Preliminary Investigation into Quantitation of Pharmaceuticals in Lake Victoria Sediments : Development of a Method for Analysis of 11 Pharmaceuticals

(1)

Linköping University | The Department of Physics, Chemistry and Biology Master Thesis, 30 hp | Programme Field of Study: Physics/Chemistry/Biology Spring Semester 2021

Preliminary Investigation into

Quantitation of Pharmaceuticals in

Lake Victoria Sediments

Development of a Method for Analysis of 11

Pharmaceuticals

Robert Lundberg

Examiner, Henrik Kylin Supervisor, Joyanto Routh

(2)

Abstract

Although Lake Victoria is threatened by pollution there is a lack of knowledge about pharmaceuticals contaminants drained into the lake from large cities bordering the lake. Hence, the purpose of this project was to develop, validate and apply a method for analysis of pharmaceutical compounds accumulating in the Lake Victoria sediments. A simple quantitative method for 11 pharmaceuticals combining accelerated solvent extraction, solid phase extraction, trimethylsilylation derivatization, and gas chromatography mass spectrometry was developed, partly validated, and applied to 18 surface sediments and a sediment core dated using the 210_{Pb method. The results showed the} presence of the pharmaceuticals estriol, gemfibrozil, metoprolol, ketoprofen, naproxen, 17α-ethinylestradiol and estrone concentrated around the regions Napoleon Gulf and Thurston Bay with accumulation rates decreasing towards the top of the sediment core. Nonetheless, a randomness in the distribution of these compounds behooves a systematic assessment investigating not only the provenance of these compounds but also further investigations to errors meaning that this study should be treated as a preliminary investigation.

(3)

Abbreviations

210_Pb sup Supported 210Pb 210_Pb xs Excess 210Pb AR Absolute recovery ACN Acetonitrile

AFRs Alternative flame retardants AMIX Analyte standard mixture ASE Accelerated solvent extraction

BSTFA N, O-bis(trimethylsilyl) trifluoroacetamide CE Collision energy

CF-CS Constant flux - constant sedimentation CIC Constant initial concentration

CID Collision-induced dissociation CRS Constant rate of supply

DCM Dichloromethane

EMHSLU Essential medicines and health supplies list for Uganda EtOAc Ethyl acetate

EE Extraction efficiency

GC Gas Chromatography

GC-MS Gas chromatography – mass spectrometry

GC-MS/MS Gas chromatography – tandem mass spectrometry HCl Hydrochloric acid

HILIC Hydrophilic interaction liquid chromatography

HPLC-MS/MS High performance liquid chromatography-tandem mass spectrometry HPU High-power ultrasound

LC Liquid chromatography LOD Limit of detection LOQ Limit of quantitation

LTM Low thermal mass technology m/z Mass-to-charge ratio

(4)

MeOH Methanol

MTBSTFA tert-butyldimethylsilyl-N-methyltrifluoroacetamide MS1 A first mass analyzer

MS2 A second mass analyzer MS/MS Tandem mass spectrometry OFN Octafluoronaphthalene

PAHs Polycyclic aromatic hydrocarbons PBDEs Polybrominated diphenyl ethers PCBs Polychlorinated biphenyls PCPs Personal care products PFAs Perfluoroalkyl substances

PIE Pharmaceuticals in the environment POPs Persistent organic pollutants

PSA Primary secondary amine PSD Particle size distribution RSD Relative standard deviation SINT Internal standard

SINT, DIL Diluted internal standard SREC Recovery standard

SREC, DIL Diluted Recovery Standard SIM Selected ion monitoring SEM Scanning electron microscopy SPE Solid-phase extraction STPs Sewage treatment plants TBDMS tert-butyldimethylsilyl TMCS Trimethylchlorosilane TMS Trimethylsilyl

UAE Ultrasonic assisted extraction

(5)

1. Introduction

Pharmaceuticals are essential for treating diseases both in humans and in the production of animal products (Azzouz & Ballesteros, 2012). The increasing population, particularly senior- and middle-class citizens, and the growth of the pharmaceutical industry can be directly correlated highlighting the symbiotic relationship that has evolved in the past years (Kumirska, et al., 2019). It was realized first in the 1980s, due to technological developments, that trace levels of pharmaceuticals were ending up in aquatic environments (Nantaba, et al., 2020; Sadkowska, et al., 2017). Consequently, there is an increasing scientific and societal interest in pharmaceuticals in the environment (PIE) (Nantaba, et al., 2020; Sadkowska, et al., 2017; Yu & Wu, 2012).

Sources of PIE range from release during manufacturing, distribution, and disposal of

pharmaceuticals along with sewage. Additional sources are from farm animals and manure which add to the growing pool. (Azzouz & Ballesteros, 2012; Kumirska, et al., 2019) Discarded and excreted pharmaceuticals reach the aquatic environment through sewage treatment plants (STPs) which are therefore the primary source of PIE. Therefore, trace analysis of PIE can be used to evaluate if more purification steps are needed in STPs to meet the safe drinking water standards. (Nantaba, et al., 2020; Yu & Wu, 2012)

1.1. Current State of Knowledge

Industrial sources of pharmaceuticals are plentiful around Lake Victoria. Although Africa has seen a growth in pharmaceutical use like the rest of the world, the knowledge about PIE and their potential threat from exposure is not well known. (Nantaba, et al., 2020) A testament to how threatened Lake Victoria is by pollution is all the types of pollutants besides pharmaceuticals that have been detected near it. The atmosphere above Lake Victoria is polluted by polycyclic aromatic hydrocarbons (PAHs) (Arinaitwe, et al., 2012), polybrominated diphenyl ethers (PBDEs), alternative flame retardants (AFRs) (Arinaitwe, et al., 2014), polychlorinated biphenyls (PCBs) (Arinaitwe, et al., 2018) and pesticides (Arinaitwe, et al., 2016). Lake Victoria water is polluted by persistent organic pollutants (POPs) (Kandie, et al., 2020), personal care products (PCPs), pesticides (Arinaitwe, et al., 2020) and perfluoroalkyl substances (PFAs) (Arinaitwe, et al., 2021). Fish in Lake Victoria are polluted by PFAs and trace metals (Arinaitwe, et al., 2020). Sediments from Lake Victoria are contaminated by POPs (Arinaitwe, et al., 2016), PAHs (Kerebba, et al., 2017), trace metals (Ribbe, et al., 2021), and pharmaceuticals (Nantaba, et al., 2020).

1.1.1. Pharmaceuticals Found in Lake Sediments

Pharmaceuticals generally equilibria slightly pushed from the water to the sediment (Fairbairn, et al., 2015). Most common pharmaceutical classes can be found in the environment since antibiotics, analgesics, anti-inflammatories, lipid regulators, β-blockers, antiepileptics, contraceptives, steroids and related hormones have been detected in the environment (Azzouz & Ballesteros, 2012; Nantaba, et al., 2020).

Antibiotics that have been detected in sediments are triclosan, chloramphenicol, and florfenicol. Analgesic and anti-inflammatory drugs that have been detected in water sediments are,

acetaminophen, acetylsalicylic acid, diclofenac, diflunisal, flunixin, flurbiprofen, ibuprofen,

ketoprofen, naproxen, niflumic acid and salicylic acid. Lipid regulators and β-blockers that have been detected in water sediments are gemfibrozil, metoprolol, and propranolol. Steroids and related hormones that have been detected in water sediments are 17α-ethynylestradiol, 17β-estradiol, diethylstilbestrol, estriol, and estrone. Additionally, antiparasitic pyrimethamine, anticonvulsant carbamazepine, and antidepressant clomipramine have also been found in aquatic sediments. (Azzouz & Ballesteros, 2012; Kumirska, et al., 2019; Sadkowska, et al., 2017; Yu & Wu, 2012)

(8)

2

Although pharmaceutical contaminants in sediment have not been reported in Lake Victoria, one study has detected a few pharmaceutical classes in water samples from Lake Victoria but with a focus on antibiotics. Antibiotics detected in Lake Victoria are trimethoprim, sulfamethoxazole,

sulfamethazine, sulfacetamide, oxytetracycline, tetracycline, erythromycin, azithromycin, ciprofloxacin, levofloxacin, norfloxacin and enoxacin with sulfamethoxazole, trimethoprim, and oxytetracycline being the most prevalent. Analgesic and anti-inflammatory drugs detected in Lake Victoria are acetaminophen, diclofenac, and ibuprofen with ibuprofen being the most prevalent. The anticonvulsant carbamazepine and the β-blockers atenol and metoprolol were also detected in Lake Victoria water, with atenolol being the more prevalent β-blocker. The Lake Victoria regions

Murchison Bay, Thurston Bay, Waiya Bay, and Napoleon Gulf showed a decreasing amount of PIE in that order. The most threatening of the detected pharmaceuticals were sulfamethoxazole followed by oxytetracycline, erythromycin, and diclofenac based on the exposure and toxicity effects of these compounds. (Nantaba, et al., 2020)

Only some of these results are following the essential medicines and health supplies list for Uganda (EMHSLU), which lists oxytetracycline, tetracycline, erythromycin, azithromycin, ciprofloxacin, levofloxacin, norfloxacin, acetaminophen, diclofenac, ibuprofen, carbamazepine, and atenolol. Of the other compounds detected in sediments worldwide EMHSLU also lists chloramphenicol, acetylsalicylic acid, salicylic acid, propranolol, pyrimethamine, and carbamazepine. (Ministry of Health, 2016). Nonetheless, EMHSLU does not list sulfamethoxazole (Ministry of Health, 2016) which, as far as known, is the most threatening PIE in Lake Victoria (Nantaba, et al., 2020).

1.2. Current Technical Knowledge of Extraction and Detection

Constantly evolving, the more than 50-year-old analytical techniques gas chromatography (GC) and liquid chromatography (LC) are still state-of-the-art of analytical tools for detecting specific

compounds in complex mixtures. This encompasses, for example monitoring and routine analysis of biological, industrial, and environmental samples. (Sadkowska, et al., 2017)

1.2.1. Analysis of PIE

PIE is most often analyzed with high performance liquid chromatography-tandem mass spectrometry (HPLC-MS/MS) because of its compatibility with the analytes (Azzouz & Ballesteros, 2012; Kumirska, et al., 2019; Sadkowska, et al., 2017; Yu & Wu, 2012). Although gas chromatography-tandem mass spectrometry (GC-MS/MS) requires derivatization of the analytes to make them more volatile, it is preferred since it is more common, cheap, sensitive, specific (Azzouz & Ballesteros, 2012; Kumirska, et al., 2019), resistant to matrix effects and environmentally friendly (Sadkowska, et al., 2017; Yu & Wu, 2012).

Extraction techniques such as Soxhlet extraction, pressurized liquid extraction, and microwave-assisted extraction (MAE) can be used for solid samples together with clean-up through solid-phase extraction (SPE) to analyze samples with complex matrices such as soil and sediment samples (Azzouz & Ballesteros, 2012). A typical combination is MAE together with SPE using either a reversed-phase Oasis-HLB sorbent or a normal-phase silica gel sorbent (Azzouz & Ballesteros, 2012; Kumirska, et al., 2019; Yu & Wu, 2012).

1.3. Aims and Objectives

There are various of PIE near Lake Victoria (Nantaba, et al., 2020). Given the potential toxicity of these compounds, their accumulation in sediments and water column poses a challenge to people. Therefore, there is an urgent need to trace PIE distribution in the lake to take appropriate measures before treated water from industries and sewage is discharged into the lake. Much new knowledge can be gained about Lake Victoria sediments and potentially threatening PIE in Lake Victoria by

(9)

3

analyzing pharmaceuticals in Lake Victoria sediments. Furthermore, there is a need to develop trace analysis methods for detecting PIE and a need for such a method to be validated (Azzouz &

Ballesteros, 2012). There is great potential for GC-MS/MS as an instrument for this purpose (Azzouz & Ballesteros, 2012; Kumirska, et al., 2019; Sadkowska, et al., 2017) such as the GC-MS/MS

instrument at the environmental research platform TEMAM where this project was done. Therefore, this work is aimed to develop, validate, and apply a method for GC-MS/MS analysis of common pharmaceuticals in sediment samples from Lake Victoria. Additionally, this data would then be correlated with age from 210_{Pb dating.}

1.3.1. Previous Studies

Since the purpose of this project was to develop a method for GC-MS/MS analysis, mainly the few studies that analyzed pharmaceuticals with gas chromatography – mass spectrometry (GC-MS) were used as a background.

1.3.1.1. Development of Methods for Analysis of Pharmaceuticals in Sediment Samples

Before GC-MS/MS analysis samples have previously been extracted using MAE (Azzouz & Ballesteros, 2012), the quick, easy, cheap, effective, rugged, and safe (QuEChERS) procedure (Kumirska, et al., 2019), or ultrasonic-assisted extraction (UAE) (Kumirska, et al., 2019; Yu & Wu, 2012). Clean-up was then done with solid-phase extraction (SPE) using Oasis-HLB sorbent (Azzouz & Ballesteros, 2012; Kumirska, et al., 2019), silica gel (Kumirska, et al., 2019) or envi-carb cartridges (Yu & Wu, 2012). Then the sample preparation was finished with derivatization using the reagent N,

O-Bis(trimethylsilyl)trifluoro-acetamide with 1% trimethyl trichlorosilane (BSTFA +1% TMCS) (Azzouz & Ballesteros, 2012; Kumirska, et al., 2019) or tert-butyldimethylsilyl-N-methyltrifluoroacetamide (MTBSTFA) (Yu & Wu, 2012). Prepared samples were then analyzed with GC-MS in selected ion monitoring (SIM) mode using triphenyl phosphate as an internal standard (Azzouz & Ballesteros, 2012), 2-methylanthracene as an internal standard (Kumirska, et al., 2019) or radiolabeled analytes (Yu & Wu, 2012) as surrogate standards.

1.3.1.2. Validation and Application of Developed Methods for Analysis of Sedimental Pharmaceuticals Mostly similar validation parameters have been used to validate previous methods for GC-MS/MS methods for the analysis of sedimental pharmaceuticals; limit of detection (LOD) linearity range, correlation coefficient (R2_{), intra-day precision (relative standard deviation (RSD)), and recovery} (Azzouz & Ballesteros, 2012; Kumirska, et al., 2019; Yu & Wu, 2012). In a couple of cases the inter-day precision was evaluated as well (Azzouz & Ballesteros, 2012; Yu & Wu, 2012). In another case, the intra-day accuracy was evaluated (Kumirska, et al., 2019). Other studies focused on the extraction efficiency (EE) besides absolute recovery (AR) (Kumirska, et al., 2019), and the limit of quantitation (LOQ) (Kumirska, et al., 2019; Yu & Wu, 2012).

Using MAE, SPE, BSTFA + 1% TMCS and GC-MS gave 91-101% recoveries, a LOD between 0.8-5.1 ng/kg, linearity ranges between 3.1-20 000 ng/kg, correlation coefficients between 0.99-1, intra-day precisions between 4.8-6.9% and inter-day precision between 4.2-6.0% (Azzouz & Ballesteros, 2012). Absolute recoveries showed that UAE extraction was better than QuEChERS, and extraction

efficiencies showed that silica gel was better than the Oasis-HLB sorbent. Hence, the final method for further validation, in that case, was the UAE-silica gel column-GC-MS(SIM) method. The UAE-silica gel column-GC-MS(SIM) with BSTFA + 1% TMCS derivatization gave recoveries between 52-113%, a LOQ between 1.0-5.0 ng/g, a LOD between 0.3-1.0 ng/g, linearity ranges between 1-1200 ng/g, correlation coefficients between 0.99-1, intra-day precisions between 0.8-10% and intra-day accuracies from 81-120%. (Kumirska, et al., 2019) Using UAE, Envi-carb SPE, MTBSTFA derivatization and GC-MS gave recoveries between 58-100%, a LOQ between 4.7-39 ng/g, a LOD between 1.4-11 ng/g, linearity ranges between 2.0-2000 pg, correlation coefficients between 0.98-0.99, intra-day precisions

(10)

4

between 3.7-8.5% and inter-day precision between 4.3-13% (Yu & Wu, 2012). Analysis of PIE gave concentrations in the range of 8.5-3100 ng/kg (Azzouz & Ballesteros, 2012) 1.9-7 ng/g (Kumirska, et al., 2019), 1.0-9.0 ng/g (Kumirska, et al., 2013) and 11-2400 ng/g (Yu & Wu, 2012).

For derivatization, a thorough comparison between acylation, sequential derivatization, cyclized silyl derivatization, tert-butyldimethylsilyl (TBDMS) derivatization and trimethylsilyl (TMS) derivatization showed that trimethylsilylation using the reagent BSTFA + 1% TMCS (Magda, et al., 2011).

Furthermore, a chemometric optimization of the derivatization resulted in the optimal parameters BSTFA + 1% TMCS: pyridine: ethyl acetate (EtOAc) (2:1:1, v/v/v) for 30 min at 60 ◦C (Kumirska, et al., 2013).

1.3.1.3. 210_{Pb Dating}

For 210_{Pb dating two different methods have been used; one based on ingrown}222_{Rn (Routh, et al.,} 2004; Routh, et al., 2007) and one based on 209_{Po added as a tracer (Choudhary, et al., 2009;}

Choudhary, et al., 2009). For the 222_{Rn method a Ge-well detector coupled to InSpector was used to} determine total 210_{Pb and parent supported levels of}210_{Pb by measuring the 46.5 keV gamma peak} and the 214_{Pb (352 keV) and}214_{Bi (609 keV) peaks respectively (Routh, et al., 2004; Routh, et al.,} 2007).

The more complicated 209_{Po method required leaching with aqua regia after}209_{Po was added.} Leached samples were filtered and mixed with HCl followed by polonium nucleotide deposition on copper disks using ascorbic acid and alpha counting. Supported and excess 210_{Pb was calculated from} the asymptote and the difference between the total and supported 210_{Pb, respectively. (Choudhary,} et al., 2009; Choudhary, et al., 2009) A variation of the 209_{Po method did not use aqua regia but used} additional treatment with an oxidizing agent before deposition (Das, et al., 2008; Minu, et al., 2018; Ranjan, et al., 2011).

1.3.2. The Strategy of This Project

Because of the analytical equipment and standards available at TEMAM, the developed method consisted of extraction using an accelerated solvent extraction (ASE), followed by silica gel electrophoresis (Kumirska, et al., 2019), BSTFA + TMS Derivatization (Azzouz & Ballesteros, 2012; Kumirska, et al., 2013; Kumirska, et al., 2019) and GC-MS/MS analysis (Azzouz & Ballesteros, 2012; Kumirska, et al., 2019; Yu & Wu, 2012).

The 11 commonly found pharmaceuticals 17α-ethynylestradiol, acetaminophen, diclofenac, estriol, estrone, gemfibrozil, ibuprofen, ketoprofen, metoprolol, naproxen, and β-Estradiol were targeted with 2-methylanthracene and triphenyl phosphate used as an internal standard and a recovery standard, respectively. These 11 pharmaceuticals cover four classes of compounds, namely analgesic and anti-inflammatory drugs, lipid regulators and β-blockers, and steroids and related hormones. Due to the similar analytes, derivatization reagents and instrument used in a previous study, a previously optimized temperature program (Kumirska, et al., 2019) and derivatization procedure (Kumirska, et al., 2013) were used. Since a method for total extraction on the ASE instrument had already been optimized, what remained was to optimize the evaporation programs, the mass spectrometry parameters, establish a calibration curve and finally to validate and apply the method. Calibration curve samples were made to cover the concentration range that pharmaceuticals in sediments show in previous studies (Azzouz & Ballesteros, 2012; Kumirska, et al., 2013; Kumirska, et al., 2019). Validation parameters used in previous studies (Kumirska, et al., 2019; Yu & Wu, 2012) were mostly similar and therefore the same simple parameters were used in this project. For the application of the method, 22 surface sediment samples and one sediment core from the Lake

(11)

5

Victoria locations were analyzed (Table B – 1, Table B – 2, and Figure B – 1 in Appendix B). By dating the different depths of the sediment core with 210_{Pb analysis, both a spatial and temporal distribution} of pharmaceutical concentrations was obtained across Lake Victoria.

1.3.2.1. Delimitations

Approximately 3000 unique pharmaceutical compounds are used in Europe (Kumirska, et al., 2019). There are more pharmaceuticals in the EMHSLU list (Ministry of Health, 2016) that has been detected in water sediments around the world (Azzouz & Ballesteros, 2012; Kumirska, et al., 2019; Sadkowska, et al., 2017; Yu & Wu, 2012). However, only selected compounds were chosen, representing some commonly used pharmaceuticals. For practical reasons, only every other depth sample in the sediment core was analyzed and analyses were only done in singlicates, lowering the precision and accuracy of the results. Additionally, inter-day accuracies and precisions were not examined and only a three-point calibration curve was used within a narrow concentration range. The cost of deuterated standards led to the use of a cheaper recovery standard that does not compromise the derivatization process and provides a consistent output.

1.3.3. Method Description

Even though the utilized methodology involved ASE, SPE, BSTFA + 1% TMCS silylation and GC-MS/MS techniques a few previously used methods are also described shortly. Nonetheless, the utilized methodologies are described more thoroughly in the theory section.

1.3.3.1. MAE

Digestion of complex matrices can be done using high temperatures, high pressures, concentrated acids, and oxidation reactions so that the matrix is completely dissolved into a liquid. MAE uses partial digestion to extract the compounds of interest from the matrix. Automatically handling many samples with little consumption of solvents makes MAE a green extraction method. However, the simple microwave radiation requires extensive grinding to fine particles beforehand. (Duarte, et al., 2013)

1.3.3.2. QuEChERS

QuEChERS is a standard extraction method for complex matrices in fields ranging from pesticide and pharmaceutical to mycotoxin and PAH analysis. The QuEChERS procedure consists of homogenizing the complex matrix with a blender, cleaning it with MgSO4 d-SPE in acetonitrile (ACN) using vortexing and centrifuging, and then mixing the supernatant with primary secondary amine (PSA) to prevent polar interferences. The primary and secondary amine functional groups of PSA can be seen in Figure 1. (González-Curbelo, et al., 2015; Harris & Lucy, 2016) As opposed to the benefits its name implies, QuEChERS suffers from poor recoveries (Lehotay, et al., 2005).

Figure 1 The sorbent primary secondary amine (PSA) with the primary amine to the right and the secondary amine in the middle.

1.3.3.3. Ultrasonic assisted extraction (UAE)

UAE uses ultrasound, frequencies above the acoustic frequencies heard by humans, to achieve rapid extraction even with soft and energy-saving conditions making it a green extraction method. Not even the applied high-power ultrasound (HPU) with low frequencies between 16-100 kHz has enough power to interact with molecules. Nonetheless, the ultrasound interacts indirectly with molecules by

(12)

6

pressurizing the liquid medium in the water bath and creating microbubbles. The implosion of the microbubbles during compression improves the mass-transfer of compounds from the matrix to the extraction solvent. (Dominguez & Munoz, 2017)

1.3.3.4. ASE

ASE extracts solid samples with the help of different organic solvents by using high pressures and temperatures. After injection of solvents, the pressurized sample cell is moved into a heated oven for extraction, followed by purging of the extraction thimble with a gas pulse. (Galbiati & Teli, 2016; Richter, et al., 1996)

1.3.3.5. Solid Phase Extraction (SPE)

SPE separates the components of a sample through their partitioning between a small amount of stationary phase and a small amount of mobile phase. After conditioning the column with the eluting solvents and loading the sample, solutes are loaded. Solutes are eluted from weak to strongly bound analytes by using stronger solvents, i.e., solvents with more similar interactions to the stationary phase. (Harris & Lucy, 2016)

1.3.3.6. Silylation Derivatization

The most common silylation reagents are BSTFA and MTBSTFA (Schummer, et al., 2009). BSTFA + TMCS is a trimethylsilylation reaction where BSTFA replaces active hydrogen sites, e.g., -OH, -COOH, =NH, -NH, and -SH groups, with TMS groups to make compounds less polar and more volatile. The reaction of BSTFA as a silyl donor is catalyzed by the presence of the weaker silyl donor TMCS. (Blau & Halket, 1993; Knapp, 1979) MTBSTFA derivatization is also a silylation reaction, but as its name suggests it replaces active hydrogens with a TBDMS group instead of the TMS group that BSTFA supplies (Schummer, et al., 2009).

1.3.3.7. GC-MS/MS

In GC-MS/MS gas chromatographic separation is coupled with tandem mass-spectrometric detection. Evaporated analytes are eluted one at a time from the gas chromatograph column, based on

hydrophobicity and volatility, into the detector. In the tandem mass spectrometer, the analytes are ionized after which the characteristic fragments of one of their characteristic fragments are selected for detection. (Sparkman, et al., 2011)

1.3.3.8. Statistical Models

A few quantifiable validation parameters accompany calibration. Accuracy is a measure of the agreement between a measured concentration and the actual concentration of reference material. Precision is a measure of the agreement between the results of replicate measurements. The difference between accuracy and precision is illustrated in Figure 2. The LOD is the lowest

concentration that gives an analytical signal and is easily distinguishable from the background noise. The LOQ is the lowest concentration enabling precise quantitation. Correlation shows how much error there is in the linear regression by how far from 1 the correlation coefficient is. (Miller & Miller, 2011)

(13)

7

Figure 2 Illustration of the difference between accuracy and precision where the true value is in the middle of the target and the black dots represent replicate measurements.

1.3.3.9. 210_{Pb Dating}

210_{Pb dating measures the lead activity of a sample to date the sample based upon the corresponding} sediment accumulation rate (Pittauerová, 2013).

1.4. Ethical Discussion

Ethically designed experiments should apply green chemistry by using routes with safer chemicals, renewable materials, more catalysis, less waste, less derivatization, and less energy consumption (Mulvihill, et al., 2011) The concern of power consumption of analytical instruments has led to the concepts of green LC and GC. Some examples of approaches to achieve greener analyses are reducing power consumption by reducing analysis times with fast GC, using temperature

programming and low thermal mass technology (LTM), and using heated nickel-clad fused columns with more efficient heating and cooling. (Sadkowska, et al., 2017) Another significant ethical concern in chemistry is safety which can be dealt with by assessing the risks of the designed experiments and following standard laboratory protocols. Any inevitable waste is disposed of using proper labeling. (Harris & Lucy, 2016)

1.4.1. Consequences of Environmental Pharmaceutical Contamination

Pharmaceuticals have been shown to negatively affect the internal organs and cell proliferation mechanisms of aquatic organisms (Nantaba, et al., 2020). Even trace levels of PIE are sufficiently active to have environmental effects, such as spreading antibiotic resistance (Azzouz & Ballesteros, 2012; Nantaba, et al., 2020). However, other effects of chronic exposure to trace levels of

pharmaceuticals remain mostly unknown and challenging to study (Yu & Wu, 2012). Nevertheless, its negative effects on humans and animals are very likely (Nantaba, et al., 2020; Yu & Wu, 2012). Due

(14)

8

to these negative consequences from the presence of PIE the development and application of methods for the analysis of PIE is ethical since it serves to prevent PIE.

1.5. Societal Relevance

As the largest freshwater lake in Africa, it is important to minimize any threats to this source of fish and water caused by the industrial, urban, and agricultural growth around it. (Nantaba, et al., 2020) After entering the environment, pharmaceuticals and their metabolites readily accumulate as trace components in soil and sediments (Azzouz & Ballesteros, 2012). Developing and applying a method for PIE analysis in Lake Victoria can provide crucial knowledge (Nantaba, et al., 2020) and validating such methods shows that it is fit for its purpose (Haslam, et al., 2011). Legislation regarding PIE is lacking (Sadkowska, et al., 2017) and the new knowledge can help develop PIE legislation since it is often based on industrial data (Oelkers, 2021).

(15)

9

2. Process

The initial plan of the project was done by making a Gantt chart in Microsoft Excel as a weekly schedule that can be seen in Figure A – 1 in Appendix A. Each part of the project was divided into as many practical tasks as possible to manageably estimate the time required for each task. During the project, the actual time spent each day was distributed between the relevant tasks to evaluate the process continuously. Significant milestones were highlighted in the Gantt chart as key points for systematic evaluation. A systematic evaluation was done during the whole project by planning, executing, analyzing the results of, and moving on from each task if possible and changing the plan otherwise. The project was fulfilled when the laboratory work had been presented in a written report and the written report had been presented orally. The report was meant to be a testament to the capabilities of the utilized analytical instrument and an addition to the knowledge bank of

(16)

10

3. Theory

3.1. Sample Preparation Methods

Sample preparation was done using ASE extraction, silica gel chromatography clean-up (purification) and silylation derivatization. Chemical compounds that are targeted for analysis are called analytes.

3.1.1. ASE

ASE is a method for the automatic extraction of solid samples with organic solvents to transfer the analytes from the solid sample to the solvent. As an example, the ASE 350 utilized in this project can be seen in Figure 3. Using very high pressures and temperatures, ASE can achieve similar recoveries as other state-of-the-art extraction method despite requiring much less time and solvent (Galbiati & Teli, 2016; Richter, et al., 1996). Extraction fluid is injected into a sample cell with a solid sample inside and after a static extraction the extract is expelled with a gas purge.

In general, increasing the temperature to over 100 °C theoretically results in a 10-fold increase of both the solubility and diffusion rates of compounds. Diffusion rates are also increased by the increased mass transfer and concentration gradient caused by the introduction of fresh solvent during a static extraction. The higher

temperature also reduces interference from matrix effects by facilitating disruption of interactions between analytes and the sample matrix. Furthermore, high temperature facilitates penetration of the sample matrix by decreasing the viscosity and surface tension of the solvent. These benefits of higher temperatures can be increased by increasing the temperatures to levels above the boiling point of analytes which is possible because of the high pressure used. Further benefits of the high pressure are reduced matrix effects from forcing the solvents into the sample and improving the solubilization of air bubbles. (Richter, et al., 1996)

3.1.2. Silica Gel Chromatography

Silica gel chromatography is a type of hydrophilic interaction liquid chromatography (HILIC) using silica gel as the stationary phase. Since the interaction is hydrophilic, a normal-phase separation is used to retain polar substances more than nonpolar ones. Normal-phase means a polar stationary phase is used as opposed to reversed-phase which uses a nonpolar stationary phase. Nonpolar substances can easily be washed away before the polar ones are eluted. Analytes are distributed between the stationary and mobile phases depending on the polar, hydrophobic, and ion-exchange interactions that different phases enable. Silica gel is the most common stationary phase in HILIC, especially for pharmaceutical analysis, in which case partitioning, adsorption and ion exchange are the most effective retention mechanisms. (Jandera, 2017)

3.1.3. Derivatization for gas Chromatographic Analysis

Silylation is the most common derivatization reaction used for gas chromatographic analysis. BSTFA in combination with TMCS is the most common trimethylsilylation reagent. Catalyzed by TMCS, BSTFA will efficiently replace all active hydrogens with TMS groups to make polar compounds less polar, more volatile, and more thermally stable. In the presence of the weak silyl donor TMCS, functional groups with active hydrogens will rapidly perform nucleophilic attacks on the silicon atoms of the stronger silyl donor BSTFA. The transition state formed through the nucleophilic attack will expel the active hydrogen bound to a group on the silyl donor to create the silylation product. To complete this reversible reaction, the silyl donor leaving group must be more basic than the

Figure 3 An ASE 350 instrument with a top carousel where the solid samples are placed and a bottom carousel where the extracts are collected.

(17)

11

functional group carrying the active hydrogen. The structures in Figure 4 show how the leaving group, circled in red, is much more basic for BSTFA than for TMCS since the biproduct of silylation with TMCS is the strong acid HCl. Examples of functional groups with active hydrogens are the OH, -COOH, =NH, -NH, and -SH groups. (Blau & Halket, 1993; Knapp, 1979)

Figure 4 Structure of N, O-bis(trimethylsilyl) trifluoroacetamide (BSTFA), to the left, and trimethylchlorosilane (TMCS), to the right, with their leaving groups circled in red.

3.2. Analytical Methods

Samples prepared through extraction and derivatization were analyzed using the MS and GC-MS/MS systems. The 210_{Pb dating method was used for establishing the sediment chronology. In} addition, statistical tools were used for correlating the data.

3.2.1. Gas Chromatography

After injection of a liquid or gaseous sample, GC as a separation method vaporizes its components in a heated injector port. As they are being pushed through a heated column by a carrier gas towards a detector the vaporized analytes are separated based on their different vapor pressures and detected individually. Separation occurs between the stationary phase inside the column and the carrier gas which is the mobile phase. Each compound has a characteristic vapor pressure; the portion of the compound being in a gaseous state at any given time. Detection occurs when compounds eluted from the column enter a detector coupled to the GC. (Harris & Lucy, 2016)

3.2.1.1. Injection

When a sample is injected into the heated injection port using a syringe the sample first reaches a silanized glass liner. The liner filters the sample from decomposition products and debris. After passing through the liner, the sample is injected using a split or the splitless mode to inject only a fraction of or the whole sample, respectively. (Harris & Lucy, 2016) Since this project covers trace analysis, it uses splitless injection.

Unlike a split liner, a splitless liner has no mixing chamber at the bottom to discard part of the sample and delivers 80% of the sample onto the column instead of 0.2-2%. It takes one minute for the sample to leave the injector port because of the gas flow. Therefore, splitless injection uses an injection temperature around 230 °C, approximately 100 °C lower than for split injection, to prevent decomposition. After this injection time, the sample needs to be focused (concentrated in space) to obtain sharp peaks using the techniques solvent trapping and cold trapping. Solvent trapping uses an initial column temperature at least 40 °C below the solvent’s boiling point to condense it and allow analytes to collect in it before starting the separation. Cold trapping uses an initial column

temperature at least 150 °C below the boiling point of the analytes, lowering their vapor pressures enough to collect them before starting the separation. (Harris & Lucy, 2016; Sparkman, et al., 2011)

(18)

12 3.2.1.2. Separation

Separation is most often done on a tubular 15-100 m coiled column consisting of an outer polyimide coated SiO2 wall, an inner 0.1-5 μm thick stationary liquid phase and a 0.1-0.53 mm diameter hollow core. Polysiloxane-based stationary phases are commonly used since these are bound to the columns and crosslinked to prevent degradation. (Harris & Lucy, 2016; Poole, 2012)

Retention on the column depends on the thermodynamics of each analyte. An increased affinity for the stationary phase and a decreased vapor pressure both shifts the equilibrium towards the

stationary phase, increasing the time spent immobilized in the stationary phase. Each analyte spends the same amount of time in the mobile phase, moving at the same speed as the carrier gas, but different time frames in the stationary phase. All analytes are constantly equilibrating between the stationary and mobile phase to satisfy the new equilibria around the analyte band. As vaporized analytes are always pushed forward by the carrier gas the entire analyte band is successively moved forward. (Harris & Lucy, 2016; Sparkman, et al., 2011) The total time an analyte spends in the column, between injection and elution, is called the retention time.

When analyzing a thermodynamically complex mixture with a constant column temperature the so-called general elution problem arises meaning that thermodynamically different analytes cannot be separated and eluted within a reasonable time. Therefore, a temperature program is used to

increase the column temperature and the vapor pressure of the analyte during the chromatography. Thus, each analyte is eluted at its own time, with a suitable temperature and reasonable separation time, as the carrier gas carries it into the detector. (Harris & Lucy, 2016; Sparkman, et al., 2011)

3.2.2. Mass Spectrometry

Mass spectrometry can selectively detect analytes eluted from a chromatographic process. A mass spectrometer ionizes the vaporized analytes as they exit the gas chromatograph, accelerates them, and separates them based on their mass-to-charge ratio (m/z). The mass spectrometric detector then gives a mass spectrum, a plot of abundance against each detected m/z value, for each time point of the chromatography in an ion chromatogram. Thus, mass spectrometers consist of an ion-source, a mass separator or mass analyzer, and a detector. (Harris & Lucy, 2016)

3.2.2.1. Electron Ionization

When coupled to GC, mass spectrometers most often use an electron ionization ion source. Electron ionization uses a hot filament to emit electrons commonly accelerated through 70 V to get an energy of 70 eV when they collide with incoming analytes. The initial collision removes an electron from the analytes, creating unstable radical molecular ions that fragment into characteristic fragments. Several electronic potentials accelerate and focus the ion fragments into the mass analyzer. (Harris & Lucy, 2016; Sparkman, et al., 2011)

3.2.2.2. Mass analyzers

The most common mass analyzer is the transmission quadrupole analyzer (Harris & Lucy, 2016; Sparkman, et al., 2011) originally consisting of four parallel cylindrical rods (Harris & Lucy, 2016). By varying the oscillating electronic fields applied to these rods, different m/z-values can be scanned with only one m/z-value having a stable enough trajectory to pass through the quadrupole without colliding. (Harris & Lucy, 2016; Campbell, 2012)

3.2.2.3. Tandem Mass Spectrometry (MS/MS)

MS/MS couples a first mass analyzer (MS1) with a second mass analyzer (MS2). Often, a collision cell is introduced between the two mass analyzers where the ion fragments are further fragmented through collision-induced dissociation (CID) by collision with an inert gas. (Campbell, 2012)

(19)

13 3.2.2.4. Detection

After being separated with a quadrupole, an electron multiplier detects the ions, to normalize their electrical response, where they release electrons upon collision. The electrons are accelerated towards the end of the electron multiplier with an increasingly positive electric potential. Repeatedly colliding with the wall of the electron multiplier the electrons are multiplied one to a hundred million times to produce the analysis signal. (Harris & Lucy, 2016)

3.2.2.5. Data Acquisition Modes

The standard mass spectrometric data acquisition mode is a full scan mode of a wide range of m/z values, resulting in a total ion chromatogram with the abundance of all the ions. However, SIM that only measures one or a few m/z values is more sensitive and selective because more time is spent measuring analyte-specific m/z values. Multiple reaction monitoring (MRM) transitions double this selectivity by selecting analyte specific ions twice to increase the signal-to-noise ratio even further. (Harris & Lucy, 2016)

3.2.3.

210

_{Pb Dating}

Excess 210_{Pb in sediment dating is based on the three assumptions that the sediment accumulation is} chronologically ordered, the time marker is immobilized and preserved, and the time marker supply rate is constant and known. The uranium isotope 238_{U decays into a series continuing with radium} isotope 226_{Ra, the radon isotope}222_{Rn and the lead isotope}210_{Pb. From an equilibrium with}226_Ra there is 210_{Pb inherent to the sediment material called supported}210_{Pb (}210_{Pbsup) but there is also} unsupported excess 210_{Pb (}210_{Pbxs). Assuming the sedimentation rate and flux of}210_{Pbxs are constant} the 210_{Pbxs concentration decreases exponentially with increasing sediment depth because of}

radioactive decay and can be used to quantify the accumulation rate. Since 210_{Pb chronology gives no} direct sample dating the 210_{Pbxs profile is instead interpreted using constant flux - constant}

sedimentation (CF-CS) fitting or Constant rate of supply (CRS) and constant initial concentration (CIC) algorithms. (Pittauerová, 2013)

The CF-CS model is appropriate for stable environments, giving a 210_{Pbxs concentration that decreases} exponentially with increasing cumulative dry mass, since it assumes constant atmospheric 210_Pbxs deposition and accumulation rates. By fitting the 210_{Pbxs profile to the following equation where C is} the activity in one layer, C (0) is the activity in the surface layer, λ is the 210_{Pb decay constant}

(0.03122 ± 0.00028 yr-1_{), m is the cumulative dry mass per unit area above the layer and r is the mass} accumulation rate (g·cm-2_∙yr-1_{) the age of the layer can be determined by dividing the m-value of the} layer with the r-value. (Pittauerová, 2013)

𝐶 = 𝐶(0) ∙ 𝑒−𝜆𝑚/𝑟

If the 210_{Pbxs flux can be assumed constant, but if the sedimentation rate is not, then the CRS model is} more appropriate than the CF-CS model. Using this model, the age, t, of sediment at the depth m can be with the following equation where λ is the 210_{Pb decay constant (0.03122 ± 0.00028 yr}-1_{), A (0) is} the integral of the activity over the whole sediment profile and A (i) is the integral of the activity below the depth m. However, if the activity does not reach zero at the bottom of the 210_{Pbxs profile,} the integrals will be inaccurate, and the sample ages will be overestimated. (Pittauerová, 2013)

𝑡 =1 𝜆𝑙𝑛(

𝐴 (0) 𝐴 (𝑖))

Finally, assuming the initial activity is constant regardless of the mass accumulation rate , the age, t, of a layer, can be calculated with the following equation where λ is the 210_{Pb decay constant (0.03122}

(20)

14

± 0.00028 yr-1_{), C (0) is the surface layer activity, and C is the activity of the current layer.} (Pittauerová, 2013)

𝑡 =1 𝜆∙ 𝑙𝑛

𝐶 (0) 𝐶

The SI unit for radioactive activity is the Becquerel, Bq, which equals one disintegration per second. Other standard units for radioactive activity are the Curie, Ci, which equals 3.7 · 1010_{Bq and}

disintegrations per minute, dpm, which equals one-sixtieth of a Bq. (Pittauerová, 2013)

3.3. Statistical Models

The more complex statistical tools used are the linear regression, the correlation coefficient, and the standard deviation.

3.3.1. Linear Regression

Assuming a linear relationship between an analytical signal, y, and an analyte concentration, x, a linear regression can be used to fit a straight line to this relationship and define it with an equation of the form y = a + bx. To best account for experimental errors the squares line is used. The least-squares line minimizes the sum of the least-squares y-residuals, the deviations of the calibration points from the linear line along the y-axis. Thus, this model assumes that all errors occur in the y-direction. The intercept of the least-squares line can be calculated by subtracting the slope, b, times the mean value of all the x-values, from the mean value of all the y-values. The slope of the least-squares line, b, can be calculated with the following equations where xi is an individual x-value, x̅ is the mean value of all the x-values, yi is an individual y-value, y̅ is the mean value of each y-value. (Miller & Miller, 2011)

𝑏 =∑ [(𝑥𝑖 𝑖− 𝑥̅)(𝑦𝑖− 𝑦̅)] ∑ (𝑥𝑖 𝑖− 𝑥̅)2

3.3.2. Linear Correlation

To evaluate the errors along the y-direction and how linear the calibration plot is the product-moment correlation coefficient, r, is usually calculated with the following equation. (Miller & Miller, 2011)

𝑟 = ∑ [(𝑥𝑖 𝑖− 𝑥̅)(𝑦𝑖− 𝑦̅)] {[∑ (𝑥𝑖 𝑖− 𝑥̅)2][∑ (𝑦𝑖 𝑖− 𝑦̅)2]}1/2

An r-value of -1, 0, and 1 describes a perfect negative correlation, no correlation, and a perfect positive correlation, respectively. Since values can only be in the range -1 ≤ r ≤ 1, the closer the r-value is to ± 1, the more linear the calibration is. However, Microsoft Excel gives the coefficient of determination, R2_{, which should be as close to 1 as possible and can be calculated with the following} equation where SS is short for the sum of squares and ŷ is the y-value calculated from the same x-value as the corresponding y-x-value using the regression equation. (Miller & Miller, 2011)

𝑅2₌𝑆𝑆 𝑑𝑢𝑒 𝑡𝑜 𝑟𝑒𝑔𝑟𝑒𝑠𝑠𝑖𝑜𝑛

𝑡𝑜𝑡𝑎𝑙 𝑆𝑆 =

∑ (ŷ𝑖 𝑖− 𝑦̅)2

∑ (y𝑖 𝑖− 𝑦̅)2

3.3.3. Relative Standard Deviation

The relative standard deviation (RSD) can compare different types of data series and can be

calculated with the following equation where s is the standard deviation and x̅ is the mean of all the data values. (Miller & Miller, 2011)

(21)

15

𝑅𝑆𝐷 (%) = 100 ∙𝑠 𝑥̅

First, the mean value and the standard deviation can be calculated using the following equations , where n is the number of values. (Miller & Miller, 2011)

𝑥̅ =∑ 𝑥𝑖 𝑖 𝑛

(22)

16

4. Materials and Methods

4.1. Materials

Sediment samples were obtained by drilling in Lake Victoria according to the detailed sample information in Table B – 1, Table B – 2, and Figure B – 1 in Appendix B. The sediment samples consisted of 22 surface sediment samples and 18 samples of different depths from a sediment core. Half of the sediment shipped from Africa were already freeze -dried and the other half had to be freeze-dried. Before use, all the sediment samples were preserved in a freezer container. Dichloromethane (DCM) of LiChroSolv quality from Merck (Darmstadt, Germany) was used for extraction, needle cleaning, quality assurance, and preparation of stock solutions. Ethanol A (96%) from Solveco (Rosersberg, Sweden) was used for rinsing. EtOAc of LiChroSolv quality from Merck (Darmstadt, Germany) was used for silica gel chromatography and derivatization. Hexane of

LiChroSolv quality from Merck (Darmstadt, Germany) was used for silica gel chromatography, quality assurance and rinsing. Methanol (MeOH) of LiChroSolv quality from Merck (Darmstadt, Germany) was used for extraction, silica gel chromatography, stock solution preparation and rinsing. Anhydrous pyridine of 99.8% purity from Merck (Darmstadt, Germany) was used for derivatization.

BSTFA + 1% TMCS of LiChropur quality from Merck (Darmstadt, Germany) was used for GC derivatization. Silica gel 60 from Merck (Darmstadt, Germany) was used for silica gel

chromatography. General-purpose grade Ottawa quartz sand from Fisher (Loughborough, UK) was used as blank for method development, method validation and quality control.

Octafluoronaphthalene (OFN) EI checkout standards of the concentrations 2 and 100 fg/μL were used to evaluate the detection limits of the GC-MS/MS instrument. Nitrogen 5.0 gas from Nippon Gases (Köping, Sweden) was used for extraction.

The utilized pure standards were 17α-ethynylestradiol (≥98%), 2-methtylanthracene (97%),

acetaminophen (98-102%), analytical standard diclofenac sodium salt, estriol (≥97%), estrone (≥99%), analytical standard gemfibrozil, ibuprofen (≥98%), ketoprofen (≥98%), pharmaceutical secondary standard metoprolol tartrate, pharmaceutical secondary standard naproxen, analytical standard triphenyl phosphate and β-Estradiol (≥98%) from Merck (Darmstadt, Germany). Triphenyl phosphate and 2-methylanthracene were also used as a recovery standard and as an internal standard,

respectively. The standards were used to optimize MRM transitions, and the calibration curve.

4.1.1 Stock solutions

Analyte stock solutions (3 g/L) were prepared by dissolving 30 mg of each analyte standard in 10 mL MeOH. An analyte standard mixture, AMIX, (3 ng/μL) was prepared by mixing 10 μL of each analyte stock solution and diluting to 10 mL in MeOH. The stock solutions SREC and SINT were prepared by dissolving 15 mg of the recovery standard and the internal standard respectively in 10 mL DCM. Working stock solutions SREC, DIL and SINT, DIL were prepared by separately diluting 100 μL of SREC and SINT respectively to 10 mL in DCM. A 2 mol/L HCl solution was prepared by diluting 20 mL 37% HCl to 100 mL in a measuring flask.

4.2. Instrumentation

A Mettler PK 4800 balance from Mettler Toledo (Stockholm, Sweden) was used for weighing before and after freeze-drying and a Mettler AE200 balance from Mettler Toledo (Stockholm, Sweden) was used for weighing samples before extraction. Samples were dried using a FreeZone freeze-dryer from Labconco (Kansas City, USA). An RP 1211 MC refrigerator from Electrolux (Linköping, Sweden) was used for storing the samples. MilliQ water was prepared with a Purelab Chorus 1 from Elga Labwater (Paris, France). A TS 8136 drying oven from Termaks (Bergen, Norway) and a MORE

(23)

17

THAN HEAT 30-3000˚C oven from Nabertherm (Lilienthal, Germany) were used for drying and combusting, respectively.

Extraction was done with a Dionex ASE 350 using ASE 350 stainless steel extraction cell bodies, end caps and cellulose extraction filters from Thermo Scientific (Göteborg, Sweden). Evaporation of volumes larger than 4 mL was done with Syncore Platform and accessories connected to a V-700/710 Vacuum Pump and V-850/855 Vacuum Controller from Buchi (Flawil, Switzerland). Silica gel

chromatography was also done with the Syncore Platform, vacuum pump and controller using the Syncore SPE Accessories from Buchi (Flawil, Switzerland).

Replacement Teflon frits from Supelco (Bellefonte, USA) were used to pack silica gel columns. Mixing was done with a REAX 2000 vortex from KEBO-Lab (Neuhausen am Rheinfall, Switzerland).

Centrifugation was done using an Avanti J-E centrifuge from Beckman Coulter (Bromma, Sweden) equipped with a JA-14.50 rotor from Beckman Coulter (Bromma, Sweden). Incubation was done using an AQUAline AL 25 water bath from LAUDA (Burgwedel, Germany).

GC-MS/MS analyses were done with an Intuvo 9000 GC System connected to a 7693 Autosampler and a 7010B GC/MS Triple Quad mass spectrometer with 70 eV electron ionization from Agilent Technologies (Santa Clara, USA). Vacuum for the GC-MS/MS was obtained with a RV5 Rotary Vane Vacuum Pump from Edwards (Lomma, Sweden). GC-MS analyses were done with a 6890N Network GC System connected to a 7683 Series Autosampler and a 5973 inert Mass Selective Detector with 70 eV electron ionization from Agilent Technologies (Santa Clara, USA). Vacuum for the GC-MS was obtained with a 1.5 vacuum pump from Edwards (Lomma, Sweden). Chromatographic separation was done with HP-5MS UI columns (30 m x 0.25 mm ID x 0.25 µm film thickness). 210_{Pb dating was} done with an Ortec dual-alpha spectrometer (Oak Ridge, USA) interfaced with a multi-channel analyzer.

4.3. Sample Preparation

4.3.1 Freeze-Drying

Before extraction, the sediment samples were freeze-dried to be able to put them in the extraction sample cells. Holes were made in the bag of a wet sediment sample and then that bag was placed in another bag and the whole package was weighed. After being frozen overnight, the bags were placed in the freeze drier with the outer bag slightly opened. Freeze -drying was done at high vacuum and – 50 °C for a weekend. The bags along with the sample after freeze-drying and empty bags were weighed after which the wet and dry sample masses, that can be found in Table C – 1 and Table C – 2 in Appendix C, were calculated. Sample masses were calculated simply by subtracting the masses of the empty bags from the masses of the bags with the samples in them.

4.3.2 ASE

Figure 5 shows the bottom, middle and top part of an ASE sample cell. The bottom of a sample cell was screwed on to push down and insert a frit from the top using the included tool. Approximately 3 g of freeze-dried soil was weighed into the sample cell and 1 μl of SREC, DIL was injected onto the soil in the sample cell. The top of the sample cell was screwed on after which the cell was inserted into the holder in the top carousel of the ASE 350 instrument (Figure 3). For each loaded sample cell an extract bottle was placed in the corresponding position in the bottom carousel of the ASE 350 instrument (Figure 3). All loaded sample cells were entered into a sequence in the software Chromeleon 7 followed by extraction with the solvent mixed in the proportion of 9:1 DCM: MeOH. Each extraction was done with an initial rinse of 5 mL solvent followed by two static 30-minute extraction cycles in standard mode using an

Figure 5 An ASE sample cell.

(24)

18

oven temperature of 100 °C, a 60% rinse volume, and a 60 s purge time. Before transferring to 4 mL vials, the extracts were evaporated on the Buchi Synchore instrument using the program 800-650 mbar (4 min); 650 mbar (90 min); 650-200 mbar (4 min); 200 mbar (22 min).

4.3.3. Silica Gel Chromatography

Extracts in 4 mL vials were evaporated until dry using nitrogen gas and redissolved in 1 mL n-hexane. Glass cartridges were loaded with 0,75 g silica gel in-between two Teflon frits and inserted into the top of the Buchi Synchore SPE accessory. A continuous manual pressure of 800 mbar was selected on the vacuum controller. Preconditioning of the silica gel column was done with 4 mL MeOH, 4 mL EtOAC and 6 mL n-hexane in that order. Extracts redissolved in n-hexane were loaded onto the column after which the SPE accessory was only opened to give 1 drop per second. A neutral fraction was eluted with 6 mL n-hexane, after which the Buchi tubes were changed to elute a polar fraction. The polar fraction was eluted with 6 mL EtOAC followed by 6 mL MeOH in the same Buchi tube , and at that point, the silica gel column was opened fully to dry the solvent. Before transferring to 4 mL vials, the neutral and polar fractions were evaporated simultaneously in different Buchi tubes using the program 800-205 mbar (2 min); 205-205 mbar (11 min); 205-150 mbar (2 min); 150 mbar (45 min). Neutral fractions were transferred directly to GC vials while the polar fractions were subjected to derivatization.

4.3.4. Derivatization

Polar fractions were evaporated until dry using nitrogen gas and redissolved in a reaction mixture consisting of 50 μL BSTFA + 1% TMCS, 25 μL anhydrous pyridine, 24 μL EtOAc and 1 μL SINT, DIL. The reaction mixture was vortexed for 1 minute at maximum speed after which the 4 mL vial was placed in a 60 °C water bath for 30 minutes. After cooling to room temperature, the reaction mixture was transferred to a GC vial for analysis.

4.4. Analytical Methods

4.4.1 GC-MS/MS and GC-MS

(25)

19

Table 1 GC-MS/MS and GC-MS parameters used for all analyses.

GC parameters

He quench gas 2.25 mL/min

N2 collision gas 1.5 mL/min

Injection volume 1 μL

Splitless pressure 12.247 psi

Injector temperature 200 °C

Total injector flow 63 mL/min

Septum purge flow Off

Split flow 0 mL/min (0.75 min); 62 mL/m in

Gas saver 0 mL/min (3 min); 15 mL/min

Oven 100 °C (1 min); 4 °C/min; 300 °C (4 min)

Column flow rate He, 1 mL/min

MS/MS and MS parameters

Ion source temperature 230 °C

Quadrupole temperature 150 °C

Ionization type Positive EI (70 eV)

Collision energy (CE) 15 V

Solvent delay 5 min

Peak width 0.8 s

Gain 100

Dwell time 10 ms

For quantification, the developed and validated methods used the MS parameters in Tables 2 and 3. However, a full scan version of this method was also used to get characteristic ions for the analytes. In the full scan method, the MRM transitions were replaced with an MS1 scan in the m/z range 40-600. This full scan method used a step size of 0.1 amu, a threshold of 10 and a scan time of 382 ms.

4.4.2

210

_{Pb Dating}

Approximately 1 g of soil spiked with 209_{Po was treated with 6 M HCl for 30 min at 95 °C. After} cooling, 1 mL 30% hydrogen peroxide was added together with 1 drop of octanol four times with 30-minute intervals. The oxidized samples were treated with 6 M HCl for 4 hours, cooled overnight, filtered, and partly evaporated. The remaining sample was combined with 100-200 mg ascorbic acid in 50 mL E-pure water to reach a pH of 1. In a 95 °C oven on copper disks, the samples were counted for 60 000 s with the spectrometer.

4.4.2.1 210_{Pb Data Calculations}

Porosity was calculated using the following equation. The density of the dry sediment core was assumed to be 2,3 g/cm3_{which is common in highly porous and organic rich lacustrine sediments} (Klump, 2021).

𝑃𝑜𝑟𝑜𝑠𝑖𝑡𝑦 = 1

1 +_{𝐷𝑒𝑛𝑠𝑖𝑡𝑦 ∙ (𝑊𝑒𝑡 𝑀𝑎𝑠𝑠 − 𝐷𝑟𝑦 𝑀𝑎𝑠𝑠)}𝐷𝑟𝑦 𝑀𝑎𝑠𝑠 Using the calculated porosity, the mass was calculated using the following equation.

𝑀𝑎𝑠𝑠 (𝑔/𝑐𝑚2_{) = 𝑇ℎ𝑖𝑐𝑘𝑛𝑒𝑠𝑠 (𝑐𝑚) ∙ 𝐷𝑒𝑛𝑠𝑖𝑡𝑦 (𝑔/𝑐𝑚}3_{) ∙ (1 − 𝑃𝑜𝑟𝑜𝑠𝑖𝑡𝑦)}

Going down from the top layer, cumulative masses were calculated as the sum of the mass of the current sample and the cumulative mass of the previous sample. Similarly, midpoint cumulative masses were calculated as the sum of the mass of the current sample and half the cumulative mass of the previous sample. The sample activity was plotted against the midpoint cumulative mass to

(26)

20

obtain the sediment accumulation rate. Analyte accumulation rates were calculated by multiplying the amount of an analyte in a layer with the CIC based age of that layer.

4.5. Method Development

4.5.1. Optimization of MRM Transitions and GC-MS Ions

Each analyte stock solution was derivatized separately following the derivatization procedure but by adding 1 μL of the respective analyte stock solution instead of 1 μL SINT. Since they are unaffected by derivatization, 1 μL of SREC, DIL and SINT, DIL was diluted to 100 μL in DCM and transferred to GC vials. Each prepared pure standard was analyzed using the full scan version of the GC-MS/MS method, after which a m/z value with high abundance was selected for each standard as pre cursor ions. Prepared pure standards were analyzed again by selecting the precursor ions in the first quadrupole, fragmenting it, and doing the same full scan in the second MS2 quadrupole. After the ion scans, a high m/z value with high abundance was selected for each standard as product ions to complete the optimized MRM transitions. The MRM method was finalized by creating retention time windows for each standard starting in the middle between the observed retention time of the standard and the previously eluting standard. A control of the MRM method was done by analyzing 5 μl of derivatized AMIX. The same samples and methodology were used to develop the GC-MS method not using the MS2 scan since a GC-MS only utilizes one quadrupole.

4.5.2. Establishment of a Calibration Curve

Calibration solutions were prepared by spiking 3 g of blank sand with 1, 5 and 10 μl AMIX in duplicates and performing the entire sample preparation procedure. After analysis with the optimized MRM method, a linear equation was fitted in Microsoft Excel. Each calibration curve correlates the quotient between the analyte peak area and the internal standard peak area with the analyte concentration in the GC vial.

4.6. Method Validation

AR and EE were calculated by preparing the four sample types Pre-Extr, Post-Extr, Standard and Non-Spiked in duplicates. Pre-Extr was prepared by spiking 3 g of blank sand with 5 μl AMIX before

extraction and therefore the two Pre-Extr samples were the same as the middle point samples of the calibration curve. Post-Extr was prepared by spiking an extract of 3 g blank sand with 5 μl AMIX directly after the ASE extraction. Standard was prepared by derivatizing 5 μl of AMIX. Non-spiked was prepared by simply extracting a blank sand sample. AR and EE values were calculated for each analyte from their peak areas according to the following equations.

𝐴𝑅 = [(𝑃𝑃𝑟𝑒−𝐸𝑥𝑡𝑟− 𝑃𝑁𝑜𝑛−𝑆𝑝𝑖𝑘𝑒𝑑) 𝑃⁄ 𝑆𝑡𝑎𝑛𝑑𝑎𝑟𝑑] × 100%

𝐸𝐸 = [(𝑃𝑃𝑟𝑒−𝐸𝑥𝑡𝑟− 𝑃𝑁𝑜𝑛−𝑆𝑝𝑖𝑘𝑒𝑑) (𝑃⁄ 𝑃𝑜𝑠𝑡−𝐸𝑥𝑡𝑟− 𝑃𝑁𝑜𝑛−𝑆𝑝𝑖𝑘𝑒𝑑)] × 100%

The correlation was calculated from the calibration curve. If the correlation of the calibration curve was good, the linearity range was determined as the calibration curve range. Intra-day accuracy was calculated from the calibration curve as the difference between measured and known

concentrations. Intra-day precision was calculated as the RSD of the calibration curve points. The LOD was determined as thrice the standard deviation of the background noise and then the LOQ was calculated. Matrix effects were evaluated by observing the signals from blank soil samples and comparing their m/z values with the optimized method.

(27)

21

4.7. Method Application

Freeze-dried soil samples were analyzed in singlicates. As mentioned, the 40 sediment samples in Table B – 1, Table B – 2 and Figure B – 1 in Appendix B were analyzed. Thus, 11 pharmaceuticals were searched for in 22 surface sediment samples and 18 sediment core samples using the optimized methodology.

4.8. Laboratory Routines

Before every use, all glassware was rinsed with water, dried at 65 °C for 15 minutes, combusted at 450 °C for 4 hours and then rinsed with MeOH. Between uses, all glassware was covered with aluminum foil to prevent contamination from air particles. ASE sample cells were rinsed with water, dried at 65 °C for 15 minutes and then rinsed with MeOH between uses. Plastic was avoided as much as possible to prevent leaks of plastic material components; plastic caps, although Teflon lined, were rinsed with deionized water and MilliQ water between uses.

Before every five real sediment samples, 3 g of blank sand was put through the entire extraction and analysis procedure to always handle a blank along with the real samples. Before every five analyses on the GC-MS, pure DCM was analyzed followed by pure hexane using a simple solvent method with the same MS parameters as the full scan method but with the temperature program 100 °C (1 min); 30 °C/min; 300 °C (0 min). Between every 40 analyses of real samples the GC injection needle was taken out and cleaned manually with DCM. After every GC analysis, the septum of the GC vial was changed and between analyses the GC vials were kept in a refrigerator. A recovery standard and an internal standard were used to compensate for and evaluate losses.

4.8.1. Quality Assurance and Control

Although instrumental difficulties and the project timeframe prevented routine quality assurance and control it is important to state that it was nonetheless planned for this project. Either deuterated or 13_{C-labeled standards, very similar to the target compounds, added before the extraction as} surrogate standards were planned besides an internal standard added before analysis. Then the analytes would have been quantified using the surrogate standards and the recoveries of the analytes would have been evaluated using the differences between the response s of the surrogate standards and the response of the internal standard. Furthermore, routine analyses of proper blanks such as procedure blanks and field blanks were also planned but omitted.

(28)

22

5. Results

5.1. Process Analysis

Method development on the GCMS/MS system was begun while the N2 leak was being resolved in the ASE 350 instrument. After the problems with the ASE 350 were resolved, we encountered difficulties with the autotuning on the GCMS/MS. At this point, time was allocated to extract and derivatize all the sediment samples to run them at the same time when the GCMS/MS would be operational. Thus, the three key milestones of having completed the method development, validation, and application were done simultaneously as one process.

During the analyses of different pharmaceutical standards (which initially worked well during method development), the GCMS/MS started showing low sensitivity. Autotuning showed a sensitivity and precision unacceptably lower than when the instrument was installed. Figure 6 shows an autotune report from 2018-05-16 when the instrument was installed, a), and two consecutive autotune reports from 2021-03-23, b) and c), showing the loss in sensitivity and precision since installation of the GC-MS/MS.

Figure 6 An autotune report from 2018-05-16 when the instrument was installed, a), and two consecutive autotune reports from 2021-03-23, b) and c), showing the loss in sensitivity and precision since installation of the GC-MS/MS.

(29)

23

Troubleshooting was done on the GCMS/MS based on guidance from the instrument manufacturer - Agilent Technologies. JetClean, a proprietary methodology and self-cleaning technique of Agilent, was used (see parameters in Table 2). This procedure did not resolve the problem and company technicians were called in to resolve the issues with instrument sensitivity. We switched to include manual cleaning of the ion source and replacing old, contaminated instrument parts (e.g., the gas filter, lenses, the filament, and the ion multiplier).

Table 2 Utilized aggressive JetClean parameters based upon the Agilent JetClean Operating Manual.

Source cleaning

Operation Clea n only

Hydrogen flow (mL/min) 3.53

Filament 2

Emission (μA) 10

Source temperature (°C) 350

Quadrupole temperature (°C) 150

Duration (min) 10

Post cleaning stabilization duration (min) 10

Solvent delay (min) 0.25

Gain 0.2

Scan type MS2 sca n

MS2 scan range (m/z) 50-300

Step size (amu) 0.1

Threshold 0

Scan time (ms) 1300

Nonetheless, the stability of the autotuning remained inadequate. Therefore, focus was changed to adapt the developed method to an Agilent 6890 GC-MS to obtain the results. Using the method parameters outlined in Table 4 to analyze calibration standards, with the concentrations 0.03, 0.15 and 0.3 ng/μL, calibration curves and linear regressions were plotted (Figure 7). Although

pharmaceutical concentrations in sediment samples are presented with the unit ng/g the unit ng/μL was used for the calibration curves since extract volumes, unlike the mass of extracted sediment samples, were always the same. The quotient between the analyte peak area and the internal standard peak area was plotted against the extract concentration since the recovery standard could not be reliably detected on the GC-MS.

Preliminary Investigation into Quantitation of Pharmaceuticals in Lake Victoria Sediments : Development of a Method for Analysis of 11 Pharmaceuticals