New tools for sample preparation and instrumental analysis of dioxins in environmental samples

(1)

New tools for sample preparation and instrumental analysis of dioxins in environmental samples

Lan Do

Department of Chemistry Doctoral Thesis

Umeå 2013

(2)

This work is protected by the Swedish Copyright Legislation (Act 1960:729)

ISBN: 978-91-7459-684-7 Cover picture: Lan Do

Electronic version available at http://umu.diva-portal.org/

Printed by: VMC, KBC, Umeå University Umeå, Sweden 2013

Papers I and II are reprinted with permission from Elsevier and The Royal Society of Chemistry, respectively

(3)

To my families and my älskling

(4)

(5)

(6)

Abstract

Polychlorinated dibenzo-p-dioxins (PCDDs) and dibenzofurans (PCDFs), two groups of structurally related chlorinated aromatic hydrocarbons, are of high concern due to their global distribution and extreme toxicity. Since they occur at very low levels, their analysis is complex, challenging and hence there is a need for efficient, reliable and rapid alternative analytical methods. Developing such methods was the aim of the project this thesis is based upon.

During the first years of the project the focus was on the first parts of the analytical chain (extraction and clean-up). A selective pressurized liquid extraction (SPLE) procedure was developed, involving in-cell clean-up to remove bulk co-extracted matrix components from sample extracts. It was further streamlined by employing a modular pressurized liquid extraction (M-PLE) system, which simultaneously extracts, cleans up and isolates planar PCDD/Fs in a single step. Both methods were validated using a wide range of soil, sediment and sludge reference materials. Using dichloromethane/n-heptane (DCM/Hp; 1/1, v/v) as a solvent, results statistically equivalent to or higher than the reference values were obtained, while an alternative, less harmful non-chlorinated solvent mixture - diethyl ether/n-heptane (DEE/Hp;

1/2, v/v) – yielded data equivalent to those values.

Later, the focus of the work shifted to the final instrumental analysis. Six gas chromatography (GC) phases were evaluated with respect to their chromatographic separation of not just the 17 most toxic congeners (2,3,7,8-substituted PCDD/Fs), but all 136 tetra- to octaCDD/Fs. Three novel ionic liquid columns performed much better than previously tested commercially available columns. Supelco SLB-IL61 offered the best overall performance, successfully resolving 106 out of the 136 compounds, and 16 out of the 17 2,3,7,8-substituted PCDD/Fs. Another ionic liquid (SLB-IL111) column provided complementary separation. Together, the two columns separated 128 congeners. The work also included characterization of 22 GC columns’

selectivity and solute-stationary phase interactions. The selectivities were mapped using Principal Component Analysis (PCA) of all 136 PCDD/F’s retention times on the columns, while the interactions were probed by analyzing both the retention times and the substances’ physicochemical properties.

Key words: PCDD/Fs, dioxins, pressurized liquid extraction, PLE, selective SPLE, modular M-PLE, soil, sediment, sludge, gas chromatography, new stationary phases, multivariate data analysis, selectivity, interaction.

(7)

Table of contents

(

PCDD/Fs) 1

1.2 Analytical challenges in studies of PCDD/Fs 3

1.3 Aims of this thesis 4

2 Pressurized liquid extraction (PLE) 5

2.1 Principles of operation 5

2.2 Selective pressurized liquid extraction (SPLE) 6

2.2.1 Paper I: Method development 6

2.2.1.1 Solvent selection 7

2.2.1.2 Optimization of extraction parameters 8

2.2.1.3 Validation 9

2.2.1.4 Outlook 10

2.3 Modular pressurized liquid extraction (M-PLE) 10

2.3.1 Paper II: Method development 11

2.3.1.1 Optimization of the carbon trap 12

2.3.1.2 M-PLE performance with low sample intake 12 2.3.1.3 M-PLE performance with high sample intake 14

2.3.1.4 Outlook 15

3 Gas chromatography (GC) 17

3.1 Principles 17

3.2 Summary of Paper III 18

3.2.1 New stationary phases with specificity for dioxins 18

3.2.2 Column combinations 23

3.2.3 Outlook 24

3.3 Summary of Paper IV 25

3.3.1 Multivariate data analysis 25

3.3.2 Column characterization 26

3.3.2.1 First PCA modeling of the retention times 26 3.3.2.2 PLS modeling of the physicochemical properties 27 3.3.2.3 Second PCA modeling of the retention times and physicochemical

properties 28

3.3.2.4 Suitable GC × GC column combinations for dioxin analysis 28

3.3.3 Outlook 28

4 Detection and quantification 30

5 Conclusions and future prospects 31

6 Acknowledgements 32

7 References 34

(8)

List of papers

This thesis is based on the following papers:

I Do, L., Lundstedt, S., Haglund, P. “Optimization of selective pressurized liquid extraction for extraction and in-cell clean-up of PCDD/Fs in soils and sediments”, Chemosphere 90 (2013) 2414-2419

II Do, L., Hoang, X.T., Lundstedt, S., Haglund, P. “Modular pressurized liquid extraction for simultaneous extraction, clean-up and fractionation of PCDD/Fs in soil, sediment and sludge samples”, Analytical Methods 5 (2013) 1231-1237

III Do, L., Liljelind, P., Zhang, Z., Haglund, P. “Comprehensive profiling of 136 tetra- to octa- polychlorinated dibenzo-p-dioxins and dibenzofurans using ionic liquid columns and column combinations”, Submitted manuscript

IV Do, L., Geladi, P., Haglund, P. “Multivariate data analysis to characterize GC columns for dioxin analysis”, Manuscript

Author contribution:

Paper I: The author contributed extensively to the planning of the experiments, performed all of the experimental work, and wrote the paper.

Paper II: The author contributed extensively to the planning of the experiments and supervised Hoang X. T., a Master’s student who performed some of the experimental work. The author also performed some of the experimental work and wrote the paper.

Paper III: The author was heavily involved in the planning of the experiment, performed most of the experimental work and wrote the paper. The author supervised a Master’s student Mamoon H. A., who also was involved in the experiments and the data evaluation.

Paper IV: The author was involved in the planning of the experiment, contributed substantially to the data evaluation, and wrote the paper. The multivariate data analysis was done by Paul Geladi.

(9)

Abbreviations

AhR Aryl hydrocarbon receptor CRM Certified reference material CCF Central composite face DCM Dichloromethane

DCM/Hp Dichloromethane/n-heptane DEE Diethyl ether

DEE/Hp Diethyl ether/n-heptane

ELISA Enzyme-linked immunosorbent assay GC Gas chromatography

GC/HRMS Gas chromatography/high resolution mass spectrometry GC × GC Comprehensive two-dimensional gas chromatography

GC × GC/MS Comprehensive two-dimensional gas chromatography/mass spectrometry

GLC Gas-liquid chromatography GSC Gas-solid chromatography Hp n-heptane

HRMS High resolution mass spectrometry IS Internal standard

M-PLE Modular pressurized liquid extraction m/z Mass-to-charge

PAHs Polycyclic aromatic hydrocarbons

PBDD/Fs Polybrominated dibenzo-p-dioxins and dibenzofurans PBDEs Polybrominated diphenylethers

PC Principal component

PCA Principal component analysis PCDDs Polychlorinated dibenzo-p-dioxins PCDFs Polychlorinated dibenzofurans

PCDD/Fs Polychlorinated dibenzo-p-dioxins/furans PCBs Polychlorinated biphenyls

PLE Pressurized liquid extraction

PLE-C Pressurized liquid extraction with integrated carbon trap PLS Partial least square

POPs Persistent organic pollutants RSD Relative standard deviation

SPLE Selective pressurized liquid extraction TEF Toxic equivalency factor

TEQ Toxic equivalent TOF Time-of-flight

WHO World Health Organization

(10)

(11)

Introduction

1 Introduction

1.1 Polychlorinated dibenzo-p-dioxins/ dibenzofurans (PCDD/Fs)

O

Cl_x Cl_y

Cl_x O Cl_y

1 2

3 4 7

6 9 8

1 2

3 4 6

7 8

9

Figure 1. General structural formula and substitution positions of PCDDs (left) and PCDFs (right).

Polychlorinated dibenzo-p-dioxins (PCDDs) and dibenzofurans (PCDFs), so called

‘dioxins’, are structurally similar planar compounds that feature two chlorine substituted benzene rings connected by one (furan) or two (dioxin) oxygen bridges (Figure 1). The three-ring heterocyclic backbone structure can be substituted with one to eight chlorine atoms in eight different positions, which are denoted 1-4 and 6- 9 in Figure 1. This yields a total of 75 possible dioxin and 135 possible furan congeners. Congeners with the same number of chlorines comprise a homologous group; differences in the distribution of chlorines within such groups give rise to different isomers.

The toxicity of the PCDD/Fs is highly dependent on the locations of the substituted chlorines. Of the 210 PCDD/Fs, the 17 congeners with chlorines in the 2,3,7,8 positions are the most toxic, with half-lives of 2-7 years in humans [2]. Their toxicity arises from their ability to bind to the aryl hydrocarbon receptor (AhR). This has a number of adverse consequences, including the induction of reproductive disorders, immunotoxicity, cancer, weight loss, and acute chloracne among others. Because the AhR- mediated response is different for each congener, the concept of toxic equivalency factors (TEFs) was developed to assess their toxicity. These TEF values relate the toxicity of each congener to the toxicity of 2,3,7,8- tetrachlorodibenzo-p-dioxin (2,3,7,8-TeCDD), which is considered the most toxic congener [1]. The TEF value of 2,3,7,8-TeCDD is equal to

1. All of the other congeners consequently have TEF values of less than 1, with the exception of 1,2,3,7,8-PeCDD, which also has a TEF of 1. The congeners’ TEF values

Table 1. WHO-TEF values for PCDD/Fs [1].

Isomers TEF-2005

2,3,7,8-TeCDF 0.1 1,2,3,7,8-PeCDF 0.03 2,3,4,7,8-PeCDF 0.3 1,2,3,4,7,8-HxCDF 0.1 1,2,3,6,7,8-HxCDF 0.1 2,3,4,6,7,8-HxCDF 0.1 1,2,3,7,8,9-HxCDF 0.1 1,2,3,4,6,7,8-HpCDF 0.01 1,2,3,4,7,8,9-HpCDF 0.01

OCDF 0.0003

2,3,7,8-TCDD 1 1,2,3,7,8-PeCDD 1 1,2,3,4,7,8-HxCDD 0.1 1,2,3,6,7,8-HxCDD 0.1 1,2,3,7,8,9-HxCDD 0.1 1,2,3,4,6,7,8-HpCDD 0.01

OCDD 0.0003

(12)

Introduction

were most recently updated in 2005, by the World Health Organization (WHO) (Table 1). The total dioxin-like toxicity of a sample is expressed in units of toxic equivalents (TEQ), and is determined by multiplying the concentration of each dioxin congener by its TEF value and summing the resulting products. TEQ values are commonly used to compare the dioxin toxicities of different samples, but one should be aware that the TEQ and TEF concepts were originally developed for assessing risks associated with human food consumption, under the assumption that the dioxin- containing materials would be ingested orally. In soil/sediment samples, where contaminants may adsorb strongly to the sample matrix, the TEQ value may overestimate the sample’s overall toxicity because it does not properly reflect the bioavailability of the contaminants.

Natural sources of PCDD/Fs such as forest fires, volcanoes and biological processes [3, 4] make only minor contributions to the total dioxin load. Most dioxins are formed unintentionally during:

i) combustion processes in which organic and inorganic compounds are burned together in the presence of chlorine. This can occur during the incineration of municipal solid waste, biomass burning and during accidental fires or backyard burning [5-8].

ii) chemical processes such as the bleaching of pulp with chlorine gas [9, 10], the production of chlorine via the chloralkali process using graphite electrodes [10, 11], and the production of organochlorine chemicals such as polychlorinated biphenyls (PCBs) and chlorophenols (in which case dioxins are formed as byproducts) [12, 13].

Due to technological improvements and regulations, industrial emissions of PCDD/Fs have been substantially reduced, and the current primary sources of these compounds are waste incineration and sinter plants. However, emissions from non- industrial sources such as domestic solid fuel combustion are barely decreased, and may be responsible for the majority of PCDD/F emissions in the near future [14].

Because PCDD/Fs are highly persistent and undergo long-range transport, they have been identified in all environmental compartments that have been studied across the globe. Soils and sediments are considered to be the main repositories of PCDD/Fs due to their low water solubility and low volatility. Consequently, soils and sediments at old industrial sites often retain serious levels of these compounds decades after their initial contamination. High dioxin levels (0.14 - 3000 µg/kg d.w.

TEQ) have been observed in contaminated soils at former Swedish sawmill sites, where chlorophenol agents that were contaminated with dioxins had been used to preserve wood [15]. These contaminated sites pose a risk to humans and animals living in their immediate vicinity, but also contribute to the global spread of dioxins throughout the environment. Figure 2 illustrates the various pathways by which human beings may be exposed to dioxins: they can enter the food chain via vegetables or terrestrial organisms and then become biomagnified in predators. Alternatively, the pollutants may associate with colloidal particles that are subsequently transported from the contaminated soils into the groundwater and nearby rivers,

(13)

Introduction

where the pollutants may be stored in sediments or taken up by aqueous organisms [16]. The dioxins may also evaporate into the air or adsorb onto aerosol and dust particles. All of these processes may result in human exposure and could thus threaten human health. Consequently, there is a need for efficient measurement techniques that can be used to monitor dioxin levels in several compartments of the environment.

Figure 2. Potential pathways of dioxin exposure [17].

1.2 Analytical challenges in studies of PCDD/Fs

The quantitative analysis of PCDD/Fs in environmental samples is challenging due to their low concentrations and the common presence of multiple interfering compounds whose concentrations may be orders of magnitude greater. Complex multi-step protocols are typically required to determine such trace components.

Important steps include i) extraction of the target analytes from the matrix, ii) removal of co-extractable organic materials such as lipids, sulfur and humic materials, iii) fractionation of the dioxins to remove other interfering compounds, iv) separation of target analytes from non-target compounds and sources of interference using a gas chromatography (GC) column, and finally v) detection of the target analytes using a selective mass spectrometer. Traditional protocols for dioxin analysis usually involve an initial Soxhlet extraction followed by a multi-column clean-up process and analysis by gas chromatography/high resolution mass spectrometry (GC/HRMS) [18-21]. However, such approaches are very labor intensive, costly as well as time consuming, and also use large quantities of purified organic solvents.

Dioxin analysis using GC/HRMS typically costs $500–$1000 per sample, and takes around 25h in total [22, 23]. In addition, because there are 136 PCDD/Fs, with four to eight chlorines per molecule, all of which have relatively similar physicochemical properties, conventional GC columns (such as the DB-5) cannot fully separate all of the isomers present within typical samples. In particular, the 17 2,3,7,8-substituted dioxins and furans are not readily separated from the other isomers. It is therefore

(14)

Introduction

usually necessary to perform a confirmatory analysis using a second GC column. As such, there is a pressing need for more efficient methods.

1.3 Aims of this thesis

The aim of the work presented in this thesis was to develop new methods for dioxin analysis, with two key goals. The first goal was to establish a method for high- throughput sample preparation based on modular pressurized liquid extraction (M- PLE), a coupling system that permits the simultaneous, single-step extraction, clean- up and fractionation of PCDD/Fs. Papers I and II describe the development of such a method and its application in the analysis of environmental solid samples. The second goal was to improve the GC separation of PCDD/Fs. Six GC columns, including three coated with ionic liquid phases, were evaluated with respect to their chromatographic separation of the 136 tetra- to octaCDD/Fs. This also involved the characterization of the column selectivity and interaction of the solutes on 22 GC columns during the chromatographic separation. Comparative studies along these lines are described in Papers III and IV.

(15)

Pressurized liquid extraction (PLE)

2 Pressurized liquid extraction (PLE)

2.1 Principles of operation

PLE is an automated extraction technique that uses organic solvents at elevated temperatures and pressures to extract organic pollutants from solid matrices [24, 25].

PLE systems consist of a stainless steel extraction cell into which the sample is placed, an oven, a pump, a gas tank, solvent bottles, collection vials and various pressure control valves. Figure 3 shows the general setup of such a system. During a typical extraction, the cell is:

1) loaded into the oven

2) filled with the organic solvent

3) heated and pressurized to predefined levels

4) extracted in several static cycles; at the end of each static cycle, the extract is transferred to the collection vial and fresh solvent is pumped through the cell to initiate the next static cycle until finish

5) purged with nitrogen gas to discharge residual solvent from the cell into the collection vial

6) depressurized

Figure 3. General setup of a pressurized liquid extraction system (adapted with permission from Pouralinazar et al., 2012 [26]).

High extraction temperatures increase the rates of diffusion and mass transfer during the extraction process as well as the solubility of the analyte in the extraction solvent and the rate of analyte desorption from the matrix (by disrupting analyte- matrix interactions). High pressures help to maintain the solvent in the liquid state at higher temperatures while also increasing the efficiency of sample wetting and matrix penetration, thereby enhancing extraction efficiency. The use of high temperatures and pressures also reduces the time required for the extraction process and the

(16)

volume of solvent used. Moreover, the method is amenable to automation and offers considerable flexibility. Interest in PLE has therefore grown considerably over the 18 years since its introduction, and it is increasingly regarded as a promising alternative to many tedious and solvent-consuming extraction techniques such as Soxhlet extraction and sonication [24, 25, 27, 28].

2.2 Selective pressurized liquid extraction (SPLE)

The exhaustive extraction achieved when using PLE is valuable because it means that quantitative analyte recovery can be achieved in only a few minutes via a static extraction process. However, PLE is also relatively unselective, and the resulting extracts are rich in co-extracted materials that need to be removed using complex clean-up procedures [29]. In 1996, Dionex reported that the introduction of alumina into PLE cells eliminated fat from extracted biota samples [30]. Since then, the concept of SPLE, i.e. PLE extraction with integrated clean-up achieved by adding specific matrix retainers, has become increasingly popular. A number of papers have been published describing strategies for extracting persistent organic pollutants (POPs) from food/feed or abiotic samples with in-cell clean-up [31-41]. For example, Wiberg and Sporring et al. [34-36] used sulfuric acid-impregnated silica to remove lipids from food and feed samples during PCB extraction. Chuang et al. [37]

developed a SPLE strategy for extracting PCDD/Fs from sediment and soil samples in which multilayer adsorbents are used to retain the sample matrix. Ong et al. [40]

utilized silica to retain humic matter and other polar co-extractants in an SPLE process for extracting polycyclic aromatic hydrocarbons (PAHs) from soil samples.

Lundstedt et al. [31] developed this approach further by combining in-cell purification with fractionation in order to separate the PAHs from their oxygenated derivatives. Similarly, Poerschmann et al. [41] used SPLE to fractionate neutral lipids from polar phospholipids. In all of these cases, the results obtained using SPLE were not only equivalent to or even better than those obtained by traditional methods, but the SPLE approach was also preferable in terms of time and cost.

2.2.1 Paper I: Method development

The purpose of Paper I was to develop an SPLE method for the extraction and in- cell clean-up of PCDD/Fs in soils and sediments. The developments were done on the ASE200 with 22 mL cell. The SPLE cell was packed with multiple silica layers (20%

KOH-silica, neutral silica and 40% H2SO4-silica). The dried sample was mixed with Celite (1:1, w/w) and placed in the cell on top of the silica, as shown in Figure 4.

During the extraction process, PCDD/Fs along with other POPs and organic compounds were extracted from the matrix and flushed through the multilayer silica plug, which retained polar and hydrolysable compounds, removing them from the final extract. The extracts were then treated with activated copper and passed

(17)

through a carbon column to separate the planar compounds (PCDD/Fs) from the non-planar ones (other POPs) prior to GC/HRMS analysis.

Figure 4. SPLE cell packing (Paper I).

2.2.1.1 Solvent selection

The choice of solvent is important to obtain a quantitative extraction of the target analytes. For the samples investigated, binary solvent mixtures proved to be more effective than single solvents. This is presumably because binary mixtures have some of the properties of both of their constituents, enabling them to dissolve a wide range of compounds with different polarities [42, 43]. Dichloromethane (DCM) has been identified as an efficient extraction solvent for dioxins [37, 44], and is suitable for SPLE because it does not react with the sulfuric acid in the multilayer silica plug [45].

However DCM can have adverse effects on human health and is harmful to the environment [46]. Therefore, a less toxic alternative with similar physicochemical properties to DCM was sought. After a preliminary investigation, diethyl ether (DEE) was identified as a suitable alternative. Two binary solvent mixtures were evaluated in the studies reported in Paper I: dichloromethane/n-heptane (DCM/Hp) and diethyl ether/n-heptane (DEE/Hp).

The next step was to determine the optimal proportions of DCM/Hp and DEE/Hp in terms of maximizing the amount of PCDD/Fs extracted while minimizing the amount of co-extracted material. The certified soil CRM-529 was used as reference sample and was extracted with six solvent mixtures: DCM/Hp (1/10, 1/4, 1/1, v/v) and DEE/Hp (1/10, 1/4, 1/1, v/v). The SPLE results obtained with each mixture were then compared to those achieved using a reference method based on Soxhlet extraction followed by external clean-up and fractionation columns [47] (Figure 5).

The performance of each mixture was evaluated based on the PCDD/F concentration in the SPLE extract as a percentage of that achieved using the Soxhlet method (%

extracted). It was apparent that solvent mixtures containing lower proportions of the polar solvent afforded less efficient extractions (Figure 5), meaning that it is necessary to include a polar solvent in order to achieve exhaustive dioxin extraction, and that an excessively high n-heptane (Hp) content will reduce the solvent strength.

The best SPLE extraction efficiency was achieved using DCM/Hp (1/1, v/v). The

Sample

40% H2SO4-silica Active silica 20% KOH-silica 2 filter papers

(18)

extraction efficiency achieved using DEE/Hp (1/1, v/v) was similar to that obtained with DEE/Hp (1/4, v/v). However, SPLE with DEE/Hp (1/1, v/v) caused severe leaching of the acid from the H2SO4-silica and should therefore be avoided.

Consequently, DEE/Hp (1/2, v/v) and DCM/Hp (1/1, v/v) were selected as the optimal solvent mixtures.

Figure 5. PCDD/F extraction percentages (% extracted) achieved using SPLE with six solvent mixtures (relative to the reference Soxhlet method). Error bars indicate 95% confidence intervals based on triplicate analyses.

2.2.1.2 Optimization of extraction parameters

A central composite face (CCF) design was used to optimize three extraction parameters: temperature, number of extraction cycles and extraction time per cycle.

Pressure was not included as an optimization parameter because several previous studies have shown that it has negligible effects on extraction efficiency [24, 48-51].

Temperature was found to be the most important parameter, and had a strong positive effect on the extraction efficiency (i.e. more dioxins are extracted at higher temperatures). The second most significant parameter was the extraction time, which also had a positive influence. The extraction time required depends on the sorption of the analytes to the sample matrix. For “aged” samples in which the analytes have penetrated deeply into the matrix and are therefore strongly sorbed, longer extraction times are required to achieve complete desorption. The third parameter, number of extraction cycles had no significant effect on the extraction efficiency and so only two cycles were used to minimize the quantity of co-extracted material.

0 20 40 60 80 100

% Extracted

(19)

a) b)

Figure 6. PCDD/F extraction percentages (% extracted, relative to the reference Soxhlet method) achieved during the optimization of SPLE settings for a) DCM/Hp (1/1, v/v) and b) DEE/Hp (1/2, v/v) plotted as a function of the extraction temperature and extraction time

Results of the optimization can be visualized using response surface plots. Figure 6 shows the response surfaces for the ‘% extracted’ as functions of the extraction temperature and time. The optimal domain (shown in red) that provides the highest extraction efficiencies covers a fairly large area, which corresponds to the optimal conditions. According to these models, the optimal extraction temperatures are 148°C for SPLE methods using DCM/Hp (1/1, v/v) and 160°C for SPLE methods using DEE/Hp (1/2, v/v).

However, it was subsequently found that extraction temperatures above 120°C should be avoided because they cause the leaching of H2SO4 from the 40% H2SO4- silica followed by water formation from the neutralization of free H2SO4 by KOH- silica during the extraction. The extractions were therefore performed at lower temperatures in order to avoid this problem. The final extraction conditions selected were two cycles of 11 minutes each at 110°C for SPLE methods using DCM/Hp (1/1, v/v) and two cycles of 12 minutes each at 110°C for SPLE methods using DEE/Hp (1/2, v/v).

2.2.1.3 Validation

The optimized conditions were validated by applying them to three certified reference materials (CRMs): the soil CRM-529, the clay CRM-530 and the sediment WMS-01. The SPLE results were compared to either the certified values or those achieved using a Soxhlet based method, both in terms of individual congener concentrations and total TEQs. In all cases, the relative standard deviations (RSDs) for the triplicate extractions were below 12%, which was within acceptable limits. The accuracy (trueness) of the TEQ values of SPLEDCM/Hp and SPLEDEE/Hp compared to the certified TEQ values was +11% and +8% for the clay sample, +8% and -7% for the sediment sample, +8% and -10% for the soil sample, respectively. The congener concentrations determined using the SPLE methods generally agreed well with the certified and Soxhlet values with the exception of those highly chlorinated (Hp-Oc)

% Extracted % Extracted

(20)

congeners, for which the SPLEDCM/Hp values tended to be higher than the reference values. This indicates that SPLE methods using optimized conditions provide even better extraction efficiencies than the certified/Soxhlet methods. It was also found that the mean of the concentrations for each congener obtained by SPLEDEE/Hp was slightly lower than those determined using SPLEDCM/Hp. This suggests that DCM/Hp mixtures provide somewhat greater SPLE extraction efficiencies than DEE/Hp, which probably reflects differences in solvent selectivity between the more polarizable DCM and the slightly basic DEE. However, DEE/Hp may be the preferable option due to its lower environmental impact.

2.2.1.4 Outlook

Our SPLE method provides numerous advantages relative to conventional methods for dioxin analysis, which involve Soxhlet extraction and open column clean-up. It combines two steps (extraction and bulk matrix removal), resulting in a more automated, faster, and more cost-efficient procedure with reduced solvent consumption and scope for high throughput (a series of 24 samples can be extracted).

While SPLE with in-cell clean-up is not a new approach and there have been several publications in this area, most of them have focused on extracting POPs in biological samples, which is much more straightforward than extracting soil and sediment samples. This is because of the strong sorption of organic pollutants in aged samples with a closed pore-structure that limits molecular diffusion. The SPLE method developed in this work is novel and efficient (having been optimized using chemometrics), and reliably extracts dioxins from these difficult matrices. When used in combination with congener-specific GC/HRMS analysis, it provides a wealth of information that can be used for risk assessment. Conversely, alternatives such as the SPLE method with enzyme-linked immunosorbent assay (ELISA) detection, developed by Chuang et al. [37] and used for solid samples, do not provide any information on the relative proportions or TEQ contributions of individual congeners.

2.3 Modular pressurized liquid extraction (M-PLE)

Despite the advantages of SPLE, it still requires a subsequent carbon column to separate dioxins from other interfering compounds. An alternative approach is to use pressurized liquid extraction with integrated carbon trap (PLE-C), which fractionates chemical mixture based on the planarity of their constituents directly on activated carbon. For example, it can be used to separate non-planar ortho-substituted PCBs from planar PCDD/Fs. PLE-C has proved to be useful for sample preparation when determining PCDD/Fs in biotic and abiotic samples [23, 52-54]. However, the resulting extracts usually contain additional co-extracted materials from the matrix that must be removed via an external clean-up step using a multilayer silica column or some other appropriate method. We hypothesized that by combining shape-

(21)

selective PLE-C and SPLE with in-cell clean-up, it might be possible to streamline the extraction and purification process and to take advantage of the synergistic properties of both techniques. This approach, which combines extraction, clean-up and fractionation in a single step is called M-PLE. The first M-PLE was developed using the Dionex ASE200 system by Erik Spinnel and Peter Haglund [45, 55] in 2008. The procedure has primarily been used to separate PCBs from chlorinated/brominated dioxins in biological samples, with some success. However, the M-PLE extraction cell used in the original protocol was rather small (possible maximum capacity is 22 ml in the ASE200), which limited the quantity of absorbent material that could be used for in-cell clean-up. Consequently, the original protocol did not completely eliminate co-extracted materials.

2.3.1 Paper II: Method development

The purpose of this study was to optimize the M-PLE approach for use with non- biological samples and non-chlorinated solvents. The new protocol designed to use the new Dionex ASE350 system, which has a larger cell volume than the ASE200 and should thus avoid the size limitations mentioned above.

The M-PLE strategy is illustrated in Figure 7. Two extraction cells were connected using an in-house manufactured adapter consisting of two cylindrical stainless steel guides with threading identical to the original PLE end-caps, and a central PEEK disc located in the center. The guides align the two extraction cell compartments, and the compression unit of the PLE system presses the extraction compartments against the PEEK disc to create a seal. This setup proved to be leak free. Figure 7 also outlines the developed extraction protocol. Cell 1 is filled with the sample and multiple layers of silica, as described in Paper I. Cell 2 is filled with a carbon-Celite mixture. The internal volumes of Cell 1 and Cell 2 can be adjusted to suit the situation at hand:

cells of 1, 5, 11 and 22 mL are available for the ASE200 instrument while cells of 1, 5, 10, 34, and 66 mL can be used with the ASE350. The extraction protocol involves three steps. In step one (forward elution), the whole system is extracted with DCM/Hp (1/1, v/v) or DEE/Hp (1/2, v/v) at which matrix components are retained by the acid-base silica, while persistent compounds such as PCBs and dioxins pass from Cell 1 to Cell 2. In Cell 2, planar compounds such as PCDD/Fs and non-ortho PCBs adsorb onto the carbon, while non-planar compounds such as ortho- chlorinated PCBs pass through the carbon trap and elute in the first fraction (Fraction 1). In step two, Cell 1 is removed, after which Cell 2 is sealed with a fresh end cap and inverted. In step three, the inverted cell is extracted with toluene, yielding a second fraction (Fraction 2) that contains the dioxins.

(22)

Figure 7. Schematic illustration of the cell packing sequence and M-PLE protocol that was developed for the analysis of PCDD/Fs in solid samples.

2.3.1.1 Optimization of the carbon trap

A good balance between the carbon trap capacity and the elution strength of the extraction solvent is essential for effective fractionation. A simple experiment was therefore conducted to optimize the carbon trap: Cell 1 was filled with clean sand (no silica was added) and spiked with a mixture of internal standards (IS) that included 17 ¹³C-labelled 2,3,7,8-substituted PCDD/Fs and four ¹³C-labelled non-ortho PCBs.

Cell 2 was then filled with various carbon/Celite mixtures ranging from 0.5-15%

carbon (w/w). The two solvent mixtures and the associated optimized extraction conditions from Paper I were reused, with one difference: a higher extraction temperature was used in the M-PLEDEE/Hp experiments to enhance the extraction efficiency. The adsorbent capacity was estimated by varying the amount of carbon in the trap to determine the minimum amount of carbon required to retain all of the PCDD/Fs while minimizing the retention of coplanar non-ortho PCBs. The results indicated that the lower the carbon content of the trap, the greater the breakthrough of non-ortho PCBs and dioxins. However, no combination of carbon content and eluent composition provided a complete separation of coplanar PCBs from PCDD/Fs.

Both compound classes were quantitatively trapped when the carbon content of Cell 2 was greater than 1%. With a carbon content of 1%, non-ortho PCBs started to break through but the PCDD/Fs were strongly retained in the carbon. With a carbon content of 0.5%, some PCDD/Fs started to pass through the trap.

2.3.1.2 M-PLE performance with low sample intake

Traps containing 1% carbon/Celite were used for M-PLE extraction of samples with high or moderate PCDD/F contents. In such cases, it is sufficient to use 1 g of sample

Celite

Sample H2SO4-silica Activated silica

KOH-silica

Adaptor

Carbon/Celite

Cell 1

Cell 2

Fraction 1 (non-planar compounds)

1. Forward elution

Fraction 2 (planar compounds, incl.

PCDD/Fs) 2a. Dissemble 2b. Backward elution

(23)

material in the extraction. Analyses of a wide range of solid matrices, including the industrial reference soil CRM-529, the reference sediment WMS-01, and two intercalibration samples (Soil C and Sludge C) yielded measured concentrations that agreed well with the certified/Soxhlet/consensus values. The RSDs were below 21%

for all of the triplicate measurements. The precision of the TEQ values ranged from 2- 9%; this satisfies the requirements outlined in EC regulation 1883/2006, which pertains to dioxin residues in solid samples [56].

Table 2. TEQ values (pg/g) for four samples extracted using traditional Soxhlet or M-PLE methods.

CRM-529 WMS-01 Soil C Sludge C Certified/Consensus - 59 ± 20 140 ± 35 46 ± 17

Soxhlet 7500 ± 260 64 ± 5 - -

M-PLEDCM/Hp 7400 ± 160 68 ± 3 160 ± 6 45 ± 3 M-PLEDEE/Hp 6700 ± 300 68 ± 5 140 ± 5 37 ± 3

Although the M-PLE and Soxhlet results were in good agreement with respect to overall PCDD/F concentrations, the Hp-OCDD/F concentrations determined using the SPLE method with DCM/Hp (1/1, v/v) were somewhat greater than those determined using the Soxhlet protocol as reported in Paper I. The same effect was observed when using the M-PLE method with DCM/Hp in Paper II, which suggests that the higher concentrations measured in the previous experiments were accurate.

This implies that the optimized M-PLE protocol using DCM/Hp as the extraction solvent may be more efficient than the Soxhlet based method. If so, this is presumably due to the higher temperatures and pressures used in the PLE-methods, and the more selective extraction achieved when using a binary solvent. When performing M-PLE with DEE/Hp (1/2, v/v), the extraction temperature was increased from 110°C to 140°C because preliminary investigations demonstrated that the extraction efficiency at 110°C was unacceptably low. In order to prevent acid leaching at this higher temperature, the composition of the multilayer silica plug was changed: whereas the protocol outlined in Paper I calls for 6 g of KOH–silica, 1 g of silica, and 2.5 g of 20% H2SO4-silica, the revised M-PLEDEE/Hp protocol uses 6 g of KOH-silica, 4 g of silica, and 2.5 g of 20% H2SO4-silica. The use of a larger silica plug was made possible by the bigger cell of the ASE350 instrument. Although the M- PLEDEE/Hp protocol was reasonably efficient, it was less so than the M-PLEDCM/Hp

protocol. Nevertheless, it yielded results that agreed well with the certified/Soxhlet values.

The M-PLE approach was also tested on fly ash #1879, which is an in-house reference material used for accreditation monitoring in the lab. A 1 g fly ash sample was loaded directly into the M-PLE cell without acid treatment and then extracted with DCM/Hp (1/1, v/v) at 110°C. Under these conditions, the extracted quantities of all congeners were significantly lower than the reference values obtained using the traditional Soxhlet method (Figure 8), even after increasing the extraction temperature to 130°C or 150°C quantitative dioxin recovery was not achieved,

(24)

meaning that the M-PLE conditions were not sufficiently exhaustive to extract all of the dioxins from a carbon-rich matrix such as fly ash.

Figure 8. Concentrations (pg/g) of PCDD/Fs extracted from fly ash by the reference Soxhlet method and by the M-PLE method with DCM/Hp (1/1, v/v) at three extraction temperatures: 110°C, 130°C and 150°C.

2.3.1.3 M-PLE performance with high sample intake

In environmental analysis, it is common to use samples of more than 1 g in cases where the PCDD/F content is low or the samples are heterogeneous. However, high sample intakes (≥ 2 g) proved to be a challenge for the current M-PLE strategy, resulting in significant breakthrough of IS. This effect became more pronounced as the sample intake increased (Figure 9a). The level of breakthrough observed when using the DEE/Hp method, which has a higher extraction temperature, was even worse than that observed with the DCM/Hp protocol. This was probably due to increased competition for the adsorption sites in the carbon plug. To increase the trap’s tolerance, its carbon content was raised from 1% carbon/Celite to 3%. Under these conditions, samples of up to 8 g could be extracted (Figure 9b).

0 2000 4000 6000 8000 10000 12000 14000 16000 18000

Reference Temp 110C Temp 130C Temp 150C

(25)

a) b)

Figure 9. Distributions (%) of ¹³C-PCDD/Fs in the first and second fractions obtained using the M-PLE method with a) 1% carbon/Celite and b) 3% carbon/Celite.

Intercalibration Sediment C was used to validate the 3% carbon/Celite trap with sample masses of 2-8 g using the ASE350 instrument. The M-PLE results obtained in this way agreed well with the Soxhlet results, with some exceptions for highly chlorinated congeners. The overall TEQs were within 95% uncertainty of the Soxhlet values, and no breakthrough of IS was observed. Consequently, this protocol should be suitable for use with high sample intakes. However, one should be aware of that this approach may result in more extensive extract contamination, potentially necessitating an additional clean-up step. It should therefore only be used when needed.

2.3.1.4 Outlook

The M-PLE procedure is an automated version of the original Stalling method for dioxin analysis, which involves Soxhlet extraction followed by clean-up on gravity-fed columns filled with multiple layers of silica, acid- or base-modified silica, and activated carbon. The new protocol has the advantage that extraction and clean-up are performed in a single step rather than three. As such, the M-PLE method has all the advantages of the SPLE method described in Paper I but is even more amenable to automation due to the incorporation of a carbon trap. Compared to the pioneering M-PLE work of Erik Spinnel [45, 55], this M-PLE protocol is more robust and therefore applicable to a wider range of solid samples. In addition, it can be used with environmentally friendly non-chlorinated extraction solvents (DEE/Hp). Other automated extraction and clean-up methods for dioxin analyses are commercially available at present, such as the PowerPrep system (FMS) [57-59]. However, these use costly disposable adsorbent cartridges, consume large quantities of pure solvent,

0%

20%

40%

60%

80%

100%

2 g 5 g 10 g 15 g 2 g 4 g 8 g DCM/Hp DEE/Hp

Fr 1 Fr 2

1% carbon /Celite

0%

20%

40%

60%

80%

100%

2 g 4 g 8 g 2 g 4 g 8 g

DCM/Hp DEE/Hp

Fr 1 Fr 2

3% carbon /Celite

(26)

have low sample throughput (six parallel samples) and are tricky to operate. These drawbacks are counterbalanced by the fact that the PowerPrep system (FMS) can be used for multi-class analysis and can be integrated into different sample preparation units, making it possible to leave samples unattended for several hours. In principle, our M-PLE method should also permit multi-class analysis, since the first M-PLE fraction (which contains ortho-PCBs and other non-planar compounds) could be analyzed by comprehensive two-dimensional GC mass spectrometry (GC × GC/MS) analysis while the second M-PLE fraction (which contains non-ortho PCBs and other planar compounds) can be analyzed using GC/HRMS or GC × GC/MS, depending on the user’s interests.

(27)

Gas chromatography (GC)

3 Gas chromatography (GC)

3.1 Principles

In gas chromatography (Figure 10), the sample to be separated is vaporized in a hot injector and then carried through a chromatographic column by a stream of an inert gaseous mobile phase (carrier gas) [60]. The sample components are separated based on differences in the solutes’ vapor pressures and/or the intensity of their interactions with the stationary phase interactions, causing what is known as retention. There are two types of GC stationary phases available: gas-solid chromatography (GSC) in which a solid adsorbent serves as the stationary phase, and gas-liquid chromatography (GLC) in which a thin-film liquid stationary phase is coated onto the wall of a capillary column or spread on an inert support. Nowadays, most environmental analyses of POPs are conducted using coated GLC capillary columns for chromatographic separation.

Figure 10. Schematic of a gas chromatograph.

Models describing the retention [61] and retention index [62] values for specific compounds in GLC systems have been presented in the literature. In brief, non-polar stationary phases interact with solutes via weak dispersive and induced dipole forces;

the retention of analytes on a non-polar phase is therefore largely determined by their vapor pressure, which is strongly dependent on the column temperature. In contrast, polar stationary phases have functional groups that can form strong interface interactions such as dipole-dipole, dipole-induced dipole, ion-dipole or ion- induced dipole interactions. Consequently, analyte retention on polar columns is governed by both vapor pressure and solute-stationary phase interactions. For complex environmental samples, especially those that contain compounds with large numbers of isomers (e.g. dioxins, PCBs, or PBDEs…) having somewhat identical mass spectra, GC separation is an essential component of the analytical chain and has a profound impact on the likelihood of successfully identifying and quantifying each individual isomer.

Detector Injector

Column

Column oven Syringe

(28)

3.2 Summary of Paper III

3.2.1 New stationary phases with specificity for dioxins

Paper III describes a study in which several novel stationary phases were evaluated based on their ability to separate the 136 tetra- to octaCDD/F congeners with special emphasis on the seventeen 2,3,7,8-substituted ones. The 74 mono- to tri-CDD/F congeners were not included in the study because several of them are not commercially available and they are of limited toxicological interest. The analysis was performed by injecting standard mixtures of dioxins directly into the GC/HRMS instrument. In total, 22 columns were compared. For six columns, we obtained data by conducting experiments, while data for the remaining 16 were obtained from the scientific literature (Table 3). Ionic liquids consist of asymmetrically substituted N or P cations (e.g. imidazolium, pyrrolidinium, or pyridinium ions) counterbalanced with inorganic anions (e.g. Cl^-, PF^6-, or BF^4-) [63-68]. The ionic liquid nature such as extreme polarity and diverse solvation interactions (hydrogen bonds, dispersion interactions, interactions with  electrons, etc.) makes them capable of separating a wide range of compounds. Ionic liquid stationary phases have been commercially available since 2009 and have found applications in the analysis of biodiesels [69], fatty acid methyl esters [67, 70], fragrances [70], and in separating some PCBs and PAHs [63]. However, they have not previously been used to analyze dioxins. We therefore included three ionic liquid columns in our study, along with two shape- selective columns (one coated with a liquid crystal, LC-50, and one with cyclodextrins, DEXcst) and one highly efficient (low bleed) non-polar column (DB- XLB).

Of the six columns tested in our laboratories, the three ionic liquid columns (SLB- IL111, SLB-IL76, and SLB-IL61) achieved better chromatographic separation than the LC-50, DEXcst and DB-XLB columns (Table 3). The superiority of the ionic liquid columns was especially pronounced for the separations of TeCDD/Fs and PeCDD/Fs (Figures 11-13). The SLB-IL61 offered the best overall performance, successfully resolving 106 out of 136 compounds, and also resolving 16 of the 17 2,3,7,8- substituted PCDD/Fs. The SLB-IL111 and SLB-IL76 resolved or partially separated 100 congeners, but the SLB-IL111 separated more of the 2,3,7,8-PCDD/F congeners (14) than the SLB-IL76 (12). Overall, the most notable of the new columns were the SLB-IL111 and SLB-IL61. Of the columns whose performance was evaluated based on literature data, the only one that could match the ionic liquid columns was the Smectic liquid crystal column, which resolved 103 congeners and separated 12 2,3,7,8-substituted PCDD/Fs. Unfortunately, the Smectic column is no longer on the market.

(29)

Table 3. Performance of the tested columns (grey background) and several literature columns (white background) in the separation of 136 PCDD/F congeners. The best separations achieved for each class of compounds are those in boxes.

Column Source Phase PCDFs PCDDs PCDD/Fs 2,3,7,8-

PCDD/Fs Supelco polarity scale^c Ref ++ +- -- ++ +- -- ++ +- -- ++ +- --

DB-XLB Agilent Non-polar, proprietary 31 19 37 20 11 18 51 30 55 12 2 3 8⁺

DEXcst Restek Chiral, proprietary CD^a 30 19 38 18 11 20 48 30 58 7 4 6 ? LC-50 Restek Liquid crystal 28 13 46 21 10 18 49 23 64 10 2 5 ? SLB-IL61 Supelco Polar, Ionic liquid 53 12 22 34 7 8 87 19 30 15 1 1 61 SLB-IL76 Supelco Polar, Ionic liquid 50 13 24 27 10 12 77 23 36 12 0 5 76 SLB-IL111 Supelco Polar, Ionic liquid 54 14 19 27 5 17 81 19 36 13 1 3 111 DB-1 Agilent Non-polar, 100%

dimethyl

16 11 60 16 13 20 32 24 80 5 7 5 5 [71]

VF-Xms Agilent Non-polar, proprietary 30 12 45 22 8 19 52 20 64 14 2 1 8⁺ [72]

VF-5ms Agilent Non-polar, 5% phenyl 30 10 47 24 7 18 54 17 65 13 1 3 8 [72]

DB-5ms Agilent Non-polar, 5% phenyl 28 14 45 18 12 19 46 26 64 9 5 3 8 [72]

DB-5 Agilent Non-polar, 5% phenyl 17 11 59 15 9 25 32 20 84 6 5 6 8 [71]

5Sil MS Restek Non-polar, 5% phenyl 29 15 43 25 5 19 54 20 62 13 2 2 8 [73]

Dioxin2 Restek Non-polar, proprietary 21 16 50 13 11 25 34 27 75 7 6 4 8 [74, 75]

BPX-DXN SGE Non-polar, proprietary 23 13 51 21 7 21 44 20 72 10 4 3 8 [76]

Equity-5 Supelco Non-polar, 5% phenyl 20 18 49 23 5 21 43 23 70 7 6 4 8 [72]

DB-17 Agilent Semi-polar, 50% phenyl 26 20 41 19 6 24 45 26 65 11 3 3 21 [71]

DB-210 Agilent Polar, 50%

trifluoropropyl

18 22 47 16 9 24 34 31 71 9 3 5 34 [71]

DB-225 Agilent Polar, 50% cyanopropyl, 50% phenyl

25 26 36 25 10 14 50 36 50 10 3 4 40 [71]

CPS-1 Discont.^b Polar, 75% cyanopropyl, 25% phenyl

43 9 35 28 4 17 71 13 52 11 2 4 60 [71]

SP-2331 Supelco Polar, 90% cyanopropyl, 10% phenyl

44 10 33 24 5 20 68 15 53 13 2 2 76 [71]

CP-Sil 88 Agilent Polar, 100% cyanopropyl 42 11 34 24 7 18 66 18 52 12 3 2 81 [71]

Smectic Discont. Liquid crystal 42 17 28 28 16 5 70 33 33 12 0 5 ? [71]

a Cyclodextrin added in 14% cyanopropylphenyl/86% dimethyl polysiloxane b Production discontinued

c Based on McReynolds constants and normalized against the Supelco SLB-IL100, which is similar in polarity to TCEP (1,2,3-tris[2- cyanoethoxypropane]), the most polar of the traditional stationary phases [65]

8⁺ means the polarity is slightly greater than 8

++ Peak well separated, valley 85-100%; +- Peak partially separated, valley 5 - 85%; -- Peak coeluted