Atomic and Molecular Insight into How PFOA Reduces α‑Helicity, Compromises Substrate Binding, and Creates Binding Pockets in a Model Globular Protein

Abstract

Per- and polyfluoroalkyl substances (PFAS) pose significant health risks due to their widespread presence in various environmental and biological matrices. However, the molecular-level mechanisms underlying the interactions between PFAS and biological constituents, including proteins, carbohydrates, lipids, and DNA, remain poorly understood. Here, we investigate the interactions between a legacy PFAS, viz. perfluorooctanoic acid (PFOA), and the milk protein β-lactoglobulin (BLG) obtained using a combination of experimental and computational techniques. Circular dichroism studies reveal that PFOA perturbs the secondary structure of BLG, by driving a dose-dependent loss of α-helicity and alterations in its β-sheet content. Furthermore, exposure of the protein to PFOA attenuates the on-rate constant for the binding of the hydrophobic probe 8-anilino-1-naphthalene sulfonic acid (ANS), suggesting potential functional impairment of BLG by PFOA. Steered molecular dynamics and umbrella sampling calculations reveal that PFOA binding leads to the formation of an energetically favorable novel binding pocket within the protein, when residues 129−142 are steered to unfold from their initial α-helical structure, wherein a host of intermolecular interactions between PFOA and BLG’s residues serve to insert the PFOA into the region between the unfolded helix and beta-sheets. Together, the data provide a novel understanding of the atomic and molecular mechanism(s) by which PFAS modulates structure and function in a globular protein, leading to a beginning of our understanding of altered biological outcomes.

Graphical Abstract

INTRODUCTION

Per- and polyfluoroalkyl substances (PFAS), the ubiquitous and “forever” class of chemicals, which number over 7000 compounds, have been found in human blood, human breast milk, neonates, and children.^1–5 The association of PFAS compounds with adverse human health outcomes has emerged as a major societal and governmental health concern.^6–9 PFAS-associated toxicity includes reproductive effects (decreased fertility), elevated levels of blood pressure, late onset of development in children, risk for cancers, and complications due to compromised immunity. PFAS is also a known obesogen.

Elsewhere, 21 PFAS compounds found upon post-mortem analysis revealed that in the kidney and lung, perfluorobutanoic acid was the most frequently found PFAS (and at highest concentrations with median values residing at 263 and 807 ng/g in the kidney and lung, respectively).^1–3,5,10 In the liver and brain, perfluorohexanoic acid showed the maximum levels (median: 68.3 and 141 ng/g, respectively), while perfluorooctanoic acid (PFOA) was the most prevalent PFAS in bone (median: 20.9 ng/g).^5,11,12 In the serum of fishery workers, PFAS levels ranged between 0.01 and 10,400 ng/mL, while those exposed to contaminated drinking water presented with levels ranging between 0.5 and 35.5 ng/mL, suggesting that occupational exposure can result in PFAS burdens 1−4 orders of magnitude higher than the general population.^13,14 PFAS has been detected in serum from the maternal chord and in newborn blood, demonstrating that these chemicals are capable of passing through the placental barrier. Estimates of the efficiency of placental transfer range between 30% and 79%.^15,16 Breast milk has been shown to account for 83−99% of PFAS total daily intake for infants.^17,18 Nevertheless, the initial underestimation of their environmental impact together with their widespread use has caused in the last decades an increased level of environmental contamination from poly- and perfluorinated compounds and, in particular, from PFAS.^19,20 Although, several efforts have been made for PFAS detection and removal from water^21–24 and soil,^25–27 addressing the consequences of their historical use remains a critical challenge.

Considering their pernicious presence in human fluids, an understanding of the impact of PFAS exposure to matrix constituents such as proteins in milk and blood is of interest. The potential unraveling of PFAS-dependent structural perturbations and/or functional alterations may correlate with adverse clinical outcomes associated with PFAS exposure. It can result in the identification of targets along the sequalae to pathology and provide PFAS-specific information, leading to the design of more benign alternatives.

Here, we have examined the interactions between the milk protein β-lactoglobulin (BLG) and the legacy PFAS PFOA. BLG is a small globular milk protein comprising 162 amino acids containing a hydrophobic calyx (barrel) comprising two sets of β-sheets.²⁸ The calyx serves as a binding pocket for hydrophobic ligands, including retinol (vitamin A), fatty acids (palmitic, docosahexanoic, oleic, and others), and organic molecules 8-anilinonaphthalene-1-sulfonic acid (ANS) and sodium dodecyl sulfate (SDS).^29–34 It is interesting to note that in a series of elegant experiments, both the secondary structure and the interactions of BLG with α-lactalbumin were examined against a number of ligands including surfactants and nano-curcumin. While BLG secondary structure was found to be affected by the addition of surfactants and other ligands, the role of nano-curcumin unearthed in the protein:protein interactions provides valuable information in food technology.^35,36 Functional roles attributed to BLG include that of ligand transport, neonatal vision and brain development, modulation of immune responses, and potential enhancement of cell proliferation.^37,38 Despite its ban in 2019, PFOA continues to pervade the biome via its persistent presence in soil, water, animals, burn sites (for firemen training), and plants, even though significant efforts have been made for their removal from soil³⁹ and water⁴⁰ using various technologies. Their entry into the food chain occurs directly through ingestion of PFAS-contaminated water, food, and beverages that in turn may have been in contact with PFAS-containing substances.^41–44

PFOA exposure has been linked to increased incidence of cancers including testicular, kidney, prostate, thyroid, bladder, and breast.⁴⁵ In animal models, PFOA administration impacted reproduction, altered growth, and was hepatotoxic.⁴⁶

In the recent past, there have been efforts to identify the molecular underpinnings driving PFAS-associated cellular and organismal outcomes. For example, PFAS binds to human serum albumin (HSA), which helps explain its systemic distribution and agency for toxicity.^47–51 Furthermore, crystallographic studies implicated a number of amino acids in HSA interacting with PFOA through −H···F−C, C−F···π, and C−F···X bonds.⁵⁰ A single high-affinity binding site, along with three other weaker binding sites, was disclosed in the report. Not surprisingly, the PFOA:HSA binding interactions mimicked those made between the protein and short-chain fatty acids.⁵⁰ Other studies also demonstrated the types of impacts that PFAS has on important proteins. The impacts PFAS compounds have on proteins are multifaceted, as summarized in Table 1. The diverse effects of PFAS include alterations to both the secondary and tertiary structures of proteins, or in some instances, proteins may serve as carriers for certain PFAS molecules.

Table 1.

Impact of PFAS on Several Proteins from Other Studies, Utilizing Circular Dichroism, Fluorescence Spectroscopy Experiments, or High-Performance Tandem Mass Spectrometry

effect on secondary structure
protein	PFAS type	[PFAS](mM)	α-helicity	β-sheet content	effect on tertiary structure	carrier	ref
Hb	PFNA	1 to 1000	increased by ∼3%	decreased by ∼5%	N/A	N/A	57
β-gal	PFNA	1 to 1000	increased by ∼3%	increased by 7%	N/A	N/A	57
Hb	PFOS	1000	decreased by 13.8%	increased from 8.6% to 17.8%	N/A	N/A	58
hL-FABP	PFOA	100	increased by 2%	decreased by 6%	N/A	N/A	52
catalase	PFOA	4000	decreased from 32.9 to 31.6	increased from 15.6% to 18%	activity decreased to 71.81%	N/A	55
Acot1	PFOA	200	increased from 12.2 to 15	decreased from 66.1% to 52.6%	N/A	N/A	53
lysozyme	PFOA	1000	decreased from 26.4 to 19.2	increased from 32.5% to 41.6%	N/A	N/A	56
BSA	PFOS	0 to 2000	increased by 7%	increased by ∼2%	destroyed	N/A	59
albumin	PFOS, PFOA, PFHxS, PFNA		N/A	N/A	N/A	yes	51

The consequences of protein exposure to PFAS were previously observed to vary depending on the protein type and the PFAS molecule being studied. In proteins like human liver fatty acid-binding protein (hL-FABP),⁵² acyl-CoA thioesterase 1 (Acot),⁵³ and bovine serum albumin (BSA),⁵⁴ PFOA has been observed to reduce α-helical content by 2−10%, accompanied by changes in β-sheet content. Similarly, PFOA decreases α-helical content in proteins like catalase⁵⁵ and lysozyme.⁵⁶ However, perfluorononanoic acid (PFNA) induces a 3% increase in the α-helical content in hemoglobin (Hb),⁵⁷ and PFOS leads to a 13.8% decrease in α-helical content in Hb.⁵⁸ PFOA is also known to induce tertiary structural alterations in proteins. For example, in catalase,⁵⁵ PFOA decreases α-helical content by 1.3% and also causes a substantial 71.8% reduction in enzyme activity. In BSA, PFOA exposure results in the complete disruption of tertiary structure accompanied by a 7% increase in α-helical content.⁵⁴ Several studies ^51,59–62 have established “functional” connections between PFAS and albumin proteins, with albumin serving as a carrier for PFAS such as perfluorooctanesulfonic acid (PFOS), PFOA, perfluorohexanesulfonate (PFHxS), and PFNA.⁵¹

Besides impacting proteins, studies have found that PFOA ingestion led to decreased levels of methylated DNA in juvenile female mice.^63,64 It also altered the expression of regulators involved in cell cycling and genes associated with anti-apoptosis, modulated levels of genes involved in splicing, and mTor signaling pathways, leading to a mechanistic model for associated hepatotoxic outcomes.^65–67 Furthermore, both PFOS and PFOA were found to associate with LDL, cholesterol, and TAGs, which may provide a basis for PFAS-related cardiovascular diseases. Elsewhere, PFOA heightened glucose and lipid indices in both the serum and liver, providing further clarity into its hepatotoxic effects.^68–71

Despite the emerging inroads, atomic and molecular insights into how fluorinated alkyls perturb biological constituents remain understudied. Therefore, using BLG and PFOA as representative models of globular proteins and PFAS, we have examined both structural and functional consequences of BLG exposed to PFOA. The experimental data reveal a PFOA dose-dependent diminution in the structurally critical α-helix coupled with alterations in the β-sheet content. Reduced on-rate constants for the recruitment of the hydrophobic probe ANS were also discovered, suggesting a PFOA-dependent compromise in function (potentially resulting from the structural perturbations). Docking studies and molecular dynamics (MD) simulations of the protein “soaked” in PFOA unearth multiple PFOA binding sites, with the strongest residing within the protein calyx and interacting with several hydrophobic residues. Data from the umbrella sampling (US) method to explore the mechanism by which PFOA perturbs the BLG secondary structure suggested that PFOA molecules bind to the partially unwound helix, stabilizing its (unwound) coiled state.

Along with possible PFOA-dependent functional consequences leading to deleterious biological effects, broader PFAS-dependent and PFAS-specific cellular and organismal dyshomeostasis is discussed.

2. RESULTS AND DISCUSSION

The interactions between PFOA and BLG, and their impact on BLG structure and substrate binding, were comprehensively investigated using a combination of experimental techniques (circular dichroism and fluorescence spectroscopy) and computational methodologies (molecular dynamics simulations and free energy calculations).

2.1. PFOA Induces Secondary Structure Perturbations in BLG.

To examine the dose-dependent effect of PFOA on the secondary structure of BLG, we performed circular dichroism (CD) spectroscopy experiments. BLG CD spectra obtained as a function of increasing PFOA dose are shown in Figure 1a. The native protein minimum at 218 nm shifts toward the right and decreases in absorption intensity as a function of PFOA concentration, indicating PFOA dose-dependent perturbations in the secondary structure of the globular protein.

Figure 1.

PFOA induces secondary structure perturbations in BLG. (a) CD spectra of BLG as a function of the increasing PFOA concentration (0−10 μM). Chemical structure of PFOA is shown in the inset. (b) Change in α-helicity content (red, bottom), β-sheet content (blue, middle), and turns and coils content (orange, top, labeled as others) for BLG with varying PFOA concentration. (c) Summary of α-helix content as a function of PFOA concentration (0−10 μM).

The CD spectra were deconvoluted to determine the secondary structure component changes with the addition of PFOA, as shown in Figure 1b. Native BLG comprises 9.1 ± 0.3% helix and 30.5 ± 1.6% β-sheet, with the remaining 60.3 ± 1.5% comprising turns, coils, and loops, in agreement with the previous literature.⁷² By contrast, exposure of the protein to increasing concentrations of PFOA (0−10 μM) results in a complete loss of α-helicity at PFOA concentrations of ≥8 μM (Figure 1b). Concomitantly, PFOA-induced changes in β-sheet content are observed up to the tested concentration of 10 μM. The alterations in the composition of other secondary structural elements are also shown (Figure 1b). The changes in the helical content as a function of PFOA dose are reported numerically in Figure 1c. In conclusion, upon carefully examining the secondary structural components of BLG in the presence of 10 μM PFOA, it is found that the protein exhibits a non-native structure with 0% α-helix and reduced β-sheet content. Additionally, we do not see any consistent increase/decrease in the β-sheet content of BLG upon titrating it with PFOA. Nevertheless, the reduction in the percentage of α-helix suggests the formation of a non-native intermediate rather than a moltenglobule state, further suggesting a nonhierarchical unfolding of the protein in the presence of PFOA (0 to 10 μM).

2.2. PFOA Compromises On-Rate of 8-Anilino-1-naphthalenesulfonic Acid Binding to BLG Calyx.

Next, we examined the influence of PFOA on BLG binding to the fluorescent molecular probe ANS, which is known to favor binding to hydrophobic ligand binding pockets⁷³ including the calyx region of BLG (formed by eight stranded antiparallel β-sheets, flanked on one side by the structurally critical α-helix). The fluorescence quantum yield of ANS is sensitive to solvent polarity, where its fluorescence emission intensity becomes quenched and blue-shifted when in polar or aqueous environments in contrast to hydrophobic environments.^74,75

As shown in Figure 2, we examined the fluorescence of ANS upon its addition to a solution of BLG alone or PFOA-incubated BLG (50 μM PFOA). Initially, the ANS fluorescence intensity is weak since ANS is added directly into the aqueous (polar) solution. Over time, as some of the ANS binds to BLG present in the solution and occupies the hydrophobic environment, its fluorescence quantum yield increases, which also increases with time as more ANS molecules find hydrophobic binding regions. The tangents to the initial increase in the ANS fluorescence emission intensity upon addition to the protein are found to be PFOA-dependent. Their numerical values (slopes) reflect the on-rate constants for ANS binding to the protein and are estimated at 4.5 × 10⁻³ RFU/s (absence of PFOA), 4.1 × 10⁻³ RFU/s when the protein is exposed to 10 μM PFOA, and 1.4 × 10⁻³ RFU/s in the presence of 50 μM PFOA. The PFOA dose-dependent diminution in on-rate constants for the ANS:BLG complex suggests that PFOA either occupies the ANS-binding site in the protein, viz., within the calyx, and/or perturbs the hydrophobic binding site, leading to attenuated ANS on-rate constants (correlating with PFOA dose).

Figure 2.

PFOA attenuates on-rate of ANS binding to BLG calyx. (a) Time dependence of the ANS fluorescence intensity growth curves upon the exposure of BLG to the ANS fluorophore. The inset shows the ANS structure and the schematic representation of the systems examined in fluorescence spectroscopy, viz., a control BLG solution and PFOA-soaked BLG solution. (b) Tangents to the initial slopes of the curves in panel (a). RFU stands for relative fluorescence unit.

2.3. PFOA Exhibits Strong Binding to BLG’s Calyx.

To examine the interactions between PFOA and BLG with atomistic resolution, we carried out molecular docking calculations and molecular dynamics simulations of PFOA:BLG complexes. The PFOA molecule was first docked onto BLG by scanning the entire protein surface. Representative docked structures of PFOA binding to BLG with their respective docking scores (ranging between −3.5 and −7.5 kcal/mol) are shown in Figure 3a. The PFOA molecule with the most favorable docking score of −7.5 kcal/mol is bound within the calyx of BLG, indicating that the calyx interior has the greatest affinity for the molecule. In addition, PFOA has other weaker binding sites on the BLG surface and distant from the calyx, whose docking scores are found to be −4.5 kcal/mol.

Figure 3.

PFOA binding to BLG. (a) Snapshot of representative docked structures of PFOA and BLG (cyan). PFOA molecules are shown in van der Waals representation and colored according to docking score values. (b) A snapshot of BLG colored according to the percentage of time spent by PFOA molecules within 3 Å of BLG (left). A space-filled representation of BLG, with the PFOA-binding residues shown in surface representation and the bound PFOA molecules labeled by their assigned IDs (right). (c) Binding energies evaluated using the molecular mechanics generalized Born-surface area (MMGB-SA) approach for PFOA molecules and BLG, averaged over the last half of the production run. (d) Percentage time bound of amino acid residues found within 3 Å of PFOA with ID 1 (calyx). (e) PFOA molecule with ID 1 binding to BLG (left). Key amino acid residues interacting with PFOA in the calyx of BLG (right). (f) Distance between the nitrogen atom of LYS60 (BLG) and carboxylic acid carbon (PFOA) over the course of a 2.1 μs trajectory. (g) The N−C distance at t = 0 and average N−C distance from the last 50% of the production run. The inset snapshot shows the initial and final N−C distances from the production run. PFOA carbon atoms are depicted in cyan, fluorine in green, oxygen in red, nitrogen in blue, and carbon atoms of BLG in cyan.

To examine the evolution of the docked complex over time, the structures with the most favorable docking scores were used as the initial structures for carrying out MD simulations. The initial system contained seven PFOA molecules at various sites on BLG, with their affinities reflected by their individual docking scores. In Figure 3b, a snapshot of the equilibrated system is shown after a 2.1 μs long production run. Protein residues are color-coded based on the percentage of time that any atom of PFOA molecules spends within 3 Å of the protein residues (any atom of the protein residues). Four out of seven PFOA molecules stay bound to BLG even after 2.1 μs of equilibration, as shown in Figure 3b. Their corresponding binding free energies are reported in Figure 3c. The PFOA molecule (ID 1) stays bound in the calyx region of BLG throughout the 2.1 μs trajectory, with the most favorable binding energy of −25.57 kcal/mol relative to the other molecules. The other PFOA molecules (IDs 2, 3, and 4), initially bound to their binding sites on BLG, unbind, diffuse in the aqueous solution, and then return to the original binding site regions over the course of the MD trajectory. The observed behavior indicates that PFOA has favorable binding to those sites, but the binding is significantly weaker than in the calyx (∼−10 kcal/mol). Overall, PFOA molecules exhibit relatively strong binding to BLG and to multiple sites on its surface.

Next, we characterized the atomic-level interactions for the most favorable PFOA:BLG binding mode using the atoms of PFOA and the amino acid residues of the BLG’s calyx as interacting candidates. Figure 3d shows a plot of amino acid residues (N90, L39, L58, I84, L87, K69, V41, I71, and K60) that are within 3 Å of PFOA for more than 50% of the trajectory as a function of the percentage of time they are bound to PFOA. Figure 3e shows the modes of interaction between PFOA and the calyx residues within a van der Waals distance of 3 Å for PFOA via a snapshot from the last frame of the trajectory. Several hydrophobic amino acid residues, viz., L39, V41, L58, I71, A86, and L87, participate in hydrophobic interactions with the PFOA molecule. Additionally, there are directional contacts of PFOA with polar and basic residues, viz., N90, K60, and K69. The positively charged amine group of K60 has a strong electrostatic interaction with PFOA’s negatively charged carboxylic acid group. This interaction between K60 and the PFOA molecule facilitates stronger and stable binding of the fluorinated carboxylic acid to the protein. The distance between the amino group of K60 and the carboxylic acid group of PFOA remains mostly constant, with an average value of 3.94 Å during the last half of the trajectory (Figure 3f,g). The computational results demonstrating PFOA favoring binding within the hydrophobic calyx site of the BLG protein agree with it (PFOA) being both a hydrophobic and lipophilic ligand. While the cooperativity of binding of PFOA to BLG is an interesting aspect of the binding process, addressing it remains beyond the scope of our MD simulations. Theoretically speaking, and considering the ability of PFOA to perturb BLG’s secondary structure, cooperativity is likely to have an adverse effect on the functionality of BLG, which is also reflected in the experimentally observed ∼3× reduction of the on-rate constant of BLG in the presence of PFOA (Figure 2b).

To interpret the results of the fluorescence experiments in Figure 2, we examined the binding of the ANS fluorescent probe to BLG using molecular dynamics simulations. Previous structural studies of ANS:BLG only relied on docking⁷⁶ and do not appear to include MD simulation studies. Therefore, our simulations were expected to provide a more detailed picture of ANS:BLG interactions. We simulated three independent systems of ANS:BLG complexes, with ANS initially binding to the calyx. Over the course of 1 μs simulations, ANS showed multiple binding modes on the BLG surface, as shown in Figure S1a,b. The results of three independent simulations reveal that ANS has transient binding to three sites, i.e., at the top of the calyx (C1), near the α-helix (C2), and near the calyx (C3). Previous fluorescence experiments showed that ANS has two binding constants when interacting with BLG,⁷⁷ indicating two possible binding sites on BLG, which were hypothesized to be near the calyx and near the α-helix. This hypothesis was reinforced by previous docking studies, which reproduced the ANS binding site near the calyx.⁷⁶ In our studies here, the two previously established binding sites^73,76,77 were reproducible in our MD simulations.

To examine the impact of PFOA upon binding of ANS to BLG, we simulated one system with both ANS and PFOA initially placed in the calyx. In the presence of PFOA, ANS binds near the α-helix (labeled previously as a C2 site) for close to a hundred percent of the trajectory time (Figure S1b).

We note that our computational results are based on the classical force field approximations for the interactions between fluorine (halogen) atoms in PFOA and the rest of the system, including BLG protein, solvent, and ions and other ligands. Describing halogen bonds accurately is an active area of research, with some recent improvements in parametrizations within CHARMM General Force Field (CGenFF)⁷⁸ and polarizable force fields⁷⁹ to better account for halogen bonding. Other all-atom force fields and MD simulations have been shown to reproduce several thermodynamic properties of fluorinated compounds that correlate strongly with experimental results.⁸⁰ In the CHARMM all-atom force field used in the present work, halogen atom models have a small partial negative atomic charge with relatively large favorable dispersion contributions. Previous studies showed that these models lead to favorable interactions of halogens with hydrogen bond donors, but with relative insensitivity to orientation.^78,81 Since the halogen bonds in the CHARMM force field have been examined in detail, we expect that our modeling also exhibits the reported errors. However, as halogen parametrization has been explored, investigated, and reported on by others and most of our modeling conclusions are also supported by our experimental studies, we expect that our overall conclusions are largely valid. Furthermore, besides halogen interactions, PFOA−BLG interactions are largely influenced by the (better described) electrostatic interactions of PFOA’s negatively charged carboxylic acid end with the positively charged protein amino acids. Finally, the analysis of contacts and distances in the main identified binding site of PFOA (calyx) shows that only one out of seven contacts (fluorine of PFOA and N90 atom) and only one out of 15 fluorine atom interactions (on a PFOA molecule) can be classified as a halogen rather than hydrophobic interaction.^82,83

2.4. PFOA Creates a New Binding Pocket and Inserts upon Unwinding of the BLG α-Helix.

While CD studies showed that PFOA leads to complete loss of α-helicity in BLG (Figure 1), observing such large secondary structural changes is unlikely in several microsecond-long MD simulations. Therefore, we performed several short (200 ns) steered molecular dynamics (SMD) simulations⁸⁴ to unwind BLG’s α-helix by an applied force and then tracked the behavior of PFOA molecules present in the solution. Notably, as seen in Figure 4a, the forced unwinding of the α-helix leads to single or multiple PFOA molecules inserting between the beta strands of BLG and the newly created coil, thus creating a new and primarily hydrophobic pocket suitable for strong PFOA interactions. Then, PFOA molecules were observed engaging in complex hydrophobic interactions and some hydrogen-bonding interactions with BLG residues, as illustrated in Figure S2. Upon a comprehensive analysis of the binding strengths of the inserted PFOA molecules within the newly formed pocket, it was observed that one of the inserted PFOA molecules possessed stronger binding energy compared to the PFOA situated near the calyx, as depicted in Figure S2. This indicates that PFOA exhibits a greater affinity toward the newly formed pocket than the established calyx site.

Figure 4.

PFOA insertion upon unwinding of the BLG α-helix. (a) Scheme of SMD calculations. In the initial state, BLG has a stable helix, one PFOA binds stably in the calyx, and other PFOA molecules bind transiently to BLG surface sites. In the final state, BLG has an unwound helix and several PFOA molecules inserted underneath the newly created coil. (b) Free energy profile obtained from the US calculations (left y-axis, lines) and the calculated % α-helicity (right y-axis, scatter plot) as a function of RMSD. The free energy profile of the control system (BLG) is shown in black, and the free energy profile of the BLG system with PFOA is shown in green. (c) Representative snapshots from US calculation windows at RMSD values of 1, 2.5, 4, and 5.75 Å. BLG is shown in cartoon representation in red, where the α-helix residues (residues 129 to 142) are highlighted in yellow. The atoms of PFOA are shown as van der Waals spheres, where C, F, and O atoms are shown in cyan, green, and red, respectively.

To explore the mechanism by which PFOA leads to the loss of helicity in BLG, we next performed US calculations to determine the free energy profile of BLG with varying α-helicity composition in the absence and presence of PFOA. These calculations were designed to obtain the free energy profile that can determine if the PFOA presence leads to spontaneous (favored) loss of helicity in BLG. The examined α-helix included BLG residues 129−142, and the intact helix corresponds to 100% helicity, whereas the fully unwound helix corresponds to 0% helicity. US calculations used the root-mean-square deviation (RMSD) as the reaction coordinate, where the RMSD of zero represented the intact helix from the crystal structure and the RMSD of 5 to 6 Å represented the fully unwound coil (Figure 4a).

The results of the US calculations were used to reconstruct the free energy profile as a function of the reaction coordinate (RMSD) or, correspondingly, the percentage of the α-helicity, shown in Figure 4b. The free energy profile was reconstructed from 23 separate RMSD windows, which showed good overlap (Figure S3). This profile also showed acceptable convergence (Figure S4). In the absence of PFOA molecules, BLG is shown to possess one stable minimum state, in which residues 129−142 exist in the α-helical conformation, in agreement with the crystal structure of the BLG protein.⁸⁵ The barrier to undergoing the α-helix to coil transition is as high as 23 kcal/mol. However, in the presence of seven PFOA molecules in solution, of which one to three PFOA molecules are inserted in the newly formed pocket, the BLG structure exhibits two shallow minima, in which residues 129−142 exist in the α-helical conformation (RMSD ∼ 0.75 Å) and in the random coil state (RMSD ∼ 4.6 Å). The observed minima structures could potentially correspond to the intermediate states of BLG during the unfolding process. Importantly, these favorable states of random coil corresponding to the second free energy minimum are observed only when one or multiple PFOA molecules are inserted between the BLG body and the random coil (Figure 4c), similarly to that in the SMD simulations. In the presence of PFOA, the barrier to α-helix unwinding drops to ∼5 kcal/mol, where reduction of the energy barrier for the helix to coil transition is also likely associated with PFOA partly inserting underneath the partially unfolded helix. Therefore, our results confirm that PFOA lowers the barrier to α-helix unwinding in BLG, creates new binding pockets in which it inserts, and stabilizes a BLG state in which residues 129−142 assume a random coil conformation.

The strength of the binding of the inserted PFOA molecules was also examined in the umbrella sampling calculation trajectories. Table S2 lists molecular mechanics generalized Born-surface area (MMGB-SA) binding energies of PFOA molecules, either in the calyx or in the new binding site created by the unfolding of the α-helix, obtained from umbrella sampling calculations. In most cases, PFOA in the calyx site exhibits more favorable binding energy than PFOA molecules in the new pocket. Nevertheless, there are still several instances where PFOA molecules in the new binding pocket bind with comparable affinity (such as within the calyx). Since US calculations have mostly multiple PFOA molecules binding to the new binding pocket, this result is expected, as the energies corresponding to binding modes of multiple versus single molecules in binding pockets are not directly comparable. Overall, the results in Table S2 also confirm significant and strong binding of PFOA molecules to the new binding pocket.

3. CONCLUSION

In this study, we investigated the interactions between PFOA and the BLG protein using a combination of experimental and computational techniques. The results from circular dichroism spectroscopy revealed a complete loss in α-helicity and alterations in β-sheet content of BLG protein when the protein was incubated with PFOA. Furthermore, the on-rate constant for the binding of the hydrophobic fluorescence probe ANS was decreased by ∼4×, suggesting a potential functional impairment due to PFOA, possibly stemming from the structural changes mentioned earlier. Note that ANS binding metrics is used to “sample” the structural integrity of the hydrophobic (calyx) binding pocket of BLG and is used as a mimic not only for retinol but also fatty-acid and other ligands binding to the protein.^29–34 The mechanism of perturbation of PFOA by BLG was revealed using steered molecular dynamics simulations, which demonstrated the formation of an energetically favorable novel binding pocket near the α-helix region (residues 129 to 142), formed as a result of α-helix unwinding. Upon the formation of this pocket, the PFOA molecules interact with the residues initially forming the α-helix and the neighboring residues constituting the calyx via hydrophobic and hydrogen bond interactions.

The free energy landscapes of BLG conformations with and without the helical segment were obtained with umbrella sampling calculations. This landscape, schematically summarized in Figure 5, confirmed that the formation of the new binding pocket in BLG becomes energetically favorable in the presence of PFOA molecules, as PFOA inserts into the newly formed pocket and stabilizes the unfolded state of the protein−ligand complex in which BLG has no helix. The presence of PFOA also significantly reduces the free energy associated with the transition state, ΔG^#, i.e., the energy barrier between the BLG conformations with the folded helix (folded state) and without the helix (unfolded state). Furthermore, the PFOA molecules exhibited stronger binding affinity toward the new binding site than to the “calyx” site, which has been known to host hydrophobic ligands.

Figure 5.

Scheme of the free energy landscape of BLG protein conformations in the presence of PFOA.

To the best of our knowledge, our study is the first to report that PFAS compounds can induce the formation of novel binding pockets in globular proteins. This study can help pave the way for further exploration of other PFAS compounds by examining whether they also exhibit tendencies to unfold secondary structure elements, potentially forming new binding pockets.

4. EXPERIMENTAL MATERIALS AND METHODS

4.1. Experimental Studies.

4.1.1. Proteins and Chemicals.

Samples of bovine BLG isoform A (purity of ≥99%) were purchased from Sigma-Aldrich. The protein solutions were prepared in a 1 mM phosphate buffer (with pH 7.0). PFOA was purchased from Lab 261. All other chemicals were of analytical grade.

4.1.2. Circular Dichroism Measurements.

CD measurements were conducted with a Jasco J-1500 spectrometer, capturing spectra from 190−260 nm at a scan speed of 50 nm/min and a 1 nm bandwidth. A quartz cuvette with a path length of 0.01 cm was used. Baseline measurements in the solvent were conducted using 1 mM phosphate buffer. The CD spectrum of native BLG (120 μM in a 1 mM phosphate buffer) was collected. The CD spectra of BLG exposed to PFOA (concentrations ranging from 1 to 10 μM) were also recorded. For quantitative analyses, BestSel software was employed to assess α-helices, β-sheets, turns, and unordered structures in BLG proteins in our samples.

4.1.3. Fluorescence Kinetics Measurements of ANS Binding to BLG.

For fluorescence kinetic measurements, 40 μM delipidated BLG (in 1 mM phosphate buffer at pH 7.0) was placed in a cuvette prior to addition of ANS. ANS fluorescence was recorded over a period of 60 s (excitation at 350 nm and emission at 540 nm). The experiment was repeated with BLG incubated in 10 and 50 μM PFOA for 5 min at room temperature. Data were analyzed using the initial slope of the ANS fluorescence intensity growth curve, following the smoothing of the raw data (dumping factor of 0.9).

4.2. Computational Studies.

4.2.1. Molecular Docking Calculations and Molecular Dynamics Simulations.

In computational studies, we examined the binding of a PFOA to BLG. The initial structure of PFOA was obtained from the RCSB database (ligand ID 8PF),⁸⁶ which was deprotonated in GaussView,⁸⁷ since PFOA exists in a deprotonated form at neutral pH. The initial structure of the BLG protein was extracted from the crystal structure 1GX8.⁸⁵ PFOA molecules were docked onto BLG using the AutoDock Vina software.^88,89 In our docking procedure, the grid box was centered at various positions on a grid with varying x-, y-, and z-coordinates with an interval of 5 nm. The docking was performed by scanning PFOA binding to the whole protein surface, using the automated procedure within our in-house Linux shell script (https://github.com/vukoviclab/docking-scan ),⁹⁰ which returns the ligand poses and their corresponding docking scores. The obtained PFOA ligand poses were analyzed using the gmx-cluster algorithm in the GROningen MAchine for Chemical Simulations (GROMACS) package.⁹¹ The centroid structures of PFOA were extracted from each major cluster using the gmx cluster, and they served as the starting point for the MD simulations.

In MD simulations, four systems were modeled in total: control (BLG only), PFOA system (BLG + seven PFOA molecules), control with ANS (BLG + ANS), and control with PFOA and ANS (BLG + PFOA + ANS). All four systems were solvated using TIP3P water molecules and were neutralized with sodium ions. Table S1 lists the numbers of molecules and the sizes of the constructed systems. All-atom MD simulations were conducted using the open-source NAMD 2.13 package⁹² and CHARMM36 force field parameters.^93,94 The simulations were performed under Langevin dynamics in the isothermal−isobaric NPT ensemble with the temperature and pressure set to 310 K and 1 bar, respectively. The Langevin coefficient (γ_Lang) was set at 1 ps⁻¹. A time step of 2 fs was used, and Coulomb and van der Waals nonbonded interactions were computed every one and two time steps, respectively, for all atoms within a 12 Å cutoff distance. Long-range Coulomb interactions were calculated using the particle-mesh Ewald (PME) method with periodic boundary conditions⁹⁵ applied. First, all four systems were minimized for 20,000 steps. After minimization, solvent molecules were equilibrated for 1 ns, during which the protein and ligand were restrained using harmonic forces with a spring constant of 1 kcal/(mol·Å²). Following solvent equilibration, the systems were simulated for 2.1 μs-long production runs. In these production runs, harmonic force restraints with a spring constant of 1 kcal/(mol·Å²) were applied to one of the central atoms of the protein, preventing significant thermally induced translation within the periodic unit cells and no other restraints. Three independent runs were set up for control with the ANS and PFOA system and were run for 1 μs each. The amino acids of BLG in the crystal structure, prior to and after interacting with PFOA molecules, were examined for chirality and trans/cis configurations using a VMD plugin.⁹⁶ No PFOA-induced changes in their chirality were found, and all amino acids were in the expected trans configuration.

4.2.2. Umbrella Sampling Calculations.

To examine the mechanism of perturbation induced by PFOA in BLG’s secondary structure, we employed the umbrella sampling method⁹⁷ with the weighted histogram analysis method (WHAM),⁹⁸ utilizing the collective variables (colvars) module⁹⁹ of NAMD 2.13. The reaction coordinate was defined as the RMSD of α-helix backbone atoms (residues 129 to 142) with respect to a reference structure, where the reference structure is the one with an intact α-helix. An example setup for US calculations is posted on the project GitHub site (https://github.com/vukoviclab/PFAS-protein-MD-analysis). First, we performed an SMD calculation, where a harmonic pulling of α-helix atoms was performed with a spring constant of 12 kcal/(mol·Å²) to attain an RMSD of 6 Å over 200 ns. Next, two analogous US runs were set up for the control system (BLG only) and the PFOA system (BLG with 1−3 PFOA molecules in the vicinity of residues 129−142). Each system had a changing reaction coordinate, which varied in 23 US windows centered at RMSD = 0.25, 0.5, 0.75, 1, 1.25, 1.5, 1.75, 2, 2.25, 2.5, 2.75, 3, 3.25, 3.5, 3.75, 4, 4.25, 4.5, 4.75, 5, 5.25, 5.5, 5.75, and 6 Å. The force constant used in each US window was 50 kcal/(mol·Å²). Each window was run for 100 ns, totaling 4600 ns overall simulation time. The WHAM algorithm¹⁰⁰ was used to reconstruct the free energy profile. For each window, Monte Carlo bootstrap error analysis was also performed along with the WHAM algorithm (with num_MC_trials set to 50). The histograms derived from the US windows, essential for reconstructing the free energy profile, showed a suitable overlap, as shown in Figure S3. The compositions of the systems used in US simulations are described in Table S1.

4.2.3. Binding Energy Calculations.

The relative binding free energies between BLG and the PFOA molecules binding to several positions on BLG’s surface were calculated using the MMGB-SA method. The following equation was used to calculate the MMGB-SA free energies of the extracted configurations from the PFOA system:

$$G_{\text{tot}}=E_{\text{MM}}+G_{\text{solv}-\text{p}}+G_{\text{solv}-\text{np}}-T\Delta S_{\text{conf}}$$ (1)

where the term $E_{\text{MM}}$ represents the sum of bonded and Lennard-Jones energy terms and $G_{\text{solv}-\text{p}}$, $G_{\text{solv}-\text{np}}$ correspond to the polar and nonpolar contributions of the solvation energy. $T\text{Δ}{S}_{\text{conf}}$ is the conformational entropy, which was not considered in these calculations. The terms $E_{\text{MM}}$, $G_{solv-\text{p}}$, and $G_{solv-np}$ were calculated using the NAMD 2.13 package generalized Born implicit solvent model where the value of the dielectric constant of the solvent was 78.5. In NAMD calculations, the $G_{\text{solv}-\text{np}}$ term was determined using a linear function of the solvent accessible surface area (SASA) with a probe radius of 1.4 Å, expressed as $G_{\text{solv}-\text{np}}=\gamma$·SASA, where the surface tension, $\gamma$, was specified as 0.00542 kcal/(mol·Å²). The binding free energies for the system of interest were computed according to the following expression:

$$\langle G_{\text{MMGB}-\text{SA}}\rangle =\langle G_{\text{tot}}(\text{PFOA}-\text{BLG})-G_{\text{tot}}(\text{PFOA})-G_{\text{tot}}(\text{BLG})\rangle$$ (2)

The angular brackets denote the time average.

4.2.4. BLG Residues Bound to PFOA Molecules.

To examine the stability of binding between BLG and PFOA molecules, we calculated the percentage of time PFOA molecules spend near the protein’s surface. First, the protein residues that lie within 3 Å of PFOA molecules were extracted. Next, for the obtained residues, the percentage of time they spend in the proximity of BLG was calculated. The calculation was done for all of the frames in the trajectory. Lastly, the beta column of the PDB obtained from the last frame of the trajectory was colored according to the percentage of time the PFOA molecules spend near BLG. The percentage time calculation was done using our tcl script, also available on our GitHub project site.

A similar calculation was done for the amino acid residues that lie within 3 Å of the PFOA molecule bound in the calyx of BLG. First, the residues found within 3 Å of PFOA were extracted. Next, for the extracted residues, the number of frames were counted for which the residues make contacts with PFOA. The fraction of time for each residue was obtained by dividing the count by the total number of frames and later converted to percentage time bound. The frame count was calculated using a tcl script, and the percentage time bound plot was generated using Python3.

4.2.5. Distance Analysis of PFOA and K60.

To examine the electrostatic interactions between PFOA and the amino acid residues in the calyx of BLG, we calculated the distance between the carboxyl carbon of PFOA and the nitrogen atom of K60 in BLG. Upon visually observing the MD trajectory of the PFOA system, the carboxyl oxygen of the PFOA molecule in the calyx was seen to participate in electrostatic interaction with the hydrogen atom of K60’s amine group. To further investigate this interaction, we calculated the distance between the carboxyl carbon of PFOA and the nitrogen atom of K60 in BLG, and carbon and nitrogen atoms were deliberately chosen to calculate the distances to avoid deviations in the distances due to σ bond rotations. The distances were calculated over the course of 2.1 μs-long production run trajectories using a tcl script. An average was calculated using the distances for the last 50% of the trajectory using bash as described below:

$$d_{\text{C}(PFOA)-\text{N}(\text{K}60)}(t)=\frac{1}{0.5t_{total}}\mathbf{\text{∑}}_{t_{50\%}}^{t_{100\%}}(r_{\text{C}(PFOA)}(t)-r_{\text{N}(\text{K}60)}(t))$$ (3)

Tcl and bash codes for this analysis are available on the GitHub project site.

4.2.6. Percent α-Helicity Calculation.

To obtain a correlation between the percentage α-helicity and the free energies obtained from the US, we calculated the percentage α-helicity for each US window. The calculation was done using VMD plugins incorporated in a tcl script where first the three helical content types, namely, α-helix (H), 3−10 helix (G), and pi helix (I), were identified for the residues of interest (residues 129 to 142). The fraction of helical content was calculated by dividing the count of H, G, and I atoms by the number of all the selected residues, and then each fraction was converted to percentages. The calculation was performed using the last frame of the MD trajectory.