Article pubs.acs.org/est
Partitioning of Polybrominated Diphenyl Ethers to Dissolved Organic Matter Isolated from Arctic Surface Waters Maya L. Wei-Haas,† Kimberly J. Hageman,‡ and Yu-Ping Chin*,† †
School of Earth Sciences, The Ohio State University, 125 South Oval Mall, Columbus, Ohio 43210 United States Department of Chemistry, University of Otago, Dunedin 9016, New Zealand
‡
S Supporting Information *
ABSTRACT: Polybrominated diphenyl ethers (PBDEs) are a class of brominated flame retardant that is distally transported to the Arctic. Little is known about the fate of PBDEs in Arctic surface waters, especially in the presence of dissolved organic matter (DOM). DOM has been shown to interact with hydrophobic organic contaminants and can alter their mobility, bioavailability, and degradation in the environment. In this study, the partitioning of six PBDE congeners between Arctic DOM (isolated via solid phase extraction) and water was measured using the aqueous solubility enhancement method. Measured dissolved organic carbon (DOC)−water partition coefficient (KDOC) values were nearly an order of magnitude lower than previously reported values for the same PBDE congeners in soil or commercial organic matter, ranging from 103.97 to 105.16 L kg−1 of organic carbon. Measured results compared favorably with values calculated using polyparameter linear free energy models for Suwannee River fulvic acid. Log KDOC values increased with increasing PBDE hydrophobicity. Slightly lower than expected values were observed for the highest brominated congeners, which we attribute to steric hindrance. This study is the first to comprehensively measure KDOC values for a range of PBDE congeners with DOM isolated from Arctic surface waters.
■
INTRODUCTION Polybrominated diphenyl ethers (PBDEs) are a class of brominated flame retardant used in a variety of common household goods such as foam and electronics.1 Recent concern about the production of PBDEs emerged in the last several decades due to mounting awareness of their toxic effects to the liver and thyroid and on neurodevelopment and endocrine functioning.2,3 Further, PBDEs can undergo distal transport to the Arctic and are ubiquitous in sediments, soil, water, and air.2,4−6 A multitude of studies have documented elevated concentrations of PBDEs and its metabolites in a variety of Arctic wildlife.4,7−11 A summary of many of these studies can be found at the Arctic Monitoring and Assessment Program Web site.12 Some congeners are also believed to biomagnify, so although the concentrations in water may not directly be toxic, the concentrations of PBDEs are magnified at alarmingly high rates for predators at the top of the food chain.4,13 Yet little is currently known about the fate of PBDEs in the aqueous Arctic environment. The large reported octanol−water partition coefficient (Kow) values of PBDEs (ranging from 106 to >108 for tri- to decabromodiphenyl ethers)14 make these substances highly susceptible to partitioning to dissolved organic matter (DOM). This process may alter the transport, bioavailability, and overall fate of these substances in the aquatic environment.15−18 Oligotrophic lakes that are high in DOM and low in suspended © 2014 American Chemical Society
solids are abundant throughout the Arctic, making DOM partitioning a significant process influencing the fate of hydrophobic organic contaminants (HOCs). Extensive previous research on the partitioning of other HOCs to DOM suggests that the magnitude of the dissolved organic carbon (DOC)-water equilibrium partition coefficient (KDOC) for a given HOC depends on the DOM’s composition, for example, aromaticity, molecular weight, and H/C and O/C atomic ratios, etc.15,19−24 However, Akkanen et al.,15 observed no such correlation for one BDE congener (2,2′,4,4′ tetrabromo diphenyl ether), which yields uncertainty in the relationship between PBDEs and DOM properties. In general, KDOC increases with greater aromaticity of the DOM and hydrophobicity of the HOC,20,25 yet limited evidence suggests that this might not be true for PBDEs. Research to date that has examined the binding of PBDE congeners to DOM has been limited in scope. The majority of PBDE KDOC values reported in the literature were measured using soil-extracted organic matter,26 geologic materials (e.g., Leonardite humic acid),27 or commercially available humic acid (HA).28 Multiple studies have previously demonstrated that commercial humic acids are poor surrogates for aquatic Received: Revised: Accepted: Published: 4852
December March 17, March 28, March 28,
6, 2013 2014 2014 2014
dx.doi.org/10.1021/es405453m | Environ. Sci. Technol. 2014, 48, 4852−4859
Environmental Science & Technology
Article
DOM.20,29,30 In addition, soil humic acids consistently yield KDOC values for HOCs that are considerably greater than those measured using aquatic DOM.19 To our knowledge, only two other studies of PBDE partitioning have utilized DOM isolated from natural waters, yet both of these studies only measured PBDE KDOC at one DOM concentration.15,31 Similarly, Wang et al.26 reported KDOC values for PBDEs that were indirectly determined from sediment sorption isotherms. Therefore, there remains a necessity for rigorously measured PBDE partition coefficients using DOM extracted from aquatic systems. Moreover, no previous studies have evaluated the partition coefficients using DOM isolated from Arctic surface waters to examine the fate of distally transported PBDEs. Single and polyparameter linear free energy relationships (spLFERs and pp-LFERs, respectively) have been developed to predict HOC partitioning to DOM but have achieved varying success. Sp-LFERs use direct linear associations to predict partition coefficients from specific compound properties such as KOW.25 Sp-LFERs are straightforward to use but are limited in scope because they do no consider the influences and complexities of solute and sorbent properties on partitioning. Pp-LFERs offer a solution to this problem by utilizing multiple fitting parameters to describe specific and nonspecific solutesorbent interactions.32,33 With rigorous parametrization, ppLFER models are a powerful means to predict the sorption behavior of HOCs in the environment. Until recently (e.g., Stenzel et al.),34 calculating PBDE KDOC values using pp-LFER models was impossible due to a lack of solute descriptors. In our study, KDOC was measured by the solubility enhancement method for six PBDE congeners with DOM isolated from Arctic surface waters. KDOC for three of the PBDEs (BDE-47, BDE-99, and BDE-153; Figure 1) that are commonly detected in
The partition coefficients reported in this study serve as an important step in determining the fate and bioavailability of PBDEs in the aqueous Arctic environment.
■
EXPERIMENTAL SECTION Study Site and DOM Isolation. Water samples were bailed at three sites near Toolik Field Station, located on the North Slope of the Brooks Range in Arctic Alaska (Supporting Information (SI) Section S1; Figure S1). The three collection sites were Toolik Lake, Imnavait Creek, and a meltwater pool near the Sagavanirktok (SAG) River. DOC of whole waters collected from these sites ranged from ∼3 mg C/L (Toolik Lake) up to ∼20 mg C/L (SAG seep). Water samples (60 L − 80 L per site) were immediately acidified to pH ∼3 and then filtered using High Capacity Ground Water 0.45 μm Versapor membrane filters (acrylic copolymer on a nonwoven support; Pall Gelman, Port Washington, NY), which were precleaned with a minimum of 5 L of Milli-Q water (pH ∼3). DOM was then extracted from the acidified water samples using Agilent Mega Bond Elute PPL solid phase extraction (SPE) cartridges (5 g styrene divinylbenzene solid phase; Santa Clara, CA) using methods modified from Dittmar et al.35 Previous research indicates that PPL SPE has the greatest capacity to isolate compounds of a large range of polarities and was used in the present study in an attempt to capture a representative sample of organic matter from Arctic surface waters.35 Details regarding the isolation procedure can be found in the SI (Section S3). Chemicals. The six investigated PBDE congeners, 2,4,4′tribromodiphenyl ether (BDE-28), 2,2′,4,4′-tetrabromodiphenyl ether (BDE-47), 3,3′,4,4′-tetrabromodiphenyl ether (BDE-77), 2,2′,4,4′,5-pentabromodiphenyl ether (BDE-99), 2,2′,4,4′,5,5′hexabromodiphenyl ether (BDE-153), and 2,3,3′,4,4′,5,6heptabromodiphenyl ether (BDE-190), (Figure 1) were obtained as individual standards (50 μg mL−1 in isooctane) from Accustandard (New Haven, CT). Select physicochemical properties of each congener are given in the SI (Table S1). Highresolution gas chromatography grade hexane (Omnisolv) was used for liquid−liquid extractions of PBDEs from water and was obtained from EMD Millipore (Billerica, MA). HPLC-grade methanol was obtained from Spectrum Analytical (Agwam, MA). All water used in this study was ultrapure deionized water from a Milli-Q water system (EMD Millipore). ACS grade sodium hydroxide (NaOH) and hydrochloric acid (HCl) were used for pH adjustments and were obtained from GFS Chemicals (Columbus, OH) and EMD Millipore, respectively. DOM Characterization. Absorbance of DOM solutions was measured from 200 to 600 nm on an Agilent Cary 1 UV−vis spectrophotometer (Santa Clara, CA) and baseline corrected for Milli-Q water. Molar absorptivity at 280 nm (ε280) was calculated as the UV absorbance at 280 nm normalized to the DOC concentration in mol C L−1. Fluorescence intensity was analyzed at 240−450 nm excitation and 300−600 nm emission on an Agilent Cary Eclipse spectrofluorometer (Santa Clara, CA) with a 1 cm path length. Samples were diluted to minimize innerfilter/light screening effects (Aλ254 < 0.05).36 Excitation emission matrices (EEMs) were corrected against a Milli-Q blank and normalized to the Raman peak area and DOC of each sample. All EEMs were corrected for instrument variation using instrumentspecific correction files.37 Because we lacked the number of samples to create a robust parallel factor analysis (PARAFAC) model specific to our instrument, we opted to descriptively analyze our EEMs. The fluorescence index (FI) is a complementary measurement to ε280 and was calculated as the
Figure 1. Molecular structures of investigated PBDE congeners.
environmental samples were measured using DOM isolated from three water bodies on the North Slope of Alaska and two International Humic Substance Society (IHSS) standards, Suwannee River natural organic matter (NOM) and Pony Lake fulvic acid (FA). Suwannee River NOM and Pony Lake FA will serve as a reference for allochthonous and autochthonous “end member” compositions. KDOC values were measured for the remaining three congeners (BDE-28, BDE-77, and BDE-190) with Suwannee River and/or one Arctic DOM isolate to elucidate the role of analyte structure in the partitioning process. These values were compared to pp-LFER calculated KDOC values using solute descriptors established by Stenzel et al.34 and phase descriptors measured by Neale et al.32 for Suwannee River FA. 4853
dx.doi.org/10.1021/es405453m | Environ. Sci. Technol. 2014, 48, 4852−4859
Environmental Science & Technology
Article
ratio of emission intensity at 470 nm to that at 520 nm at a fixed excitation wavelength of 370 nm.36,37 The Suwannee River NOM and Pony Lake FA standards served as allochthonous and autochthonous end-member DOM and were obtained from the IHSS (St. Paul, MN).38 The former is aquatic DOM derived from higher plant precursors and has a highly aromatic composition. In contrast, Pony Lake FA originates from a highly eutrophic Antarctic lake devoid of higher plants. Since DOM in most environments is of an intermediate composition, these standards allowed us to assess how Arctic DOM composition and measured PBDE KDOC values compared to these geochemical end members. Details about isolation procedures and site descriptions for the IHSS DOM standards may be found online.38 Choice of Method for Measuring KDOC of PBDEs. A number of methods have been developed for measuring KDOC, including ones that utilize equilibrium dialysis,15,39 solid phase microextraction (SPME),26,40,41 polydimethlysiloxane (PDMS) disks,42 and solubility enhancement.17,19,20,23 The choice of approach is largely dictated by the properties of the analyte and DOM. For example, equilibrium dialysis is a static equilibrium method ideally suited for high molecular weight DOM;39,43 however, lower molecular weight moieties can potentially “leak” across the membrane because a significant fraction of the DOM have molecular weights less than the dialysis membrane cutoff, making interpretation of the data possibly difficult.44,45 KDOC values were measured using the solubility enhancement method, which has demonstrated to be consistent and robust for HOCs with log KOW ≥ ∼4.17,19,20,30 The effectiveness of the solubility enhancement approach can be influenced by DOC properties and concentration when used with more polar compounds. However, these properties have minimal influence on the results when used with highly hydrophobic compounds, like PBDEs (see SI, Table S1 for PBDE KOW values). In addition, the solubility enhancement approach is superior for analytes, such as PBDEs, that tend to sorb to vessel walls, since analyte loss can compromise measured KDOC values. This is particularly important for methods, such as those involving SPME or PDMS, that rely on mass balance calculations, unless these approaches are used in the “dosing” mode where the analytes are preloaded on to the solid phase. Yet we were not equipped to conduct the experiments via SPME or PDMS dosing without considerable effort to test and ground truth these methods. In contrast, the solubility enhancement method uses aqueous analyte concentrations that are well above their solubility in pure water and therefore eliminates the issue of loss due to sorption. We found that PBDEs undergo significant sorption to vessel walls and Teflon septa in aqueous solutions below solubility over periods of 1 day (SI, Section S4, Figure S3). The drawback of the solubility enhancement method, however, is that it is labor intensive and requires careful handling and measurement of the actual dissolved compound. Solubility Enhancement Experiments. Stock solutions (∼25 mg C L−1) of freeze-dried DOM from each site were made by dissolving the DOM in Milli-Q water, adjusting the pH to 7−8 with ∼2 M NaOH, and filtering the solutions through a combusted Whatman (Piscataway, NJ) 0.7 μm GF/F filter. DOM stock solutions were diluted to six concentrations, ranging from 2 to 25 mg C L−1 in Milli-Q water. DOM, quantified as DOC, was measured using a Shimadzu total organic carbon analyzer, TOC-VCPN (Columbia, MD). Individual PBDE congeners dissolved in isooctane were plated on the bottom of 20 mL amber glass vials and the solvent was allowed to evaporate;
the plated PBDE mass exceeded that required to reach maximum aqueous solubility for each congener by a factor of ∼2.5 for BDE28 and a factor of ∼8 for higher brominated congeners (for aqueous solubility values see SI, Table S1). Each vial was then filled with 10 mL of one of the six DOM solutions. The experiment was also repeated in the absence of DOM (Milli-Q water only) to measure the aqueous solubility for each of the PBDE congeners. No preservatives were added to avoid any possible side reactions. Although we acknowledge that some degradation of the DOM may occur during equilibration, DOC was not measured until after the full equilibration period, allowing us to account for any changes in concentration. Reaction vials were allowed to equilibrate for a minimum of 5 days (∼23 °C) with gentle daily manual agitation as determined from kinetics experiments (SI, Figure S4). The experiments were conducted at room temperature, since thermodynamic calculations suggest that temperature effects for HOC-DOM partitioning are small.46 An aliquot of each aqueous solution was then removed from the reaction vial and the PBDEs were extracted with hexane. Each hexane extract was spiked with 2,2′,4,4′,6,6′ hexachlorobiphenyl (PCB-155) as an internal standard. PBDE concentrations in the hexane extracts were determined using a Thermo Trace gas chromatograph (GC) Ultra (West Palm Beach, FL), with an electron capture detector (ECD). The GC was run in splitless mode with an injector temperature of 265 °C, a 2 min splitless time, and an ECD detector temperature of 350 °C. The PBDE congeners were separated using a 15-m fused silica column (RTX-1614, 5% phenyl methyl, 0.25 mm i.d. × 0.10 μm film thickness) and a 5-m Siltek-deactivated fused silica guard column (both from Restek, Bellefonte, PA). The GC oven was programmed to start at 100 °C, hold for 1 min, ramp to 300 °C at 15 °C per minute, and hold for 3 min. Sample concentrations were calculated using internal calibration curves. KDOC (units of L kg−1 organic carbon) was calculated by fitting the data to the following linear equation:19 Sw*/Sw = KDOC[DOC] + 1
(1)
where SW* is the apparent solubility (in the presence of DOM) and SW is the measured aqueous solubility of the HOC (in pure Milli-Q water). Thus, if the data is plotted in the form of SW*/SW vs [DOC], then the KDOC is the slope of the linear relationship (eq 1) and calculated based on the following equation: KDOC = [(Sw*/Sw ) − 1]/[DOC]
(2)
Measured aqueous solubility for each congener in Milli-Q water was within a factor of ∼2 times the reported literature values (SI, Table S1). More details about the solubility enhancement procedure and data processing may be found in the SI (Section S5). Calculating KDOC Values Using pp-LFER Models. The following pp-LFER model, known as Abraham’s solvation model, is used to predict partition coefficients in which both phases are condensed.47,48 log(K ) = eE + sS + aA + bB + vV + c
(3)
The upper and lower case letters denote solute and phase descriptors, respectively, for each of the following: E is the excess molar refraction (based on the refractive index of the compound at 20 °C), S is the polarizability, A is the hydrogen donor or acidity constant, B is the hydrogen acceptor or basicity constant, and V is the McGowan molar volume. Since PBDEs are not hydrogen donors, the term for this solute descriptor (A) was set 4854
dx.doi.org/10.1021/es405453m | Environ. Sci. Technol. 2014, 48, 4852−4859
Environmental Science & Technology
Article
to zero for KDOC calculations (SI, Table S2). Goss49 developed an alternative relationship that describes partitioning between any two phases: log(K ) = lL + sS + aA + bB + vV + c
Table 1. Summary of Results for the Spectral Characteristics of the Studied DOM
(4)
Toolik Lake Imnavait Creek SAG Seep Pony Lake FA Suwannee River NOM
In this model, the excess molar refraction (E) has been replaced by the L term, which is the hexadecane/air partition coefficient. Hereafter, we refer to eqs 3 and 4 as ESABV and LSABV models, respectively. These models are essentially interchangeable if both phases are condensed.49 The choice of model for PBDEs is largely dependent upon the availability and quality of calibrated terms. Stenzel et al.34 established solute descriptors for seven PBDE congeners based on GC retention times and liquid/liquid partition coefficients. These include four of the congeners analyzed in this study (BDE-28, BDE-47, BDE-99, and BDE153) in addition to 2,2′4,4′6-pentabromodiphenyl ether (BDE100), 2,2′,4,4′,5,6′-hexabromodiphenyl ether (BDE-154), and 2,2′,3,4,4′,5′,6′-heptabromodiphenyl ether (BDE-183). We calculated KDOC for these congeners using the ESABV and LSABV phase descriptors for Suwannee River FA that were determined based on the data presented in Neale et al.32 The training set for Suwannee River FA (n = 34) included a range of polar and nonpolar compounds (log KDOC = 1.29−3.53).32 The refractive index for the E term used in the ESABV model was calculated using ACD Chemsketch (Version 12.00; Toronto, Ontario) since PBDEs are solid at 20 °C and no experimental refractive index exists. All solute and phase descriptors used for KDOC calculations can be found in the SI (Table S2 and S3).
FIa
ε280 (L mol−1 cm−1)
1.53 1.50 1.42 1.46 1.26
230 232 265 294 400
FI is calculated as the ratio of emission at 470 to 520 nm at a fixed excitation of 370 nm.
a
DOM pool than the XAD stationary phase used for the Pony Lake DOM isolation. In depth discussion of these trends can be found in the SI (Section S7). Overall FI and ε280 results indicate that despite preferential isolation for different DOM moieties, our Arctic DOM samples and IHSS standards represent a diversity of organic matter composition present in surface waters. Trends in Measured KDOC Values. Suwannee River NOM, which has the greatest aromatic content of the studied DOM, has significantly larger KDOC for BDE-47 and -99 compared to KDOC values measured using the less aromatic Arctic isolates. But interestingly, we observed little correlation between the spectral properties of our DOM samples and partition coefficients (Figure 2). Previous studies on the binding of HOCs (polycyclic
■
RESULTS AND DISCUSSION DOM Characterization. The fluorescence spectroscopy analysis of the Arctic DOM and IHSS standards in the present study reflects a range of compositions (SI Figure S5). Qualitative interpretation of our EEMs reveals the presence of a blue-shifted fluorophore (P) in the Toolik Lake sample (maximum peak intensity ∼280ex 350em) and has previously been demonstrated to represent microbially derived (autochthonous) DOM (SI Figure S5).50 The DOM isolates from all other localities display prominent humic-like (allochthonous) fluorophores at 420− 450em from 320 to 350ex (H1) and 240−260ex (H2).50 FI and ε280 values have been used in past studies to assess the nature of the DOM precursor materials.36,37 Low FI and high ε280 are indicative of allochthonous or lignin-derived DOM (i.e., high aromaticity, molecular weight) whereas high FI and low ε280 reflect autochthonous or microbial precursors such as phytoplankton and prokaryotic organisms (i.e., low aromaticity, low molecular weight).36,51,52 Because exact FI values may vary between instruments due to differences in optical design and light source, these values are most meaningful when examined as a trend among multiple samples and standards.53 As expected, Suwannee River NOM had the lowest FI (1.26) and greatest ε280 (400 L mol−1 cm−1), indicating high allochthonous (aromatic) content (Table 1). Meanwhile, Pony Lake FA had a higher (1.46) FI and lower ε280 (294 L mol−1 cm−1), indicative of greater autochthonous or microbial influence (low aromaticity) (Table 1). The Arctic DOM FI (range: 1.42−1.53) and ε280 values (range: 230−265 L mol−1 cm−1) are more similar to Pony Lake FA than Suwannee River NOM. This is highly surprising given that these water bodies are located in the tundra, which is dominated by higher plants such as sedges and tussock. We suspect that our observations could result from the PPL SPE isolation since this method captures significantly more of the
Figure 2. Summary of measured log KDOC values for PBDE congeners using Arctic and IHSS standard DOM. DOM is listed in order of increasing aromatic character based on measured spectral properties. Error bars indicate ±95% confidence intervals (n = 22−54). Data for BDE-153 with Imnavait creek was not presented due to its high degree of scatter; the slope was indistinguishable from zero. Exact values for all KDOC are presented in Table 2.
aromatic hydrocarbons, PCBs, etc.) to DOM reported correlations between KDOC and DOM properties such as polarity (as quantified by O/C ratios), aromaticity, and molecular weight.17,19,20,22 However, a study conducted by Akkanen et al.15 corroborates our observations, reporting few significant correlations between the KDOC for BDE-47 and a wide variety of properties measured via spectroscopic techniques and 13CCMAS, NMR. One previous study reported greater partitioning for BDE-47 and -99 to HA compared to FA,31 but this is not surprising given that HOC partitioning to humic acids is typically higher than for fulvic acids or other aquatic DOM isolates.19 Our results indicate that aromaticity alone is insufficient to describe PBDE log KDOC and other factors, such as polarity and molecular weight, may prove significant. Furthermore, interactions between PBDEs and DOM appears to be complex and may involve multiple molecular interactions.15,20,24 4855
dx.doi.org/10.1021/es405453m | Environ. Sci. Technol. 2014, 48, 4852−4859
Environmental Science & Technology
Article
Log KDOC values for the analyzed congeners generally increased with increasing bromine substitution (Figures 2 and 3). However, log KDOC values for SAG Seep and Pony Lake are
Figure 4. Comparison of experimental (purple circles) to calculated KDOC values using the ESABV (red squares) and LSABV (green squares) pp-LFER models for Suwannee River FA. Previously determined KDOC values with aquatic DOM are displayed for comparison (orange ×; exact values can be found in the SI, Table S5). Error bars represent ±95% confidence intervals. When not displayed, error bars are smaller than the symbols. Exact values for all measured and calculated KDOC can be found in Table 2.
Figure 3. Solubility enhancement experiment results for three PBDE congeners (BDE-47, BDE-99, and BDE- 153) using SAG Seep DOM. The slope (m) of each line and correlation coefficient (r2) is listed for each congener. Assuming that Sw* = Sw when DOC is equal to 0 (i.e., the y-intercept equals 1), the slope of each line is equal to the KDOC (see eq 2). The dashed lines represent the ±95% confidence interval for each slope (BDE-99, n = 23; BDE-153, n = 27; BDE-47, n = 54).
compounds with DOM extracted from a variety of surface waters (log KDOC up to ∼6.9; n = 263 data points, 39 compounds).54 Thus, despite the lack of very hydrophobic compounds in the Neale et al.32 training set, both the ESABV and LSABV models for Suwannee River FA appear to accurately predict partitioning of compounds with a wide range of hydrophobicities, including PBDEs. In spite of this success there remains the need for additional reliable data sets to calibrate pp-LFERs for other types of DOM. The calculated PBDE KDOC values further indicate that despite the restriction of training set data to Suwannee River FA, the solute and phase descriptors accurately describe HOC partitioning to DOM originating from diverse aquatic sources. Similarly, Neale et al.32 observed that calculated KDOC values (ESABV) for several PCB congeners were comparable (
