Determination of Energies and Sites of Binding of ... - ACS Publications

Oct 28, 2010 - the estimation of the free binding energies adopting WHAM-umbrella sampling and semiempirical methodologies. The binding sites so ...
0 downloads 0 Views 6MB Size
Download PDF
14860

J. Phys. Chem. B 2010, 114, 14860–14874

Determination of Energies and Sites of Binding of PFOA and PFOS to Human Serum Albumin Matteo Salvalaglio, Isabella Muscionico, and Carlo Cavallotti* Department of Chimica, Materiali e Ingegneria Chimica “G. Natta”, Politecnico di Milano, Via Mancinelli 7, 20131 Milano, Italy ReceiVed: July 15, 2010; ReVised Manuscript ReceiVed: October 4, 2010

Structure and energies of the binding sites of perfluorooctanoic acid (PFOA) and perfluorooctane sulfonate (PFOS) to human serum albumin (HSA) were determined through molecular modeling. The calculations consisted of a compound approach based on docking, followed by molecular dynamics simulations and by the estimation of the free binding energies adopting WHAM-umbrella sampling and semiempirical methodologies. The binding sites so determined are common either to known HSA fatty acids sites or to other HSA sites known to bind to pharmaceutical compounds such as warfarin, thyroxine, indole, and benzodiazepin. Among the PFOA binding sites, five have interaction energies in excess of -6 kcal/mol, which become nine for PFOS. The calculated binding free energy of PFOA to the Trp 214 binding site is the highest among the PFOA complexes, -8.0 kcal/mol, in good agreement with literature experimental data. The PFOS binding site with the highest energy, -8.8 kcal/mol, is located near the Trp 214 binding site, thus partially affecting its activity. The maximum number of ligands that can be bound to HSA is 9 for PFOA and 11 for PFOS. The calculated data were adopted to predict the level of complexation of HSA as a function of the concentration of PFOA and PFOS found in human blood for different levels of exposition. The analysis of the factors contributing to the complex binding energy permitted to outline a set of guidelines for the rational design of alternative fluorinated surfactants with a lower bioaccumulation potential. 1. Introduction Fluorinated surfactants such as perfluorooctanoic acid (PFOA) and perfluorooctane sulfonate (PFOS) are widely diffused chemical additives largely employed in different fields of the process industry. Their most common use is as surfactants in the emulsion polymerization of perfluoro polymers such as poly(tetrafluoroethylene) or fluorinated elastomers, though they are also widely adopted in the textile, paper, and leather industry, in the production of food packaging, and in electronic materials manufacturing.1,2 Another application of these compounds is related to the formulation of antifire foams. Their most innovative use is in the biological field, either in protein extraction processes or in the selective encapsulation of controlled-release therapeutic agents.3 The chemical inertia that characterizes the perfluoroalkyl chain, which is the source of many technologically desirable properties, is also responsible for their slow depletion kinetics and lead them to be prone to accumulation in living organisms.1,4 Because of this, the bioaccumulation potential of perfluoro surfactants has been object of scientific investigation for several decades. Nowadays, the accumulation of PFOA and PFOS in a variety of living organisms, comprising water mammals, fishes, birds,4-11 and humans,12,13 has been incontrovertibly demonstrated. Recently, Conder and co-workers14 have evaluated the potential bioaccumulation of different perfluoro surfactants, investigating its dependence with respect to alkyl chain length and number of sulfonic groups. It was found that the accumulation of the perfluorinated surfactants takes place preferentially in the blood and in the liver, highlighting the proteinophilic nature of these compounds. For what concerns the half-life time * Corresponding author. Tel.: +39-02-23993176. Fax.: +39 02 23993176. E-mail: [email protected].

of PFOA and PFOS in the human body, there is not yet a consensus among the published data: 0.9-4.4 years,15 4 years,16,17 and 3.5 years18,19 are the values measured for PFOA, and 8.7 years20 and 4.8 years21 are those obtained for PFOS. It can, however, be considered as established the slower depletion of PFOS with respect to PFOA. The toxicity of perfluorinated surfactants has been investigated through in vitro and in vivo experiments.22 It was observed that the exposition to perfluorinated surfactants determines hepatomegaly, hepatocellular hypertrophy, and vacuolization both in rats23-25 and in primates.26-28 PFOA has been recently recognized as a carcinogenic agent for animals,16 while PFOS has been proved to be correlated to the insurgence of hepatocellular and thyroid adenomas.29 The health issues related to the toxic effects of PFOA and PFOS have driven the US and EU institutions to restrict their production and diffusion. The US government has produced a restrictive regulation that has almost completely inhibited the PFOS production since 2002 (U.S. EPA SNUR), while the European parliament has produced a directive in 2006 that regulates the amount of PFOS allowed in preparations (0.005% by mass) or in finite products (0.1% by mass).30 Regarding PFOA, the US EPA has recently introduced a regulation plan for its complete retirement from the market before 2015,31 while the EU is currently assessing the risks associated with its use, encouraging the adoption of alternative compounds in the manufacturing processes. This scenario points out the need for a rational development of PFOS/PFOA analogues with a lower bioaccumulation potential. In this framework, the focus of this work is the study of the interaction between the two most adopted perfluorinated surfactants, PFOA and PFOS, and human serum albumin (HSA), the protein present in the highest concentration in human blood. In fact, though an investigation of the effects of PFOS and

10.1021/jp106584b  2010 American Chemical Society Published on Web 10/28/2010

Binding of PFOA and PFOS to Human Serum Albumin PFOA on the lipid membranes has been recently published,32 a rational description of the proteinophilic interaction that determines their accumulation in the organism is still missing. Since HSA is one of the main carriers for pharmaceuticals in the human organism, several experimental studies have been conduced in order to clarify its binding properties.6,33-36 HSA is able to bind chemicals with both its hydrophobic and polar groups. In addition, it is also reported its capability to form covalent adducts between its cysteine residue not involved in disulfide bonds and thiol groups. Both primary and secondary bonds can be stereospecific. HSA also shows some esteraselike enzymatic activity.37 Two main reversible binding sites, able to bind a large variety of compounds, have been reported in the literature for HSA. The first binding site (BSI), formerly known as “warfarin binding site”, is located in the IIA subdomain and is constituted by a hydrophobic pocket surrounded by positively charged amino acids. Its flexibility is able to justify the capability to bind a large number of molecules such as bicarbossilic acids, negatively charged heterocycles, long-chain fatty acids, indole derivatives, and aldeids.33 Also some metals, like Cu and Ni,37,38 can bind to this site. The second relevant binding site (BSII) is known as “indole-benzodiazepine binding site” and is located in the IIIA subdomain. BSII is smaller and more rigid than BSI, with binding properties mostly influenced by the Arg 410 and Tyr 411 residues. This site binds preferentially aromatic carboxylic acids though it exhibits a high selectivity also for anions of long-chain fatty acids (C8). In the literature, other secondary binding sites, strongly liganddependent, have been reported. Of particular interest are the complexes found by Bhattacharya and co-workers for fatty acids.34 The HSA binding sites known to the literature are summarized in Figure 1. Several experimental investigations of the complexes formed by PFOA or PFOS with HSA have been recently published. Han et al.39,40 have determined, through the combination of different experimental techniques (ESI MS, NMR, desalting columns), the average number of PFOA molecules complexed by HSA and RSA (rat serum albumin) in different conditions and the related mean affinity constant. They were also able to demonstrate that the accumulation of PFOA in the blood is due to the formation of specific interactions with proteins. More recently, Chen and Guo41 measured the binding constants of both PFOA and PFOS to a specific HSA binding site through fluorescence spectroscopy. Finally, Li et al.42 found that the fluorinated alkyl chain contributes significantly to the binding strength of surfactants to HSA. The interaction of PFOA with HSA was studied also by Wu and co-workers employing fluorospectrometry, isothermal titration calorimetry, and circular dichroism; these experimental studies highlighted that the HSA-PFOA interaction follows a two-step Langmuir sequence and that the favorite binding site is located in the protein hydrophobic core.43 A similar array of experimental techniques was employed by Zhang et al. in order to assess the binding of PFOS to HSA; their findings confirmed that PFOS bioaccumulates more than PFOA and acts as a mild denaturating agent, producing an inhibition of the natural transport function of HSA in the blood. The partial denaturation of the protein receptor may be in part responsible for the extremely high complexation ratio (45 PFOS molecules per HSA) found in this work.44 The aim of this work is to determine which are the binding sites of PFOA and PFOS on HSA and to classify them on the basis of their interaction energy. The adopted computational approach, described in section 2, consists of molecular dynamic (MD) simulations performed with a force field in part developed

J. Phys. Chem. B, Vol. 114, No. 46, 2010 14861

Figure 1. Summary of HSA binding sites. The labels indicate the numeration of the fatty acids binding sites of Bhattacharya and collaborators.34 (a) Blue surfaces correspond to (PDB id in parentheses): decanoic acid (1E7E), dodecanoic acid (1E7F), tetradecanoic acid (1E7G, 1HK4, 1HK5), hexadecanoic acid (1E7H), octadecanoic acid (1E7I), and myristate (2BXI, 2BXK, 2BXL, 2BXM, 2BXN, 2BXO, 2BXP, 2BXQ). (b) Red surfaces correspond to the binding sites of nonfatty acid ligands reported by Ghuman et al.:35 azapropazone (2BX8, 2BXI, 2BXK), D 3-carboxy-4-methyl-5-propyl-2-furanpropanoic acid (2BXA), oxyphenbutazone (2BXB), phenilbutazone (2BXC, 2BXP, 2BXQ), warfarin (2BXD), diflunisal (2BXE), diazepam (2BXF), ibuprofen (2BXG), indoxyl sulfate (2BXH), indomethacin (2BXK, 2BXM, 2BXQ), 3,5-2 diiodosalicilic acid (2BXL), iodipamide (2BXN), oxyphenbutazone (2BXO), 4Z,15E-bilirubin-IX-2alpha (2VUE), fusidic acid (2VUF).

in this work. The energetic and structural analysis of the results of the simulations is reported in section 3, together with an extensive comparison with literature data on HSA fatty acid binding sites. The calculated data are finally compared with experimental HSA complexation data and used to predict HSA complexation levels in blood for different concentrations of perfluorinated surfactants. 2. Theoretical Methods The interactions between fluorinated surfactants and HSA have been investigated by employing a simulation strategy that combines docking and molecular dynamics simulations in a rational workflow. Since the major structural characteristic of the surfactants under investigation is their peculiar helical structure, which cannot be reproduced using standard force

14862

J. Phys. Chem. B, Vol. 114, No. 46, 2010

Salvalaglio et al.

Figure 2. Schematic representation of the adopted molecular modeling workflow.

fields, the determination and implementation of tailored force field parameters arose as a necessity. This task was carried out in part through the implementation and validation of an Amber compatible version45 of a force field developed for perfluoroalkyl compounds46 and in part through the ex novo development of atomic partial charges and torsional potentials able to correctly describe the structural features of the surfactants polar moieties. The free energy evaluation for the protein-surfactant 1:1 complexes was performed following a multilevel strategy consisting of three different computational approaches of growing accuracy: the molecular mechanics generalized Born surface area (MMGBSA) approach developed by Kollman and co-workers,47 based on the postprocessing of MD simulations using the GB implicit solvation model, the linear interaction energy (LIE) approach developed by Aqvist,48,49 which assumes a linear scaling between free interaction energy and electrostatic and van der Waals interaction energies, and the application of the weighted histogram analysis method (WHAM) to the postprocessing of umbrella sampling (US) simulations. Their role and employment are explained in the following sections, while the adopted simulation and analysis strategy is summarized in Figure 2. 2.1. Force Field. Perfluorinated carbon compounds are characterized by several peculiar properties, such as the capability to form halogen bonds50 and a general tendency to organize their structures in helical shapes. This structural behavior was first evidenced through X-ray crystallography by Bunn and Howells51 for poly(tetrafluoroethylene) and recently confirmed by several theoretical studies performed on n-perfluoroalkanes.46,52 In the present work, such spatial organization was confirmed also for PFOA and PFOS by quantum chemistry structural optimization calculations performed in implicit water at the B3LYP/6-31G(d,p) level using the implicit integral equation formalism polarized continuum model (IEFPCM) (see Figure S1-S3 and Tables S1-S3 in the Supporting Information).53 In order to reproduce such a complex geometry with a molecular mechanics force field, an appropriate set of parameters was implemented in an Amber ff03 compatible formalism. The

starting point for the development of the tailored force field was a recently published set of parameters46 that were determined in order to study the origin of helicity in perfluorinated n-alkanes. The literature force field was converted to the functional form implemented in the Amber force field as summarized in Tables S7 and S8 in the Supporting Information. The torsional potentials of sulfur and oxygen on a CxFy chain are missing in the literature and were therefore determined in this work through a least-squares fit of the rotational PES computed at the B3LYP/6-31G(d,p) level following the same procedure described in the literature45,54-56 for the implementation of Amber-compatible parameters. The calculations of the full torsional potentials were performed through the constrained geometry optimization of the dihedral PES scanned at 20° steps. Each dihedral potential was calculated by subtracting to the total force field energy determined by setting to zero the potential under investigation the corresponding DFT rotational energy barrier. Since the target molecules of the present study (PFOA and PFOS) have several unknown torsional potentials each, at a first stage the potentials were determined for a training set of small molecules (C2F5CO2- and C3F9CO2- for PFOA; C4F9SO3- and CF3SO3- for PFOS), which were then adopted as first guess for the estimation of the effective torsional potentials of PFOA and PFOS. The torsional potentials determined employing the described procedure were then fitted to a cosine sum expression and implemented in Amber-like parameters sets.57,58 In order to create a reliable add-on module to be used with the Amber force field, two new atom types were introduced in the library for carbon and fluorine, named FG and CF, respectively. The two frcmod modules adopted for the simulations, PFOA.frcmod and PFOS.frcmod, are reported in section S5 and S6 in the Supporting Information. The atomic charges of PFOA and PFOS were calculated employing the RESP56,59 formalism. Electrostatic potentials were determined at the B3LYP/6-31G(d,p) level while the RESP fitting was performed with the RESP utility of the Amber package.60 The RESP atomic charges so calculated are reported in Table S6 in

Binding of PFOA and PFOS to Human Serum Albumin the Supporting Information. All the DFT calculations were performed using the Gaussian 03 computational suite.53 2.2. Docking. Docking was adopted to determine at a first level of approximation a list of the possible binding sites of HSA for perfluorinated surfactants. HSA is a globular protein constituted by three domains, each divided in two subdomains. The predominant secondary structure observed in the folded wild-type HSA is the R helix. The three domains are interconnected by flexible loops, while the overall structure is crosslinked through 17 disulfide bridges that involve 34 of its 35 cysteine residues. According to the experimental investigation performed by Leggio and co-workers61 that has shed light on one of the unfolded states of HSA observed at low pH,61 the globular heart-shaped structure is the HSA’s most stable tertiary structure in a pH range comprised between 7 and 8 (blood pH) and was therefore adopted as reference HSA structure for docking (PDB id 1N5U).62 Three different grid volumes were considered, each including one of the three HSA domains. This choice was determined by the need to maintain a small grid spacing for the search box and at the same time to perform the docking calculation without neglecting any portion of the protein surface. The main carbon chain of the two fluorinated surfactants was set as completely rotatable, leaving the carbon chain dihedral angles as active degrees of freedom to the conformational minimum energy search algorithm. All docking calculations were performed with the Autodock 3.0 package63 adopting three different algorithms: Lamarckian genetic, simulated annealing, and local search. For each algorithm 10 structures were collected, resulting in a total of 30 structures for each search box. This procedure led to the identification of 90 potential structures for each of the two surfactants investigated. A visual inspection refinement was carried out in order to organize the identified structures in homogeneous groups, which were defined on the basis of the location of the binding site on the protein surface. This organized the initial 180 structures into 24 and 30 groups of potential 1:1 complex structures for PFOA and PFOS, respectively. Each group was composed of up to 10 structures. In order to choose a representative complex structure for each group, a single structure MMGBSA64 calculation was performed. The lowest energy structure of each group was chosen as starting structure for a MD simulation. 2.3. Molecular Dynamics Simulations. MD simulations were used to refine the structures and evaluate the energies of the complexes which first guess structures were determined through docking. All the simulations were performed in explicit TIP3P water in a cubic box using a nonbonded cutoff of 12 Å. Periodic boundary conditions were applied and long-range electrostatic interactions were evaluated using the particle mesh Ewald method.64 MD simulations were performed through a sequence of calculations designed in order to produce an equilibrated starting system to be used as input for a production run. As a first step, the solvent molecules added to the initial configuration were relaxed with a 2000-cycle minimization. At this stage, the solute was restrained with a harmonic potential of the form k(∆x)2 with a force constant k ) 500 kcal mol-1 Å-2, eliminating any nonphysical contact between the solute and the solvent. After that, the system was minimized without restraints for 1500 cycles. These two minimization steps were followed by a simulated annealing simulation of 20 ps at constant volume, used to raise the temperature of the system from 0 to 300 K. A weak restraint was imposed on the complex (k ) 10 kcal mol-1 Å-2) to avoid excessive structural fluctuations. The last step was a 100 ps run at constant pressure, which helped to relax the water density. The configuration of the

J. Phys. Chem. B, Vol. 114, No. 46, 2010 14863 system so determined was used as starting structure for the production MD run, performed at constant pressure and temperature (1 atm, 300 K) for 2 ns. The Shake algorithm was used to constrain all the covalent bonds involving hydrogen, which allowed to use a time step of 2.0 fs.64 Structures to be used for successive analysis were extracted every 500 fs. Two additional 8 ns simulations of PFOA and PFOS in explicit water were carried out in order to obtain the ensemble-averaged interaction potential of the surfactants with water following the same protocol described above. All MD simulations were performed with the Amber 8 computational suite.60 All the structure pictures reported in this work were produced using PyMol 1.365 and VMD 1.8.2.66 The interaction energies between solute and environment were determined with the Anal program of the Amber 8 computational suite. 2.4. Energetic Analysis. The energetic analysis of the protein-surfactant complexes was performed at different levels of theory adopting a hierarchical strategy. In principle, the best approach to determine free interaction energies from MD simulations would be the thermodynamic integration (TI) or the free energy perturbation (FEP) of the molecular Hamiltonian along a given reaction path. These approaches are, however, too demanding from a computational standpoint in view of the large number of possible binding sites to be investigated. A reasonable compromise between computational load and quality of the results was therefore found by parametrizing a fast and flexible semiempirical method (LIE), adopting energies determined with a rigorous statistical thermodynamics evaluation of the potential of mean force (PMF) based on US simulations postprocessed with the WHAM algorithm. In particular, the parametrization of LIE was performed on the free binding energy of complexation of the Trp 214 binding site of HSA, for which both experimental and computational data are available. 2.4.1. Potential of Mean Force. The potential of mean force (PMF) W(ξ) for a target HSA-PFOA binding site was calculated over a dissociation pathway ξ defined as the distance between the carboxylic carbon of PFOA (C7 according to the internal nomenclature given to the PFOA home-built library) and the nitrogen atom of the polyaromatic ring of the Trp 214 residue (NE1 atom according to the standard Amber nomenclature of the Trp residue) as

〈F(ξ)〉 [ 〈F(ξ*)〉 ]

W(ξ) ) W(ξ*) - kBT ln

(1)

The average energy distribution function F(ξ) was built using the US approach. This method consists in postprocessing an ensemble of MD simulations performed applying a harmonic biasing potential wi centered on the pivot atom of one of the dissociating moieties (C7 of PFOA in this case), which was moved along the dissociation coordinate for each simulation. This produces a series of biased probability distributions (one for each simulation window, defined as steps along ξ) that, once merged, gives the overall PMF. The PMF for each simulations window can be written as

〈F(ξ)〉 ] - w (ξ) + F [ 〈F(ξ*)〉

W(ξ) ) W(ξ*) - kBT ln

i

i

(2)

where Fi is the free energy associated with the introduction of the biasing potential in the ith window. The local biased probability distributions resulting from each US simulation

14864

J. Phys. Chem. B, Vol. 114, No. 46, 2010

Salvalaglio et al.

window were merged adopting the WHAM method, which was implemented here as defined by eqs 3 and 4 Nw

〈F(ξ)〉 )

Nw

∑ ni〈F(ξ)〉(i) × [ ∑ nje-[w (ξ)-F ]/k T]-1

(3)

∫ dξ e-w (ξ)/k T〈F(ξ)〉

(4)

j

i)1

e-Fi/kBT )

j

B

j)1

i

B

Equation 3 relates the total unbiased probability 〈F(ξ)〉 to the biased probability distributions 〈F(ξ)〉(i), while eq 4, which derives directly from eq 2, defines the free energy as a function of the dissociation coordinate for each window.67 Since the free energies Fi and the distribution function F(ξ) are interdependent, eqs 3 and 4 were solved using an iterative scheme. An initial Fi guess was thus employed to determine a first guess unbiased probability distribution through eq 3, which was then used to recalculate Fi using eq 4. The procedure is repeated until convergence. In the implemented algorithm, the convergence condition is considered satisfied when the difference between the Fi values of two successive steps is lower than an arbitrary threshold. The PMF calculated in the present work was determined adopting a threshold difference of 10-4 kcal/mol. An extensive series of umbrella simulations was performed in order to achieve a sufficiently dense sampling over the mentioned pathway. In order to restrain the sampling in each window, a harmonic potential k∆x2 was employed with a k value of 30 kcal mol-1 Å-2. The degree of freedom was mapped in the range 4.9-35 Å, with a spacing ranging between 0.2 and 0.5 Å between adjacent windows. The standard simulation time and the number of sample points for each window were 0.5 ns and 2500, respectively. A generalized Born (GB) model64 was used to account implicitly for the effect of the solvent. 2.4.2. Linear Interaction Energy. The LIE49,68 method was employed in order to determine efficiently the association free energies of the complexes found with the docking procedure and relaxed with the explicit water MD simulations. The LIE approach is based on the main assumption that the free energy of interaction of a molecule with the surrounding environment can be linearly correlated to both electrostatic and van der Waals interaction energies. This hypothesis permits the calculation of the free energy of complexation directly from the mean energy of interaction of the target ligand with the environment in a free and a bound state. The free state can be conveniently represented as a system in which the studied molecule (e.g., PFOA, PFOS) is surrounded only by the solvent, while the bound state is represented by the same molecule involved in the interaction with the receptor. The free energy of binding is then defined as

for organic molecules is 0.18. Since in this case the energetic analysis was carried out for compounds with a fluorinated apolar moiety, the R value is not known in the literature and was thus refitted over the free binding energy of PFOA with HSA determined from the PMF for the Trp 214 binding site. 2.4.3. Scatchard Analysis. The predictive capability of the adopted computational protocol was tested by comparing calculated and experimental Scatchard plots. The motivation of the seminal paper written by Scatchard in 1949 was to provide an answer to three questions concerning the nature of the interaction between proteins and ligands. The first two were “how many” ligands can be carried by a protein receptor and “how tightly bound” they are.69 The answer consisted of a simple model, whose pillars were the assumption that each binding site has the same average interaction energy with the same ligand and that the free complexation energy diminishes by a coefficient, determined through Debye-Huckel theory, which is proportional to the square of the charge of the ligand. A regression of the model parameters over experimental data allows to determine the average free interaction energy and the total number of binding sites. The theory was so successful that it is still routinely applied nowadays. In order to compare the binding free energies here calculated, which are site specific, with experimental data collected in the Scatchard plots, the original Scatchard equations have been modified as described in the following. This development was made possible by the availability of the answer to the third question posed by Scatchard, concerning “where” the ligand is located on the receptor protein surface, which was given by the results of the docking and MD calculations described in the previous sections. The complexation reactions involving a protein P and a generic ligand L can be described through a series of successive reactions as

P + L a P1 P + 2L a P2 l P + NL a PN where P1 represents all the possible 1:1 complexes, P2 the 2:1 complexes, and so on until PN is the concentration of the N:1 complexes, with N defined as the number of protein binding sites. The P1 concentration is related through thermodynamic equilibrium to the complexation reactions involving N possible binding sites as N

P1 ) PAL

∑ e-∆G /RT i

(6)

i)1

vdw ele ele 〈∆G〉 ) R(〈Vvdw bound〉 - 〈Vfree 〉) + β(〈Vbound〉 - 〈Vfree〉)

(5) where the Vele and VVdW terms are respectively electrostatic and the van der Waals interaction energies averaged on the MDgenerated ensemble of configurations that represent the given state (indicated as subscript). The R and β coefficients in eq 5 are weighting factors dependent on the chemical properties of the solute.49,68 The β factor was theoretically derived from perturbation theory and, for a molecule that carries a net charge, it has a value of 0.5. The R factor is usually empirically determined by fitting over experimental data. Its standard value

The P2 concentration can then be defined considering all the possible 2:1 complexation reactions that involve as receptor one of the N possible 1:1 complexes. Considering for example the 2:1 complex in which the two ligands are bound in sites A and B, the PA,B concentration can be expressed as

PA,B ) PALe-∆GB/RT ) PBLe-∆GA/RT ) PL2e-(∆GA+∆GB)/RT (7) Thus, the total P2 concentration is

Binding of PFOA and PFOS to Human Serum Albumin N

P2 ) PL2

N

∑ ∑

e-∆Gi1,i2/RT

J. Phys. Chem. B, Vol. 114, No. 46, 2010 14865

(8)

i1)1 i2)i1+1

where ∆Gi1,i2 ) ∆Gi1 + ∆Gi2. Iterating the procedure, the generic PN concentration can be expressed as N

Pn ) PLn

TABLE 1: Intramolecular Distances of the Central C-C-C Fragment of a Set of Perfluoroalkanesa

N

∑ ∑

N

∑

···

i1)1 i2)i1+1

e-∆Gi1,i2,...,in-1,in/RT

A-G

A-F

C-G

C-E

C-F

mean err %

3.870 3.84 3.87 3.87 3.86

3.090 3.08 3.09 3.09 3.10

2.720 2.75 2.74 2.73 2.72

2.920 3.02 2.93 2.92 2.93

2.810 2.87 2.82 2.82 2.82

1.54 0.25 0.14 0.22

a

In the first row are reported the published values taken as reference. The column headers refer to interatomic distances and follow the labeling of Figure 3.

in)in-1+1

where n ∈ [1, N]

Jang et al. C6F10 C8F18 C10F22 C20F42

46

(9)

with N

∆Gi1,i2,...,iN-1,iN )

∑ ∆Gi j)1

j

(10)

The definition of the average complexation ratio nj is thus straightforward N

∑ nxn

nj )

(11)

n)1

Figure 3. Central fragment of a perfluoroalkane molecule. The labels are the same as used in Table 1.

where

TABLE 2: Dihedral Angles Calculated for Perfluorodecane and Average Dihedral Angles Determined for the Two Helical Minima Structures of the Perfluorocarbon Chainsa

N

xn ) Pn /

∑ Pn

t(

n)1

is the fractional population of n:1 complexes. In order to take into account the crowding effect of the protein, as proposed in the original Scatchard analysis, a wn2 penalty term for the generic nth complexation reaction was introduced. For a charge spread uniformly over the surface of a sphere, the w coefficient can be expressed using Debye-Huckel theory as69

w)

2 2

ez 1 k 2DkT b 1 + κa

(

)

Φ1 Φ2 Φ3 Φ4 Φ5 Φ6 Φ7 t( 〈Φ〉 h( 〈Φ〉

(12)

where e is the electron charge, z the valence of the ligand, D the dielectric constant of the medium, k the Boltzmann constant, T the temperature, b the radius of the sphere that excludes small ions to a radius a, and κ the reverse of the Debye length. The a and b parameters used in the calculations are those suggested by Scatchard:69 29.5 and 27.5 Å, respectively. The definition of average complexation ratio expressed in eq 11 was used to compare the results calculated in the present study with experimental data. 3. Results and Discussion 3.1. Force Field Validation. The force field implementation was first tested comparing the minimum energy geometries optimized in vacuum for three perfluoroalkanes, C6F14, C8F18, and C10F20, with those determined by Goddard and co-workers using a similar computational protocol.46 The comparison is performed in Table 1 in terms of distances calculated between the A, C, E, F, and G atoms of the CF chain, named according to the convention displayed in Figure 3.46

Jang et al.46

tailored force field

C10F22 165.90 163.50 163.60 163.70 163.60 163.50 165.90 Average Dihedral Angles 164.24 60.43

166.05 162.88 163.03 163.13 163.03 162.88 166.05 163.86 58.49

a In the second column are reported the reference values while in the third those determined with the implemented force field.

The capability of correctly describing the peculiar helical shape assumed by the fluorinated carbon chain was further checked by reproducing the specific sequence of dihedral angles that characterize a perfluorodecane molecule in its most stable configurations. The CCCC potential exhibits in fact two minima that define the two most stable structures for the backbone structural conformers: t is the main minimum (164°) and h the secondary minimum (64°). Since the CCCC potential is symmetric a total of four minimum-energy structures is possible, namely t( and h(, differing for helical orientation (() and minimum dihedral angle (64 or 164°). The minimum energy dihedral angles calculated in the t conformation and the average dihedral angle measured in the two minimum-energy configurations are compared to those calculated by Jang and co-workers46 in Table 2. The force field test was then completed through the comparison of the torsional potential profiles of the dihedral angles that contains oxygen and sulfur atoms in PFOA and PFOS with those determined

14866

J. Phys. Chem. B, Vol. 114, No. 46, 2010

Salvalaglio et al.

Figure 4. Comparison between DFT and FF torsional potentials for CCCC, FCCO, and FCSO dihedrals. The DFT potentials were calculated at the B3LYP/6-31G(d,p) level in vacuum while the FF potentials were calculated through the tailored force field implemented in the present work.

at the B3LYP/6-31G(d,p) level. The absolute error for the torsional potentials is smaller than 1.5 kcal/mol, while the error on the minima position is of a few degrees. This error level is comparable to the average accuracy found in similar works in which Amber compatible parameters were derived, as well as to that of the original Amber parametrization.57,70 In Figure 4 the potential energy surfaces of the FCSO, CCCC, and FCCO dihedral angles calculated with the implemented force field are compared to those determined at the DFT level. It can be observed that the absolute difference between DFT and force field (FF) energies is higher for the CCCC dihedral angle (modeled with the literature derived force field) than for the FCCO or the FCSO dihedral angles (parametrized in this work). Despite this, the CCCC torsional potential was successful in reproducing known structures for perfluoroalkanes (as discussed above and shown in Tables 1 and 2), supporting the conclusion that the level of approximation reached with the classical force field development is adequate for the structural description of PFOA and PFOS. All the DFT/FF comparisons of the torsional potentials derived in this work are reported in paragraph S.5 of the Supporting Information together with the minimum-energy geometry determined for each molecule used to develop or to test the force field. 3.2. Docking. The systematic docking procedure followed by a MMGBSA analysis allowed to identify 24 and 30 possible structures for the HSA-PFOA and HSA-PFOS complexes, respectively. The binding structures of all the complexes so determined are sketched in Figure 5. In both cases, several binding sites among those sketched in Figure 5a for PFOA and 5c for PFOS were already known to the literature37 and are displayed separately in Figure 5b for PFOA and 5d for PFOS. In particular, we found that all the sites that demonstrated a high affinity to fatty acids36,71 were individuated as potential binding sites for PFOA, while only a few of them corresponded to those determined for PFOS. This result suggests that the similarity between PFOA and an alkyl fatty acid allows both compounds to access the same pool of binding sites, although it does not give any insight on the relative stability of the complexes. The ensemble of the docking structures individuated for PFOA and PFOS, however, did not include only fatty acids sites. It was thus found that some structures are superimposed or very close to the binding sites

of well-known pharmaceutical compounds such as warfarin, thyroxine, indole, and benzodiazepin.35 The nice agreement found between computational results and experimental data on the location of the possible HSA binding sites indicates that the adopted computational protocol is reliable and predictive, since all the already known binding sites were identified by the computational protocol. It was thus with some interest that we approached the energetic investigation of the complexes involving putative binding sites not known to the literature. The binding sites under investigation were labeled C, D, E, F, H, K, M, N, P, Q, R, V, W, and X for PFOA and B, D, E, H, I, J, K, L, O, P, R, S, T, U, Y, Z, AB, and AC for PFOS. 3.3. Energetic Analysis. All the 54 complexes identified through the docking/MMGBSA protocol were relaxed through molecular dynamics simulations in order to determine structures and energies of interactions at a level of theory higher than that adopted by the docking algorithms. The free binding energy of complexation was thus first determined from the potential of mean force calculated through US for binding site J. The calculated interaction energy (-8.0 kcal/mol) is in good agreement with the experimental value measured by Chen and co-workers (-7.45 kcal/mol). The PMF calculated for binding site J is reported in Figure 6. As it can be observed, the potential has a first minimum at about 13 Å, after which it rises again until a well-defined plateau is reached at 30 Å. The PMF profile can be interpreted in terms of the environment encountered by the PFOA molecule while it moves along the dissociation coordinate. The drifting of PFOA from binding site J to the solution involves the transition through a predominantly hydrophobic channel. The nonpolar character of the cavity makes the charged fluorinated compound unable to perform specific interactions, thus determining the constantly increasing trend observed in the PMF profile. The local minimum at about 13 Å is determined by the establishment of an interaction between PFOA and a network of positively charged amino acids constituted by the side-chain terminal guanidinium groups belonging to Arg 218 and Arg 222 and the backbone amine group of Asn 294. The final plateau that starts at 30 Å is determined by the loss of contact between PFOA and HSA and thus corresponds to the solvated state. The interaction energy calculated for the Trp 214 binding site was then adopted to fit the R scaling coefficient of LIE

Binding of PFOA and PFOS to Human Serum Albumin

J. Phys. Chem. B, Vol. 114, No. 46, 2010 14867

Figure 5. (a) Structures of the 24 complexes individuated for HSA-PFOA through docking; (b) PFOA molecules in fatty acids/pharmaceuticals binding sites; (c) structures of the 30 complexes individuated for the HSA-PFOS with docking; (d) PFOS molecules in fatty acid/pharmaceutical binding sites.

Figure 6. Potential of mean force evaluated as a function of a reaction path describing the transition of a PFOA surfactant molecule from the J binding site to a fully solvated state. The free energy of complexation was evaluated as the difference of the PMF between the two states and has an absolute value of 8.0 kcal/mol. The chosen reaction coordinate is reported in the inset.

theory, which was used to calculate the ∆G of interaction of the other 53 complexes. The fitted R parameter is 0.67. The interaction free energies and reaction energies were then calculated for all the other binding sites using LIE theory and are summarized in Tables 3 and 4 for PFOA and PFOS, respectively. The interaction between PFOS and HSA is dominated by apolar interactions. The electrostatic contribution to the com-

plexation free binding energy is in fact often slightly positive or close to zero while the van der Waals contribution can be as high as -10 kcal/mol. This can be mostly ascribed to the long perfluorinated carbon chain, which contains one CF2 group more than PFOA. It can be observed that the PFOS interaction free energies are slightly higher (as absolute values) with respect to those determined for PFOA, in agreement with the experimental evidence of the higher bioaccumulation potential of PFOS. Also, the binding sites with the lowest free energies correspond to those with the lowest reaction enthalpies, thus indicating that entropic effects, though important, are secondary with respect to enthalpic contributions and giving support to the reliability of the computational binding site identification, as enthalpy changes are less susceptible to systematic errors than LIE free energies. The LIE interaction free energies predict a slightly lower affinity of PFOS for the Trp 214 binding site (site W, -7.79 kcal/mol) with respect to PFOA, which is qualitatively in agreement with what found experimentally for PFOS, though the measured interaction energy, -6.01 kcal/mol,41 is significantly smaller. A possible explanation of the origin of this underestimation is that site W is positioned in proximity of site V, which is the HSA binding site for PFOS with the highest interaction energy. It is thus reasonable that the interaction energy measured through fluorescence spectroscopy might be diminished by the mutual repulsion between two adsorbed PFOS molecules if the two sites are contemporarily populated. A comparison of two PFOS molecules adsorbed in the V and W

14868

J. Phys. Chem. B, Vol. 114, No. 46, 2010

Salvalaglio et al.

TABLE 3: LIE Free Energies of Complexation and Reaction Enthalpies Calculated for the PFOA-HSA Complexes Identified by the MMGBSA Refined Docking Procedurea label

∆VEEL (kcal/mol)

∆VVDW (kcal/mol)

∆H (kcal/mol)

∆G (kcal/mol)

A B C D E F G H I J K L M N O P Q R S T U V W X

7.06 4.63 6.04 -2.86 5.87 -3.37 0.06 10.94 3.52 -1.61 9.95 6.83 14.45 3.66 -1.88 2.99 0.60 6.23 11.07 0.64 33.24 5.83 4.81 1.77

-12.62 -8.27 -6.90 -6.96 -9.12 -4.66 -7.72 -9.12 -7.17 -10.74 -13.19 -12.54 -9.11 -9.25 -7.90 -5.39 0.03 -14.40 -11.04 -12.23 -13.55 -12.06 -3.55 -4.79

-5.56 -3.64 -0.86 -9.82 -3.25 -8.03 -7.66 1.82 -3.65 -12.35 -3.25 -5.71 5.34 -5.58 -9.78 -2.40 0.62 -8.17 0.03 -11.59 19.69 -6.24 1.26 -3.02

-4.93 -3.23 -1.61 -6.10 -3.17 -4.81 -5.14 -0.64 -3.04 -8.00 -3.87 -4.99 1.12 -4.37 -6.24 -2.12 0.32 -6.54 -1.87 -7.88 7.54 -5.17 0.02 -2.32

The adopted LIE scaling parameters are R ) 0.67 and β ) 0.5. In the second and third columns are reported the electrostatic and van der Waals interaction energies. a

sites, which highlights their proximity, is sketched in Figure 7. Further arguments supporting this thesis are given in section 3.6. 3.4. Structural Analysis of PFOA-HSA Complexes. The analysis of structure and characteristic features of the HSA binding sites for PFOA was focused on the five having a complexation free energy higher than 6 kcal/mol. In Figure 8 is reported a sketch of the binding structure of each complex, in which are highlighted the involved amino acid residues. In order to rationalize the relative importance of polar and apolar contributions, van der Waals and electrostatic binding energies have also been reported in Table 3 together with the free energy of complexation estimated with the LIE approach. The binding site location on the HSA surface is discussed below site by site. A comparison with the binding sites individuated through X-ray experiments for fatty acids by Bhattacharya and co-workers34 is reported as well. Complex D does not correspond to any known fatty acid binding site, but has a certain similarity with binding site FA5, first individuated by Ghuman and co-workers.35 FA5 is located in the IIIB domain of HSA and is able to host oxyphenyl butazone, a molecule that differs from PFOA both from a chemical and a structural standpoint. The complex D structure is characterized by an ionic interaction between the carboxylic group of PFOA and the terminal amino group of the Lys 413 side chain, and is further stabilized by a contemporary apolar interaction between the fluorocarbon chain of PFOA and the Val 493 and Lys 538 residues. In particular, the fluorinated chain interacts significantly with the aliphatic portion of the Lys side chain. The complexation free energy is -6.09 kcal/mol. Complex T is superimposed with fatty acid site 1. The polar head interacts with both Arg 117 and Arg 186, while the fluorinated chain interacts with a network of aromatic rings and

TABLE 4: LIE Free Energies of Complexation and Reaction Enthalpies Calculated for the PFOS-HSA Complexes Identified by the MMGBSA Refined Docking Procedurea label

∆VEEL (kcal/mol)

∆VVDW (kcal/mol)

∆H (kcal/mol)

∆G (kcal/mol)

A AA AB AC AD B C D E F G H I J K L M N O P Q R S T U V W X Y Z

5.22 2.84 5.33 13.39 7.76 5.00 3.74 3.04 9.29 8.14 0.80 2.38 0.08 7.64 2.68 10.11 13.16 1.51 -1.40 9.61 6.04 7.64 3.04 8.44 11.55 3.02 8.44 7.37 3.61 3.44

-15.69 -10.76 -2.39 -11.02 -13.97 -10.91 -11.73 -7.40 -12.37 -10.64 -11.35 -6.24 -9.97 -9.43 -3.82 -10.15 -14.76 -11.39 -8.96 -10.23 -13.50 -8.93 -7.55 -9.10 -10.39 -15.40 -17.92 -14.65 -8.24 -10.49

-10.47 -7.92 2.94 2.38 -6.21 -5.91 -7.99 -4.35 -3.08 -2.50 -10.56 -3.86 -9.89 -1.79 -1.14 -0.04 -1.60 -9.88 -10.36 -0.62 -7.46 -1.29 -4.51 -0.67 1.16 -12.38 -9.48 -7.28 -4.64 -7.05

-7.90 -5.79 1.07 -0.69 -5.48 -4.81 -5.99 -3.44 -3.64 -3.06 -7.21 -2.99 -6.64 -2.50 -1.22 -1.75 -3.31 -6.88 -6.70 -2.05 -6.03 -2.16 -3.54 -1.88 -1.19 -8.81 -7.79 -6.13 -3.72 -5.31

a The adopted LIE scaling parameters are R ) 0.67 and β ) 0.5. In the second and third columns are reported the electrostatic and van der Waals interaction energies.

Figure 7. Details of HSA-PFOS binding sites V and W.

apolar groups constituted by the side chains of Tyr 138, Tyr 161, Ile 142, His 146, Phe 149, Phe 157, Leu 182, and Leu 185. The binding site involved in this complex is one of the most favored for the complexation of short fatty acids, though also palmitate was crystallized in this site.37 The complexation free energy is -7.88 kcal/mol. Complex O shares with the fatty acids binding site 4 the charged residue host, Arg 348, placed between the IIB and IIIA

Binding of PFOA and PFOS to Human Serum Albumin

J. Phys. Chem. B, Vol. 114, No. 46, 2010 14869

Figure 8. Details of the binding sites for the five HSA-PFOA most stable complexes.

domains of HSA. The apolar portion of the molecule is, however, located mostly on the opposite side with respect to the fatty acid site. The interactions established by the apolar chain involve Arg 472, Lys 475, Asn 483, Ala 490, and Leu 491. The complexation free energy is -6.24 kcal/mol. The R complex is characterized by a rather simple structure: the surfactant polar moiety is involved in electrostatic interactions with the two lysine residues, Lys 432 and Lys 436, while the chain establishes some favorable contacts with Tyr 452. The complexation free energy is -6.54 kcal/mol. The J complex corresponds to fatty acid site 8. It is located between domains II and III of HSA and is mainly characterized by the establishment of van der Waals interactions between the fluorinated chain and Lys 199, Lys 195, Arg 218, Arg 222, and Trp 214. Also, electrostatic interactions involving the carboxylic group of the surfactant are present and contribute to the free binding energy, though to a lesser extent with respect to van der Waals energies. This is a well-known HSA binding site, since it is placed exactly in the core of the heart-shaped protein and is the only one that presents a fluorescent residue: Trp 214. This site, considered by Bhattacharya a low-affinity binding spot for carboxylic acids, seems to play an important role for the binding of PFOA. Both the umbrella sampling calculations performed in the present work and the experimental results obtained by Chen and Guo41 confirm the stability of the J complex. The calculated complexation free energy is the highest among those calculated, -8.0 kcal/mol, and is mostly determined by van der Waals interactions.

Summarizing, the structural analysis of the binding sites reveals that PFOA binds the protein both through polar and nonpolar interactions. A first group of stable complexes (R, T) are stabilized through the establishment of strong van der Waals interactions between the fluorinated backbone chain and typically apolar amino acids, with a binding character which can be considered typical of uncharged compounds. In the remaining complexes, both electrostatic and the van der Waals interactions play a relevant role. The amphiphilic nature of the surfactant is responsible for the interactions that can characterize sites with an apolar core surrounded by positive charges. A relevant example of this kind of interaction sites is the above-mentioned binding site J. The hydrophobic core of this site is able to host the fluorinated chain while the charged amino or guanidinium group placed at the interface with the solution interacts with the charged polar head of PFOA. 3.5. Structural Analysis of PFOS-HSA Complexes. The binding site location on the HSA surface is discussed below site by site for binding sites with complexation energies higher than 6 kcal/mol. The sketches of the binding structures are reported in Figure 9. In the N complex, PFOS binds HSA in a region that corresponds to fatty acid site 1, in subdomain IB. The position assumed by PFOS is very close to the experimental structure found by Bhattacharya for short-chain fatty acids.34 At the end of the MD simulation, the sulfonic group interacts with Lys 114 and Lys 190 while the fluorinated carbon chain interacts with Leu 115, Leu 182, Arg 117, Pro 118, Tyr 138, and Tyr

14870

J. Phys. Chem. B, Vol. 114, No. 46, 2010

Salvalaglio et al.

Figure 9. Details of the binding sites for the seven HSA-PFOS most stable complexes.

161. It was interesting to observe that this complex undergoes a quite important reorganization during the MD simulation: the polar head in fact produces a distortion in the local arrangement of the side chains and loses in part the favorable polar contacts that were present in the docked structure. This reorganization permits the establishment of apolar contacts. As described above, this complex was found to be stable also when the ligand is PFOA and corresponds to the HSA-PFOA T complex. The complexation free energy is -6.88 kcal/mol. Complex A is located in subdomain IIIB in correspondence to fatty acid site 5. The spatial organization of the binding site found for fatty acids is recalled here by the conformation assumed by PFOS. The polar head interacts with Tyr 401 and Lys 525 while the fluorinated chain interacts with Phe 507, Phe 509, Leu 532, Leu 575, and Met 548, showing a remarkable

affinity for such a hydrophobic chemical surrounding. The complexation free energy is -7.90 kcal/mol. In complex V the surfactant interacts with a region of the protein surface placed between the adjacent fatty acid sites 6 and 8. The structural features of this complex are in substantial agreement with the known binding structure found for fatty acids. While the polar head is stabilized by the electrostatic interactions with Lys 199 and Arg 484 in site 8, the apolar chain interacts with Phe 206, Arg 209, Glu 354, Leu 347, Lys 351, and Ser 480, which belong to site 6. This cooperation between neighboring binding sites was observed also for some fatty acids.34 The complexation free energy is -8.81 kcal/mol. Complex G was determined starting the simulations from a structure located in site 9. The relaxation through molecular dynamics produced a partial reorganization of the complex:

Binding of PFOA and PFOS to Human Serum Albumin while the polar interactions between the sulfonic group and Arg 186 and Arg 428 were maintained, the fluorinated chain moved to form more favorable interactions with Leu 179, Pro 180, Asp 183, Asp 187, Glu 184, Glu 518, and Lys 519. The complexation free energy is -7.21 kcal/mol. Complex I involves a binding site that has not yet been identified in the literature, which is adjacent to another favored complex, named O. In both complexes the binding interaction takes place in subdomain IB. Both complexes are characterized by a high mobility: in complex I the polar head interacts with Arg 114, Arg 186, Lys 190, and Lys 519 while the chain interacts with Val 116, Glu 425, Glu 520, Ile 523, and Lys 524. In complex O the polar interactions involve Arg 114 and Arg 186 while the chain maintains a contact with Asn 109. The complexation free energies of complex I and O are -6.64 and -6.70 kcal/mol, respectively. Also for the PFOS-HSA complexes one of the most important binding configurations involves Trp 214 (PFOS site W), though in a different arrangement from that found for PFOA or known for fatty acids. In this case, the sulfonic group interacts with Asn 295, Arg 218, Arg 222, and Lys 444, while Trp 214 is part of the molecular surface composed by Lys 195, Leu 198, Asp 451, and Ser 454 and interacts with the fluorinated chain. The complexation free energy is -7.79 kcal/mol. 3.6. Scatchard Analysis and Prediction of PFOA/PFOS Behavior in Blood. The estimated free complexation energies were finally employed in order to perform a site-specific Scatchard analysis, devoted to the evaluation of the complexation ratio and of the distribution of the ligands over the available binding sites. The first parameter that was extrapolated from the Scatchard analysis of the computational results is the limiting complexation ratio, evaluated when the concentration of PFOA and PFOS in solution is equal to their critical micelle concentration (25 mM PFOA, 8 mM PFOS72), which represents the upper limit for the number of ligand molecules allowed to be simultaneously bound to HSA. This parameter was calculated considering two limiting ideal situations: when the surfactant molecules interact with the protein without mutual influence and when the energetic penalty due to the crowding effect of a single protein molecule is taken into account through the wn2 penalty term, in which the w coefficient is calculated within the Debye-Huckel approximation (eq 12). The complexation ratio for PFOA-HSA complexes is 9.1 for w ) 0 and 7.4 with w calculated through eq 12. This last value is in good agreement with the 6 to 9 PFOA ligands per protein molecule experimentally measured and reported by Han et al.,39 while it is slightly smaller than the 13 molecules found by Wu and co-workers.43 For the PFOS-HSA complexes, the limiting complexation ratios are higher: 11.5 for w ) 0 and 8.7 for w calculated using eq 12. To test further the reliability of the developed computational framework, the complexation data measured by Han and co-workers for PFOA-HSA were determined and are compared with experimental data in Figure 10 for different values of w. The agreement is good for the simulations performed with w ) 0, while in the other case the experimental data are overestimated. A perfect match between experimental and calculated data for w calculated with Debye theory is obtained decreasing all the interaction free energies by 0.7 kcal/mol, which is well within the uncertainty of the present calculations. In these conditions the complexation ratio is 9.4. The same computational analysis protocol was adopted to determine the complexation ratio of PFOS with HSA in the experimental conditions explored by Chen and Guo41 in order

J. Phys. Chem. B, Vol. 114, No. 46, 2010 14871

Figure 10. Comparison between the binding affinity and number of binding sites measured by Han et al.39 using microdesalting column separation and those here calculated for the same ligand and protein concentrations (50 µmol).

to determine the PFOS binding energy in the Trp 214 binding site (site W). The calculations, performed considering the repulsive w contribution, predicted that the complexation ratio is comprised between 2.4 and 7.0. This means that in all the investigated conditions the binding site with the highest energy, site V, will be populated. This supports the hypothesis advanced in section 3.3 that the measured binding free energy of PFOS in site W, -6.0 kcal/mol, might be decreased by repulsive interactions established with a second PFOS molecule adsorbed in the nearby V site. The second stage of the Scatchard analysis was centered on the analysis of the relative contribution to the overall binding of HSA of each binding site. At this stage the relative population of each complex (calculated at a 2 mM surfactant concentration and normalized to 1) has been calculated as a function of the binding ratio (n) and of the occupied binding sites. It was found that, in the surfactant concentration range under investigation, the relative population of the binding sites exhibits an almost negligible dependence on the number of ligands per protein molecule n. The population of the binding sites for each one of the possible 1:1 complexes is shown in Figure 11. This analysis highlights that the most probable binding sites for PFOA-HSA complexes are those marked with the D, J, O, R, and T labels (Figure 11a). The PFOS-HSA complexes exhibit a slightly broader distribution than that observed for PFOA-HSA complexes, which corresponds to a larger number of sites with a nonnegligible probability of being populated (6 for PFOS vs 5 for PFOA). The most probable binding sites for PFOS are A, G, I, N, V, and W (see Figure 11b). The developed model was finally employed to study the complexation of HSA by PFOA and PFOS in conditions that correspond to the surfactant levels measured in the human blood in different situations. The calculations were performed for a mean HSA concentration of 50 g/L for concentrations of PFOA and PFOS that reflect three situations: (i) the average values measured in the pool of north American population monitored in the work of Olsen and co-workers (35 ng/mL PFOS, 5 ng/ mL PFOA73), (ii) the average values measured in individuals living in close proximity to a PFOA or PFOS production site (0.329 ppm, measured in the Washington, DC, area in proximity of a DuPont factory74), and (iii) the average value measured in chemical plant employees who work directly in contact with perfluorinated surfactants (PFOA 0.899 ppm, PFOS 0.941 ppm).75 The results of the calculations are summarized in Table 5. In all cases the excess of protein with respect to the surfactant

14872

J. Phys. Chem. B, Vol. 114, No. 46, 2010

Salvalaglio et al.

Figure 11. PFOA (a) and PFOS (b) normalized populations of the binding sites of 1:1 surfactant-HSA complexes determined considering a 5 × 10-5 M protein concentration and a 2 × 10-3 M surfactant concentration. The histograms report on the x axis the label of each binding site and on the y axis the fraction of 1:1 complexes. The most populated binding sites are reported in the inset.

TABLE 5: Percentage of Complexed HSA Calculated Using the Scatchard Analysis for Three Surfactant Concentrations: (i) Average Values Measured in a Selected Pool of the American Population,73 (ii) Average Value Measured in Individuals Living in Close Proximity of a PFOA/PFOS Production Site,74 and (iii) Average Value Measured in Chemical Plants Employees75 ligand conc. (ppm)

complexed HSA (%)

i ii iii

0.005 0.329 0.899