J. Pestic. Sci.

47(1), 8–16 (2022)

DOI: 10.1584/jpestics.D21-030

Regular Article

Criterion of molecular size to evaluate the bioaccumulation

potential of chemicals in fish

Chiyoko Miyata,1,2,* Yoshihide Matoba,1 Makiko Mukumoto,1 Yoshiaki Nakagawa2 and

Hisashi Miyagawa2
1
 Sumitomo Chemical Co., Ltd. Environmental Health Science Laboratory, 1–98 Kasugadenaka 3-chome, Konohana-ku, Osaka 554–8558, Japan
2
 Division of Applied Life Sciences, Graduate School of Agriculture, Kyoto University, Kyoto 606–8502, Japan

(Received June 6, 2021; Accepted October 27, 2021)

To evaluate the bioaccumulation potential of chemicals in fish, a molecular-size descriptor,

Dmax aver, has been used as a weight of evidence under the EU REACH. The Dmax aver value, however,
is estimated on the basis of 3-D structures of possible stable conformers in a vacuum using OASIS
software that requires expertise upon parameter input. We developed a method to calculate the
3-D conformers in water, which is more suitable for bioaccumulation potential evaluation in an
aquatic environment, by introducing MD simulation. By examining the relationship of the calcu-
lated molecular size of 1665 chemicals with their reported BCF values, we found that 17.1 Å of
Dmax aver or 15.6 Å of Dmax min was a threshold of molecular size in water to predict the low bioac-
cumulation (i.e., BCF<5000) of a chemical. Setting this threshold as a new standard would reduce
the number of animal tests without compromising the quality of safety evaluation.

Keywords: molecular size, quantum chemical calculation, molecular dynamics simulation, bioaccumulation, bioconcentration.

tion. Those having a low potential to permeate biological mem-

Introduction branes or little chance of direct and indirect exposure by living
Persistent, bioaccumulative, and toxic chemicals (PBTs) and per- organisms in the aquatic environment are also exempted from
sistent organic pollutants (POPs) are controlled under various the bioconcentration test.4,5) This test is conducted to determine
national and international regulatory frameworks, including the the bioconcentration factor (BCF) of a chemical in fish. The
Stockholm Convention,1,2) due to their negative effects on the BCF is defined by the ratio of the steady-state concentration of
environment. Bioaccumulation is the most critical issue for hu- a chemical in fish to that in water after the fish take the chemi-
mans and environmental organisms from the viewpoint of the cal in water through gill membranes during a specific time pe-
food chain. In Japan, under the Japanese Chemical Substances riod.6,7) If a chemical has a BCF ≥5000, it is regarded as highly
Control Law (CSCL), since 1974, an aqueous bioconcentration bioaccumulative under the Stockholm Convention and the
test has been required in principle for every chemical.3) In the CSCL.1)
EU, the bioconcentration test is required for chemicals produced Currently, typical protocols of the bioconcentration test are
in quantities greater than 100 tons per year under REACH (Reg- relatively laborious and time consuming. They also require large
istration, Evaluation, Authorization and Restriction of Chemi- numbers of experimental fish, which is undesirable from an ani-
cals) unless the chemical has a low potential for bioaccumula- mal welfare standpoint. Thus, alternative methods of assessing
the bioaccumulation potential of chemicals have been applied
using their physicochemical parameters, such as the octanol–
* To whom correspondence should be addressed. water partition coefficient and the molecular weight (MW).8)
E-mail: [email protected] Some of these methods focus on the permeation process
Published online January 11, 2022 through biological membranes when organisms take in a chemi-
  © Pesticide Science Society of Japan 2022. This is cal, and they assume its molecular size as an uptake-constraint
an open access article distributed under the Creative Commons Attri- factor.9–12) The basic idea is that bulky molecules cannot perme-
bution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND ate the membranes: for example, a chemical with MW >800 is
4.0) License (https://creativecommons.org/licenses/by-nc-nd/4.0/) approved as low bioaccumulative, and a bioconcentration study
Vol. 47, No. 1, 8–16 (2022) Criterion of molecular size to evaluate bioaccumulation potential  9

is not required under the CSCL.2,13) Mekenyan et al. have pro- domly selected and rotated by every 120 degrees. The quantum
posed a Dmax aver value as one of the steric factors to characterize calculation was stopped if the number of the generated structures
the bulkiness of a molecule14,15); Dmax aver >17.4 Å is approved as exceeded 5 million. The number of 5 million was determined by
one indicator in a weight of evidence assessment, a method for considering the calculation cost. It covers that of all conformers
making decisions by considering multiple sources of informa- having 14 dihedral angles in a molecule where every angle was
tion,16) for limited bioaccumulation potential (i.e., BCF <5000) rotated by every 120 degrees (314=4,782,969). To verify whether
under REACH.17) the number of 5 million was sufficient, five different initial 3-D
The Dmax aver value is an arithmetic mean diameter of the structures with Dmax values different from each other by >3 Å
smallest spheres accommodating the locally stable conform- were submitted to HyperChem as very flexible chemicals with
ers. This value can be obtained only by using commercial soft- more than 14 dihedral angles.
ware OASIS, developed by Mekenyan et al.17,18) Although the The free energy gradient method was applied to the geometry
Dmax aver value is a leading indicator to describe the steric prop- optimization to obtain a locally stable conformer. This optimiza-
erty of a chemical, this indicator alone is considered insufficient tion was performed in two steps. In the first step, the conformers
to exempt a chemical from bioconcentration testing.16,17) Due to generated by MCMM were roughly optimized by the molecular
some problems, countries other than the EU have not approved mechanics method using an MM+ force field21) with a Polak–Ri-
it. The most critical problem appears to be that OASIS estimates biere Conjugate Gradient.22) In the second step, each of the lo-
the Dmax aver value of a chemical based on its conformations in a cally stable conformers by the molecular mechanics method was
vacuum,14,15) although the bioaccumulation in fish occurs in an re-optimized by the semi-empirical molecular orbital method
aquatic environment. Since the simulation of molecular struc- using AM1/6-31G+15) with the Polak–Ribiere Conjugate Gradi-
tures in water requires enormous computational cost in a quan- ent. The root-mean-square gradient convergence threshold of
tum chemical calculation that considers the interactions with the Polak–Ribiere Conjugate Gradient was set to 0.01 kcal/Å mol
surrounding water molecules, Mekenyan and coauthors selected in each optimization.
locally stable conformers within 20 kcal/mol of the most stable After these optimizations, locally stable conformers with po-
one in a vacuum in the OASIS calculation to consider the struc- tential energies 20 kcal/mol higher than that of the most stable
tural differences between in a vacuum and in water.15) However, one were excluded in accordance with the method for Dmax aver
scientific discussions have not validated this selection, such as a calculation in OASIS.17) Conformers with a similar molecular
comparison with actual Dmax aver of conformers in water. More- size were also excluded. The molecular similarity was validated
over, in calculating Dmax aver values with OASIS software, expert by a root-mean-square distance calculation between the two
knowledge is required for inputting parameters in the genetic al- atoms at both ends of the conformers. When differences in the
gorithm, such as the size of the permanent population, the num- root-mean-square distance were less than 0.1 Å, these conform-
ber of children, the mutation/crossover ratio, and the generation ers were considered similar, and one of them was selected to ex-
times.19) clude others. These 3-D conformers obtained by HyperChem are
In this study, we developed a new method of calculating described as “complete locally stable conformers” in the follow-
Dmax aver values in a vacuum that requires no expert knowledge. ing sections.
The method combined a commercial program to generate a set The locally stable conformers that emerged around the gentle
of energetically stable conformers in a vacuum with an origi- bottom in the same valley on the potential energy surface were
nal program to calculate Dmax aver from the generated conformer further excluded by an original Excel program using Visual
set. Then, we reproduced the dynamic behavior of molecules in Basic for Applications (VBA) because these similar conformers
water using the molecular dynamics (MD) simulation method, were likely to provide a biased weight in calculating the Dmax aver
and the calculated Dmax aver values were compared with those in a value. In the program, the root-mean-square distances between
vacuum. Finally, based on reliable BCF values of 1665 chemicals, all corresponding two atoms of a pair of locally stable conform-
we proposed a new threshold of Dmax value in water to judge the ers were calculated according to Eq. (1),
low bioaccumulation potential (i.e., BCF <5000) of chemicals in
an aquatic environment.
d= (Δx A
ij − ΔxijB )2 (1)
Materials and methods i, j

1.  Calculation method where Δx Ai,j and Δx Bi,j indicate the distance between atoms i and
1.1.  Conformers in a vacuum j for locally stable conformers A and B, respectively. A smaller
An initial 3-D structure of a test chemical was drawn using d value indicates a higher similarity between two locally stable
CHEM3D Ver.18 (PerkinElmer Informatics, USA). This initial conformers. Thus, of similar conformers within a small d value,
structure was submitted to the Monte Carlo Multiple Minimum one representative conformer was selected to exclude others
(MCMM) search algorithm20) of HyperChem 8.0.6. (HyperCube, from the complete locally stable conformers. The maximum
USA) to generate 3-D conformers by a quantum chemical calcu- number of conformers for Dmax calculation was changed from
lation, in which the dihedral angles of each molecule were ran- 10 to 200 in this study by varying the d value, and Dmax aver val-
 10  C. Miyata et al. Journal of Pesticide Science

ues based on the selected conformers were compared with those x2. This operation was repeated y times. Then the moving dis-
given by OASIS. The conformers selected by specifying the d tance was changed to (hxi+vxi)/22, and the point was moved by
value are described as “filtered stable conformers.” (hxi+vxi)/22 to the atom farthest from the arrival point by repeat-
1.2.  Conformers in water ing. This operation was repeated y times. In this way, the point
Three-dimensional structures of each test chemical in water gradually reached the true center of the sphere by decreasing the
were generated by MD simulations using Material Studio 2018 moving distance to (hxi+vxi)/2n. The number of y was set at 20,
(BIOVIA, USA). The simulation was carried out in a cube con- based on the convergence conditions investigated for this meth-
taining 2000 water molecules with a density of 1.0 g/cm3 by the od. We judged that the point had arrived at the center when the
Forcite module. The cubic size was set at 39.1 Å in length, which moving distance got to less than 1×10−6 Å.
was twofold longer than the limit of the Dmax aver of the low bio-
accumulation potential (i.e., 17.4 Å). The initial 3-D structure 2.  Test chemicals
was drawn using CHEM3D. The cubic box containing the test Dmax values of 69 chemicals of various sizes and shapes were cal-
chemical and water molecules was energy-minimized by the culated for the conformers in a vacuum or water (Table 1). These
Smart algorithm in the Forcite module with a Condensed-Phase chemicals were described in a report published by the UK Envi-
Optimized Molecular Potentials for Atomistic Simulation Stud- ronmental Agency18) and consisted of biphenyl ethers, polychlo-
ies (COMPASS) II force field.23) Convergence tolerance for mini- rinated alkanes, polysulfides, styrylphenols, primary alcohols,
mization was Medium, with energy of 0.001 kcal/mol, a maxi- pigments, and 29 chemicals with various skeletons.
mum force of 0.5 kcal/mol·Å, and a maximum displacement of Test chemicals with measured BCF values were collected from
0.015 Å. After the energy minimization, the MD simulation was the database of the Japanese National Institute of Technology and
performed with a time step of 1.0 fs by using an NTP ensemble Evaluation (NITE). Bioconcentration tests with 750 chemicals
in the Forcite module of Material Studio, where a Nosé-Hoover were conducted by the Japanese Ministry of Economy, Trade and
thermostat and a Berendsen barostat24–26) were used to keep the Industry (METI) from 1974 to 1986,29) and tests with 1118 chem-
temperature and pressure at 298 K and 1×105 Pa, respectively. icals were conducted by applicants under the CSCL from 1987 to
The MD simulations in water for a flexible chemical were per- 2017.30) These tests were performed under the Organisation for
formed for 10, 100, 200, 500, and 2000 ps. The snapshots were Economic Co-operation and Development (OECD) Principles of
collected every 1 ps. The 3-D coordinates of all of the atoms in Good Laboratory Practice31) following OECD Test Guideline 305
each conformer were used for the following Dmax calculation. (I).6) The BCF values were used as described in these test results,
The effect of the simulation time was evaluated by the repeat- except for the following cases: When more than two BCF values
ability of the calculated Dmax value and of the range of fluctua- were available for a single chemical, the highest was selected for
tion during calculation. conservative analysis. When a BCF value was given for a mixture
1.3.  Calculation of Dmax values of different chemicals, the original reports and references were
The coordinates of all atoms of the conformers calculated in sec- examined, and only the value that could be assigned to a single
tions 1.1 and 1.2 were used to calculate the diameter of a sphere component was used. Otherwise, the values were excluded from
covering the van der Waals radii27) of all atoms in a conformer the analysis. Inorganic chemicals and heavy metal–containing
(i.e., Dmax) using an original Excel program with VBA. In this organic chemicals were excluded from the investigation because
program, the center of the sphere was first determined,28) and their uptake mechanism to fish differs from passive diffusion
then the distance between the sphere’s center and the edge of through the gill membrane.32,33) For ionic chemicals in water,
the van der Waals radius of the farthest atom was calculated. the dissociated structures were used for the calculation to avoid
The Dmax value was obtained by doubling the distance. The Dmax overestimating their molecular size. Perfluoroalkyl chemicals
value was calculated for each of the complete locally stable con- have been reported to bioaccumulate in organisms with the as-
formers in a vacuum, of the filtered stable conformers in vacu- sistance of specific binding proteins in the blood.34–36) For these
um, or of the conformers in water. Dmax aver, Dmax min, and Dmax max substances, Dmax values are not relevant as the uptake-constraint
values of a test chemical were determined by averaging the Dmax factor; hence, they were excluded from the investigation. Some
values arithmetically and by selecting the minimum and maxi- chemicals (BCF <100 and MW <100) were excluded from the
mum among the Dmax values, respectively. calculation because they were too small and not relevant to con-
The center of the sphere was determined as follows: a mov- cerns of high bioaccumulation. As a result, Dmax values in water
ing point was first set at the origin of the 3-D coordinate, and were determined for 1665 chemicals.
it was moved toward the farthest atom (x1) from the origin by
a distance of (hx1+vx1)/2, where hx1 was the distance between the
Results and discussion
origin and the center of the farthest atom (x1), and vx1 was the 1. Dmax values in a vacuum
van der Waals radius of the atom. Next, the point was further Dmax values in a vacuum were estimated by HyperChem and
moved from there toward the farthest atom (x2) by a distance of the original program for 69 chemicals listed in the UK report.18)
(hx2+vx2)/2, where hx2 was the distance between the arrival point The total number of rotational dihedral angles in each chemi-
and the center of x2, and vx2 was the van der Waals radius of cal ranged from 1 to 39 (Fig. 1A). Five minutes to two weeks
Vol. 47, No. 1, 8–16 (2022) Criterion of molecular size to evaluate bioaccumulation potential  11

Table 1.  Classification and typical structures of 69 chemicals18) used to calculate Dmax values in a vacuum or in water

Chemical groups Typical chemical structures No. of chemicals

Biphenyl ethers 11

Polychlorinated alkanes 22

Polysulphides 6

Styrylphenols 5

Primary alcohols 2

Pigments 3

Other chemicals 29

were required to obtain the complete locally stable conformers A long-chain chlorinated alkane ID33 (C30H49Cl13) has
by changing the respective dihedral angles. The numbers of the the highest rotational freedom and, therefore, has the larg-
complete locally stable conformers ranged from 1 to 717 (Fig. est number of 3-D conformers among the chemicals exam-
1B), depending on the number of the dihedral angles in the ined. Dmax values ranged from Dmax min=17.8 Å (the smallest) to
molecule that gave rise to higher freedom of motion. Dmax max=38.9 Å (the largest), as shown in Fig. 2. The most stable

Fig. 1.  Numbers of dihedral angles (A) and the complete locally stable conformers (B) for 69 chemicals with ID numbers described in the UK report.18)
 12  C. Miyata et al. Journal of Pesticide Science

using this method without any expert knowledge.

For the 69 chemicals written about in the UK report, Dmax
values calculated from the complete locally stable conformers
were correlated with those by OASIS18) with a slope of linear
regression equation of 0.992 and a coefficient of determination
(R2) of 0.985 (Fig. 3A). However, when the potential energy sur-
face gradient was gradual, some similar conformers could con-
verge into different multiple locally stable conformers satisfying
the convergence threshold (i.e., 0.01 kcal/Å mol). If such con-
formers were deemed to be independent, their Dmax values could
substantially affect the calculation of Dmax aver, making it unreli-
able by overweighting their contribution. To eliminate the effect
of highly similar conformers on the calculation of Dmax aver, the
filtering conditions for excluding them were examined. By ad-
justing the d value as defined by Eq. (1), the maximum number
of filtered stable conformers was changed from 10 to 200. The
Fig. 2. Three-dimensional models of locally stable conformers and their
Dmax values for chemical ID33 (C30H49Cl13).
maximum number of complete locally stable conformers in this
chemical set was 717. The calculation of Dmax aver for each set
of filtered stable conformers revealed that 100 conformers had
conformer of ID33 has a Dmax stable value of 23.2 Å, whereas the the highest agreement with Dmax aver given by OASIS, where the
average Dmax value (Dmax aver) was 26.4 Å, based on the complete slope of the linear regression equation was 0.996, and the coef-
locally stable conformers. ficient of determination (R2) was 0.996 (Fig. 3C). When the con-
The initial 3-D structure of each chemical was submitted to formers were filtered to 200, the degree of agreement was lower,
HyperChem to generate a set of up to 5 million conformers. In probably due to the incomplete removal of similar conformers
order to verify whether the number of 5 million was sufficient, (Fig. 3D). On the other hand, when the number of filtered con-
flexible long-chain chlorinated alkanes (i.e., ID29 to ID33) with formers was less than 50, some of the structures that impor-
more than 14 dihedral angles in a molecule were selected from tantly contributed to Dmax aver were likely excluded, also lowering
the 69 test chemicals. Five initial structures that had Dmax values the degree of agreement (Fig. 3B shows the result using 30 con-
different from each other by >3 Å were submitted to HyperChem formers). The differences between the Dmax aver values calculated
to generate the stable conformers and calculate their Dmax aver val- from the 100 filtered stable conformers and those calculated by
ues for ID29 to ID33. Table 2 shows the ranges of the number of OASIS were within ±0.60 Å for the 69 chemicals, while, taking
the complete locally stable conformers, energy of the most stable the complete locally stable conformers into consideration, the
conformer, and Dmax aver values calculated from the complete lo- differences were within ±1.17 Å. Thus, the maximum number of
cally stable conformers, which were derived from five different filtered stable conformers for each chemical was set to 100 in the
input structures of each test chemical. While the number of the following investigations.
complete locally stable conformers varied depending on the shape To summarize the results in a vacuum, the computational pa-
of the potential energy surface and/or the convergence condi- rameters are clearly defined and unequivocal in this study, and the
tions of each input structure, the differences in the energies and calculation results are not affected by the initial conformation of a
Dmax aver values were within 0.3 kcal/mol and ±0.04 Å, respectively. chemical. Everyone can perform the same quality of calculations
As a consequence, we confirmed that 5 million was an appropri- as OASIS without expertise, which is definitely advantageous.
ate number of random conformer generation even for test chemi-
cals with more than 14 dihedral angles, and the Dmax aver value 2. Dmax values in water
was independent of a structure input into HyperChem. This high The effect of simulation time on the Dmax calculation was exam-
repeatability suggests that everyone can obtain the same results ined using the most flexible chemical ID33 (C30H49Cl13) among

Table 2.  Repeatability of Dmax aver calculation in a vacuum by HyperCHEM starting from five different 3-D strctures of each test chemical

No. of No. of complete locally Energy of most stable Dmax aver of complete locally
Chemical ID Formula
dihedral angles stable conformers conformer (kcal/mol) stable conformers (Å)
ID29 C22H37Cl9 27 648–659 32.2–32.3 21.77±0.02
ID30 C24H40Cl10 30 657–666 37.5–37.8 23.40±0.03
ID31 C26H43Cl11 33 660–674 39.9–40.1 23.63±0.02
ID32 C28H46Cl12 36 661–672 36.2–36.3 24.87±0.04
ID33 C30H49Cl13 39 698–717 50.3–50.6 26.35±0.02
Vol. 47, No. 1, 8–16 (2022) Criterion of molecular size to evaluate bioaccumulation potential  13

Fig. 3. Comparison of Dmax aver calculated by OASIS with those calculated in this study from the complete locally stable conformers (A) and from the
filtered stable conformers with the numbers of 30 (B), 100 (C), and 200 (D).

the 69 chemicals. Seven different 3-D structures (A to G) of the ranged from 25.2 (structure C) to 35.3 Å (structure G), and
chemical ID33 were created by CHEM3D, whose Dmax values the difference was 10.1 Å. These results indicated that both the
were 32.1, 28.2, 25.0, 25.8, 30.3, 23.2, and 35.3 Å, respectively. Dmax min and Dmax max values were strongly affected by the initial
Of these, structure F represented the smallest among the avail- structures of a test chemical in the case of the 10 ps simulation,
able initial structures, while G represented the largest. The MD and such a short simulation time was insufficient to cover a wide
simulations for each structure in water were performed for a range of possible structural changes of a chemical having a flex-
specified time up to 2000 ps. Figure 4 shows the averages and the ible structure like ID33. By extending the simulation time from
ranges of Dmax values after simulation times of 10, 100, 200, 500, 10 to 2000 ps, the differences of Dmax min, Dmax max, or Dmax aver
and 2000 ps for each initial structure. The conformers for the values became smaller, ultimately being less than 0.1 Å between
Dmax calculation were sampled every 1 ps in principle, although any pair of two initial structures after 500 ps or more. In fact,
the conformers for the 10 ps simulation were sampled every the standard deviations of Dmax min, Dmax max, and Dmax aver among
0.1 ps to increase the number of the conformers. The small- these seven structures in the 10 ps simulation were 3.50, 4.03,
est Dmax values (Dmax min) in the 10 ps simulation ranged from and 3.84 Å, respectively; however, those in the 500 ps simula-
23.2 (structure F) to 32.2 Å (structure G), and the difference tion decreased to 0.07, 0.09, and 0.03 Å, respectively. Eventually,
was 9.0 Å. On the other hand, the largest Dmax values (Dmax max) Dmax values ranged from 23 to 35 Å depending on the simula-
tion time; however, the maximum or minimum values of Dmax
obtained after 500 ps simulations were unchanged by extending
the simulation time as in the case of chemical ID33. When sev-
eral other test chemicals were examined in the same manner, all
simulations converged in a shorter simulation time than that of
ID33. Therefore, the simulation time was set to 500 ps in the fol-
lowing investigations, taking into account the calculation costs.

3.  Comparison of Dmax values in a vacuum and in water

Dmax values of the 69 chemicals in water were compared with
those in a vacuum as calculated in this study (Fig. 5). Although
the maximum Dmax values (Dmax max) in water were almost the
same as those in a vacuum (Fig. 5A), the minimum Dmax values
Fig. 4. Average (black square) and range (vertical bar) of Dmax values
in water during calculation times of 10, 100, 200, 500, and 2000 ps when
(Dmax min) in water were different from those in a vacuum in the
starting from seven different 3-D structures (A to G) of chemical ID33 case of polychlorinated alkanes with a high degree of freedom
(C30H49Cl13). (open circles in Fig. 5B). For example, the differences of Dmax min
 14  C. Miyata et al. Journal of Pesticide Science

Fig. 5. Comparison of Dmax max (A), Dmax min (B), and Dmax aver values (C) in water based on MD simulations with those in a vacuum based on quantum
chemical calculations for 69 chemicals (closed circles) including ID12 to ID33 (open circles) and ID D2 and D3 (open squares). A line with slope 1 is
shown in each figure.

values in water and in a vacuum for ID12 to ID33 were 1.8 to Under the CSCL, chemicals with MW >800 are approved as
6.5 Å, where the values in water were larger than those in a vac- low bioaccumulative and are not required to be tested in fish.2,13)
uum. These chemicals can adopt highly compact conformations When applying this standard, only 116 of the examined 1665
in a vacuum, but such compact conformations would hardly be chemicals (open triangles in Fig. 6) are exempted from biocon-
formed in water due to their interaction with water molecules. centration testing using fish. In this study, the Dmax values de-
As a result, the average values of Dmax (Dmax aver) in a vacuum rived from realistic 3-D conformers in water were considered
were not in good agreement with those in water because some of more appropriate than MW as the indicator of molecular bulki-
the stable conformers in a vacuum cannot exist in the aqueous ness that may govern the penetration of biological membranes.
environment (Fig. 5C). Similarly, two pigments, ID D2 and D3 Based on the Dmax value in water calculated in this study, it was
(open squares in Fig. 5C; see their structures in Table 1), with expected that 436 chemicals with Dmax aver >17.1 Å could be ex-
higher molecular weights (i.e., around 700) also showed differ- cluded from testing.
ent Dmax aver in a vacuum and in water. In general, Dmax aver values While the Dmax aver values have conventionally been deemed
in water were larger than those in a vacuum for bulky chemicals. the criterion of low bioaccumulation, the Dmax min values more
These results indicate that the MD calculations help correctly es- likely reflect on the stereo structures of compact conformers that
timate the 3-D stereo structures of molecules in water, especially can be more closely associated with gill membrane permeation.
for some flexible or bulky molecules. Thus, from the viewpoint of screening assessment, the use of
Dmax min values as thresholds seems more conservative and eas-
4.  Evaluation of bioaccumulation potential with the Dmax value ily accepted by regulatory authorities. Figure 7 shows the rela-
in water tionship between the measured BCF values and the Dmax min val-
Mekenyan et al. selected the conformers in a vacuum that can ues in water for the same set of chemicals as in Fig. 6. Here, the
exist within 20 kcal/mol of the most stable conformer without maximum Dmax min value for the chemicals with BCF ≥5000 was
any validations, and the 17.4 Å of Dmax aver based on these con- 15.6 Å. The number of chemicals regarded as having low bioac-
formers has been proposed as an indicator of low bioaccumula-
tion.16,17) In this study, we calculated the molecular sizes in water
for chemicals that had been subject to bioconcentration studies
and tried to find a size threshold to predict the low bioaccumu-
lation of chemicals.
Figure 6 shows the relationship between the measured BCF
values and the Dmax aver values in water for the 1665 chemicals,
including 116 substances with MW >800 (open triangle), to-
gether with a dotted line at BCF=​5000. The maximum Dmax aver
value for the chemicals with BCF ≥5000 was 17.1 Å among the
1665 chemicals. The number of chemicals with Dmax aver >17.1 Å
was 436, and all of them had low bioaccumulation (i.e., BCF
<5000). Therefore, a criterion of the low bioaccumulation in
terms of Dmax aver in water could be set as >17.1 Å, which was
in good agreement with the Dmax aver in a vacuum of 17.4 Å ap-
proved under REACH.17) This result indirectly justifies the ap-
Fig. 6. Relationship between Dmax aver values in water and measured
proximation by Mekenyan et al. to predict the Dmax values in BCF values for 1665 chemicals (closed circles) including substances
water, where only the conformers in a vacuum within 20 kcal/ >800 MW (open triangles). The dotted line denotes the threshold of
mol of the most stable one were selected for the calculation. BCF=5000.
Vol. 47, No. 1, 8–16 (2022) Criterion of molecular size to evaluate bioaccumulation potential  15

Analysis using the measured BCF values of 1665 chemicals re-

vealed that chemicals in water with Dmax aver >17.1 Å or Dmax min
>15.6 Å had BCF values of less than 5000. This threshold would
contribute to reducing animal testing without diminishing the
conventional screening accuracy based on molecular weight.

References
  1) D. Wijk, R. Chénier, T. Henry, M. D. Hernando and C. Schulte: In-
tegrated Approach to PBT and POP Prioritization and Risk Assess-
ment. Integr. Environ. Assess. Manag. 5, 697–711 (2009).
  2) S. Jacobi, B. Brown, M. Claessens, I. Colombo, M. Embry, J. Frank-
lin, B. Hidding, M. L. Paumen, N. McGrath, C. Miyata, K. Wood-
burn and H. Vrijhof: Information to be considered in a weight-of-
evidence-based PBT/vPvB assessment of chemicals (Annex XIII of
REACH). ECETOC Special Report No. 18 (2014).
Fig. 7. Relationship between Dmax min values in water and measured  3) Japan METI: “Act on the Evaluation of Chemical Substances and
BCF values for 1665 chemicals (closed circles) including substances Regulation of Their Manufacture, etc. (Chemical Substances Con-
>800 MW (open triangles). The dotted line denotes the threshold of trol Law).” Ministry of Economy, Trade and Industry, Tokyo, Japan.
BCF=5000. http://www.japaneselawtranslation.go.jp/law/detail/?id=3350&
vm=02&re=01 (Accessed 10 Sep., 2020).
  4) Regulation (EC) No 1907/2006-Registration, Evaluation, Authorisa-
cumulation by the Dmax min threshold was 357, which was some-
tion and Restriction of Chemicals (REACH) https://osha.europa.
what less than that by the Dmax aver threshold used in this study.
eu/en/legislation/directives/regulation-ec-no-1907-2006-of-the-
This result also confirmed that the Dmax min threshold was more european-parliament-and-of-the-council (Accessed 10 Sep., 2020).
conservative than that of the Dmax aver.  5) ECHA: REGULATION (EC) No 1907/2006 OF THE EUROPEAN
This study deduced two thresholds, Dmax aver >17.1 Å and PARLIAMENT AND OF THE COUNCIL of 18 December 2006
Dmax min >15.6 Å, of low bioaccumulation evaluation. These (2006).
values are determined based on the realistic molecular size in  6) OECD guidance for testing chemicals: Bioaccumulation in Fish:
water, so they are considered more reliable thresholds than the Aqueous and Dietary Exposure (2012).
conventional values that assume the molecular size in a vacuum.   7) Guideline of Chemical Substances Control Law: Testing methods for
new chemical substances, Bioconcentration test using fish (Water
In addition to reflecting more realistic structures, the 3-D ste-
exposure method) (2018) (in Japanese).
reo structures of some chemicals were larger in water than in a
  8) J. W. Nichols, M. Bonnell, S. D. Dimitrov, B. I. Escher, X. Han and N.
vacuum (Fig. 5). This means that even the chemicals with Dmax I. Kramer: Bioaccumulation Assessment Using Predictive Approach-
values below the threshold in a vacuum may have values above es. Integr. Environ. Assess. Manag. 5, 577–597 (2009).
the threshold in water in some cases; thus, the use of the Dmax   9) S. Nakai, K. Takano and S. Saito: Bioconcentration Prediction under
value in water can be more advantageous for reducing the num- the Amended Chemical Substances Control Law of Japan. R&D
ber of fish experiments. Between the two threshold values based Report SUMITOMO KAGAKU 2006-I (2006).
on Dmax aver and Dmax min, it seems appropriate to use the former, 10) Y. Sakuratani, Y. Noguchi, K. Kobayashi, J. Yamada and T. Nishi-
since it reflects all conformers that can exist in water. However, hama: Molecular size as a limiting characteristic for bioconcentration
in fish. J. Environ. Biol. 29, 89–92 (2008).
from a regulatory point of view, setting a margin on the thresh-
11) A. Opperhuizen, E. W. Volde, F. A. P. C. Gobas, D. A. K. Liem and J.
old is preferable due to safety concerns. Therefore, we conclude
M. D. Steen: Relationship between bioconcentration in fish and steric
that the more conservative threshold based on Dmax min is more factor of hydrophobic chemicals. Chemosphere 14, 1871–1896 (1985).
suitable. In any case, setting these new thresholds in water 12) F. Grisoni, V. Consonni, S. Villa, M. Vighi and R. Todeschini: Molec-
would contribute to finding a more accurate screening meth- ular size cutoff criteria for screening bioaccumulation potential: Fact
od for determining low bioaccumulative chemicals with BCF or fiction? Chemosphere 127, 171–179 (2015).
<5000 and reducing animal testing. 13) A. V. Weisbrod, L. P. Burkhard, J. Arnot, O. Mekenyan, P. H. How-
ard, C. Russom, R. Boethling, Y. Sakuratani, T. Traas, T. Bridges, C.
Conclusion Lutz, M. Bonnell, K. Woodburn and T. Parkerton: Workgroup report:
Review of fish bioaccumulation databases used to identify persistent,
To evaluate the bioaccumulation potential in an aquatic envi-
bioaccumulative, toxic substances. Environ. Health Perspect. 115,
ronment, Dmax values of chemicals in water were calculated by
255–261 (2007).
an MD simulation. Obtained Dmax aver values in water were gen- 14) S. D. Dimitrov, N. C. Dimitrova, J. D. Walker, G. D. Veith and O. G.
erally comparable to those in a vacuum, but some of the most Mekenyan: Bioconcentration potential predictions based on molecu-
compact conformers in a vacuum (i.e., Dmax min) were not repro- lar attributes—An early warning approach for chemicals found in
duced in water for 69 chemicals in the UK report. Thus, calcu- humans, birds, fish and wildlife. QSAR Comb. Sci. 22, 58–68 (2003).
lating 3-D conformations in water seemed to be helpful for a 15) S. D. Dimitrov, N. C. Dimitrova, J. D. Walker, G. D. Veith and O. G.
more accurately screening bioaccumulation potential in fish. Mekenyan: Predicting bioconcentration factors of highly hydropho-
 16  C. Miyata et al. Journal of Pesticide Science

bic chemicals. Effects of molecular size. Pure Appl. Chem. 74, 1823– 26) H. J. C. Berendsen, J. P. M. Postma, W. F. van Gunsteren, A. DiNola
1830 (2002). and J. R. Haak: Molecular-dynamics with coupling to an external
16) ECHA: Guidance on information requirements and chemical safety bath. J. Chem. Phys. 81, 3684–3690 (1984).
assessment. Part C: PBT/vPvB assessment. Version 2.0. European 27) A. Bondi: van der Waals Volumes and Radii. J. Phys. Chem. 68, 441–
Chemicals Agency, Helsinki, Finland. (2014). 451 (1964).
17) ECHA: Guidance on information requirements and chemical safety 28) K. Ishihata: JPSJ Magazine. 43, 1009–1015 (2002).
assessment. Chapter R.11: PBT assessment. Version 3.0. European 29) National Institute of Technology and Evaluation: The result of bio-
Chemicals Agency, Helsinki, Finland. (2017). concentration. (judged from 1974 to 1986). https://www.nite.go.jp/
18) D. N. Brooke and M. J. Crookes: Calculation of molecular dimen- data/000022750.xlsx (Accessed 31 May., 2021).
sions related to indicators for low bioaccumulation potential. Sci- 30) National Institute of Technology and Evaluation: The result of bio-
ence report SCHO0109BPGT-E-P. Environment Agency, Bristol, UK. concentration (judged from 1987 to 2017). https://www.nite.go.jp/
(2009). data/000108118.xlsx (Accessed 31 May., 2021).
19) O. Mekenyan, D. Dimitorov, N. Nikolova and S. Karabunarliev: Con- 31) OECD Principles of Good Laboratory Practice (1997).
formational Coverage by a Genetic Algorithm. J. Chem. Inf. Comput. 32) National Institute of Technology and Evaluation: Prediction of bio-
Sci. 39, 997–1016 (1999). concentration by category approach. https://www.nite.go.jp/chem/
20) G. Chang, W. C. Guida and W. C. Still: An internal-coordinate Monte qsar/category_approach.html (in Japanese). (Accessed 31 May., 2021).
Carlo method for searching conformational space. J. Am. Chem. Soc. 33) Y. Ikenaga, Y. Sakuratani and J. Yamada: Evaluation of Bioconcen-
111, 4379–4386 (1989). tration Factors by Category Approach for Chemicals. National In-
21) A. Hocquet and M. Langgard: An Evaluation of the MM+ Force stitute of Technology and Evaluation. https://www.nite.go.jp/data/
Field. J. Mol. Model. 4, 94–112 (1998). 000064116.pdf (Accessed 31 May., 2021).
22) L. Grippo and S. Lucidi: A globally convergent version of the Polak- 34) J. W. Martin, S. A. Mabury, K. R. Solomon and D. C. Muir: Progress
Ribiere conjugate gradient method. Math. Program. 78, 375–391 toward understanding the bioaccumulation of perfluorinated alkyl
(1997). acids. Environ. Toxicol. Chem. 32, 2421–2423 (2013).
23) H. Sun: COMPASS: An ab Initio force-field optimized for condensed- 35) C. A. Ng and K. Hungerbühler: Bioconcentration of perfluorinated
phase applicationssoverview with details on alkane and benzene alkyl acids: How important is specific binding? Environ. Sci. Technol.
compounds. J. Phys. Chem. B 102, 7338–7364 (1998). 47, 7214–7223 (2013).
24) S. Nosé: A unified formulation of the constant temperature molecular- 36) C. A. Ng and K. Hungerbühler: Bioaccumulation of perfluorinated
dynamics methods. J. Chem. Phys. 81, 511–519 (1984). alkyl acids: Observations and models. Environ. Sci. Technol. 48,
25) W. G. Hoover: Canonical dynamics: Equilibrium phase-space distri- 4637–4648 (2014).
butions. Phys. Rev. A 31, 1695–1697 (1985).