Quantitative Chemical Proteomics Reveals Interspecies Variations on Binding Schemes of L-FABP with Perfluorooctanesulfonate

Abstract

Evaluating interspecies toxicity variation is a long-standing challenge for chemical hazard assessment. This study developed a quantitative interspecies thermal shift assay (QITSA) for in situ, quantitative, and modest-throughput investigation of chemical−protein interactions in cell and tissue samples across species. By using liver fatty acid binding protein (L-FABP) as a case study, the QITSA method was benchmarked with six per- and polyfluoroalkyl substances, and thermal shifts (ΔT_m) were inversely related to their dissociation constants (R² = 0.98). The QITSA can also distinguish binding modes of chemicals exemplified by palmitic acid. The QITSA was applied to determine the interactions between perfluorooctanesulfonate (PFOS) and L-FABP in liver cells or tissues from humans, mice, rats, and zebrafish. The largest thermal stability enhancement by PFOS was observed for human L-FABP followed by the mouse, rat, and zebrafish. While endogenous ligands were revealed to partially contribute to the large interspecies variation, recombinant proteins were employed to confirm the high binding affinity of PFOS to human L-FABP, compared to the rat and mouse. This study implemented an experimental strategy to characterize chemical−protein interactions across species, and future application of QITSA to other chemical contaminants is of great interest.

INTRODUCTION

Nearly 300 million tons of >80,000 chemical contaminants enter the environment worldwide every year.^1,2 Hazard assessment of chemical contaminants requires the evaluation of toxicities across species including humans and other mammals, fishes, and invertebrates.³ This is typically achieved through interspecies extrapolation of toxicity data from a small number of laboratory model species. Across species, unfortunately, differential chemical sensitivity of orders of magnitude may exist, ^4,5 which poses challenges to protecting susceptible species. Intrinsic binding of chemical contaminants to molecular targets (i.e., proteins) has been suggested as a major reason⁴⁻⁶ for interspecies toxicity variations. For example, planar polychlorinated biphenyls show 800-fold stronger toxicity in chicken (Gallus gallus) than the common tern (Sterna hirundo) due to the variation of two amino acids in the ligand-binding domain of an aryl hydrocarbon receptor (AhR).^4,7 Thus, investigation of the physical interactions between chemical contaminants and their protein targets is promising to inform interspecies toxicity variations in a high-throughput way. Computational toxicology has been recently proposed to predict chemical–protein interactions across hundreds of species,⁸⁻¹⁰ but the uncertainty of prediction is challenging to control without comparisons to empirical results. Implementation of a complementary experimental strategy to evaluate interspecies toxicity variations is of great interest.

Bioassays with recombinant proteins or reporter systems are frequently used to characterize the interactions between chemical contaminants and protein targets.⁴⁻⁶ Indeed, large interspecies variations were revealed with this strategy for several nuclear receptor-mediated toxicities.^4,11 However, classic bioassays are laborious and prone to technical issues (e.g., protein insolubility) and are limited by low throughput. More importantly, heterologous protein expression systems such as E. coli, yeast, and mammal cell lines are not always physiologically relevant. Post-translational modifications (PTMs), chaperones, and co-factors in the host cellular environment are essential to modulate the functions of many proteins, including nuclear receptors,¹² but are missing in most in vitro systems, potentially introducing artifacts.¹³ Thus, an alternative methodology is needed to “in situ” characterize chemical−protein interactions across species, in a high-throughput and quantitative way.

Probe-based chemical proteomics, e.g., activity-based protein profiling (ABPP), has made important progress in recent years to identify protein targets of small molecules on a proteome-wide level.¹⁴⁻²¹ However, the application of probe-based chemical proteomics to quantitative chemical−protein interactions is limited by its intrinsic insensitivity to binding affinity. Methods for probe-free chemical proteomics have been recently developed based on protein thermal stability, conformation changes, or protease accessibility from our group and others.²²⁻²⁶ Probe-free chemical proteomics is time-effective to evaluate many chemicals of interests, without the need for laborious chemical probe synthesis. Furthermore, as exemplified by the thermal shift assay,^22,20 the thermal stability enhancement of target proteins is quantitatively associated with the dissociation constant (K_d) of ligands (Figure 1A). This suggests the potential of thermal stability-based chemical proteomics for in situ and quantitative characterization of ligand binding in cells or even tissues.

Figure 1.

PFOS enhances the thermal stability of hL-FABP. (A) Ligand binding enhances the thermal stability of proteins by decreasing the Gibbs free energy. ΔG_(total) indicates the total Gibbs free energy to denature holoprotein; ΔG_(U) indicates the Gibbs free energy to denature apoprotein; ΔG_(L) indicates the Gibbs free energy reduced by ligand binding; R is the gas constant (8.314 J K⁻¹ mol⁻¹); and T is the temperature on the Kelvin scale. K_d indicates the dissociation constant of ligand binding. (B) PFOS enhances the thermal stability of hL-FABP in crude E. coli lysates, as evidenced by SDS-PAGE. Lesser abundances of denatured proteins were detected in the PFOS treatment group. Red arrows indicate the band of hL-FABP (∼14.2 kDa). (C) Confirmation of the enhanced thermal stability of hL-FABP by PFOS, with western blotting. (D) Workflow of QITSA to determine chemical−protein interactions in cells or tissues across species. Targeted proteomics was adopted to quantify the fraction of denatured proteins.

Per- and polyfluoroalkyl substances (PFASs) are a class of chemical contaminants of public concern due to their²⁷⁻²⁹ persistence and toxicity. Large interspecies variations on the toxicokinetics of PFASs were observed across mammals, rodents, fish, and birds.^28,30−32 Liver fatty acid binding protein (L-FABP) is a critical target regulating the toxicokinetics of PFASs with a high binding affinity,³³⁻³⁷ but interspecies interaction between PFASs and L-FABP remains unknown despite its sequence conservation across vertebrate species.³⁸ In this study, we developed a quantitative interspecies thermal shift assay (QITSA) for in situ determination of PFAS and endogenous L-FABP interactions across species including humans, rats, mice, and zebrafish.

MATERIALS AND METHODS

Chemicals and Reagents.

Perfluorooctanesulfonic acid (PFOS), perfluorobutanesulfonic acid (PFBS), perfluorohexanesulfonic acid (PFHxS), perfluorobutanoic acid (PFBA), perfluorooctanoic acid (PFOA), and perfluorononanoic acid (PFNA) were obtained from the U.S. Environmental Protection Agency (EPA). The hL-FABP pNIC28-Bsa4 plasmid (plasmid #42344) and pTrc99A-Rat FABP1 plasmid (plasmid #13577) were purchased from Addgene (Cambridge, MA, USA). Recombinant mouse L-FABP was purchased from R&D system (Minneapolis, MN, USA). The high-speed plasmid mini kit was purchased from Geneaid Biotech Ltd. BL21 Escherichia coli was purchased from BioLabs Inc. (Whitby, ON, CA). HepG2 was obtained from ATCC (HB8065). Triethyl ammonium bicarbonate buffer (TEAB), tris(2carboxyethyl) phosphine hydrochloride (TCEP), iodoacetamide, cOmplete Mini EDTA-free protease inhibitor cocktail tablets, and dithiothreitol (DTT) were purchased from SigmaAldrich (St. Louis, MO, USA). 4-(2-Hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES) and phosphate-buffered saline tablets were obtained from BioShop Canada Inc. (Burlington, ON, Canada). Antimycotic, low-glucose Dulbecco’s minimal essential medium (DMEM), fetal bovine serum (FBS), isopropyl-β-d-thiogalactopyranoside (IPTG), 1-anilinonaphthalene-8-sulfonic acid (1,8-ANS), ultrapure water, methanol, and acetonitrile (HPLC grade) were obtained from Fisher Scientific (Ottawa, ON, CA). Formic acid (mass spectrometry grade) was purchased from EMO Millipore (St. Louis, MO, USA). Trypsin Gold (mass spectrometry grade) from Promega (Madison, WI, USA) was used for protein digestion. All other reagents are the highest grade available.

On-Gel or Proteomics-Based QITSA.

Chemical incubation and thermal denaturation of proteins were performed as described previously with minor modifications.²³ Briefly, 800 μg of recombinant hL-FABP expressed in E. coli was aliquoted into 1.5 mL tubes, and the proteins were diluted to 1 mg/mL with TS buffer (20 mM Tris, 150 mM NaCl, pH 7.4). PFOS was added into the lysate to a final concentration of 100 μM, followed with incubation on ice for 1 h with shaking. The control group was incubated with the same volume of DMSO (0.1%). Subsequently, the lysates were split into new tubes with 50 μL in each. Heat treatment was carried out in a block heater with seven-point serial temperatures starting from 55 °C (temperature points were 55, 60, 65, 70, 75, 80, and 85 °C), with two replicates for each temperature. Relatively high temperatures at 55−85 °C were selected to denature L-FABPs, mainly due to its high thermal stability compared to most human proteins.³⁹ The workflow is shown in Figure 1. All samples were then centrifuged at 14,000g for 30 min at 4 °C. The supernatant was discarded after centrifugation, while the protein pellets were washed with 50 μL of TS buffer and resuspended with 20 μL of loading buffer. The denatured proteins were subjected to SDS-PAGE to determine hL-FABP (14.2 kD). Also, the results were confirmed by western blot with the anti-His antibody. More details about expression of recombinant hL-FABP, SDS-PAGE, and western blot are provided in the Supporting Information (SI).

For cell lysates or liver tissue samples, targeted proteomics was employed for QITSA to determine endogenous L-FABPs. Cell lysates or liver tissue extracts were incubated with 100 μM PFOS and heated using the same protocol of recombinant L-FABP as described above. The pellets were digested into peptides with trypsin and subjected to LC−MS-based proteomics analysis (see more details in the SI) to determine the thermal shift curves of endogenous L-FABPs in the presence or absence of PFOS.

Statistical Analyses.

Statistical analyses were performed mainly via GraphPad Prism or R studio. The FASTA file of Homo sapiens was obtained from the Uniprot database. MaxQuant algorithms (http://maxquant.org/, version 1.6.2.3) were utilized to process the proteomic raw files to identify and quantify proteins. More details about MaxQuant searching are provided in the SI.

To calculate ΔT_m, peptide intensities were first normalized to the intensity of the same peptides at the highest temperature in the control group (DMSO in this case). In GraphPad Prism, the dataset of QITSA was fit to a Boltzmann sigmoidal curve with eq 1:

$$Y=\text{bottom}+\left(\text{top}-\text{bottom}\right)/\left(1+\exp \left(\left(T_{\text{m}}-X\right)/\text{slope}\right)\right)$$ (1)

where Y is the percentage versus the maximal peptide intensity, X is the temperature, bottom is the baseline percentage at low temperature, top is the maximal percentage, slope is the steepness of the curve, with larger values denoting shallower curves, and T_m is the melting temperature of the protein (the temperature at which 50% of the protein is precipitated). The fit values of each parameter in the equation were provided by GraphPad Prism for subsequent data analysis. ΔT_m was calculated with the following equation:

$$\Delta T_{\text{m}}=T_{\text{mPFCs}}-T_{\text{mDMSO}}$$ (2)

In some of the PFAS treatment groups, proteins were not fully precipitated, leading to a biased estimate of maximal percentage. In this case, the T_m of treatment groups needs to be adjusted with eq 3, and T_mc is the calibrated T_m value:

$$T_{\text{mc}}=X_{50}=T_{\text{m}}-\log \left(\frac{\text{top}-Y_{50}}{Y_{50}-\text{bottom}}\right)\times \text{slope}$$ (3)

As for shotgun proteomics, the abundance ratios between proteins from the DMSO group and the same proteins from PFOS treatments at the same temperature were calculated below:

$${\text{ratio}}_{i,j,k}=\text{abundance}\_{\text{DMSO}}_{i,j,k}/\text{abundance}\_{\text{PFOS}}_{i,j,k}$$ (4)

where abundance_DMSO_i,j,k represents the abundances of the ith protein from the kth replicate of DMSO groups, at the jth temperature, and abundance_PFOS_i,j,k represents the abundances of the ith protein from the kth replicate of PFOS treatment, at the jth temperature.

To further account for the fluctuations during sample preparation and nLC−MS/MS analysis, the peak abundances were further normalized by the ratios of L-FABP, which has been determined by targeted proteomics with high reproducibility.

$$\text{normalized}\_{\text{ratio}}_{i,j}=\left(\sum _{k=1}^n{\text{ratio}}_{i,j,k}\right)/n\times \left(\sum _{k=1}^n\frac{{\text{FABP}}_{j,k}}{{\text{FABP\_ref}}_{j,k}}\right)/n$$ (5)

where FABP_ref_j,k represents the ratio of the L-FABP from the kth replicate at the jth temperature, determined by targeted proteomics; FABP_j,k represents the ratios of the L-FABP from the kth replicate at the jth temperature determined by shotgun proteomics; and n is the number of replicates (n = 2).

Due to the limited number of data points for robust curve fitting, the average ratios were used to rank target proteins:

$$\text{rank}\_{\text{ratio}}_i=\left(\sum _{j=1}^m\text{normalized}\_{\text{ratio}}_{i,j}\right)/m$$ (6)

where rank_ratio_i represents the finalized ratio of the ith protein for ranking and m is the number of temperature points. For shotgun proteomics, six temperature points (55−80 °C) were involved in ratio calculation.

RESULTS AND DISCUSSION

Establishing a Fast QITSA Method with Targeted Proteomics.

To test if PFAS can enhance the thermal stability of L-FABP, perfluorooctanesulfonate (PFOS), the most ubiquitous and bioaccumulative PFAS, was used as a submicromolar hL-FABP ligand.³⁴ As a proof of concept, Histagged hL-FABP was expressed in E. coli and then the crude cell lysates were incubated with PFOS or DMSO (negative control). After heating the extracts to 60−85 °C, hL-FABP from DMSO treatment was gradually denatured following elevated temperatures, completely precipitated at 75 °C (Figure 1B). In contrast, hL-FABP from PFOS treatment was not denatured until 80 °C, while other E. coli proteins showed similar thermal denaturation trends between control and PFOS treatment as evidenced by SDS-PAGE. Western blotting with anti-His antibody confirmed the resistance of hL-FABP to thermal denaturation after incubation with PFOS (Figure 1C). The results demonstrated that interaction with PFOS enhanced the thermal stability of hL-FABP by >5 °C. Encouraged by the results from recombinant hL-FABP, a QITSA workflow by in situ monitoring the thermal stability of endogenous hL-FABP in the cell or tissue lysates from different species was proposed by employing LC−MS-based proteomics (Figure 1D).

Additional studies examined whether PFOS could stabilize endogenous hL-FABP in human cells. Several cells including HEK293, MCF7, BEAS-2B, and HepG2 cells were tested, and hL-FABP was only detected in HepG2 by nano-LC−MS/MS-based (nLC−MS/MS) shotgun proteomics (data not shown). Notably, a high abundance of hL-FABP was detected in HepG2, e.g., peak intensity is 1.6 × 10⁸ for peptide AIGLPEELIQK. This is consistent with previous studies that hL-FABP is one of the most abundant cytosolic proteins found in the liver.⁴⁰ Note that the objective of QITSA is to elucidate interspecies variations for a selected protein, rather than target identification at the proteome-wide level commonly pursued by chemical proteomics.²³ Therefore, targeted proteomics was chosen due to its better reproducibility and sensitivity. Considering the long running time of nLC (>120 min per sample), to achieve high throughput, regular HPLC−MS/MS (∼15 min per sample) was used to monitor endogenous hL-FABP. Recombinant hL-FABP was digested, and the tryptic peptides were used as standards. Among 12 peptides detected with a sequence coverage of 81.9%, two unique peptides with strong instrumental signals were selected for subsequent targeted proteomics (details are shown in Table S1). The collision energy was optimized to 20 eV to obtain selective MS² fragments, as exemplified by peptides VIQNEFTVGEE-CELETMTGEK and AIGLPEELIQK (Figure S1). With these tryptic peptides, a targeted proteomics method was established to monitor hL-FABP in a 15 min LC gradient, with parallel-reaction monitoring (PRM) mode (Figure 2A). This method was then applied to monitor endogenous hL-FABP in HepG2 cells, and the two selected peptides were detected under both MS¹ and MS², with >1000 signal-to-noise (S/N) ratios (Figure 2B). nLC−MS/MS has been widely used for proteomics due to its enhanced ionization efficiency,⁴¹ but it is also prone to retention time and instrumental response fluctuations. This study took advantage of the high abundance of hL-FABP, using regular HPLC to achieve higher reproducibility and throughput (Figure 2C). Great reproducibility was achieved as the relative standard deviation of three independent runs is less than 5%, and the standard deviation of retention time is less than 0.1 min (Figure S2).

Figure 2.

Development of a fast targeted-proteomics method to analyze hL-FABP. (A) Representative chromatograms of two selected peptides from recombinant hL-FABP at both MS¹ and MS² levels. (B) Detection of hL-FABP in HepG2 cell lysates as evidenced by retention times, MS¹ and MS² spectra. (C) Targeted rather than shotgun proteomics was employed for QITSA considering its better reproducibility and shorter time. Abbreviations: PRM−parallel-reaction monitoring; DDA−data-dependent acquisition.

The QITSA method was then used to monitor the thermal stability of endogenous hL-FABP in HepG2 cell lysates after heating to 55−85 °C. Increasing amounts of hL-FABP were detected in pellets along with heating, following a sigmoid curve (Figure S3A). Denaturation of hL-FABP started at 65 °C, and complete denaturation was achieved around 75 °C, consistent with the thermal stability of recombinant hL-FABP as mentioned above. Sufficient reproducibility was achieved for three replicates (Figure S3B), as well as inter-day batches (Figure S3A). Note that label-free proteomics was used here, without any chemical labeling such as tandem mass tags or dimethyl labeling,^42,43 illustrating the high reproducibility of HPLC−MS/MS. Compared to antibody-or fluorescent dye-based methods widely used for biophysics studies, the use of LC−MS offers the sensitivity and selectivity advantage to directly quantify endogenous L-FABP from tissues and cell lysates for in situ chemical−protein interaction studies.

QITSA Is HighLy Quantitative in Determining Binding Affinities of PFASs.

The QITSA method was then applied to monitor the thermal stability of endogenous hL-FABP in HepG2 after incubation with PFOS. Consistent with the results from recombinant hL-FABP, the thermal stability of endogenous hL-FABP in HepG2 cell lysates was significantly enhanced by PFOS (Figure 3B). Such enhancement was evidenced by both peptides VIQNEFTVGEECELETMTGEK and AIGLPEELIQK selected for targeted proteomics (Figure S4). To further test the selectivity of the QITSA method, the same tryptic peptides were injected into the nLC−MS/MS for shotgun proteomics. Among 2695 proteins detected, FABP1 (L-FABP) was identified as the top hit (see Materials and Methods for details), together with two other putative protein targets including serum albumin (ALBU) and acryl carrier protein (NDUFAB1) (Figure 3C). Notably, all three proteins are lipid binding proteins, which is not surprising considering the similar structures between PFOS and endogenous fatty acids. Indeed, serum albumin has been well-documented to be bound by PFOS with relatively high affinities.⁴⁴ This is consistent with the enhanced thermal stability of serum albumin by PFOS (Figure 3D). This confirms that PFOS selectively enhances the thermal stability of its protein targets including hL-FABP.

Figure 3.

In situ determination of the interactions of PFASs with endogenous hL-FABP in HepG2. (A) Structures of three perfluoroalkyl carboxylic acids (PFCAs), and three perfluoroalkyl sulfonates (PFSAs) selected for the test. K_d values are shown for each PFAS, determined by Zhang et al. with a fluorescence displacement method and recombinant hL-FABP.³⁴ (B) Enhancement of thermal stability of hL-FABP in HepG2 cell lysates by different PFASs. The percentage of denatured proteins from each treatment is determined by targeted proteomics after centrifugation (n = 2). (C) Shotgun proteomics reveals L-FABP as the top hit of PFOS in HepG2 cell lysates, together with serum albumin and NDUFAB1. (D) Thermal stability of serum albumin in HepG2 is enhanced by PFOS.

To further test the breadth of the QITSA method, it was applied to five other PFASs with differential hL-FABP binding affinities (K_d ≈ 16.2 to >1000 μM) as determined by a classic fluorescence displacement bioassay with recombinant hL-FABP.³⁴ Three carboxylates (PFBA, PFOA, and PFNA) and two sulfonic acids (PFBS and PFHxS) with different fluoroalkyl chain lengths were selected (structures are shown in Figure 3A). A minor but significant thermal stability enhancement was observed for PFBS, a weak hL-FABP ligand (K_d = 1034 μM), while a larger thermal shift was observed for PFHxS with a higher hL-FABP binding affinity (K_d = 85.7 μM) (Figure 3B). This demonstrates that QITSA efficiently monitors the enhancement of thermal stability of hL-FABP even by weak ligands (e.g., PFBS), while a more pronounced thermal shift was observed for stronger interactions. A similar binding affinity-dependent thermal enhancement was also observed for three perfluoroalkyl carboxylates. Both PFNA (K_d = 16.2 μM) and PFOA (K_d = 50.4 μM) largely stabilized hL-FABP, but no thermal stability was observed for PFBA (K_d > 1000 μM).

To further investigate if the K_d-dependent thermal shift, as mentioned above, is quantitative, the thermal shift (ΔT_m) of hL-FABP was quantitatively defined by calculating the temperatures to denature 50% of hL-FABP in control and treatments (Figure 4A), as previously adopted in the CETSA method.^22,23 The ΔT_m values of six PFASs were calculated to be −0.67 ± 0.68, 0.35 ± 0.66, 4.43 ± 0.74, 5.42 ± 0.43, 6.37 ± 0.67, and 8.04 ± 1.81 °C for PFBA, PFBS, PFHxA, PFOA, PFOS, and PFNA, respectively, following the order of their binding affinities to hL-FABP.

Figure 4.

QITSA chemical proteomics distinguishes binding affinity and binding mode. (A) ΔT_m is used to quantify the thermal shift of proteins by PFASs. The temperatures to denature 50% of hL-FABP in HepG2 cell lysates from control (i.e., DMSO) or PFAS treatments are calculated by dose−response curve fittings. (B) ΔT_m values of six PFASs are negatively correlated with (R² = 0.98) their binding affinities (K_d) determined by an independent, recombinant protein-based assay.³⁴ Palmitic acid (PA) derives from the regression curve. (C) Palmitic acid shows weaker enhancement of the thermal stability of hL-FABP in HepG2 cell lysates compared to PFOS.³⁴ (D) For PFAS binding to hL-FABP with the same binding mode, the QITSA method quantifies their binding affinities; as for palmitic acid, the first palmitic acid enters the interior binding site and induce conformation changes followed by the binding of the second palmitic acid. The different binding mode of palmitic acid from PFASs may induce different unfolding states (K_u2).

According to thermodynamics calculation, ΔT_m is inversely related to log-transformed K_d values of PFASs, as described by eq 7 (see the SI for details):

$$\ln \left(K_{\text{d}}\right)=\ln \left(\frac{\left[L_{\text{F}}\right]}{K_{\text{U}}-1}\right)\approx a\times \Delta T_{\text{m}}+\ln \left[L_{\text{F}}\right]$$ (7)

where K_U is the equilibrium constant to unfold protein, [L_F] is the concentration of a free ligand, and “a” is a combined term related to the entropy and enthalpy of L-FABP. Note that 100 μM PFASs ([L_F]) was used to completely saturate hL-FABP for thermal stability determination to avoid the confounding impacts from free proteins. The relationships between K_d and ΔT_m could be applied to the environmentally relevant concentration of PFOS for toxicity postulations.

After thermodynamics calculation, strong inverse correlations were found between ΔT_m and log-transformed K_d values of all six PFASs (R² = 0.98, Figure 4B). Particularly, a differential thermal shift was observed between PFOA (ΔT_m = 5.42 ± 0.43 °C) and PFNA (ΔT_m = 8.04 ± 1.81 °C), even if their K_d values are only differentiated by 3.1-fold. This demonstrates that QITSA is a quantitative method in determining the binding affinities of PFASs toward endogenous hL-FABP. Note that QITSA provides a stronger correlation than current in silico strategies including published molecular dynamics (MD) results⁴⁵ (Figure S5) in quantifying K_d values of PFAS (R² = 0.98 with QITSA vs R² = 0.72 with MD), supporting QITSA as a modest-throughput and accurate experimental method to evaluate the interactions between chemical contaminants and their protein targets.

QITSA Is Able to Differentiate Binding Modes.

Two binding sites exist on hL-FABP, but only one of the binding sites is accessible for PFASs.⁴⁶ This is different from the binding modes of fatty acids, the endogenous ligands of L-FABP, by occupying both two binding sites with differential affinities.⁴⁷ To investigate if the QITSA method could distinguish binding modes, HepG2 cell lysates were spiked with palmitic acid, the most abundantly found saturated fatty acid with resolved crystal structures complexed with hL-FABP.⁴⁷ Significant thermal stability enhancement by palmitic acid was observed, with a ΔT_m value of 2.86 ± 0.44 °C, 2.2-fold smaller than PFOS (Figure 4C). This is unexpected as the binding affinity of palmitic acid (K_d = 0.63 μM) to hL-FABP is 28-fold higher than PFOS (K_d = 18.5 μM) as determined by classic bioassays.³⁴ When palmitic acid is plotted with PFASs according to K_d and ΔT_m, an obvious deviation from the log-linear regression was observed (Figure 4B).

Previous crystal structure and NMR studies^47,48 reported that the first fatty acid located in the interior binding cavity induces conformation changes of hL-FABP, while the second fatty acid at the exterior binding site does not. Together with the information, a mechanistic model was proposed (Figure 4D) where the conformation change induced by the palmitic acid paves a different path to unfolding hL-FABP at a high energetic transition state (i.e., partially unfolded), which shifts equilibrium toward denaturation and hence smaller ΔT_m. Since PFASs only bind to a single binding site,^34,46 the holo hL-FABP remains the same as apoprotein to be denatured, leading to a bigger thermal shift. The results demonstrate that QITSA can distinguish binding modes of chemicals if they induce differential protein conformation changes, together with their binding affinities. Protein conformation changes induced by chemicals are critical in determining the biological functions of proteins, e.g., nuclear receptors and L-FABP, by recruiting different co-factors.⁴⁹⁻⁵¹ Thus, the application of QITSA to investigate the binding schemes, including binding affinity and binding modes, between PFASs and L-FABP, is of great interest.

Large Interspecies Variations on PFOS-FABP Interactions.

The QITSA method was applied to investigate the potential interspecies variations on binding schemes between PFOS and L-FABP across rat (Rattus norvegicus), mouse (Mus musculus), and zebrafish (Danio rerio), three major model animals used for toxicology studies. Liver tissue samples from the rat, mouse, and zebrafish were homogenized and lysed to extract endogenous L-FABP. Slightly different from the QITSA protocol used for human cells due to the lack of recombinant proteins as standards when we were trying to develop the LC−MS method, shotgun proteomics was used to identify L-FABP. Then, retention time, exact MS¹, and fragment spectra information were transferred to establish a PRM-based targeted proteomics method (see SI). With this strategy, L-FABP was detected in all species with sufficient S/N ratios (Figures S6 and S7).

The QITSA method was then applied to characterize the thermal stability of L-FABP in liver tissue lysates from the mouse, rat, and zebrafish after spiking with PFOS (Figure 5A). As a highLy conserved protein, L-FABP from the four species showed similar thermal stability with ΔT_m at 73−76 °C in the absence of PFOS. This confirmed the similar heat capacity (ΔC_p) of homologue proteins across species and the use of the QITSA method for quantitative assessment of interactions. Compared to the large enhancement of the thermal stability of hL-FABP by PFOS (ΔT_m = 7.32 ± 1.81 °C), only a minor thermal shift was detected for mL-FABP (ΔT_m = 2.01 ± 0.61 °C) and rL-FABP (ΔT_m = 1.90 ± 0.73 °C). No significant thermal stability enhancement was detected for zL-FABP (ΔT_m = −0.92 ± 0.62 °C). This demonstrates large interspecies variations on in situ thermal enhancement of LFABP by PFOS across the human, mouse, rat, and zebrafish. The order of thermal stability enhancement of L-FABPs by PFOS across species is accordant with its half-lives in humans (3.3−5.4 years),^52,53 mice (30−36 days),⁵⁴ rats (24−83 days),^55,56 and zebrafish (9.3 days).⁵⁷ This suggested that L-FABP might partly contribute to the interspecies variations on PFOS toxicokinetics, together with well-documented organic anion transporters.⁵⁸

Figure 5.

Large interspecies variations on interactions between PFOS and L-FABPs across human, rat, mouse, and zebrafish. (A) Thermal shift of L-FABP induced by PFOS is most pronounced in humans followed by the mice, rats, and zebrafish before (black) or after dialysis (red). (B) Thermal stability of rL-FABP in liver tissues (black) is reduced after removal of endogenous ligands by dialysis (red) or co-incubation with rL-FABP (blue). (C) Validation of the weakly enhanced thermal stability of recombinant rL-FABP and mL-FABP.

Interspecies Variations Are Only Partially Explained by the Presence of Endogenous Ligands.

The L-FABP has been reported to bind endogenous metabolites including cholesterol, vitamins, and fatty acids.⁵⁹ To investigate if endogenous ligands contribute to the small thermal stability enhancement of L-FABP by PFOS in nonhuman species, dialysis was used to remove endogenous ligands from L-FABP before heat treatment. The thermal shift curve of hL-FABP in HepG2 cells after dialysis is similar to crude extracts (Figure 5A). This implies that the majority of hL-FABP in HepG2 cells is free, not bound by endogenous ligands. The same dialysis protocol was applied to rat liver tissue extracts. The thermal stability curve of rL-FABP in the absence of PFOS was downshifted, clearly demonstrating the occupation of L-FABP by endogenous ligands in rat liver tissues. Previous studies predicted that a large fraction of L-FABP should be bound by endogenous lipids,³⁴ and this study provides the first experimental evidence by directly monitoring in situ interactions between L-FABP and endogenous ligands. To further confirm this, rat tissue lysates were co-incubated with recombinant hL-FABP. A downshift of thermal stability of rL-FABP was also observed, likely due to the competition of endogenous ligands by high concentrations of hL-FABP (Figure 5B). A similar downshift of denaturation of mL-FABP was also observed, although such occupation by endogenous ligands was not observed for zL-FABP. Thus, removal of ligands by dialysis is essential to exclusively link ΔT_m to PFOS binding schemes.

Further, QITSA was performed on the liver tissues from all three nonhuman species after dialysis. Consistent with the results above, the thermal denaturation curves of rL-FABP and mL-FABP were downshifted after dialysis, in the presence of PFOS, but relatively unchanged for zL-FABP (Figure 5A). The ΔT_m values of FABPs for PFOS after dialysis were determined to be 14.71 ± 2.09, 2.32 ± 1.21, 1.92 ± 0.89, and −1.00 ± 0.90 °C for the human, mouse, rat, and zebrafish FABP, respectively. The results demonstrate that while endogenous ligands impact the overall thermal stability of L-FABP, the general order of thermal enhancement is consistent before and after dialysis, as human > mouse > rat > zebrafish (Figure 5A) and large interspecies variations on intrinsic binding schemes between PFOS and L-FABP. The low thermal stability enhancement of rL-FABP or mL-FABP by PFOS was further validated by subjecting recombinant proteins expressed in E. coli to QITSA analysis, while the zL-FABP was not available. ΔT_m values of 3.91 ± 1.32 and 5.61 ± 0.56 °C were observed for recombinant rL-FABP and mL-FABP after spiking with PFOS (Figure 5C), above the 1.92 ± 0.89 and 2.32 ± 1.21 °C obtained from tissue lysates, but were still much smaller than hL-FABP (14.71 ± 2.09 °C). This confirmed the QITSA results that endogenous ligands only partially contribute to the lower thermal stability enhancement of rL-FABP or mL-FABP by PFOS, while differential intrinsic binding of PFOS plays a more important role.

Large Thermal Stability Enhancement of hL-FABP by PFOS Was Attributed to High Binding Affinity.

The L-FABP is highly conserved across fishes and mammals, but with significant sequence variations.³⁸ Thus, in comparing the protein sequence similarity of L-FABP across the four investigated species, mL-FABP shows the closest sequence identity (84.3%) to hL-FABP followed by rL-FABP with 82.7% sequence identity, whereas only 64.6% sequence identity was observed between the human and zebrafish (Figure 6A,B). The sequence similarity of L-FABP is consistent with their thermal stability shift induced by PFOS, demonstrating that intrinsic sequence variations of L-FABP may impact interactions between PFOS and L-FABP across species.

Figure 6.

Interspecies variations supported by sequence analysis, in silico molecular docking, and recombinant protein-based assays. (A) Sequence similarity of L-FABP in mice, rats, and zebrafish relative to humans. Note that the protein sequence phylogenetics of L-FABP is accordant with their ΔT_m. (B) Amino acid sequence alignment of L-FABP from four species. The high affinity binding sites to endogenous lipids are labeled in red color. (C) Correlations of K_d values of six PFAS and one palmitic acid with ΔG calculated by Autodock4. (D) Comparison between ΔG (kcal/ mol) calculated by Autodock4, and ΔT_m ( °C) for PFOS and L-FABP across four species. (E) Dose-dependent competition of fluorescence signals of 1,8-ANS (40 μM) in 1 μM L-FABP by PFOS. Fluorescence intensity was normalized to vehicle controls (DMSO). (F) Modified QITSA method was performed at 75 °C with 5 μg/mL L-FABPs and treated with PFOS over a range of concentrations (100, 50, 25, 12.5, and 6.25 μM).

As binding modes and binding affinities might both impact the thermal shift of proteins as mentioned above, molecular docking was employed to further distinguish these two factors. First, it was determined whether molecular docking can quantify the binding affinities of PFASs and palmitic acid. A positive linear regression curve (R² = 0.73 and p = 0.015) was observed between K_d and ΔG calculated by the Autodock4 (Figure 6C). Notably, the binding affinity of palmitic acid was also correctly predicted by molecular docking, which cannot be distinguished by the QITSA method due to its different binding modes. This highlights the advantages to combining QITSA and in silico docking to investigate both binding affinities and modes of chemical contaminants.

Then, molecular docking was conducted to calculate the interactions between PFOS and L-FABP across rat, mouse, and zebrafish. As crystal structures are not available for mL-FABP and zL-FABP, I-TASSER (https://zhanglab.ccmb.med.umich.edu/I-TASSER/) was used to reconstruct their 3D structures using hL-FABP (PDB ID: 3STM) as a homologue template.⁶⁰ As shown in Figure 6D, the hL-FABP was predicted to exhibit the lowest ΔG (−4.42 kcal/mol), compared to the mouse (−3.56 kcal/mol), rat (−4.09 kcal/mol), and zebrafish (−3.63 kcal/mol). This demonstrates that binding affinity should be at least partly responsible for the biggest thermal stability enhancement of hL-FABP by PFOS.

To confirm the results, we obtained recombinant mL-FABP and rL-FABP and determined their binding affinities with PFOS using the fluorescence displacement method with 1-anilinonaphthalene-8-sulfonic acid (1,8-ANS). The K_d values of ANS binding to three L-FABPs are similar, at 3.45, 4.03, and 5.39 μM for hL-FABP, mL-FABP, and rL-FABP, respectively. Consistent with the molecular docking results, while strong fluorescence competition by PFOS was observed for hL-FABP, the competition was weak for mL-FABP or rL-FABP (Figure 6E). To further confirm that this is not due to synergistic binding modes between PFOS and ANS, we used a modified QITSA method by investigating the thermal stability of LFABP at different concentrations of PFOS (Figure 6F). Strong thermal stability enhancement by increasing PFOS concentrations was observed for hL-FABP (EC₅₀ = 7.26 μM), while such concentration-dependent trends were ∼8 times weaker for mL-FABP (EC₅₀ = 34.3 μM) or rL-FABP (EC₅₀ = 57.7 μM). These results confirmed that binding affinities should be primarily responsible for the interspecies variations between human and nonhuman species. However, recombinant zL-FABP is not available, and investigation of the mechanism for its weak interaction with PFOS is warranted in future studies.

Implications.

Characterizing interspecies variations on chemical contaminant toxicities is a long-standing unresolved question. This study proposed a novel experimental strategy, by in situ and quantitative investigation of chemical−protein interactions across species. In contrast to classic bioassays, tissue samples were directly used for QITSA without the need for laborious protein expression, and thus modest throughput was achieved. With the established QITSA, four species and one chemical were tested within 1 week. While this study is focused on L-FABP, it is straightforward to apply QITSA to other chemical−protein interactions of interest even with low abundances, e.g., AhR with dioxins.

Note that two versions of the QITSA method could be applied: the concentration-dependent method is able to distinguish binding affinities, while the temperature-dependent method is able to cover both binding affinities and modes. The ability of QITSA to distinguish binding modes as well as binding affinities is attractive as both could be important for interspecies toxicities. Thus, QITSA is proposed as a high throughput and first-run method to screen interactions across species to narrow down targets of interest, and subsequent modeling and biochemistry validation (as done here) are incorporated to further elucidate the mechanisms.