Most Plastic Products Release Estrogenic Chemicals: A Potential Health Problem That Can Be Solved

Background: Chemicals having estrogenic activity (EA) reportedly cause many adverse health effects, especially at low (picomolar to nanomolar) doses in fetal and juvenile mammals. Objectives: We sought to determine whether commercially available plastic resins and products, including baby bottles and other products advertised as bisphenol A (BPA) free, release chemicals having EA. Methods: We used a roboticized MCF-7 cell proliferation assay, which is very sensitive, accurate, and repeatable, to quantify the EA of chemicals leached into saline or ethanol extracts of many types of commercially available plastic materials, some exposed to common-use stresses (microwaving, ultraviolet radiation, and/or autoclaving). Results: Almost all commercially available plastic products we sampled—independent of the type of resin, product, or retail source—leached chemicals having reliably detectable EA, including those advertised as BPA free. In some cases, BPA-free products released chemicals having more EA than did BPA-containing products. Conclusions: Many plastic products are mischaracterized as being EA free if extracted with only one solvent and not exposed to common-use stresses. However, we can identify existing compounds, or have developed, monomers, additives, or processing agents that have no detectable EA and have similar costs. Hence, our data suggest that EA-free plastic products exposed to common-use stresses and extracted by saline and ethanol solvents could be cost-effectively made on a commercial scale and thereby eliminate a potential health risk posed by most currently available plastic products that leach chemicals having EA into food products.


Technical Details of the Robotic MCF-7 Assay
Rationale: Chemicals with EA bind to ERs (either α, β, or ERRs) and activate the transcription of estrogen-responsive genes, which leads to proliferation of MCF-7 cells. Anti-EA is measured as an ability to reduce MCF-7 cell proliferation by the natural estrogen, 17β-estradiol (E2). MCF-7 cells are perhaps the most widely used in vitro model for studying estrogen action and were derived from human breast cancer cells. In the years since these landmark experiments, MCF-7 cell proliferation has been used as the in vitro benchmark ("gold standard") for determining EA or anti-EA and for characterizing the EA** of many environmental chemicals (Leusch et al. 2010;Soto et al. 1995).
Equipment: EpMotion robotic workstations placed in a Labconco Class II Biosafety Hood with a 254nm fluorescent fixture are used to seed cells in 96-well plates, prepare serial dilutions of the test chemicals in tissue-culture medium, change media and prepare 96-well plates for DNA quantification assays. DNA quantification, based on the diphenylamine reaction, is conducted on a Bio-Tek PowerWave X 96-well plate reader spectrophotometer.

Materials:
Bovine insulin, calf thymus DNA type I, Hoechst dye 33258, EDTA, Hanks' balanced salt solution (HBSS), and E2 are obtained from Sigma Chemical Company (St. Louis, MO). We use fetal bovine serum to maintain the cells, RPMI Medium 1640 with phenol red for cell maintenance and RPMI medium 1640 without phenol red for EA and anti-EA assays. L-Glutamine, Minimum Essential Medium (MEM) non-essential amino acids, antibiotics, anti-mycotics, and lyophilized trypsin are obtained from Invitrogen (Carlsbad, CA).

Cell preparation:
Estrogen sensitive MCF-7 cells were obtained from Dr. V. Craig Jordan, Northwestern University Medical School (now at Georgetown Medical School). Stocks of these estrogen-responsive MCF-7 cells are stored in liquid nitrogen. MCF-7 cells are also maintained at 37°C in RPMI with nonessential amino acids, 10 µg/mL phenol red, 10 mM HEPES, 6 ng/mL insulin, 100 units/mL penicillin, 100 µg/mL streptomycin, and 10% fetal bovine serum (maintenance medium). Because the responsiveness of MCF-7 cells can drift, cells are propagated for 2-3 months and then replaced with cells derived from the primary source kept in liquid N 2 storage. An aliquot of cells maintained at 37°C are grown for two days in phenol-free RPMI media containing nonessential amino acids, 10 mM HEPES, 6 ng/mL insulin, 100 units/mL penicillin, 100 µg/mL streptomycin, 1% charcoal-stripped fetal bovine serum, and 4% charcoal-stripped bovine calf serum (EA-free medium). Using the epMotion 5070 unit, MCF-7 cells are seeded at 5000 cells per well in 0.2 mL of EA-free medium in 96-well cell culture plates. The cells are adapted for 3 days in EA-free medium prior to adding test chemicals.
We routinely test each new batch of charcoal-stripped serum for estrogenstimulated MCF-7 cell growth. Unless E2-stimulated cell growth is ≥ 4 times the negative control growth in seven days and the cell growth rate is similar for the vehicle control (0.5% EtOH) and ICI 182,780, we do not use that batch of serum.
DNA quantification: Briefly, 60 µL of a 1:5 ratio mixture of 0.16% acetaldehyde and 20% v/v perchloric acid, and then 100 µL of 4% diphenylamine in glacial acetic acid, are added to each well. A standard curve is prepared by adding samples of 0.125 to 3 µg DNA/well in 10 µL of HBSS. After incubating the plate overnight at 37°C, the absorbance is measured as the difference between OD 595 and OD 700 in a Bio-Tek EL x 808 plate reader. The output is converted to µg DNA/well by using a third degree polynomial curve fit of the standard curve; the correlation coefficient of the standard curve fit is typically greater than 0.99. Each robotic assay includes a DNA standard for calibration. DNA absorbance at 1 mg/mL is set at 50 absorbance units. The DNA concentration of MCF-7 cells accurately estimates the number of cells (Taylor et al. 1995). An estimate of cell number/well is obtained by dividing the total amount of DNA per well by 20 pg of DNA per MCF-7 cell.

Treatment of test chemicals or extracts (also see description in text of main paper):
A test chemical is dissolved in ethanol (EtOH) and/or culture media (saline) at a stock concentration of 10 -1 M or to the highest soluble concentration and then diluted 6-12 times in 2-10x steps in EA-free saline.
We cut plastic items to be extracted into 4 mm square pieces so as to standardize extraction methods and make comparisons among plastic items more valid. Cutting the plastic item has only a small effect on the surface area. For example, a typical plastic resin sheet or item has a thickness of 0.05 cm and a polymer density of 1 g/cc, so that the area per gram is 42 cm 2 for 2.5 cm squares. For 4 mm squares, the surface area per gram is 50 cm 2 . Additionally, the migration of typical additives (e.g. antioxidants) is a very weak function of surface area -the rate limiting step for most common migrants is the removal of the extracted chemical from the boundary region between the plastic and the extracting solvent (or the rate of hydrolysis of the extracted material in the case of hindered phenolic antioxidants).
Plastic resins or products are extracted by first sterilizing finely cut plastic in a glass test tube placed about two feet from a 254 nm fluorescent fixture in a Labconco Hood for 30 minutes to kill bacteria before adding sterile solvent, sealing the test tube, removing them from the hood and placing them in a 37°C incubator for 72 hours.
The extracts are performed using one of three protocols: • Culture media (about 0.3M saline) is used as a more polar solvent in a ratio of 1 g plastic/1.5 mL media that, following the extraction period, is then removed from the vial and diluted 2x with EA-free medium containing 2% charcoal-stripped FBS and 8% of charcoal-stripped calf serum as the first test concentration. • EtOH is used as a less polar solvent in a ratio of 1 g plastic/1 mL and, following the extraction period is then removed from the vial and diluted 100x with culture medium as the first test concentration. • Some 1 g/ml EtOH extracts are first concentrated 20x and then diluted 100x with culture medium as the first test concentration.
Aliquots of the extracts are then diluted 4-8 times with culture medium typically using a 5x dilution factor to obtain the 2 nd through 4 th (or up to 8 th ) test concentrations of the extract.
We add each test chemical or extract at each concentration in triplicate or quadruplicate to 96-well plates containing MCF-7 cells in EA-free culture media for our EA assay or in 10 -12 M E2 for our anti-EA assay. The position of test and control chemicals on a 96-well plate depends, in part, upon whether or not the test chemical or extract is volatile. Volatile chemicals are tested on a separate 96-well plate to avoid cross-contamination of adjacent wells. We often pre-screen test chemicals and extracts for volatility before performing an EA or anti-EA assay.
Cells in 96-well plates are incubated with test chemicals or extracts and controls at 37°C. The media and its contents are changed every other day for 6 days. At the end of the 6 day exposure, the media is removed, the wells washed once with HBSS, and then assayed to quantify amounts of DNA/well using a microplate modification of the Burton diphenylamine assay (Burton 1956;Natarajan 1994).
EA Confirmation: Estrogenic stimulation (as opposed to nonspecific stimulation) of MCF-7 proliferation (measured as increased amount of DNA per well) induced by the test chemical or extract is confirmed as estrogenic in an EA confirmation study: If the stimulation of MCF-7 proliferation by a test chemical or extract is suppressed by coincubation with an anti-estrogen (ICI 182,780 at 10 -7 -10 -8 M), the estrogenic activity of the test chemical or extract is then confirmed. Otherwise, the EA of the chemical or extract is NOT confirmed and the stimulation of MCF-7 cell proliferation is not via an ER-mediated mechanism.
Quantitative Calculation of EA: For test chemicals or extracts incubated for six days, the DNA/well produced by any dilution of a test chemical or extract is expressed as a percentage of the maximum DNA/well produced by the maximum response to 17βestradiol (E2, positive control) corrected by the DNA response to the vehicle (negative) control as given by Equation 1: For estrogenic test chemicals, the concentration needed to obtain half-maximum stimulation of cell proliferation (EC50, in M) is calculated from best fits to dose-response data that meet a well-defined set of criteria (see below) by Michaelis-Menton kinetics using GraphPad Prism. The estrogenicity of extracts is calculated as the relative maximum %E2 (%RME2) defined as a percentage of the maximum DNA/well produced by an extract at any dilution with respect to the maximum DNA/well produced by E2 at any dilution corrected by the DNA response to the vehicle (negative) control as given by Equation 2: If a test chemical has a positive response, but an EC50 can not be calculated because all criteria are not met, then the estrogenicity of the test chemical is simply characterized as EA positive or by its %RME2. Note that EC50s as calculated using Equation 1 give a measure of binding affinity (but not response amplitude), whereas %RME2s as calculated using Equation 2 give a measure of response amplitude (but not binding affinity).
The average of three or four replicates for each concentration are plotted as relative EA (% E2) with E2 = 100% and negative control (extract solution) = 0%. The standard deviations for the replicates are so small that they fall within space taken up by plots of averages of data for each concentration. The EC50 of E2 (50%E2) in this assay is historically 10 -13 M -10 -12 M depending on starting cell density/well and the maximum E2 response is attained at ~10 -11 M.
The EA of a test chemical or extract is considered detectable if it produces cell proliferation that is greater than 15% of the maximum response to E2 (aka > 15%RME2), which is greater than three standard deviations from the historic control baseline response (about 10 -15 M), i.e. a rather conservative measure of EA detectability. MCF-7 cell proliferation to 10 -11 M E2 is typically used as the positive reference control in assays of extracts to insure that E2 is indeed at maximum value. Note that for assays of extracts (Figure 1 D-F in main paper), we typically dilute 10 -9 M E2 100x to obtain 10 -11 M E2. Note that negative %E2 values can occur at concentrations at which a test chemical or extract induces less proliferation than the standard negative control; %RME2 values greater than 100% can occur when the chemical or extract induces more cell proliferation than the maximum produced by E2.
If a test chemical or extract does not induce MCF-7 cell proliferation or the stimulated cell proliferation cannot be inhibited by ICI 182,780, and all positive controls (E2 and/or EA-positive plastic resin control) show detectable EA, then the test chemical is said to have no detectable EA.

EA Assay Acceptance Criteria:
1) The negative control is the vehicle control. The vehicle control must not stimulate cell growth. 2) E2 must stimulate cell growth by at least 3 fold.
3) The EC50 of E2 must be within 2.5 SD of the historical mean established by CCi and have an R 2 (coefficient of determination) value ≥ 0.9 calculated by Hill kinetics. 4) The EC50 of a test chemical cannot be calculated unless the dose-response curve of a test chemical meet the following criteria: at least one data point on each EC50 plot of E2 control must be 10% -50% of the maximum response and at least one data point must be 50% -100% of the maximum response. At least two data points must be between 10% of the minimum (vehicle control) response, and an R 2 (coefficient of determination) value ≥ 0.8 calculated by Hill kinetics. 5) A weak plastic positive control must stimulate cell growth by at least 2 fold. 6) The stimulation of cell proliferation must be inhibited by ICI 182,780 at 10 -7 M -10 -8 M.
Calculation of anti-EA: As per ICCVAM recommendations, our robotic anti-EA assay uses 2x10 -12 M E2 to obtain an EC80 level of E2-stimulated MCF-7 cell proliferation (DNA E2 ). [Our historical data show that the EC80 of E2 is about 2x10 -12 M in our robotic EA assay] The data are plotted as relative anti-EA with ICI = 100% and negative control = 0%. If there is no cellular toxicity at higher concentrations, the anti-EA of a test chemical or extract can then be expressed as the IC50 concentration (in M) needed to reduce the E2-stimulated cell proliferation by 50% (IC50). When an IC50 cannot be calculated (dose-response curves do not meet the anti-EA acceptance criteria described below), the (normalized) anti-EA of a test chemical or extracts from plastics is expressed as percentage of normalized anti-EA of ICI 182,780 (ICI, positive control) as calculated by Equation 3:

Equation 3
Normalized %ICI = g/well) Plots of a single test run showing IC50 data for test or control chemicals are calculated as the mean of at least three replicates for each concentration. If a test chemical or extract does not inhibit E2-induced MCF-7 cell proliferation, and all positive controls (ICI, and/or dibenzoanthracene) show detectable anti-EA then the test chemical or extract is said to have no detectable anti-EA. If inhibition of E2-stimlated MCF-7 cell proliferation by a test chemical cannot be reversed by the presence of higher concentration of E2, then the test chemical or extract is considered to have no detectable anti-EA. Such inhibition is almost-certainly caused by cytotoxicity that is examined in a separate assay.

Anti-EA Assay Acceptance Criteria:
1) The negative control is the vehicle control. The vehicle control must not inhibit 2x10 -12 M E2-stimulated cell growth. 2) E2 must stimulate cell growth by at least 3 fold.
3) The IC50 of ICI 182,780 calculated by measuring the reduction in cell proliferation produced by 2x10 -12 M E2 by ICI in dilutions of 10 -7 M to 10 -13 M. The IC50 of ICI 182,780 must be within 2.5 SD of our historical mean established and have an R 2 (coefficient of determination) value ≥ 0.9 calculated by Hill kinetics. The reduction of E2-stimulated cell growth by ICI must be ≥ 3 fold 4) The IC50 of this test chemical can not be calculated unless the dose-response curve of a test chemical meets the following criteria: At least one data point on each IC50 plot must be 10% -50% of the maximum response and at least one data point must be 50%-100% of the maximum response; At least two data points must be between the minimum (vehicle control) response and the maximum response (10 -7 M ICI), and an R 2 (coefficient of determination) value ≥ 0.8 calculated by Hill kinetics 5) A test chemical or extract is considered to be anti-EA positive without a calculated IC50 if its dose-response curve does not meet the above criteria, but its reduction of cell proliferation is reversed by 100 -1000 fold increases in [E2].

MATERIALS AND PROCESSES IN PLASTICS PRODUCTION
Though outwardly simple, a plastic product, such as a recently-purchased baby bottle, uses many chemicals, chemical reactions, manufacturing processes, and decorating techniques. Plastic parts are typically made from resins using manufacturing processes (Supplemental Material, Figure 1) that collectively incorporate into the plastic part a great variety of chemicals, many of which might be expected to exhibit EA because they have physical-chemical properties, often an insufficiently-hindered phenol group, that enable them to bind to ERs (Supplemental Material, Table 1). Supplemental Material, Figure 1 is a simplified schematic of the production of only one component part of a plastic product, e.g., the bottle component of a contemporary baby bottle. Other components, such as the cap, gasket, and nipple have their own resin formulations and processing techniques, and hence increase the number of chemicals and processes used to make the final product. That is, each plastic part is usually a rather unique combination of 5-30 chemicals. A more complex plastic product (e.g., a baby bottle) often has 30 -100 or more different chemicals (some unintended, such as catalyst residues), all of which can typically leach from the product --any of which might have EA** --and none of which are typically revealed to the public because the chemical composition of the product is proprietary.
There are two principal classes of plastics, thermoplastics and thermosets. Thermoplastics are polymers which can be melted and solidified many times while thermosets are polymers whose final molded form is irreversible. Thermoplastics are the most prevalent plastics in the U.S. for food and beverage packaging due to their cost, performance, and ease of processing. Some examples of thermoplastic resin types are: COC, COP, Co-PET, HDPE, LDPE, PC, PE, PES, PET, PETG, PLA, PP, and PPCO (see main paper for definition of each abbreviation). Thermoset polymers include liquid silicone rubbers (used for bottle nipples), phenolics, epoxies, and polyurethanes. Baby bottles usually have component parts from each class.

Supplemental Material, Figure 1: Production of a typical plastic part
Legend. Diagram of steps and materials used to make a typical thermoplastic part starting with raw materials. A thermoplastic resin producer will begin with various monomers, link them together into long chains ("polymers"), which are subsequently mixed with small quantities of various additives (antioxidants, plasticizers, clarifiers, etc) and melted, mixed, extruded, and pelletized to form a resin. Resins are either used as-is or, more commonly, mixed with other resins, additives, colorants, lubricants, and/or extenders to form plastic compound resins (e.g., polymer blends and pre-colored polymers). Next, a compounder will combine additives that impart color, clarity, impact resistance, flexibility, and other desired final product qualities with dispersing agents and a carrier resin in specialized extruders to form a "masterbatch" material, which is essentially a concentrated form of the additives, suitable for adding to base resins. An end processor will then typically blend one or more base resins or compounds with one or more masterbatches to engineer the final product's look, feel, and performance before melting the blend for shaping. The final shaping or conversion usually involves molding and/or extrusion processes that expose the blend to heat and shear stresses. Once formed, the product is then decorated by printing and overcoating, which add polymer inks, pigments, and other additives to the chemical mix whose composition is proprietary and almost never revealed.
Thermoplastics used for most bottle component parts are made by polymerizing a specific monomer or monomers in the presence of catalysts into a high molecular weight chain known as a thermoplastic polymer. The resulting polymer is mixed with small quantities of various additives (antioxidants, plasticizers, clarifiers, etc) and melted, mixed, extruded, and pelletized to form a base thermoplastic resin. Base resins are either used as-is (e.g., BPA-based PC non-BPA-based PPCO, and non-BPA-based PPHO) or, more commonly, mixed with other resins, additives, colorants, and/or extenders to form plastic compounds (e.g., polymer blends and pre-colored polymers). Plastic products are then made by conversion processes (e.g., molding) using one or more plastic compounds or resins to form a finished plastic part that can be subjected to finishing processes that may utilize inks, adhesives, etc. (all of which often use chemicals having EA or anti-EA), to make a finished product.
Almost all of the 5-30 chemicals that are used to make a component plastic part can leach from any plastic product because polymerization is almost always incomplete (leaving residual unincorporated monomers) and/or because most additives (e.g., antioxidants) are not chemically part of the polymeric structure (Begley et al. 2005;De Meulenaer and Huyghebaert 2004;Le et al. 2008). Unless the selection of chemicals is carefully controlled, some of those chemicals will almost certainly have EA** -especially since many phenolics are typically used in all steps of manufacturing and decorating processes. An EA-free resin may end up in an EA**-containing product because chemicals in various finishing or marking materials had EA**.
Even when using all materials that initially test EA-free**, the stresses of manufacturing can change chemical structures or create chemical reactions to convert an EA-free** chemical into a chemical having EA**. Furthermore, after the product is purchased and put to its intended use, multiple common-use stresses such as UV light, microwave radiation, moist heat, or freezing can also convert EA-free** polymers or additives into chemicals exhibiting EA**. Accounting for each and all such factors may appear obvious or mundane, but are essential to producing EA-free** parts and products. Only recently have a few of these factors been explicitly considered individually, such as avoiding the use of individual chemicals having easily-detectable EA** (e.g., BPA, some phthalates). These factors have not been considered in aggregate by any scientific publication, government agency or commercial entity to date of which we are aware. Table 1 describes some quantitative structure-activity relationships (QSAR) of chemicals that help us predict whether a chemical will bind to mammalian ERs. Table 1: Physical-chemical properties of chemicals that affect binding to ERs. Based in part on Fang et al. 2003. 1. The molecular weight. The MW of phenol, 94 daltons, is assumed to be the lowest limit for a xenobiotic to bind to an ER, whereas 1000 daltons is assumed to be the upper limit of ER binding. Most chemicals having MW's >= 600 Daltons will not bind to the ERs. However, the degradation products of many chemicals exposed to stresses (heat, moisture, UV) experienced by plastic in normal use may bind to ERs. This result is often seen for chemicals having phenolic groups bound to linear carbon chains. 2. A ring structure. If a chemical contains no ring structure, it is exceedingly unlikely to be an ER ligand. A more rigid ring (e.g., benzene, triazine) structure favors ER binding. If a chemical has a non-aromatic ring structure, it is unlikely to be an ER ligand, assuming that it does not contain an O, S, N or other Hbonding atom. However, a non-aromatic chemical may transform into an aromatic structure in the presence of heat, catalysts, exposure to light, or other stresses. 3. A phenolic ring. The presence of a single phenolic (or hydroxyl triazine) ring is much more significant than any other structural or physical-chemical feature. The phenolic 3-OH group acts primarily as H-bond donor, although it can also act as an acceptor. The H-donor ability of the 3-OH group is especially affected by the nature of immediately adjacent ortho or para groups. The H-bond donor ability for several ortho-or para-substituted phenols show the trend: phenol > 2-methylphenol = 2-t-butylphenol > 2,6-dimethylphenol > 2,6-di-t-butylphenol, in which 2,6-di-t-butylphenol is not an H-bond donor, consistent with the lack of binding activity observed for 4,4'-methylenebis(2,6-di-t-butylphenol). A chemical with a phenolic structure is likely to bind to ER, but the degree of potency is dependent on the presence of key structural features described in 4-7 below. If a xenobiotic has a benzene ring without an -OH group, it can still bind to the ER, although its binding potential is then heavily dependent on the presence of key structural features described in 6-7 below. Furthermore, OH, Cl, and other groups that have appropriate characteristics described in 6-7 below are easily added to hexane, hexene, or benzene rings under stress conditions experienced by many plastic products in normal use.

Supplemental Material,
4. An H-bond donor mimicking the 17β-OH. 5. The distance between the oxygen atoms at 3-OH and 17β-OH. The spacing of two -OH groups at either end of a planar, and primarily hydrophobic, chemical is critical for ER binding. Chemicals containing only one phenolic group are likely to have weak to medium affinity for the ER. Most strong to medium ER ligands contain two -OH groups with an O-O distance ranging from 9.7 -12.3 Å. 6. Steric hydrophobic centers mimicking 7α and 11β steric configuration of 17β-estradiol. The precise steric size and orientation of the hydrophobic groups is as important as a 17β-OH. Chemicals containing a phenolic ring separated from another benzene ring with 0-3 bridge atoms will most likely be an ER ligand. The volume of the ER ligand-binding pocket (450 Å 3 ) is about twice as that of 17β-estradiol (245 Å 3 ). The length and breadth of the 17β-estradiol skeleton are well matched by the geometry of the ligand binding pocket, but there are large unoccupied cavities at the 7α and 11β -positions of 17β-estradiol. The positions of these cavities allow steric groups of certain sizes to fit, and are of great importance for various xenoestrogens, some of which do not have a benzene ring or OH groups. 7. The hydrophobicity of the entire molecule. The ligand-binding pocket, determined by X-ray crystallography of ERs, has a 3D "cross"-like shape with center and vertical ends mainly hydrophobic, and polar groups located at opposite ends of the horizontal cavity. When a direct comparison can be made in cases when properties 1-6 above are held constant, then increased hydrophobicity usually produces greater ER affinity. Xenoestrogens with groups lacking the most effective O-O distance require greater hydrophobicity to attain the same binding affinity exhibited by 17β-estradiol or xenoestrogens with 3α and 17β -OH groups.
Supplemental Material, Tables 2A-C summarize data on the percent of commercially available food-containing plastic products (%D) that have detectable EA (15%RME2) not yet exposed to common-use stresses. The products are grouped by whether the plastic container had contents (removed before analyzing the plastic container), and whether the container had no contents at the time of purchase. The products are also grouped by different categories of: 1) Plastic resins used to manufacture the product. [As discussed above, the manufacturing process can easily add EA-containing chemicals to an EA-free resin.]. 2) Type of plastic product [bags, baby bottles, water bottles, etc].
3) Type of retail outlet where the product was purchased [large retail chains not specializing in organic products vs. large retail chains specializing in organic products advertised for their health benefits].
One advantage of this particular study is that it realistically represents variability in release of chemicals having EA expected for purchase of commercially available plastic products. As described in the main paper, on average whether or not a product had contents before purchasing does not on average affect whether or not the product had detectable EA. That does not mean that the contents do not affect the release of chemicals having detectable EA in individual products. For example, on average the contents may decrease the ability to detect chemicals having EA from some products (perhaps by extracting them). Conversely, chemicals having EA in the contents can be absorbed by the polymer of the container --perhaps because the contents contained a common food antioxidant like BHA or BHT having EA. Most of the contents of such a bottle can be quickly extracted. The BHA or BHT absorbed in the walls of the container will be extracted much more slowly. Table 1 in the main paper and Supplemental Material, Table 2 summarize the detailed data presented in Supplemental Material, Table 5. Because of its length (over 20 pages), we have placed Supplemental Material, Table 5 at the end of this Supplemental Material. Note that each data point in Supplemental Material, Table 5 is the %RME2 derived from a dose-response curve for an extract of each of 455 plastic items, some of which (117) are analyzed using two or more protocols. Table 1 in the main paper summarizes data for all these 455 unstressed plastic products that had one or more assays whether or not the container ever had contents. Table 2 in the main paper summarizes the data for 102 of these unstressed plastic items that had two assays whether or not the item ever had contents. Supplemental Material, Tables 2A-C summarize the data for the 455 unstressed plastic products for each assay type according to whether the product had contents or did not have contents at the vtime of purchase.
Supplemental Material, Tables 2A-C shows that that most (71%) unstressed plastic items released chemicals having reliably-detectable EA in one or more extraction protocols, independent of whether or not the item had contents. This conclusion holds for the average numbers of items having detectable EA for different resin types, product types, or retailer types. There is no significant difference (p > 0.05) in the percentage of items having EA that had contents (69%, n=296) versus no contents (76%, n=160). The percentage of samples having EA or the mean %RME2 in individual subcategories of items also did not vary in any consistent manner between items with contents versus items without contents. Most importantly, items without contents in all categories exhibited detectable EA in at least one protocol, including 78% of items made from HDPE (n=18), 57% from PP (n=14), and 100% from PET (n=6). Given all these results, we presented the data for all items tested in Tables 1-2 in the main paper as representative for plastic items with or without contents. As noted in the main paper, this lack of significant difference in average percentage having detectable EA between plastic items with and without contents does not imply that the contents do not affect the total EA or specific chemicals having EA released by individual plastic items.
Supplemental Material, Tables 2A-C also shows that the leaching of a chemical having EA was significantly (p < 0.01) more likely to be detected if both polar (saline) and non-polar (EtOH) solvents were used (61%) than if only one solvent were used (15% or 31%) independent of whether or not the container had contents. This conclusion holds for products made from different resins, different product types, and products sold by different retailers.

Supplemental Material, Tables 2A-C: Number (N) and percent of unstressed plastic items with or without contents having EA (%D)
Supplemental Material,

With Content Without Content
Items tested only by concentrated EtOH assay

Concentrated EtOH With Content Without Content
Items tested by concentrated EtOH assay and at least one other assay Supplemental Material,

With Content Without Content
Items tested only by saline assay

Saline With Content Without Content
Items tested by saline assay and at least one other assay

Monomers and Additives in Plastics
Monomers: Figure 3 in the main paper shows the EA** of monomers for the most common thermoplastics. The pure ("barefoot") resins of most non-transparent flexible thermoplastics do not exhibit EA**, in large part because they do not have benzene or phenolic rings. In contrast, most HC thermoplastics have EA**. For a polymer to be HC it must have limited or no ability to crystallize and it must have a high glass transition temperature (T g ) well above room temperature. To have limited or no ability to crystallize, polymers typically have limited backbone symmetry, limited flexibility in the main chain segments and lack intermolecular polar or hydrogen bonds. To have a high T g , polymers typically have limited backbone symmetry, short main chain segments, bulky side groups, high intermolecular forces, strong hydrogen bonding and high molecular weight. Since polymers made from phenolic-based monomers usually have an amorphous structure and a high T g , their chemistry has been well studied and utilized to make HC polymers. However, as illustrated in Figure 3 in the main paper, COP/COC HC polymers produced from cyclic olefin monomers can also have a crystalline structure, high T g and can be both EA-free** and non-toxic. Their chemistry is less well explored than HDPE, PP, PET, PS, PLA, and PC.
Antioxidant Additives: Additives are chemicals that are introduced in small quantities to enhance a plastic's physical properties (e.g., plasticizers and slip agents), visual characteristics (e.g., clarifiers and colorants), or processability (e.g., antioxidants). Antioxidant additives are almost always a required component of almost all polymer resins, while other additives (e.g., colorants and slip agents) are optionally introduced during the conversion process. Antioxidants are the most critical additives because they prevent or minimize plastic degradation due to oxidation of a polymer to form carboxyl groups that break molecular chains (chain scission), which leads to discoloration, loss of surface gloss, surface cracking, and lowering of tensile strength (Schwarzenbach et al. 2001). The oldest and most common antioxidants deemed suitable for food contact belong to a chemical class known as hindered phenols. Organophosphites are a class of antioxidant commonly used with hindered phenols to provide synergistic oxidation protection, as well as whitening. Most organophosphites in their unaltered state should not bind to ERs. However, most organophosphites are also hydrolytically unstable and produce phenols when exposed to water (Schwarzenbach et al. 2001) and, therefore, usually cannot be used to produce EA-free** plastics. Supplemental Material, Table 3 shows the %RME2 of four unstressed antioxidants commonly used in plastic resins.

Colorant Additives:
The addition of colorants to a plastic formulation during the conversion process allows a firm to differentiate its product not only by its shape and design, but also by its color. Color can also provide the plastic product with additional UV protection. Supplemental Material, Table 4 shows that many common colorants test positive for EA** (> 15%RME2).
There are two basic types of colorants used in plastic products, inorganic and organic. QSAR analyses based on Supplemental Material, Table 1 predict, and our testing data confirm, that inorganic colorants typically do not exhibit EA**. The widespread use of inorganic colorants in plastic products is limited due to issues of availability, processability, or FDA-published toxicities. Organic colorants are more readily available, but many exhibit EA due to cyclic aromatic moieties. We have tested many colorants, and have found more than 40 inorganic and organic colorants that do not have EA** or FDA-published toxicities. Inorganic colorants are generally supplied as concentrated compounds, where a finely powdered inorganic pigment, usually a metal oxide, is dispersed in a plastic carrier. The carriers generally are a mix of low molecular weight polymers, waxes, and dispersing agents, all of which to be potentially usable to make an EA-free product must be tested to be EA-free (<15%RME2) using our MCF-7 assay.
Other Additives: Although not as critical to plastic manufacturing processes as antioxidants or colorants, other additives are utilized to improve the physical properties, optical properties (e.g. clarifiers) and marketability of a finished plastic product. A converter may use a purge compound to clean out the equipment and allow the converter to more efficiently transition between materials and products. However, most purge compounds used by converters are based on polystyrene and have EA. Converters may also use items such as mold release agents and mold cleaners. We have identified purge and mold release agents that have or do not have detectable EA.
In summary, even when using an EA-free resin, the resulting finished product can exhibit EA** due to additives used during conversion and/or finishing processes. Hence, EA-free** additives must be used in all stages of the manufacturing process. As discussed above, even when EA-free** monomers and additives are used, the final plastic item must still be tested for release of chemicals having EA** because manufacturing processes or common-use stresses can convert EA-free** chemicals to chemicals having EA**.  , Tables 5A-T   Supplemental Material, Tables 5A-S: %RME2 for individual samples with and without contents of a given classification (resin type, functional use, class of retailer) and the mean %RME2 ± SD for all samples having EA (>15%RME2) for each extraction protocol for that given classification.

Summary of Supplemental Materials
Supplemental Material, Table 5T: mean %RME2 ± SD for all samples having EA (>15%RME2) for each extraction protocol for all classification with contents, without contents, or with or without contents.
For each table in Supplemental Material, Tables 5A-S, sample numbers (first column: n 1 -n x ) are assigned in order of increasing EA (second column: %RME2 values), and the last column gives whether the item had contents at the time of purchase (y = yes, n = no). The same sample number (1, 2, 3, etc.) for different extraction protocols usually does not refer to the same item. The %RME2 for each sample is calculated (Equation 2) from its %E2 dilution curve relative to E2 (Equation 1) and confirmed by ICI suppression (see Fig. 1D-F in main paper). The mean and standard deviation (SD) is calculated for %RME2 values for all samples in each category having detectable %RME2 > 15%. N= number of samples for each calculation; Mean D = mean of samples having detectable EA; SD D = SD of all samples having detectable EA; N D = N number of samples having detectable EA.
The first set of Mean D ± SD D and N D values given for each extraction protocol for each subtable (e.g., 5A, 5B, etc) are for all items having detectable EA with or without contents, the next set of Mean D ± SD D and N D values are for items having detectable EA purchased with contents and the last set of Mean D ± SD D and N D values are for items having detectable EA purchased without contents.