Fourteen resins were received by CertiChem (CCi) in 2008–2012 and by University of California at Davis (UCD) in 2012. Analyses at CCi focused on assessing release of chemicals having EA from many different stressed and unstressed resins, especially three Tritan™ resins, although relatively fewer assays were performed on unstressed Tritan™ resins. In contrast, UCD focused on EA release from three unstressed Tritan™ resins. The three Tritan™ resins tested were from a family of similar resins that are synthesized from the same monomers and probably the same additives, but in different concentrations. Some resins may even been the same (e.g. EX401 and TX201), but are packaged and marketed differently. At CCi, each resin was subjected to various (not necessarily the same) combinations of extraction solvents and stresses and EA was determined by two different assays, MCF-7 and BG1Luc (see below). CCi sometimes stopped testing a resin if a particular type of extraction solvent or stress clearly showed that the resin leached significant levels of chemicals having EA. In other cases, a resin was run many times as a positive control for other studies and all such data are reported herein. That is, the aim of this study was not to perform an exhaustively complete set of responses, using the same number of replicate assays for all possible stresses and extraction conditions for each resin, but rather to survey a larger sample of PC-replacement resins and assess whether they released chemicals with EA. Resins were sometimes tested once each for three different assay conditions, rather than two or more times with one given assay condition.
Protocols to stress plastic resins
Prior to applying stresses at CCi, resin pellets were compression molded at 230°C into 4 × 4 × 1/64 inch thick plaques. Heat and moisture stresses were obtained by sealing plaques in individual crimped packets of aluminum foil and placing them into a Tuttnauer autoclave set at 134°C for 8 minutes.
Microwave stresses were obtained by placing samples into glass beakers and microwaving them in a 1200 W oven set to “high” for two minutes, allowed to sit for 30 minutes, and the cycle repeated 10 times. Alternatively, samples were placed in EA-free polypropylene tubes, microwaved on “high” setting for three minutes with a resting period of 30 minutes and this cycle repeated 5 times. Both methods produced very similar results.
Several methods of UV stressing were carried out as previously described [16, 17]:
-
(1)
Long wavelength (315-400 nm) UVA stresses: plaques were placed in a Q-Lab QUV unit containing UVA-340 nm bulbs (intended to simulate sunlight between 295 nm and 365 nm) for 80 hours at 45–50°C with no condensation.
-
(2)
Short wavelength (100-280 nm) UVC stresses: plaques were placed onto a piece of aluminum foil in a Labconco Biosafety hood approximately 24” from a germicidal fluorescent light (maximum intensity wavelength of 254 nm) for 24 hours.
-
(3)
Natural sunlight stresses: plaques were placed individually between a quartz glass plate on top and aluminum foil on a porcelain plate below. The two plates were clamped together using binder clips. To control for heat versus sunlight effects, some of these plaques were wrapped in thick aluminum foil. These plaques were placed on the roof of CCi’s facility for 1–14 days in summer.
Note that UV radiation in sunlight is often classified by wavelength [25] as UVC (100–280 nm), UVB (280–315 nm), and UVA (315–400 nm). Most UVC wavelengths do not reach the earth’s surface. However, UVC wavelengths are used in some germicidal UV devices, e.g., to sterilize baby bottles. Visible wavelengths are 400–750 nm.
Extracts of plastic resins/plaques
At CCi, unstressed or stressed plaques described above were cut into ~4 × ~4 mm pieces and 2.0-5.0 grams of these cut resin pieces were added to sterile glass test tubes and placed under a germicidal UV light for 30 minutes to sterilize the samples before adding an extraction solvent to a final concentration of 1.0 g of resin/ml. This brief exposure to UV does not alter release of chemicals having EA from plastics [16]. The extraction solvents used were a saline-based solution (saline: Roswell Park Memorial Institute RPMI-1640 medium without Phenol red), 100% EtOH, 10 or 50% aqueous EtOH, or distilled water. Some samples were extracted for 72 hours at 37°C in a static incubator; the majority of samples were extracted at 40°C for 240 hours in an incubator shaker. EtOH extracts were concentrated 10× by evaporation and then diluted 100× with estrogen-free-medium (EFM) with the highest resin concentration at 0.1 g of resin/ml. Saline extracts were diluted 1:1 with 2× EFM and then diluted 1-4× with EFM so that the highest resin concentration was 0.125 g resin/ml. Estrogen-free medium is actually estrogenic-activity (EA)-free medium and it is made “essentially free” of EA by using phenol-red-free basal medium supplemented with dextran-coated/charcoal stripped fetal bovine serum (FBS).
Unstressed resin pellets were used as received at the University of California, Davis (UCD). Unstressed resin pellets (6.0 g) were extracted with 3.5 ml 100% EtOH in a baked glass tube with Teflon-lined cap for three days at room temperature in the dark. The extract solution was transferred to a second baked glass tube, and then evaporated to dryness. The residue was dissolved in 40 μl DMSO, vortexed, and diluted with 1960 μL of EFM. This protocol produced stock solutions of ~3.0 g resin/ml in medium containing 1% DMSO. The stock solutions were serially diluted such that the highest test concentration was 1.5 g resin equivalent/ml. The highest g resin concentration/ml for BG1Luc assays run at UCD was 4–15 times higher than the highest g resin concentration/ml for MCF-7 or BG1Luc assays run at CCi.
We used more than one extraction method for many reasons. First, there are no standards yet devised for such extracts. Second, our data in [16, 17] and this paper clearly showed that more than one extraction solvent was needed to better detect hydrophilic versus hydrophobic chemicals in leachates. Third, different extraction solvents require different dilution protocols. That is, isotonic saline extracts can be tested without dilution, but distilled water or various EtOH extracts must be diluted to a starting EtOH concentration of no more than 1% EtOH. [Higher EtOH concentrations kill cells in cell-based in vitro assays such as the MCF-7 and BG1 assays]. Third, evaporation of the solvent concentrates the extracted chemicals, but would lose chemicals that are more volatile than the solvent. EtOH extracts were always concentrated, in part because EtOH extracts had to be diluted (see above) while saline extracts did not have to be diluted and hence were often not concentrated. In brief, using a single protocol would have affected the results, unfortunately by hiding/not finding many chemicals that had EA in leachates.
Materials and supplies
Both cell types were grown and maintained in polystyrene T-75 flasks (BD Falcon, BD Biosciences, San Jose, CA, cat#353136) and polystyrene T-25 flasks (CytoOne, USA Scientific, Ocala, FL, cat#CC7682-4825). Media and medium supplements (RPMI -1640 Medium, DMEM, FBS, nonessential amino acids, L-glutamine, penicillin, streptomycin) were purchased from Invitrogen (Grand Island, NY, USA). Insulin was purchased from Sigma (St. Louis, MO, USA).
MCF-7 cells were seeded into 96-well flat bottom PS polystyrene plates (BD Falcon, cat#353075) and BG1Luc cells were seeded into 96-well white wall/clear bottom plates (Greiner Bio-One, Monroe, NC, cat#655098). Ethanol of 100% purity was purchased from OmniPur, EMD-Millipore, Billerica, MA, Acros Organics/Fisher Scientific, Pittsburgh, PA or Sigma-Aldrich, St. Louis, MO. Water was distilled on-site in an all-glass system and collected directly into glass before use in extractions performed in borosilicate glass tubes. ICI was obtained from Tocris Bioscience (Minneapolis, MN, USA).
BG1Luc Assays
The BG1Luc4E2 cell line (referred to here as BG1Luc) is a human ovarian cancer cell line that responds to estrogenic chemicals with the induction of firefly luciferase [26], and has been approved as a screening method for estrogenic chemicals by OECD, EPA, and ICCVAM/ NICEATM [22]. BG1Luc cells were maintained in cell culture medium, and then placed in EFM for 3 (CCi) or 5 (UCD) days. Standard BG1Luc cell culture maintenance medium is complete growth medium containing phenol red, FBS and various additives and thus contains estrogen and chemicals with EA that facilitate cell growth. The EFM used for detection and analysis of EA using the BG1Luc cells was of two varieties: one EFM was identical to the MCF-7 EFM (see below), the other EFM used a different phenol-red-free basal medium (DMEM) supplemented with 10% dextran-coated/charcoal-stripped FBS.
Acclimated cells were seeded at 10,000 (CCi) or 70,000 (UCD) cells per well in 100 μL EFM in 96-well plates for 24 h, followed by a 24 ± 6 h incubation with test extracts in triplicate. The use of slightly different protocols by the two laboratories meant that quantitative values for %RME2 could not be directly compared, but insured that qualitative conclusions (the presence or absence of EA in sample leachates) did not depend upon one specific protocol for the BG1Luc assay used by one laboratory. Cytotoxicity was assessed by microscopic inspection as previously described [17, 22, 23, 27]. Cell culture medium was aspirated, cells were lysed with 1% Triton X-100, 10% glycerol, 2 mM EGTA, and 1 mM DTT. Luciferase was then measured using an automated microplate luminometer (Tristar, Berthold) with the Promega Luciferase Assay System as previously described [17, 26, 27].
MCF-7 Assays
We used an MCF-7WS8 cell line, a human breast cancer cell line, in a robotic version [16, 17, 23] of the MCF-7 cell proliferation assay (aka E-screen assay) that has been used for decades to reliably assess EA [28, 29]. The assay has been undergoing validation for international use by ICCVAM/ NICEATM and has been nominated for validation by OECD. Chemicals with EA activate the ERs and ER-dependent transcription of estrogen-responsive genes, which leads to proliferation of MCF-7 cells. In brief, each test chemical or extract at each concentration was added in triplicate or quadruplicate to 96-well plates containing MCF-7 cells in EA-free culture media. MCF-7 EA-free culture medium is phenol-red-free RPMI 1640 medium supplemented with antibiotic/antimycotic solution and L-glutamine, 1% dextran-coated/charcoal stripped FBS, and 4% dextran-coated/charcoal stripped FBS.
After six days of exposure, MCF-7 cells were lysed with 1:5 (vol/vol) of 0.16% acetaldehyde/20% perchloric acid. The amount of DNA/well, an indication of cell numbers, was assayed using a microplate modification of the diphenylamine assay as previously reported [16, 17, 23]. Cytotoxicity was assessed as described above for BG1Luc assays and as previously described for MCF-7 assays [16, 17, 23].
Calculation of EA
The EA of extracts was calculated as the relative maximum %E2 (%RME2, a measure of response amplitude), i.e., a percentage of the maximum DNA/well (MCF-7 assays) or relative luminescence (BG1Luc assays) produced by an extract at any dilution relative to the maximum agonist effect produced by E2 [16, 23, 27]. Aliquots of the highest concentration of an extract were diluted four to eight times to produce up to eight test concentrations. Each test chemical or extract at each concentration was added in triplicate or quadruplicate to 96-well plates containing MCF-7 cells in EA-free culture media. Both a vehicle control (VC) and “sham” control (SC), also called a “method blank”, were run in triplicate or quadruplicate in each experiment. The VC was the vehicle solvent used for that particular assay. The SC was the vehicle solvent taken through all steps that were used to assay the test sample/test extract. The VC was set to 0% RME2. For the MCF7 assay, cell proliferation responses were normalized by the DNA response of the vehicle control according to Equation 1 [16, 23]:
(1)
For the BG1 assay, %RME2 was similarly calculated by subtracting the VC RLU (relative luminescence units) from test extract RLU, and then normalized by dividing by the VC-adjusted highest E2 RLU response. The maximum E2 response was determined by a concentration-response curve run in triplicate for 8 – 12 E2 concentrations for each assay. The maximum E2 response was set to 100% RME2 and the VC to 0%RME2. The concentration-response curve for EA of a test substance or extract was plotted with log M or log g/mL test concentrations respectively, on the X-axis and %RME2 on the Y-axis.
Typical values for SC were 0% ±10% RME2. The %RME2 of a sample SC or extract can be negative relative to VC by random variation. If more negative than -10% RME2, anti-EA activity or toxicity would be suspected. Inclusion of an SC accounts for any extraneous residual EA that might exist in the media, extraction solution, or derived from the materials used for sample preparation. If the EA of the SC was greater than 15% RME2, then the entire experiment was rejected. The SDs for the extract or SC replicates for a given assay were typically so small that they fell within the space taken up by symbols used to plot the averages of %RME2 data for each concentration.
The greatest %RME2 response of 4–8 dilutions of a test chemical or extract run in triplicate was considered detectable if it produced an effect whose average %RME2 was greater than 15% RME2, i.e., more than three standard deviations (3SD, p <0.01, Student’s t test) above the mean SC response of a triplicate run for that particular assay. Such an “EA/no EA” categorical criterion/cut-off provides a rather conservative measure of detection for positive EA in an extract and has been previously used for detection of EA in plastics or other substances [16, 17, 22, 23, 27]. For known test chemicals assayed by the BG1Luc or MCF-7 assay (e.g. E2, TPP), the concentration that produced half-maximal stimulation by the test chemical (EC50, in M) was calculated from best sigmoidal fits to dose–response data using GraphPad Prism [16, 17, 23, 27].
Agonist stimulation of MCF-7 proliferation or BG1Luc Luciferase activity was confirmed to be estrogenic (versus non-specific) if the stimulation by a test chemical or extract was suppressed by co-incubation with the pure anti-estrogen ICI at 10-7 - 10-8 M [16, 17, 23, 24]. These in vitro “confirmation assays” rarely produce false positive responses [16, 17, 23, 27] and we saw no examples of an agonist response that was not suppressed by ICI in this study.
Statistical comparisons
Each assay using MCF-7 cell or BG1-Luc cells of a given test substance (e.g., TPP) or a given extract (e.g., 100% EtOH) of a given resin (e.g., Tritan EX401) exposed to a given stress (e.g., no stress, UVA, etc.) was run on a plate that had various dilutions of the test substance or extract in different wells and also had wells containing the VC or SC. Each test solution was repeated 3–4 times in different wells of the 96 well plate. The intra-assay variation and difference between the mean ± SD of the normalized %RME2 values of the extract (or the test substance) and the VC (or SC) were analyzed by Student’s T-Test with n equal to 3 or 4. In illustrations showing %RME2 values, a dotted line shows the 15%RME2 value that was always 3SD greater than the VC or SC values for each assay.When a given resin was extracted, stressed and run in the same manner on different days (e.g., a single cell in Figures 2 or 3), the mean ± SD of the %RME2 values of these “n” repeated assays of the same type were sometimes calculated. Figures 2 and 3 present the individual %RME2 values for such assays. When an assay of the same type was run on the same resin, but the stress condition was varied (e.g., microwave stress versus QUV stress in Figures 2 and 3), then the two means and SD’s of these two sets of repeated assays were sometimes compared using a two-tailed Student’s T-test.
When different assays were run on the same resin or class of closely related resins (e.g., three Tritan resins), the data were categorized as “yes” versus “no” for EA if the %RME2 for each individual assay was equal to, or greater than, 15%RME2 versus less than 15%RME2, respectively. Such “inter-assay” assessments of the probability of detecting significant amounts of EA were analyzed by Pearson’s Chi-Squared test with Yates correction for a 2×2 matrix (0 vs 14d) and by Pearson’s Chi-squared test for a 5×2 matrix (0, 1, 2, 7 and 14d). We performed all statistical analysis using GraphPad Prism5 (GraphPad Software, San Diego, CA) and/or R. Studio. All p-values were two-tailed.
The number and type of replications described above are indicated in each figure and table and/or its legend.