Estimating GrassâSoil Bioconcentration of Munitions Compounds from

Subscriber access provided by Grand Valley State | University

Article

Estimating Grass–Soil Bioconcentration of Munitions Compounds from Molecular Structure Tifany Lorena Torralba-Sanchez, Yuzhen Liang, and Dominic M. Di Toro Environ. Sci. Technol., Just Accepted Manuscript • DOI: 10.1021/acs.est.7b02572 • Publication Date (Web): 17 Aug 2017 Downloaded from http://pubs.acs.org on August 24, 2017

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Environmental Science & Technology is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 40

Environmental Science & Technology

1

ESTIMATING GRASS–SOIL

2

BIOCONCENTRATION OF MUNITIONS

3

COMPOUNDS FROM MOLECULAR

4

STRUCTURE

5

Tifany L. Torralba–Sanchez, Yuzhen Lianga, Dominic M. Di Toro*

6

Department of Civil & Environmental Engineering, University of Delaware, Newark,

7

Delaware, 19716

8

a

9

Technology, Guangzhou, Guangdong 510006, China

Current affiliation: School of Environment and Energy, South China University of

10 11

*Corresponding author: Dominic M. Di Toro. E–mail address: [email protected];

12

phone number: (302) 831-4092

13 14

Author contributions: This manuscript was written through contributions of all

15

authors. All authors have given approval to the final version of the manuscript.

16

1

ACS Paragon Plus Environment


17

Abstract

18

A partitioning–based model is presented to estimate the bioconcentration of five

19

munitions compounds and two munition–like compounds in grasses. The model uses

20

polyparameter linear free energy relationships (pp–LFERs) to estimate the partition

21

coefficients between soil organic carbon and interstitial water and between interstitial

22

water and the plant cuticle, a lipid–like plant component. Inputs for the pp–LFERs are

23

a set of numerical descriptors computed from molecular structure only that characterize

24

the molecular properties that determine the interaction with soil organic carbon,

25

interstitial water, and plant cuticle. The model is validated by predicting concentrations

26

measured in the whole plant during independent uptake experiments with a root mean

27

square error (log predicted plant concentration - log observed plant concentration) of

28

0.429. This highlights the dominant role of partitioning between the exposure medium

29

and the plant cuticle in the bioconcentration of these compounds. The pp–LFERs can

30

be used to assess the environmental risk of munitions compounds and munition–like

31

compounds using only their molecular structure as input.

32

Keywords

33

pp–LFER; Partitioning; Plant Uptake; Cuticle; Organic Carbon

2


Page 2 of 40

Page 3 of 40


34 35 36

TOC/Abstract art

3



37 38

Page 4 of 40

Introduction Munitions compounds (MCs) are widely used in commercial and military 1-3

39

activities, and are often released into the environment

40

may bioconcentrate MCs. This causes concerns regarding the potential for

41

environmental risk due to both direct toxicity and transference of these compounds to

42

higher trophic levels. Hence the need to develop prediction models able to estimate the

43

degree to which MCs are transferred from the ambient environment into plants, i.e., the

44

extent of bioconcentration to be expected.

45

. Organisms that are exposed

Studies proposing models to predict bioconcentration of organic compounds in 4-11

46

plants from various growth media have been proposed

47

account for transformation and degradation of the parent compound within the plant

48

(e.g., metabolism, photodegradation), volatilization from leaves, and plant

49

physiological processes such as growth and water transpiration 7, 11. However, in order

50

to be applied to a specific compound, these models require parameter estimates that

51

quantify each of these mechanisms for that compound. This limits their general

52

applicability.

. Some of these models

53

Models have also been formulated assuming that the uptake of nonionic organic

54

compounds through the plant roots results in a steady state partitioning between the

55

plant components and the soil

56

between the plant and the external phase(s) where the organic compound is present (e.g.,

57

soil), and it occurs along the plant transpiration stream

58

can be modeled assuming an equilibrium between the soil solids, soil solution

59

(interstitial water), and the plant.

10, 12

. The uptake is driven by concentration gradients

4, 5, 13

. In this way, the process

60

The concentration of an organic compound available for plant root uptake in

61

soils is that freely dissolved in the interstitial water (IW) 14, 15. This concentration is the

4


Page 5 of 40


62

result of soil solid–soil IW adsorption–desorption and is controlled by both the

63

compound chemical properties and soil properties such as the mass fraction of organic

64

carbon (𝑓𝑓𝑂𝑂𝑂𝑂 ) and clay size particles (𝑓𝑓𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 ) 16-20. It has been shown for a wide variety of

65

nonionic organic compounds that sorption to soil organic carbon (OC) is the dominant

67

mechanism for 𝑓𝑓𝑂𝑂𝑂𝑂 greater than approximately 0.1 to 0.2 % 21-26. The soil OC–aqueous

68

using a log–log correlation of KOC with the octanol–water partition coefficient, KOW.

69

This assumes that n–octanol is a good surrogate for soil OC 27, 28. While these single–

70

parameter quantitative structure-activity relationships (QSARs) have been shown to

71

work reasonably well for mostly nonpolar hydrophobic organic chemicals 29, they have

72

failed for more polar compounds, compounds that interact by hydrogen–bonding, and

73

for highly hydrophobic compounds

74

models of KOC that perform well for a wide range of compound classes has been

75

identified 33.

66

phase concentrations ratio is the partition coefficient, KOC, that is commonly estimated

30-32

. Consequently, a need for comprehensive

76

Equilibrium sorption of organic contaminants by plant tissue components, such

77

as carbohydrates and lipids, has also been measured and modeled 34-36. Similarly to the

78

KOC models, n–octanol is commonly used as a surrogate for plant lipids. While practical,

79

this approach has not been able to fully characterize the interactions between organic

80

compounds and plant tissues 12, 34.

81

An alternative approach is to use a polyparameter linear free energy relationship 37, 38

82

(pp–LFER) model

. Unlike single–parameter KOW–based predictions, pp–LFERs

83

predict partitioning by explicitly considering the contributions from different types of

84

chemical interactions (e.g., hydrogen bonding, van der Waals forces) between the solute

85

and the condensed phase (e.g., soil OC, plant lipids). Thus, pp–LFERs are able to more

5



86

fully characterize the solvation properties of the condensed phase and the strength of its

87

interactions with solutes relative to that of the aqueous phase.

88

In order to achieve these results the estimation of pp–LFERs parameters require

89

a large and chemically diverse training dataset of partition coefficients. In the case of

90

plants, for example, the pp–LFERs require sufficient data to quantify both the solvation

91

properties of a specific biomass component (e.g., lipids, carbohydrates, proteins) and

92

the molecular properties of the organic contaminant of interest 9, 36, 39.

93

The purpose of this work is to predict the bioconcentration of MCs and

94

compounds with similar chemical structural functionalities (Table S1 in Supporting

95

Information – SI), which are hereafter referred to as munition–like compounds (MLCs),

96

in plants based on the partitioning between soil solids, IW, and plant. This is achieved

97

using pp–LFERs for predicting the dissolved IW concentration in the soil using KOC,

98

and for predicting the sorption to plant tissue using the partitioning between IW and

99

plant cuticle, a lipid–like component. This procedure is validated by predicting

100

concentrations in plant biomass compiled from published uptake assays in an

101

independent dataset (95 observations). The pp–LFERs employed in this work use only

102

molecular structure to compute the required model parameters. Therefore, they can also

103

be used to evaluate the bioconcentration potential for proposed compounds early in the

104

development stage of new MCs.

105

Methodology

106

Basis for modeling approach

107

The model is designed to predict the bioconcentration of MCs for plants, more

108

specifically, for grasses, in situations where estimates of the parameters required in more

6


Page 6 of 40

Page 7 of 40


109

detailed uptake models, e.g., plant water fluxes, growth rates, and contaminant

110

transformation/metabolism rates are not available. The model estimates the upper–

111

bound of the concentration in the plant when exposed to the bioavailable MC in the soil

112

IW using equilibrium partitioning (EqP). The EqP concept has been widely applied to

113

assess the bioconcentration of organic compounds in organisms including fish, worms,

114

and plants 10, 12, 40-46.

115

In the case of grasses, the MCs available in the soil IW are transported in solution

116

through the plant roots and along the plant transpiration stream. Eventually, as shown

117

by Chiou et al. 10, nonionic organic compounds that are relatively water soluble, as is

118

the case for MCs, reach equilibrium between the external and internal aqueous phases

119

(see SI for further details). In the same vein, we have shown

120

bioconcentration factors (BCFs) for MCs in barley reach steady state under constant

121

exposure concentrations in sand culture experiments. The observed BCFs are

122 123

comparable to plant–water partition coefficients (𝐾𝐾𝑃𝑃𝑃𝑃 ) measured by equilibrating

124

experiments 47, 48. These results support the use of the EqP model to predict the uptake

125

of MCs by grasses from IW.

47, 48

that the

sectioned barley biomass and the exposure solution used in the sand culture uptake

126

Figure 1 illustrates MC partitioning between soil and plant components and

127

water. When the root achieves equilibrium with the soil IW and simultaneously with the

128

plant internal water, both the internal and external aqueous phases are at equilibrium. In

129

this case, the root–water partitioning can be bypassed and the MC concentration in the

130

plant internal water is assumed to be equal to that in the soil IW. Therefore, partitioning

131

between soil OC and plant cuticle determines the MCs bioconcentration in grasses.

7



132 133 134 135 136 137 138

Figure 1

Schematic diagram of the basis for the modeling approach: Equilibrium partitioning. Left–right arrows indicate partitioning of MC (or MLC) 𝑖𝑖 between two phases. 𝐾𝐾𝑖𝑖𝑂𝑂𝑂𝑂 = soil organic carbon–water partition coefficient of MC 𝑖𝑖 (Lwater kgOC-1), 𝐾𝐾𝑖𝑖𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 = root–water partition coefficient of MC 𝑖𝑖 (Lwater kgroot-1), and 𝐾𝐾𝑖𝑖𝐶𝐶𝐶𝐶𝐶𝐶 = plant cuticle–water partition coefficient of MC 𝑖𝑖 (Lwater kgcuticle-1).

Organic carbon is usually assumed to determine the mobility of MCs in soils 49-

139 140

51

141

fOC > 1 – 2% 25, 52, 53. This is larger than the usual boundary for hydrophobic compounds

142

fOC = 0.1 – 0.2 % as discussed above 21-26.

143

. It has been shown that OC is the dominant phase for partition of MCs in soils with

Plant cuticle is an extracellular hydrophobic membrane composed of 54, 55

144

interconnected long–chain fatty acids and alkyls

145

fruits, leaves, and stems

146

hydrophobic membrane appears to be the principle site where sorption of MCs occurs.

147

Contributions from plant carbohydrates were also examined before being excluded (see

148

“Results and Discussion”).

56

that coats plant organs such as

, and protects and waterproofs the plant surface. This

8


Page 8 of 40

Page 9 of 40


149

Uptake through plant leaves from the atmosphere is not included in the model.

150

The model might also be successfully applied to other similar nonionic organic

151

compounds, however, the validation presented in this study is limited to MCs and MLCs

152

only.

153

Polyparameter Linear Free Energy Relationship (pp–LFER) Models

154

The pp–LFER models for partitioning between soil OC and water, and between

155

plant cuticle and water used in this work are based on the Abraham polyparameter model

156 157

37

log 𝐾𝐾𝑆𝑆𝑆𝑆 = 𝑐𝑐 + 𝑒𝑒𝑒𝑒 + 𝑠𝑠𝑠𝑠 + 𝑎𝑎𝑎𝑎 + 𝑏𝑏𝑏𝑏 + 𝑣𝑣𝑣𝑣

(1)

158

where KSW is the partition coefficient between a solvation phase (e.g., OC, plant cuticle)

159

and water, and the right hand side of Eq. (1), are the parameters that account for the free

160

energy contributions from different types of molecular interactions. The uppercase

161

letters in Eq. (1) are solute (e.g., MC, MLC) descriptors and the lowercase letters

162

quantify the complementary effect of the solvation phase on the corresponding

163

interaction. The 𝑒𝑒𝑒𝑒 term represents the dispersion interactions that are predominant

164

between nonpolar (no permanent multipole moments) molecules and that are not

165 166

captured by the 𝑣𝑣𝑣𝑣 term. The 𝑠𝑠𝑠𝑠 term encompasses dipole–dipole or dipole–induced

167

donation and acceptance of hydrogen bonds, which are bonds between certain types of

dipole plus some polarizability interactions. The 𝑎𝑎𝑎𝑎 and 𝑏𝑏𝑏𝑏 terms account for the

169

hydrogen atoms and highly electronegative atoms in polar molecules. 𝑎𝑎𝑎𝑎 is solvent

170

accounts for both the energy required for cavity formation and part of the London

171

dispersion interactions, and 𝑐𝑐 is a regression constant 37, 57.

168

acceptor (𝑎𝑎)–solute donor (𝐴𝐴) and 𝑏𝑏𝑏𝑏 is solvent donor (𝑏𝑏)–solute acceptor (𝐵𝐵). 𝑣𝑣𝑣𝑣

9



172

Page 10 of 40

Plant Cuticle–Water pp–LFER

173

Platts and Abraham 9 fitted a pp–LFER model to a dataset of plant cuticle–water

174

partition coefficients, KCut, for tomato (Lycopersicum esculentum Mill) cuticle, for 62

175 176

volatile organic compounds (-0.77 < log KOW < 6.25) yielding

177 178 179 180 181 182

log 𝐾𝐾𝐶𝐶𝐶𝐶𝐶𝐶 = −0.415 + 0.596𝐸𝐸 − 0.413𝑆𝑆 − 0.508𝐴𝐴 − 4.096𝐵𝐵 + 3.908𝑉𝑉

(2)

𝑅𝑅 2 = 0.981; SD = 0.236; F = 566. KCut = plant cuticle–water partition coefficient (Lwater

kgcuticle-1), 𝑅𝑅 2 = coefficient of determination, SD = regression standard deviation, and F

= Fischer’s F statistic. The magnitude of the terms in Eq. (2) indicate that cuticle is more competitive for solutes than water through 𝜋𝜋- and n- electron pairs dispersion interactions (𝑒𝑒 = 0.596 > 0), and via cavity formation in the cuticle that requires much

less free energy than in water (𝑣𝑣 = 3.908 > 0). This model also indicates that the

184

cuticle is less polar/polarizable (𝑠𝑠 = −0.413 < 0) and accepts hydrogen bonds (𝑎𝑎 =

185

−0.508 < 0) and donates hydrogen bonds (𝑏𝑏 = −4.096 < 0) much less readily than water 9.

186

In order to broaden the chemical and plant species diversity and to incorporate

187

MCs and MLCs functionalities, such as aromatic compounds with multiple C-NO2

188

groups and non-aromatic cyclic structures with N-NO2 bonds (SI Table S1), into the

189

training set used by Platts and Abraham 9, additional KCut data (SI Table S5) were

190

collected and included in deriving the plant cuticle–water pp–LFER. Sources for these

191

data are described below.

183

192

Since new chemicals and new plant cuticle data were added to the dataset, it was

193

necessary that the solute parameters (uppercase letters in Eq. (1)) be obtained from

194

consistent sources. Therefore, three different sources for solute descriptors were

195

compared (SI Table S7 to Table S10): (i) Absolv estimated Abraham Parameters

10


Page 11 of 40


58

196

(Absolv–AP) from the Absolv software module of ACD/PhysChem Suite

197

Experimentally derived Abraham Parameters (Exp–AP) from Liang et al.

198

UFZ–LSER database 60, and (iii) Quantum Chemically estimated Abraham Parameters

199

(QCAP) from Liang 61, 62. These three sources were selected because Absolv–AP have

200

been widely used and are recommended by Platts and Abraham

201

structure, and Exp–AP and QCAP have been shown to successfully predict KSW for a

202

wide variety of organic compounds including MCs and MLCs

203

plant cuticle–water pp–LFERs were obtained using a multiple linear regression (MLR)

204

with Absolv–AP, Exp–AP, and QCAP as the solute parameters and the full KCut dataset

205

(SI Table S5) as the independent variable. The MLRs were performed using the lm

206

function of the R software for statistical computing 63.

207 208 209

210 211

9

59

, (ii)

and the

for any organic

59, 61, 62

. Three updated

Soil Organic Carbon–Water pp–LFER KOC ’s were predicted using the pp–LFER developed by Kipka and Di Toro 64 log 𝐾𝐾𝑂𝑂𝐶𝐶 = 0.670 (±0.088) + 1.075 (±0.061)𝐸𝐸 − 0.277 (±0.083)𝑆𝑆 − 0.363 (±0.100)𝐴𝐴 − 1.697 (±0.085)𝐵𝐵 + 1.468 (±0.077)𝑉𝑉

(3)

for 440 compounds; RMSE = 0.48. KOC = soil organic carbon–water partition

coefficient (Lwater kgOC-1), values in parenthesis = ± the standard error, and RMSE =

212

root mean square error of prediction. This model was built using a large and chemically

213

diverse dataset of nonionic organic compounds with Absolv–AP for the solute

214

parameters. A comparison of the performance of this model relative to other available

215

pp–LFERs to predict independent log KOC literature observations for MCs is presented

216

in the SI. The model by Kipka and Di Toro 64 yielded similar or smaller RMSEs than

217

those by the other pp-LFERs using either experimental or estimated solute descriptors

11



218

(see SI). In general, experimental descriptors are more accurate than estimated ones to

219

calibrate pp–LFERs. In this manuscript, however, the aim is to illustrate that if

220

descriptors are computed with versatile and robust methods (e.g., quantum chemical

221

computations), they can be used for the calibration of pp–LFERs obtaining similar

222

estimation errors than those with experimental descriptors, as shown below in Results

223

and Discussion. Moreover, the use of estimated descriptors expands the applicability of

224

pp–LFERs to large collections of compounds or proposed MCs early in the development

225

stage, for which no data is available. These are scenarios where the experimental

226

determination of descriptors is limited by time and feasibility constrains.

227

The solvent parameters in Eq. (3) reveal that the soil OC phase has similar

228

solvation capabilities to those shown by plant cuticle in Eq. (2). The solute descriptors

229

used to apply Eq. (3) for the prediction of concentrations in the plants were the

230

appropriate QCAP reported by Liang 61, 62 (SI Table S14).

231

Experimental Data

232

Two types of experimental data were collected from the literature: (i) reported

233

KCut values, and (ii) measurements of concentrations in plant biomass made during

234

uptake assays where plants were exposed to MCs, or MLCs, in the growth medium. The

235

former dataset was added to the training set used by Platts and Abraham 9, while the

236

latter served to validate the partitioning–based bioconcentration model. The MCs,

237 238

MLCs, and plant species included in the datasets are listed in Table 1.

12


Page 12 of 40

Page 13 of 40


239 240

Table 1

Munitions Compounds (MCs), Munitions–Like Compounds (MLCs), and plant species included in the datasets. Abbreviation

Name MCs and MLCsa

TNT

2,4,6-trinitrotoluene

2,4-DNT

2,4-dinitrotoluene

2,4-DNAN

2,4-dinitroanisole

RDX

hexahydro-1,3,5-trinitro-1,3,5-triazine

HMX

octahydro-1,3,5,7-tetranitro-1,3,5,7-tetrazocine

4-NAN

4-nitroanisole

2-M-5-NPYNE

2-methoxy-5-nitropyridine

2,5-DM-4-NANE

2,5-dimethoxy-4-nitroaniline Plant Species: KCut datasetb

L. esculentum

Lycopersicum esculentum Mill

F. elastica

Ficus elastica Roxb. var. decora

C. annuum

Capiscum annuum L.

C. aurantium

Citrus aurantium L.

P. laurocerasus

Prunus laurocerasus L.

G. biloba

Ginkgo biloba L.

J. regia

Juglans regia L.

S. lycopersicum

Solanum lycopersicum

M. domestica

Malus domestica

S. tuberosum

Solanum tuberosum

V. heyneana

Vitis heyneana Roem. et Schult

L. multiflorum

Lolium multiflorum Lam.

L. arundinaceium

Lolium arundinaceium

F. rubra

Festuca rubra L.

S. oleracea

Spinacia oleracea

H. vulgarec

Hordeum vulgare L. Plant Species: Uptake datasetd

13



C. esculentus

Cyperus esculentus

L. perenne

Lolium perenne

M. sativa

Medicago sativa

Z. mays

Zea mays

G. max

Glycine max

S. Sudanese

Sorghum Sudanese

T. aestivum

Triticum aestivum

P. vulgaris

Phaseolus vulgaris

Page 14 of 40

B. rapa Brassica rapa Further details including chemical properties are listed in SI Table S1 b As reported in corresponding source(s). Data in SI Table S5 c Also in uptake dataset d As reported in corresponding source(s). Data in SI Table S13 a

241 242

Plant Cuticle–Water Partition Coefficients (KCut) Data

243

Partition coefficients between plant cuticle and water are commonly determined

244

by individually equilibrating either isolated cuticular membranes (CM) or cuticle

245

matrices (MX, the dewaxed CM)

246

compound (𝑖𝑖) and calculating the cuticle partition coefficient

56, 65, 66

247

248

𝐾𝐾𝑖𝑖𝐶𝐶𝐶𝐶𝐶𝐶 =

with an aqueous solution of an organic

𝐶𝐶𝑖𝑖𝐶𝐶𝐶𝐶 𝑜𝑜𝑜𝑜 𝑀𝑀𝑀𝑀 𝐶𝐶𝑖𝑖𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊

(4)

where 𝐶𝐶𝑖𝑖𝐶𝐶𝐶𝐶 𝑜𝑜𝑜𝑜 𝑀𝑀𝑀𝑀 = concentration of compound 𝑖𝑖 in CM or MX (mg kgdwt-1; dwt: dry

250

weight), and 𝐶𝐶𝑖𝑖𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊 = concentration of compound 𝑖𝑖 in the water phase (mg L-1). The

251

of epicuticular wax in the cuticle component proved to have no significant effect on the

252

resulting KCut (SI Table S6) 67.

249

KCut dataset (SI Table S5) includes values obtained with CM or MX since the presence

14


Page 15 of 40


253

In addition to KCut from isolated cuticle components, values from experiments

254

performed with whole plant biomass were also included after normalization of the

255 256

plant–water partition coefficient by the mass fraction of cuticle

257

𝐾𝐾𝑖𝑖𝐶𝐶𝐶𝐶𝐶𝐶 =

𝐾𝐾𝑖𝑖𝑃𝑃𝑃𝑃 𝑓𝑓𝐶𝐶𝐶𝐶𝐶𝐶

(5)

where 𝐾𝐾𝑖𝑖𝑃𝑃𝑃𝑃 = plant–water partition coefficient of compound 𝑖𝑖 (Lwater kgdwt plant-1) and

259

𝑓𝑓𝐶𝐶𝐶𝐶𝐶𝐶 = dry weight fraction of cuticle in the plant (kgcuticle kgdwt plant-1). A total of 143

260

pp–LFER training set (SI Table S5). The KCut for compounds containing dissociable

261

groups were either from experiments performed in solutions buffered below the pKa of

262

the compound or had been corrected for the degree of dissociation using acid

263

dissociation constants as detailed in Kerler and Schonherr 68. The corrected KCut were 9

264

observations for 3 compounds in the training set.

258

265

experimental KCut for undissociated organic compounds were compiled for the cuticle

Data from Plant Uptake Assays

266

A dataset was compiled from measurements reported in published uptake assays

267

with grasses and other plants belonging to closely related families exposed to MCs, or

268

MLCs, in the growth medium (SI Table S13). In addition to experiments performed in

269

spiked or contaminated field soil (hereafter referred to as "soil"), assays using either

270

coarse quartz sand (99%, 0.85–1.27 mm effective diameter particles, hereafter referred

271

to as "sand") or aqueous solutions as the growth medium were also included in the

272

dataset. The inclusion of these datasets had two purposes: (i) compare the predictions

273

relative to those in soil exposures, and (ii) test only the plant cuticle–water pp–LFER

274

without the need for a soil OC–water pp–LFER. The full dataset includes the

15



Page 16 of 40

275

concentration in the plant, concentration in the growth medium, exposure time, and dry

276

weight fraction of OC in the soil when applicable (SI Table S13).

277

Concentrations in the plant were for the whole plant or only for the aboveground

278

plant parts when available; measurements in fruits or flowers (e.g., corn kernels or

279

tassels) were not included. Concentrations below reported analytical quantification

280

limits or without clarification on whether they were expressed on a dry or fresh weight

281

basis were excluded. Data from studies not reporting either the soil OC or organic matter

282

content were also excluded (excluded data in SI Table S18).

283 284

Estimation of Concentrations in Plants Observed in Independent Uptake Assays

285

Concentrations in grasses and closely related plants reported in published uptake

286

assays were predicted using models of the appropriate partition coefficients. The

287

concentration of MCs and MLCs available for plant root uptake in soil growth medium

288 289

was estimated using

290 291 292 293 294

𝐶𝐶𝑖𝑖𝐼𝐼𝐼𝐼 =

𝐶𝐶𝑖𝑖𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝐶𝐶𝑖𝑖 = 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝐾𝐾𝑖𝑖𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝐾𝐾𝑖𝑖𝑂𝑂𝑂𝑂 𝑓𝑓𝑂𝑂𝑂𝑂

(6)

where 𝐶𝐶𝑖𝑖𝐼𝐼𝐼𝐼 = concentration of MC (or MLC) 𝑖𝑖 in the growth medium interstitial water

(mg L-1), 𝐶𝐶𝑖𝑖𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 = concentration of MC 𝑖𝑖 in the soil solids (mg kgdwt-1), 𝐾𝐾𝑖𝑖𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 =

soil–water partition coefficient of MC 𝑖𝑖 (Lwater kgdwt soil-1), and 𝑓𝑓𝑂𝑂𝑂𝑂 = dry weight fraction of organic carbon in the soil (kgOC kgdwt soil-1).

Values for 𝐶𝐶𝑖𝑖𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 were those reported by the sources as the concentrations

295

at the beginning of the exposure or a steady state concentration when available. Values

296

for 𝑓𝑓𝑂𝑂𝑂𝑂 were also obtained from the sources. A factor of 0.50 was used to convert soil

297

organic matter content (𝑓𝑓𝑂𝑂𝑂𝑂 ) to 𝑓𝑓𝑂𝑂𝑂𝑂 (i.e., 𝑓𝑓𝑂𝑂𝑂𝑂 = 0.5 𝑓𝑓𝑂𝑂𝑂𝑂 ) when needed 69. 𝐾𝐾𝑖𝑖𝑂𝑂𝑂𝑂 ’s were 16


Page 17 of 40


298

estimated using the pp–LFER in Eq. (3) described previously and the appropriate QCAP

299

from Liang 61, 62 (SI Table S14).

300 301

302

The concentration of MCs and MLCs in plant biomass was estimated using 𝐶𝐶𝑖𝑖𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 = 𝐾𝐾𝑖𝑖𝑃𝑃𝑃𝑃 𝐶𝐶𝑖𝑖𝐼𝐼𝐼𝐼 = 𝐾𝐾𝑖𝑖𝐶𝐶𝐶𝐶𝐶𝐶 𝑓𝑓𝐶𝐶𝐶𝐶𝐶𝐶 𝐶𝐶𝑖𝑖𝐼𝐼𝐼𝐼

(7)

where 𝐶𝐶𝑖𝑖𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 = concentration of MC 𝑖𝑖 in the plant biomass (mg kgdwt-1). 𝐶𝐶𝑖𝑖𝐼𝐼𝐼𝐼 directly

303

observed in experiments performed in either sand or aqueous solutions, or predicted

304 305

using Eq. (6) for soil experiments. Due to the scarcity of 𝑓𝑓𝐶𝐶𝐶𝐶𝐶𝐶 values for the species in

306

abundant plant families in the dataset: 0.18 and 0.21 kg kgdwt-1 for Poaceae and

the uptake assays, a single 𝑓𝑓𝐶𝐶𝐶𝐶𝐶𝐶 from the literature was used for each of the most

308

Fabaceae, respectively (SI Table S15). An average of these two 𝑓𝑓𝐶𝐶𝐶𝐶𝐶𝐶 values (0.20 kg

309

were estimated using the pp–LFER in Eq. (8) described below and the QCAP from

310

Liang 61, 62 (SI Table S14).

311

Results and Discussion

307

312

kgdwt-1) was used for the species belonging to the other closely related families. 𝐾𝐾𝑖𝑖𝐶𝐶𝐶𝐶𝐶𝐶 ’s

Plant Cuticle–Water pp–LFER

313

The predicted KCut using Eq. (2) versus observed KCut are shown in Figure 2 (SI

314

Table S7 to Table S9). The accuracy of the predictions varied with the source of the

315

Abraham solute descriptors. The predicted KCut for the nitramine MCs, RDX and HMX,

316

using Absolv–AP were seven orders of magnitude smaller than the observed KCut

317

(Figure 2A). A reason for these large underpredictions might be the absence of the

318

nitramine (N-NO2) functional group in the Absolv fragment descriptors set 70. Missing

319

fragments has been identified as a major drawback for the group contribution approach

320

in the estimation of solute descriptors 61, 62, 71. N-NO2 is a highly electronegative group

17



321

and it increases the potential reactive sites of these MCs

322

descriptor needs to be included.

18


Page 18 of 40

72

. Therefore the fragment

Page 19 of 40


323 324 325 326 327 328 329 330 331 332 333 334

Figure 2

pp–LFER–predicted KCut versus observed KCut for organic compounds (SI Table S5 and Table S7 to Table S9). Predictions made using the pp–LFER model from Platts and Abraham 9, Eq. (2), for the full KCut dataset collected from the literature. Solute descriptors, uppercase letters for Eq. (2), are: (A) Absolv–AP 58; (B) Exp–AP from either Liang et al. 59 or the UFZ– LSER database 60, and Absolv–AP when Exp–AP were not available; and (C) Adjusted QCAP 61, 62. RMSE: root mean square error of prediction (log predicted - log observed) for all compounds included in the full KCut dataset (SI Table S5). The RMSEs for only MCs and MLCs were = (A) 3.154; (B) 0.544; and (C) 0.507. The solid line indicates the best agreement (unity), dashed lines are spaced at 1 log unit from unity.

19



335

The predictions for RDX and HMX improved by more than six orders of

336

magnitude when using Exp–AP (Figure 2B) and the overall RMSE for all compounds

337

decreased from RMSE = 1.167 (Figure 2A) to RMSE = 0.382 (Figure 2B). The

338

experimental derivation of molecular properties generally results in high–quality solute

339

descriptors. However, its application to large collections of compounds or proposed

340

MCs early in the development stage, for which no data is available, is limited by time

341

and feasibility constrains. Consequently, an alternative method for the determination of

342

solute descriptors that depends only on molecular structure was also tested.

343

QCAP’s are derived from quantum chemically computed solvent–water

344

partition coefficients and molecular polarizability 61, 62. The solute descriptors: 𝑆𝑆, 𝐴𝐴, and

346

𝐵𝐵 are simultaneously estimated using 65 quantum chemically computed solvent–water

347

partition coefficients obtained from a continuum solvation model. 𝐸𝐸 and 𝑉𝑉 are

independently obtained from quantum mechanical estimated molecular polarizability

348

and molecular volume.

349

solvent–water partitioning data. Therefore, they are incompatible with the solvent

350

parameters in existing pp–LFERs. This inconsistency is corrected by adjusting QCAP

351

using large datasets of Exp–AP (details in Liang 61, 62). These adjusted QCAP were used

352

for all the compounds to predict the observed KCut in Figure 2C. The RMSE = 0.494,

353

which is competitive relative to that using Exp–AP.

345

354

61, 62, 73

. QCAP are not generated using only experimental

Plant Cuticle–Water pp–LFER including MCs and MLCs

355

The dataset employed by Platts and Abraham 9 to develop the KCut pp–LFER in

356

Eq. (2), and used to make the predictions in Figure 2, did not contain MCs or MLCs.

357

Therefore, pp–LFERs were developed using the expanded KCut dataset compiled from

358

the literature, which also includes a more heterogeneous group of plant species. The

20


Page 20 of 40

Page 21 of 40


359

results are shown in Figure 3 and the pp–LFERs are listed in SI Table S11. The same

360

solute descriptors were used as in Figure 2A and B. However, for Figure 3C, QCAP

361

were used instead of the adjusted QCAP since QCAP have been found to be superior

362

when a new pp–LFER is being created 61, 62.

21



363 364 365 366 367 368 369 370 371 372 373 374

Figure 3

pp–LFER–fitted KCut versus observed KCut for organic compounds (SI Table S5, Table S7, Table S8, and Table S10). Calculations made using the general pp–LFER, Eq. (1), fitted to the full KCut dataset collected from the literature. Solute descriptors, uppercase letters for Eq. (1), are: (A) Absolv–AP 58; (B) Exp–AP from either Liang et al. 59 or the UFZ–LSER database 60, and Absolv–AP when Exp–AP were not available; and (C) QCAP 61, 62. RMSE: root mean square error of prediction (log fitted - log observed) for all compounds included in the full KCut dataset (SI Table S5). The RMSEs for only MCs and MLCs were = (A) 1.261; (B) 0.478; and (C) 0.421. The solid line indicates the best agreement (unity), dashed lines are spaced at 1 log unit from unity.

22


Page 22 of 40

Page 23 of 40


375

The KCut pp–LFER obtained using Absolv–AP improved the overall RMSE

376

(1.167 to 0.778; Figure 2A and Figure 3A, respectively), but misfits for many of the

377

compounds resulted (Figure 3A). The KCut pp–LFER obtained using Exp–AP slightly

378

changed the already good overall fit (RMSE = 0.382 to 0.332; Figure 2B and Figure 3B,

379

respectively).

380

A good fit was also obtained when QCAP were used for all the compounds. The

381

resulting KCut pp–LFER satisfactorily captured the overall variations in the dataset

382

(Figure 3C) yielding a RMSE = 0.386. The resulting plant cuticle–water pp–LFER built

383 384

using QCAP is

385

log 𝐾𝐾𝐶𝐶𝐶𝐶𝐶𝐶 = −0.593 (±0.098) + 0.433 (±0.086)𝐸𝐸 + 0.900 (±0.164)𝑆𝑆 − 0.587 (±0.100)𝐴𝐴 − 5.409 (±0.253)𝐵𝐵 + 3.442 (±0.203)𝑉𝑉

(8)

2 using 77 compounds; N = 143 data points; RMSE = 0.386; 𝑅𝑅𝐴𝐴𝐴𝐴𝐴𝐴. = 0.939; SE = 0.394;

387

2 𝐹𝐹 = 437.6. Values in parenthesis = ± the standard error, 𝑅𝑅𝐴𝐴𝐴𝐴𝐴𝐴. = adjusted 𝑅𝑅 2 and SE =

388

The system parameters in Eq. (8), which are the lowercase variables in Eq. (1),

389

quantify the extent to which compounds partition to cuticle relative to water. The signs

390

of the system parameters in Eq. (8) are consistent with those obtained by Platts and

391

Abraham

392

term, 𝑠𝑠, which is negative 𝑠𝑠 = −0.413 in Eq. (2) and positive 𝑠𝑠 = 0.900 in Eq. (8). The

386

regression residual standard error.

9

(Eq. (2)) for all molecular interactions except the dipolarity/polarizability

394

𝑠𝑠 > 0 in Eq. (8) suggests that cuticle is a stronger solvation phase than water when

395

water is usually stronger than many environmental and biological phases through

396

polarizability interactions

393

interacting with polar/polarizable solutes, which is an unexpected result. This is because

74

, i.e., 𝑠𝑠 < 0. However, the ability of a pp–LFER to

23



Page 24 of 40

397

accurately characterize the properties of a solvation phase such as cuticle, largely relies

398

on the quality and diversity of the solute descriptors used in its calibration dataset 74.

399

While the solute descriptors used in Platts and Abraham 9 (Absolv–AP and/or

400

Exp–AP) produced an excellent agreement to their observed values, the training set used

401

in Eq. (8) covers a considerably wider descriptor space in 𝐸𝐸 and 𝑆𝑆 values than that by

402

Eq. (2) (range 𝐸𝐸: 4.29 vs. 3.26; and range 𝑆𝑆: 2.33 vs. 1.76; respectively. SI Table S12).

404

The high end of the 𝑆𝑆 range for Eq. (8) comes largely from the MCs and MLCs. These

405

which can behave through both specific and nonspecific molecular interactions 75. The

406 407

𝑠𝑠 = 0.900 (Eq. (8)) suggests that the cuticle system has a higher ability than water to

408

the partition coefficient to increase. This might be due in part to the polar characteristics

409

of the fatty acids in the cutin and cutan components of this lipid–like plant component

410

54, 55, 76

411

target lipid model (TLM) for the prediction of acute toxicity in aquatic organisms

412

which uses lipid as the site of action for the sorption of contaminants.

403

compounds revealed the strength of the cuticle interaction with more polar solutes,

attract polar solutes through dipolarity/polarizability–type interactions and hence cause

. It is interesting to note that a positive 𝑠𝑠 is also found in the pp–LFER of the 77

,

413

Besides a larger array of plant species in the calibration set, the good quality and

414

wider diversity of solute descriptors supported the choice of Eq. (8) over Eq. (2) for the

415

prediction of concentrations in plants from published uptake assays.

416 417

Estimation of Concentrations in Plants Observed in Independent Uptake Assays

418

Predictions were made based on the partitioning between soil OC, IW, and plant

419

cuticle, as described previously. The KOC model, Eq. (3), was used to estimate the

420

concentration of the MC, or MLC, in the soil IW (i.e., exposure concentration), and the

24


Page 25 of 40


421

KCut model, Eq. (8), was used to predict the corresponding concentration in the plant

422

biomass. The final equation used to predict the plant concentrations is Eq. (9) in Table

423

2, which contains the equations that comprise the bioconcentration model. An example

424

is presented in SI Table S16. Predicted versus observed concentrations in plants for five

425

MCs and two MLCs are shown in Figure 4 (SI Table S13 and Table S17).

426 427

Table 2

Equations for the prediction of concentrations in plants exposed to MCs, or MLCs, in soil and sand or water culture.

A partitioning–based plant bioconcentration modela: 𝐶𝐶𝑖𝑖𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 =

Var.

𝐾𝐾𝑖𝑖𝐶𝐶𝐶𝐶𝐶𝐶 𝑓𝑓𝐶𝐶𝐶𝐶𝐶𝐶 𝐶𝐶𝑖𝑖𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝐾𝐾𝑖𝑖𝑂𝑂𝑂𝑂 𝑓𝑓𝑂𝑂𝑂𝑂 Equation

𝐾𝐾𝑖𝑖𝑂𝑂𝑂𝑂 : log 𝐾𝐾𝑂𝑂𝑂𝑂 = 0.670 + 1.075𝐸𝐸 − 0.277𝑆𝑆 − 0.363𝐴𝐴 − 1.697𝐵𝐵 + 1.468𝑉𝑉 𝐶𝐶𝑖𝑖𝐼𝐼𝐼𝐼 : 𝐶𝐶𝑖𝑖𝐼𝐼𝐼𝐼 =

𝐶𝐶𝑖𝑖𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝐾𝐾𝑖𝑖𝑂𝑂𝑂𝑂 𝑓𝑓𝑂𝑂𝑂𝑂

(9) # (3) (6)

𝐾𝐾𝑖𝑖𝐶𝐶𝐶𝐶𝐶𝐶 : log 𝐾𝐾𝐶𝐶𝐶𝐶𝐶𝐶 = −0.593 + 0.433𝐸𝐸 + 0.900𝑆𝑆 − 0.587𝐴𝐴 − 5.409𝐵𝐵 + 3.442𝑉𝑉 (8) 𝐶𝐶𝑖𝑖𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 : 𝐶𝐶𝑖𝑖𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 = 𝐾𝐾𝑖𝑖𝐶𝐶𝐶𝐶𝐶𝐶 𝑓𝑓𝐶𝐶𝐶𝐶𝐶𝐶 𝐶𝐶𝑖𝑖𝐼𝐼𝐼𝐼

(7)

Var.: Variables; 𝑖𝑖: MC, or MLC, of interest; 𝐶𝐶𝑖𝑖𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 : concentration of compound 𝑖𝑖 in the plant biomass (mg kgdwt-1); 𝐾𝐾𝑖𝑖𝐶𝐶𝐶𝐶𝐶𝐶 : plant cuticle–water partition coefficient of 𝑖𝑖 (Lwater kgcuticle-1); 𝑓𝑓𝐶𝐶𝐶𝐶𝐶𝐶 : dry weight fraction of cuticle in the plant (kgcuticle kgdwt plant-1); 𝐶𝐶𝑖𝑖𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑠𝑠 : concentration of 𝑖𝑖 in the soil solids (mg kgdwt-1); 𝐾𝐾𝑖𝑖𝑂𝑂𝑂𝑂 : soil organic carbon– water partition coefficient of 𝑖𝑖 (Lwater kgOC-1); 𝑓𝑓𝑂𝑂𝑂𝑂 : dry weight fraction of organic carbon in the soil (kgOC kgdwt soil-1); 𝐸𝐸, 𝑆𝑆, 𝐴𝐴, 𝐵𝐵, and 𝑉𝑉: solute descriptors for 𝑖𝑖 (SI Table S14); 𝐶𝐶𝑖𝑖𝐼𝐼𝐼𝐼 : concentration of 𝑖𝑖 in the growth medium interstitial water (IW) (mg L-1) a

428 429

The three panels in Figure 4 contain the same predicted and observed

concentrations in the plant but with different coding by (A) compound, (B) growth

25



430

medium, and (C) plant species. These are the main factors that affect the performance

431

of the model. Measured concentrations in plants range about four orders of magnitude

432

for seven different compounds including nitroaromatic, nitramines, and nitropyridines.

433

They were predicted within an order of magnitude (Figure 4A). No bias in the

434

predictions was observed for either MCs or MLCs.

435

26


Page 26 of 40

Page 27 of 40


436 437 438 439 440 441 442 443 444 445 446 447 448

Figure 4

Predicted concentrations in the plant versus observed values from published uptake studies (Table 2, and SI Table S13 and Table S17). Color coding assigned based on: (A) MCs and MLCs (Table 1), (B) growth medium, and (C) plant species. Unfilled symbols indicate that the predicted concentration in the interstitial water (using Eq. (6)) was replaced by the corresponding MC aqueous solubility as the predicted concentration in the interstitial water exceeded solubility limits. The border color for the unfilled symbols corresponds to the color identification in each panel legend. Root mean square error of prediction (log predicted - log observed), excluding predictions bounded at aqueous solubility, RMSE = 0.429. The solid line indicates the best agreement (unity), dashed lines are spaced at 1 log unit from unity.

27



449

The data included plant uptake assays carried out in three different growth

450

media: soil, sand, and water (Figure 4B). The accuracy of the predictions increased in

451

the order of soil < sand < water as shown by the logarithmic residuals ranging (minimum

452

to maximum) from -0.897 to 1.249, -0.459 to 0.837, and -0.215 to 0.612, respectively

453

(SI Table S17). This can be due to both the complexity of the experimental procedure

454

increasing in the order of water < sand < soil and the inclusion of a KOC model to make

455

the estimations from soil, which contributes to a larger prediction uncertainty relative

456

to that for uptake assays in water or in sand. A comparison of the performance of the

457

bioconcentration model (Eq. (9)) using log KOC pp-LFERs other than the one by Kipka

458

and Di Toro 64 (Eq. (3)) is presented in the SI. The RMSEs yielded using Eq. (3) were

459

similar or smaller than those using the other pp-LFERs with either experimental or

460

estimated solute descriptors (see SI for details).

461

Predictions made using only the KCut pp–LFER (Eq. (8)), i.e., those for

462

experiments performed with sand and water as the growth medium, produced RMSEs

463

of 0.527 and 0.214, respectively. This indicates that the KCut pp–LFER model alone is

464

capable of estimating within a reasonable uncertainty the bioconcentration of

465

compounds that are available for plant root uptake using only the measured

466 467

concentration in the water (IW in the case of sand), 𝑓𝑓𝐶𝐶𝐶𝐶𝐶𝐶 , and QCAP.

468

examined before being excluded from the model. A pp–LFER was built using literature

469

data for partition to cellulose and starch from water. Carbohydrate–water partition

470

coefficients, KCh, for MCs and MLCs were small (-1.016 < (log KCh) < -0.470) and

471

accounted for maximum 3.3% of the KPW even assuming carbohydrates were the

472

predominant component of the plant. The overall estimation error of the model changed

Contribution from plant carbohydrates to the bioconcentration of MCs was also

28


Page 28 of 40

Page 29 of 40


473

only by 0.5% when carbohydrates were included as a sorption compartment (see SI for

474

details).

475

It is difficult to maintain a nearly constant exposure concentration in soils, unlike

476

in water or sand due to both the degradation of the parent compound and the

477

inhomogeneous distribution of the compound in the soil. Unfortunately, for the majority

478

of the data in Figure 4 the concentrations in the growth medium were not monitored or

479

prevented from significantly fluctuating. Only the initial concentrations in the growth

480

medium were available (SI Table S13). The loss of the parent compound led to the

481

overestimation of some of the resulting plant concentrations (Figure 4A and B),

482

especially for compounds like TNT which have been shown to be readily transformed

483

20, 78

.

484

In order to circumvent the problem of significant compound degradation, large

485

amounts of chemicals are usually applied to soil as the initial exposure treatment during

486

uptake assays (SI Table S13). However, for these soil treatments (unfilled symbols in

487

Figure 4), the concentration in the IW predicted with the KOC pp-LFER exceeded the

488

aqueous solubility of the compound. Hence, for these cases, the prediction of the

489

concentration in the plant (Eq. (7)) was made using the aqueous solubility of the

490

compound as the exposure concentration. This yielded single predictions (the horizontal

491

trends in Figure 4), especially for MCs with low aqueous solubilities, such as RDX and

492

HMX (SI Table S1).

493

Figure 4C shows the ten plant species in the dataset. No bias in the quality of

494

predictions was observed as a function of plant species. This indicates that using a

495

couple 𝑓𝑓𝐶𝐶𝐶𝐶𝐶𝐶 values representing the predominant families in the dataset is not an

496

unreasonable approximation.

29



497

Despite the lack of uniformity in the plant concentration data and some

498

uncertainty in the soil concentration over the duration of the exposure, the prediction of

499

concentrations from independent uptake assays for all three growth media using Eq. (9)

500

yielded a RMSE = 0.429 (excluding predictions at aqueous solubility). This RMSE is

501

competitive relative to that produced by a detailed dynamic model based on

502

physiological plant uptake by Trapp and Eggen 11, RMSE = 0.578 (calculated using data

503

in their Figure 2). Their model predicts concentrations in plants, such as barley and

504

carrot (leaves, steams, and/or roots; excluding fruits for comparison to this work), from

505

greenhouse experiments for nonionic polar organic compounds. Unfortunately,

506

comparison to the other models for bioconcentration of organic compounds in plants

507

cited previously was not possible as a validation to an independent dataset is often not

508

presented or it is performed as a cross–validation 8.

509

Implications

510

Bioconcentration of MCs and MLCs in grasses can be estimated based on the

511

partitioning between the growth medium solids, IW, and plant. Partitioning between soil

512

OC and IW, and between IW and plant cuticle for MCs, MLCs, and other organic

513

compounds can be predicted with pp–LFERs. Smaller estimation errors (RMSE =

514

0.386) for the KCut pp–LFER are obtained using QCAP as the solute descriptors for all

515

compounds relative to using Absolv–AP or Exp–AP. The superior quality of the QCAP

516

and the diversity of the KCut solute training dataset, enable the better characterization of

517

the solvation properties of the plant cuticle.

518

A demonstration of the prediction capabilities of the partitioning–based model

519

to estimate concentrations from independent plant uptake assays provides evidence of

520

the model’s usefulness and practicality. More accurate predictions are obtained for the

30


Page 30 of 40

Page 31 of 40


521

estimation of plant concentrations from assays performed in sand or water than those in

522

complex soils.

523

The results presented suggest that the partitioning–based plant bioconcentration

524

model proposed (Eq. (9)), which utilizes quantum chemically computed Abraham

525

parameters (QCAP) and two pp–LFERs, can be used to predict the MCs concentrations

526

in grasses from molecular structure only.

527

Acknowledgements

528

The authors thank Prof. Stanley Sandler in the Department of Chemical and

529

Biomolecular Engineering at the University of Delaware for the use of his research

530

group’s quantum–based solvation model, COSMO–SAC 2013. This project was funded

531

by the Strategic Environmental Research and Development Program (SERDP), Project

532

ER-1734.

533

Supporting Information Available

534

Selected characteristics and physicochemical properties of the MCs and MLCs.

535

Details on the equilibrium assumption for transport of organic compounds from soil into

536

plants. Comparative analysis for the performance of the log KOC pp–LFER chosen

537

relative to others available. Literature data collected and solute descriptors used for the

538

calibration and validation of the proposed models. Details on the results obtained with

539

the inclusion of carbohydrates in the bioconcentration model.

31



540

REFERENCES

541 542 543 544 545 546 547

1. Ampleman G, Thiboutot S, Lewis J, Marois A, Gagnon A, Bouchard M, Jenkins T, Ranney TA, Pennington JC. Evaluation of the contamination by explosives and metals in soils, vegetation, surface water and sediment at Cold Lake Air Weapons Range (CLAWR), Alberta, Phase II Final report. Technical Report No. DRDC-Valcartier-TR2004-204. Valcartier, Québec, Canada: Defence Research and Development Canada Valcartier, Valcartier QUE (CAN); Cold Regions Research and Engineering Lab, Hanover NH (US). 2004;1-116.

548 549 550 551

2. Jenkins, T.F.; Hewitt, A.D.; Grant, C.L.; Thiboutot, S.; Ampleman, G.; Walsh, M.E.; Ranney, T.A.; Ramsey, C.A.; Palazzo, A.J.; Pennington, J.C. Identity and distribution of residues of energetic compounds at army live-fire training ranges. Chemosphere 2006, 63 (8), 1280-1290; 10.1016/j.chemosphere.2005.09.066.

552 553 554

3. Clausen, J.L.; Korte, N. Fate and transport of energetics from surface soils to groundwater, In Chappell, M.A.; Price, C.L.; George, R.D., Eds.; American Chemical Society: Washington, DC, 2011; Vol.1069 pp. 273-316.

555 556 557

4. Briggs, G.G.; Bromilow, R.H.; Evans, A.A. Relationships between lipophilicity and root uptake and translocation of non-ionised chemicals by barley. Pestic. Sci. 1982, 13 (5), 495-504; 10.1002/ps.2780130506.

558 559 560

5. Briggs, G.G.; Bromilow, R.H.; Evans, A.A.; Williams, M. Relationships between lipophilicity and the distribution of non-ionised chemicals in barley shoots following uptake by the roots. Pestic. Sci. 1983, 14 (5), 492-500; 10.1002/ps.2780140506.

561 562

6. Travis, C.C.; Arms, A.D. Bioconcentration of organics in beef, milk, and vegetation. Environ. Sci. Technol. 1988, 22 (3), 271-274; 10.1021/es00168a005.

563 564 565

7. Trapp, S.; Matthies, M. Generic one-compartment model for uptake of organic chemicals by foliar vegetation. Environ. Sci. Technol. 1995, 29 (9), 2333-2338; 10.1021/es00009a027.

32


Page 32 of 40

Page 33 of 40


566 567 568

8. Dowdy, D.L.; McKone, T.E. Predicting plant uptake of organic chemicals from soil or air using octanol/water and octanol/air partition ratios and a molecular connectivity index. Environ. Sci. Technol. 1997, 16 (12), 2448-2456; 10.1002/etc.5620161203.

569 570 571

9. Platts, J.A.; Abraham, M.H. Partition of volatile organic compounds from air and from water into plant cuticular matrix: An LFER analysis. Environ. Sci. Technol. 2000, 34 (2), 318-323; 10.1021/es9906195.

572 573 574

10. Chiou, C.T.; Sheng, G.; Manes, M. A partition-limited model for the plant uptake of organic contaminants from soil and water. Environ. Sci. Technol. 2001, 35 (7), 14371444; 10.1021/es0017561.

575 576 577

11. Trapp, S.; Eggen, T. Simulation of the plant uptake of organophosphates and other emerging pollutants for greenhouse experiments and field conditions. Environ. Sci. Pollut. R. 2013, 20 (6), 4018-4029; 10.1007/s11356-012-1337-7.

578 579 580

12. Li, H.; Sheng, G.; Chiou, C.T.; Xu, O. Relation of organic contaminant equilibrium sorption and kinetic uptake in plants. Environ. Sci. Technol. 2005, 39 (13), 4864-4870; 10.1021/es050424z.

581 582 583 584

13. Bromilow, R.H.; Chamberlain, K. Principles governing uptake and transport of chemicals, In Plant Contamination: Modeling and Simulation of Organic Chemical Processes, Trapp, S.; McFarlane, J.C., Eds.; CRC Press, Inc.: Boca Raton, FL, 1995; pp. 37-68.

585 586 587

14. Cunningham, S.D.; Anderson, T.A.; Schwab, A.P.; Hsu, F.C. Phytoremediation of soils contaminated with organic pollutants, In Advances in Agronomy, Sparks, D.L., Ed.; Academic Press, Inc.: San Diego, CA, 1996; Vol.56 pp. 55-114.

588 589

15. Collins, C.; Fryer, M.; Grosso, A. Plant uptake of non-ionic organic chemicals. Environ. Sci. Technol. 2006, 40 (1), 45-52; 10.1021/es0508166.

590 591 592 593

16. Pennington JC, Myers TE, Davis WM, Olin TJ, McDonald TA, Hayes CA, Townsend DM. Impacts of sorption on in situ bioremediation of explosivescontaminated soils. Technical Report No. IRRP-95-1. Vicksburg, MS: U,S. Army Waterways Experiment Station, Corps of Engineers. 1995;1-54.

33



594 595 596

17. Larson, S.L.; Martin, W.A.; Escalon, B.L.; Thompson, M. Dissolution, sorption, and kinetics involved in systems containing explosives, water, and soil. Environ. Sci. Technol. 2007, 42 (3), 786-792; 10.1021/es0717360.

597 598 599 600

18. Berthelot, Y.; Valton, É; Auroy, A.; Trottier, B.; Robidoux, P.Y. Integration of toxicological and chemical tools to assess the bioavailability of metals and energetic compounds in contaminated soils. Chemosphere 2008, 74 (1), 166-177; 10.1016/j.chemosphere.2008.07.056.

601 602 603 604

19. Savard, K.; Sarrazin, M.; Dodard, S.G.; Monteil-Rivera, F.; Kuperman, R.G.; Hawari, J.; Sunahara, G.I. Role of soil interstitial water in the accumulation of hexahydro-1,3,5-trinitro-1,3,5-triazine in the earthworm Eisenia andrei. Environ. Toxicol. Chem. 2010, 29 (4), 998-1005; 10.1002/etc.113.

605 606 607 608

20. Kalderis, D.; Juhasz, A.L.; Boopathy, R.; Comfort, S. Soils contaminated with explosives: Environmental fate and evaluation of state-of-the-art remediation processes (IUPAC Technical Report). Pure Appl. Chem. 2011, 83 (7), 1407-1484; 10.1351/PACREP-10-01-05.

609 610

21. Karickhoff, S.W.; Brown, D.S.; Scott, T.A. Sorption of hydrophobic pollutants on natural sediments. Water Res. 1979, 13 (3), 241-248; 10.1016/0043-1354(79)90201-X.

611 612 613

22. Karickhoff, S.W. Semi-empirical estimation of sorption of hydrophobic pollutants on natural sediments and soils. Chemosphere 1981, 10 (8), 833-846; 10.1016/00456535(81)90083-7.

614 615 616 617

23. Schwarzenbach, R.P.; Gschwend, P.M.; Imboden, D.M. Sec. 9.3: Sorption of Neutral Organic Compounds from Water to Solid-Phase Organic Matter (POM), In Environmental Organic Chemistry, 2nd ed.; John Wiley & Sons: Hoboken, NJ, 2005; pp. 291-294.

618 619 620

24. Liu, Z.; He, Y.; Xu, J.; Huang, P.; Jilani, G. The ratio of clay content to total organic carbon content is a useful parameter to predict adsorption of the herbicide butachlor in soils. Environ. Pollut. 2008, 152 (1), 163-171; 10.1016/j.envpol.2007.05.006.

621 622 623

25. Dontsova, K.M.; Hayes, C.; Pennington, J.C.; Porter, B. Sorption of high explosives to water-dispersible clay: influence of organic carbon, aluminosilicate clay, and extractable iron. J. Environ. Qual. 2009, 38 (4), 1458-1465; 10.2134/jeq2008.0183. 34


Page 34 of 40

Page 35 of 40


624 625 626

26. Milinovic, J.; Lacorte, S.; Rigol, A.; Vidal, M. Sorption behaviour of nonylphenol and nonylphenol monoethoxylate in soils. Chemosphere 2015, 138, 952-959; 10.1016/j.chemosphere.2014.12.037.

627 628 629

27. Sabljić, A.; Güsten, H.; Verhaar, H.; Hermens, J. QSAR modelling of soil sorption. Improvements and systematics of log KOC vs. log KOW correlations. Chemosphere 1995, 31 (11–12), 4489-4514; 10.1016/0045-6535(95)00327-5.

630 631 632 633

28. Gawlik, B.M.; Sotiriou, N.; Feicht, E.A.; Schulte-Hostede, S.; Kettrup, A. Alternatives for the determination of the soil adsorption coefficient, KOC, of nonionicorganic compounds — a review. Chemosphere 1997, 34 (12), 2525-2551; 10.1016/S0045-6535(97)00098-2.

634 635 636

29. Doucette, W.J. Quantitative structure-activity relationships for predicting soilsediment sorption coefficients for organic chemicals. Environ. Sci. Technol. 2003, 22 (8), 1771-1788; 10.1897/01-362.

637 638 639 640

30. von Oepen, B.; Kördel, W.; Klein, W.; Schüürmann, G. QSAR in environmental toxicology predictive QSPR models for estimating soil sorption coefficients: potential and limitations based on dominating processes. Sci. Total Environ. 1991, 109, 343-354; 10.1016/0048-9697(91)90189-L.

641 642 643

31. Wei, D.B.; Wu, C.D.; Wang, L.S.; Hu, H.Y. QSPR-based prediction of adsorption of halogenated aromatics on yellow-brown soil. SAR QSAR Environ. Res. 2003, 14 (3), 191-198; 10.1080/10629360310000101773.

644 645 646

32. Kozerski, G.E.; Xu, S.; Miller, J.; Durham, J. Determination of soil-water sorption coefficients of volatile methylsiloxanes. Environ. Toxicol. Chem. 2014, 33 (9), 19371945; 10.1002/etc.2640.

647 648 649

33. Goss, K.; Schwarzenbach, R.P. Linear free energy relationships used to evaluate equilibrium partitioning of organic compounds. Environ. Sci. Technol. 2001, 35 (1), 19; 10.1021/es000996d.

650 651 652

34. Barbour, J.P.; Smith, J.A.; Chiou, C.T. Sorption of aromatic organic pollutants to grasses from water. Environ. Sci. Technol. 2005, 39 (21), 8369-8373; 10.1021/es0504946.

35



653 654 655

35. Zhang, M.; Zhu, L. Sorption of polycyclic aromatic hydrocarbons to carbohydrates and lipids of ryegrass root and implications for a sorption prediction model. Environ. Sci. Technol. 2009, 43 (8), 2740-2745; 10.1021/es802808q.

656 657

36. Hung, H.; Lin, T.; Chiou, C.T. Partition coefficients of organic contaminants with carbohydrates. Environ. Sci. Technol. 2010, 44 (14), 5430-5436; 10.1021/es1004413.

658 659 660

37. Abraham, M.H. Scales of solute hydrogen-bonding: their construction and application to physicochemical and biochemical processes. Chem. Soc. Rev. 1993, 22 (2), 73-83; 10.1039/CS9932200073.

661 662 663

38. Goss, K. Predicting the equilibrium partitioning of organic compounds using just one linear solvation energy relationship (LSER). Fluid Phase Equilibr. 2005, 233 (1), 19-22; 10.1016/j.fluid.2005.04.006.

664 665 666

39. Hale, S.E.; Cornelissen, G.; Arp, H.P.H. Comment on 'Partition coefficients of organic contaminants with carbohydrates'. Environ. Sci. Technol. 2011, 45 (3), 11581158; 10.1021/es103754p.

667 668 669

40. Meador, J.; Casillas, E.; Sloan, C.; Varanasi, U. Comparative Bioaccumulation of Polycyclic Aromatic-Hydrocarbons from Sediment by 2 Infaunal Invertebrates. Mar. Ecol. Prog. Ser. 1995, 123 (1-3), 107-124; 10.3354/meps123107.

670 671 672

41. Jager, T. Mechanistic approach for estimating bioconcentration of organic chemicals in earthworms (Oligochaeta). Environ. Toxicol. Chem. 1998, 17 (10), 2080-2090; 10.1897/1551-5028(1998)0172.3.CO;2.

673 674

42. Welke, B.; Ettlinger, K.; Riederer, M. Sorption of Volatile Organic Chemicals in Plant Surfaces. Environ. Sci. Technol. 1998, 32 (8), 1099-1104; 10.1021/es970763v.

675 676 677

43. Kraaij, R.; Mayer, P.; Busser, F.; Bolscher, M.; Seinen, W.; Tolls, J. Measured porewater concentrations make equilibrium partitioning work - A data analysis. Environ. Sci. Technol. 2003, 37 (2), 268-274; 10.1021/es020116q.

678 679 680

44. Arnot, J.A.; Gobas, F.A.P.C. A review of bioconcentration factor (BCF) and bioaccumulation factor (BAF) assessments for organic chemicals in aquatic organisms. Env. Rev. 2006, 14 (4), 257-297; 10.1139/A06-005.

36


Page 36 of 40

Page 37 of 40


681 682 683 684

45. Dodard, S.G.; Sarrazin, M.; Hawari, J.; Paquet, L.; Ampleman, G.; Thiboutot, S.; Sunahara, G.I. Ecotoxicological assessment of a high energetic and insensitive munitions compound: 2,4-Dinitroanisole (DNAN). J. Hazard. Mater. 2013, 262, 143150; 10.1016/j.jhazmat.2013.08.043.

685 686 687

46. Kuo, D.T.F.; Di Toro, D.M. Biotransformation model of neutral and weakly polar organic compounds in fish incorporating internal partitioning. Environ. Toxicol. Chem. 2013, 32 (8), 1873-1881; 10.1002/etc.2259.

688 689 690

47. Torralba-Sanchez, T.L. Bioconcentration of munitions compounds in plants and worms: experiments and modeling. Doctoral Dissertation, University of Delaware, Newark, DE, USA, 2016.

691 692 693

48. Torralba–Sanchez, T.L.; Kuo, D.T.F.; Allen, H.E.; Di Toro, D.M. Bioconcentration Factors and Plant–Water Partition Coefficients of Munitions Compounds in Barley. Manuscript submitted for publication. 2017.

694 695 696

49. Singh, N.; Berns, A.E.; Hennecke, D.; Hoerner, J.; Koerdel, W.; Schaeffer, A. Effect of soil organic matter chemistry on sorption of trinitrotoluene and 2,4-dinitrotoluene. J. Hazard. Mater. 2010, 173 (1–3), 343-348; 10.1016/j.jhazmat.2009.08.090.

697 698 699

50. Alavi, G.; Chung, M.; Lichwa, J.; D’Alessio, M.; Ray, C. The fate and transport of RDX, HMX, TNT and DNT in the volcanic soils of Hawaii: A laboratory and modeling study. J. Hazard. Mater. 2011, 185 (2–3), 1600-1604; 10.1016/j.jhazmat.2010.10.039.

700 701 702

51. Clausen, J.L.; Scott, C.; Osgerby, I. Fate of Nitroglycerin and Dinitrotoluene in Soil at Small Arms Training Ranges. Soil Sediment Contam. 2011, 20 (6), 649-671; 10.1080/15320383.2011.594108.

703 704 705

52. Sharma, P.; Mayes, M.A.; Tang, G. Role of soil organic carbon and colloids in sorption and transport of TNT, RDX and HMX in training range soils. Chemosphere 2013, 92 (8), 993-1000; 10.1016/j.chemosphere.2013.03.028.

706 707 708

53. Miglino, A.N. Modeling the adsorption and desorption of munition constituents on soils using reversible and resistant components. Doctoral Dissertation, University of Delaware, Newark, DE, USA, 2015.

37



709 710 711

54. Bargel, H.; Koch, K.; Cerman, Z.; Neinhuis, C. Structure-function relationships of the plant cuticle and cuticular waxes - a smart material? Funct. Plant Biol. 2006, 33 (10), 893-910; 10.1071/FP06139.

712 713

55. Dominguez, E.; Cuartero, J.; Heredia, A. An overview on plant cuticle biomechanics. Plant Sci. 2011, 181 (2), 77-84; 10.1016/j.plantsci.2011.04.016.

714 715 716

56. Schreiber, L.; Schonherr, J. Water and solute permeability of plant cuticles: measurement and data analysis. Springer-Verlag Berlin Heidelberg: Berlin, Germany, 2009.

717 718

57. Arey, J.; Green, W.; Gschwend, P. The electrostatic origin of Abraham's solute polarity parameter. J. Phys. Chem. B 2005, 109 (15), 7564-7573; 10.1021/jp044525f.

719 720

58. Advanced Chemistry Development Inc. ACD/PhysChem Suite. Absolv module. 2013, v. 5.0.

721 722 723

59. Liang, Y.; Kuo, D.T.F.; Allen, H.E.; Di Toro, D.M. Experimental determination of solvent-water partition coefficients and Abraham parameters for munition constituents. Chemosphere 2016, 161, 429-437; 10.1016/j.chemosphere.2016.07.028.

724 725 726 727

60. Ulrich, N., Endo, S., Brown, T.N., Watanabe, N., Bronner, G., Abraham, M.H., Goss, K.-U., UFZ-LSER database v. 3.2 [Internet], Leipzig, Germany, Helmholtz Centre for Environmental Research-UFZ. 2016 [accessed on 16.11.2016]. Available from http://www.ufz.de/lserd.

728 729 730

61. Liang, Y. Determining Abraham solute and system parameters for neutral organic compounds using quantum chemical methods. Doctoral Dissertation, University of Delaware, Newark, DE, USA, 2016.

731 732 733

62. Liang, Y.; Xiong, R.; Sandler, S.L.; Di Toro, D.M. Quantum Chemically Estimated Abraham Solute Parameters Using Multiple Solvent-Water Partition Coefficients and Molecular Polarizability. Manuscript submitted for publication. 2017.

734 735

63. R Development Core Team R: A language and environment for statistical computing. 2015, v. 3.2.0.

38


Page 38 of 40

Page 39 of 40


736 737 738

64. Kipka, U.; Di Toro, D.M. A linear solvation energy relationship model of organic chemical partitioning to particulate organic carbon in soils and sediments. Environ. Toxicol. Chem. 2011, 30 (9), 2013-2022; 10.1002/etc.611.

739 740 741

65. Holloway, P.J.; Baker, E.A. Isolation of plant cuticles with zinc chloridehydrochloric acid solution. Plant Physiol. 1968, 43 (11), 1878-1879; 10.1104/pp.43.11.1878.

742 743 744 745

66. Riederer, M.; Schonherr, J. Accumulation and transport of (2,4dichlorophenoxy)acetic acid in plant cuticles: I. Sorption in the cuticular membrane and its components. Ecotoxicol. Environ. Saf. 1984, 8 (3), 236-247; 10.1016/01476513(84)90027-7.

746 747 748

67. Sabljic, A.; Gusten, H.; Schonherr, J.; Riederer, M. Modeling plant uptake of airborne organic-chemicals .1. Plant cuticle water partitioning and molecular connectivity. Environ. Sci. Technol. 1990, 24 (9), 1321-1326; 10.1021/es00079a004.

749 750 751

68. Kerler, F.; Schonherr, J. Accumulation of Lipophilic Chemicals in Plant Cuticles Prediction from Octanol Water Partition-Coefficients. Arch. Environ. Contam. Toxicol. 1988, 17 (1), 1-6.

752 753

69. Pribyl, D.W. A critical review of the conventional SOC to SOM conversion factor. Geoderma 2010, 156 (3–4), 75-83; 10.1016/j.geoderma.2010.02.003.

754 755 756

70. Platts, J.A.; Butina, D.; Abraham, M.H.; Hersey, A. Estimation of molecular linear free energy relation descriptors using a group contribution approach. J. Chem. Inf. Comput. Sci. 1999, 39 (5), 835-845; 10.1021/ci980339t.

757 758 759

71. Hou, T.J.; Xia, K.; Zhang, W.; Xu, X.J. ADME evaluation in drug discovery. 4. Prediction of aqueous solubility based on atom contribution approach. J. Chem. Inf. Comput. Sci. 2004, 44 (1), 266-275; 10.1021/ci034184n.

760 761 762

72. Kandhakumar, G.; Jothi, M.; Kumaradhas, P. Probing the effect of nitro groups in nitramine based energetic molecules: a DFT and charge density study. Mol. Simulat. 2016, 42 (3), 173-185; 10.1080/08927022.2015.1031130.

763 764 765

73. Lin, S.; Sandler, S.I. A priori phase equilibrium prediction from a segment contribution solvation model. Ind. Eng. Chem. Res. 2002, 41 (5), 899-913; 10.1021/ie001047w. 39



766 767 768

74. Endo, S.; Goss, K. Applications of polyparameter linear free energy relationships in environmental chemistry. Environ. Sci. Technol. 2014, 48 (21), 12477-12491; 10.1021/es503369t.

769 770 771 772

75. Schwarzenbach, R.P.; Gschwend, P.M.; Imboden, D.M. Sec. 3.2: Molecular Interactions Determining the Partitioning of Organic Compounds Between Different Phases, In Environmental Organic Chemistry, 2nd ed.; John Wiley & Sons: Hoboken, NJ, 2005; pp. 59-63.

773 774

76. Yeats, T.H.; Rose, J.K.C. The Formation and Function of Plant Cuticles. Plant Physiol. 2013, 163 (1), 5-20; 10.1104/pp.113.222737.

775 776 777

77. Kipka, U.; Di Toro, D.M. Technical basis for polar and nonpolar narcotic chemicals and polycyclic aromatic hydrocarbon criteria. III. A polyparameter model for target lipid partitioning. Environ. Toxicol. Chem. 2009, 28 (7), 1429-1438; 10.1897/08-364.1.

778 779 780

78. Esteve-Nunez, A.; Caballero, A.; Ramos, J.L. Biological degradation of 2,4,6trinitrotoluene. Microbiol. Mol. Biol. Rev. 2001, 65 (3), 335-352; 10.1128/MMBR.65.3.335-352.2001.

40


Page 40 of 40

Estimating GrassâSoil Bioconcentration of Munitions Compounds from

Recommend Documents

Estimating GrassâSoil Bioconcentration of Munitions Compounds from