A Reliable and Cost-Effective Method for Determination of Endocrine-Disrupting Compounds in Coastal Waters, Suspended Particulate Matter, and Sediments by Ultrafast Liquid Chromatography Coupled to Photodiode Array and Fluorescence Detectors

Analytical methods for determining 14 endocrine-disrupting compounds (EDCs) in coastal waters, suspended particles, and sediment samples were successfully performed by ultrafast liquid chromatography with photodiode array and fluorescence detections (UFLC-PDA-FLD). Solid-phase extraction (SPE) and ultrasound-assisted extraction (USE) were used for sample preparation. Two chromatographic methods have been developed. An isocratic separation method was used to separate bisphenol A (BPA) and steroids and another gradient elution method to separate phthalates and alkylphenols. The detection by fluorescence was used for alkylphenols, BPA, and steroids and photodiode array (PDA) for phthalates. Limits of detection (LOD) ranged from 0.41 (4-tert-octylphenol (4tOP)) to 63 ng L (dibutylphthalate (DBP)), 0.41 (4tOP) to 63.2 ng g dried weight (dw) coastal waters, and solid samples (suspended particulate matter (SPM) and sediment samples), respectively. Recoveries ranged from 52 (diethylphthalate (DEP)) to 116% (DBP) for water, from 54 (DEP) to 108% (estrone (E1)) for SPM, and from 62 (4-n-nonylphenol (4nNP)) to 117% (4-n-octylphenol (4nOP)) for sediment samples. Finally, with the minimization of reagents and energy, the proposed methods were applied to samples collected from Todos os Santos Bay (BTS), Bahia, Northeastern Brazil.


Introduction
Coastal ecosystems play an important role in receiving hydrophobic contaminants from different routes. In general, marine areas are the final destination of pollutants from land runoff, atmospheric deposition, and discharges from industrial, agricultural, and domestic effluents. [1][2][3][4][5] Due to their hydrophobic characteristics, once organic compounds enter into the water bodies, they tend to be preferentially associated with suspended particulate matter (SPM) and sediments as well as being absorbed by biota. [6][7][8][9][10][11][12] This tendency is likely to be even more pronounced in marine waters due to their inherently high salt content. Those characteristics make the hydrophobic organic compounds easily migrate to other compartments, binding to organic matter. 1,[13][14][15] In turn, once in the biota, hydrophobic compounds may take part in the food web, circulating among different trophic levels as they may bioaccumulate and/or biomagnify.
Depending on the chemical structure and other physicochemical characteristics, hydrophobic organic compounds are likely to have different (eco)toxicological effects on wildlife and human health. 7,[16][17][18] Indeed, there is a large group of hydrophobic organic chemicals known or suspected to cause some level of endocrine dysfunctions in living beings. They are classified as endocrine-disrupting compounds (EDCs), which have received increasing interest in the last decades due to their increasing occurrence and persistence in the environment. Some EDCs can bind to hormone receptors in organisms and disrupt regular suspended particulate matter, and sediment, among others), many different isolations, preconcentration, and clean-up steps are proposed in the literature. [26][27][28][30][31][32][33] For analysis of aqueous samples, most of the sample preparation methods often employ solid-phase extraction (SPE), 10,20,21,26,34,35 liquid-phase microextraction (LPME), 33,36,37 and stir bar sorptive extraction (SBSE). 29 In terms of sample complexity, sample preparation methods for solid environmental matrices become more difficult and troublesome. 27 Solid sample preparation methods generally include ultrasoundassisted extraction (USE), [38][39][40] microwave-assisted extraction (MAE), 41 and pressurized liquid extraction (PLE), 42 followed by one or more of the clean-up steps. Cleaning up is often done by SPE, commonly employing C18, silica, or Florisil cartridges. 27 Nonetheless, for each technique, both advantages and disadvantages are considered in analyzing EDCs in environmental matrices. Indeed, the authors kindly suggest to the readers to refer to very interesting reviews published by Salgueiro-González et al. 26,27 in order to deal with those in more details, since they are out of the scope of the present study.
EDC analysis is largely done by either gas chromatography coupled to flame ionization (GC-FID) 9,26,27 or mass spectrometer (GC-MS or GC-MS/MS) 9,13,26,27 detectors as well as liquid chromatography with ultraviolet (LC-UV), 9,10 fluorescence (LC-FLD), 9,10 or mass spectrometer (LC-MS, LC-MS/MS) 9,12 detectors. Although GC-MS is a robust instrumental technique with adequately low limits of detection (LOD) and limits of quantification (LOQ) for determinations of plasticizers and hormones in environmental samples, it is still necessary to have a derivatization step in the procedure. This is due to the fact that separation is improved when silylation or acylation is done in order to derivatize the hydroxyl groups of the EDCs. This step is important for reducing polarity, increasing volatility, and thermal stability of the analytes, all interesting characteristics for GC-based analyses. However, derivatization may generally result in important analyte losses during analyses (i.e., no derivatization reaction is able to reach a 100% yield), which is not affordable in most situations when determining EDCs in environmental complex matrices. On the other hand, liquid chromatography-based methods generally do not require a derivatization step, which could potentially simplify and become a faster way for the EDCs determination (and reflect in lower LOD/LOQ values). 10,20,21,[26][27][28]30,34,35,[43][44][45][46][47][48] This study aimed to develop sensitive, simple, reliable, and cost-effective analytical methodologies for the determination of plasticizers and steroid hormones in marine water, suspended particulate matter (SPM), and sediment samples by ultrafast liquid chromatography coupled to a photodiode array (PDA) and fluorescence detectors (UFLC-PDA-FLD). In this work, we chose to study fourteen target EDCs, namely BPA, 4-n-octylphenol (4nOP), 4-tert-octylphenol (4tOP), 4-n-nonylphenol (4nNP), E1, E2, E3, EE2, dimethylphthalate (DMP), diethylphthalate (DEP), dibutylphthalate (DBP), butylbenzylphthalate (BBP), di-(2-ethylhexyl)phthalate (DEHP), and di-n-octylphthalate (DnOP), which are widely distributed in the aquatic systems. Furthermore, we optimized and validated our methods according to International Union of Pure and Applied Chemistry (IUPAC) guidelines [49][50][51][52] regarding response function, linear range, linearity, LOD, LOQ, selectivity, precision, matrix effect, accuracy, and application to real samples. Indeed, the proposed methods were successfully applied to coastal waters, SPM, and sediment samples collected from Todos os Santos Bay (BTS), Bahia, Northeastern Brazil.

Sample collection
We collected coastal waters, SPM, and sediments from Todos os Santos Bay, Bahia, Brazil, to develop and validate the analytical methods. Considering there are no commercially available standard reference materials (SRM) for EDCs, for the method development, we collected waters from the Ondina beach (OND), Salvador, Bahia ( Figure 1). This beach was chosen for collecting water samples at the method development and validation steps since it is close to the University, and it is known 10 this site has EDCs levels below the method LOD. In order to proceed with the method development, water samples were enriched with a known amount of the EDCs standards. We collected water samples from 10 cm below the surface into 4 L amber bottles previously cleaned, as stated in the QA/QC description (presented in SI section). Prior to collection, the bottles were rinsed three times with the ambient water sample. After collection, samples were kept in ice and rapidly transported to the laboratory. In the lab, the water samples were vacuum filtered (vacuum pump WP611560, Millipore Corporation, USA) through glass fiber filters (calcinated at 400 °C for 4 h, 47 mm diameter, 0.7 µm pore size, Whatmann, UK). The dissolved fractions of the water samples were analyzed immediately after preparation.
The samples for suspended particulate matter (SPM) were collected from Ribeira bay (RIB) (Figure 1), a site that has relatively large amounts of SPM. The amount of filtered water varied according to the SPM mass on filters in order to obtain filter masses of at least ca. 15 mg dried weight (dw). Filters were dried at room temperature in desiccators until they reach constant masses and then kept in the freezer for up to 3 months until analyses.
For sediments, tests were performed using a sample previously collected from Paraguaçu Estuary (sample 4) 52 that presents low EDC concentrations. Surface sediment (first 5 cm) was collected by using a Van Veen dredge. The sample was collected with a metal spoon covered with previously decontaminated aluminum foil, stored in DCM-cleaned and calcinated aluminum containers, and kept in the freezer until analysis.
Water and SPM (n = 3) simple samples from several localities along the Todos os Santos Bay, namely: Alagados (ALA), Caboto (CAB), Mataripe (MAT), São Francisco do Conde (SFC), and Terminal Marítimo (TMA) (Figure 1), were used for method validation. For sediments, composite samples were collected from the Subaé Estuary ( Figure 1). These samples were collected as described above and were analyzed according to the optimized procedure for EDCs.

Development, optimization, and validation of an analytical method for EDCs
Analytical method development was done through univariate procedures. For this purpose, we considered the following reasoning: (i) extraction of EDCs in coastal waters (dissolved fraction) by using SPE as sample preparation method, (ii) extraction of EDCs in SPM by microwave-assisted extraction (MAE) and ultrasoundassisted extraction (UAE), (iii) extraction of EDCs in sediments by USE, and (iv) separation and determination of EDCs by UFLC-PDA-FLD. Validation was done following the IUPAC recommendations for calibration curves, repeatability, selectivity, sample matrix effect, precision, accuracy, limit of detection (LOD), limit of quantification (LOQ), and analysis of real samples (which were collected in Todos os Santos Bay, Bahia, Brazil). [48][49][50][51][53][54][55][56][57] Water sample preparation The preconcentration apparatus used for coastal waters sample preparation is a modified version of the apparatus proposed by Sodré et al. 58 This system can preconcentrate up to 4 L water samples through commercial SPE cartridges. It is composed by a set of 4 SPE preconcentration lines manufactured with polytetrafluoroethylene (PTFE) pieces, sphere valves, metal nipples, and stainless-steel tube adaptors. This SPE set is connected to a reservoir to collect water coming through SPE cartridges by using a vacuum pump ( Figure S1, SI section).
We tested two types of cartridges: C18 (6 mL polypropylene tubes with 1 g adsorbent, Supelco, USA) and Oasis HLB (6 mL glass tubes with 200 mg adsorbent, Waters, USA). In regard to the Oasis HLB cartridges, firstly we did not condition them before use (as instructed by the manufacturers in their certificate). In this way, analyte recovery tests using Oasis HLB cartridges were performed without any prior conditioning, and even then, the analytes were poorly retained. So, after that, we tested different forms of conditioning for both cartridges. For each type of cartridge, we carried out four different SPE cartridge conditioning tests: (i) 15 mL ACN/MeOH (1:1), (ii) 15 mL ACN followed by 15 mL MeOH, (iii) 15 mL acetone followed 15 mL MeOH, and (iv) 15 mL MeOH. Each test was done at 1.5 mL min -1 and then 15 mL ultrapure water (also at 1.5 mL min -1 ). At this step, we used coastal water samples enriched with a final concentration of 50 µg L -1 of E2, E3, EE2, BPA, 4nOP, 4tOP, and 4nNP, 250 µg L -1 final concentration for E1, and 10 mg L -1 final concentration for each phthalate. Then, we passed the 1 L water sample, which is the preconcentration step by itself. After that, we rinsed the cartridges with 15 mL ultrapure water to remove sea salt and then dried them at room temperature. For eluting the EDCs from cartridges, we tested a fractionated elution (2 times of 2 mL followed by 4 times of 1 mL MeOH) 10 and analyzed them as separate fractions. We observed the complete EDCs removal from the cartridge after elution of 5 mL MeOH. In order to assure EDCs' complete elution, cartridges were eluted with 6 mL MeOH. Next, the extracts were filtered through syringe filter Millex units (cellulose membrane, 0.22 µm pore size, 15 mm diameter, Millipore Corporation, Bedford, USA). Next, the extracts were injected into a UFLC-PDA-FLD system. After the tests with cartridges and the conditioning method, we also assessed if acidifying the water samples would improve extraction. We tested pH values at 1, 2, 4, and 6 by using an HCl 1.0 mol L -1 solution to adjust the pH of the enriched water samples. All tests were done in triplicates.

SPM sample preparation
We tested SPM extractions through microwave-assisted extraction (MAE) (Multiwave Pro, Anton Paar, USA) and ultrasound-assisted extraction (USE) (ultrasonic bath, model T760DH, Elma, Germany) procedures. In both cases, we tested 6 different extracting solvents: ACN 100%, DCM 100%, ACN: For MAE, we tested each extracting solution with 3 different heating programming: (i) rising from room temperature to 60 °C for 10 min, staying at 60 °C for 30 min, and cooling back to room temperature in 20 min (total runtime: 60 min); (ii) rising from room temperature to 80 °C for 10 min, staying at 80 °C for 30 min, and cooling back to room temperature in 20 min (total runtime: 60 min); and (iii) rising from room temperature to 110 °C for 10 min, staying at 110 °C for 30 min, and cooling back to room temperature in 20 min (total runtime: 60 min). Thus, we tested 3 extraction times (10, 20, and 30 min) at 120 W in the USE experiments.
During MAE and USE tests, SPM filters were cut into small pieces with cleaned scissors and either transferred to extraction tubes (for MAE) or 10 mL amber flasks (for USE). For both procedures, 7 mL of extraction solvents were added on top of the filter pieces. This extraction solvent volume was set by the minimum volume of MAE as determined by the manufacturer and in order to keep the same solvent-to-filter proportion in USE. After extraction, samples were centrifuged at 2000 rpm for 10 min and the supernatants were collected. 59 Next, we cleaned up the extracts by passing them through Florisil cartridges (3 mL, 500 mg adsorbent mass, Agilent Technologies, USA) 60 previously conditioned with 2 mL MeOH. Following, the extracts were filtered through syringe filter Millex units (cellulose membrane, 0.22 µm pore size, 15 mm diameter, Millipore Corporation, Bedford, USA) and directly transferred to vials to be then injected into the UFLC-PDA-FLD system. All solvent tests and blank tests were done in triplicate.

Sediment sample preparation
All the frozen sediment samples were lyophilized (lyophilizer Alpha 1-4 LDplus, Christ, Germany) and ground with a ball mill (8000 D, SPex sample prep., USA). An aliquot of 1 g sediment mass was extracted with 7 mL of solvent extraction mix, which presented the best results for extractions. Then, the sediment extract was cleaned up and analyzed by the best SPM results procedure. All experiments and blanks were done in triplicates. After optimizing the sample preparation procedures for water, SPM, and sediment matrices, we performed the recovery tests. We added an N 2 stream extract drying step for these tests, followed by an extract resuspension to 500, 250, and 500 µL of MeOH for water, SPM, and sediment samples, respectively.

Development and validation of a chromatographic method by UFLC-PDA-FLD
The significant differences in phthalates' physical and chemical characteristics compared to the other EDCs led us to develop two different chromatographic methods to determine the 14 EDCs in this study. The first method was adjusted for the phthalates and alkylphenols, and the second one was used for the steroid hormones and BPA. Both methods used the same column (Shim-pack XR-ODS, 2 mm inner diameter, 150 mm length, 2.2 µm particle size, and 8 nm pore size). For each method, we tested: (i) different eluents, (ii) different eluent gradient programming, (iii) variation in the eluent flowrate, and (iv) variation of the column temperature.
In evaluating different eluents, we tested two sets of binary eluents: methanol and acidified ultrapure water (2% v v -1 acetic acid) and acetonitrile and acidified ultrapure water (2% v v -1 acetic acid). Individual and mix analytical standards were injected for each eluent system in order to adjust the best eluent gradient programming. The best separation condition was obtained after the tests with the gradient programming and the retention time for each compound. After that, tests were performed varying the eluent flowrate from 0.1 to 0.4 mL min -1 , by incrementing 0.05 mL min -1 at a time.

Sample preparation
Sample preconcentration using coastal waters started by testing two types of SPE cartridges (Oasis HLB and C18), followed by the cartridge conditioning step, and then the sample acidification test. Recovery levels are presented in Table 1.
The best recovery levels were reached using C18 cartridges conditioned with 15 mL ACN followed by 15 mL MeOH. Thus, even though Oasis HLB cartridges have been developed specifically for EDCs, they did not show better results than C18 cartridges. It may be due to the fact C18 cartridges present less polarity and more active sites than the Oasis HLB cartridges (copolymer of n-vinylpyrrolidone and divinylbenzene), favoring better retention of the studied low polarity EDCs. In general, reverse phase functionalized silica SPE cartridges are generally conditioned with water miscible-solvents, such as methanol. [62][63][64][65][66] Within this type of adsorbent, methanol works wetting its surface and penetrating the alkyl bonds, making active sites available for interactions with the analytes. In turn, acetonitrile is used to improve the adsorbent conditioning by removing the possible interferents present in the cartridges with more efficiency.
Rinsing cartridges with ultrapure water after the salted water samples concentration step is very important to avoid chromatograph problems, such as column pore obstruction or salt precipitation in the injector. Additionally, not removing the salt content before sample elution may cause analyte losses due to the salting-out effect when eluting the sample into the vial, resulting in low recovery levels.
In the attempt of improving recoveries and preserving samples after field collection and filtration, we tested sample acidification at different pH values. The chemical equilibria of EDCs in aqueous solutions can be controlled by the pH of the media. In this equilibrium, at low pH, the ionization of the hydroxyl group is suppressed, favoring the molecular form of the compound (i.e., the hydroxyl group remains protonated instead of donating the H + ). In this way, the protonated form (the molecular form) of the EDCs interacts more effectively with the alkyl groups from the C18 cartridges and the chromatographic column. As a result, the best recoveries were reached at pH 2 (Table S1, SI section). For the same reasons, we chose to work with acidified ultrapure water in the binary eluent system. By using acidified water in the eluent system, we could preserve eluents for a longer time and improve separation. Therefore, considering the presented results here, the optimized conditions for seawater sample preparation were: 1 L acidified seawater sample (at pH 2), a preconcentration step with C18 cartridges, conditioning of the C18 cartridge with 15 mL ACN followed by 15 mL MeOH, salt removal by rinsing cartridges with 15 mL ultrapure water prior the EDCs elution, followed by analytes elution with 6 mL MeOH, and analysis by UFLC-PDA-FLD ( Figure S2, SI section).
In order to establish the optimal conditions for SPM sample preparation, we experimented six extraction solvents (ACN 100%, DCM 100%, ACN:DCM 1:1 (v v -1 ), ACN:DCM 1:3 (v v -1 ), ACN:DCM 3:1 (v v -1 ), and ACN:MeOH 1:1 (v v -1 )) with MAE and USE. Results for MAE and USE are presented in Figures 2 and 3, respectively. Results for MAE ( Figure 2) presented higher variability than USE, with lower recovery levels for DMP and DEP in all extracting solutions. Considering that DMP and DEP present high vapor pressures, they were probably lost by volatilization during the MAE procedure.  In addition, some compounds presented recovery levels higher than the accepted range (50-120% for complex environmental samples), which is possibly due to the baseline signal increase for alkylphenols. The best results from MAE were reached with the ACN/DCM 1:3 (v v -1 ) mix with heating at 60 °C, in which lower losses of the most volatile elements were observed.
In general, USE ( Figure 3) presented better results than MAE, principally for DMP and DEP in almost every extracting solution. Among the extracting solutions evaluated, ACN/DCM 3:1 (v v -1 ) presented the best results for EDCs in SPM. It is probably because DCM presents good efficiency, dissolving non-polar species, whereas ACN provokes the decrease of the extracting mix vapor pressure and favors the dissolution of intermediary polarity compounds. Regarding the sonication time, there are no significant differences between the tests, with good results at 10 min.
Considering the results presented in Figures 2 and 3, the optimal conditions for the SPM sample preparation were 7 mL ACN/DCM 3:1 (v v -1 ) under sonication (120 W) for 10 min. Next, the extracts were centrifuged at 2000 rpm for 10 min, and then they were cleaned up through Florisil cartridges. Next, the extracts were filtered through Millex units, gently dried under N 2 stream, and finally resuspended into 250 µL MeOH. The final extracts were injected into the UFLC-PDA-FLD system afterward ( Figure S2). Once having established the optimized sample preparation method for SPM, we applied the same procedure for the extraction of EDCs from sediments because SPM and sediment generally have similar compositions and will provide similar responses to the extraction processes. 67

Chromatographic method
Firstly, we tried to develop an eluent gradient program that could satisfactorily separate the 14 EDCs in a single run with the best possible resolution. As a starting point, we considered the chromatographic conditions from Lisboa et al. 10 since they also used an ultrafast liquid chromatograph, although in their study phthalates were not considered (only BPA, alkylphenols, and steroid hormones). However, considering in the present work we used a UFLC system with different specifications from Lisboa et al. 10 we could not repeat their separation conditions. Therefore, we optimized the gradient programs presented here. Initially, we used an eluent variation from 0% ACN to 100% ACN with acidified water (2% acetic acid) by 20 min. We separated the steroid hormones (E1, E2, E3, and EE2), BPA, and alkylphenols (4nOP, 4tOP, and 4nNP) with fluorescence within this program detection and phthalates by PDA detection. By injecting individual standards, we identified their elution order and retention times. From modifications in the elution gradients, we could separate the 14 EDCs within 16 min. However, this elution gradient induced too much oscillation in the chromatogram baseline from the fluorescence detector interfering with some alkylphenols' detection. In order to overcome this situation, we decided to do two separated elution programming to enable good separations of all analytes, as stated in detail in Table S2 (SI section).
The optimization of the eluent programming, the flow rate, and the column temperature for each method was done univariately. For both methods, the chosen eluents were ACN and acidified ultrapure water (2% acetic acid) at 0.4 mL min -1 , and 60 °C as the column temperature. Method 1 is an isocratic eluent programming at 39% ACN for 6.5 min total runtime for analysis of BPA and hormones, which are detected by fluorescence (λ exc = 208 nm and λ em = 306 nm). Method 2 is a gradient eluent programming, which starts at 66% ACN and it is kept by 3 min, then it rises from 66-71% ACN until 3.8 min, then it rises again from 71 to 100% ACN during 1 min, 100% ACN is kept during 2.2 min, followed by a reduction from 100 to 66% ACN during 0.5 min. The total runtime was 10.5 min. Alkylphenols' detection was done by fluorescence (λ exc = 208 nm and λ em = 306 nm) and phthalates detection was done by PDA (λ max = 245 nm). In Figure 4, there are standard solution chromatograms for EDCs (50 µg L -1 alkylphenols, BPA, E2, E3, and EE2; 250 µg L -1 for E1; and 10 mg L -1 for phthalates). For the method 1, the elution order is E3 (t R = 1.45 min), BPA (t R = 3.54 min), E2 (t R = 4.49 min), EE2 (t R = 5.76 min), and E1 (t R = 6.17 min). In the method 2 the elution order is DMP (t R = 1.21 min), DEP (t R = 1.65 min), 4tOP (t R = 4.04 min), BBP (t R = 4.19 min), DBP (t R = 4.72 min), 4nOP (t R = 6.18 min), 4nNP (t R = 8.15 min), DEHP (t R = 10.0 min), and DnOP (t R = 10.2 min). The elution order and the respective retention time were obtained by injecting individual standards. Both methods show good peak separations with no co-elutions. Even though we needed two chromatographic methods, we could observe that these optimized methods were efficient to separate and detect the 14 EDCs by UFLC-PDA-FLD. The main advantages of these methods are (i) no derivatization steps for determinations of BPA, hormones, and alkylphenols (which could reflect in losses and LOD/LOQ increase) when the analysis method would be GC-MS, (ii) reduction of sample contamination risk of phthalates from septa and other connections from the GC-MS instrumentation, (iii) shorter analysis time when compared to conventional LC methods, 62,68-71 and (iv) drastic reduction of eluents and other solvents during analysis, which generates a much smaller amount of waste at the end.

Method validation
Method validation was carried out by following criteria suggested by IUPAC [48][49][50][51] and also found in Domingos et al., 53 Lisboa et al., 10 Santos et al., 54 and Nascimento et al. [55][56][57] In this way, we validated the proposed methods regarding calibration curve, limit of detection (LOD), limit of quantification (LOQ), linear range, linearity, accuracy, selectivity, matrix effect evaluation, precision, and tests with real samples. In addition, to assess selectivity of the chromatographic methods we investigated if there were interfering peaks around the analyte peaks when comparing a blank environmental sample chromatogram (a sample absent of analytes) with a standard-enriched environmental sample chromatogram. [48][49][50][51]54,[72][73][74] Figure S3 (SI section) shows comparisons between chromatograms of coastal water samples absent of analytes and standard-enriched. In turn, Figures S4-S7 (SI section) show comparisons among chromatograms of SPM and sediment samples absent of analytes and standard-enriched SPM and sediment samples. Figures S3-S7 show some ghost peaks in the chromatograms, but none of them interfere with the analytes' peaks detection. Therefore, the proposed methods are considered to be selective for the studied EDCs.
Linear regression equations were obtained by external calibration curves, and linearity was assessed based on the coefficient of determination, R 2 . In turn, the linear range was considered to be ranged from the respective LOD for each compound to the highest injected standard concentration. As a result, a good linear fit was observed for the studied EDCs over the full range of the calibration curves, with R 2 > 0.995 for all species (Table 2). Linearity and linear range of calibration curves have met the minimum criteria established by IUPAC for obtaining suitable adjustments and reliable results.
The LOD and the LOQ were calculated considering the data of the calibration curves. 52,[53][54][55][56][57]73,74 Briefly, LOD = 3 × (s/a) and LOQ = 10 × (s/a), where "s" is the linear coefficient standard deviation and "a" is the angular coefficient from the respective analytical curve for a given compound. Then, the LOD and LOQ for the water were recalculated considering the nominal volume of 1.0 L, including the 500 µL concentration step during sample preparation, as shown in Table S2 (SI section). Accordingly, LOD and LOQ for solid samples were calculated by considering an average 1 g dw sediment mass to 500 µL of final extraction (Table 2). Method LOD values varied from 0.412 ng L -1 (4tOP) to 63.2 ng L -1 (DBP) for seawater and from 0.412 to 63.2 ng g -1 dw for solid samples, respectively, to 4tOP and DBP. These LODs and LOQs for the determination of 14 EDCs in environmental samples are considered adequate since they are similar to those found in the literature (Tables S3-S4, SI section).
The sample matrix effect was evaluated by comparing the angular coefficient (a) from two linear regression curves made with and without standard additions. One set of the curves was made with 6 standard concentration levels added either to coastal water sample (a standard + water ), to SPM sample matrix (a standard + SPM ), or sediment sample matrix (a standard + sediment ). The second curve was done with standards dissolved in methanol (a standard ) and directly injected into the UFLC-PDA-FLD system. If there is no significant sample matrix effect in the analysis, the ratio a standard + sample /a standard should be close to 1, and there would not need to make quantification by standard addition calibration curve, which is difficult and time-consuming. As presented in Table S5 (SI section), among all samples, some compounds (for Table 2.  example, EE2, 4tOP, and 4nOP) presented values that differed slightly from 1, indicating that proposed methods are virtually free of significative sample matrix effects and, therefore, standard-addition calibration curves are not needed. Method precision was assessed in terms of retention time and peak area repeatability. Precision was evaluated by injecting standards at three concentration levels (10 replicates each) within the same day (intra-day precision or repeatability) and in 5 successive days (inter-day precision or intermediate precision), expressed as RSD (%) ( Table S6, SI section). RSD values for both inter-day and intra-day precision were below 6.7% for peak area and below 0.31% for retention time. As RSD values up to 20% for instrumental analysis are generally well accepted, the proposed method is precise. Since there are no commercially available standard reference materials (SRM) for endocrine disrupting-chemicals in any sample matrices, we evaluated method accuracy by performing recovery tests. We added a known amount (3 concentration levels) of EDC standards to coastal waters, SPM, and sediment samples for the recovery tests. Standard additions were (A) 10 µg L -1 (alkylphenols, BPA, hormones), 50 µg L -1 estrone, and 2 mg L -1 phthalates; (B) 25 µg L -1 (alkylphenols, BPA, hormones), 125 µg L -1 estrone, and 5 mg L -1 phthalates; and (C) 50 µg L -1 (alkylphenols, BPA, hormones), 250 µg L -1 estrone, and 10 mg L -1 phthalates. After the standard addition, samples were preconcentrated and analyzed as described previously ( Figure S2, SI section).
The last step of method validation is to apply the new analytical methods to the samples collected at Todos os Santos Bay. Results for SPM and waters are presented in Table 3, whereas results for sediments are presented in Table 4. Each sample was analyzed in triplicates. From the list of the studied EDCs, the only species not detected in any sample was 4nOP. All the other species were detected at least once. Hormones were detected at low concentrations within the analyzed samples. Considering E1, E2, and E3 are natural steroid hormones released by humans via domestic effluents, they may have been degraded, since emission, by microorganisms. 3,6,7,35,75 The BPA, 4nNP, DMP, DEHP, and DnOP were the most frequently detected compounds among the evaluated samples. Figure S8 (SI section) shows chromatograms of coastal waters, SPM, and sediment samples collected in the BTS.
The determination methods proposed in this study present good improvements if compared to other studies.   First of all, without using mass spectrometry detectors (MS or MS/MS) coupled to liquid chromatographs, we could reach LODs and LOQs either as low as or lower than LODs from other studies in the literature, as stated in Tables S3  and S4, and references therein. Our LODs/LOQs were low enough to achieve the detection of the 14 analytes at typical environmental levels of coastal environments. Hence our method using UFLC-PDA-FLD is adequate to be employed in environmental monitoring programs.
As for the results presented for the dissolved fraction, there is a sample of the three collected at each point. We chose not to analyze the field replicates because they have low variability. 10 For the sediment and MPS, three bench replicates were made that demonstrated low variability (RSD < 15% for MPS and RSD < 10% for sediment). The set formed by extraction and determination methods guarantees a quick and efficient analysis of the EDCs in three important environmental sample matrices. With this, it is possible to measure the levels of contamination that these products present in coastal environments and gain insight into the processes that control their distribution and fate.

Conclusions
In order to assess the occurrence and dynamics of EDCs in coastal environments, analytical methodologies were developed for sample preparation (extraction, clean up, and pre-concentration) of water, sediment, and SPM, as well as chromatographic analysis by a UFLC-DAD-FLD system. Those methods approach the Principles of the Green Analytical Chemistry. In addition, we pursued procedures that use reduced sample size, fast chromatographic analyses, and being reliable and cost-effective. In this way, the proposed methods could be eligible as a substitute to conventional, time-consuming methods (such as Soxhlet and liquid-liquid (LLE) extractions) in routine analysis or standardized protocols in the future.

Supplementary Information
Supplementary data are available free of charge at http://jbcs.sbq.org.br as PDF file.

Author Contributions
Gabriel C. de Souza was responsible for conceptualization, data curation, investigation, software, validation, visualization, writing original draft, writing-review and editing; Cristiane S. Fahning for conceptualization, investigation, software, validation, visualization; Vanessa Hatje for formal analysis funding acquisition, project administration, resources, visualization, writing original draft, writing-review and editing; Gisele O. da Rocha for conceptualization, data curation, formal analysis funding acquisition, project administration, resources, visualization, writing original draft, writing-review and editing.