Environ. Sci. Technol. 2004, 38, 6263-6270
Toward More Useful In Vitro Toxicity Data with Measured Free Concentrations M I N N E B . H E R I N G A , * ,† RICHARD H. M. M. SCHREURS,† FRANS BUSSER,† PAUL T. VAN DER SAAG,‡ BART VAN DER BURG,§ AND JOOP L. M. HERMENS† Institute for Risk Assessment Sciences (IRAS), Utrecht University, Yalelaan 2, 3584 CL Utrecht, The Netherlands, Hubrecht Laboratory (Netherlands Institute for Developmental Biology), Uppsalalaan 8, 3584 CT Utrecht, The Netherlands, and BioDetection Systems B.V., Badhuisweg 3, 1031 CM Amsterdam, The Netherlands
In vitro assays and computer models are promising alternatives for in vivo animal testing, but the power of these alternative methods to predict in vivo risk is still very limited. One step forward is to make the outcome of in vitro assays (such as median effect concentrations (EC50 values)) independent of assay conditions such as protein content. Here we show that measured free concentrations of chemicals in the in vitro assay medium result in systemindependent EC50 values. We introduce a very simple method to measure free concentrations in miniature test systems using negligible depletion solid-phase microextraction. The generated data are much more suitable for extrapolation to in vivo, provide unbiased input for computational methods (for example, quantitative structureactivity relationships), and can shed an entirely different light on the activity of environmental contaminants.
Introduction The latest policies on risk assessment of chemicals, for example, the REACH initiative in the European Union (1), demand more testing and at the same time a reduction in use of laboratory animals. In vitro assays are popular in risk assessment as a fast, cheap, and ethically more accepted alternative to animal tests. Miniature in vitro systems are used to generate quantitative data for the dose that produces a certain biological effect, but their predictive power for in vivo effects is still limited. The concentration added to the cell culture (i.e., the nominal concentration) is usually taken as dose parameter, assuming this is what the cells are exposed to. However, the amount of test compound added to an in vitro test system can be partly “lost” for production of an effect through various pathways (Figure 1). The occurrence and severity of these “losses” can vary between test conditions and test compounds, depending, for example, on the amount of cell membrane present (i.e. cell number), the serum (protein) content in the test medium, and the volatility and * Corresponding author telephone: +31-30-2535328; fax: +3130-2535077; e-mail:
[email protected]. † Utrecht University. ‡ Netherlands Institute for Developmental Biology. § BioDetection Systems B.V. 10.1021/es049285w CCC: $27.50 Published on Web 10/16/2004
2004 American Chemical Society
hydrophobicity of the compound. Serum content of the cell culture medium is known to vary between laboratories, and this can lead to a different outcome of the same test in different laboratories (2). Such a variety in outcome complicates a comparison of the potencies of biologically active chemicals. These, and other artifacts in in vitro assays have already been recognized by others (3-7). Additionally, anomalous effects observed in, for example, mixture studies could be explained by interactions at the exposure levels, while they are now often explained via a complex biological mechanism because the actual exposure is not studied. One step in the enhancement of the predictive power of in vitro assays is to make their outcome (such as median effect concentrations (EC50 values)) independent of assay circumstances such as protein and cell content (3, 8). The dose that really produces the effect is the target dose: the concentration of compound at the target site (9). For chemicals that are transported via diffusion processes, the freely dissolved concentration in the vicinity of the target is a good representative of the target dose. Figure 2 shows an example of how the test conditions (e.g., serum content) can influence the free concentration. Measurement of the free concentration will therefore increase the reliability and precision of toxicity data. To perform measurements in such miniature systems, a suitable miniature sampling technique is necessary and “negligible depletion” solid-phase microextraction (nd-SPME) represents such a technique. The SPME technique (10) is a solid-phase-based extraction method. The nd-SPME method, as developed by Vaes et al. (11), takes advantage of the tiny volume of the extraction phase. The extraction phase is the (hydrophobic) polymer coating of an optical glass fiber (with a coating thickness of 7 µm), as also used in telecommunication cables. Organic molecules that are freely dissolved in an aqueous solution will partition to this solid phase when a fiber is placed in the solution. When the volume of the polymer is small enough in comparison with the solution volume, only a negligible amount of these free molecules will be absorbed by the fiber. The free concentration will then virtually remain constant, leaving other partitioning and binding equilibria undisturbed. The concentration of a compound in the fiber coating will then be directly related to the free concentration in the aqueous solution. In practice, free concentrations in a matrix are determined from concentrations on fibers and a calibration curve. This calibration curve is established by plotting concentrations in the fiber versus aqueous concentrations in standards without matrix where total and free concentrations are similar. The application of disposable pieces of the glass fibers strongly enhances the applicability of this ndSPME method because it can be combined with, in principle, each analytical technique (12, 13) and also enabled us to apply it in an in vitro assay setup with culture plates containing wells. As nd-SPME had not yet been applied in in vitro culture plates, an experimental setup was developed, and the kinetics of uptake into the fiber in this new setup was studied by creating absorption profiles. Furthermore, we have studied the effects of serum protein binding (one of the “loss” processes; Figure 1) on the outcome of an in vitro assay. The in vitro assay we chose as an example for our study was an estrogenicity reporter gene assay with stable hERβ transfectants of HEK293 cells (14). For both estradiol and octylphenol, a known xeno-estrogen, dose-response curves were produced at different serum concentrations in the medium, while the free concentrations of these compounds VOL. 38, NO. 23, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
6263
FIGURE 3. Schematic representation of the experimental setup of nd-SPME for in vitro assays. FIGURE 1. Schematic representation of the various processes that decrease the availability of a test compound for its target in a typical in vitro assay.
FIGURE 2. Effect of serum content on freely dissolved fraction of estradiol. The free fraction (ff) was calculated using the equation ff ) 1/(1 + Ka fupPt), with Ka as the affinity constant (7.2 × 10-5 M-1 for estradiol and BSA; 24), fup as the fraction unoccupied protein, and Pt as the total concentration of protein (estimated at 1 × 10-4 M in 100% fetal calf serum; 23). were measured using nd-SPME. From curves based on nominal concentrations, as well as from curves based on free concentrations, EC50 values were derived and compared with each other. Relative potencies were calculated from the obtained EC50 values to assess whether these were dependent on protein concentration, as Gu ¨ lden and co-workers found a dependency on cell content (15). Also, an effort was made to calculate the free concentrations with a simple binding model and compare these to the measured free concentrations.
Materials and Methods Apparatus and Reagents. [2,4,6,7-3H]17β-Estradiol (37 MBq/ mL/3.52 TBq/mmol) was purchased from New England Nuclear (Boston, MA) and used within 2 months after purification to ensure radiochemical purity (>94%). For the absorption profile, a different batch of [2,4,6,7-3H]17βestradiol was used (37 MBq/mL/3.33 TBq/mmol) 3 months after purification so the radiochemical purity in this experiment was estimated at 92-96%. 4-n-Octylphenol (99%) was purchased from Aldrich (Steinheim, Germany), PCB 138 (2,2′,3,4,4′,5-hexachlorobiphenyl; 99%) was from RiedeldeHae¨n (Seelze, Switzerland), and 2,3,4-trichlorophenol (TCP; 99%) was from Aldrich. Ethyl acetate (99.8%), ethanol (99.8%), and the derivatization agent MTBSTFA (N-methyl-N-(tertbutyldimethylsilyl)trifluoroacetamide; >97%) were purchased from Lab-Scan (Dublin, Ireland), Riedel-deHae¨n, and Fluka (Buchs, Germany), respectively. The 7-µm polyacrylate (PA) and 7-µm poly(dimethylsiloxane) (PDMS) fiber was purchased from Supelco (Bellefonte, PA). The selection of the fiber coating was based on the outcome of preliminary experiments in which equilibration time and sensitivity were studied with the two fiber coatings (polyacrylate and PDMS). Cell Culture. Human embryonal kidney 293 (HEK293) cells were obtained from the American Type Culture Col6264
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 23, 2004
lection (ATCC; Rockville, MD). The cells were maintained in a 1:1 mixture of Dulbecco’s modified Eagle’s medium and Ham’s F12 medium (DF medium), prepared specially with extra nonessential amino acids and bicarbonate by Life Technologies (Gaithersburg, MD). The medium was supplemented with 7.5% fetal calf serum (FCS; Bodinco, Alkmaar, The Netherlands), 0.2% penicillin-streptomycin (5000 U/ml and 5000 µg/mL, respectively; Life Technologies), 50 µg/mL hygromycine (Sigma, St. Louis, MO), and 200 µg/mL neomycine (Sigma). Cells were kept in an atmosphere of 7.5% CO2 and at 37 °C. Gene Expression Assay with Stable ERβ Reporter Cell Line. The generation of stable hERβ transfectants of HEK293 cells is described elsewhere (14). Cells were rinsed twice with phosphate-buffered saline (PBS) without calcium and magnesium, trypsinized, and suspended in phenol red-free DF medium supplemented freshly with 0.2% penicillin-streptomycin and 5% dextran-coated charcoal-stripped FCS (DCC-FCS). The cells were plated in 24-well tissue culture plates (Corning Inc., Corning, NY) with 80 000 cells and 200 µL of suspension per well. After 48 h, the medium was replaced by 2.0 (estradiol assay) or 2.5 (octylphenol assay) mL/well of fresh phenol-red free medium containing the desired percentage of DCC-FCS and the desired concentration of test compound. The test compound was added to the medium in an ethanol solution of a 1000-fold higher concentration than ultimately desired in the medium. Plate lids containing the SPME fibers were then placed at a noted time. Exactly 24 h later, the lids with fibers were removed, and the fibers were processed as described elsewhere. The medium was then removed from the cells, and 200 µL of lysis solution was added to each well (1% (v/v) Triton X-100, 25 mM glycylglycine, 15 mM MgSO4, 4 mM EGTA, and 1 mM dithiothreitol). After 15 min of gentle swirling at 4 °C, 50 µL of lysate from each well was transferred to a well in a black 96-well plate (Corning Inc.) as well as six 50-µL samples of the lysis solution for controls. Luciferase activity was measured with the LucLite luciferase reporter gene assay kit (Packard Instruments, Groningen, The Netherlands) according to the manufacturer’s instructions, using 50 µL of LucLite solution per well. Activity was measured for 6 s with a Topcount liquid scintillation counter (Packard Instruments). In Vitro SPME Procedure. Holes of 0.5 cm diameter were drilled centrally in five lids of 24-well plates, above all 24 wells. A Viton sheet (Rubber BV, Hilversum, The Netherlands) was cut into pieces fitting inside the upstanding rims on top of the lids and fastened at the edges to the lids with Scotch tape. Viton was chosen for its inertness, sufficient elasticity, and relatively low cost. The desired fiber was cut into pieces of ∼3 cm long, which were pierced through the Viton with the help of a syringe needle. The length of fiber sticking out underneath the Viton was adjusted manually so that 5 mm of the fiber would be exposed to the solution in the well during the experiment (Figure 3). Prior to a cellular assay, the lids with fibers were sterilized by overnight exposure to UV light. After exposure to the test solution, the fibers were pulled out of the Viton and analyzed. Extraction and Analysis of Estradiol and Octylphenol. [2,4,6,7-3H]17β-Estradiol in fibers was analyzed similarly as
described in Heringa et al. (12). In short, the fibers were transferred to a scintillation vial and left for desorption with 3.8 mL of Ultima GOLD (Packard Bioscience Co., Meriden, CT) for at least 3 h before counting the activity in a liquid scintillation counter (Minaxi Tri-carb 4000 LSC; Packard Bioscience Co.). Aqueous samples were analyzed for [3H]estradiol by transferring them to scintillation vials, adding 3.8 mL of Ultima GOLD, and directly counting radioactivity. All samples were calibrated with a series of different volumes of several [2,4,6,7-3H]estradiol dilutions in ethanol. Octylphenol was desorbed from a fiber by inserting the whole fiber exposed-side down in a 250-µL glass insert in a 1.7-mL glass vial. The 20 µL of ethyl acetate was added to the fiber, followed by subjection to an ultrasonic bath (Transsonic T 460; Elma Hans Schmidbauer, Singen, Germany) for 15 min. This procedure was found to extract close to 100% of the octylphenol because a second extraction step delivered less than the detection limit (
