Binding of Perfluorocarboxylates to Serum Albumin: A Comparison of

Dec 29, 2009 - Department of Chemistry, Union College, Schenectady, New York 12308, and Department of Civil and Environmental. Engineering, Stanford ...
2 downloads 0 Views 348KB Size
Download PDF
Anal. Chem. 2010, 82, 974–981

Binding of Perfluorocarboxylates to Serum Albumin: A Comparison of Analytical Methods Laura A. MacManus-Spencer,*,† Monica L. Tse,† Paul C. Hebert,† Heather N. Bischel,‡ and Richard G. Luthy‡ Department of Chemistry, Union College, Schenectady, New York 12308, and Department of Civil and Environmental Engineering, Stanford University, Stanford, California 94305 Perfluorochemicals are globally pervasive contaminants that are persistent, bioaccumulative, and toxic. Perfluorocarboxylic acids (PFCAs) with 8-13 carbons accumulate in the liver and blood of aquatic organisms; PFCAprotein interactions may explain this accumulation pattern. Here, the interactions between PFCAs with 8-11 carbons and serum albumin are examined using three experimental approaches: surface tension titrations, 19F NMR spectroscopy, and fluorescence spectroscopy. Surface tension titrations indicate complex formation at high (mM) PFCA concentrations. Secondary association constants ranging from 102 to 104 M-1 were determined from 19F NMR titrations at high PFCA: albumin mole ratios. Fluorescence measurements indicate that PFCA-albumin interactions alter the protein conformation at low PFCA:albumin mole ratios (up to 5:1) and suggest two binding classes with association constants around 105 and 102 M-1. While 19 F NMR and fluorescence provide both qualitative and quantitative information about PFCA-albumin interactions, surface tension provides only qualitative information. Limitations associated with instrumentation and methods require high PFCA concentrations in both surface tension and 19F NMR experiments; in contrast, fluorescence allows for analysis of a wider range of PFCA concentrations and PFCA:albumin mole ratios. Results from this study indicate that fluorescence, though an indirect method, offers a more comprehensive picture of the nature of PFCA-albumin interactions. Manufactured since the 1940s, fluorinated alkyl substances have many industrial and commercial applications, including firefighting foams, textile and paper coatings, and insecticides.1 Perfluoroalkyl acids (PFAs) are found in such products at varying concentrations and in recent years have emerged as global contaminants that are persistent,2,3 bioaccumulative,4-7 and * To whom correspondence should be addressed. Phone: (518) 388-6153. Fax: (518) 388-6795. E-mail: [email protected]. † Union College. ‡ Stanford University. (1) Kissa, E. Fluorinated Surfactants and Repellants; Marcel Dekker, Inc.: New York, 2001. (2) Hekster, F. M.; Laane, R. W. P. M.; de Voogt, P. Rev. Environ. Contam. Toxicol. 2003, 179, 99–121. (3) Schultz, M. M.; Barofsky, D. F.; Field, J. A. Environ. Eng. Sci. 2003, 20, 487–501.

974

Analytical Chemistry, Vol. 82, No. 3, February 1, 2010

potentially toxic.2,8-11 Unlike other persistent organic pollutants, PFAs do not preferentially accumulate in lipids and fatty tissue but rather in body compartments with high protein content, including the liver, kidneys, and serum.12-16 Additionally, a recent study showed that a lipid-normalized biota-sediment accumulation factor (BSAF) does not accurately describe the accumulation of PFAs in Lumbriculus variegatus from sediment.17 The ability of tissue and serum proteins to bind PFAs18-22 may explain this accumulation behavior. In the mid-1950s, researchers investigated the binding of perfluorooctanoic acid (PFOA) to bovine and human serum albumins for the purposes of protein precipitation and protection against denaturation.23-25 Upon later recognition of PFAs as persistent global contaminants, investigators returned (4) Giesy, J. P.; Kannan, K. Environ. Sci. Technol. 2001, 35, 1339–1342. (5) Olsen, G. W.; Huang, H.-Y.; Helzlsouer, K. J.; Hansen, K. J.; Butenhoff, J. L.; Mandel, J. H. Environ. Health Perspect. 2005, 113, 539–545. (6) Martin, J. W.; Smithwick, M. M.; Braune, B. M.; Hoekstra, P. F.; Muir, D. C. G.; Mabury, S. A. Environ. Sci. Technol. 2004, 38, 373–380. (7) Martin, J. W.; Whittle, D. M.; Muir, D. C. G.; Mabury, S. A. Environ. Sci. Technol. 2004, 38, 5379–5385. (8) Butenhoff, J.; Costa, G.; Elcombe, C.; Farrar, D.; Hansen, K.; Iwai, H.; Jung, R.; Kennedy, G., Jr.; Lieder, P.; Olsen, G.; Thomford, P. Toxicol. Sci. 2002, 69, 244–257. (9) Jones, P. D.; Newsted, J. L.; Giesy, J. P. Organohalogen Compd. 2003, 62, 311–314. (10) Kennedy, G. L., Jr.; Butenhoff, J. L.; Olsen, G. W.; O’Connor, J. C.; Seacat, A. M.; Perkins, R. G.; Biegel, L. B.; Murphy, S. R.; Farrar, D. G. Crit. Rev. Toxicol. 2004, 34, 351–384. (11) Levitt, D.; Liss, A. Toxicol. Appl. Pharmacol. 1986, 86, 1–11. (12) Vanden Heuvel, J. P.; Kuslikis, B. I.; Van Rafelghem, M. J.; Peterson, R. E. J. Biochem. Toxicol. 1991, 6, 83–92. (13) Martin, J. W.; Mabury, S. A.; Solomon, K. R.; Muir, D. C. G. Environ. Toxicol. Chem. 2003, 22, 196–204. (14) Kannan, K.; Yun, S. H.; Evans, T. J. Environ. Sci. Technol. 2005, 39, 9057– 9063. (15) Verreault, J.; Houde, M.; Gabrielsen, G. W.; Berger, U.; Hauks, M.; Letcher, R. J.; Muir, D. C. G. Environ. Sci. Technol. 2005, 39, 7439–7445. (16) Guruge, K. S.; Taniyasu, S.; Yamashita, N.; Wijeratna, S.; Mohotti, K. M.; Seneviratne, H. R.; Kannan, K.; Yamanaka, N.; Miyazaki, S. J. Environ. Monit. 2005, 7, 371–377. (17) Higgins, C. P.; Luthy, R. G. Environ. Sci. Technol. 2007, 41, 3254–3261. (18) Luebker, D. J.; Hansen, K. J.; Bass, N. M.; Butenhoff, J. L.; Seacat, A. M. Toxicology 2002, 176, 175–185. (19) Vanden Heuvel, J. P.; Kuslikis, B. I.; Peterson, R. E. Chem.-Biol. Interact. 1992, 82, 317–328. (20) Han, X.; Snow, T. A.; Kemper, R. A.; Jepson, G. W. Chem. Res. Toxicol. 2003, 16, 775–781. (21) Jones, P. D.; Hu, W.; De Coen, W.; Newsted, J. L.; Giesy, J. P. Environ. Toxicol. Chem. 2003, 22, 2639–2649. (22) Messina, P. V.; Prieto, G.; Ruso, J. M.; Sarmiento, F. J. Phys. Chem. B 2005, 109, 15566–15573. (23) Ellenbogen, E.; Maurer, P. H. Science 1956, 124, 266–267. (24) Klevens, H. B.; Ellenbogen, E. Discuss. Faraday Soc. 1954, 18, 277–288. (25) Nordby, G. L.; Luck, J. M. J. Biol. Chem. 1956, 219, 399–404. 10.1021/ac902238u  2010 American Chemical Society Published on Web 12/29/2009

to protein binding to explain accumulation patterns, elimination rates, and toxicity. Luebker et al., using a competitive binding assay with a fluorescent ligand, found that PFOA and perfluorooctanesulfonic acid (PFOS) interfere with fatty acid binding by liver fatty acid binding protein (L-FABP) in the low micromolar range.18 Vanden Heuvel et al. noted that 80% of 14C-labeled perfluorodecanoic acid ([14C]PFDA, 100 µM) was bound to bovine serum albumin (BSA; 80 µM) after 60 min.19 Reporting binding constants on the order of 103 M-1, Han et al. suggested that more than 90% of PFOA would be bound to serum albumin in human blood.20 In direct binding studies, Jones et al. showed that PFOS was more than 80% bound at BSA concentrations above about 1 g/L.21 From such experiments, binding to proteins is likely important in the uptake and tissue distribution of PFAs.10 Serum albumin26,27 is an ideal protein to use in studies of PFCA-protein binding. In humans, serum albumin concentrations range from 35 to 50 g/L, making it the most abundant protein in the blood;27 albumins are also found in tissues and bodily secretions.28 Serum albumin binds a variety of molecules, particularly fatty acids, and transports them between tissues and organs.27 As the sequences of human serum albumin (HSA) and BSA are 76% conserved,27 BSA is commonly substituted for HSA in experiments due to its availability and lower cost. Several analytical methods have been used to qualitatively and quantitatively study the binding of anionic organic ligands, including fatty acids and some PFAs, to proteins such as serum albumin. Traditional methods include equilibrium dialysis29 and surface tension;30-32 more recent studies have used calorimetry,33 mass spectrometry,20,21 absorbance spectroscopy,22,34 fluorescence spectroscopy,35,36 ion-selective electrodes,37,38 ζ-potential measurements,34 and NMR spectroscopy.20,39,40 Binding affinities, measured as the association constant, Ka, have been reported in the literature for PFOA binding to bovine, human, and rat serum albumins and range from 102 to 105 M-1.20,24,38 The number of PFOA molecules bound per protein, n, ranges from 1 to ∼250020,24,25,38 depending on the conditions and the method of analysis. Given the wide variations in binding values reported in the literature, the purpose of this study is to compare three simple analytical methods to determine the nature and strength of (26) (27) (28) (29) (30) (31) (32) (33) (34) (35) (36) (37) (38) (39) (40)

Kragh-Hansen, U. Pharm. Rev. 1981, 33, 17–53. Peters, T., Jr. Adv. Protein Chem. 1985, 37, 161–245. He, X. M.; Carter, D. C. Nature 1992, 358, 209–215. Green, H. O.; Moritz, J.; Lack, L. Biochim. Biophys. ActasLipids Lipid Metab. 1971, 231, 550–552. Tribout, M.; Paredes, S.; Gonzalez-Manas, J. M.; Goni, F. M. J. Biochem. Biophys. Methods 1991, 22, 129–133. Knox, W. J., Jr.; Parshall, T. O. J. Colloid Interface Sci. 1970, 33, 16–23. Nishikido, N.; Takahara, T.; Kobayashi, H.; Tanaka, M. Bull. Chem. Soc. Jpn. 1982, 55, 3085–3088. Deep, S.; Ahluwalia, J. C. Phys. Chem. Chem. Phys. 2001, 3, 4583–4591. Sabin, J.; Prieto, G.; Gonzalez-Perez, A.; Ruso, J. M.; Sarmiento, F. Biomacromolecules 2006, 7, 176–182. Lissi, E.; Abuin, E.; Lanio, M. E.; Alvarez, C. J. Biochem. Biophys. Methods 2002, 50, 261–268. De, S.; Girigoswami, A.; Das, S. J. Colloid Interface Sci. 2005, 285, 562– 573. Rendall, H. M. J. Chem. Soc., Faraday Trans. 1 1976, 72, 481–484. Messina, P.; Prieto, G.; Dodero, V.; Ruso, J. M.; Schulz, P.; Sarmiento, F. Biopolymers 2005, 79, 300–309. Kronis, K. A.; Carver, J. P. Biochemistry 1982, 21, 3050–3057. Sauter, N. K.; Bednarski, M. D.; Wurzburg, B. A.; Hanson, J. E.; Whitesides, G. M.; Skehel, J. J.; Wiley, D. C. Biochemistry 1989, 28, 8388–8396.

Figure 1. Structures and names of the perfluorocarboxylic acids included in this study.

perfluorocarboxylic acid (PFCA) binding to bovine and human serum albumins. Due to the lack of published data for PFCAs other than PFOA, four PFCAs of varying chain length have been selected for this study. PFCAs with 8-11 carbons were chosen due to their environmental prevalence and aqueous solubility. The structures of these PFCAs are shown in Figure 1, with each represented as the carboxylate anion, the relevant form at physiological pH.41,42 MATERIALS AND METHODS Chemicals. PFOA (99.9%), perfluorodecanoic acid (PFDA; 99.9%), and perfluoroundecanoic acid (PFUnA; 102.2%) were from Aldrich Chemical Co. (Milwaukee, WI). Perfluorononanoic acid (PFNA; 99.2%) was from Fluka through Sigma-Aldrich (St. Louis, MO). Anhydrous sodium hydrogen phosphate (Na2HPO4) and sodium dihydrogen phosphate monohydrate (NaH2PO4 · H2O) were from Fisher Chemical (Fairlawn, NJ). Deuterium oxide (D2O; >99%) was from Cambridge Isotope Laboratories (Andover, MA). BSA (>99% by agarose gel electrophoresis, essentially fatty acid and globulin free) and HSA (>99% by agarose gel electrophoresis, essentially fatty acid and globulin free) were from Sigma-Aldrich. The molecular weight used for BSA was 66 430,43 and that for HSA was 66 248.44 Surface Tension Measurements. Surface tension measurements were performed using a Fisher semiautomatic model 21 Tensiomat tensiometer with a 6 cm platinum iridium du Nou¨y ring. Stock solutions of PFCAs were prepared in Milli-Q water or 50 mM pH 7.4 sodium phosphate buffer in 50 mL polypropylene centrifuge tubes. A sonicating bath (40 °C) was used to assist in the dissolution of the PFCAs. Stock solutions of BSA, which were prepared in 50 mM pH 7.4 sodium phosphate buffer and stirred at 4 °C until dissolved, were used immediately and prepared fresh daily. Dilutions of PFCAs in Milli-Q water or buffer were prepared in 50 mL polypropylene centrifuge tubes and were analyzed immediately. PFCA-BSA mixtures were prepared in 50 mL polypropylene centrifuge tubes, mixed well, allowed to sit overnight at 4 °C prior to analysis, and allowed to warm to room temperature before analysis. All surface tension measurements were made in polypropylene sample containers. The platinum iridium ring was cleaned with 50:50 methanol-acetone and flamed between each measurement. 19 F NMR Spectroscopy. 19F NMR spectra were acquired on a Varian Inova 600 MHz spectrometer using a 1H/19F{X} (inverse proton/fluorine detection, broad-band decoupling) 5 mm probe with a z-axis gradient or a Bruker AMX 500 MHz spectrometer using a dual 1H/13C 5 mm probe tuned to the (41) Burns, D. C.; Ellis, D. A.; Li, H.; McMurdo, C. J.; Webster, E. Environ. Sci. Technol. 2008, 42, 9283–9288. (42) Goss, K.-U.; Arp, H. P. H. Environ. Sci. Technol. 2009, 43, 5150–5151. (43) Hirayama, K.; Akashi, S.; Furuya, M.; Fukuhara, K. Biochem. Biophys. Res. Commun. 1990, 173, 639–646. (44) Putnam, F. W., Ed. The Plasma Proteins, 2nd ed.; Academic Press: New York, 1975; pp 133-181.

Analytical Chemistry, Vol. 82, No. 3, February 1, 2010

975

Figure 2. Left: Surface tension curves for PFNA in 50 mM pH 7.4 sodium phosphate buffer in the absence of BSA (solid line) and presence of 10 µM BSA (dashed line) or 25 µM BSA (dotted line). Arrows indicate the direction of increasing BSA concentration. Right: At a constant PFNA concentration of 2 mM, the addition of BSA increases the surface tension in a concentration-dependent manner. Duplicate sets of data are shown. Data are fit to eq 1. Error bars represent the standard deviation of five replicate measurements. 19

F frequency (470.5 MHz). Samples of BSA, HSA, PFOA, and PFNA were prepared in 50 mM pH 7.4 sodium phosphate buffer containing 10% D2O. In a given experiment, the concentration of protein was held constant at 12, 23, 35, or 47 µM, while the PFCA concentration was varied from 180 to 5000 µM (PFOA) or from 182 to 1820 µM (PFNA). To determine the delay time, spin-lattice relaxation times (T1) for resonances of PFOA and PFNA in buffer were measured by the inversionrecovery technique and were found to be around 1 s. Onedimensional (1D) 90° pulse spectra were collected with a 25 kHz sweep width, a 5 s pulse delay, 16K data points, and 64-2048 scans. Two-dimensional (2D) 19F homonuclear correlation spectroscopy (COSY) spectra were recorded with 512 complex t1 increments, 1024 t2 points, and 16 scans for each FID. Chemical shifts were referenced to the fluorines on the carbon adjacent to the carboxylate group of the PFCA, which was set to 0.0 ppm (Figure 3). This assignment is different from a previous assignment by Han et al. for PFOA,20 who we believe misassigned this resonance to fluorines near the middle of the perfluoroalkyl chain. 19F NMR resonances for PFNA referenced to the internal standard trifluoroacetic acid are also shown in the Supporting Information (Figure S5). Fluorescence Spectroscopy. Steady-state fluorescence spectra were acquired on a Horiba Jobin Yvon Fluorolog fluorometer or a Photon Technology International (PTI) QuantaMaster spectrofluorometer. For all experiments, the excitation wavelength was 295 nm, and the emission was monitored from 300 to 600 nm (1 nm increments, 0.1 s integration time). The excitation and emission monochromator slit widths were varied between 1 and 5 nm for different albumin concentrations but were held constant during each experiment. In all plots of albumin fluorescence emission, the fluorescence is represented as the emission peak area from 305 to 475 nm or from 305 to 570 nm (constant during each experiment). Stock solutions of PFCAs were prepared in 50 mM pH 7.4 sodium phosphate buffer as described above for surface tension experiments. Stock solutions of BSA and HSA were also prepared as described above. PFCA-protein mixtures (varying protein concentration between 1 and 10 µM and PFCA concentrations between 100 nM and 1 mM) were prepared and handled as described above. All spectra were collected using a sealed quartz cuvette. 976

Analytical Chemistry, Vol. 82, No. 3, February 1, 2010

Time-resolved fluorescence data were acquired on a PTI LS100 luminescence system. For all experiments, the excitation wavelength was 295 nm, and the emission was monitored at 350 nm. The start and end delays were set to 90 and 350 ns, respectively. The integration time was 0.1 s, and 10 scans were averaged for each measurement. A cuvette containing buffer was used to account for the lamp pulse profile by measuring scatter at 295 nm. The same samples analyzed by steady-state fluorescence were used to measure fluorescence lifetimes. Lifetime data were fit to a biexponential decay. Software. Kaleidagraph (Synergy Software) graphical analysis software was used to prepare all data plots and fit data to linear and nonlinear models. Safety Considerations. Standard safety procedures were used when handling all materials. RESULTS AND DISCUSSION Surface Tension. The surface active nature of PFCAs was exploited to investigate their binding to BSA. If PFCAs bind to BSA, the addition of BSA to a PFCA solution should result in an increase in the surface tension as the concentration of free PFCA in solution is reduced. However, BSA itself is surface active, and changes in surface tension upon addition of BSA are dependent on the concentration of PFCA tested (Figure 2, left). For example, the addition of BSA reduced the surface tension in an additive manner at PFNA concentrations well below its critical micelle concentration (cmc), measured here (Figure S1, Supporting Information) to be about 2 mM (Figure 2, left). BSA had little effect over the PFNA concentration range where surface tension decreases proportionally with increasing PFNA concentration. For PFNA concentrations at or above its cmc, the addition of BSA caused a concentration-dependent increase in surface tension (Figure 2, right). This suggests that the addition of BSA disrupts normal surfactant behavior, possibly through the formation of surfactant-protein aggregates, as has been described for other surfactants.36,45 Similar results were obtained for PFOA, PFDA, and PFUnA (Figure S2, Supporting Information). To further study the formation of PFCA-albumin aggregates, PFCA concentrations at or just above the cmc were selected for (45) Rieger, M. M., Rhein, L. D., Eds. Surfactants in Cosmetics, 2nd ed.; Dekker: New York, 1997.

titration by BSA. Data from these titrations were fit to a logistic function: γ ) γ1 +

γ2 - γ1 log [BSA] 1+ log [BSA]0

(

(1)

)

p

where γ1 and γ2 are the initial and final surface tension values, respectively, [BSA]0 is the BSA concentration at which the inflection point occurs, and p is related to the slope of the curve. Table 1 summarizes the experimental values of [BSA]0 and p for the four PFCAs tested. Because each PFCA was tested at a different concentration at or just above its cmc, the ratio of [PFCA]t to [BSA]0 was computed; these ratios indicate that 50% of the total change in surface tension occurred by a PFCA:BSA mole ratio of about 4:1 (Table 1). There is also a general increase in the slope factor, p, with increasing length of the perfluorocarbon tail, indicating a greater change in surface tension per unit increase in the BSA concentration as the perfluorocarbon tail length increases. The observed relationship between p and perfluorocarbon chain length suggests the strength and/or extent of the PFCA-albumin associations observed at these mole ratios increases with the chain length. Notably, PFNA does not follow these trends (Table 1); the source of this deviation is currently unknown. 19 F NMR Spectroscopy. 19F NMR is a useful method to study the binding of PFCAs to proteins due to the unique signals that arise from the perfluorocarbon tail and the absence of any signal from the protein. PFCA binding is accompanied by a shift in and broadening of the fluorine resonance peaks (representative data for PFNA shown in Figure 3). The resonance most sensitive to binding is that due to the fluorines on the carbon adjacent to the carboxylate headgroup, labeled as R in Figure 3. The assignment of this resonance and the others is based on a two-dimensional 19F COSY experiment (Figure S3, Supporting Information). It has previously been shown that the predominant correlations in 19F-19F COSY experiments are those that arise from the through-space coupling of fluorines separated by four bonds (i.e., two carbons apart),46,47 as opposed to the correlations observed in 1H-1H COSY experiments that are due to protons on adjacent carbons. 19F-19F COSY has previously been used to correctly assign the 19F NMR resonances of PFOA;48 the assignment of the R resonance of PFNA is based on this knowledge. The theory underlying the use of NMR to quantitatively characterize ligand-protein binding has been described preTable 1. Summary of Data from PFCA-BSA Surface Tension Titrations PFCA

Na

[PFCA]tb (mM)

[BSA]0c (mM)

p

[PFCA]t/ [BSA]0

PFOA PFNA PFDA PFUnA

1 3 1 2

10 2.0 1.0 0.44

2.5 ± 0.1 0.20 ± 0.04 0.24 ± 0.02 0.091 ± 0.009

16 ± 1 14 ± 4 20 ± 4 24 ± 5

4.0 10 4.2 4.8

a Number of experiments conducted. b Total PFCA concentration for the experiment. c The error given is the standard error (95% confidence) from Kaleidagraph (when only one experiment conducted) or the propagated standard error for multiple data sets (when two or more experiments conducted).

viously.39,40 Ligand resonances can shift and broaden due to exchange of the ligand between a binding site on the protein and the solvent. Under conditions in which the fraction of bound ligand is relatively small, a reasonable assumption given the high PFCA: protein mole ratios required in the 19F NMR experiments, eq 2 describes the change in the chemical shift of a sensitive resonance upon binding to protein:

[PFCA]t )

n[protein]t ∆δBapp - Kd ∆δ

(2)

where [PFCA]t and [protein]t are the total PFCA and protein concentrations in the system, respectively, ∆δ is the change in the chemical shift of the monitored resonance at a given [PFCA], ∆δBapp is the apparent change in the chemical shift of the monitored resonance in the bound state, and n is the number of PFCA molecules bound per protein molecule. The value of Kd is then extracted as the negative y-intercept of a plot of [PFCA]t vs 1/∆δ, as is shown for PFNA binding to BSA in Figure 3; the average measured Kd value corresponds to an association constant, Ka, of (8 ± 5) × 103 M-1. 19F NMR titration data for PFOA binding to BSA and for PFNA binding to BSA and HSA are summarized in Table 2 (see Figure S4, Supporting Information, for data plots). These results show that PFNA binds more strongly to BSA than PFOA, which suggests a positive correlation between binding strength and PFCA chain length. The data also show that PFNA appears to bind to both BSA and HSA with similar affinity. Fluorescence Spectroscopy. PFCA-albumin interactions were studied by fluorescence spectroscopy using the native tryptophan (Trp) fluorescence of the proteins. BSA contains two Trp residues (212 and 134),27 whereas HSA contains only one (Trp-214), the conserved residue.49 For both proteins, binding resulted in a decrease in fluorescence intensity and a shift in the wavelength of maximum fluorescence emission, λEM,max. Figure 4 shows representative data for PFOA and PFNA binding to BSA (see Figure S6, Supporting Information, for PFNA binding to HSA). At low PFCA:albumin mole ratios (as high as 5:1), λEM,max remains essentially constant. However, a clear blue shift is observed at higher mole ratios up to the cmc (Figure 4c,d for BSA; Figure S6, Supporting Information, for HSA), indicating a change in the environment around the Trp residue(s). Trp residues inside proteins are characterized by a shorter wavelength λEM,max, around 330 nm, while those on or near the surface of a protein are characterized by a longer wavelength λEM,max.50 In addition, the fluorescence quantum yield of a Trp residue may be as low as 0.07-0.10 in a nonpolar interior site and as high as 0.32-0.44 in a more polar near-surface, but not completely exposed, site.50 The blue shift and decrease in intensity thus suggest the Trp transitions from a more polar to a less polar environment.51 (46) Petrakis, L.; Sederholm, C. H. J. Chem. Phys. 1961, 35, 1243–1248. (47) Mallory, F. B. J. Am. Chem. Soc. 1973, 95, 7747–7752. (48) Buchanan, G. W.; Munteanu, E.; Dawson, B. A.; Hodgson, D. Magn. Reson. Chem. 2005, 43, 528–534. (49) Dugaiczyk, A.; Law, S. W.; Dennison, O. E. Proc. Natl. Acad. Sci. U.S.A. 1982, 79, 71–75. (50) Burstein, E. A.; Vedenkina, N. S.; Ivkova, M. N. Photochem. Photobiol. 1973, 18, 263–279. (51) Lakowicz, J. R. Principles of Fluorescence Spectroscopy, 2nd ed.; Kluwer Academic/Plenum Publishers: New York, 1983.

Analytical Chemistry, Vol. 82, No. 3, February 1, 2010

977

Figure 3. 19F NMR data for the titration of BSA with PFNA in 50 mM pH 7.4 sodium phosphate buffer. Example spectra are shown on the left, along with the structure of PFNA illustrating the resonance monitored (R). Data from the titrations of 12 (circles), 23 (squares), 35 (tilted squares), and 47 (triangles) µM BSA with PFNA are plotted. As described in the text, the intercept of a plot of [PFNA] vs 1/∆δ is equal to -Kd. Table 2. Data from

19

F NMR Titration Experiments Kaa (M-1)

PFCA

BSA

HSA

PFOA PFNAb

(6.3 ± 0.8) × 102 (80 ± 50) × 102

(200 ± 300) × 102

a The error given is the standard error (95% confidence) from Kaleidagraph. b The error given for the PFNA-BSA titration is the standard deviation from four measurements at different BSA concentrations.

At very high PFCA:albumin mole ratios, where the PFCA concentration exceeds its cmc, λEM,max increases, eventually returning to its initial value (Figure 4c,d for BSA; Figure S6, Supporting Information, for HSA). As is evident for PFOA, a small amount of the initial fluorescence intensity is also recovered at the highest PFOA concentration (Figure 4b). On the basis of measured fluorescence quantum yields of Trp residues in a variety of proteins, fully exposed residues have fluorescence quantum yields that are higher than those of residues in nonpolar interior sites but lower than those of near-surface residues.50 The observed red shift and slight increase in intensity thus signify a transition of the Trp residue(s) from a nonpolar to a more polar environment. The results presented here for BSA and HSA are consistent with conformational changes in the protein upon PFCA binding that cause a change in the environment surrounding the Trp residue(s). The initial transition is from a somewhat polar environment near the surface of the protein to the nonpolar interior. The subsequent red shift and slight increase in fluorescence intensity indicate complete exposure of the Trp residue(s) to the aqueous environment due to PFCA-induced protein unfolding at high PFCA concentrations. Similar solvatochromic effects have been observed for the binding to BSA of other anionic surfactants, such as octyl, decyl, and dodecyl sulfate,52,53 and would be expected from the conformational changes caused by fatty acids binding to albumin.26,54 In addition to qualitative observations supporting conformational changes upon PFCAs binding to serum albumin, the (52) Moriyama, Y.; Ohta, D.; Hachiya, K.; Mitsui, Y.; Takeda, K. J. Protein Chem. 1996, 15, 265–272. (53) Steinhardt, J.; Krijn, J.; Leidy, J. G. Biochemistry 1971, 10, 4005–4015. (54) Spector, A. A. J. Lipid Res. 1975, 16, 165–179.

978

Analytical Chemistry, Vol. 82, No. 3, February 1, 2010

fluorescence data may be quantified to estimate association constants for the PFCA-albumin complexes. Others have quantified the binding of surfactants to albumins by monitoring the native protein fluorescence,35,36 though with little to no attention paid to the origins of the observed fluorescence changes. For example, in this study, a decrease in the fluorescence lifetimes of BSA and HSA is observed in the presence of PFCAs (shown for PFNA and PFUnA with BSA in Figure S7, Supporting Information). By traditional definitions, a decrease in lifetime is the effect of collisional quenching, and modeling of the system requires consideration of both time-resolved measurements to characterize collisional quenching and steady-state data to characterize the combined effects of collisional and static quenching.51 However, for a dynamic protein such as serum albumin,27 many factors affect the probability of fluorescence.50 Feldman et al. offer a more appropriate model in which a decrease in fluorescence lifetime is not necessarily associated with collisional quenching, but signifies any dynamic process, such as resonance energy transfer, collisional quenching, or the formation of a nonfluorescent excitedstate complex with a lifetime shorter than that of the uncomplexed fluorophore.55 As the data presented here are consistent with the transition of the Trp residue(s) in albumin to an interior site in the protein, neither collisional quenching by PFCAs nor the formation of nonfluorescent Trp-PFCA complexes satisfactorily explains the observed decrease in fluorescence intensity. Instead, we propose that the transition of the Trp residue(s) to an interior site upon PFCA binding increases the probability of quenching by dynamic processes within the protein. Assuming these dynamic processes are solely responsible for the decrease in fluorescence intensity and the rates of these processes are proportional to PFCA binding, the following two-site Stern-Volmer model may be used: f2 f1 F + ) F0 1 + K1[PFCA] 1 + K2[PFCA]

(3)

where F0 and F are the fluorescence intensities in the absence and presence of PFCA, respectively, f1 and f2 are the fractional (55) Feldman, I.; Young, D.; McGuire, R. Biopolymers 1975, 14, 335–351.

Figure 4. BSA (10 µM) fluorescence decreases upon titration with PFNA (a) or PFOA (b) in 50 mM pH 7.4 sodium phosphate buffer. The shifts in the wavelength of maximum emission as a function of PFNA (c) or PFOA (d) concentration are illustrated for varying BSA concentrations (1-10 µM).

Figure 5. BSA (1-10 µM, left) and HSA (1 or 10 µM, right) fluorescence (F) decreases upon titration with PFNA in 50 mM pH 7.4 sodium phosphate buffer. Data are fit by eq 3.

contributions of each site, and K1 and K2 are the Stern-Volmer constants. As the observed decrease in fluorescence intensity is proportional to PFCA binding, the Stern-Volmer constants are estimates of PFCA-albumin association constants. Due to the indirect nature of the fluorescence method, the equilibrium PFCA concentration is approximated by the total PFCA concentration, a shortcoming addressed in ongoing studies. Representative Stern-Volmer plots for PFNA binding to BSA and HSA are shown in Figure 5, and those for PFOA, PFDA, and PFUnA binding to BSA are in Figure S7, Supporting Information. Table 3 contains fit parameters for each PFCA-protein combination. As mentioned earlier, HSA contains one Trp residue while BSA contains two, yet data for both proteins are fit well by the twosite model. We propose that the two terms correspond to two classes of binding sites on the proteins, one with higher affinity than the other, and that binding primarily affects the fluorescence properties of the conserved Trp. The primary Stern-Volmer constants for PFCAs (K1) range from 0.3 × 105 to 6 × 105 M-1 for both BSA and HSA, which are of the same order of magnitude as binding constants reported for C8 (0.34 × 105 M-1)

and C10 (1.0 × 105 M-1) fatty acids binding to HSA.54 There is a general increase in both K1 and K2 with increasing perfluorocarbon chain length (Table 3 and Figure S9, Supporting Information). As can be seen in Table 3 and Figure S8 (Supporting Information), there is an inverse dependence of K1 on the concentration of BSA, particularly evident for PFNA. Such an inverse dependence has been observed for other ligands binding to serum albumin; the authors of one such prior study suggested both protein aggregation and ligand or protein impurities as potential explanations, though acknowledging the need for further research.56 The nature of this inverse dependence, whether an experimental or mathematical artifact, is a subject of exploration in ongoing studies. Nature of PFCA-Albumin Interactions. A detailed discussion of the nature of PFCA-albumin interactions is beyond the scope of this paper. However, the three methods described here offer complementary information about PFCA-albumin interactions and insight into the changes experienced by both the ligand (56) Bowmer, C. J.; Lindup, W. E. Biochim. Biophys. Acta 1980, 624, 260–270.

Analytical Chemistry, Vol. 82, No. 3, February 1, 2010

979

Table 3. Stern-Volmer Parameters for PFCAs Binding to BSA or HSA f2a

K2a (M-1)

R2

PFOA-BSA (0.33 ± 0.04) × 105 (1.5 ± 0.2) × 105

0.49 ± 0.01 0.52 ± 0.01

(0.5 ± 0.1) × 102 (0.8 ± 0.1) × 102

0.997 0.996

0.405 ± 0.008 0.46 ± 0.03 0.47 ± 0.02 0.47 ± 0.02 0.55 ± 0.04 0.51 ± 0.02

PFNA-BSA (0.49 ± 0.04) × 105 (0.7 ± 0.2) × 105 (1.0 ± 0.1) × 105 (1.7 ± 0.3) × 105 (2.3 ± 0.7) × 105 (6 ± 1) × 105

0.606 ± 0.008 0.56 ± 0.03 0.55 ± 0.02 0.53 ± 0.01 0.46 ± 0.03 0.51 ± 0.02

(3.0 ± 0.2) × 102 (7 ± 1) × 102 (7.4 ± 0.9) × 102 (3.5 ± 0.5) × 102 (4 ± 1) × 102 (9 ± 2) × 102

0.999 0.995 0.998 0.995 0.979 0.992

10 2b 1b

0.46 ± 0.02 0.67 ± 0.09 0.65 ± 0.09

PFDA-BSA (1.8 ± 0.3) × 105 (1.4 ± 0.5) × 105 (3 ± 1) × 105

0.57 ± 0.02 0.3 ± 0.1 0.4 ± 0.1

(8 ± 1) × 102 5 × 102 5 × 102

0.994 0.988 0.969

10 1b

0.64 ± 0.06 0.72 ± 0.04

PFUnA-BSA (3 ± 1) × 105 (0.20 ± 0.06) × 105

0.40 ± 0.05 0.28 ± 0.04

(10 ± 1) × 102 0.1 × 102

0.976 0.989

10b 1

0.67 ± 0.07 0.64 ± 0.09

PFNA-HSA (0.3 ± 0.2) × 105 (0.5 ± 0.2) × 105

0.36 ± 0.06 0.38 ± 0.09

0.1 × 102 (1 ± 4) × 102

0.943 0.976

[BSA] (µM)

f1a

10 1

0.52 ± 0.01 0.52 ± 0.01

10 8 6 4 2 1

K1a (M-1)

a The error given is the standard error (95% confidence) from Kaleidagraph. b The K2 values for these data sets were fixed at the values shown to allow Kaleidagraph to best fit the data. The K2 values were chosen by optimizing R2.

and protein. This information may be considered in both a qualitative and a quantitative fashion. From a qualitative perspective, the observed fluorescence trends at low PFCA:albumin mole ratios suggest conformational changes upon binding that lead to the transition of the Trp residue(s) from a more polar to a less polar environment. On the basis of the crystal structure of native HSA,57 the conserved Trp (214) is located in an R helix in domain IIA near the interface with domain I. Upon fatty acid binding, domains I and III pivot about points near their interfaces with domain II,58 changing the environment around Trp-214. Given their structural similarity to fatty acids, PFCAs would be expected to bind in the same primary binding sites, causing similar conformational changes. At higher PFCA:albumin mole ratios, surface tension data indicate significant interactions between PFCAs and the protein; fluorescence data at similar mole ratios suggest protein unfolding, likely due to these PFCA-albumin interactions. Together, the results from these experimental techniques are consistent with the unfolding of the protein and association of micellar aggregates along the peptide chain, like the “string of pearls” description suggested for other surfactants.45 While interactions at such high PFCA: albumin mole ratios are not relevant in biological systems, they offer some mechanistic insight into surfactant-protein interactions. Quantitative consideration of the data presented here provides a measure of the strength of PFCA-albumin associations. Fluorescence titrations of BSA and HSA with PFCAs of varying chain length suggest primary sites with Ka on the order of 105 M-1 and secondary sites with Ka on the order of 102 M-1. Previous studies have not clearly explored the effect of PFCA chain length on binding strength; here a positive correlation is observed between Ka and the perfluorocarbon chain length for both primary and secondary binding, a result that is (57) Sugio, S.; Kashima, A.; Mochizuki, S.; Noda, M.; Kobayashi, K. Protein Eng. 1999, 12, 439–446. (58) Curry, S.; Mandelkow, H.; Brick, P.; Franks, N. Nat. Struct. Biol. 1998, 5, 827–835.

980

Analytical Chemistry, Vol. 82, No. 3, February 1, 2010

important for predicting the bioaccumulation and pharmacokinetics of PFCAs of varying chain length. 19F NMR titrations conducted for two PFCAs with both BSA and HSA yield association constants of 102-104 M-1, which, within error, are similar in magnitude to the K2 values obtained from the fluorescence titrations. This suggests that, at the high PFCA: albumin mole ratios required for the 19F NMR experiments, the primary sites are already saturated and only the weaker secondary binding is observed. The 19F NMR data for PFOA and PFNA support the positive correlation between Ka and perfluorocarbon chain length for secondary binding. The primary association constants estimated here (105 M-1) are similar in magnitude to the primary Hill binding constant reported by Messina et al. for PFOA-albumin complexes.38 Ongoing experiments in a complementary study using equilibrium dialysis and nanoelectrospray mass spectrometry will provide further information on primary association constants at low PFCA:albumin mole ratios. CONCLUSIONS As mentioned earlier, many experimental techniques have been employed to study the interactions of PFCAs and other anionic organic ligands with serum albumin both qualitatively and quantitatively, often without regard to their limitations. In this study, three widely used experimental methodssfluorescence spectroscopy, 19F NMR spectroscopy, and surface tension measurementsswere compared to clarify the strengths and limitations of each method when applied to PFCAs. Given the complicating factor that both PFCAs and serum albumin are surface active, surface tension data alone do not fully elucidate the nature of PFCA-albumin interactions or offer quantitative association constants. However, taken together with data from other techniques such as fluorescence spectroscopy, surface tension measurements may help to explain changes in protein conformation as a function of PFCA concentration, though only at concentrations well beyond physiological relevance.

19

F NMR spectroscopy should allow more accurate determination of binding constants when compared to indirect methods, as it involves the direct observation of bound and unbound PFCAs. However, determining the y-intercept of the data plot, as opposed to the slope, introduces a large experimental error in the method. The method is also very sensitive to instrumental parameters, spectral processing, and sample preparation; any inconsistencies add to the error in Kd. The net effect is a potentially large relative error in Kd, which is then retained upon calculation of Ka. Limitations associated with both the instrumentation and the method require the use of PFCA:albumin mole ratios that significantly exceed physiological relevance. Consequently, binding strength information obtained from NMR titrations must be interpreted with caution due to the high intrinsic error and the inability to characterize binding at low mole ratios. Monitoring changes in native protein fluorescence as a function of PFCA binding allows greater sensitivity than the prior two methods, enabling the study of low PFCA:albumin mole ratios that approach environmental and physiological relevance. Fluorescence spectroscopy provides the most complete qualitative and quantitative information about the binding of the three techniques studied. The greatest limitation of this technique is that it is indirect. Consequently, fluorescence changes must be interpreted thoughtfully, and suitable approximations must be employed to fully characterize the binding.

Of the three techniques studied here, fluorescence spectroscopy is the most efficient. This technique both offers qualitative insight into PFCA-albumin interactions and provides estimated binding constants that agree with those obtained by established methods such as equilibrium dialysis. It also allows for the study of the widest range of PFCA:albumin mole ratios. Ongoing research focuses on realizing the full potential of this technique through thoughtful application of approximations to obtain a complete quantitative characterization of perfluoroalkyl acidalbumin interactions. ACKNOWLEDGMENT We thank Susan Kohler for assistance in performing the 19F NMR titrations at Union College. We thank Jennifer Field (Oregon State University) and Dale Bacon and Mark Ellefson (3M) for analytical standards. This research was supported by the U.S. National Science Foundation under Grants 0201955 and 0216458, a grant from Stanford University’s Woods Institute for the Environment, and Union College. SUPPORTING INFORMATION AVAILABLE Additional information as noted in text. This material is available free of charge via the Internet at http://pubs.acs.org. Received for review October 4, 2009. Accepted December 10, 2009. AC902238U

Analytical Chemistry, Vol. 82, No. 3, February 1, 2010

981