and Polyfluoroalkyl Substances (PFASs) - ACS Publications

Subscriber access provided by UNIVERSITY OF TOLEDO LIBRARIES

Ecotoxicology and Human Environmental Health

Predicting Relative Protein Affinity of Novel Per- and Polyfluoroalkyl Substances (PFASs) by An Efficient Molecular Dynamics Approach Weixiao Cheng, and Carla Ng Environ. Sci. Technol., Just Accepted Manuscript • DOI: 10.1021/acs.est.8b01268 • Publication Date (Web): 13 Jun 2018 Downloaded from http://pubs.acs.org on June 14, 2018

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 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 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.

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 30

Environmental Science & Technology

1

Predicting Relative Protein Affinity of Novel Per- and Polyfluoroalkyl

2

Substances (PFASs) by An Efficient Molecular Dynamics Approach

3

Weixiao Cheng and Carla A. Ng*

4

Department of Civil and Environmental Engineering, University of Pittsburgh, Pittsburgh,

5

Pennsylvania 15261, USA

6

* Address correspondence to: Carla A. Ng, Department of Civil & Environmental Engineering,

7

University of Pittsburgh, 3700 O’Hara St, Pittsburgh, PA 15261. Tel.: 412-383-4075. Fax: 412-

8

624-0135. E-mail: [email protected]

9

1

ACS Paragon Plus Environment


10

ABSTRACT

11

With the phasing out of long-chain per- and polyfluoroalkyl substances (PFASs), production of a

12

wide variety of alternative PFASs has increased to meet market demand. However, little is

13

known about the bioaccumulation potential of these replacement compounds. Here, we

14

developed a modeling workflow that combines molecular docking and molecular dynamics

15

simulation techniques to estimate the relative binding affinity of a total of 15 legacy and

16

replacement PFASs for human and rat liver-type fatty acid binding protein (hLFABP and

17

rLFABP). The predicted results were compared with experimental data extracted from three

18

different studies. There was good correlation between predicted free energies of binding and

19

measured binding affinities, with correlation coefficients of 0.97, 0.79, and 0.96, respectively.

20

With respect to replacement PFASs, our results suggest that EEA and ADONA are at least as

21

strongly bound to rLFABP as perfluoroheptanoic acid (PFHpA), and as strongly bound to

22

hLFABP as perfluorooctanoic acid (PFOA). For F-53 and F-53B, both have similar or stronger

23

binding affinities than perfluorooctane sulfonate (PFOS). Given that interactions of PFASs with

24

proteins (e.g., LFABPs) are important determinants of bioaccumulation potential in organisms,

25

these alternatives could be as bioaccumulative as legacy PFASs, and are therefore not necessarily

26

safer alternatives to long-chain PFASs.

27

2


Page 2 of 30

Page 3 of 30


28 29

3



30

INTRODCUTION

31

Per- and polyfluoroalkyl substances (PFASs) are a family of chemicals that have been widely

32

used in a variety of industrial and consumer applications, from fire-fighting foams to food

33

contact materials to apparel.1-6 It is estimated there are more than 3000 PFASs currently on the

34

global market.7, 8 The characteristic carbon-fluorine bonds of PFASs make them extremely

35

resistant to chemical and thermal degradation.9 The most commonly used PFASs, such as

36

perfluorooctanoic acid (PFOA) and perfluorooctanesulfonic acid (PFOS) have been detected

37

ubiquitously in the environment, wildlife and in humans.9-12 Experimental studies have indicated

38

that long-chain PFASs (defined as C7 and longer perfluoroalkyl carboxylic acids (PFCAs) and

39

C6 and longer perfluoroalkanesulfonic acids (PFSAs)) accumulate in the human body, with

40

biological half-lives estimated to be several years.13 Moreover, toxicological studies in animals

41

have found long-chain PFAS exposure causes toxic effects on reproduction and development,

42

and on the nervous, endocrine, and immune systems.14-21

43

The concern about their persistence, bioaccumulation and toxicity has led to a phasing out of the

44

production of long-chain PFASs for the majority of uses.22-24 To take their place, manufacturers

45

have started using alternatives that include shorter-chain homologues of PFOA and PFOS, as

46

well as perfluoroether carboxylic acids (PFECAs) and perfluoroether sulfonic acids (PFESAs).23-

47

26

48

identified in a recent review paper by Wang et al.23 These alternatives are increasingly detected

49

in the environment and appear to accumulate in some aquatic organisms.27-30 However, the

50

identity and frequency of use of many other alternative PFASs remains largely unknown, leading

51

scientists to employ extensive non-target analytical techniques to puzzle out the structures

52

present in complex environmental mixtures of PFAS.31-34 In addition, little is known about the

A number of those fluorinated alternatives used in industrial and consumer products have been

4


Page 4 of 30

Page 5 of 30


53

potential impacts of alternative PFASs on humans and the environment; there is a particular lack

54

of information on the bioaccumulation potential and toxicity of PFECAs and PFESAs.23

55

Given the large number of PFASs and the scarce knowledge about their potential hazards, a rapid

56

and reliable method to predict the behavior of these chemicals in the environment and organisms

57

would be of great benefit. In our previous work, we developed mechanistic physiologically based

58

pharmacokinetic (PBPK) models that explicitly consider binding with serum albumin, liver-type

59

fatty acid binding protein (LFABP), and organic anion transporters to predict the

60

bioaccumulation of PFCAs and PFSAs in different tissues of both fish and rat.35, 36 The success

61

of our models demonstrated that the interaction of PFASs with proteins plays an essential role in

62

determining their bioaccumulation potential in organisms, and thus could be used as a proxy for

63

bioaccumulation assessment. However, the protein binding parameters used to build the models

64

were limited to a small subset of PFASs (e.g., PFOA and PFOS).37

65

To provide insights into the bioaccumulation potential of novel PFASs and generate more

66

protein binding parameters for PBPK models, we proposed an in-silico method based on

67

molecular dynamics (MD) simulations to predict PFAS-protein interactions. Specifically, we

68

employed the molecular mechanics combined with Poisson-Boltzmann surface area (MM-

69

PBSA) continuum solvation method to calculate binding affinities between ligands (i.e., PFASs)

70

and proteins. As a starting point, we focused on LFABPs in this study because of they have high-

71

quality 3-dimensional crystal structures available as well as experimental binding affinity data

72

with different PFASs, which can be used for method evaluation. In the MM-PBSA method, the

73

free energy of binding, ∆Gbind, for a chemical reaction: P + L = PL (P denotes the protein and L

74

the ligand) is calculated from:

75

= − −

5


(1)


76

where the free energy of a state (i.e., GP, GL, and GPL) is derived from post-processing an

77

ensemble of representative protein-ligand snapshots generated from MD simulations38. This

78

method is more computationally efficient than rigorous alchemical perturbation methods (e.g.,

79

free energy perturbation and thermodynamic integration methods), but more robust compared to

80

molecular docking based on scoring functions.39 It is worth noting that MM-PBSA is a

81

continuum solvation method and involves several thermodynamic approximations, which makes

82

the absolute binding energies unreliable.39 However, many studies have demonstrated that MM-

83

PBSA is able to successfully predict the relative binding affinities of ligands.38-43 The primary

84

goal of this study is to rank the binding affinities of PFASs for LFABPs. Given the large number

85

of PFASs potentially in commerce, an efficient method like MM-PBSA would be of great

86

benefit.

87

To test the performance of MM-PBSA for LFABP-PFAS interactions, we considered 15 PFASs

88

with different functional head groups and fluorinated carbon chain lengths including 10 legacy

89

PFASs (7 PFCAs: PFBA, PFPA, PFHxA, PFHpA, PFOA, PFNA, and 2m-PFOA; and 3 PFSAs:

90

PFBS, PFHxS, and PFOS) and 5 alternatives (3 PFECAs: ADONA, GenX, and EEA; and 2

91

PFESAs: F-53 and F-53B). The 2-dimensional structures of these chemicals are shown in the

92

Supporting Information (SI) Figure S1. The binding affinities of these chemicals were evaluated

93

for 2 different LFABPs (hLFABP and rLFABP for human and rat LFABP, respectively) which

94

have been previously experimentally determined.44-46 For estimation of ∆Gbind, the following

95

workflow was developed: the initial structure of the LFABP-PFAS complex was generated from

96

molecular docking, a powerful tool to predict the binding mode between a protein and a ligand;47

97

based on the complex structure, the MD simulation was then carried out; finally, MM-PBSA was

98

used to calculate the ∆Gbind.

6


Page 6 of 30

Page 7 of 30


99

100

MATERIALS AND METHODS

101

LFABPs and Ligands. The 3-dimensional crystal structures were obtained from the Protein

102

Data Bank (PDB, http://www.rcsb.org) for hLFABP (PDB code: 3STM48) and rLFABP (PDB

103

code: 1LFO49). These structures were selected because of their high resolution and completeness

104

of key residues, as discussed in our previous docking study of LFABP interaction with PFASs.50

105

For PFAS ligands, the 3-dimensional structures of PFOA and PFOS were extracted from 5JID

106

(PDB code for crystal structure of human transthyretin in complex with PFOA51) and 4E99 (PDB

107

code for crystal structure of human serum albumin in complex with PFOS52), respectively. The

108

other 3-dimensional structures were constructed from scratch using Avogadro (v1.2.0)53 and

109

exported in pdb file format.

110

Molecular Docking. All 15 PFAS ligands were docked to both hLFABP and rLFABP with

111

Autodock Vina (v1.1.2),54 as described in our previous study.50 Briefly, both protein and ligand

112

structures produced above were first preprocessed using AutoDock Tools (v1.5.6),55 the output

113

pdbqt files were then used for docking. For each protein, the binding site boundaries were

114

determined using the Grid menu in AutoDock Tools. According to the 3-dimensional structure of

115

rLFABP, the binding cavity is a flattened rectangular box (roughly 13 × 9 × 4 Å).49 Since there

116

are no available dimensions for hLFABP, we assumed it has a similar binding pocket as

117

rLFABP. For rLFABP, although there are two ligands in the binding cavity of the experimental

118

complex 1LFO, they are not independent of each other. That is, the secondary binding site (for

119

the ligand in the solvent-accessible location) would not exist until the primary binding site (for

120

the ligand buried inside the pocket) is filled. For hLFABP, only one ligand lies in the binding

121

cavity (i.e., the one buried inside the pocket) of 3STM. For simplification, we only considered 7



122

the primary binding site for both LFABPs. The ligands docked to the protein were assumed to be

123

in their deprotonated forms, given their low pKa.7, 56, 57 The docking experiments output binding

124

free energies and docking poses for each ligand-protein complex. The 3 binding modes with

125

lowest energies (strongest associations) and distinct conformations were chosen as initial

126

structures for MD simulations (Table S2 and S3). As a result, a total of 90 ligand-protein

127

structures were generated.

128

To assess the success of the docking experiment, we redocked the fatty acid ligands from the

129

crystalized complexes (i.e., PDB code 3STM and 1LFO) back into their corresponding receptors

130

and measured the root-mean-square deviation (RMSD) between the original crystal structure of

131

the complex and the docked ligands using PyMol;58 the results (Table S1) showed that Autodock

132

Vina can successfully predict the bound conformations of the ligands and LFABP with

133

reasonable accuracy (RMSD < 2.5 Å). In addition, our previous study also demonstrated

134

Autodock Vina can redock PFOS to human serum albumin with RMSD smaller than 2 Å.50

135

MD Simulations. The system setup and simulation of the 90 ligand-protein complexes were

136

performed with the Amber 14 suite.59 For system setup, the ff14SB force field60 was used for

137

proteins and the general AMBER force field (GAFF)61 was used for ligands. The atomic partial

138

charges of ligands were derived by AM1-BCC (AM1-bond charge correction), which is an

139

efficient method to reproduce HF/6-31G* RESP charges.62 The whole complex system was

140

explicitly solvated in a cubic box of TIP3P water molecules with a minimal distance of 12 Å

141

from solute atoms to box edges. Na+ counterions were added to neutralize the systems. Periodic

142

boundary conditions were employed for all simulations. Long-range electrostatic interactions

143

were handled by the particle mesh Ewald (PME) method.63 The cutoff for nonbonded

144

interactions was set to 8 Å. 8


Page 8 of 30

Page 9 of 30


145

The simulation was carried out by the GPU accelerated pmemd module.64 First, the solvent

146

molecules were subjected to 2000 steps of energy minimization, while the solute was constrained

147

with the harmonic force constant of 500 kcal mol-1 Å-2 to eliminate nonphysical contact between

148

solute and solvent. Next, the whole system was minimized without restraint for 1500 cycles.

149

Then, the system was heated from 0 to 300 K in 20 ps at constant volume; a weak harmonic

150

force constant of 10 kcal mol-1 Å-2 was added on the complex. After the heating phase, the

151

density of the system was adjusted to 1 g/cm3 at constant pressure (1 bar) for 100 ps with

152

restraint (10 kcal mol-1 Å-2) on the complex. Finally, the system was equilibrated at constant

153

temperature (300 K) and pressure (1 bar) for 2 ns, the temperature was controlled by Langevin

154

dynamics with a collision frequency of 2 ps-1,59 and the pressure was controlled by the isotropic

155

position scaling protocol.59 Under the same condition as in the equilibration phase, the

156

production run for the complex system was performed for 24 ns. The SHAKE bond length

157

constraints were used to allow a larger timestep of 2 fs.65 The trajectories were sampled at a time

158

interval of 16 ps to ensure each snapshot is statistically independent.39, 66 All simulations were

159

run on an AMBER GPU Certified MD workstation (Exxact Corporation, CA, USA). The

160

analysis of trajectories is provided in Figure S2.

161

MM-PBSA Calculations. Free energy calculations were conducted using the MMPBSA.py

162

program in Amber 14.38 Specifically, ∆Gbind between PFAS ligands and LFABPs were calculated

163

as follows:

164

= − −

(2)

165

where GComplex, GLFABP, and GPFAS are the free energies of complex, LFABPs, and PFAS ligands,

166

respectively. The free energy (G) of each state was estimated from the following sum:67-69

9



167

= < + + + + − >

Page 10 of 30

(3)

168

where the brackets indicate an average over MD trajectories. Inside the brackets, the first three

169

terms are molecular mechanical energy terms for bonded, electrostatic and van der Waals (vdw)

170

interactions, respectively. Gpolar is the polar solvation free energy, which was calculated using the

171

Poisson-Boltzmann (PB) implicit solvent method (which is a differential equation based on the

172

Poisson continuum dielectric model and Boltzmann distribution for the ions in the solvent).70

173

Gnonp is the nonpolar contribution, which was determined from a linear relation to solvent-

174

accessible surface area. The last term is the absolute temperature (T) multiplied by the entropy

175

(S), which was estimated by normal-mode analysis using the nmode program in Amber 14.59

176

Specifically, the entropy change is calculated based on the change in three types of molecular

177

motions (i.e., translational, rotational, and vibrational motion) of the system when a ligand binds

178

to the receptor.71 For the thermodynamic variables to control the calculation, the recommended

179

values for Amber 14 were employed (e.g., the external and internal dielectric constants are 80.0

180

and 1.0, respectively).59 Finally, a single trajectory protocol (STP) approach was used for the free

181

energy calculation.38 That is, the calculation of Gcomplex, GLFABP, and GPFAA were based on a

182

single MD trajectory of the complex system, rather than on 3 trajectories generated from 3

183

separate MD simulations. We employed the STP method based on the following considerations.

184

First, since no study is available indicating that the binding of PFASs to LFABPs causes a

185

significant conformation change, we assumed the proteins and ligands are comparable in the

186

bound and unbound states. Second, all ligands in our study have similar chemical structures (i.e.,

187

fluorinated carbon chain and functional head group). Finally, The STP method is less

188

computationally expensive, and it also leads to a cancellation of the Ebond term in equation 3,

189

which improves the precision of the results tremendously.38, 39

10


Page 11 of 30


190

As described in the MD simulation section, a total of 1500 independent snapshots (24 ns

191

production divided by the timestep of 16 ps) were generated for an individual complex system

192

(i.e., one of the 90 protein-ligand structures). These MD snapshots were evenly divided into 6

193

groups (i.e., the first 4 ns or 250 snapshots is group one, the second 4 ns or 250 snapshots is

194

group two and so forth, each group could be considered as an independent simulation phase). In

195

each group, the binding free energy was calculated using equations 2 and 3 and then averaged

196

over all snapshots; for the entropy calculation, because it is very computationally expensive,38

197

only 10 snapshots in each group were considered for normal-mode analysis. Those 10 snapshots

198

were collected as a subset of the total of 250 snapshots in each group based on an interval of 25

199

snapshots. All MM-PBSA calculations were carried out on Bridges, part of the Pittsburgh

200

Supercomputing Center (www.psc.edu).

201

Free Energy Decomposition. To gain insights into the contributions to the binding free energy,

202

energy decomposition on a per-residue basis was performed using the decomp program in Amber

203

14.38 The per-residue decomposition scheme can decompose calculated free energies into

204

specific residue contributions based on the Poisson-Boltzmann implicit solvent model.72, 73 The

205

contribution of each residue in LFABP to the total free energy of the complex system was

206

estimated.

207

Data Analysis. The final binding free energies estimated by MM-PBSA were averaged over 6

208

MD simulation phases and 3 binding modes for each LFABP-PFAS pair. The standard error of

209

the mean was then calculated. Finally, the correlation analysis for predicted free energies versus

210

carbon chain lengths and predicted free energies versus experimental binding affinities were

211

conducted. For comparison and correlation analysis, the experimental data represented as

11



212

equilibrium dissociation constant (Kd) with units of µM were translated into free energy of

213

binding (∆Gbind, in kcal/mol) as follows:74, 75

= !"#

$ %&

214

where R is gas constant (1.987 cal K-1 mol-1), T is temperature (which is assumed to be 300 K),

215

and c0 is the standard state concentration (1 M).

216

217

RESULTS

218

Method Evaluation. To evaluate the effectiveness of the method, we conducted correlation

219

analysis between the predicted ∆Gbind to the experimental ∆Gbind derived from three different

220

studies (Figure 1).44-46 For hLFABP, MM-PBSA performance varied between different

221

experimental studies. In comparison with the Sheng et al. study (Figure 1A),46 the correlation

222

coefficient was excellent (r = 0.97), while the correlation analysis between computational results

223

and the Zhang et al.45 study indicated a coefficient of 0.79 (Figure 1B), which is also acceptable.

224

The binding free energies between rLFABP and PFAS ligands from the simulation correlate very

225

well with experimental data (Figure 1C, r = 0.96). The results also show that the predicted

226

absolute binding energies of PFASs are generally lower than corresponding experimental values.

227

However, it should be emphasized that our study is primarily focused on the relative binding

228

affinities of PFASs rather than their absolute binding strengths.

229

Molecular Docking. The docking experiment was conducted mainly to predict the interaction

230

between LFABP and PFAS to find potential binding modes (the binding affinities, which we

231 12


Page 12 of 30

Page 13 of 30


232 233

Figure 1. Correlations between the average ∆Gbind calculated by MM-PBSA and the experimental values for (A)

234

hLFABP from Sheng et al.,46 (B) hLFABP system from Zhang et al.,45 and (C) rLFABP from Woodcroft et al.,44

235

Error bars indicate the standard error for predicted ∆Gbind.

236

13



237 238

Figure 2. The interactions between hLFABP and PFAS ligands. The results were generated from molecular docking

239

and plotted with Autodock Tools.55

240

14


Page 14 of 30

Page 15 of 30


241

expected to be less accurate than via MMPBSA based on MD, were also estimated by the

242

scoring function, as indicated in Table S1).47 The interaction structures for the best binding mode

243

of each LFABP-PFAS complex are indicated in Figure 2, and Figures S3 and S4. As shown, the

244

interactions between PFASs and the two LFABPs are substantially different. For hLFABP, the

245

residues closely interacting with PFASs include ARG 122, SER 39, SER 124, PHE 50, ILE 109,

246

ILE 41, and LEU 91 (a glossary for the residues is shown in Table S2); while for rLFABP, the

247

close contact residues consist of ARG 122, TYR 120, MET 74, LEU 28, and TYR 54 (except for

248

the rLFABP-PFBS interaction, where the close contact residues are ARG 122, SER 39, and SER

249

124). With respect to hydrogen (H) bonding interactions (Table 1), all PFAS ligands formed H

250

bonding with hLFABP, and most interactions occurred between ligands and residues of ARG

251

122, SER 124, or SER 39. On the other hand, only a few instances of H bonding occurred

252

between PFAS ligands and rLFABP, and the major residue participating in H bonding

253

interactions is TYR 120.

254

In addition, for both LFABPs, the predicted binding modes were similar among ligands with

255

different functional head groups (i.e., carboxyl and sulfonate groups). The binding of alternative

256

PFASs (which contain ether groups in their structures) and legacy PFASs (which contain no

257

ether group) to LFABPs show little difference in terms of conformations. The only notable

258

differences were observed for 2m-PFOA and GenX, both of which have branched structures. As

259

indicated in Figure 2, the carboxyl group in 2m-PFOA and GenX mainly interacted with THR

260

102 and SER 100, not ARG 122, SER 124 and SER 39, which were the major residues

261

interacting with the head group of other PFASs.

262

MM-PBSA. The average ∆Gbind calculated based on MM-PBSA and five energy components

263

(i.e., vdw, electrostatic, polar and nonpolar solvation energy, and entropy) for each LFABP15



Page 16 of 30

264

Table 1. The protein residues interacting with the PFAS ligands through H-bond and those having dominant energy

265

contributions to ∆Gbind of each LFABP-PFAS complex (“-” indicates no H-bond).

Ligands PFBA PFPA PFHxA PFHpA PFOA PFNA PFBS PFHxS PFOS EEA GenX ADONA 2m-PFOA F-53 F-53B

H-bond interaction ARG 122, SER 124 ARG 122, SER 124 ARG 122, SER 124 ARG 122, SER 124 ARG 122, SER 124 ARG 122, SER 124 ARG 122, SER 39 ARG 122 ARG 122, SER 124 ARG 122, SER 39 THR 102 ARG 122, SER 124 SER 100 ARG 122, SER 124 ARG 122, SER 124

hLFABP Largest energy contribution ARG 122, SER 39, ILE 52 ARG 122, VAL 83, PHE 50 ARG 122, SER 39, SER 124 ARG 122, SER 39, ILE 52 ARG 122, SER 39, ILE 52 ARG 122, SER 39, ILE 52 ARG 122, SER 124, LEU 9 ARG 122, SER 124, SER 39 ARG 122, SER 124, ILE 52 ARG 122, SER 39, ASN 111 ARG 122, ASN 111, THR 73 ARG 122, SER 39, SER 124 ARG 122, SER 100, ASN 111 ARG 122, PHE 50, SER 39 ARG 122, SER 124, SER 39

H-bond interaction TYR 120 ARG 122, SER 39 TYR 120 TYR 120 TYR 120

266

16


rLFABP Largest energy contribution SER 57, LYS 58, LYS 32 ARG 122, TYR 55, ILE 53 ARG 122, ILE 53, LYS 58 ARG 122, ILE 60, MET 74 ARG 122, TYR 55, ILE 60 ARG 122, ILE 60, ILE 53 ARG 122, SER 100, LEU 71 ARG 122, ASN 111, LEU 51 ARG 122, ILE 60, ILE 53 ARG 122, MET 74, ILE 60 ARG 122, MET 74, ILE 53 ARG 122, MET 74, TYR 55 ARG 122, TYR 120, ILE 60 ARG 122, SER 124, ILE 53 ARG 122, TYR 55, ILE 60

Page 17 of 30


267

PFAS pair are present in Tables S5 and S6. As indicated, the predicted free energies of ligands

268

interacting with hLFABP and rLFABP range from -15.76 to -2.21 kcal/mol and from -11.26 to -

269

3.74 kcal/mol, respectively. For each ligand, the predicted binding affinities with hLFABP are

270

generally higher than that of rLFABP. For both LFABPs, the most significant contribution to the

271

binding affinity is the electrostatic interaction, but this large change of electrostatic interaction

272

upon binding is compensated by the polar solvation energy. The nonpolar solvation energies are

273

very small and show a minor variation among ligands, thus having insignificant contribution to

274

the ∆Gbind.

275

Figure 3 shows the distribution of vdw, the sum of electrostatic and polar solvation energy, and

276

entropy changes for each ligand-protein system. The sum of electrostatic and polar solvation

277

energy is shown instead of the separate contributions, because both energy terms are strongly

278

anti-correlated (r = -0.96). An obvious pattern was observed between vdw and carbon chain

279

length: as carbon chain length increases, the vdw interaction energy decreases. The entropy term

280

also indicated a similar trend, but the relationship is not as strong as vdw. With respect to

281

electrostatic and polar solvation energy, wild fluctuations were observed in both LFABP

282

systems. In particular, the sum of electrostatic and polar solvation energy for PFHxA bound to

283

hLFABP is much lower compared with other ligands.

284

A correlation analysis was conducted between predicted ∆Gbind and carbon chain length. As

285

shown in Figure 4, for both LFABP systems, the ∆Gbind indicates negative relationships with

286

carbon chain length. A strong correlation is observed for PFCAs versus rLFABP, and PFSAs

287

versus both hLFABP and rLFABP, with correlation coefficients of -0.93, -1.0, and -0.72,

288

respectively. The correlation for PFCAs versus hLFABP is relatively weak (r = -0.41), and the

17



289 290

Figure 3. The distributions of the energy components of ∆Gbind including the sum of electrostatic interactions and

291

polar solvation free energy (ELE + PB), van der Waals energy (vdw), and entropy changes upon binding. Pink

292

represents the hLFABP system, while blue is the rLFABP system.

293

18


Page 18 of 30

Page 19 of 30


294 295

Figure 4. The distribution of ∆Gbind (mean ± standard error) for different LFABP-PFAS complexes and the

296

correlation analysis between ∆Gbind and carbon chain length. Pink represents hLFABP, and blue is rLFABP.

297

19



298

major reason for this is due to the much lower predicted ∆Gbind for PFHxA, which can be further

299

attributed to the much more favorable electrostatic interaction and polar solvation energy

300

between PFHxA and hLFABP (Figure 3). In terms of predicted ∆Gbind for novel PFASs, 2

301

PFESAs exhibit comparable or stronger binding affinities than PFOS for both LFABPs. The

302

∆Gbind of EEA and ADONA are similar with PFNA when bound to hLFABP, and similar with

303

PFHpA when bound to rLFABP, while the ∆Gbind of GenX is weaker compared to that of

304

PFHxA, which has the same carbon number as GenX. Finally, 2m-PFOA has a comparable

305

binding affinity with PFOA for both LFABPs.

306

Free Energy Decomposition. The contribution of each residue in the LFABPs to the binding

307

free energy was determined based on a per-residue decomposition scheme. For each residue, all

308

free energy components in Equation 3 except entropy and nonpolar solvation energy (the

309

calculation of both terms were not available in the decomp program in Amber 14) were

310

calculated (Tables S7 and S8). In each LFABP-PFAS complex system, the residues contributing

311

most to the total free energy were determined (they account for 44 % to 85 % contribution

312

among all protein residues). As shown in Table 1, for hLFABP the residues such as ARG 122,

313

SER 39, SER 124, and ILE 52 contribute significantly to ∆Gbind among all ligands, while for the

314

rLFABP system, the residues showing strong contributions include ARG 122, ILE 60, ILE 53,

315

TYR 55, and MET 74.

316

317

DISCUSSION

318

In this study, we developed a workflow combining molecular docking and MM-PBSA based on

319

MD simulation techniques to predict the binding affinity of legacy and replacement PFASs for

20


Page 20 of 30

Page 21 of 30


320

LFABPs. Experimental data from three different studies44-46 were used to evaluate this approach,

321

and the performance is excellent for predicting PFAS binding affinity to rLFABP (r = 0.96). For

322

hLFABP, predictions are different between Zhang et al.45 (r = 0.79) and Sheng et al.46 (r = 0.97).

323

Both studies used fluorescence displacement assays to measure the dissociation constant.

324

However, the binding affinity results (expressed as Kd values in unit of µM) of Sheng et al. were

325

3 to 8 times higher than those of Zhang et al., which reveals the variation among different

326

experimental studies. Given that available experimental datasets for LFABP-PFAS complex are

327

very limited, we call for a broadening of experimental work on protein-PFAS interactions, which

328

will make validation of the predicted results more reasonable. The available data in those three

329

studies cover most traditional PFASs and two novel PFASs (i.e., GenX and F-53B). The

330

satisfactory performance of the MM-PBSA method we used demonstrates its ability to rank the

331

binding affinity of both legacy and alternative PFASs.

332

This approach also provides mechanistic understanding of how the molecular structures of

333

PFASs influence their protein binding behavior. For legacy PFASs (i.e., PFCAs and PFSAs), as

334

carbon chain length increases, the binding affinity also increases. Further analysis for each

335

energy component of ∆Gbind showed that the strong relationship between carbon chain length and

336

binding affinity was mainly caused by the vdw interaction energy and entropy change upon

337

binding, both of which indicate a close correlation with carbon chain length (Figure 3). The sum

338

of electrostatic interaction and polar solvation energy terms, on the other hand, seem to fluctuate

339

around a certain value; the extremely low value for PFHxA bound to hLFABP may be because

340

the simulation time is not long enough for the hLFABP-PFHxA complex system. For most

341

alternative PFASs, the addition of ether groups actually increased their binding affinities to

342

LFABPs in comparison with their perfluoroalkyl counterparts with same carbon numbers.

21



343

Binding free energy component analysis indicated that the largest influence of the “novel” PFAS

344

composition is that introducing oxygen in their backbone increases their chain length. A longer

345

chain length leads to greater vdw interactions with the proteins and more favorable entropy

346

changes (a larger molecule has greater molecular motions and thus higher entropy).71 Therefore,

347

the introduction of ether groups in PFASs of a given initial chain length could lead to a stronger

348

binding free energy (Figure 4). It is interesting to note that distinct from other novel PFASs,

349

GenX has a branched structure similar to 2m-PFOA, which imparts some special behavior. Due

350

to its structure, GenX (and 2m-PFOA) showed a significantly different binding mode from linear

351

PFASs; instead of interacting with ARG or SER residues, the head group of GenX mainly

352

interacted with THR residue through H-bonding (Figure 2). Furthermore, although inserting an

353

oxygen atom, the binding affinity of GenX was comparable with or even a little weaker than

354

PFHxA (which has the same number of carbons as GenX). This implies that the binding affinity

355

is closely related to the backbone chain length, not the total carbon number including branches.

356

Our results suggest that EEA and ADONA bind at least as strongly to rLFABP as PFHpA, and to

357

hLFABP at least as strongly as PFOA. For F-53 and F-53B, both have similar or stronger

358

binding affinities than PFOS. Based on our toxicokinetics model, these alternatives could be as

359

bioaccumulative as legacy PFASs. In addition, toxicological studies of F-53B have shown a

360

similar or stronger toxicity compared with PFOS.26, 46 The toxic effects (e.g., hepatotoxicity,

361

genotoxicity, and developmental toxicity) of other alternatives were also reported and

362

summarized by Wang et al.24 Taken together these bioaccumulation and toxicity results indicate

363

that those substances are not necessarily safer alternatives to legacy PFASs, particularly when

364

the backbone chain length is greater than 6.

22


Page 22 of 30

Page 23 of 30


365

Implications for risk assessment of PFASs. Given the vast number of PFASs (more than 3000)

366

potentially on the market and our limited resources (e.g., time and cost), it is not feasible to

367

evaluate all PFASs individually through experiments.7 Therefore, in-silico methods based on

368

computational biology hold great promise for the hazard and risk assessment of non-tested

369

PFASs. The MM-PBSA approach based on MD simulation we illustrate here provides reliable

370

relative protein binding affinity prediction for legacy and alternative PFASs, and thus can be

371

used for large-scale screening of protein-PFAS interactions. In addition, the binding affinities

372

generated from this approach can be further used as parameters for our previous PBPK models,

373

which were developed by considering the interactions between PFASs and proteins including

374

serum albumin, LFABPs, and membrane transporters.35, 36 The combination of MD simulation

375

and PBPK modeling will provide a flexible framework that can be used to evaluate the

376

bioaccumulation behaviors of non-tested PFASs and support their risk assessment.

377

378

AUTHOR INFORMATION

379

Corresponding Author

380

*Address: 3700 O’Hara St, Pittsburgh, PA 15261 USA. Tel: 412-383-4075; Fax: 412-624-0135;

381

E-mail: [email protected].

382

Notes

383

The authors declare no competing financial interest.

384

385

ACKNOWLEDGMENTS

23



386

The authors gratefully acknowledge the use of Bridges of the Pittsburgh Supercomputing Center

387

in the completion of this work.

388

REFERENCES

389

(1) Buck, R. C.; Franklin, J.; Berger, U.; Conder, J. M.; Cousins, I. T.; De Voogt, P.; Jensen, A. A.; Kannan, K.;

390

Mabury, S. A.; van Leeuwen, S. P. Perfluoroalkyl and polyfluoroalkyl substances in the environment: terminology,

391

classification, and origins. Integr. Environ. Assess. Manage. 2011, 7 (4), 513-541.

392

(2) D’eon, J. C.; Mabury, S. A. Is indirect exposure a significant contributor to the burden of perfluorinated acids

393

observed in humans? Environ. Sci. Technol. 2011, 45 (19), 7974-7984.

394

(3) Kannan, K. Perfluoroalkyl and polyfluoroalkyl substances: current and future perspectives. Environ. Chem.

395

2011, 8 (4), 333-338.

396

(4) Place, B. J.; Field, J. A. Identification of novel fluorochemicals in aqueous film-forming foams used by the US

397

military. Environ. Sci. Technol. 2012, 46 (13), 7120-7127.

398

(5) Liu, X.; Guo, Z.; Folk IV, E. E.; Roache, N. F. Determination of fluorotelomer alcohols in selected consumer

399

products and preliminary investigation of their fate in the indoor environment. Chemosphere 2015, 129, 81-86.

400

(6) Kotthoff, M.; Müller, J.; Jürling, H.; Schlummer, M.; Fiedler, D. Perfluoroalkyl and polyfluoroalkyl substances

401

in consumer products. Environ. Sci. Pollut. Res. 2015, 22 (19), 14546-14559.

402

(7) Wang, Z.; DeWitt, J. C.; Higgins, C. P.; Cousins, I. T., A never-ending story of per-and polyfluoroalkyl

403

substances (PFASs)? Environ. Sci. Technol. 2017, 51 (5), 2508-2518.

404

(8) OECD (Organisation for Economic Co-operationand Development), Summary report on the new comprehensive

405

global database of Per- and Polyfluoroalkyl Substances (PFASs). Publications Series on Risk Management No.

406

39, 2018. http://www.oecd.org/officialdocuments/publicdisplaydocumentpdf/?cote=ENV-JM-

407

MONO(2018)7&doclanguage=en

408

(9) Lau, C.; Anitole, K.; Hodes, C.; Lai, D.; Pfahles-Hutchens, A.; Seed, J. Perfluoroalkyl acids: a review of

409

monitoring and toxicological findings. Toxicol. Sci. 2007, 99 (2), 366-394.

410

(10) Krafft, M. P.; Riess, J. G. Per-and polyfluorinated substances (PFASs): Environmental challenges. Curr. Opin.

411

Colloid Interface Sci. 2015, 20 (3), 192-212.

24


Page 24 of 30

Page 25 of 30


412

(11) Olsen, G. W.; Lange, C. C.; Ellefson, M. E.; Mair, D. C.; Church, T. R.; Goldberg, C. L.; Herron, R. M.;

413

Medhdizadehkashi, Z.; Nobiletti, J. B.; Rios, J. A. Temporal trends of perfluoroalkyl concentrations in American

414

Red Cross adult blood donors, 2000–2010. Environ. Sci. Technol. 2012, 46 (11), 6330-6338.

415

(12) Wu, M.; Sun, R.; Wang, M.; Liang, H.; Ma, S.; Han, T.; Xia, X.; Ma, J.; Tang, L.; Sun, Y. Analysis of

416

perfluorinated compounds in human serum from the general population in Shanghai by liquid chromatography-

417

tandem mass spectrometry (LC-MS/MS). Chemosphere 2017, 168, 100-105.

418

(13) Olsen, G. W.; Burris, J. M.; Ehresman, D. J.; Froehlich, J. W.; Seacat, A. M.; Butenhoff, J. L.; Zobel, L. R.

419

Half-life of serum elimination of perfluorooctanesulfonate, perfluorohexanesulfonate, and perfluorooctanoate in

420

retired fluorochemical production workers. Environ. Health Perspect. 2007, 115 (9), 1298-1305.

421

(14) Abbott, B. D.; Wolf, C. J.; Das, K. P.; Zehr, R. D.; Schmid, J. E.; Lindstrom, A. B.; Strynar, M. J.; Lau, C.

422

Developmental toxicity of perfluorooctane sulfonate (PFOS) is not dependent on expression of peroxisome

423

proliferator activated receptor-alpha (PPARα) in the mouse. Reprod. Toxicol. 2009, 27 (3-4), 258-265.

424

(15) Ankley, G. T.; Kuehl, D. W.; Kahl, M. D.; Jensen, K. M.; Linnum, A.; Leino, R. L.; Villeneuve, D. A.

425

Reproductive and developmental toxicity and bioconcentration of perfluorooctanesulfonate in a partial life‐cycle test

426

with the fathead minnow (Pimephales promelas). Environ. Toxicol. Chem. 2005, 24 (9), 2316-2324.

427

(16) Austin, M. E.; Kasturi, B. S.; Barber, M.; Kannan, K.; MohanKumar, P. S.; MohanKumar, S. M.

428

Neuroendocrine effects of perfluorooctane sulfonate in rats. Environ. Health Perspect. 2003, 111 (12), 1485-1489.

429

(17) Fair, P. A.; Driscoll, E.; Mollenhauer, M. A.; Bradshaw, S. G.; Yun, S. H.; Kannan, K.; Bossart, G. D.; Keil, D.

430

E.; Peden-Adams, M. M. Effects of environmentally-relevant levels of perfluorooctane sulfonate on clinical

431

parameters and immunological functions in B6C3F1 mice. J. Immunotoxicol. 2011, 8 (1), 17-29.

432

(18) Macon, M. B.; Villanueva, L. R.; Tatum-Gibbs, K.; Zehr, R. D.; Strynar, M. J.; Stanko, J. P.; White, S. S.;

433

Helfant, L.; Fenton, S. E. Prenatal perfluorooctanoic acid exposure in CD-1 mice: low-dose developmental effects

434

and internal dosimetry. Toxicol. Sci. 2011, 122 (1), 134-145.

435

(19) Shi, G.; Cui, Q.; Pan, Y.; Sheng, N.; Sun, S.; Guo, Y.; Dai, J. 6: 2 Chlorinated polyfluorinated ether sulfonate, a

436

PFOS alternative, induces embryotoxicity and disrupts cardiac development in zebrafish embryos. Aquat. Toxicol.

437

2017, 185, 67-75.

438

(20) Wan, H.; Zhao, Y.; Wei, X.; Hui, K.; Giesy, J.; Wong, C. K. PFOS-induced hepatic steatosis, the mechanistic

439

actions on β-oxidation and lipid transport. Biochim. Biophys. Acta, Gen. Subj. 2012, 1820 (7), 1092-1101.

25



440

(21) Wang, M.; Chen, J.; Lin, K.; Chen, Y.; Hu, W.; Tanguay, R. L.; Huang, C.; Dong, Q. Chronic zebrafish PFOS

441

exposure alters sex ratio and maternal related effects in F1 offspring. Environ. Toxicol. Chem. 2011, 30 (9), 2073-

442

2080.

443

(22) Vestergren, R.; Cousins, I. T. Tracking the pathways of human exposure to perfluorocarboxylates. Environ. Sci.

444

Technol. 2009, 43 (15), 5565-5575.

445

(23) Wang, Z.; Cousins, I. T.; Scheringer, M.; Hungerbühler, K. Fluorinated alternatives to long-chain

446

perfluoroalkyl carboxylic acids (PFCAs), perfluoroalkane sulfonic acids (PFSAs) and their potential precursors.

447

Environ. Int. 2013, 60, 242-248.

448

(24) Wang, Z.; Cousins, I. T.; Scheringer, M.; Hungerbuehler, K. Hazard assessment of fluorinated alternatives to

449

long-chain perfluoroalkyl acids (PFAAs) and their precursors: status quo, ongoing challenges and possible solutions.

450

Environ. Int. 2015, 75, 172-179.

451

(25) D’Agostino, L. A.; Mabury, S. A. Identification of novel fluorinated surfactants in aqueous film forming foams

452

and commercial surfactant concentrates. Environ. Sci. Technol. 2013, 48 (1), 121-129.

453

(26) Wang, S.; Huang, J.; Yang, Y.; Hui, Y.; Ge, Y.; Larssen, T.; Yu, G.; Deng, S.; Wang, B.; Harman, C. First

454

report of a Chinese PFOS alternative overlooked for 30 years: its toxicity, persistence, and presence in the

455

environment. Environ. Sci. Technol. 2013, 47 (18), 10163-10170.

456

(27) Zhao, P.; Xia, X.; Dong, J.; Xia, N.; Jiang, X.; Li, Y.; Zhu, Y. Short-and long-chain perfluoroalkyl substances

457

in the water, suspended particulate matter, and surface sediment of a turbid river. Sci. Total Environ. 2016, 568, 57-

458

65.

459

(28) Wen, W.; Xia, X.; Hu, D.; Zhou, D.; Wang, H.; Zhai, Y.; Lin, H. Long-Chain Perfluoroalkyl acids (PFAAs)

460

Affect the Bioconcentration and Tissue Distribution of Short-Chain PFAAs in Zebrafish (Danio rerio). Environ. Sci.

461

Technol. 2017, 51 (21), 12358-12368.

462

(29) Lam, J. C.; Lyu, J.; Kwok, K. Y.; Lam, P. K. Perfluoroalkyl substances (PFASs) in marine mammals from the

463

South China Sea and their temporal changes 2002–2014: Concern for alternatives of PFOS? Environ. Sci. Technol.

464

2016, 50 (13), 6728-6736.

465

(30) Shi, Y.; Vestergren, R.; Zhou, Z.; Song, X.; Xu, L.; Liang, Y.; Cai, Y. Tissue distribution and whole body

466

burden of the chlorinated polyfluoroalkyl ether sulfonic acid F-53B in crucian carp (Carassius carassius): Evidence

467

for a highly bioaccumulative contaminant of emerging concern. Environ. Sci. Technol. 2015, 49 (24), 14156-14165.

26


Page 26 of 30

Page 27 of 30


468

(31) Yao, Y.; Zhao, Y.; Sun, H.; Chang, S.; Zhu, L.; Alder, A. C.; Kannan, K. Per-and Polyfluoroalkyl Substances

469

(PFASs) in Indoor Air and Dust from Homes and Various Microenvironments in China: Implications for Human

470

Exposure. Environ. Sci. Technol. 2018, 52 (5), 3156-3166.

471

(32) Yeung, L. W.; Stadey, C.; Mabury, S. A. Simultaneous analysis of perfluoroalkyl and polyfluoroalkyl

472

substances including ultrashort-chain C2 and C3 compounds in rain and river water samples by ultra performance

473

convergence chromatography. J. Chromatogr. 2017, 1522, 78-85.

474

(33) Sun, M.; Arevalo, E.; Strynar, M.; Lindstrom, A.; Richardson, M.; Kearns, B.; Pickett, A.; Smith, C.; Knappe,

475

D. R. Legacy and emerging perfluoroalkyl substances are important drinking water contaminants in the Cape Fear

476

River Watershed of North Carolina. Environ. Sci. Technol. Lett. 2016, 3 (12), 415-419.

477

(34) Strynar, M.; Dagnino, S.; McMahen, R.; Liang, S.; Lindstrom, A.; Andersen, E.; McMillan, L.; Thurman, M.;

478

Ferrer, I.; Ball, C. Identification of novel perfluoroalkyl ether carboxylic acids (PFECAs) and sulfonic acids

479

(PFESAs) in natural waters using accurate mass time-of-flight mass spectrometry (TOFMS). Environ. Sci. Technol.

480

2015, 49 (19), 11622-11630.

481

(35) Ng, C. A.; Hungerbühler, K. Bioconcentration of perfluorinated alkyl acids: how important is specific binding?

482

Environ. Sci. Technol. 2013, 47 (13), 7214-7223.

483

(36) Cheng, W.; Ng, C. A. A Permeability-Limited Physiologically Based Pharmacokinetic (PBPK) Model for

484

Perfluorooctanoic acid (PFOA) in Male Rats. Environ. Sci. Technol. 2017, 51 (17), 9930-9939.

485

(37) Ng, C. A.; Hungerbühler, K. Bioaccumulation of perfluorinated alkyl acids: observations and models. Environ.

486

Sci. Technol. 2014, 48 (9), 4637-4648.

487

(38) Miller III, B. R.; McGee Jr, T. D.; Swails, J. M.; Homeyer, N.; Gohlke, H.; Roitberg, A. E. MMPBSA. py: an

488

efficient program for end-state free energy calculations. J. Chem. Theory Comput. 2012, 8 (9), 3314-3321.

489

(39) Genheden, S.; Ryde, U. The MM/PBSA and MM/GBSA methods to estimate ligand-binding affinities. Expert

490

Opin. Drug Discovery 2015, 10 (5), 449-461.

491

(40) Chen, F.; Liu, H.; Sun, H.; Pan, P.; Li, Y.; Li, D.; Hou, T. Assessing the performance of the MM/PBSA and

492

MM/GBSA methods. 6. Capability to predict protein–protein binding free energies and re-rank binding poses

493

generated by protein–protein docking. Phys. Chem. Chem. Phys. 2016, 18 (32), 22129-22139.

27



494

(41) Hou, T.; Wang, J.; Li, Y.; Wang, W. Assessing the performance of the MM/PBSA and MM/GBSA methods. 1.

495

The accuracy of binding free energy calculations based on molecular dynamics simulations. J. Chem. Inf. Model.

496

2010, 51 (1), 69-82.

497

(42) Rastelli, G.; Rio, A. D.; Degliesposti, G.; Sgobba, M. Fast and accurate predictions of binding free energies

498

using MM‐PBSA and MM‐GBSA. J. Comput. Chem. 2010, 31 (4), 797-810.

499

(43) Sun, H.; Li, Y.; Tian, S.; Xu, L.; Hou, T. Assessing the performance of MM/PBSA and MM/GBSA methods. 4.

500

Accuracies of MM/PBSA and MM/GBSA methodologies evaluated by various simulation protocols using PDBbind

501

data set. Phys. Chem. Chem. Phys. 2014, 16 (31), 16719-16729.

502

(44) Woodcroft, M. W.; Ellis, D. A.; Rafferty, S. P.; Burns, D. C.; March, R. E.; Stock, N. L.; Trumpour, K. S.; Yee,

503

J.; Munro, K. Experimental characterization of the mechanism of perfluorocarboxylic acids' liver protein

504

bioaccumulation: The key role of the neutral species. Environ. Toxicol. Chem. 2010, 29 (8), 1669-1677.

505

(45) Zhang, L.; Ren, X.-M.; Guo, L.-H. Structure-based investigation on the interaction of perfluorinated

506

compounds with human liver fatty acid binding protein. Environ. Sci. Technol. 2013, 47 (19), 11293-11301.

507

(46) Sheng, N.; Cui, R.; Wang, J.; Guo, Y.; Wang, J.; Dai, J. Cytotoxicity of novel fluorinated alternatives to long-

508

chain perfluoroalkyl substances to human liver cell line and their binding capacity to human liver fatty acid binding

509

protein. Arch. Toxicol. 2018, 92 (1), 359-369.

510

(47) Morris, G. M.; Lim-Wilby, M., Molecular docking. Methods Mol. Biol. 2008, 443, 365-382.

511

(48) Sharma, A.; Sharma, A. Fatty acid induced remodeling within the human liver fatty acid-binding protein. J.

512

Biol. Chem. 2011, 286 (36), 31924-31928.

513

(49) Thompson, J.; Winter, N.; Terwey, D.; Bratt, J.; Banaszak, L. The Crystal Structure of the Liver Fatty Acid-

514

binding Protein A COMPLEX WITH TWO BOUND OLEATES. J. Biol. Chem. 1997, 272 (11), 7140-7150.

515

(50) Ng, C. A.; Hungerbuehler, K. Exploring the use of molecular docking to identify bioaccumulative

516

perfluorinated alkyl acids (PFAAs). Environ. Sci. Technol. 2015, 49 (20), 12306-12314.

517

(51) Zhang, J.; Begum, A.; Brännström, K.; Grundström, C.; Iakovleva, I.; Olofsson, A.; Sauer-Eriksson, A. E.;

518

Andersson, P. L. Structure-based virtual screening protocol for in silico identification of potential thyroid disrupting

519

chemicals targeting transthyretin. Environ. Sci. Technol. 2016, 50 (21), 11984-11993.

520

(52) Luo, Z.; Shi, X.; Hu, Q.; Zhao, B.; Huang, M. Structural evidence of perfluorooctane sulfonate transport by

521

human serum albumin. Chem. Res. Toxicol. 2012, 25 (5), 990-992.

28


Page 28 of 30

Page 29 of 30


522

(53) Hanwell, M. D.; Curtis, D. E.; Lonie, D. C.; Vandermeersch, T.; Zurek, E.; Hutchison, G. R. Avogadro: an

523

advanced semantic chemical editor, visualization, and analysis platform. J. Cheminform. 2012, 4 (1), 17.

524

(54) Trott, O.; Olson, A. J. AutoDock Vina: improving the speed and accuracy of docking with a new scoring

525

function, efficient optimization, and multithreading. J. Comput. Chem. 2010, 31 (2), 455-461.

526

(55) Morris, G. M.; Huey, R.; Lindstrom, W.; Sanner, M. F.; Belew, R. K.; Goodsell, D. S.; Olson, A. J. AutoDock4

527

and AutoDockTools4: Automated docking with selective receptor flexibility. J. Comput. Chem. 2009, 30 (16), 2785-

528

2791.

529

(56) Prevedouros, K.; Cousins, I. T.; Buck, R. C.; Korzeniowski, S. H. Sources, fate and transport of

530

perfluorocarboxylates. Environ. Sci. Technol. 2006, 40 (1), 32-44.

531

(57) Gomis, M. I.; Wang, Z.; Scheringer, M.; Cousins, I. T. A modeling assessment of the physicochemical

532

properties and environmental fate of emerging and novel per-and polyfluoroalkyl substances. Sci. Total Environ.

533

2015, 505, 981-991.

534

(58) Schrodinger, LLC, The PyMOL Molecular Graphics System, Version 1.8. 2015.

535

(59) Case, D. A.; Darden, T. A.; Cheatham, T. E., III; Simmerling, C. L.; Wang, J.; Duke, R. E.; Luo, R.; Merz, K.

536

M.; Wang, B.; Pearlman, D. A.; Crowley, M.; Brozell, S.; Tsui, V.; Gohlke, H.; Mongan, J.; Hornak, V.; Cui, G.;

537

Beroza, P.; Schafmeister, C.; Caldwell, J. W.; Ross, W. S.; Kollman, P. A. Amber 14. 2014, University of

538

California, San Francisco.

539

(60) Hornak, V.; Abel, R.; Okur, A.; Strockbine, B.; Roitberg, A.; Simmerling, C. Comparison of multiple Amber

540

force fields and development of improved protein backbone parameters. Proteins 2006, 65 (3), 712-725.

541

(61) Wang, J.; Wolf, R. M.; Caldwell, J. W.; Kollman, P. A.; Case, D. A. Development and testing of a general

542

amber force field. J. Comput. Chem.2004, 25 (9), 1157-1174.

543

(62) Jakalian, A.; Jack, D. B.; Bayly, C. I. Fast, efficient generation of high‐quality atomic charges. AM1‐BCC

544

model: II. Parameterization and validation. J. Comput. Chem.2002, 23 (16), 1623-1641.

545

(63) Darden, T.; York, D.; Pedersen, L. Particle mesh Ewald: An N⋅ log (N) method for Ewald sums in large

546

systems. J. Chem. Phys. 1993, 98 (12), 10089-10092.

547

(64) Le Grand, S.; Götz, A. W.; Walker, R. C. SPFP: Speed without compromise—A mixed precision model for

548

GPU accelerated molecular dynamics simulations. Comput. Phys. Commun. 2013, 184 (2), 374-380.

29



549

(65) Case, D. A.; Cheatham, T. E.; Darden, T.; Gohlke, H.; Luo, R.; Merz, K. M.; Onufriev, A.; Simmerling, C.;

550

Wang, B.; Woods, R. J. The Amber biomolecular simulation programs. J. Comput. Chem.2005, 26 (16), 1668-1688.

551

(66) Genheden, S.; Ryde, U. How to obtain statistically converged MM/GBSA results. J. Comput. Chem. 2010, 31

552

(4), 837-846.

553

(67) Kollman, P. A.; Massova, I.; Reyes, C.; Kuhn, B.; Huo, S.; Chong, L.; Lee, M.; Lee, T.; Duan, Y.; Wang, W.

554

Calculating structures and free energies of complex molecules: combining molecular mechanics and continuum

555

models. Acc. Chem. Res. 2000, 33 (12), 889-897.

556

(68) Srinivasan, J.; Cheatham, T. E.; Cieplak, P.; Kollman, P. A.; Case, D. A. Continuum solvent studies of the

557

stability of DNA, RNA, and phosphoramidate− DNA helices. J. Am. Chem. Soc. 1998, 120 (37), 9401-9409.

558

(69) Homeyer, N.; Gohlke, H. Free energy calculations by the molecular mechanics Poisson− Boltzmann surface

559

area method. Mol. Inform. 2012, 31 (2), 114-122.

560

(70) Hadden, J. A.; Tessier, M. B.; Fadda, E.; Woods, R. J. Calculating Binding Free Energies for Protein–

561

Carbohydrate Complexes. Methods Mol. Biol. 2015, 431-465.

562

(71) Poland, D. Statistical Mechanics (McQuarrie, Donald A.). J. Chem. Educ. 1977, 54 (10), A428.

563

(72) Gohlke, H.; Kiel, C.; Case, D. A. Insights into protein–protein binding by binding free energy calculation and

564

free energy decomposition for the Ras–Raf and Ras–RalGDS complexes. J. Mol. Biol. 2003, 330 (4), 891-913.

565

(73) Metz, A.; Pfleger, C.; Kopitz, H.; Pfeiffer-Marek, S.; Baringhaus, K.-H.; Gohlke, H. Hot spots and transient

566

pockets: predicting the determinants of small-molecule binding to a protein–protein interface. J. Chem. Inf. Model.

567

2011, 52 (1), 120-133.

568

(74) Caldwell, G. W.; Yan, Z., Isothermal titration calorimetry characterization of drug-binding energetics to blood

569

proteins. Methods Pharmacol. Toxicol. 2004, 123-149.

570

(75) Kastritis, P. L.; Bonvin, A. M. On the binding affinity of macromolecular interactions: daring to ask why

571

proteins interact. J R Soc Interface 2013, 10 (79), 20120835.

572

30


Page 30 of 30

and Polyfluoroalkyl Substances (PFASs) - ACS Publications

Recommend Documents