Relative toxicological ranking of eight polybrominated diphenyl ether congeners using cytotoxicity, chemical properties and exposure data

Food and Chemical Toxicology 108 (2017) 74e84

Contents lists available at ScienceDirect

Food and Chemical Toxicology journal homepage: www.elsevier.com/locate/foodchemtox

Relative toxicological ranking of eight polybrominated diphenyl ether congeners using cytotoxicity, chemical properties and exposure data Sabrina Tait a, *, 1, Monia Perugini b, Cinzia La Rocca a, 1 a b


, Viale Regina Elena 299, 00161 Rome, Italy Dept. of Food Safety and Veterinary Public Health, Istituto Superiore di Sanita
di Bioscienze e Tecnologie Agro-alimentari e Ambientali, Teramo University, Localita
Piano d'Accio, 64100 Teramo, Italy Facolta

a r t i c l e i n f o

a b s t r a c t

Article history: Received 26 February 2017 Received in revised form 22 June 2017 Accepted 20 July 2017 Available online 21 July 2017

Polybrominated diphenyl ethers are widely used flame retardants which persist and diffuse in the environment thus entering the food chain. Eight congeners, most relevant for human exposure (BDE-28, 47, 99, 100, 153, 154, 183 and 209), were analyzed in vitro and in silico to derive a relative toxicological ranking. Cytotoxicity was assessed on human liver (HepG2) and colon (DLD-1) cell lines, by three assays (MTS, ATP and DNA content) in a range of realistic concentration (1pM - 10 nM). Jejunum and Caco-2 passive absorptions were calculated in silico. Exposure estimates were calculated using EFSA database. By ToxPi we integrated the overall data. No reduction of DNA content was observed, supporting absence of cytotoxicity. Otherwise, hormetic effects were exerted by all the congeners, except BDE-183. BDE-28, 47, 99, 100 differently affected the ATP content inducing a dose-related increase in HepG2 and depletion in DLD-1. Jejunum coefficients did not differ among congeners, whereas a higher Caco-2 coefficient indicates rapid absorption of BDE-28. ToxPi relative rankings support the toxicological relevance of BDE-153 and 28 congeners for their potential hazard; the inclusion of exposure data in young and adult populations shifted BDE-209 and BDE-47 as top ranked due to their widespread occurrence in the diet. © 2017 Elsevier Ltd. All rights reserved.

Keywords: Polybrominated diphenyl ethers Cytotoxicity ToxPi Toxicological ranking Gastrointestinal absorption Human exposure

1. Introduction Polybrominated diphenyl ethers (PBDEs) are a class of chemicals mainly used as flame retardants in several household and commercial products such as textiles, furniture and electronic devices, or used as additives to plastics. PBDEs share structural similarities with polychlorinated biphenyls (PCBs) with 209 possible congeners differing in the number and position of bromine atoms (EFSA, 2011). PBDEs are lipophilic and stable in the environment where they diffuse and persist. The consequent phenomenon of bioaccumulation and bio-magnification along the food chain raised concern on their potential adverse effects on wildlife and human health. Indeed, neurodevelopmental, endocrine and liver adverse effects have been documented for a number of PBDEs (Legler, 2008; Costa et al., 2014). Moreover, the increasing detection of several PBDE congeners in human compartments such as blood, cord * Corresponding author. E-mail address: [email protected] (S. Tait). 1 Current affiliation: Toxicology Unit, Center for Gender-Specific Medicine, Isti
, Viale Regina Elena 299, 00161 Rome, Italy. tuto Superiore di Sanita http://dx.doi.org/10.1016/j.fct.2017.07.041 0278-6915/© 2017 Elsevier Ltd. All rights reserved.

blood, placenta, breast milk, liver and adipose tissue demonstrates the hazard these compounds may represent, especially for vulnerable population groups as fetuses and children (Fromme et al., 2016). In Europe, diet represents the main route of exposure for humans with fish and meat providing the prevalent contribute (EFSA, 2011; U.S. Environmental Protection Agency, 2010). On the contrary, in U.S.A. the most relevant route of exposure for humans, accounting for the 90% of the total exposure estimate, is represented by dust ingestion in indoor environments like houses and offices, due to leaching from electronic devices (i.e. televisions, PC cabinets) or textiles, where PBDEs are added in variable amounts but not chemically bound (U.S. Environmental Protection Agency, 2010). On the basis of prevalence data in food, the European Food Safety Authority (EFSA) recommended to collect toxicological data on eight most relevant congeners, i.e. BDE-28, 47, 99, 100, 153, 154, 183 and 209 (EFSA, 2011). To date, only some of these congeners have been investigated for their cytotoxic effects using human in vitro models (Hamers et al., 2006; Huang et al., 2010; Llabjani et al., 2011; Wang et al., 2012; Souza et al., 2013, 2016, Pereira

S. Tait et al. / Food and Chemical Toxicology 108 (2017) 74e84

et al., 2013) and no study compared the cytotoxicity of all the eight congeners in the same experimental system. Due to the lack of a comparison scale, such as the Toxicity Equivalent Factors established for dioxin-like compounds (Van den Berg et al., 2006), EFSA performed the risk assessment for separated congeners which also differ in the mode of action (EFSA, 2011). The aim of the present study is to investigate potential toxicological differences among the eight dietary relevant PBDE congeners and to apply a method to classify their toxicological relevance by integrating results from in vitro cytotoxicity data, i.e. EC50/IC50 values or Benchmark doses, with chemical data on lipophilicity, gastrointestinal absorption and estimated half-lives, as well as with exposure intake data in different subpopulation groups. The cytotoxic potential of these compounds was investigated on two human cell lines representative of the digestive system, involved in metabolism and absorption of food nutrients and contaminants, namely the human hepatoma cell line (HepG2) and the adenocarcinoma cell line (DLD-1). This in vitro system previously proved to be useful to highlight different PCBs pleiotropic effects (Stecca et al., 2016). Hepatotoxicity is one of the main adverse effects ascribed to PBDEs (EFSA, 2011) and apoptotic effects induced by BDE-47, 99 and 209 have been observed on HepG2 (Hu et al., 2007, 2014; Souza et al., 2013; Wang et al., 2012). Otherwise, information about effects on colon cells is available only for BDE-209 in vitro (Curcic ndez et al., 2014) and for BDE-47 in a fish model (Barja-Ferna et al., 2013). For a more realistic assessment of potential human health effects, we considered concentrations
10 nM according to maximum estimated environmentally relevant PBDEs levels to be used in in vitro studies (Wei et al., 2010) which are in the range of mean occurring PBDEs concentrations in food (EFSA, 2011). Therefore, HepG2 and DLD-1 cells were treated with PBDE congeners’ concentrations ranging from 1 pM to 10 nM. We performed a cytotoxicity assessment of each congener on both cell lines by three methods, i.e. the metabolic MTS and ATP assays and the CyQuant® assay to determine the total amount of DNA. In silico calculation of Jejunum and Caco-2 passive absorption rates was also performed. To estimate the relative toxicity of the eight PBDE congeners evaluated, we used the Toxicological Prioritization Index (ToxPi) tool (Reif et al., 2013) a powerful platform which allow the integration of different sources of information to derive a single weighed score for each chemical. In the model, we included obtained cytotoxicity data, chemical properties derived by our in silico calculation or publicly available, as well as exposure intake data from public repositories. As a result, we obtained provisional toxicological ranks of the eight dietary relevant PBDE congeners for young and adult subpopulation groups. 2. Materials & methods 2.1. Chemicals

75

F12 and RPMI1640 media (Life Technologies, Paisley, UK), both without phenol red and supplemented with 10% fetal bovine serum (Lonza, Basel, Switzerland), 2 mM L-glutamine (Lonza), 100 U/ml penicillin and 100 mg/ml streptomycin (Life Technologies). Cells were maintained in a humidified Steri-Cult 200 Incubator (Forma Scientific, Marietta, OH, USA) at 37  C and 5% CO2. 2.3. Cell lines treatment with PBDEs Three different cytotoxicity assays were performed: 1) the colorimetric MTS assay to determine the amount of metabolically active cells able to reduce tetrazolium salts to formazan by dehydrogenases enzymes (Berridge et al., 2005); 2) the bioluminescent ATP assay to determine the intracellular amount of ATP as a parameter of proliferation/apoptosis (Crouch et al., 1993); 3) the fluorimetric CyQuant® assay to determine the total amount of DNA as directly proportional to cell number (Jones et al., 2001). According to the assay, 5000 (MTS and CyQuant®) or 2000 (ATP) cells/ well were plated in transparent 96 flat-bottomed multiwells for MTS and CyQuant® Assays and in white 96 flat-bottomed multiwells for ATP Assay. Cells were incubated overnight in a humidified incubator at 37  C to permit their adhesion. Medium was then replaced with fresh medium added with ten-fold serial dilutions of each PBDE congener in the range 1pMe10 nM, in triplicated wells, or with medium with vehicle alone (DMSO) as control, incubating cells for 72 h at 37  C. Final DMSO concentration did not exceed 0.2%. Three independent experiments were performed for each assay. Values were normalized with respect to control cells set at 100%. 2.4. MTS assay The Cell Titer 96® Aqueous One Solution reagent (Promega, Madison, WI, USA) was used to perform the MTS assay according to the provided protocol. Briefly, PBDEs treated cells in 96 flatbottomed multiwells were added, after 72 h incubation, with 20 ml of MTS reagent to each well, incubating 60 min at 37  C. Cell viability was determined by reading absorbance at 490 nm by a Victor 3 Multilabel Reader (PerkinElmer, Waltham, MA, USA). 2.5. ATP assay The ApoSENSORtrade; Cell Viability Assay (Biovision, Milpitas, CA, USA) was used to assess intracellular ATP levels following manufacturer's protocol. Briefly, after 72 h incubation, PBDEs treatment medium or control medium were removed from 96 flatbottomed white multiwells. Cells were added with 100 ml/well Nuclear Releasing Buffer incubating for 5 min at room temperature to lyse them. Next, 10 ml/well of ATP Monitoring Enzyme was added reading luminescence in a Victor 3 Multilabel Reader (PerkinElmer). 2.6. CyQuant® assay

PBDE congeners 28, 47, 99, 100, 153, 154, 183 and 209 were purchased by Wellington Laboratories (Ontario, Canada). Dr. Roberta Galarini and Dr. Arianna Piersanti from Istituto Zooprofilattico Abruzzo e Molise (in the frame of the Italian Ministry of Health funded project RF-2010-2311608) provided each congener as dry powder, following nitrogen flushing treatment to eliminate the organic solvent from the standard solutions. Upon arrival in our lab, each congener was dissolved in DMSO to obtain 50 mg/ml standard solutions which were stored at 4  C.

The CyQuant® Direct Cell proliferation Assay (Life Technologies) was used to assess DNA content according to manufacturer's protocol. Briefly, PBDEs treated cells in 96 flat-bottomed multiwells were added, after 72 h incubation, with 100 ml (equal volume as culture media) of 2X Detection Reagent. Upon incubation for 60 min at 37  C, fluorescence was red from bottom with a Victor 3 Multilabel Reader (PerkinElmer) using a standard green filter.

2.2. Cell lines

2.7. Calculation of EC50/IC50 values and benchmark doses

HepG2 and DLD-1 cell lines were grown, respectively, in DMEM/

The GraphPad Prism v5.01 software (GraphPad Software Inc., La

76

S. Tait et al. / Food and Chemical Toxicology 108 (2017) 74e84

Jolla, CA, USA) was used to calculate and visualize the best curve fit for each data set (i.e. each assay on each cell line), on the basis of higher R2 and lower p-values. According to the dose-response, logistic or bell-shaped curves were fitted; relative effective (EC50) and inhibitory (IC50) concentrations, corresponding to halfway the bottom and top responses of each curve, were then calculated without constraints on asymptotes. To derive more conservative reference concentrations, especially in the case of biphasic effects where less accurate EC50 and IC50 values may be obtained, we also calculated benchmark dose (BMD) values for each experimental data set, i.e. the dose corresponding to one standard deviation change from mean values of control untreated cells. Since BMD may be calculated only for data with a dose-response relationship, values observed at high doses deviating from the initial dose-response were conveniently excluded from the analysis. The BMDS Wizard, based on the BMDS v2.6.0.1 software (U.S. EPA, National Center for Environmental Assessment, Research Triangle Park, NC, USA), was used to calculate BMD values; upon input of each data set (including numerosity, mean and standard deviation values for each treatment group), the software performs 14 different curve modelling and provides a summary of the calculated parameters. On the basis of higher p-values and Akaike Information Criteria (AIC), we selected the best curve fit for each data set, annotating the corresponding BMD and lower confidence limit (BMDL). 2.8. In silico calculation of gastrointestinal passive absorption Due to the characteristics of the chemicals analyzed and the digestive tissues from which the in vitro models derive, we calculated predicted Jejunum and Caco-2 permeability coefficients (in cm/sec), as descriptors of passive absorption, using ACD/Percepta 2015 (Advanced Chemistry Development, Inc., Toronto, ON, Canada, www.acdlabs.com) (Reynolds et al., 2009). By entering PBDE congeners’ chemical structures, Jejunum coefficients were calculated with default settings, whereas Caco-2 coefficients were calculated applying the consensus algorithm using available octanol/water partition coefficients (log Kow) (U.S. Environmental Protection Agency, 2010; Environment Canada, 2010) and setting pH ¼ 7.4 and 500.000 rpm stirring rate as conditions.

(Supplementary Table 1). 2.10. Toxicological prioritization index calculation The ToxPi standalone GUI v1.3 software (Reif et al., 2013) was used to calculate ToxPi scores and relative toxicological ranking of the eight PBDE congeners. Following inclusion of information from different sources, the tool applies an algorithm to perform a weighed combination of all data, for each compound, providing as a result dimensionless values, i.e. the ToxPi scores. Higher ToxPi values correspond to more toxicologically relevant compounds. In our model, we included data from three different categories: Assays, Chemical Properties and Exposure. For the domain Assays, we used the EC50, IC50 or BMD values obtained from our cytotoxicity data, expressing concentrations as mM (conventional ToxCast-style) and scaling values as -log10(concentration)þ6 values within ToxPi. Negative responses, i.e. no extrapolated/available values, were conventionally assigned a 106 value, which became zero upon scaling. Couples of EC50/IC50 values derived by each curve fit were included in the same slice, thus obtaining three slices for the assays performed in HepG2 cells and three slices for the assays performed in DLD-1 cells, all equally weighed. For the domain Chemical Properties, we used the available log Kow (U.S. Environmental Protection Agency, 2010; Environment Canada, 2010) as descriptors of lipophilicity, the in silico calculated Jejunum and Caco-2 permeability coefficients and the estimated elimination half-lives of the eight PBDEs in humans, as reviewed by EFSA (EFSA, 2011) for a total of four equally weighed slices. All the coefficients included in the Chemical Properties domain are summarized in Table 1. Finally, for the Exposure domain we used the calculated exposure estimates for young or adult population groups in the eight food categories corresponding to as many equally weighted slices. The obtained ToxPi scores are visualized as circles composed of as many slices as the included components. The distance from the center of each slide is proportional to the normalized value for the corresponding data set (i.e. values obtained for each compound divided by the maximum value observed across compounds for that category). Thus, the farther is the slice from the center the more potent is the effect in that category. 2.11. Statistical analysis

2.9. Calculation of dietary exposure estimates For each PBDE congener, we collected minimum (lower bound, LB) and maximum (upper bound, UB) mean occurrence values (in ng/g wet weight) across eight main food categories, according to the FoodEx classification (Eggs and egg products, Milk and dairy products, Meat and meat products, Fish and other seafood, Animal and vegetable fats and oils, Products for special nutritional use, Vegetables and vegetable products, Food for infants and small children), as reported by EFSA (EFSA, 2011). Such values were then multiplied with median chronic consumption values of same food categories (in g/kg bw per day) by different population groups (i.e., infants, toddlers, children, adolescents, adults, elderly) derived by stratification of the EFSA Comprehensive European Food Consumption Database (https://www.efsa.europa.eu/en/foodconsumption/comprehensive-database). We thus obtained LB and UB (in ng/kg bw per day) dietary exposure estimates for each population group. Although relevant for human exposure, the house dust ingestion contribute was not considered due to unavailability of occurrence and ingestion data in all subpopulation groups. Calculated chronic exposure estimates values for the eight PBDE congeners across the food categories in the different population groups are available as Supplementary Material

The statistical analysis was performed with the JMP 9.0 Software (SAS Institute Inc., Cary, NC, USA). All the data were normally distributed and with homogeneous variances as assessed, respectively, with the Shapiro Wilk and the Levene tests. Thus, statistical differences among treatment groups and control cells were evaluated by the Analysis of the Variance (ANOVA) followed by post-hoc Dunnett's pair-wise comparisons where appropriate. Results with p < 0.05 were considered significant. 3. Results 3.1. Cytotoxicity of PBDEs on HepG2 cells The eight PBDE congeners considered in this study determined quite different dose-responses on HepG2 cells according to the assay and the compound analyzed (Fig. 1). In Table 2 are reported EC50 or IC50 values calculated on the basis of fitted curves, corresponding to concentrations giving 50% of the maximum response. When biphasic curves were obtained, both EC50 and IC50 values were calculated. Similarly, BMD values are also reported. All the congeners induced a dose-response increase in ATP intracellular content, although with different curve trends.

S. Tait et al. / Food and Chemical Toxicology 108 (2017) 74e84

77

Table 1 In silico calculated and publicly available PBDE congeners’ chemical properties. Congener

log Kowa

Jejunum permeability (cm/sec)c

Caco-2 permeability (cm/sec)c

Estimated elimination half-lives (days)d

BDE-28 BDE-47 BDE-99 BDE-100 BDE-153 BDE-154 BDE-183 BDE-209

5.94 6.81 7.32 7.24 7.9 7.82 8.27 8.7b

0.000857 0.000832 0.00081 0.00081 0.00079 0.00079 0.000772 0.000726

0.000148 0.00004 0.000014 0.000016 0.000004 0.000004 0.000002 0

1100 664 1040 573 2380 1214 94 15

a b c d

Log Kow values available from EPA opinion (U.S. Environmental Protection Agency, 2010). Updated log Kow value from Environment Canada (Environment Canada, 2010). Jejunum and Caco-2 permeability coefficients calculated with ACD/Percepta 2015. Estimated elimination half-lives available on EFSA opinion (EFSA, 2011).

Noteworthy, such ATP increase was not always reflected by a parallel increase in DNA content, and thus to an augmented cellular proliferation. In fact, except BDE-183 which determined almost the same increase of ATP and DNA content in the range of concentrations tested, all the other compounds exerted divergent effects on these two end points. In particular, both BDE-28 and 100 increased the DNA content, reaching a maximum at 1010 M and then decreasing, but only BDE-28 exerted such induction in a significant manner at 1010 (p ¼ 0.008) and 109 M (p ¼ 0.0371); meanwhile, the ATP content always rose with the concentration being significantly higher compared to controls in HepG2 treated with BDE-28 1010 to 108 M (p ¼ 0.0382, p ¼ 0.0013 and p ¼ 0.0004, respectively for BDE-28 1010, 109 and 108) and with BDE-100 109 and 108 M (p ¼ 0.0175 and p ¼ 0.0016, respectively). Thus, a minimum in ATP/DNA content was observed for BDE-28 at 1011 M and for BDE-100 at 1010 M (Fold change, FC ¼ 0.79 and 1.02, respectively), whereas the maximum ATP/DNA ratio was induced at 108 M by both congeners (FC ¼ 1.20 and 1.60, respectively). Otherwise, in HepG2 treated with BDE-99, 153 and 154, ATP decreased when DNA increased and vice versa. A significant difference with respect to control cells was observed only for the ATP content being increased by BDE-99 at 108 M (p < 0.0001). Maximum variation in ATP/DNA ratio occurred at 1011 M of BDE99 (FC ¼ 1.21), 109 and 108 M of BDE-153 (FC ¼ 0.74 and 1.25 respectively) and 1012 M of BDE-154 (FC ¼ 0.78). Parallel trends were observed for BDE-47 and BDE-209 only at the lower doses, then ATP continued to increase with the dose and the DNA content remained almost equal or decreased, respectively for BDE-47 and 209. Both congeners, at 108 M, induced a significant ATP increase compared to controls (p ¼ 0.0016 and p ¼ 0.0395, respectively for BDE-47 and 209); accordingly, a maximum ATP/ DNA ratio was reached (FC ¼ 1.44 and 1.77, respectively). The MTS was the less sensitive assay since it did not change significantly across the range of concentrations tested apart a slight biphasic response observed in BDE-28-treated cells with a significant increase at 1010 M (p ¼ 0.0221), as well as a significant increase in cells treated with BDE-47 at 1010 M (p ¼ 0.0076), BDE209 at 109 and 108 M (p ¼ 0.0343 and p ¼ 0.0171, respectively) and BDE-153 at 1012 M (p ¼ 0.0135). 3.2. Cytotoxicity of PBDEs on DLD-1 cells Conversely to what observed in HepG2, a decrease in ATP intracellular content was significantly induced in DLD-1 at 108 M by the lower brominated PBDE congeners BDE-28, 47 and 99 (p < 0.0001 for all), by BDE-100 at both 109 and 108 M (p < 0.0001 for both) (Fig. 2) and by BDE-183 at 109 M (p ¼ 0.039). Only BDE-209 increased the ATP content at the highest concentration (p ¼ 0.0242).

The DNA content decreased only slightly, although not significantly, at the highest dose of BDE-28, 153 and 100 treatments, whereas it remained unchanged upon BDE-47, 183 and 209 exposures. A biphasic, but not significant, change in DNA content was observed across the BDE-99, 100 and 154 ranges of concentrations. Thus, the ATP/DNA ratio reached minimum values at 108 M for all the four lower brominated congeners (FC ¼ 0.41 for BDE-28 and 47; FC ¼ 0.44 and 0.52, respectively, for BDE-99 and 100), and at 109 M for BDE-100 (FC ¼ 0.44). Otherwise, a maximum in ATP/DNA ratio was reached at 108 M of BDE-209 (FC ¼ 1.33). The MTS assay showed only a slight increase in cells treated with 109 M BDE-47 (p ¼ 0.0112) and with 109-1010 M BDE-99 (p ¼ 0.0141 and p ¼ 0.0112, respectively). Otherwise, a significant decrease was detected at 109 and 108 M of BDE-153 (p ¼ 0.0236 and p ¼ 0.0167, respectively), at 109 M of BDE-154 (p ¼ 0.0003) and at 10-8 M of BDE-209 (p ¼ 0.0284). 3.3. In silico calculation of passive absorption coefficients The in silico calculated Jejunum permeability coefficients did not show relevant differences among the eight PBDE congeners considered (Table 1). On the other hand, the calculated Caco-2 passive absorption rates showed progressively lower values with increasing degree of bromination. Indeed, the highest permeability coefficient was observed for the lower brominated BDE-28. 3.4. Toxicological ranking of the 8 PBDE congeners The ToxPi tool (Reif et al., 2013) was used to get a toxicological ranking of the 8 PBDE congeners by a tiered approach. As a first step, we used our cytotoxicity data by including calculated EC50/ IC50 or BMD values (Tables 2 and 3). As shown in Fig. 3, two different rankings were obtained, due to the different algorithms applied in the model fits and to the number of available EC50/IC50 or BMD values. However, in both cases, BDE-153 had the highest and BDE-28 the second ToxPi score. Overall the ranking order did not reflect the degree of bromination. Comparing the two toxicological ranking, the congeners 154, 99 and 183 differed only for a position; on the other hand, BDE-47 and 209 were the last two congeners when considering EC50/IC50 values and in the middle of the ranking when considering BMD values. The congener which differed more was BDE-100, appearing in the third or in the last positions, according to EC50/IC50 or BMD inclusion, respectively. The addition of the Chemical Properties in the ToxPi framework, i.e. the log Kow values, the calculated Jejunum and Caco-2 permeability coefficients as well as the estimated elimination half-lives (Table 1), increased the overall score values but did not affect the ranking order (Fig. 4), apart the reversal in the relative position of

78

S. Tait et al. / Food and Chemical Toxicology 108 (2017) 74e84

Fig. 1. Effects of BDE-28, 47, 99, 100, 153, 154, 183 and 209 congeners treatment on MTS reduction (red), ATP (blue) and DNA (green) content in HepG2 cells. Data are expressed as percentage of control cells (±SEM); lines represent the corresponding best fit model. Each experiment was performed in triplicate and repeated three times. Asterisks (colored according to the curve) indicate the level of significance vs controls: *p < 0.05, **p < 0.01, ***p < 0.001. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

S. Tait et al. / Food and Chemical Toxicology 108 (2017) 74e84

79

Table 2 EC50, IC50 (calculated with GraphPad Prism) and BMD values (calculated with BMDS Wizard), according to best fit models obtained for each assay, for each congener, in HepG2 cells. congener

MTS EC50

BDE-28 BDE-47 BDE-99 BDE-100 BDE-153 BDE-154 BDE-183 BDE-209 a b

4.57E-12 ND ND ND eb ND ND ND

ATP IC50 3.18E-10 ND ND ND 3.44E-12 ND ND ND

BMD a

ND ND ND ND 1.23E-12 ND ND 7.10E-11

DNA

EC50

IC50

BMD

EC50

IC50

BMD

1.28E-10 6.52E-09 8.17E-08 9.69E-11 1.32E-11 9.90E-12 1.40E-11 3.96E-09

e e ND e ND ND e e

2.80E-11 1.32E-10 ND 5.32E-11 3.00E-12 6.49E-12 7.66E-11 9.59E-10

1.32E-11 ND 6.98E-11 6.17E-11 6.57E-11 6.15E-11 8.03E-12 e

5.70E-10 ND e 1.15E-09 3.16E-09 ND e 6.81E-06

3.83E-13 2.13E-12 9.91E-11 3.35E-11 3.47E-10 2.41E-11 3.60E-13 2.50E-10

ND: values Non Defined due to the absence of convergence for any non-linear fit tested. Values not calculated due to the monotonic fit of the corresponding curve.

BDE-47 and 183 congeners when considering EC50/IC50 values (Fig. 4A). As a final step, we included human exposure data calculated considering PBDEs concentrations detected in food and the daily chronic consumption data for different subpopulation groups (Supplementary Table 1). Both LB and UB occurrence data were evaluated (EFSA, 2011); since very minor differences were observed, only results considering UB occurrence data are shown for brevity (Fig. 5). Although the consumption of some food categories differed a little, the proportion in the intake of the 8 congeners in infants, toddlers, children and adolescents was similar, thus we obtained an identical toxicological ranking for these population groups (Fig. 5A and B, respectively including EC50/IC50 or BMD values); similarly we obtained the same ranking for adult and elderly subgroups (Fig. 5C and D, respectively including EC50/IC50 or BMD values). Due to the widespread occurrence of BDE-209 in the different food categories, it shifted to the top of the toxicological ranking for both young and adult populations, regardless the cytotoxicity set of values considered. Similarly, BDE-47 rose in the ranking; in particular, when including BMD values it reached the second position for both young and adults groups, whereas when including EC50/IC50 values the toxicological relevance was higher in the younger group. Interestingly, although a lower occurrence in foods, comparable in young and adult groups, BDE-153 and BDE-28 were always within the first four ranking positions. The order BDE99 > 100 > 183 was observed in all the toxicological ranking, with almost same scores between population groups. Otherwise, BDE154 was more variable, being toxicologically more relevant when considering BMD values, especially in adults. 4. Discussion In the present study, we evidenced different toxicological potentials of the eight PBDE congeners most relevant for human exposure through diet (EFSA, 2011; U.S. Environmental Protection Agency, 2010), in a range of realistic occurring concentrations. The data integration from different sources provided relative toxicological ranking of such PBDE congeners. The combined use of the HepG2 and DLD-1 human cell lines derived, respectively, by liver and colon tissues, proved to be a useful tool to highlight different tissue-dependent effects of chemicals, as previously observed for PCBs (Stecca et al., 2016). The eight PBDE congeners did not significantly decrease cell vitality in the range of concentrations tested, as indicated by the CyQuant® assay, the more accurate as regard correlation with proliferation (Quent et al., 2010) and which is not affected by cellular metabolic changes (Jones et al., 2001). On the contrary, the BDE-28, 99, 100, 153, 154 congeners showed an hormetic effect in

both cell lines, i.e. inducing cell proliferation at lower doses followed by a decline at higher doses. The congeners BDE-47 and 209 showed such effect only in DLD-1 and HepG2 cells, respectively. The hormetic effect of BDE-47 in HepG2 cells has been previously reported, in the same range of concentrations (Wang et al., 2012). The apparently dissimilar results may be attributed to the assays used, differing in the magnitude of the output response; indeed, we observed a slight significant increase at 109 M in the MTS assay but not an evident biphasic curve. Hormetic effects induced by BDE-47, 153, 183 and 209 congeners were observed also in MCF7 breast cancer cells at same concentrations (Llabjani et al., 2011). Thus, overall, our results confirm and extend the previous evidence for PBDEs as inducers of hormetic effects, in HepG2 and DLD-1 cells, in the pico-nanomolar range of concentrations. Noteworthy, the ATP assay did not correlate with the other two cytotoxic tests performed, in particular the CyQuant® which measures the DNA content. Instead, the ATP content appeared a sensitive end-point of PBDE congeners’ effect, especially for the lower brominated compounds. More interestingly, the ATP intracellular level was increased in HepG2 and decreased in DLD-1 cells by the congeners BDE-28, 47, 99 and 100, whereas BDE-209 increased the ATP content in both cell lines at 108 M. Other authors previously supported that uncorrelated results, obtained with different cytotoxicity methods, may be determined by different underlying mechanisms of the studied compounds. This may especially occurs for metabolic assays such as MTS and ATP measuring different mitochondrial parameters, i.e. the dehydrogenases activity and the intracellular ATP amount, respectively (Weyermann et al., 2005; Ulukaya et al., 2008). Induction of ATP and other metabolites involved in the maintenance of the cellular energy state was observed in mussels and earthworms treated with BDE-47 in the nanomolar range, suggesting a possible increase in energy requirements associated with impaired osmosis (Ji et al., 2013a, 2013b). Otherwise, ATP depletion attributable to mitochondrial membrane permeability impairment was evidenced in rats liver mitochondria exposed to BDE-47, 99 or 100 in the micromolar range (Pereira et al., 2013; Pazin et al., 2015). In both cases, mitochondria homeostasis was affected thus, direct consequences on the cell cycle may be speculated (Renner et al., 2003). Indeed, a decrease in the percentage of cells in G0/G1 phases, supportive of an anti-apoptotic effect, was observed in HepG2 upon treatment of BDE-47 at 1010-109 M (Wang et al., 2012). Since different outcomes were exerted on HepG2 and DLD-1 cells by same compounds, alternative effects on apoptosis and mitochondrial integrity may occur according to the target tissue; thus, comparative investigations on these end-points in the nanomolar range are worth exploring. The integration of the calculated EC50/IC50 values within the ToxPi framework tool pointed out the hexa-brominated BDE-153

80

S. Tait et al. / Food and Chemical Toxicology 108 (2017) 74e84

Fig. 2. Effects of BDE-28, 47, 99, 100, 153, 154, 183 and 209 congeners treatment on MTS reduction (red), ATP (blue) and DNA (green) content in DLD1 cells. Data are expressed as percentage of control cells (±SEM); lines represent the corresponding best fit model. Each experiment was performed in triplicate and repeated three times. Asterisks (colored according to the curve) indicate the level of significance vs controls: *p < 0.05, ***p < 0.001. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

S. Tait et al. / Food and Chemical Toxicology 108 (2017) 74e84

81

Table 3 EC50, IC50 (calculated with GraphPad Prism) and BMD values (calculated with BMDS Wizard), according to best fit models obtained for each assay, for each congener, in DLD1 cells. congener

MTS EC50

BDE-28 BDE-47 BDE-99 BDE-100 BDE-153 BDE-154 BDE-183 BDE-209 a b

a

ND 1.80E-10 4.62E-10 e 1.20E-11 e e e

ATP IC50 ND 7.52E-11 e e 1.73E-10 1.65E-10 e e

BMD 4.47E-13 4.39E-09 6.12E-10 e 1.44E-13 2.27E-10 e 4.23E-11

EC50 b

e e e e e 6.77E-10 3.16E-09 4.80E-08

DNA IC50

BMD

EC50

IC50

BMD

1.40E-07 7.82E-09 1.79E-09 5.43E-10 3.73E-10 e 7.79E-11 e

7.33E-09 1.49E-09 8.99E-10 2.72E-10 1.43E-08 6.32E-10 4.20E-11 4.03E-09

e e 3.16E-12 2.70E-11 e e e e

6.62E-10 e 3.88E-07 2.31E-09 2.62E-10 2.67E-11 e e

3.51E-09 9.76E-12 8.97E-12 ND 7.77E-10 3.21E-11 e e

ND: values Non Defined due to the absence of convergence for any non-linear fit tested. Values not calculated due to the monotonic fit of the corresponding curve.

Fig. 3. ToxPi slices profiles of the eight PBDE congeners calculated by including A) EC50/IC50 values or B) BMD values. The legend on the bottom left indicates which data each slice represents. The distance from the center of each slide is proportional to the normalized values for the corresponding data set. External black dotted arches and internal white dotted arches represent, respectively, upper and lower 95% confidence limits. The corresponding ToxPi scores are reported below each congener's name.

congener as the most toxicologically relevant, in the context of our in vitro model. The second higher score was obtained for BDE-28, a mono-brominated congener, followed by BDE-100, 154, 99, 183, 47 and 209. Such toxicological ranking cannot be compared to previous evidence because not all the eight BDE congeners have been tested for cytotoxicity in same experimental models (Llabjani et al., 2011; Pereira et al., 2013; Huang et al., 2010; Souza et al., 2013, 2016). The only in vitro study evaluating potential differences in endocrine disrupting potencies of seven out of the eight PBDE congeners herein studied (except BDE-154), indicated BDE-100 as the more hazardous compound, followed by BDE-28, 183, 153, 47, 209, 99, as derived by approximatively ranking the provided hierarchical clustering (Hamers et al., 2006). Thus, despite the difference in the range of concentration tested (pico-nanomolar versus nano-micromolar range), the assays performed and the in vitro models used, overall the two studies agree on BDE-28 and 100 congeners as toxicologically high ranked and on BDE-47 and 209

congeners as less toxic. The use of the BMD approach is still limited in in vitro studies, however, its application is advisable in toxicological as well as in epidemiological studies for a better characterization of potential risks for human health (Hardy et al., 2017). Indeed, the BMD is a more conservative reference concentration for the risk assessment of toxic compounds, compared to EC50 and IC50 values, since it corresponds to an earlier response level, i.e. one standard deviation or 5% difference from control mean values, according to the method used (Davis et al., 2011; Hardy et al., 2017). In our model, when considering BMD values, BDE-47 and 209 were higher in the ranking, BDE-100 was in the last position, but BDE-153 and 28 were still top ranked, thus confirming their toxicological relevance. All the PBDE congeners tested are not markedly different for either the log Kow values, due to the high lipophilicity, or to the Jejunum coefficients. Else, BDE-153 has the longer estimated halflife excretion of about 6.5 years (EFSA, 2011), which justifies its

82

S. Tait et al. / Food and Chemical Toxicology 108 (2017) 74e84

Fig. 4. ToxPi slices profiles of the eight PBDE congeners calculated by adding chemical properties coefficients to A) EC50/IC50 values or B) BMD values. The legend on the bottom left indicates which data each slice represents. The distance from the center of each slide is proportional to the normalized values for the corresponding data set. External black dotted arches and internal white dotted arches represent, respectively, upper and lower 95% confidence limits. The corresponding ToxPi scores are reported below each congener's name.

frequent detection in human compartments (Bramwell et al., 2016), and BDE-28 has the highest calculated Caco-2 passive absorption coefficient. Our in silico calculation is in agreement with a very recent report on absorption profiles of BDE-28, 47, 99 and 153 determined in vitro on Caco-2 cells, where lower brominated congeners, especially BDE-28, showed higher absorption rates (Kang et al., 2016), thus being more easily transported in and out the cells by transport proteins. The elevated gastrointestinal absorption of BDE-28 compared to the other congeners, associated with a long estimated half-life excretion (about 3 years), raises the concern on its bioavailability and the potential hazard posed for human health. The inclusion of the chemical properties in the ToxPi framework did not change the toxicological ranking and further supported the toxicological hazard of BDE-153 and 28 congeners. For a more comprehensive risk assessment, we included in the ToxPi framework also exposure estimates data through diet of the eight PBDE congeners for different population groups. Only two ToxPi rankings were obtained, corresponding to young (i.e. infants, toddlers, children and adolescents) and adult (i.e. adult and elderly) individuals, for each cytotoxicity dataset used (EC50/IC50 or BMD). This occurred because the ToxPi algorithm assigns a score after ordering the values, for each variable, ranging from 0 (not active) to 1 (more active). Therefore, this implies that proportions in the intake of food categories within young or adult subgroups, and thus exposure to PBDEs, are the same even though absolute quantities differ. Due to the very high occurrence of BDE-209 in almost all food commodities, it jumped to the top of all the toxicological rankings. A noticeable score increase was observed also for BDE-47 and 99, the other two prevalent congeners (EFSA, 2011; U.S. Environmental Protection Agency, 2010). Noteworthy, although a lower assumption by food intake, BDE-153 and 28 were always found within the first four positions due to the relevant contribution of the other two components, i.e. Assay and Chemical Properties data. Only minor differences were observed between rankings for young and adults,

with BDE-47 or BDE-99 being more relevant for young groups when considering, respectively, the EC50/IC50 or the BMD cytotoxicity dataset. The present approach demonstrated how different sources of information, i.e. in vitro, in silico and publicly available, may be integrated to provide provisional toxicological indexes useful for further focused research (Reif et al., 2013). In this context, our results prompt to better investigate the toxicological hazard of BDE153 and 28 congeners. The power of the model is that it may be implemented with additional and continuously increasing toxicity data, including in vivo evidence, on these dietary relevant PBDEs. The present findings may represent a useful contribution to the toxicological risk assessment of PBDEs, proposing a prioritization ranking in young (BDE-209 > 47 > 153 > 28 > 99 > 154 > 100 > 183) and adult (BDE-209 > 47 > 153 > 28 > 154 > 99 > 100 > 183) population groups considering the more conservative BMD values. 5. Conclusions By testing the cytotoxicity of eight PBDE congeners with three different assays (i.e., MTS, ATP and CyQuant) on a hepatic (HepG2) and an enteric (DLD-1) cell lines, we evidenced the propensity of the PBDE congeners, except BDE-183, to induce hormetic effects in a range of dietary relevant concentrations. Moreover, the ATP intracellular content appeared a sensitive end-point of PBDEs’ mode of action, with tissue-dependent alterations. By in silico calculations we determined that BDE-28 has the highest Caco-2 coefficient, thus indicating a rapid passive absorption of this congener. By performing integration of cytotoxicity, chemical and exposure data within the ToxPi tool we derived provisional toxicological rankings of the eight dietary relevant PBDE congeners in young and adults. Among the four more toxicologically relevant congeners, further investigations on BDE-153 and 28 effects are suggested besides the better known BDE-209 and 47.

S. Tait et al. / Food and Chemical Toxicology 108 (2017) 74e84

83

Fig. 5. ToxPi slices profiles of the eight PBDE congeners calculated by including EC50/IC50 values (A,C) or BMD values (B,D), chemical properties coefficients and UB daily exposure estimates in young (AeB) or in adult population subgroups (CeD). The legend on the bottom indicates which data each slice represents. The distance from the center of each slide is proportional to the normalized values for the corresponding data set. External black dotted arches and internal white dotted arches represent, respectively, upper and lower 95% confidence limits. The corresponding ToxPi scores are reported below each congener's name.

Acknowledgements

Appendix A. Supplementary data

This work was supported by the Italian Ministry of Health [grant number RF-2010-2311608].

Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.fct.2017.07.041.

84

S. Tait et al. / Food and Chemical Toxicology 108 (2017) 74e84

Transparency document Transparency document related to this article can be found online at http://dx.doi.org/10.1016/j.fct.2017.07.041. References  ndez, S., Míguez, J.M., Alvarez-Otero, Barja-Ferna R., 2013. Histopathological effects of 2, 20 , 4, 40 -tetrabromodiphenyl ether (BDE-47) in the gills, intestine and liver of turbot (Psetta maxima). Ecotoxicol. Environ. Saf. 95, 60e68. Berridge, M.V., Herst, P.M., Tan, A.S., 2005. Tetrazolium dyes as tools in cell biology: new insights into their cellular reduction. Biotechnol. Annu. Rev. 11, 127e152. Bramwell, L., Glinianaia, S.V., Rankin, J., Rose, M., Fernandes, A., Harrad, S., PlessMulolli, T., 2016. Associations between human exposure to polybrominated diphenyl ether flame retardants via diet and indoor dust, and internal dose: a systematic review. Environ. Int. 92, 680e694. Costa, L.G., de Laat, R., Tagliaferri, S., Pellacani, C., 2014. A mechanistic view of polybrominated diphenyl ether (PBDE) developmental neurotoxicity. Toxicol. Lett. 230, 282e294. Crouch, S., Kozlowski, R., Slater, K., Fletcher, J., 1993. The use of ATP bioluminescence as a measure of cell proliferation and cytotoxicity. J. Immunol. Methods 160, 81e88. Curcic, M., Durgo, K., Kopjar, N., Ancic, M., Vucinic, S., Antonijevic, B., 2014. Cadmium and decabrominated diphenyl ether mixture: in vitro evaluation of cytotoxic, prooxidative and genotoxic effects. Environ. Toxicol. Pharmacol. 38, 663e671. Davis, J.A., Gift, J.S., Zhao, Q.J., 2011. Introduction to benchmark dose methods and US EPA's benchmark dose software (BMDS) version 2.1.1. Toxicol. Appl. Pharmacol. 254, 181e191. EFSA, 2011. Scientific opinion on polybrominated diphenyl ethers (PBDEs) in food. EFSA J. 9 (5), 2156. Environment Canada, 2010. Ecological State of the Science Report on Decabromodiphenyl Ether (decaBDE) - Bioaccumulation and Transformation. 2010. €lkel, W., 2016. Brominated flame retarFromme, H., Becher, G., Hilger, B., Vo dantseExposure and risk assessment for the general population. Int. J. Hyg. Environ. Health 219, 1e23. Hamers, T., Kamstra, J.H., Sonneveld, E., Murk, A.J., Kester, M.H., Andersson, P.L., Legler, J., Brouwer, A., 2006. In vitro profiling of the endocrine-disrupting potency of brominated flame retardants. Toxicol. Sci. 92, 157e173. Hardy, A., Benford, D., Halldorsson, T., Jeger, M.J., Knutsen, K.H., More, S., Mortensen, A., Naegeli, H., Noteborn, H., Ockleford, C., 2017. Update: use of the benchmark dose approach in risk assessment. EFSA J. 15. Hu, X., Xu, Y., Hu, D., Hui, Y., Yang, F., 2007. Apoptosis induction on human hepatoma cells Hep G2 of decabrominated diphenyl ether (PBDE-209). Toxicol. Lett. 171, 19e28. Hu, X., Zhang, J., Jiang, Y., Lei, Y., Lu, L., Zhou, J., Huang, H., Fang, D., Tao, G., 2014. Effect on metabolic enzymes and thyroid receptors induced by BDE-47 by activation the pregnane X receptor in HepG2, a human hepatoma cell line. Toxicol. Vitro 28, 1377e1385. Huang, S.C., Giordano, G., Costa, L.G., 2010. Comparative cytotoxicity and intracellular accumulation of five polybrominated diphenyl ether congeners in mouse cerebellar granule neurons. Toxicol. Sci. 114, 124e132. Ji, C., Wu, H., Wei, L., Zhao, J., Lu, H., Yu, J., 2013a. Proteomic and metabolomic analysis of earthworm Eisenia fetida exposed to different concentrations of 2, 20 , 4, 40 -tetrabromodiphenyl ether. J. proteomics 91, 405e416. Ji, C., Wu, H., Wei, L., Zhao, J., Yu, J., 2013b. Proteomic and metabolomic analysis reveal gender-specific responses of mussel Mytilus galloprovincialis to 2, 20 , 4, 40 -tetrabromodiphenyl ether (BDE 47). Aquat. Toxicol. 140, 449e457. Jones, L.J., Gray, M., Yue, S.T., Haugland, R.P., Singer, V.L., 2001. Sensitive determination of cell number using the CyQUANT® cell proliferation assay. J. Immunol. Methods 254, 85e98.

Kang, Y., Pan, W., Liang, S., Zeng, L., Zhang, Q., Luo, J., Guo, X., 2016. In vitro digestion and absorption of BDE-28, -47, -99 and -153 in indoor dust and its implication in risk assessment. Ecotoxicol. Environ. Saf. 134, 11e16. Part 1. Legler, J., 2008. New insights into the endocrine disrupting effects of brominated flame retardants. Chemosphere 73, 216e222. Llabjani, V., Trevisan, J., Jones, K.C., Shore, R.F., Martin, F.L., 2011. Derivation by infrared spectroscopy with multivariate analysis of bimodal contaminantinduced dose-response effects in MCF-7 cells. Environ. Sci. Technol. 45, 6129e6135. Pazin, M., Pereira, L.C., Dorta, D.J., 2015. Toxicity of brominated flame retardants, BDE-47 and BDE-99 stems from impaired mitochondrial bioenergetics. Toxicol. Mech. methods 25, 34e41. Pereira, L.C., Souza, A.O., Dorta, D.J., 2013. Polybrominated diphenyl ether congener (BDE-100) induces mitochondrial impairment. Basic & Clin. Pharmacol. Toxicol. 112, 418e424. Quent, V., Loessner, D., Friis, T., Reichert, J.C., Hutmacher, D.W., 2010. Discrepancies between metabolic activity and DNA content as tool to assess cell proliferation in cancer research. J. Cell. Mol. Med. 14, 1003e1013. Reif, D.M., Sypa, M., Lock, E.F., Wright, F.A., Wilson, A., Cathey, T., Judson, R.R., Rusyn, I., 2013. ToxPi GUI: an interactive visualization tool for transparent integration of data from diverse sources of evidence. Bioinformatics 29, 402e403. Renner, K., Amberger, A., Konwalinka, G., Kofler, R., Gnaiger, E., 2003. Changes of mitochondrial respiration, mitochondrial content and cell size after induction of apoptosis in leukemia cells. Biochimica Biophysica Acta (BBA)-Molecular Cell Res. 1642, 115e123. Reynolds, D.P., Lanevskij, K., Japertas, P., Didziapetris, R., Petrauskas, A., 2009. Ionization-specific analysis of human intestinal absorption. J. Pharm. Sci. 98, 4039e4054. Souza, A.O., Tasso, M.J., Oliveira, A., Pereira, L.C., Duarte, F.V., Oliveira, D.P., Palmeira, C.M., Dorta, D.J., 2016. Evaluation of polybrominated diphenyl ether toxicity on HepG2 cells-hexa-brominated congener (BDE-154) is less toxic than tetra-brominated congener (BDE-47). Basic & Clin. Pharmacol. Toxicol. 119, 485e497. Souza, A., Pereira, L., Oliveira, D., Dorta, D., 2013. BDE-99 congener induces cell death by apoptosis of human hepatoblastoma cell lineeHepG2. Toxicol. Vitro 27, 580e587. Stecca, L., Tait, S., Corrado, F., Esposito, M., Mantovani, A., La Rocca, C., 2016. Development of an in vitro test battery model based on liver and colon cancer cell lines to discriminate PCB mixtures by transcription factors gene expression analysis. Toxicol. Vitro 34, 204e211. U.S. Environmental Protection Agency, 2010. An Exposure Assessment of Polybrominated Diphenyl Ethers (PBDE). EPA/600/R-08/086F 2010. Ulukaya, E., Ozdikicioglu, F., Oral, A.Y., Demirci, M., 2008. The MTT assay yields a relatively lower result of growth inhibition than the ATP assay depending on the chemotherapeutic drugs tested. Toxicol. vitro 22, 232e239. Van den Berg, M., Birnbaum, L.S., Denison, M., De Vito, M., Farland, W., Feeley, M., Fiedler, H., Hakansson, H., Hanberg, A., Haws, L., Rose, M., Safe, S., Schrenk, D., Tohyama, C., Tritscher, A., Tuomisto, J., Tysklind, M., Walker, N., Peterson, R.E., 2006. The 2005 World Health Organization reevaluation of human and Mammalian toxic equivalency factors for dioxins and dioxin-like compounds. Toxicol. Sci. 93, 223e241. Wang, L., Zou, W., Zhong, Y., An, J., Zhang, X., Wu, M., Yu, Z., 2012. The hormesis effect of BDE-47 in HepG 2 cells and the potential molecular mechanism. Toxicol. Lett. 209, 193e201. Wei, R., Zhao, Y., Liu, P., Qin, Z., Yan, S., Li, Y., Qin, X., Xia, X., Xu, X., Yan, M., 2010. Determination of environmentally relevant exposure concentrations of polybrominated diphenyl ethers for in vitro toxicological studies. Toxicol. Vitro 24, 1078e1085. Weyermann, J., Lochmann, D., Zimmer, A., 2005. A practical note on the use of cytotoxicity assays. Int. J. Pharm. 288, 369e376.