Volume 6, Issue 2 p. 210-224
Critical Review
Free Access

Molecular size cutoff criteria for screening bioaccumulation potential: Fact or fiction?

Jon A Arnot

Corresponding Author

Jon A Arnot

Centre for Environmental Modelling and Chemistry, Trent University, 1600 West Bank Drive, Peterborough, ON, Canada K9J 7B8

Centre for Environmental Modelling and Chemistry, Trent University, 1600 West Bank Drive, Peterborough, ON, Canada K9J 7B8.Search for more papers by this author
Michelle I Arnot

Michelle I Arnot

Department of Pharmacology and Toxicology, 4219 Medical Sciences Building, University of Toronto, Toronto, ON, Canada

Search for more papers by this author
Don Mackay

Don Mackay

Centre for Environmental Modelling and Chemistry, Trent University, 1600 West Bank Drive, Peterborough, ON, Canada K9J 7B8

Search for more papers by this author
Yves Couillard

Yves Couillard

Environment Canada, Science and Technology Branch, Fontaine Building, 200 Sacré-Coeur Boulevard, Gatineau, QC, Canada

Search for more papers by this author
Drew MacDonald

Drew MacDonald

Environment Canada, Science and Technology Branch, Fontaine Building, 200 Sacré-Coeur Boulevard, Gatineau, QC, Canada

Search for more papers by this author
Mark Bonnell

Mark Bonnell

Environment Canada, Science and Technology Branch, Fontaine Building, 200 Sacré-Coeur Boulevard, Gatineau, QC, Canada

Search for more papers by this author
Pat Doyle

Pat Doyle

Environment Canada, Ecological Assessment Division, 4 Reeves Place, St. John's, NL, Canada

Search for more papers by this author
First published: 04 February 2010
Citations: 32


It has been asserted that, when screening chemicals for bioaccumulation potential, molecular size cutoff criteria (or indicators) can be applied above which no, or limited, bioaccumulation is expected. The suggested molecular size values have increased over time as more measurements have become available. Most of the proposed criteria have been derived from unevaluated fish bioconcentration factor (BCF) data, and less than 5% of existing organic substances have measured BCFs. We critically review the proposed criteria, first by considering other factors that may also contribute to reduced bioaccumulation for larger molecules, namely, reduced bioavailability in the water column, reduced rate of uptake corresponding to reduced diffusion rates, and the effects of biotransformation and growth dilution. An evaluated BCF and bioaccumulation factor (BAF) database for more than 700 substances and dietary uptake efficiency data are compared against proposed cutoff values. We examine errors associated with interpreting BCF data, particularly for developing molecular size criteria of bioaccumulation potential. Reduced bioaccumulation that is often associated with larger molecular size can be explained by factors other than molecular size, and there is evidence of absorption of molecules exceeding the proposed cutoff criteria. The available data do not support strict cutoff criteria, indicating that the proposed values are incorrect. Rather than assessing bioaccumulation using specific chemical properties in isolation, holistic methods that account for competing rates of uptake and elimination in an organism are recommended. An integrated testing strategy is suggested to improve knowledge of the absorption and bioaccumulation of large substances. Integr Environ Assess Manag 2010;6:210–224. © 2009 SETAC


Globally, regulatory agencies are assessing the potential hazards and risks of new and existing substances according to persistence (P), bioaccumulation (B), and toxicity (T) criteria (Government of Canada 1999; UNEP 2001; European Parliament 2006). Criteria for bioaccumulation assessments use aquatic-based endpoints for fish, including the bioaccumulation factor (BAF), the bioconcentration factor (BCF), and the octanol-water partition coefficient (KOW), as a surrogate for lipid–water partitioning. Few measured data are available, and quantitative structure–activity relationships (QSARs) and mass balance models are required for assessments (Arnot and Gobas 2006). Molecular size parameters have also been proposed for screening bioaccumulation potential (see, e.g., EC 2003; ECHA 2008).

Table 1 lists chemical properties, including molecular size parameters, proposed for bioaccumulation assessments. The values have been derived primarily from unevaluated fish BCF data, and there is no apparent consensus in the literature. The effective cross-section or effective cross-sectional diameter (ECS or Deff) has been defined as the second smallest (or largest) estimated cross-section or diameter of the molecule (Opperhuizen et al. 1985; Schuurman 1990) and as the minimum diameter of cylinders circumscribing the molecule (Dimitrov et al. 2003). The maximum diameter (Dmax) has been defined as the largest estimated diameter (length) of the molecule such that the perpendicular diameter, i.e., Deff, is minimized (Schuurman 1990) and as the minimum diameter of spheres circumscribing the molecule (Dimitrov et al. 2003). Opperhuizen et al. (1985) proposed a loss of membrane permeation for hydrophobic substances with a Deff >0.95 nm. Gobas and Morrison (2000) have suggested that, although a size restriction to membrane permeation is plausible, it appears to be greater than 0.95 nm, and it may not exist as a sharp cutoff. Dimitrov et al. (2003) have suggested that a Dmax of 1.47 nm provides reasonable discrimination for chemicals with log BCF less than or greater than 3.5. The European Commission Technical Guidance Document states that substances with a molar mass (MW) >700 g/mol are less likely to be absorbed and bioconcentrate (EC 2003). The UK Environment Agency has reviewed computational chemistry and molecular modelling tools for estimating molecular size parameters, but this analysis did not include a critical review of bioaccumulation data (Brooke and Cronin 2009). This study also pointed out that there is some confusion regarding the exact definition of molecular property indicators used for bioaccumulation assessments as outlined in REACH (ECHA 2008), particularly with respect to molecular diameters (Brooke and Cronin 2009).

Table 1. Chemical properties that have been proposed for screening bioaccumulation potential
Parameter and criteria Proposal Data used Reference
Deff >0.95 nm Loss of membrane permeation Unevaluated BCF and dietary exposure data

Opperhuizen et al. 1985

equation image <0·1 mg/L and equation image <10 mg/L No bioaccumulation test required for pigments Unevaluated BCF data for organic pigments and dyes

Anliker et al. 1988

Log KOW >3 and Deff >1.05 nm and MW >450 g/mol Possibly no bioaccumulation test required, very low probability for organic colorant accumulation Unevaluated BCF data for organic pigments and dyes

Anliker et al. 1988

Dmax >1.47 nm Log BCFs <∼3.5 Unevaluated BCF data

Dimitrov et al. 2003

Dmax, average = 1.7 nm; >2.5 nm Probability 0.5 for crossing cell membranes; probability near 0 for crossing cell membranes Unevaluated BCF data

Dimitrov et al. 2005

MW >700 g/mol Less likely to bioaccumulate Unevaluated BCF data

EC 2003

5 H-bond donors, 10 H-bond acceptors, MW >500 g/mol and log KOW >5 Poor absorption (Lipinski's “rule of 5” for drug development) Pharmaceuticals

Escher and Weisbrod 2006

SO <0.2 mg/L; or log KOW <4; or Deff >1.05 nm, or Dmax >1.5 nm No potential for bioaccumulation for organic pigments and possibly other substances such as dyes and surfactants Unevaluated BCF data

Horrocks et al. 2006

MW >550 g/mol, Dmax >2.0 nm, and Deff >1.1 nm Log BCF <3.69 Unevaluated BCF data

Sakuratani et al. 2008

MW >550 g/mol, Dmax >2.9 nm, and Deff >1.4 nm Log BCF <3.0 Unevaluated BCF data

Sakuratani et al. 2008

  • Deff = effective diameter; Dmax = maximum diameter; MW = molar mass; Curn:x-wiley:15513777:media:IEAM20090511:tex2gif-stack-1 = solid solubility in water; Curn:x-wiley:15513777:media:IEAM20090511:tex2gif-stack-2 = solid solubility in octanol; SO = solubility in octanol; BCF = bioconcentration factor.

The general objective of the present study is to determine whether proposed molecular size cutoff criteria (or indicators) are supported by available data and whether it is scientifically credible to include these criteria for screening assessments of bioconcentration and bioaccumulation. Theories and models for bioaccumulation and mechanisms of absorption by passive diffusion are first reviewed. A large BCF and BAF database evaluated for data quality and dietary exposure data are used to test for relationships with proposed molecular size criteria. Uncertainties associated with using BCF data for deriving molecular size criteria are highlighted, particularly for hydrophobic substances. Research from the pharmacological and nanoparticle literature is also considered. Finally, recommendations are provided to improve methods for screening bioaccumulation potential in aquatic species through the development of integrated testing strategies.


The concept of retarded uptake

The hypothesis that a molecular size cutoff exists is probably based on a mental picture of the absorption process at the gill or in the gastrointestinal tract (GIT). To be absorbed, the chemical must be capable of diffusing through aqueous and organic (lipid) layers. The tightly packed lipid layer can be viewed as impermeable to molecules larger than a certain size, i.e., the membrane acts as a filter, allowing small molecules to diffuse freely but blocking or retarding the diffusion of large molecules. Several correlations have been developed for diffusivity, as reviewed by Reid et al. (1987). These authors show an inverse dependence on solute molar volume raised to a power in the range 0.6 to 0.71, so doubling the molar volume is expected to reduce diffusivity by a factor of 1.50 to 1.64. Thus, a tenfold increase in molar volume will cause a decrease in diffusivity by a factor of only about 4 to 5. Direct measurements of diffusion rates in biological systems are difficult; however, a promising approach is to use in vitro systems that simulate biological transport (Kwon et al. 2006; Kwon and Escher 2008). The most readily available sources for evidence of transport (or lack of transport) are BCF and BAF data, for which the quantity of chemical absorbed is measured in relation to concentrations in the water. A low BCF or BAF can be attributed to slow or blocked diffusion; however, when interpreting such data, it is essential to consider other factors that may influence the uptake rate and possible accumulation, specifically bioavailability, biotransformation, and the role of sequential transport resistances. These factors are considered below.

Bioconcentration, bioaccumulation, and bioavailability

Figure 1 illustrates the major routes of chemical uptake and elimination processes in a fish. Bioaccumulation is the net result of these competing process rates, as discussed in detailed reviews on bioconcentration, biomagnification, and bioaccumulation (Gobas and Morrison 2000; Mackay and Fraser 2000; Barber 2003, 2008). The BCF (L/kg) and BAF (L/kg) for fish can be measured as the steady-state ratio of the chemical concentration in the organism (mol/kg) to the chemical concentration in the water (mol/L), i.e., CF/CW (OECD 1996; Gobas and Morrison 2000). The BCF is measured under controlled laboratory conditions in which the organism is exposed only to a chemical in the water, and thus the chemical is absorbed at the gill surface and through the skin. The BAF is measured in the natural environment, which includes exposure from the water and the diet (absorption of the chemical in the GIT). Dietary exposure can be the dominant route of exposure and uptake for hydrophobic chemicals (see, for example, Qiao et al. 2000). The BCF and BAF share the same major routes of chemical elimination, such as respiratory loss, fecal egestion, and metabolic biotransformation. Growth dilution is a pseudoelimination process, in that the substance is not eliminated from the organism but rather the concentration is reduced as a result of increased mass (or volume). Growth is an important consideration when interpreting bioaccumulation data for hydrophobic, slowly biotransformed substances. Organisms can be fed well in laboratory tests, resulting in relatively fast growth rates (Gobas and Morrison 2000; Jonker and van der Heijden 2007). Dermal absorption and elimination rates are generally considered negligible compared with the much faster rates of chemical absorption and elimination at the surface of the gill.

Details are in the caption following the image

Conceptual model of bioaccumulation in fish and bioavailability in the water column.

Only the truly dissolved chemical concentration in the water (CWD) is bioavailable for diffusive uptake by fish (McCarthy and Jimenez 1985; Barron 1990; Schrap and Opperhuizen 1990; Gobas and Morrison 2000). The total water concentration (CWT), as is usually measured by solvent extraction in laboratory and field studies, includes both the truly dissolved bioavailable fraction and a fraction that is sorbed to organic matter in the water column that is not bioavailable for uptake. The bioavailable solute fraction (BSF = CWD/CWT) for hydrophobic substances (log KOW greater than approximately 6) can be very low under standard laboratory (BCF) test conditions and in the environment (BAF) because of the presence of particulate and dissolved organic matter in the water (Gobas et al. 1989; Gobas and Morrison 2000; Jonker and van der Heijden 2007; Parkerton et al. 2008). BCF and BAF measurements for hydrophobic chemicals based on the total water concentration therefore underestimate the degree of bioaccumulation when based on the quantity of chemical that is actually bioavailable to the organism, because the water concentration in the denominator is erroneously high (Arnot and Gobas 2003; Jonker and van der Heijden 2007). For chemicals with the potential to dissociate, the degree of dissociation can also influence the fraction of chemical that is bioavailable, insofar as evidence suggests that only (or predominantly) the neutral species is able to diffuse through gill membranes (Erickson et al. 2006a, 2006b). The concept of bioavailability in the water column is critical for the interpretation of absorption and bioaccumulation data for hydrophobic chemicals and for proposing molecular size restrictions.

A related issue is the occasional attempt to facilitate absorption by increasing the concentration of the test chemical using cosolvents, which can result in exposures to concentrations exceeding the solubility in water. The use of such unrealistic conditions casts doubt on the validity of the results; for example, in bioconcentration tests using supersaturated conditions, there may be intestinal absorption of colloidal chemical particles.

Chemical absorption, membrane permeation, and the 2-film model

Chemical absorption rates at epithelial surfaces can be expressed in a general sense as
equation image(1)
where A (mol/h) is the chemical absorption rate, G (m3/h) is the contact (or flow) rate of the exposure medium to the epithelial surface, C (mol/m3) is the chemical concentration in the exposure medium, and E (unitless) is the chemical absorption efficiency. For fish, the subscripts W and D can be used to refer to absorption from the water and from digestion (GIT), respectively. The chemical absorption efficiency (EW or ED) expresses the amount of substance that is actually absorbed relative to the amount of bioavailable chemical that is brought into contact with the epithelial surface (Gobas and Morrison 2000). Under a specific set of conditions, such as a particular fish exposed to chemicals in the water, the gill flow rates are the same for all chemicals; therefore, differences in rates of absorption are the result of differences in C and E. Assuming a unit exposure concentration, differences in chemical uptake rates reflect differences in E and thus possible differences in diffusion rates.

Passive diffusion is considered the predominant mechanism for the transport of substances across epithelia for most pharmaceuticals and environmental organic contaminants, although facilitated, active, paracellular, and phagocytosis (pinocytosis and endocytosis) transport mechanisms can be important for certain substances (DeVito 2000). Passive diffusion rather than a facilitated process controls the absorption of hydrophobic persistent organic pollutants (POPs; Kelly et al. 2004). For diffusion through epithelial tissue barriers, the substance must be transported through a series of aqueous and organic phases (Flynn and Yalkowsky 1972; Gobas et al. 1986b). For example, for transport from the GIT contents to the blood, the absorption pathway involves crossing an aqueous stagnant layer, a layer of mucus, an apical cell membrane, the cell contents, the basal cell membrane, the basement membrane, the intercellular space between the basement membrane and the capillary, and the capillary wall (Hayton 1980). Several mathematical models have been developed to describe passive absorption (see, for example, Flynn and Yalkowsky 1972; McKim et al. 1985; Gobas et al. 1986b; Balon et al. 1999; Abraham et al. 2002; Yalkowsky et al. 2006). Essentially all of these models incorporate elements of chemical partitioning, solubility, and molecular size.

Flynn and Yalkowsky (1972) proposed a simple and robust diffusion mass transfer model and presented empirical evidence to characterize the permeation of nonelectrolytes across aqueous and organic phases (or diffusion layers or diffusion barriers) separating two compartments. The overall permeability of a substance through a series of diffusion layers (P, cm/sec) is the reciprocal of the total resistance through these layers (RT, sec/cm). The permeability equation for the diffusion layer–membrane system is derived, assuming independent and additive resistances of the individual layers (Flynn and Yalkowsky 1972; Gobas et al. 1986a), as
equation image(2)
where RW and RM are the inherent resistances in water and membrane layers, respectively, and can be regarded as reciprocals of mass transfer coefficients or velocities. The membrane–water partition coefficient (KMW) is required to account for the different concentrations at the interfaces, i.e., for differences in the substance's affinities for water and the membrane. According to Fick's First Law (diffusion within a homogeneous phase), each individual resistance is equal to the distance of diffusion, i.e., thickness in each layer (h; cm), divided by the diffusion coefficient of the substance in that layer (D; cm2/sec). With the assumption that the different aqueous layers are compositionally equivalent (e.g., unstirred water layers on either side of the cell, cytoplasm) the resistance to diffusion in the combined aqueous layers is
equation image(3)
The resistance to diffusion in the membrane (organic layers) is expressed as
equation image(4)

When adding these resistances, as in Equation 2, the partition coefficient must be included. In general, diffusive resistances are largest in phases in which the concentrations are lowest. When KMW is low, concentrations in the membrane are low, and resistance is higher in this phase. For substances with low KMW values, the permeation rate is largely controlled by diffusion through the membrane, and the flux is proportional to KMW. When KMW is large (as applies to hydrophobic substances) the concentration in the aqueous phase is lower, and the aqueous resistance controls the overall resistance to permeation. Thus, as KMW increases, the rate of membrane permeation switches from membrane layer control to aqueous layer control. Under complete aqueous diffusion layer control, rates of membrane permeation are fairly constant (Flynn and Yalkowsky 1972; Gobas et al. 1986a; Kwon et al. 2006; Kwon and Escher 2008). The octanol–water partition coefficient is often used as a surrogate for KMW because there are more data available for KOW.

Figure 2A shows data for measured gill uptake efficiencies, EW, based on total water concentrations (McKim et al. 1985) and a two-film model for EW (blue line in Figure 2A; Gobas and Mackay 1987). For substances with log KOW values less than about 3, resistance in the membrane controls EW, and, for substances with log KOW values greater than about 3, resistance in the aqueous layer controls EW (McKim et al. 1985; Gobas et al. 1986a). For chemicals with log KOW values >6, reduced chemical bioavailability in the water can also reduce chemical uptake at the gill, as illustrated by including model estimates for the BSF for nondissociating organic substances (red line in Figure 2A; Arnot and Gobas 2004). The combination of the 2 models (black line in Figure 2A) convincingly describes the parabolic relationship observed in the measured data set. Figure 2B illustrates the 2-film model relationship for ED from measured data for different species for POPs (Kelly et al. 2004). Chemical absorption efficiency in the GIT shows declines for substances with log KOW greater than about 7, as diffusion becomes controlled by dissolution in aqueous layers (Gobas et al. 1988). These 2-film models are used in many mechanistic bioaccumulation models and provide a convincing, theoretically justified, quantitative basis for estimating uptake rates.

Details are in the caption following the image

Measurements and models (A) characterizing the dynamics of chemical absorption efficiency (EW) in fish gills as a function of the octanol–water partition coefficient (KOW). Circles represent measured data reported by McKim et al. (1985) based on total water concentrations after solvent extraction. The blue line represents a two-film resistance model estimate for EW (Gobas and Mackay 1987), and the red line represents model estimates for the bioavailable solute fraction (BSF) (Arnot and Gobas 2004). The black dashed line represents the combined model estimates of mass transfer dynamics through membranes and bioavailability in the water phase on the overall gill uptake efficiency. BSF was calculated assuming 0.5 mg/L of dissolved organic carbon and 0.5 mg/L of particulate organic carbon in the water. OECD 305E BCF test guidelines allow for a total of 2 mg/L of total organic carbon, a level that would further lower the BSF estimates (OECD 1996). Two-film model estimates (B) for the dietary uptake efficiencies (ED) for persistent contaminants in various species as a function of KOW as reviewed by Kelly et al. (2004).

Ostensibly, substances that are absorbed by passive diffusion are soluble in organic (lipid membranes) and aqueous (cytoplasm, unstirred water layers) phases so that concentration gradients can be established. Hence, membrane–water partitioning, rather than explicit solubility in one phase or the other, determines a substance's ability to diffuse through cell membranes. Flynn and colleagues (1974) showed that diffusivity in aqueous solutions for a homologous series of substances is observed to decrease with increasing partial molal volume of the substance. However, this relationship is not linear, and with incremental increases in molal volume the reductions in diffusivity were proportionally less.

The use of KOW as a surrogate for KMW may not always be appropriate. Octanol is not necessarily a satisfactory surrogate for lipid membranes, but it appears to give an order of magnitude estimate of KMW. Octanol fails to mimic the pronounced interfacial character of the bilayer membrane structure and ionic interactions between membrane lipids and solutes, particularly for polar and charged molecules that can have a greater affinity to the surface of the membrane. Liposomal systems have been shown to be less discriminating than octanol, especially for lower-KOW substances (Balon et al. 1999). Polyparameter, linear-free energy relationships (ppLFERs) may provide better estimates for KMW and absorption, insofar as there may be greater resolution in treating key parameters involved in partitioning individually (Abraham et al. 2002). The degree of dissociation can also influence the mass transfer of electrolytes. For bioconcentration in fish, evidence suggests that the ionic species can support the re-equilibration of the neutral species in aqueous layers, thus effectively increasing mass transfer in aqueous phases during membrane permeation (Erickson et al. 2006a, 2006b).

Relative resistances in gill, skin, and intestinal tissues

It is noteworthy that there are considerable differences in resistances and thus permeabilities of gill, skin, and intestinal tissues. The transepithelial electrical resistance (TEER, Ωcm2) is a measure of epithelial tissue properties, integrity, and general permeability to ions and electrons (Pasternak and Miller 1996). TEER often correlates with the flux of small molecules across the tissue and can be used to classify the tightness or leakiness of epithelial tissues (Paranjpe and Sinko 2002). Lower TEER values indicate greater permeability and presumably greater absorption potential via paracellular routes, i.e., tight junctions, although the parameter is also generally used to quantify and qualify the integrity of membranes. Epithelia with TEER values of about 10 to 50 Ωcm2 are considered to be leaky (high permeability), whereas values >1,000 Ωcm2 are considered to be tight (low permeability, Ward et al. 2000). Table 2 lists TEER values of different epithelial tissues, illustrating a range of values that encompasses over 3 orders of magnitude. Intestinal tissues in fish and mammals can be orders of magnitude more permeable than gills and skin. An implication is that, even if a substance exhibits a low BCF because resistance at the gill is large, it may bioaccumulate readily because of the reduced resistance in the GIT. A corollary to this is that, when assessing bioaccumulation potential, a BAF is a better metric than a BCF.

Table 2. Transepithelial electrical resistance (TEER) of various epithelial tissues and cell lines
Epithelial tissue TEER (Ωcm2) Reference
Proximal renal tubule (rat, rabbit, dog) 6–7

Ward et al. 2000

Fish intestine 25–50

Trischitta et al. 1999

Mammalian small intestinea 20–100

Paranjpe and Sinko 2002

Distal colon (human) 259–319

Paranjpe and Sinko 2002

Caco-2 (human colon cell line) 230–1000

Ward et al. 2000

Calu-3 (human airway cell line) 850–1150

Paranjpe and Sinko 2002

Fish gill 3500

Ward et al. 2000

Skin (frog) 2000–8700

Ward et al. 2000

Skin (human) 20 000

Peck and Higuchi 1997

  • a Duodenum, jejenum, ileum in rats.

The key conclusion from this theoretical analysis is that the uptake of larger, more hydrophobic substances is controlled by the diffusive resistance in the aqueous phase, not the lipid phase, and uptake from food can be faster than from water (because of the higher food concentration relative to water and the greater permeability in the GIT); therefore, even if there is retarded diffusion in lipid layers of the gills, this will not necessarily be reflected in reduced bioaccumulation in the environment.


This review discusses publicly available measured and evaluated BCF and BAF data for more than 700 substances, with a focus on larger molecules that approach or exceed the proposed cutoff criteria. We also review measured data for dietary uptake efficiency, critically evaluate previous proposals for molecular size criteria, and consider insights from pharmacology and nanoparticle studies.

Measurements and model predictions for molecular size parameters are compared with the measured bioaccumulation data. Unless cited otherwise, MW and KOW values are from EPI Suite™ with measurements for KOW selected preferentially over model predictions (USEPA 2009). Dmax and Deff values were predicted by the Oasis Canadian POPs Model (v.1.1.11) (based on Mekenyan et al. 2005). For chemicals predicted to have multiple conformers (e.g., flexible molecules), averages were calculated from a maximum of 30 energetically stable conformers (i.e., Dmax, avg and Deff, avg; Dimitrov et al. 2005). The range of the molecular size predictions for the various conformers of a single molecule is illustrated for certain chemicals (e.g., range between the minimum and the maximum Dmax predictions for 30 conformers). Considerable variation in estimated values of molecular size descriptors using molecular modeling tools has been reported with deviations of approximately 50% of the average commonly predicted (Brooke and Cronin 2009). There is no practical method to assess which estimation methods are correct, because the actual dimensions of the molecules are unknown (Brooke and Cronin 2009).

BCFs and BAFs

The proposed molecular size criteria rely heavily on BCF data (Table 1); however, the BCF test is subject to potential sources of error, which can result in substantial uncertainty in reported values (Arnot and Gobas 2006; Parkerton et al. 2008) and thus in uncertain molecular size proposals. In a data quality review of 4323 fish BCFs for 770 chemicals, 42% had at least 1 major source of analytical uncertainty (Arnot and Gobas 2006). Parkerton et al. (2008) reviewed 2 chemicals (anthracene and 2,3,7,8-tetrachloro-p-dioxin) for data quality and found that only 15 to 22% of the reported BCFs were of reliable quality (“reliable with restrictions”). Possible sources of errors when interpreting BCF data include the use of radiolabelled chemicals without correcting for the parent signal, assuming constant water concentrations during the exposure period, assuming that steady state is reached, and exposing the organism at chemical concentrations that are greater than the chemical's water solubility limit (Arnot and Gobas 2006; Parkerton et al. 2008). The latter three sources are particularly relevant for sparingly soluble, hydrophobic chemicals. It is difficult to maintain constant bioavailable aqueous exposure concentrations for sustained durations for hydrophobic substances, because these substances tend to partition from the pure water phase onto test equipment (e.g., tank walls) and particulate and dissolved organic carbon in the bulk water compartment. The time to steady state can exceed test durations for slowly biotransformed hydrophobic substances (OECD 1996; Parkerton et al. 2008). Furthermore, as noted earlier, the actual BCF is underestimated if the water concentration is greater than the chemical's water solubility limit, because the bioavailable water concentration is overestimated (Arnot and Gobas 2006).

A critically evaluated bioaccumulation database was used to explore relationships between BCF and BAF measurements and molecular size parameters. Full details of the BCF and BAF database and data quality assessment methods are provided elsewhere (Arnot and Gobas 2006). Briefly, BCF data were considered reliable if the chemical concentration in the water was measured during the exposure period and was near or below water solubility limits, if chemical-specific analyses were conducted using appropriate methods (the use of radiolabeled chemicals must have been corrected for parent chemical), if sufficient exposure durations were provided to reach or approximate steady-state conditions (≥80% of steady state), and if rate constants were reported, experimental conditions were well controlled (water temperature, etc.), and no adverse toxic effects were reported. The data quality review sought to reduce uncertainty in the bioaccumulation measurements; however, uncertainty cannot be entirely eliminated.

Figure 3 shows 3664 BCF and BAF measurements for 718 organic chemicals that have been evaluated for data quality. There are few measured data available for substances with Dmax estimates >1.5 nm and Deff estimates >0.95 nm; however, it is clear that chemicals with these properties have BCF and BAF values greater than bioaccumulation criteria in regulatory programs, i.e., BCF or BAF >1,000 or 5,000. The measured BAF data are primarily for currently listed POPs (UNEP 2001), which include substances with low biotransformation capacity in fish and molecular size estimates below proposed cutoff values. The following discussion highlights examples of measured bioaccumulation end points that exceed regulatory criteria and proposed molecular size cutoff criteria or indicators.

Details are in the caption following the image

Measured bioconcentration factors (BCF; blue circles) and bioaccumulation factors (BAF based on total water concentrations [green triangles] and BAF based on truly dissolved water concentrations [yellow diamonds]) considered to be of reliable quality compared with maximum diameter (Dmax) and effective diameter (Deff) estimates, molar mass (MW), and the octanol–water partition coefficient (KOW). The solid and dashed horizontal lines represent bioaccumulation criteria of 5000 (Government of Canada 2000; UNEP 2001) and 1000 (USEPA 1999), respectively. The dashed vertical line represents proposed molecular size cutoff values (Dmax = 1.47 nm [Dimitrov et al. 2003], Deff = 0.95 nm [Opperhuizen et al. 1985], MW = 700 g/mol [EC 2003]). Maximum diameters (Dmax) and effective diameters (Deff) were estimated using the Oasis Canadian POPs model v.1.1.11. For substances with multiple conformers, the Dmax and Deff predictions shown are the averages (Dmax, avg or Deff, avg; see text for details).

Figure 3 includes fish BAFs for brominated flame retardants (BFRs) calculated from a Lake Winnipeg, Canada, food web study (Law et al. 2006). These BFRs have molecular size parameters greater than currently listed POPs, and the BAFs are based on measured estimates for the truly dissolved water concentrations. Figure 3 includes hexabromocyclododecane α- and γ-isomers (HBCD, MW = 642 g/mol; Dmax, avg = 1.2 nm; Deff, avg = 1.0 nm), bis(2,4,6-tribromophenoxy) ethane (BTBPE, MW = 688 g/mol; Dmax, avg = 1.6 nm; Deff, avg = 0.92 nm), and polybrominated diphenyl ethers (congeners 47, 99, 100, 153, 154, 209). The BAFs for decabromodiphenyl ether (DBDE, MW = 959 g/mol; Dmax, avg = 1.5 nm; Deff, avg = 0.97 nm) and decabromodiphenyl ethane (DBDPE, MW = 971 g/mol; Dmax, avg = 1.5 nm; Deff, avg = 0.97 nm) were calculated using the method detection limits in water, because the truly dissolved water concentrations were below the method detection limits (Law et al. 2006). The illustrated DBDE and DBDPE BAFs are therefore expected to be underestimated. The BFRs in this study biomagnify in the food web (biomagnification factors >1, trophic food web magnification factors >1; Law et al. 2006), and biotransformation rates in fish are predicted to be very slow (Arnot et al. 2009). Many of these larger BFRs have also been detected in organisms of terrestrial food webs such as birds and mammals, including humans (Thomsen et al. 2007; Hu et al. 2008; Gauthier et al. 2009).

Figure 3 includes measured BAFs for esters and other substances that are biotransformed in fish at rates that are orders of magnitude faster than the listed POPs and the BFRs (Arnot et al. 2008, 2009). For example, 1,2-benzenedicarboxylic acid, dinonyl ester (DnNP) has a Dmax, avg estimate of 1.9 nm (Dmax range 1.6–3.0 nm), a log BAF (wet weight) of approximately 2.3 based on total water concentrations (Gobas et al. 2003), and an estimated biotransformation rate constant of about 2/d for a 10-g fish (Arnot et al. 2009). Octaethylene glycol monotridecyl ether has a Dmax, avg estimate of 2.7 nm (range 2.0–4.8 nm), a log BCF of 1.5, and an estimated biotransformation rate constant of, 10/d in fathead minnow (Pimephales promelas, about 0.7 g; Tolls and Sijm 1999). This chemical is only moderately hydrophobic (log KOW = 4.4; Tetko et al. 2005), and the biotransformation rate is relatively fast. A moderately hydrophobic (log KOW = 4.11; Tetko et al. 2005), long-chain, anionic ethoxylate surfactant (pKa = 1.1; SPARC version 4.2; Hilal et al. 1995) has an estimated Dmax, avg of 2.5 nm (range 1.6–4.1 nm) and log BCFs ranging from 1.5 to 2.2. A biotransformation rate constant of about 17/d is predicted for a 10-g fish for this substance, which is considered very fast (Arnot et al. 2009). The low bioconcentration observed for this large substance is thus primarily a result of fast biotransformation and reduced uptake rates because of high dissociation. Reduced bioaccumulation potential for these substances in fish cannot be strictly attributed to molecular size restrictions because of factors such as fast metabolic biotransformation.

BCF measurements for other large substances that are likely to have slower biotransformation rates than the aforementioned esters and ethoxylates show higher bioaccumulation potential (log BCF or log BAF values >3.7). Log BCFs of approximately 3.8 to 4.0 for [1,1′-biphenyl]-4-ol, 3,5-bis(1,1-dimethylethyl) (Dmax, avg = 1.3 nm; Deff, avg = 1.1 nm) and 4.0 to 4.1 for methylbis(phenylmethyl)benzene (Dmax, avg = 1.5 nm; range 1.3–1.7 nm) have been reported (Japan CITI 1992). The predicted biotransformation rates of these substances in fish are moderate (Arnot et al. 2009) and estimated molecular size parameters that exceed proposed cutoff values.

Figure 3 shows 4 chemicals with MW >700 g/mol for which measured BCF and BAF data are available. From the BAF data set, these are DBDE and DBDPE with log BAFs ranging from approximately 2.1 to 4.9. From the BCF data set, a benzenesulfonic acid with MW = 881 g/mol has log BCFs of about 0.5 and 1.2. A QSAR prediction indicates a very fast biotransformation rate for this acid (Arnot et al. 2009). The substance with the second largest MW is a naphthalenesulfonic acid (MW = 787 g/mol) with log BCFs of about 1.8. A QSAR prediction indicates a moderate biotransformation rate for this acid (Arnot et al. 2009). These acids have an estimated pKa of ∼0.7 (SPARC version 4.2; Hilal et al. 1995) and are expected to be >99.9% in their ionic forms under exposure conditions. Thus, the considerable degree of dissociation is expected to reduce chemical uptake at the gill during BCF tests, and biotransformation rates are expected to account largely for the low level of bioconcentration rather than high MW.

There are limited measured bioaccumulation data for substances requiring evaluation. Organic chemicals on Canada's Domestic Substances List (DSL; n about 11 000) can be considered representative of global commercial chemical production. More than 95% of the organic chemicals on the DSL have not been measured for bioaccumulation potential (BCFs or BAFs; Arnot and Gobas 2006). The median MW for DSL organics is 230 g/mol, and approximately 1100 have an MW >600 g/mol and 330 have an MW >900 g/mol. The median MW for DSL organics that have been measured is approximately 200 g/mol. Among the chemicals that have BCF or BAF measurements, only 2% have MW >600 g/mol. Thus, less than 1% of chemicals with MW >600 g/mol have been measured for bioaccumulation potential. Reliable bioaccumulation measurements for larger organic substances are limited, and molecular size criteria (or indicators) have been derived from relatively few larger substances.

Dietary uptake efficiency studies

For hydrophobic chemicals, the diet can be a significant route of uptake (Qiao et al. 2000), and, as discussed earlier, evidence shows that the GIT is leakier than fish gills and other epithelial tissues (Table 2). Dietary tests may therefore be more relevant for testing larger molecules, and these tests are expected to mitigate some of the analytical challenges faced in conducting aqueous-based exposure tests (BCFs) for hydrophobic chemicals. A database of ED values for fish was compiled and reviewed for the present study. There are no standardized test protocols for ED; therefore, data quality evaluations are difficult. Radiolabel tests were not considered reliable unless the analysis distinguished between parent and transformation product signals. Most of the available data are for polychlorinated biphenyls (PCB), but some other substances, such as polycyclic aromatic hydrocarbons and dioxins and furans, have also been measured.

Figure 4 shows slight relationships between measured ED and molecular property descriptors (Dmax, Deff, MW, KOW); however, substances with Dmax estimates >2.5 nm have ED values ranging from approximately 10 to 35%. Large variability for measured ED is a result of different methods for dosing the animals (e.g., gavage, spiked food) and biotransformation processes in the GIT and in the body. Biotransformation in the GIT and in the liver (first pass) may result in lower ED estimates (Kleinow et al. 1998). For example, DBDE is absorbed and debrominated to less brominated diphenyl ether congeners in the GIT and liver with ED estimates ranging from <1 to 5.2% (Kierkegaard et al. 1999; Stapleton, Alaee, et al. 2004, 2006; Tomy et al. 2004). Although these absorption efficiencies are low, DBDE is bioaccumulative in aquatic and terrestrial food webs, including data from human monitoring programs (Lindberg et al. 2004; Chao et al. 2007). In mammals, DBDE is also absorbed and biotransformed into less brominated congeners that have longer whole-body, biological half-lives than DBDE (Huwe and Smith 2007; Kierkegaard et al. 2007). Larger chemicals (absorbed or possibly not absorbed) may be transformed into smaller substances that can be more bioavailable and more persistent than the parent substance. The dietary data highlight the potential for preliminary bioaccumulation screening assessment errors if molecular size cutoff criteria derived from BCF data are used.

Details are in the caption following the image

Measured dietary uptake efficiencies (ED) in fish compared with maximum diameter (Dmax) and effective diameter (Deff) estimates, molar mass (MW), and the octanol–water partition coefficient (KOW). Dark blue circles represent polychlorinated biphenyl data; light blue circles represent other chemicals. The dashed vertical line represents proposed molecular size cutoff values (Dmax = 1.47 nm [Dimitrov et al. 2003], Deff = 0.95 nm [Opperhuizen et al. 1985], MW = 700 g/mol [EC 2003]). Maximum diameters (Dmax) and effective diameters (Deff) were estimated using the Oasis Canadian POPs model v.1.1.11. For substances with multiple conformers, the Dmax and Deff predictions shown are the averages (Dmax, avg or Deff, avg; see text for details).


The Opperhuizen proposal

Opperhuizen and colleagues (1985) proposed a loss of membrane permeation in fish gills for substances with Deff (reported as effective cross-section; ECS) >0.95 nm. This proposal was based on data from their own experiments and data cited from three other papers, namely, Zitko and Hutzinger (1976), Zitko (1977), and Bruggeman et al. (1984). Table 3 lists representative chemicals and physical–chemical parameters reported in the Opperhuizen proposal and chemical property values cited from other sources illustrating some uncertainty in chemical properties. The experimental bioaccumulation data used in the Opperhuizen paper and from more recent studies are discussed below, highlighting potential errors in using unevaluated BCF data for molecular size proposals.

Table 3. Data used to support the proposal that chemicals with an effective cross-section (Deff) >0.95 nm do not bioconcentrate in fish (loss of membrane permeation)
Chemical name Abbreviation MW (g · mol−1) Deff (nm) Log KOW
1,2-Dichloronaphthalene DCN 197 0.86a 0.88b 4.9a 4.4c
1,3,5,7-Tetrachloronaphthalene Tetra-CN 266 0.93 0.93 6.4 5.6
Octachloronaphthalene OCN 404 0.98 0.99 8.4 7.7
2,3,7,8-Tetrachlorodibenzo-p-dioxin TCDD 322 0.76 0.73 >6 6.0
Octachlorodibenzo-p-dioxin OCDD 460 0.98 0.99 >6 8.0
Hexachlorobenzene HCB 285 0.87 0.96 5.8 5.5
Hexabromobenzene HBB 552 0.96 1.02 >6 6.3
Decachlorobiphenyl DCB 499 0.87 0.87 >6 8.9
2,2',3,3′4,4′-Hexabromobiphenyl HBBP 628 0.94 0.98 >6 8.0
Decabromobiphenyl DBB 944 0.96 1.02 >6 10.0
  • MW = molar mass; KOW = octanol–water partition coefficient.
  • a Values in these columns are from Opperhuizen et al. (1985).
  • b Values in this column are from Anliker et al. (1988).
  • c Values in this column are averages from Tetko et al. (2005).

Opperhuizen and colleagues (1985) measured bioconcentration in the guppy (Poecillia reticulata) for individual chloro-naphthalenes (di-, tri-, tetra-) and an industrial chloronaphthalene mixture. The reported Deffs for di-, tri- and tetra-chlorinated naphthalenes ranged from 0.86 to 0.93 nm. During the 7-d exposure period, most of the guppies exposed to the individual chemicals showed narcotic toxic effects, and up to 50% mortality of the test population was reported, suggesting chemical absorption. Later, Burreau et al. (1997) exposed Northern pike (Esox lucius L.) to some polychlorinated naphthalenes (PCNs) and some polybrominated diphenyl ethers (PBDEs) in the diet. Deff values of 0.97 nm for hexa- and hepta-CNs, 0.98 nm for octa-CN, and 0.96 nm for penta- and hexa-BDEs were reported. Dietary uptake efficiencies for these substances ranged from about 40% for octa-CN and hexa-BDE to about 75% for hexa-CN.

Zitko and Hutzinger (1976) exposed Atlantic salmon (Salmo salar) to di-, tri-, and tetrachlorinated and brominated biphenyls and hexachloro- and hexabromobenzenes in water under static conditions for 4 d and separately to spiked food over a period of 40 d. Hexabromobenzene (HBB, MW = 552) was not detected in the fish, i.e., <0.004 µg/g (the method detection limit) as a result of either route of exposure, and the authors speculated that a substance with MW ≥600 g/mol might have a limited capacity for uptake in fish. The aqueous exposure concentration for HBB was between 9 and 70 µg/L, more than an order of magnitude above the reported aqueous solubility of the substance, i.e., 0.16 µg/L (Opperhuizen 1985). Static aqueous exposures can result in significant quantities of organic matter in the test system, reducing the BSF and bioavailable concentrations during the bioconcentration test. A log BCF of approximately 3 for HBB was measured in Rainbow trout (Oncorhynchus mykiss) exposed under flow-through conditions for 75 d (Oliver and Niimi 1985). Water used in flow-through tests is expected to have less organic carbon than water used in static tests, thereby increasing bioavailability for hydrophobic chemicals such as HBB.

Zitko (1977) exposed Atlantic salmon (Salmo salar) to commercial polybrominated biphenyl mixtures (PBBs), including hepta-, octa- and nona-PBBs, in water under static conditions for 4 d and separately to spiked food over a period of at least 42 d. The author proposed, based on the observed chromatograph profiles, that the highly brominated biphenyls were debrominated in the fish, suggesting that the larger molecules were absorbed. Recent studies show definitive evidence for the debromination of highly brominated diphenyl ethers (Stapleton, Letcher, et al. 2004).

Bruggeman et al. (1984) exposed guppies to a series of PCBs and other halogenated aromatic substances (octachlorodibenzo-p-dioxin [OCDD] and HBB) in the water and the diet separately. A high degree of lethality (50%) and variability in fish tissue concentrations was reported. In the bioconcentration tests, concentrations of OCDD and HBB could not be detected in live fish; however, nonsteady-state log BCFs of about 3 and 4 were reported in dead fish for OCDD and HBB, respectively. This suggests the live fish may have absorbed and biotransformed OCDD and HBB below detection limits. In the dietary uptake experiments, OCDD and HBB levels in fish tissues were reported to be at or below the detection limits, suggesting some absorption of these chemicals. In 1990, a log BCF of 2.85 for OCDD in guppy exposed under flow-through conditions was reported, and biotransformation and reduced bioavailability were suggested as explanations for the BCF values being lower than expected based on linear regression estimates using KOW (Gobas and Schrap 1990). A log BCF of 3.37 for OCDD in guppy exposed under flow-through conditions has also been reported (Loonen et al. 1994). Niimi and Oliver (1986) conducted a dietary study for OCDD in rainbow trout (70 d exposure) and reported a dietary assimilation efficiency of 8% and a whole-body biological half-life of 15 d.

Other proposals

Dimitrov and colleagues (2003) evaluated relationships between fish BCF data and molecular size parameters (MW, Dmax, Deff) for substances with a log KOW >5 from the Syracuse Research Corporation BCFWIN database (Meylan et al. 1999). Available BCF measurements in fish and a BCF value of 5000 were used to explore threshold values that may affect bioaccumulation potential. BCF values >5000 were considered to be bioaccumulative and values less than 5000 were considered to be not bioaccumulative. The BCF data were not evaluated. The authors suggested that threshold values for MW of 700 or 1000 g/mol and Deff of 0.95 nm did not reduce the number of false-positive predictions and proposed that a Dmax of 1.47 nm discriminates between substances with a log BCF less than or greater than about 3.5. Figure 3 shows that substances that have Dmax >1.47 nm can be highly bioaccumulative. Dimitrov and colleagues (2005) later proposed that substances with Dmax, avg = 1.7 nm have approximately a 0.5 probability of crossing the cell membrane. The probability approaches 0 at Dmax, avg ≥2.5 nm (Table 1). As discussed above, the limited data available show that even substances biotransformed at fast rates with predicted Dmax, avg ≥2.5 nm are still absorbed by fish (Figure 3).

Sakuratani and colleagues (2008) explored relationships between fish BCF data and molecular size parameters (MW, Dmax, Deff) using a data set of 737 new and 441 existing chemical substances listed under the Japanese Chemical Substances Control Law (CSCL; Japan 1992). The CSCL bioconcentration test follows the 305C methods outlined by the Organisation for Economic Co-operation and Development (OECD 1992). Clear threshold values could not be determined; however, the authors found chemicals with MW >550 g/mol, Dmax >2.0 nm, and Deff >1.1 nm correspond to BCFs <5000. For a BCF threshold of <1000, the corresponding molecular size values were MW >550 g/mol, Dmax >2.9 nm, and Deff >1.4 nm.

A few examples are provided to show that, despite following OECD methods, there can still be uncertainties in the unevaluated BCF data and that BCF data should not be used in isolation for developing molecular size cutoff criteria for screening bioaccumulation potential. The CSCL database reports a log BCF for decabromobiphenyl (DBB, 943 g/mol) of about 0.5 and 0.6 in carp (Cyprinus carpio) after 42 d of aqueous exposures to water concentrations of 150 µg/L and 15 µg/L, respectively (Japan CITI 1992). The aqueous solubility is uncertain, but estimated values include 1.2 × 10−3 and 1.2 × 10−8 µg/L (USEPA 2009). Thus, the exposure concentrations may be at least 4 orders of magnitude above the solubility limit for the chemical, resulting in underestimation of the actual BCFs (Arnot and Gobas 2006). The CSCL BCF database reports log BCFs of 0.70 and 1.70 for DBDE after 42 d of exposure to water concentrations of 60 and 6 µg/L, respectively. The estimated water solubility of DBDE is ≤0.1 µg/L (USEPA 2009) suggesting the reported BCF values may underestimate actual BCFs by at least 1 order of magnitude (Arnot and Gobas 2006). DBDE is also susceptible to biotransformation (Kierkegaard et al. 1999; Stapleton, Alaee, et al. 2004; Stapleton et al. 2006; Tomy et al. 2004), which can confound the interpretation of absorption potential. DBDE BAFs from the environment exceed the CSCL BCF values and regulatory criteria (Figure 3), further illustrating the misapplication of BCF data for developing molecular size cutoff criteria for use as indicators for assessing bioaccumulation potential.

Pigments and dyes

Bioaccumulation data in fish for 23 dyestuffs and 13 organic pigments were compared with physical–chemical properties (KOW, Deff, MW; Anliker and Moser 1987; Anliker et al. 1988). The MWs for the 36 substances range from about 307 to 1127 g/mol. Six of the 13 pigments have an MW >600 g/mol, and all of the dyes are <600 g/mol. The Deffs were estimated and reported for some of these pigments and dyes and ranged from about 0.84 to 1.86 nm (determined for only 2 of the 13 pigments). Log KOW estimates range between about 0.7 and 17.4, estimates for solubility in water range between about 10−17 and 102 mg/L, and solubility in octanol ranges between <0.05 and 2430 mg/L (Anliker and Moser 1987; Anliker et al. 1988). The log BCF values range from about 0 to 1.76 for the 23 dyes and for 2 of the pigments and are from the CSCL data set. The precise identities for the different pigments and dyes reported in these studies are uncertain, because neither the names nor the CAS numbers were reported. Many of these substances have very low water solubilities, suggesting that BCF tests might have been conducted with exposure concentrations above solubility limits (see discussion above for the CSCL data set). Based on these data, bioaccumulation testing exemptions have been proposed (see Table 1; Anliker et al. 1988; Horrocks et al. 2006).

Solubility cutoffs in octanol and water

It is acknowledged that a substance with very low water solubility in both octanol and water is unlikely to establish a high concentration in fish and thus is unlikely to exert a toxic effect. The 2-film model requires that the substance be soluble in both aqueous and organic phases. Water and octanol solubility limits have been proposed as restricting the passive absorption of sparingly soluble pigments and dyes and perhaps other substances. For example, Anliker et al. (1988) propose that bioaccumulation tests are not necessary for pigments with equation image <0.1 mg/L and equation image <10 mg/L. The estimated water solubility for OCDD is about 10−10 mmol/L (Mackay et al. 1999), and OCDD is absorbed at the gills and in the diet, which suggests that water solubility would have to be <10−10 mmol/L to prevent passive diffusion. For certain organic pigments, KOW estimates based on ratios of solubilities in octanol and water (about 1–6) are much lower than QSAR estimates (about 105.8–107.5), so low absorption and bioaccumulation are expected. Low solubility in membranes (octanol) may slow permeation kinetics; however, if the solubility in water is much lower than solubility in membranes, then the ratio of solubilities, i.e., KMW, may still support diffusion and subsequent bioaccumulation. These assumptions generally suffer from a lack of definitive studies and do not preclude other possible mechanisms of chemical uptake. Certain pigments may behave as particles because of their propensity to aggregate; therefore, solubility measurements for the truly dissolved molecule in solvents such as water and octanol must be interpreted with care. The aqueous solubility of solid substances is strongly influenced by the melting point and the fugacity ratio. Substances such as pigments with melting points exceeding 350 °C must have very low solubilities in both octanol and water, casting doubt on the relevance and feasibility of measuring accurate KOW values for these substances.


Pharmaceuticals, nanoparticles (NPs, 1–100 nm), and larger fine particles released to the environment are emerging concerns for environmental health. In addition, mammals such as humans are included in regulatory objectives; therefore, pharmacological and NP studies were explored to evaluate the application of molecular size criteria or indicators for screening bioaccumulation potential.

Lipinski's “rule of 5” is often used as a guideline to identify effective pharmaceuticals that are passively absorbed and rapidly distributed to sites of therapeutic action (Lipinski et al. 2001). The rule of 5 predicts that poor absorption or permeation is more likely when there are more than 5 H-bond donors, 10 H-bond acceptors, MW >500, and log KOW >5. Lipinski's guidelines are generally consistent with the objectives of drug design (i.e., effective delivery to target sites); however, these guidelines are not well suited as exclusive indicators for identifying bioaccumulative hazards (Escher and Weisbrod 2006; Gobas et al. 2006).

Absorption mechanisms other than passive diffusion may be important for certain substances. There are believed to be >400 transport protein families implicated in the transport of chemicals (Dahl et al. 2004). Compounds such as β-lactam antibiotics, probencid, nonsteroidal anti-inflammatory drugs, and methotrexate are also transported across the plasma membrane by the organic anion transport (OAT) system (Eraly et al. 2004). Although these carrier systems did not evolve for contaminant transport, molecular mimicry between exogenous and endogenous compounds may apply to some contaminants. For example, the OAT system is suspected to influence elimination kinetics for certain perflourinated organic surfactants (Yoo et al. 2009). There are numerous instances of ligand receptor-mediated endocytosis (RME) illustrating the internalization of large molecules (Simionescu et al. 2002). Botulinum toxin is made up of 2 polypeptide segments (100 000 and 50 000 g/mol), and uptake from the GIT has been shown to occur via RME (Simpson 2004). Macromolecules such as polyvinyl pyrrolidone (40 000 g/mol), bovine serum albumin (60 000 g/mol), and immunoglobulin G (about 150 000 g/mol) and particles with diameters >200 nm and up to 2000 nm are absorbed by endocytosis (van Hoogdalem et al. 1989). The uptake of very large substances (about 150 000 nm diameter) may also occur by paracellular routes (e.g., persorption, through gaps caused by cells sloughed off; van Hoogdalem et al. 1989).

Environmental bioaccumulation assessments for NPs are currently limited; however, aquatic exposures of NPs and fine particles (20–1,000 nm) to the aquatic invertebrate Daphnia magna show absorption of these substances through the GIT (Rosenkranz et al. 2009). Most available accumulation data for NPs and fine particles are from human health studies. Nanoparticles seem to be absorbed by nonspecific and targeted (RME) endocytosis as well as by diffusion. Nanoparticle size has a role in determining the rate and degree of uptake by endocytotic mechanisms. Certain NPs with diameters of about 50 nm were taken up by mammalian cells at rates faster than other sizes, and NPs with a radius <12 nm appear to be too small to trigger rapid internalization individually and must cluster before significant uptake occurs (Chithrani and Chan 2007). Poly(lactic-coglycolicacid) (50–1000 nm, MW 40 000–75 000 g/mol) absorption in Caco-2 cell lines was maximized for particles with diameters of about 100–200 nm and decreased for particles with diameters of 50, 500, and 1000 nm (Win and Feng 2005). In vitro human lung fibroblast uptake studies showed that processes on the surface of the cell were faster than the physical transport to the cell for substances ranging in hydrodynamic diameter from 25 to 500 nm (Limbach et al. 2005). Aerosol NPs with median diameters of 22 nm are absorbed in rats via exposure to the lungs, and, after 1 h of exposure, 24% of the particles were located within and beyond the epithelial barrier (Geiser et al. 2005). The NPs found within cells were not membrane bound, and the authors proposed a transport mechanism of diffusion rather than endocytosis. Rothen-Rutishauser and colleagues (2006) examined human red blood cell absorption of NPs and larger particles (about 20–1000 nm diameter) consisting of different materials and charges. These nonphagocytic cells lack the traditional cell surface molecules involved in the endocytic pathways yet show accumulation of both negatively charged and noncharged particles (20–200 nm). The mechanism for the translocation of NPs in this study could not be identified.


We conclude that proposed hypotheses for molecular size cutoff criteria for use as bioaccumulation potential indicators are not supported by a critical review of the available data. The proposed hypotheses must be considered false, insofar as a cutoff cannot be factually demonstrated. Adopting a precautionary approach requires that a cutoff not be used to eliminate chemicals as nonbioaccumulative in preliminary screening assessments. Proposed molecular size cutoff values have increased over time with improvements in analytical methods and bioaccumulation knowledge and as more measured data have become available. For larger substances, reliable bioaccumulation measurements are still extremely limited; however, chemicals with molecular size parameters greater than proposed cutoff values are bioaccumulative according to regulatory criteria (e.g., BCF or BAF >5000). Data used to develop most cutoff criteria and indicators have not been critically evaluated. BCFs are not recommended for determining molecular size criteria. Factors such as bioavailability and metabolic biotransformation can obscure the interpretation of bioaccumulation potential, and BCF tests do not include dietary exposures. Bioavailability and dissociation in water can reduce the apparent bioaccumulation potential, giving the impression that certain chemicals are not absorbed because of steric phenomena. Substances that are appreciably biotransformed should not be included in databases used for developing molecular size criteria unless the biotransformation rates can be appropriately quantified.

Bioaccumulation and chemical absorption are complex functions of diverse physiological processes, characteristics of the absorption site, and physical-chemical properties. Molecular size influences solubility and diffusivity in water and organic phases (membranes), and larger molecules may have slower uptake rates. These same kinetic processes apply to diffusive routes of chemical elimination (i.e., slow in = slow out). Thus, bioaccumulation potential remains for substances that are subject to slow absorption processes unless they are biotransformed or eliminated by some other process at rates sufficient to reduce this potential. Even substances with low bioaccumulation potential can still pose substantial risks if emissions to the environment or toxicity are high (Arnot and Mackay 2008). Rather than applying cutoff values for single or multiple categories of chemical properties, an integrated holistic approach to account for the competing rates of uptake and elimination is recommended for assessing bioaccumulation potential. For fish, the BAF provides this information. In the absence of reliable BAF data, BAF models that include mechanistic principles of uptake and elimination such as absorption efficiency can provide estimates of bioaccumulation potential. The mass-balance BAF models and required QSARs (e.g., KOW or KMW) may not be applicable to all chemicals, because there are few measured data available for testing and refining the models. Uncertainties in the data and the models can be reduced through integrated testing strategies by obtaining quality in vivo and in vitro measurements for substances and chemical classes that are not currently well represented. Dietary exposure studies may improve bioaccumulation assessments for hydrophobic substances; however, it should be recognized that certain substances may be transformed in the GIT or in the liver more readily than other routes of exposure (e.g., gills, lungs). QSARs for biotransformation rates (Arnot et al. 2009) and molecular size parameters (Dimitrov et al. 2002) can be coupled with mass-balance bioaccumulation models to select larger, poorly biotransformed chemicals for testing chemical absorption kinetic hypotheses.


We thank the Natural Sciences and Engineering Research Council (NSERC) of Canada and Environment Canada for financial support. We thank an anonymous reviewer for helpful comments on an earlier report and James Armitage and three anonymous reviewers of the manuscript.