Environ. Sci. Technol. 2008, 42, 5202–5209
Screening Chemicals for the Potential to be Persistent Organic Pollutants: A Case Study of Arctic Contaminants TREVOR N. BROWN AND FRANK WANIA* Department of Chemistry and Department of Physical and Environmental Sciences, University of Toronto Scarborough, 1265 Military Trail, Toronto, Ontario, Canada M1C 1A4
Received February 13, 2008. Revised manuscript received April 23, 2008. Accepted April 29, 2008.
A large and ever-increasing number of chemicals are used in commerce, and researchers and regulators have struggled to ascertain that these chemicals do not threaten human health or cause environmental or ecological damage. The presence of persistent organic pollutants (POPs) in remote environments such as the Arctic is of special concern and has international regulatory implications. Responding to the need for a way to identify chemicals of high concern, a methodology has been developed which compares experimentally measured properties, or values predicted from chemical structure alone, to a set of screening criteria. These criteria include partitioning properties that allow for accumulation in the physical Arctic environment and in the Arctic human food chain, and resistance to atmospheric oxidation. At the same time we quantify the extent of structural resemblance to a group of known Arctic contaminants. Comparison of the substances that are identified by a mechanistic description of the processes that lead to Arctic contamination with those substances that are structurally similar to known Arctic contaminants reveals the strengths and limitations of either approach. Within a data set of more than 100,000 distinct industrial chemicals, the methodology identifies 120 high production volume chemicals which are structurally similar to known Arctic contaminants and/ or have partitioning properties that suggest they are potential Arctic contaminants.
Introduction Considering the large number of chemicals in commerce, the complication of varied and unknown chemical degradation pathways in the environment, and the highly selective nature of most analytical detection systems, it is conceivable, if not highly likely, that the majority of contaminants present in the environment, in wildlife, and in humans, remain unidentified. Even though there are more than 82,000 substances covered by the U.S. Toxic Substances Control Act (TSCA) (1), the U.S. Centers for Disease Control monitor less than 120 substances in human blood (2). Contaminants whose environmental occurrence has not previously been documented are neither detected, studied, monitored, nor evaluated, let alone regulated. For example, even though synthesized and used for more than half a century, the ubiquitous presence of perfluorinated alkyl-compounds in * Corresponding author
[email protected]. 5202
9
phone:
+1-416-287-7225;
e-mail:
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 14, 2008
people, wildlife, and the physical environment was not recognized until a decade ago. Now these substances are of the highest priority to chemical regulators around the world (3). Discovery of previously undetected environmental contaminants is rarely the result of a directed search, but often owes much to serendipity. Clearly, there is a need for a more rational approach with a greater promise to succeed in identifying undetected contaminants. Based on a quantitative understanding of the relationship between chemical properties and environmental behavior, it is now becoming possible to identify the properties of substances that qualify as contaminants of concern. For example, we now can predict what chemical characteristics make an organic substance susceptible to accumulation in the Arctic physical environment (4). Similar approaches seek to constrain the properties of substances bioaccumulating in human food chains (5). It should thus be possible to screen the multitude of chemicals of commerce for those that do possess such properties, and which therefore are most likely to be found in environmental and human tissue samples. Prioritizing chemicals with respect to environmental concern has been an ongoing effort for several decades. Muir and Howard (6) provide an in-depth review of screening methods and projects aimed specifically at identifying persistent organic pollutants (POPs). The motivation for such screening has intensified in response to the United Nations Environment Program’s Stockholm Convention on POPs (7), which calls for the identification and global regulation of chemicals that fulfill the four criteria of persistence, bioaccumulation, long-range transport, and toxicity. Nearly all screening initiatives include parameters that address these criteria when searching for new potential POPs (6). The task of developing an effective screening method is difficult and complex. A method must be highly discriminatory to eliminate the many chemicals that are of little concern, and must simultaneously identify those chemicals which constitute a large hazard. In order to be considered effective a method must demonstrate a strong selectivity for those chemicals which cause the highest concern, capturing as many high risk chemicals and as few low risk chemicals as possible. We present here an approach which is specifically designed to be highly selective for chemicals which may become Arctic contaminants. The rationale is that traditional dietary habits place Northern indigenous residents among the human subpopulations most vulnerable to elevated exposure to POPs (8), and many of the traditional POPs are of especially high concern to this specific population.
Methods Outline of the Approach. The approach consists of two parallel screening methodologies: one methodology screens chemicals based on substance properties and the other screens chemicals based on a structural profile of known Arctic contaminants. Since substance properties are determined by molecular structure, both methodologies might be expected to flag the same chemicals as being of high concern. However, comparing the sets of substances identified by either methodology can provide valuable insight into their relative strengths and limitations. The pathway a chemical must follow to become a contaminant in the upper trophic levels of the Arctic food chain is circuitous; emission, long-range transport, deposition, exposure, bioconcentration, and biomagnification all play a role. The primary screening methodology applied here, and in most screening exercises, examines each step along this pathway and defines a set of criteria which chemicals 10.1021/es8004514 CCC: $40.75
2008 American Chemical Society
Published on Web 06/11/2008
must meet. Here we rely on three such criteria: the chemical must have distribution properties that allow it to reach the Arctic marine food chain, it must have distribution properties that allow it to bioaccumulate in humans, and it must be sufficiently persistent in the atmosphere to reach the Arctic without degrading. This methodology rests on two main assumptions: the first is that the potential to accumulate in the Arctic marine food chain depends on chemical distribution, meaning chemicals with certain combinations of partitioning properties will have a higher potential to accumulate, and the second assumption is that long-range transport happens mainly via the atmosphere; and therefore atmospheric oxidation is the primary factor limiting longrange transport to the Arctic. Although a chemical must also be resistant to metabolism to achieve high levels in top predators of the Arctic food chain, we do not include a metabolic criterion because of the difficulty of predicting susceptibility to metabolism. The second screening methodology rests upon the observation that chemicals which have already been identified as contaminants in the higher trophic levels of the Arctic food chain must have the correct combination of substance properties to traverse the entire pathway from emission to human. A chemical profiling method is employed which compares a chemical’s structure to the structures of known Arctic contaminants, and those chemicals which match the structural profile are assumed to have a higher likelihood of being Arctic contaminants. Cases where these two paradigms overlap and where they differ are used to identify possible false positives and negatives, and to assist in classification. These screening methodologies test a chemical for the potential to be an Arctic contaminant, but to be a risk the chemical must also be emitted. Ideally usage profiles and emissions data should be consulted but this is impractical for a large number of chemicals and the data are largely unavailable, so it is assumed that chemicals with a higher production volume will have a greater likelihood of significant emissions. Compilation of Data. Identifying information (CAS number, SMILES string), physical-chemical properties (molar mass, octanol-water partition coefficient KOW, Henry’s law constant, vapor pressure, aqueous solubility), and the atmospheric oxidation half-lives for 105,584 individual chemicals were compiled and calculated using the extensive CAS/SMILES string database of the EPISuite software package (9). These data served in the first screening method. In all cases experimental values were used where available. It should be noted that EPISuite assumes that all molecules are in their neutral form, so it is not possible to assess the effects of any dissociation equilibria. Two separate sets of partitioning properties were calculated from these data, with the same log KOW used in both cases. In the first set (“HLC data set”) a value for the air-water partition coefficient KAW was derived from the Henry’s Law constant (KAW ) HLC/ (RT)) and a value for the octanol-air partition coefficient KOA was obtained by applying a thermodynamic cycle with the KOW (KOA ) KOW/KAW). If the Henry’s Law constant was unavailable then it was estimated using the ratio of vapor pressure and aqueous solubility. In the second set of partitioning properties (“VP data set”) the vapor pressure was used to estimate a value for the KOA based on an empirical equation from Xiao et al. (10), then a value for KAW was calculated from KOW (KAW ) KOW/KOA). Ideally experimental values should be used for all partitioning properties and OH oxidation rates, because the applicability of the EPISuite prediction methods to a very wide range of organic compounds is untested and doubtful (11–13). However, the list of organic substances with reliable experimental data is much smaller than the number of chemicals with SMILES strings in the EPISuite database.
Experimental values are also much less likely to be available for chemicals which have not already been identified as harmful, and screening will therefore not identify any new potential POPs. Properties predicted from chemical structure alone, such as those available in the EPISuite database, are of limited and unknown accuracy but at this time they are the most extensive available and can help identify high concern chemicals which have not yet been closely examined. For the second methodology based on structural resemblance, a list of 86 known Arctic contaminants was compiled. This list (Table S1), references for concentrations found in Arctic biota, the procedure and the criteria used in compiling the list, as well as a short description of the six major structural classes of known Arctic contaminants is given in the Supporting Information. SMILES strings were found in the EPISuite database or composed and used to quantify the 86 chemicals’ structural properties. Five different lists were used to determine if a chemical has a high production volume (HPV): the Canadian Domestic Substances List which identifies substances having an annual production volume greater than 1,000 t (metric tonnes) (14), the U.S. EPA’s HPV Challenge Program which includes chemicals that are produced or imported into the United States in quantities of 454 t (1 million pounds) or more per year (15), the HPV list compiled by the European Chemical Bureau’s European Chemical Substances Information System which includes chemicals produced in or imported to the European Union in quantities of 1,000 t or more (16), the OECD’s list of HPV chemicals produced in quantities of 1,000 t or greater in the EU or another member country (17), and any chemical listed under the U.S. Toxic Substances Control Act (TSCA) produced in a quantity of 454 t (1 million pounds) or more in any year for which data are available (18). Chemicals were further checked against three databases of registered pesticides: the Canadian Domestic Substances List (14), the U.S. EPA’s list of registered pesticides (19), and the World Health Organization’s list of current use pesticides (20). Screening for Long Range Transport to the Arctic and Bioaccumulation in Humans. By linking global transport calculations (21) with a food chain bioaccumulation model (5), Czub et al. (8) have delineated the combination of partitioning properties KAW, KOW, and KOA, which result in a high potential for organic chemicals to be transported to the Arctic and to accumulate in the Arctic human food chain. They illustrated these combinations by plotting an Arctic Contamination and Bioaccumulation Potential (AC-BAP) after seventy years for a multitude of hypothetical perfectly persistent chemicals as a function of the chemical partitioning space defined by KOA and KAW (Figure 1). The area of elevated AC-BAP is defined here as exceeding at least 10% of the maximum AC-BAP70 value (red outline in Figure 1). Over half of the 86 known Arctic contaminants fall within the defined area of elevated AC-BAP. Some, such as heavily chlorinated PCDD/Fs and PCBs, fall outside of the red outline because the global transport calculations predict a low potential to reach the Arctic for particle-bound substances with log KOA values above 11 (21, 8). This subject is explored further below. The area of elevated AC-BAP comprises the following thresholds: log KOW g 3.5 log KOA g 6 0.5 g log KAW g -7 log KAW e -1.78 × log KOA + 14.56
(1)
The selection of the 10% line is arbitrary and not very conservative, but was deemed to be a compromise between selectivity and the possibility of excluding chemicals which should be flagged as of concern. Using the 1% line instead VOL. 42, NO. 14, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5203
FIGURE 1. Red outline delineates chemical partitioning property combinations that result in a high (i.e., in excess of 10% of the maximum) potential for accumulation in the Arctic physical environment and the Arctic human food chain as described by Czub et al. (8). Values for log KOA and log KAW of 86 known Arctic contaminants (experimental values referenced in the SI, and predicted with SPARC (22) if not available) are also shown. would greatly improve the agreement between the model results and the monitoring data, as all but two of the chemicals fall within the line indicating 1% of the maximum AC-BAP70. However, doing so would essentially neglect the effect of long-range transport because the 1% line is nearly equivalent to the area of elevated Arctic bioaccumulation (8). The contours of the plot of AC-BAP suggest that the elevated area extends below log KAW ) -5. There are computational reasons for not extending the plot below this line but no mechanistic reasons, so we have extrapolated two log units into this area to ensure we capture as many potential Arctic contaminants as possible. By checking if the partitioning properties of an organic chemical fall within the red outline in Figure 1, the potential for environmental transport to the Arctic and potential for bioaccumulation in the Arctic food chain are being screened for simultaneously. Whereas the screening process is simple, the definition of the area of elevated AC-BAP is based on detailed mechanistic models and represents a sophisticated and effective exclusion method. Using an enclosed chemical space has intuitively a higher power of discrimination than using cutoff values for only one of the three partitioning properties, such as the common use of a log KOW threshold of 5 (6), which in itself may not even be valid (eq 1, (5)). The partitioning properties of the entire chemical data set were screened using the thresholds of eq 1. Both the HLC data set and the VP data sets were checked and chemicals were considered to meet the criterion of elevated AC-BAP if either data set placed the chemical within the red boundaries. Screening for Persistence in Air. In the screening step involving the partitioning space, chemical persistence is not being considered. To filter out chemicals which are quickly degraded in the atmosphere and therefore have low longrange transport potential, an atmospheric oxidation half-life cutoff is used. Hydroxyl radicals are the primary atmospheric oxidant for most substances (23), so a hydroxyl radical atmospheric oxidation half-life (OH thalf) was used as the screening parameter. Some experimental OH thalf values are available in the EPISuite database and the rest were predicted 5204
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 14, 2008
using EPISuite (9). There will be contributions from other oxidants to the total atmospheric oxidation rate but the reaction rates with OH are the most reliable (24, 25). Reaction with other oxidants will only increase the overall rate of oxidation, so OH thalf values may be viewed as the maximum half-life for a chemical in air. A cutoff value of two days was used and is consistent with the Stockholm Convention (7) and Muir and Howard (6). We should caution that this criterion could eliminate chemicals that may undergo longrange transport to the Arctic in the oceans, or have a longer oxidation half-life when sorbed to atmospheric particles. Screening for Resemblance with the Structural Profile of Known Arctic Contaminants. There is a distinct set of combinations of partitioning and persistence properties which meet all the criteria of the first screening methodology, but these properties are themselves governed by chemical structure. This implies that there may be a limited number of possible combinations of structural features that give rise tothepropertiesofanArcticcontaminant.Astructure-property relationship (SPR) was developed to predict if a chemical has the potential to be an Arctic contaminant based on chemical structure. The vast majority of SPRs rely on multiple linear regressions but that approach is impractical in this case as no continuous scale of the potential to become an Arctic contaminant exists that can be used to parametrize a regression equation. Instead, the list of 86 known Arctic contaminants is compared to the full list of chemicals and deviations from a ”typical” molecule are quantified for 3 structural parameters: halogenation, internal connectivity, and molecular size. A detailed description of the derivation and application of the structural profile of Arctic contaminants is provided in the Supporting Information.
Results and Discussion Preliminary Screening. Figure 2 shows a schematic representation of how the two screening methodologies classify the screened chemicals. From the list of 105,584 chemicals, 16,888 chemicals fall within the boundaries of elevated ACBAP using either the HLC or VP data sets (see Figure S1 in the Supporting Information). A total of 13,905 chemicals are identified by the HLC data set, 14,957 are identified by the VP data set, and 11,974 chemicals are identified by both data sets. Neither data set shows a significant bias toward overor underestimating the log KAW or log KOA values so it can not be immediately determined which data set is more accurate. A total of 16,151 chemicals have experimental or estimated OH thalf values greater than two days. There are 2,025 chemicals which both fall within the chemical space and have an OH thalf value greater than two days. There are 3,088 chemicals which match the structural profile of known Arctic contaminants. More than two-thirds of the chemicals screened (74,016 chemicals or 70.1%) did not meet any of the three screening criteria and are unlikely to reach the Arctic or bioaccumulate in humans in this region. A further 12,669 (12.0%) chemicals do not match the structural profile of known Arctic contaminants and fall outside of the area of elevated AC-BAP, but are persistent in air. These chemicals may possibly reach the Arctic environment but because the area of elevated ACBAP is strongly influenced by the potential to bioaccumulate they are unlikely to lead to significant human exposure. An additional 14,608 (13.8%) chemicals fall within the area of elevated AC-BAP but are not persistent in air and do not match the structural profile of known Arctic contaminants. It is unlikely that a significant fraction of these chemicals will reach the Arctic and become contaminants. Eliminating these three groups of chemicals from further consideration excludes 101,293 chemicals (95.9%). Chemicals Matching the Structural Profile of Known Arctic Contaminants. There are 2,266 chemicals (2.1%)
FIGURE 2. Schematic diagram of the process used to screen the 105,584 chemicals in the EPI Suite database for the potential to be contaminants which accumulate in the Arctic food chain. which, despite matching the structural profile of known Arctic contaminants, fail to meet one or both of the other two screening criteria. Of these there are 554 (0.5%) that fail to meet both the persistence in air and chemical space criterion and 255 (0.2%) that fall within the area of elevated AC-BAP but are not persistent in air. Due to low air persistence these chemicals are unlikely to reach the Arctic, however it is possible that their degradation products could be of concern. The oxidation products could be more persistent and their partitioning properties may be different, but predicting the major oxidation products of the 809 chemicals identified is beyond the scope of this study. An additional 1,457 (1.4%) chemicals match the structural profile of known Arctic contaminants and are persistent in air but fall outside of the area of elevated AC-BAP. Figure 3 shows the location in the chemical space of all 3,088 chemicals which match the structural profile of known Arctic contaminants. There are thus large numbers of substances that are structurally similar to known Arctic contaminants, but nevertheless do not possess the right combination of partitioning properties to become Arctic contaminants based on the model predictions and assumptions used in this study. A closer inspection of Figure 3 reveals three types of chemicals outside of the red boundaries. It is enlightening to explore what prevents these chemicals from being classified as potential Arctic contaminants, because it could possibly reveal the limitations of both the screening and the chemical profiling technique. Type 1: Volatile Chemicals. Many chemicals are located to the upper left of the high AC-BAP area, indicative of high log KAW (>0.5) and low log KOA (
