PS at a flow rate of 100 μL min⁻¹ was 1.57 and increased to 2.10 when the flow rate was doubled at ACN_m: 44%. The minimum PS value obtained was 1.5 at 50% ACN_m at a flow rate of 100 μL min⁻¹.

Fig. 2 Variation of peak symmetry of (A) E1, (B) E2, (C) E3, and (D) EE2 with ACN_m% and flow rate at ACN_s = 25%.

Variation of PS of E2 at a different flow rate and % ACN_m concentration is shown in Fig. 2B. The lowest PS value around 1.0 was observed at ACN_m > 47% and flow rate of 100 μL min⁻¹. The effect of flow rate was negative on the PS and more significant than that of % ACN_m. Increasing flow rate from 100 μL min⁻¹ to 200 μL min⁻¹ at 50% ACN_m resulted in raising the PS value from 1.1 to 2.7. The conditions can be determined as 50% ACN_m and flowing at a rate of 100 μL min⁻¹ for the best peak symmetry (PS = 1.0) of E2. The most significant factor in the case of the PS value of E3 was % ACN_m. Increasing ACN_m from 44% to 50% resulted in a substantial improvement in the peak symmetry from 2.3 to 1.1, respectively, at flow rate of 100 μL min⁻¹ (Fig. 2C). The effect of flow rate on the PS was not as strong as ACN_m%. Decrease in PS from 1.4 to 1.2 occurred when flow rate was increased from 100 μL min⁻¹ to 200 μL min⁻¹ at 50% ACN_m. The best peak symmetry around 1.0 for E3 can be obtained at flow rate = 200 μL min⁻¹ and 50% ACN_m.

Finally, variation of PS for EE2 with ACN_m and flow rate at ACN_s = 25% is depicted in Fig. 2D. The main factor that affects the PS value of EE2 is % ACN_m rather than the flow rate. PS value significantly decreased from 2.5 to 1.9 when % ACN_m was increased from 44% to 50% at minimum flow rate of 100 μL min⁻¹. However, the PS value varies between 1.8 and 1.9 for flow rates between 100 μL min⁻¹ and 200 μL min⁻¹ at 50% ACN_m. This slight variation in PS value for different values of flow rate can be observed even at lowest concentration of ACN_m = 44%. In summary, the best acceptable peak symmetry for EE2 can be obtained at flow rate of 100 μL min⁻¹, ACN_m = 50%.

Similar to peak symmetry, a response equation for the resolution factor was also developed. The resolution factor calculation of E1 was omitted, since it is the final peak to be eluted. ANOVA test results indicated that the most significant factor is the flow rate for all targeted hormones (p < 0.05). The lack of fit for all hormones was insignificant, which indicates reproducibility of the results (p > 0.05). Response equations developed for resolution factors of E2, E3, and EE2 (eqn (6)–(8), respectively) are given in S-6.† Regression coefficients between observed and predicted values were higher than 0.85 for E3 and E2 but it was 0.57 for EE2. However, AP for this hormone was greater than 4, which means that the model equation can be used to predict the response for any value of factors within the range of experimental design. Variation of RF values with two-level interactions of the factors was evaluated. % ACN_s was kept constant at 25%, which is the optimal concentration obtained after optimization. Fig. 3A depicts the effect of % ACN_m and flow rate on RF value of E3. Both factors significantly affected the RF value. For example, decreasing flow rate from 200 all min⁻¹ to 100 all min⁻¹ at 50% ACN_m resulted in increasing in RF value from 2.1 to 2.8. Similarly, there was an increase from 2.1 to 3.2 with the decrease in ACN_m concentration from 50% to 44%. However, the interaction effects of these factors were more significant than the main factor effects. As a result, decrease in both factors provided a substantial increase in the RF value. The maximum RF value with regard to E3 can be obtained as 4.5 at ACN_m = 44% and flow rate = 100 all min⁻¹. The relationship between factors for E2 was found as nearly linear with an insignificant curvature. Similar to RF value of E3, the combined effect of factors was more significant than the main effects of factors (Fig. 3B). The maximum RF value can be obtained as 1.3 at ACN_m = 44% and flow rate of 100 all min⁻¹. In the case of EE2, the main increase in the response can be obtained with the decrease in flow rate from 200 μL min⁻¹ to 100 μL min⁻¹ (Fig. 3C).

Fig. 3 Variation of RF values of (A) E3, (B) E2, and (C) EE2 with % ACN_m and flow rate at ACN_s = 25%.

The sum of the ranking difference method was also applied to compare the difference between observed and predicted values of resolution factors and peak symmetries given in S-5.† A downloadable program, CRRN-SRD with ties, was used⁴² to conduct the comparison. The averages of corresponding observed and predicted values were used as a golden standard as described elsewhere.⁴³ SRD_max was calculated as 144 for 17 different experimental conditions. The one with a SRD value closer to zero was defined as the better model and also proximity of SRD values for the compared pairs of results or models shows similarities of the model.⁴⁴ In our case, SRDs of observed and predicted values obtained for peak symmetry and resolution factors were calculated and given in S-5.† The differences in peak symmetry were the smallest for the values of E3 and EE2 indicating the proximity of the predicted values to the observed values. However, the SRD values for E1 and E2 were not close enough to show similarity, but they were close to zero indicating that response equations can be used for predicting peak symmetry. The best SRD results for a resolution factor were obtained with E3 and E2 where too small differences were observed for predicted and observed values. In the case of EE2, although there was a difference in SDR values, they were close to zero. Briefly, the response equation and the coefficients given in S-4 and S-6† can be used to predict peak symmetry and resolution factors for the analysis of hormones with LC-MS/MS.

Although statistical analysis indicates that the model equations can be used to predict the responses at any values of factor within the studied range, the best approach is to run an experiment at the point, which is different than the design points. S-7† depicts the observed and predicted values of two different experimental conditions of investigated factors. The results indicate that there is no substantial difference between observed and predicted values. The model can be used to predict the RF and PS values of hormone analysis within the studied ranges. An optimization tool of the software was run to determine the optimum LC elution conditions. The goals for optimizing resolution factors and peak symmetries were maximization of RS and to obtain PS between 1 and 2. The desirability (D) level selected to satisfy these goals was D = 1. The predicted optimal conditions for the defined goals were estimated as % ACN_s: 28, % ACN_m: 44, and flow rate: 137 μL min⁻¹.

3.3. Optimization of LC-MS/MS operation conditions

Under the optimal elution conditions and mobile phase composition given above, instrumental operating conditions were optimized by using the Box–Behnken experimental design method. The factors were pressure of sheath gas (SGP = X₁), ion sweep gas (ISGP = X₂), aux gas (AGP = X₃), temperature factor belonging to capillary temperature (CT = X₄), vaporizer temperature (VT = X₅), collision gas (CGP = X₆), and the spray voltage (SV = X₇). The ranges used for these factors are given in S-2.† The response was the peak area of the hormones. The significance of response equation coefficients were determined by ANOVA test. Response equations (eqn (9)–(12)) including coefficients were developed and given in S-8.† The regression coefficients for observed and predicted values as well as AP values were R > 0.90 and AP > 4, respectively for all the hormones (S-9†). The optimization tool of the software was run to determine the optimal conditions of variables for maximization of peak area of studied hormones. The conditions were determined as SGP = 32.52 Arb, ISGP = 0.41 Arb, AGP = 16.91 Arb, CT = 253.9 °C, VT = 351.88 °C, CGP = 1.0 mTorr, SV = 2740 V with the peak area of E1 = 204 298, E2 = 58 498, E3 = 87 257, and EE2 = 36 838 at 1.0 ng mL⁻¹ hormone concentrations.

The model equation can be used to evaluate variations in responses with different factors. For the construction of a 3D surface plot, one of the three factors was kept constant. In the present study, considering the 7 factors included, numerous plots can be drawn. In order to keep the process simple, only variation of peak areas of hormones with ISGP and SGP were plotted. The other factors were kept constant at the optimized values. Fig. 4A depicts the integration area of E1 for different values of ISGP and SGP. The maximum response was obtained around 2.1 × 10⁵ at SGP = 38 Arb. The effect of ISGP on peak area was important for the low values of SGP (32 Arb). The response increased from 1.9 × 10⁵ to around 2.1 × 10⁵ when ISGP was decreased from 2 Arb to 0.0 Arb. Almost the same effects of these factors can be observed for E2 (Fig. 4B) and E3 (Fig. 4C). A slight increase in the response was obtained when ISGP was decreased from 2 to 0 Arb for any values of SGPs. But, decreasing SGP provides a substantial increase in the response. On the other hand, the results were found to be quite different for EE2 (Fig. 4D). ISGP had a more significant effect on EE2 peak area than that of SGP. But, the interaction between factors was substantial as well, especially at low values of factors. The highest peak area was obtained as 4.0 × 10⁴ when SGP = 32 Arb and ISGP = 0 Arb.

Fig. 4 Surface plot of peak areas of (A) E1, (B) E2, (C) E3, and (D) EE2 at optimized conditions.

The model verification analysis was conducted for EE2, which displayed the lowest peak area. There was good agreement between observed and predicted values for EE2 at the points, which are different than design experimental points (S-10†). Finally, the optimization process was run for all the hormones and corresponding results were evaluated in terms of the sensitivity by taking into account the peak area. The aim in optimization was determination of LC-MS/MS operation conditions for maximization of integration area. Optimal conditions were determined as: SGP: 33 Arb, ISGP: 0.4 Arb, AGP: 17 Arb, CT: 254 °C, VT: 352 °C, CGP: 1.9 mTorr, and SV: 2740 V with desirability = 1. The chromatograms obtained under optimized conditions for 1.00 μg L⁻¹ mixed standard solution are given in Fig. 5. As can be followed from the chromatograms, well separated Gaussian shaped peaks were obtained for all the target hormones by using the SRM mode after optimization.

Fig. 5 (a) Total ion LC-MS-MS chromatogram of 1000 ng L⁻¹ standard mixture, and each chromatogram in SRM mode of (b) E1, (c) E2, (d) E3, and (e) EE2 with different retention time.

3.4. Calibration and validation studies

The quantitative determination was performed by injection of mixed standard solutions of the target hormones into the LC-MS/MS system. Calibration curves were constructed using standard solutions of E1, E2, E3, and EE2. The analytical merits of the method by considering limit of detection (LOD) and limit of quantification (LOQ), precision (repeatability, RSD < 20%), and (linearity, R² > 0.95) are given in Table 2.

Table 2 The analytical merits of developed methods

Instrumental method							SPE method
Analyte	IDL^a (ng L^−1)	IQLs^a (ng L^−1)	Linear range (ng L^−1)	Calibration equation	Regression coefficient (R^2)	Repeatability (% RSD)	LOD (ng L^−1)	LOQ (ng L^−1)	RSD (%)	Recovery^a (%)
a IDL: instrumental detection limit; IQL: instrumental quantification limit.
E1	1.7	5.5	5–30	y = 59.855x − 207.91	0.9933	7.9 (for 15 ng L^−1)	0.008	0.026	14.9	97.4
			20–150	y = 137.37x − 2419.1	0.9973	8.5 (for 100 ng L^−1)
			30–1000	y = 190.07x − 7079.4	0.9987	4.1 (for 150 ng L^−1)
						10.6 (for 15 ng L^−1)
E2	49.1	163.7	75–1000	y = 95.336x − 3779.8	0.9977	4.4 (for 150 ng L^−1)	0.025	0.084	11.4	89.4
						1.5 (for 600 ng L^−1)
E3	3.3	10.9	50–1000	y = 80.734x − 3165.7	0.9973	5.3 (for 100 ng L^−1)	0.241	0.803	14.2	81.4
			50–1000	y = 80.734x − 3165.7	0.9973	3.7 (for 300 ng L^−1)
EE2	23.2	77.2	75–250	y = 9.539x − 322.98	0.9818	9.7 (for 100 ng L^−1)	0.079	0.262	10.5	107.6
			150–1000	y = 21.415x − 3041.7	0.9937	1.9 (for 600 ng L^−1)

---

The calibration curves have good linearity in the concentration ranges studied with regression coefficients close to unity. The sensitivity of the method was assessed by calculating LOD and LOQ values, which are defined as the lowest concentration to exceed the mean baseline value of a water sample by three and ten times the signal/noise ratio, respectively. For stability experiments, freshly prepared 1.0 ng mL⁻¹ standard solution was injected into the LC-MS/MS system for the following four days using the same procedure. The results show that the peak area decreases much more than 50% after two days. Overall results reveal that combination of the separation and qualification power of the LC-MS/MS system enable us to determine the target hormones simultaneously in a ng L⁻¹ range, which is required for environmental samples.

3.5. Method application and comparison with previous methods

The method was applied to tap water samples spiked with 1.0 ng L⁻¹ of each estrogenic hormone. Then, the SPE procedure given in the Experimental part was applied. The peak area values obtained from LC-MS/MS measurement for spiked samples were compared with those self optimization values and the concentration ratio was found about three orders of magnitude, which constitutes the remarkable sensitivity of the method. Analytical merits of the SPE method including limit of detection (LOD) and limit of quantification (LOQ) levels along with the repeatability are given in Table 2. As can be seen from the table, the method provides sensitive and repeatable results for tap water samples with acceptable recovery values. Accuracy of the method was tested by recovery studies conducted with spiked tap water samples. Recovery values for the SPE method after addition of 1.0 ng L⁻¹ mixed standard solutions of the estrogenic hormones were between 81.4 and 107.6%.

Analytical method development for the measurement of estrogenic hormones in water and wastewater samples has received considerable attention for the last decade. The major concerns were the determination of mobile phase composition and separation methods to obtain high recovery and low LOD values. Ammonia and methanol are the most commonly used mobile phases. Formic acid and ammonia additions were also performed to improve separation. Most of the studies were conducted in LC-MS/MS negative electrospray ionization mode (ESI⁻). Isobe et al.²⁴ developed an analytical method for E1, E2, E3, and EE2 and their conjugates in lake water samples by using LC-MS/MS. Acetonitrile was the mobile phase, and triethylamine (TEA) and ammonia were used to increase efficiency of ionization. Percent recovery was in the range of 116–90 for the studied hormones. Laganà et al.³⁰ studied analysis of estrogenic hormones in different matrices including surface water and sludge treatment plant influent and effluent using LC-MS/MS with ESI⁻. A mobile phase of acetonitrile and ammonia solution (50 mmol) was used with a flow rate of 0.11 mL min⁻¹ for post column separation. Method recovery percentage was over 90% for the estrogenic hormones. Vanderford et al.²³ used methanol and 0.1% formic acid (v/v) in water as mobile phase components for the analysis of E2 and EE2. Recoveries of E2 and EE2 at 92% and 96% respectively, from surface water spiked with 10 ng L⁻¹ were reported. Similarly, Trenholm et al.²⁹ used 0.1% formic acid (v/v) in water and methanol as mobile phase components. The recoveries were E2: 92%, E1: 90%, E3: 101%, and EE2: 92%. Rodriguez et al.²⁰ studied method development for measurement of estrogenic hormones at ng L⁻¹ concentrations using LC-MS/MS in the ESI negative mode. Mobile phase components consisted of acetonitrile and pure water. The recoveries of hormones were over 91%, and 100% recovery of E1 from a water sample was achieved. LOD values of developed methods varied depending on matrix, mobile phase, and extraction methods. Recently, ionic liquid intercalated clay sorbents were used for micro solid phase extraction of steroid hormones from water samples.⁴⁵ The mobile phase was composed of 44% acetonitrile, 3% 0.20 M ammonia solution and 53% water. The developed method provided 86.9–97.7% recovery of hormones and just 10 mL of the sample was adequate for achieving a sensitive analysis. Table 3 summarizes the results of LC-MS/MS methods using different strategies for sample treatment. Relatively lower LOD values obtained in this study can be attributed to determination of optimum instrumental operation conditions through RSM.

Table 3 Comparison of the results with other studies carried out in LC-MS/MS

Sample matrix	Extraction method	Sample volume	LOD (ng L^−1)				Ref. no.
			E1	E2	E3	EE2
a NA: not available.
Surface and ground water	SPE C18	1000 mL	0.02	0.01	0.03	0.20	7
Drinking water	SPE-Lichrolut RP-18	500 mL	2.5	2.5	5.04	3.22	20
Distilled water	SPE-StrataX or HLB	200 mL	0.9	2.5	4.3	2.8	22
Surface water	SPE-Lichrolut RP-18	1000 mL	1	NA^a	NA^a	1	23
Lake water	SPE C18	1000 mL	0.1	0.3	1.5	0.2	24
Water samples	SPE C18	1000 mL	0.4	0.5	2	1	25
River water	SPE C18	500 mL	15	30	30	30	27
River water	SPE-C18	1000 mL	0.1	0.2	0.3	0.4	30
River/drinking water and WWTP effluents	SPE C18	250 mL	0.24	0.85	0.62	0.62	32
River water	SPE-Lichrolut RP-18	500 mL	1.15	2.27	1.13	7.55	33
River water	μ-SPE	10 mL	0.012	0.062	0.018	0.693	45
Tap water	SPE C18	500 mL	0.008	0.025	0.241	0.079	This study

---

4. Conclusions

EDCs are complex chemicals and their effects on human and animals are not well known yet. But, it could be a main concern to protect human and animal life soon when their effects are clearly understood. The ability to understand adverse effects of these chemicals on humans and animals is directly related with capability to determine ng L⁻¹ or pg L⁻¹ concentrations of these chemicals. For this purpose, a sensitive and selective method for the measurement of estrogenic hormones (E1, E2, E3, and EE1) with LC-MS/MS by using experimentally designed methods was developed. The applied instrumental condition optimization methodology provided better peak symmetry and higher resolution factors along with increased sensitivity (a signal gain about 20–25) compared with instrumental self-optimized conditions. The detection limits attained revealed that the proposed method can be effectively used for contaminated water samples, particularly when a small volume of sample is available. Further improvement in the sensitivity was achieved by coupling the instrumental method with a SPE procedure for attaining high pre-concentration factors. Consequently, development of methods for measurement of very low levels of hormones in wastewater will be very helpful to improve knowledge about their adverse health effects on humans and animals. Moreover, these kinds of studies will also provide supportive information to establish national or international legislation for discharge of these chemicals.

Acknowledgements

This study was supported by Dokuz Eylul University with grant number 2011.KB.FEN.010 and by Ege University with grant number 2012.FEN.043.

References

1. R. K. Bhandari, S. L. Deem, D. K. Holliday, C. M. Jandegian, C. D. Kassotis, S. C. Nagel, D. E. Tillitt, F. S. vom Saal and C. S. Rosenfeld, Gen. Comp. Endocrinol., 2015, 214, 195–219.
2. J. K. Leet, S. Sassman, J. J. Amberg, A. W. Olmstead, L. S. Lee, G. T. Ankley and M. S. Sepúlveda, Aquat. Toxicol., 2015, 158, 98–107.
3. A. C. V. Dias, F. W. Gomes, D. M. Bila, G. L. Sant'Anna Jr and M. Dezotti, Ecotoxicol. Environ. Saf., 2015, 120, 41–47.
4. P. H. Chou, Y. L. Lin, T. C. Liu and K. Y. Chen, Chemosphere, 2015, 138, 814–820.
5. S. Esteban, M. Gorga, M. Petrovic, S. González-Alonso, D. Barceló and Y. Valcárcel, Sci. Total Environ., 2014, 466–467, 939–951.
6. K. Sornalingam, A. McDonagh and J. L. Zhou, Sci. Total Environ., 2016, 550, 209–224.
7. G. Rasier, J. Toppari, A. S. Parent and J. P. Bourguignon, Mol. Cell. Endocrinol., 2006, 254, 187.
8. M. S. Souza, P. Hallgren, E. Balseiro and L. A. Hansson, Environ. Pollut., 2013, 178, 237–243.
9. D. Belhaj, R. Baccar, I. Jaabiri, J. Bouzid, M. Kallel, H. Ayadi and J. L. Zhou, Sci. Total Environ., 2015, 505, 154–160.
10. H. Hamid and C. Eskicioglu, Water Res., 2012, 46, 5813–5833.
11. E. Vulliet, L. Wiest, R. Baudot and M. F. Grenier-Loustalot, J. Chromatogr. A, 2008, 1210, 84.
12. S. Combalbert, M. L. Pype, N. Bernet and G. Hernandez-Raquet, Anal. Bioanal. Chem., 2010, 398, 973.
13. J. L. Zhao, G. G. Ying, L. Wang, J. F. Yang, X. B. Yang, L. H. Yang and X. Li, Sci. Total Environ., 2009, 407, 962.
14. B. Albero, C. Sánchez-Brunete, E. Miguel, R. A. Pérez and J. L. Tadeo, J. Chromatogr. A, 2013, 1283, 39–45.
15. T. Trinh, N. B. Harden, H. M. Coleman and S. J. Khan, J. Chromatogr. A, 2011, 1218, 1668.
16. M. J. Lopez de Alda and D. Barcelo, J. Chromatogr. A, 2001, 911, 203.
17. P. K. Zarzycki, E. Włodarczyk and M. J. Baran, J. Chromatogr. A, 2009, 1216, 7612.
18. A. Penalver, E. Pocurull, F. Borrull and R. Marce, J. Chromatogr. A, 2002, 964, 153.
19. L. Mao, C. Sun, H. Zhang, Y. Li and D. Wu, Anal. Chim. Acta, 2004, 522, 241.
20. S. Rodriguez-Mozaz, M. J. Lopez de Alda and D. Barcelo, J. Chromatogr. A, 2004, 1045, 85–92.
21. Y. H. Lin, C. Y. Chen and G. S. Wang, Rapid Commun. Mass Spectrom., 2007, 21, 1973.
22. E. Vulliet, J. B. Baugros, M. M. Flament-Waton and M. F. Grenier-Loustalot, Anal. Bioanal. Chem., 2007, 387, 2143.
23. B. J. Vanderford, R. A. Pearson, D. J. Rexing and S. A. Snyder, Anal. Chem., 2003, 75, 6265.
24. T. Isobe, H. Shiraishi, M. Yasuda, A. Shinoda, H. Suzuki and M. Morita, J. Chromatogr. A, 2003, 984, 195.
25. M. Farre, M. Kuster, R. Brix, F. Rubio, M. J. Alda and D. Barcelo, J. Chromatogr. A, 2007, 1160, 166.
26. V. Kumar, N. Nakada, M. Yasojima, N. Yamashita, A. C. Johnson and H. Tanaka, Chemosphere, 2009, 77, 1440.
27. M. Pedrouzo, F. Borrull, E. Pocurull and R. M. Marce, Talanta, 2009, 78, 1327.
28. Y. Koh, T. Chiu, A. Boobis, E. Cartmell, J. Lester and M. Scrimshaw, J. Chromatogr. A, 2007, 1173, 81.
29. R. A. Trenholm, B. J. Vanderford, J. C. Holady, D. J. Rexing and S. A. Snyder, Chemosphere, 2006, 65, 1990.
30. A. Laganà, A. Bacaloni, I. De Leva, A. Faberi, G. Fago and A. Marino, Anal. Chim. Acta, 2004, 501, 79.
31. S. Esteban, M. Gorga, M. Petrovic, S. González-Alonso, D. Barceló and Y. Valcárcel, Sci. Total Environ., 2014, 466–467, 939–951.
32. M. Kuster, M. J. Lopez de Alda, M. D. Hernando, M. Petrovic, J. Martin-Alonso and D. A. Barcelo, J. Hydrol., 2008, 358, 112.
33. M. Kuster, D. D. A. Azevedo, M. Lopez de Alda, F. R. Aquino Neto and D. Barcelo, Environ. Int., 2009, 35, 997.
34. L. Sun, W. Yong, X. Chu and J. M. Lin, J. Chromatogr. A, 2009, 1216, 5416.
35. I. Molnar, H. J. Rieger and K. Monks, J. Chromatogr. A, 2010, 1217, 3193.
36. B. Debrus, P. Lebrun, J. M. Kindenge, F. Lecomte, A. Ceccato, G. Caliaro, J. Mbay, B. Boulanger, R. Marini and E. Rozet, J. Chromatogr. A, 2011, 1218, 5205.
37. K. Monks, I. Molnar, H. J. Rieger, B. Bogati and E. Szabo, J. Chromatogr. A, 2012, 1232, 218.
38. M. A. Bezerra, R. E. Santelli, E. P. Oliveira, L. S. Villar and L. A. Escaleira, Talanta, 2008, 76, 965.
39. S. C. Ferreira, R. Bruns, H. Ferreira, G. Matos, J. David, G. Brandao and E. P. da Silva, Anal. Chim. Acta, 2007, 597, 179.
40. D. C. Montgomery, Design and Analysis of Experiments, John Willey & Sons Inc., 5th edn, 2001, vol. 104.
41. F. J. Cerino Córdova, A. M. García León, R. B. Garcia Reyes, M. T. Garza González, E. Soto Regalado, M. N. Sánchez González and I. Quezada López, Int. J. Environ. Sci. Technol., 2011, 8(4), 695.
42. K. Kollar-Hunek and K. Heberger, Chemom. Intell. Lab. Syst., 2013, 127, 139.
43. K. Heberger, TrAC, Trends Anal. Chem., 2010, 29, 101.
44. K. Heberger and K. Kollar-Hunek, J. Chemom., 2011, 25, 151.
45. C. Aftafa, F. O. Pelit, E. E. Yalcinkaya, H. Turkmen, I. Kapdan and F. Ertas, J. Chromatogr. A, 2014, 1361, 43.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6ra06526k

---