Metabolism and estrogenicity of bisphenol A and its analogues

Full text


Metabolism and estrogenicity of bisphenol A and its analogues

A comparative analysis of experimental and

computational data on metabolism of bisphenols

Suzanne Bruks

Supervisors: Ioana Chelcea and Patrik Andersson

Bachelor Thesis, 15 ECTS

Life Science/Chemistry, 180ECTS Spring term 2020




Bisphenol A (BPA) is an endocrine disturbing chemical found in several everyday consumer products. The toxicity of BPA is due to its resemblance to the hormone estrogen. Both the parental compound and its metabolites show agonistic effects on estrogen receptors (ER) within the mammalian system. With the awareness of the toxic effects of BPA the usage of substitutional chemicals, such as bisphenol S and bisphenol F, are increasing. Bisphenols does not only share structural resemblance to one another, but also characteristics, and there are reasons to believe that the replacement of BPA to other bisphenols are to consider as regrettable substitution. Around 200 chemicals with structural resemblance to BPA are today found in industry. In this study a screening of published literature have gathered up to date data on estrogenicity and metabolism of bisphenols (n=11), showing that experimental values on a lot of BPA analogues are missing. The bisphenols that had the largest quantity of published data (BPA, BPS and BPF) were used to make an comparative analysis to in silico predictions on metabolism. Three prediction programs were used (CTS, BioTransformer and Meteor Nexus) and evaluated with the hypothesis that in silico software can be used to predict metabolism of bisphenols. The result of the literature survey show that the main metabolic pathways for detoxification is phase I hydroxylation in vitro, and phase II glucuronidation in vivo. Reactive species from oxidative phase I reactions are the main source for bioactivation of bisphenols. All of the in silico tools managed to predict the main metabolites, but none of the programs managed to predict the metabolites that showed greatest estrogenicity in vitro. The software that performed best for all of the compounds was concluded BioTransformer.


List of abbreviations

BPA Bisphenol A

BPS Bisphenol S

BPF Bisphenol F

EDC Endocrine disrupting chemical

ER Estrogen receptor

QSAR Quantitative structure–activity relationship

UGT UDP-glucuronosyltransferase

BPA-G BPA-glucuronide

SULT Sulfotransferases

BPA-S BPA-sulfate

CYP Cytochrome P450 enzyme

SMILES The Simplified Molecular Input Line Entry Specification

KemI Swedish chemicals agency

CTS Chemical transformation simulator

Parent-Glu-Sul Parent- glucuronide-sulfate

Parent-OH-Glu Parent- hydroxylation- glucuronide Parent-OH-Sul Parent- hydroxylation- sulfate

HAC Hydroxycumyl alcohol

IPP Isopropenyl phenyl

MBP 4-Methyl-2,4-bis(4-hydroxyphenyl)pent-1-ene

DHB Dihydroxybenzophenone


Table of contents

Abstract ... 1

Table of contents ... 3

1. Introduction ... 4

1.1 Aim of the diploma work ... 5

2. Experimental material and method ... 6

2.1 Literature survey ... 6

2.1.1 Applied search logic ... 6

2.1.2 Literature screening ... 6

2.2 In silico prediction software ... 7

2.2.1 Chemical transformation simulator (CTS)... 7

2.2.2 BioTransformer ... 7

2.2.3 Meteor Nexus ... 7

2.3.4 Comparative analysis ... 7

3. Results and Discussion ... 9

3.1 Literature survey of bisphenol metabolism... 9

3.1.1 Phase I metabolism ... 9

3.1.2 Phase II metabolism ... 10

3.1.3 Differences in systemic clearance ... 10

3.1.4 The role of time of exposure during development ... 11

3.2 Comparative analysis of experimental and computational data for BPA, BPS and BPF ... 12

3.2.1 Experimentally identified metabolites of Bisphenol A ... 12

3.2.2 In silico predictions of BPA ... 14

3.2.3 Experimentally identified metabolites of Bisphenol S ... 17

3.2.4 In silico predictions of BPS ... 18

3.2.5 Experimentally identified metabolites of Bisphenol F ... 21

3.2.6 In silico predictions of BPF ... 22

3.3 Performance of the prediction software ... 24

3.3.1 Evaluation of the pros and cons of the in silico tools ... 25

3.4 Estrogenicity of bisphenols ... 26

3.5 Estrogenicity of metabolites of BPA, BPS and BPF ... 28

3.6 Major findings of this report ... 29

4. Conclusions ... 30

5. Appendix ... 31

6. References ... 32



1. Introduction

Bisphenol A (BPA) is an endocrine disrupting chemical (EDC) that is believed to cause a number of adverse health effects, such as reproductive disabilities, diabetes, obesity, cancer and neuropsychiatric diseases 1–4. BPA is produced mainly as a monomer for polycarbon plastic, but also as an epoxy resin, and is found in consumer products such as aluminum cans, sports equipment, toys, water bottles, receipts, DVDs and pipe lines5. Bisphenol A was introduced to the market in the 1930s as a potential estrogen pharmaceutical due to its resemblance to estradiol6. The compounds affinity to cellular estrogen receptors (ER) is referred to as its estrogenicity, and the agonistic properties are causing endocrine disrupting effects. Due to the concern of human health the regulations regarding use of BPA are getting stricter in the EU5 and the production of substitutional chemicals, such as bisphenol F (BPF), are increasing7. The substitution is often done with other bisphenols. There are about 15

compounds that share the common name of “bisphenols” followed by a letter (abbreviated BPX), these compounds are referred to as members of the bisphenolic alphabet6. The substitutional chemicals share structural resemblance to BPA and there are reasons to address this as a regrettable substitution.

In 2017 the Swedish Chemicals Agency (KemI) identified over 200 compounds in industry that shared molecular resemblance to BPA, all with two phenol groups joined together by a bridge, not limited by being part of the bisphenolic alphabet of compounds6. Quantitative structure–activity relationship (QSAR) analysis indicated that 39 of the compounds had endocrine disrupting properties, suggesting that more than the members of the bisphenolic alphabet might be of concern when alternatives to BPA are investigated.

In this study data on metabolism and estrogenicity for eleven bisphenols were found in literature.

Seven of them were detected in human urine in Saudi Arabia 2017. In that study the levels of bisphenol S (BPS) was even higher than those of BPA8. As the increasing usage of the BPA analogues result in detection in human bodies, knowledge regarding similarities in toxicological properties becomes of great interest.

BPA is not persistent in the human body but rather go through metabolism9. Bisphenol A is

metabolized primarily in vivo by the metabolic phase II group of enzymes UDP-glucuronosyltransferases (UGTs) in the intestine and liver into BPA-glucuronide (BPA-G), a compound that is believed to be ER- inactive10, and the phase II metabolism is therefore referred to as a detoxification pathway11. The main metabolic route of BPA is illustrated in figure 1 below.

Figure 1. Illustration of the main metabolic route of BPA in a pregnancy model in vivo. The scheme only include the phase II pathway of UGT enzymes. The figure is created based on the biotransformation of BPA12

1. Oral exposure into digestive tract

2. UGTs in intestine transforms some of the BPA to BPA-G.

BPA is also excreted via faeces.

3. Unconjugated BPA and BPA- G travel to the liver via the portal vein. UGTs transforms BPA to BPA-G.

4. BPA-G enter the circulatory system.

5. Some BPA-G enter the kidney for urinary excretion

6. BPA-G reach the placenta and are transformed to active BPA by β-glucuronidase.

7. BPA-G re-enter the intestine via the bile duct

8. Excretion via faeces as BPA-G


5 A concern regarding BPA metabolism by UGTs is related to BPA-Gs ability to deconjugate back into the active parent compound. Deconjugation is carried out by β-glucuronidases, which are present not only in adults but also in the foetus and the placenta13. The phase II metabolism of BPA also include

sulfotransferases (SULTs), that conjugates the parental compound into BPA-sulfate (BPA-S), which in similarity to BPA-G does not show any ER-activity14. However the metabolism of xenobiotics also include the phase I metabolism taking place in the liver by the cytochrome P450 enzymes (CYPs). This oxidative metabolism is leading to a number of different metabolites where the most common is the hydroxylated form of BPA (BPA catechol). A schematic picture of the general metabolic routes of Bisphenol A is illustrated in figure 2 below. The image shows metabolites formed by phase I and phase II enzymes respectively.

Figure 2. Illustration of some metabolic pathways identified in BPA15–18, the active enzymes in the different pathways as well as the metabolites with general names are presented.

Some of the metabolic products are shown to possess the same properties as the parental compound of being estrogenic. Metabolites of phase I metabolism of BPA assess estrogenic properties stronger and weaker than the parental compound14,15,19. Questions regarding the substitution of BPA with

structurally similar compounds should therefore also take the compounds metabolism into consideration.

1.1 Aim of the diploma work

Since data on a lot of bisphenols are missing it is interesting to investigate the similarities and

differences in bisphenol metabolism, in order to see if there are some common pathways, for instance those that are bioactivating, i.e. that show greater estrogenicity from the metabolites than the parent compound. Experimental tests for toxic effects of xenobiotics often includes studies on live animals.

Since there are a lot of bisphenols with structural resemblance in vivo experiments on all of them would include the usage of many animal models. Therefore, it would be much interesting to see if the metabolism of bisphenols can be predicted by computational tools.

The overall aim of this bachelor thesis is to make an overview of the to date data on metabolism of bisphenols (n=11), and a comparative analysis of their estrogenicity. Also the aim include a more detailed comparative analysis of the experimentally found metabolites of a few bisphenols (n=3) to in silico predictions using several software on mammalian metabolism.

The objectives include an assessment on the bioactivation properties of bisphenols, the hypothesis is that some of the metabolites from phase I metabolism will show estrogenicity while the phase II

metabolites will not. Furthermore the objectives are to analyze the performance accuracy of different in


6 silico tools considering metabolic prediction for bisphenols, the second hypothesis is that in silico tools can be used for metabolic predictions of bisphenols.


2. Experimental material and method

The process included a literature survey to find experimental data on how different bisphenols are metabolised, as well as compiling information on the estrogenicity of parental compounds and metabolites. The search was based on an initial list of over 20 analogues to BPA (see appendix)

complied based on data availability as well as exposure risk to humans. The compounds on the list were picked based on risk ranking done by the bisphenol analysis done by KemI6, but also other data on BPA substitutes.

2.1 Literature survey

The first part of this study was a screening of the published literature on metabolism and endocrine activity of bisphenols. The search database used was Google Scholar. The search also included

enzymatic rate values for the metabolic reactions such as maximum velocity (Vmax), affinity constants (km) or disappearance of parent compound over time (mass/time unit). The purpose was to investigate the kinetics of the metabolism, especially to find data on clearance time for phase I metabolites. If the estrogenic molecules have a very short persistence in the system the toxicity would be considered lower, however the data found was not comprehensive and that part of the study was excluded.

2.1.1 Applied search logic

The studies was identified using the following search strings:

• Bisphenol analogues AND metabolism

• ((Bisphenol) AND metabolism) AND rate

• Bisphenol AND (metabolism OR metabolites) -bacteria -microorganism

• (((Bisphenol) AND metabolites) NOT Bisphenol A

• biotransformation AND bisphenol

• bioactivation OR estrogenicity AND bisphenol 2.1.2 Literature screening

Articles were picked after relevance based on their title, 79 published papers were screened further by abstract. 17 papers were identified not through the google search but as references from other studies.

In total a number of 96 articles were identified. Of these 49 was fully read, and in this study 32 were used for reference data on metabolism and/or estrogenicity of bisphenol A and its structural analogues.

The process is illustrated in figure 3.

Figure 3. Flow diagram over the process of finding metabolic and estrogenic data on bisphenols.


7 From the articles structures of the identified metabolites, as well as hypothesis regarding formation pathways and reactive enzymes were studied. Bisphenol A is the most studied of the eleven bisphenols, followed by BPS, BPF and BPAF. The amount of data on metabolism determined that the in depth survey and in silico comparison would concentrate on BPA, BPS and BPF. However all data collected on the documented metabolism of the BPA analogues was used to demonstrate similarities in their metabolism and their potency to the estrogen receptor α, based on measured values of EC50 (the concentration required to induce half of maximum response of, in this case, estrogenicity) 2.2 In silico prediction software

Three different in silico tools were used for the metabolic predictions. All of the software made predictions based on mammalian models.

2.2.1 Chemical transformation simulator (CTS)

CTS only provides predictions on phase I metabolism. It is a free online software provided by United States Environmental Protection Agency (EPA)20.

The Simplified Molecular Input Line Entry Specification (SMILES) code of each parental compound (BPA, BPS, BPF) was used to generate transformation products. The program provides a number of settings for metabolic model, and for this study the parameters was set to mammalian reaction system and human phase I metabolism. The number of generations was set to two. CTS provides prediction of metabolites as well as likelihood. The program presents the metabolites in a scheme with structures and SMILE code.

2.2.2 BioTransformer

BioTransformer is an in silico tool that is freely available via The Metabolomics Innovation Centre. It provides mammalian metabolic predictions and metabolites identification for small molecules in several modules21. In this study the modules for phase I (CYPs) and phase II metabolism was used.

As for CTS the SMILES of each parental compound was typed into the program and set on phase I as well as phase II metabolism. BioTransformer presents a number of metabolites and the result may be exported as JSON, CSV or SDF file, the latter was used for this study. In the SDF file the different metabolites is presented as InChI codes. The InChI codes was transformed to structures using an additional software, Chemical Identifier Resolver22. To generate several generations of metabolites (phase I + phase II or phase I + phase I) the SMILEs for each metabolite was run in the

BioTransformer program in the same manner as the parent molecule. The search was run for phase I, phase II, phase I + phase I, phase I + phase II and phase II + phase II, to simulate metabolism for two generations of metabolites in all combinations of reactions.

2.2.3 Meteor Nexus

Nexus is a tool where the predictions are based on expert knowledge on metabolism. It presents phase I and phase II metabolites based on three different methodologies, absolute/ relative reasoning, static scoring/occurrence ratio or site of metabolism scoring. The software is purchasable via Lhasa Limited23.

The default settings of Nexus are to show three generation of metabolites, and to stop prediction after the attachment of one generation phase II functional group and site of metabolism scoring. To identify as many metabolites as possible, but still get an amount of data that was manageable, the settings were customized to predict two generations, but not to be limited in phase II products (to get the phase II + phase II metabolites) and finally set to use absolute/relative reasoning and present metabolites that are predicted to be equivocal or more likely than that. The program presents a tree of metabolites from parent compound, with structures and an estimation on likelihood. Meteor Nexus predicts metabolites from phase I redox and non-redox reactions, and phase II products from e.g glucuronidation, sulfation, methylation, acetylation and conjugation with amino acids

2.3.4 Comparative analysis

The experimental data found in literature was compared to the computational predictions. Assessment on how well the programs did was based on how many metabolites the program predicted in relation to how many of those that were seen experimentally. Example: Meteor nexus predicted 12 metabolites for BPS, whereof seven were see in literature. 7/12 = 58% of the metabolites were a match.


8 The analysis was also based on how many of the experimental metabolites the programs were able to detect. For example: Literature findings on metabolites of BPS was eight. So the seven that matched with Nexus predictions made up 7/8= 87,5% of the expected metabolites.

A score was set based on the sum of the programs performances in both these percentage calculations, to evaluate their overall prediction capacity.



3. Results and Discussion

The literature survey found result on estrogenicity and metabolism for eleven (n=11) different

bisphenols (Figure 4). In resemblance to the in silico prediction study made by KemI6, the search was not limited to members of the bisphenolic alphabet, but by the general structure of two phenolic rings linked together by a bridge. As a result within the findings benzophenone-2 (BP-2) is included.

Figure 4. Parental compounds used in the comparative study on metabolism and estrogenicity of bisphenols.

3.1 Literature survey of bisphenol metabolism

Table 1 below picture a summary of the literature findings on metabolites produced and identified in vitro and in vivo. The general structure of the different metabolites can be seen in figure 2 in the introduction.

Table 1. Summary of metabolites found for in vivo and in vitro studies on different bisphenols. Studies used for table: BPA15–18,24–28, BPS11,29–31, BPF24,32–35, BPAF24, BPZ24, BP231, BPB18, BPC24,36, TMBPA36, TBBPA37,38, BPE36


Parent ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓

Parent- catechol ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓

Parent- quinone ✓ ✓ ✓ ✓

Dimerization ✓ ✓ ✓ ✓ ✓ ✓

Ipso-reaction ✓ ✓ ✓ ✓ ✓

Parent-Glu ✓ ✓ ✓ ✓ ✓ ✓

Parent-Sul ✓ ✓ ✓ ✓ ✓ ✓

Parent-Glu-Sul ✓ ✓ ✓

Parent-OH-Glu ✓ ✓

Parent-OH-Sul ✓ ✓

3.1.1 Phase I metabolism

CYP-mediated hydroxylation is a common reaction pattern for the different bisphenols. Most of the studies above are performed in vitro with parent- catechol as the major metabolite, as the result of phase I metabolism. As BP-2 already has hydroxyl groups in ortho position, hydroxylation is not

expected for this compound. Nakamura et al. (2011) found that some of the CYP-enzymes (CYP3A4 and CYP3A5) had the ability to cleave the BPA molecule via an ipso substitution forming the two

metabolites isopropenyl phenyl (IPP) and hydroxycumul alcohol (HAC). When the metabolites were tested for estrogenic properties it showed that IPP had similar ER activity as the parent compound, while the ER-binding of HAC was 100 times higher than BPA, indicating that this metabolic pathway was bioactivating. Occurrence of the reaction was about 20% of the main oxidative pathway of catechol formation15. HAC and IPP, were found as the result of cleavage of the parental compound, and both molecules are believed to origin from a carbocation intermediate15. A similar reactive intermediate is thought to cause the dimerization into 4-Methyl-2,4-bis(4-hydroxyphenyl)pent-1-ene (MBP), as it is the result of a dimer of IPP18. A schematic illustration of possible mechanisms are presented in figure 5


10 below. Note that this scheme is slightly different from the proposed mechanism of the experimental finding of MBP, where they suggest a radical of IPP as intermediate18.

Figure 5. Suggested mechanism for formation of IPP, HAC and MBP. Three estrogenic metabolites of BPA.

To investigate if the ipso reaction took place for other bisphenols, Schmidt et al. (2013) tested the CYP oxidative reactions of BPF, BPAF, BPZ and BPC in the search for similar metabolites. They found that the ipso-substitution reaction occurred for all of the investigated bisphenols, confirming that their phase I metabolism follow the same reaction patterns. Further they showed that the ipso-substitution pathway is a minor oxidative pathway and is not limited to any certain CYP-enzyme24. To investigate the potential estrogenic effect of the result of an ipso-substitution Fic et al. (2014) tested the 4-

(hydroxymethyl)phenol product of BPF and found no ER activity39. Presenting that the reaction is not bioactivating for all bisphenols. Skledar et al. (2016) repeated the experimental procedure to

investigate any possible reactive species formed via ipso-substitution of BPS. They could not verify any such reaction11. Showing how the structural characters of different bisphenols cause deviations in their metabolism. Other concerns from the phase I metabolism was the quinone formation of BPA, that Atkinson et al. (1995) found to be DNA-binding, and the metabolites role in the toxicity of BPA was elucidated27.

No reactive intermediates or oxidative species was detected in the four studies of BPS, however Kang et al. (2014) found bioactivation when they measures the activity through rat liver S9 fractions and found that the ER activity of BPS went up with metabolism. In that study they did not propose any structure of metabolites, but they suggested a dimer formation to be the cause of the elevated ER activity40. For BPF a unique metabolite named dihydroxybenzophenone (DHB) was detected and that metabolite showed estrogenicity34.

3.1.2 Phase II metabolism

In vivo the predominant species are glucuronidation or sulfation indicating that phase II metabolism is the major pathway in living mammals. Those bisphenols that does not show any phase II metabolites in table 1 were not identified in vivo. The physicochemical characteristics determine the excretion route for the different compounds. In vivo experiments of BPS show that BPS-glucuronide is the main metabolite and that it is mainly excreted via urine30,41. The main UGT enzyme responsible for BPS conversion was found to be UGT1A9, predominantly found in the liver, while for BPF it was UGT1A10, an enzyme most active in the intestine. This is resulting in excretion of BPF-G to be predominantly found in faeces whereas BPS-G is found in urine30. Consistent with that is the findings of Cabaton et al.

(2006) where they saw that the main BPF metabolite found in urine of rats was BPF-sulfate32. Skledar et al. (2016) also found in vitro evidence that glucuronidation of BPS was a much faster reaction than the CYP mediated oxidative reactions, explaining why the phase II metabolites are the most abundant species found in vivo. They also found that, in resemblance to the UGTs, the CYPs involved in

metabolism is different for different bisphenols11.

3.1.3 Differences in systemic clearance

Recently Gayrard et al. (2019) investigated the bioavailability of BPS compared to BPA in piglets, and found that BPS was quickly absorbed after oral ingestion. However, only 41% was glucuronidated during first pass metabolism in the liver, and 57% stayed in the circulation system as BPS. Whereas only 0.5% of the BPA was left as the parent compound. Indicating that any endocrine disrupting


11 properties of BPS could originate from the parental bisphenol not being quickly metabolised. The overall clearance of BPS was 3.5 times lower than BPA41. Similar findings on BPS persistence being longer within the circulation system compared to BPA was done by Liu et al. (2019), they saw that thirty times higher percentage of unmetabolized BPS was excreted in urine than BPA42. High plasma concentration of BPS was verified by Gingrich et al. (2019) in a pregnant sheep model. Further their findings also demonstrated that the half-life of BPS was shorter than BPA and BPF, in contrast to the previous study. However the half-life of BPS in the foetus was demonstrated to be longer than the other two bisphenols. BPF was shown to present the lowest concentration in the maternal system, and also the fastest clearance. The study also verified the ability of all the three compounds to cross the placental barrier43.

3.1.4 The role of time of exposure during development

The endocrine disrupting and toxic effects of bisphenols have been concluded to have largest effect during development. Corbel et al. (2015) investigated the foetus ability to detoxify BPA, and found that BPA is successfully transformed to BPA-G late in the pregnancy, at early stages the detoxification enzymes are less active. Furthermore the study showed that the BPA-glucuronide remained trapped within the placental compartment, and the presence of β-glucuronidases leading to re-exposure to active BPA13. Also the foetus liver is found to express significant levels of CYPs and low levels of UGTs44.Foetal excretion of bisphenols would result in assembly within the amniotic fluid and the possibility for oral recirculation, making the discard of the compounds difficult. The presence of BPS and BPF within the amniotic fluid was verified by Gingrich et al. (2019) and estimations were made that BPS might be persistent within the foetal compartment a month after exposure. In that study they also found that BPA was more permissive to the placental barrier than BPS and BPF, most likely due to physiochemical properties such as solubility, size and lipophilicity43.

The overall findings indicates that bisphenols are detoxified via phase II metabolism, and that

endocrine disrupting properties may originate from parental compounds not being cleared fast enough, or phase I oxidative metabolites. The experimental data show that phase I products are further

metabolized by phase II enzymes, or that the phase II metabolism occurs at a faster rate than phase I.

Concluding that endocrine disrupting properties from these metabolites are most toxic in systems with malfunctioning or low concentration phase II enzymes, such as a foetal model.


12 3.2 Comparative analysis of experimental and computational data for BPA, BPS and BPF The literature search for metabolites resulted in nine studies on BPA, four studies on BPS and five studies on BPF. These three were the bisphenols with the largest amount of experimental data to date.

Tables 2, 4 and 6 include the identified metabolites with names and structures. The systems that detected the different molecules are listed.

3.2.1 Experimentally identified metabolites of Bisphenol A

Table 2. Names and structures of metabolites of BPA found in vivo and vitro in nine different studies.

Abbreviations used for later comparison to in silico predictions.

Name 4-[2-(4-hydroxyphenyl)propan-2-yl]phenol Structure

CAS 080-05-7

SMILES Oc1ccc(C(C)(C)c2ccc(O)cc2)cc1

Usage Polycarbonate plastics used in food containers, water bottles, sports equipment, toys, DVDs.

Epoxy resins for coating the inside of metal cans and pipe lines. Dental sealants. Thermal papers.


Metabolite Structure Abb. System

Phase I Hydroxy-BPA (catechol) Predominant in vitro

A1 CD1 mice feces in vivo25

CD1 mice liver microsomal and S9 fractions in vitro26

Peroxidase activation system in vitro 27 Rat liver S9 in vitro18

Human liver microsomes in vitro15 Human liver microsomes in vitro24

BPA-ol A2 CD1 mice liver microsomal and S9

fractions in vitro26

BPA ortho- quinone

A3 Peroxidase activation system in vitro27 Rat liver S9 in vitro18

Human liver microsomes in vitro24 MBP


A4 Rat liver S9 in vitro18

Hydroxycumyl alcohol (HAC)

A5 Human liver microsomes in vitro15 Human liver microsomes in vitro24

Isopropenyl Phenyl (IPP)

A6 Rat liver S9 in vitro18

Human liver microsomes in vitro15

Phase I + II


13 BPA-ol-


A7 CD1 mice urine in vivo25

Hydroxy-BPA- glucuronide/


A8 CD1 mice digestive tract in vivo25

methoxylated BPA-



A9 CD1 mice digestive tract in vivo25

Hydroxy-BPA- glucuronide

A10 Human microsomes in vitro28

Phase II

BPA sulfate A11 MCF7 in vitro17

CD1 mice urine, feces in vivo25


glucuronide Predominant in vivo

A12 MCF7 in vitro17

CD1 mice urine, liver, digestive tract in vivo25

Rat in situ16

CD1 mice liver microsomal and S9 fractions in vitro26

BPA sulfate/


A13 Rat in situ16


14 3.2.2 In silico predictions of BPA

Chemical transformation simulator

CTS prediction of phase I metabolites was set to two generations, providing predictions on phase I metabolism and further oxidation of the two primary metabolites. Both generation 1 metabolites are also formed experimentally. Four of the seven metabolites are not seen in vivo or in vitro studies. All of the predictions for BPA was predicted as unlikely by CTS. Figure 6 demonstrates all the found

metabolites with structures. The identified metabolites that was also found experimentally are marked with abbreviations from table 2.

Figure 6. CTS predictions on two generations of phase I metabolism of BPA. Metabolites found in literature are marked with corresponding abbreviations.


The same generation 1 metabolites from phase I metabolism as CTS predicted are also seen in BioTransformer. As the program only presents one search at a time, the SMILES of the gen1 metabolites was run through the program again searching for phase II metabolites. The software predicted both glucuronidation and sulfation of both compounds, as well as methylation of the BPA- chatechol metabolite (see figure 6 generation 2 at the far right). This metabolite is not seen in the literature studies, however looking at the structure of A9 it is likely that an intermediate with such a methylation is present. Sulfation of the phase I metabolites was not seen in literature. Glucuronidation is consistent with experimental metabolites. The parental compound was also run through phase II metabolism and both the expected products of glucuronidation and sulfation was predicted. The product of phase II was run again in the program for second generation phase II products and found the diglucuronide, disulfate and the glucuronide/sulfate double conjugation of BPA. Those metabolites are not presented within figure 7 below, to make the schematic less crowded. Biotransformer predicted 17 unique metabolites. Those metabolites that formed several isoforms were counted as the same metabolite, because of the lack in the literature on precise structures of the metabolites. Of the 17 predicted metabolites seven was found in literature. BioTransformer failed in predicting A3 as a result of oxidation of A1, in contrary to CTS.


15 Figure 7. Summary of metabolic prediction of BPA by Biotransformer. * Predicted in two isoforms, one is shown here. **Predicted in three isoforms, one is shown here. The products of II + phase II metabolism are not shown in the schematic. Metabolites found in literature are marked with corresponding abbreviations.

Meteor Nexus

The default settings of meteor nexus found only two metabolites of BPA, BPA-G and BPA-catechol.

Advanced search with the parameters explained in the experimental section gave 14 metabolites of phase I and phase II metabolism, presented below in figure 8. Since only five of the 14 metabolites was found in literature, and Nexus failed to detect the BPA-ol (A2), that both BioTransformer and CTS found, the search was customized to only include phase I metabolism and to show all metabolites, even improbable ones, this resulted in 14 new metabolites. However, the BPA-ol, or any of the other phase I products from literature, was still not found.

Figure 8. Metabolic predictions on BPA done by Meteor Nexus. To compress the picture only metabolites from 1st generation is shown in the illustration. The metabolites with several isoforms are only presented once and marked with the number of different metabolites next to their name.


16 Overall the software managed to predict the main metabolites of BPA. As CTS does not include phase II metabolism, its prediction of BPA catechol is considered as finding all the main metabolites. None of the software predicted cleavage or dimerization. A summary of the programs findings are presented in table 3 below, as well as the percentage of correctly identified metabolites. Over 50% of the metabolites predicted was never seen in literature, this was seen for all programs.

Table 3. Summary of the metabolites of BPA found by the different software, and the percentage of correctly predicted metabolites as well as percentage of experimental metabolites found.

CTS BioT Nexus

BPA- catechol ✓ ✓ ✓

BPA-ol ✓ ✓

BPA-quinone ✓


BPA-ol-glucuronide ✓

Hydroxy-BPA-glucuronide/ sulfate Methoxylated BPA- glucuronide/ sulfate

Hydroxy-BPA-glucuronide ✓ ✓

BPA sulfate ✓ ✓

BPA glucuronide ✓ ✓

BPA sulfate/ glucuronide ✓ ✓

Number of predicted metabolites 7 17 14 Percentage of correct predictions 43% 41% 36%

Percentage of metabolites found 50% 54% 38%


17 3.2.3 Experimentally identified metabolites of Bisphenol S

Table 4. Names and structures of metabolites of BPS found in vivo and vitro in four different studies.

Abbreviations used for later comparison to in silico predictions.

Name 4-(4-hydroxyphenyl)sulfonylphenol Structure

CAS 080-09-1

SMILES S(=O)(=O)(c1ccc(O)cc1)c1ccc(O)cc1

Usage Polycarbonate plastics and epoxy resins. Substitutional chemical for BPA, Thermal papers, food packaging


Metabolite Structure Abb. System

Phase I Hydroxy-BPS (catechol) Predominant in vitro

S1 Human liver microsomes in vitro29 Human liver microsomes + Yeast in vitro11 Rat hepatocytes in vitro30

Dihydroxy- BPS

S2 Human liver microsomes in vitro29 Human liver microsomes + yeast in vitro11

Phase I + II Hydroxy-BPS- glucuronide

S3 Human liver microsomes in vitro29 Human liver microsomes + yeast in vitro11

Hydroxy-BPS- sulfate

S4 Human liver microsomes in vitro29

Phase II BPS- glucuronide Predominant in vivo

S5 Human liver microsomes in vitro29 Rat hepatocytes in vitro30

Zebra fish in vivo31

Human liver microsomes + Yeast in vitro11

BPS-sulfate S6 Human liver microsomes in vitro29

Rat hepatocytes in vitro30 Zebra fish in vivo31

BPS sulfate/


S7 Rat hepatocytes in vitro30


18 BPS


S8 Rat hepatocytes in vitro30

3.2.4 In silico predictions of BPS Chemical transformation simulator

The prediction provided from CTS gave two metabolic products that was found experimentally, the hydroxylated (S1) and dihydroxylated BPS (s2). In addition it found four more 2nd generation products that were not seen in literature. In consistence with the experimental products the only phase I

metabolite of 1st generation is the BPS-catechol. CTS predicted S1 as likely, but did not do so for any of the 2nd generation products, they were all predicted as unlikely. Summary of the computational

findings is shown in figure 9.

Figure 9. Phase I metabolism of BPS in silico predictions by CTS. Metabolites found in literature are marked with corresponding abbreviations.


Both the S1 and S2 metabolites from CYP mediated phase I metabolism was predicted correctly by the software, in resemblance to CTS. However it made no predictions of sulfation from phase II

metabolism, and failed to detect several of the in vivo and in vitro metabolites due to this. The program was run for two generations on both phase I and phase II, and the diglucuronide of BPS was found from phase II + phase II metabolism. All metabolites are presented in figure 10 below.

BioTransformer only managed to detect six metabolites of BPS, five of which was seen in literature and hence missed three of the experimentally found products. The only phase II products was from UGTs, no sulfation products were predicted.


19 Figure 10. Metabolites of BPS predicted in silico by Biotransformer. * predicted in three isoforms, only one presented here. Metabolites found in literature are marked with corresponding abbreviations.

Meteor Nexus

Nexus did well in predicting the metabolites of BPS. It found a total of 12 metabolites, seven of which were seen in literature. 42% of the metabolites was not found experimentally, see table 5, however it managed to detect almost all of the metabolites in literature, only missing the dihydroxy product. Both BioTransformer and CTS managed to predict that metabolite. Findings are presented in Figure 11.

Figure 11. Metabolic prediction of BPS by meteor nexus. To compress the picture only metabolites from 1st generation is shown in the illustration. The metabolites with several isoforms are only presented once and marked with the number of different metabolites next to their name. Metabolites found in literature are marked with corresponding abbreviations.


20 Table 5 below summarise the findings of the metabolic software on BPS. BioTransformer did not

predict as many novel metabolites as the others. Nexus found almost all of the metabolites seen in literature. CTS found all of the products from phase I metabolism, but also predicted four additional metabolites.

Table 5. Summary of the metabolites of BPS found by the different software, and the percentage of correctly predicted metabolites as well as percentage of experimental metabolites found.

CTS BioT Nexus

BPS- catechol ✓ ✓ ✓

Dihydroxy BPS ✓ ✓

Hydroxy-BPS-glucuronide ✓ ✓

Hydroxy-BPS-sulfate ✓

BPS-glucuronide ✓ ✓

BPS-sulfate ✓

BPS- glucuronide/ sulfate ✓

BPS- diglucuronide ✓ ✓

Number of predicted metabolites 6 6 12 Percentage of correct predictions 33% 84% 58%

Percentage of metabolites found 100% 63% 88%


21 3.2.5 Experimentally identified metabolites of Bisphenol F

Table 6. Names and structures of metabolites of BPF found in vivo and vitro in five different studies.

Abbreviations used for later comparison to in silico predictions.

Name 4-[(4-hydroxyphenyl)methyl]phenol Structure

CAS 620-92-8

SMILES Oc1ccc(Cc2ccc(O)cc2)cc1

Usage Polycarbonate plastics and epoxy resins. Substitutional chemical for BPA, linings, lacquers, varnishes, water pipes, dental sealers.


Metabolite Structure Abb System

Phase I Ortho-hydroxy- BPF (catechol) Predominant in vitro

F1.1 Human liver microsomes + NADPH and glutathione in vitro24

Human and rat- S9 fractions and microsomes in vitro33

HepG2 in vitro34 Meta- hydroxy-

BPF (catechol)

F1.2 Human and rat- S9 fractions and microsomes in vitro33

HepG2 in vitro34

Dihydroxy BPF F2 Human and rat- S9 fractions and

microsomes In vitro33

BPF dimer F3 Human and rat- S9 fractions and

microsomes in vitro33

Dihydroxybenzo phenone (DHB)

F4 Human cells HepG2 in vitro34


(hydroxymethyl) phenol

F5 Human liver microsomes + NADPH and glutathione

In vitro24 Phase I + II

Hydroxy-BPF- sulfate

F6 Human and rat- S9 fractions and microsomes

In vitro33 Phase II

BPF-glucuronide F7 Rat in vivo32

HepG2 in vitro35

Human and rat- S9 fractions and microsomes in vitro33


22 BPF-sulfate

Predominant in vivo

F8 Rat in vivo32 HepG2 in vitro35

Human and rat- S9 fractions and microsomes in vitro33

BPF-disulfate F9 Human and rat- S9 fractions and

microsomes In vitro33

3.2.6 In silico predictions of BPF Chemical transformation simulator

As presented in figure 12 below CTS only found one product for the 1st generation of phase I

metabolism. This product was not seen within the experimental metabolites. However, the oxidized product of the hydroxylated metabolite, DHB (F4), was found to be a major metabolite in rat urine34, indicating that the metabolite indeed is formed in vivo as well. The other product from 2nd generation phase II is not seen in literature. CTS was only able to predict one of the phase I metabolites of BPF that is found experimentally. The metabolite of generation 1 was predicted as likely, and the F4 product from generation 2 was predicted as probable.

Figure 12. Illustration of phase I metabolism of BPF in two generations, predicted by CTS. Metabolites found in literature are marked with corresponding abbreviations.


The software succeeded better than CTS in the prediction, finding the metabolite with the hydroxylated phenol (catechol). The program was set to search for two generations of metabolites (phase I, phase I + phase I, phase I + phase II, phase II + phase II). The result of the first three searches can be seen in figure 12 below, phase II + phase II metabolites are not presented in order to make the overview easier.

The metabolites missing from figure 13 were diglucuronidation, the combined metabolite of glucuronidation and sulfation, as well as the disulfate metabolite seen in literature (F9).


23 Figure 13. Metabolites predicted by biotransformer. * predicted in two isoforms, only one presented here. **

predicted in three isoforms, only one presented here. Metabolites found in literature are marked with corresponding abbreviations.

Meteor Nexus

The default settings gave three metabolites (BPF-G, BPF-OH, BPF-OH-G) With the advanced settings explained in the method section Nexus found a large number of metabolites (52), this is far more than for the other studied bisphenols. Still, it scored less than BioTransformer in number of correct

predictions, see table 7. The large number of metabolites made illustration difficult and the generation 2 metabolites are therefore not presented in figure 14.

Figure 14. Metabolic predictions on BPF done by Meteor Nexus. To compress the picture only metabolites from 1st generation is shown in the illustration. Metabolites found in literature are marked with corresponding abbreviations.

Overall the predictions for BPF were uneven. CTS was the program that predicted the lowest amount of novel metabolites, however it also managed to only detect one out of five of the literature metabolites from phase I metabolism, see table 7. Nexus found as many as 52 metabolites, where only six was experimentally verified. Giving an accuracy of only 12% for BPF, the far lowest score of all predictions.

It did manage to find 66% of the metabolites, but BioTransfomer found 78% and still had a higher amount of correct predictions. The predictions are summarized in table 7 below.


24 Table 7. Summary of the metabolites of BPF found by the different software, and the percentage of correctly predicted metabolites as well as percentage of experimental metabolites found.

CTS BioT Nexus

BPF- catechol ✓ ✓

Dihydroxy BPF ✓

BPF dimer

DHB ✓ ✓ ✓


Hydroxy-BPF-sulfate ✓ ✓

BPF-glucuronide ✓ ✓

BPF-sulfate ✓ ✓

BPF-disulfate ✓ ✓

Number of predicted metabolites 3 16 52 Percentage of correct predictions 33% 44% 12%

Percentage of metabolites found 20% 78% 66%

3.3 Performance of the prediction software

Looking at tables 3, 5 and 7 the most unitary predictions was seen for BPA, however over half of the predictions from all programs were never seen in literature. None of the programs managed to find even 2/3 of the expected metabolites for BPA. The amount of specific metabolites from phase I metabolism seen for BPA can explain this. BPS was the compound where the programs overall managed to predict the most metabolites correctly. Comparing this to the literature and the low amount of phase I products it becomes clear in comparison to BPA that it is the prediction of phase I products that are the most difficult for the programs. Cleavage and dimerization are seen in both BPA and BPF and none of the software were able to predict such reactions. BPF was the compound that the programs were most uneven in their predictions, however BioTransformer were most successful, finding 78% of the metabolites and had a score of accurate predictions of 44%. To evaluate the performance from tables 3, 5 and 7 both of the parameters of percentage of correct/expected predictions and the percentage of literature metabolites found needed to be addressed. It was interesting to see if any program were able to find the expected metabolites in combination of not predicting a high amount of metabolites not seen in literature. To evaluate this the percentages were summed together and converted into a score, see figure 15 below.

Figure 15. Diagram of the score sum of percentage of metabolites correctly predicted by the in silico tools on metabolism and the percentage of predicted metabolites that was expected based on experimental findings.



53 95






0 20 40 60 80 100 120 140 160


sum of percentage score

In silico prediction performance

CTS Biotransformer Nexus


25 Data from table 3, 5 and 7. For the programs that predicted isomers the isomers have been counted as the same compound due to the lack of precise positioning of the functional groups within the literature findings.

The programs found the most comparable result for BPS, and BPF was the compound that showed the greatest diversity. BPS getting the highest overall score is consistent with the theory that the most difficult predictions comes from phase I metabolites and the fact that no experimentally found

oxidative species other than hydroxylated was presented in table 4 for BPS. For all the compounds the program that got the highest prediction score was BioTransfomer.

3.3.1 Evaluation of the pros and cons of the in silico tools

CTS is beneficial in being easy to use and understand as consumer, it does not require learning time and it presents the findings in a pedagogical way. All the results are presented at once with a nice presentation and flow chart. It can give predictions up to four generations and each metabolite can be magnified and details on the compound such as SMILES and names are presented. The program presents likelihood. It may also be used for other systems than mammalian. The disadvantage of the program is the limitation to phase I metabolism, however in the case of estrogenicity of bisphenols the ER-active metabolites are all of phase I origin. Still it does not provide a comprehensive picture of metabolism of bisphenols, and did not manage to find all of the phase I products seen in literature.

BioTransformer did better on the predictions of metabolites. The software is easy to understand and provides data in three different file-types. The data includes a lot of information, such as name(s), InChI code, InChI key, what sort of reaction that has occurred and which enzymes are responsible for the reaction. The program is able to predict both phase I and phase II metabolites giving it an

advantage over CTS. The number of possible generations of metabolism have no limit as the produced metabolites can be used for a further run in the program. BioTransformer did not manage to find all of the metabolites, but it made good predictions. However the program has limitations as the structures of the metabolites are not presented and additional software has to be used to illustrate the molecules.

The benefits of Meteor Nexus is its large capacity, and the ability to do advanced and specific searches.

It provides a nice schematic picture (tree) of metabolism from the parent compound. For each metabolite it presents which metabolic pathway that has been used (glucuronidation, methylation, sulfation etc). The advanced settings can be set to stop predictions of phase II metabolism after one such reaction to limit the amount of possible metabolites. However the program does predict a large number of metabolites and navigation becomes complex. The advanced settings was set to a general search for this study due to the massive number of metabolites that was presented. The program might be the most useful if more time is spent on understanding the search logics however metabolites might still be missed and the program did not manage to predict all metabolites even though it gave a lot of suggestions. The amount of data produced becomes overbearing and in a sense BioTransformer has an advantage of presenting one metabolite at a time.

All of the programs predicted several metabolites that was not found in literature. It is important to address that this does not mean that these metabolites is not formed in vivo or in vitro, nearly that they have not been identified. Some of the studies have performed targeted analysis, and found only those metabolite that they were looking for. Other possibilities of metabolites not being found experimentally could be if they are short lived or work as intermediates. Furthermore, metabolism studies are done mainly on liver cell-lines in vitro or by measuring urine or faeces in vivo. This may be limiting as metabolism occurs in the rest of the body as well and other metabolic products may be possible. The systems used for the experimental detection of metabolites are also different and include both rat, human and zebra fish. The cell lines of these different species differ from each other and can give different results. The performance of the programs can therefore not be judged solely based on scoring such as in this study, but it still gives an idea to weather some sort of reactions are missed by the programs.


26 3.4 Estrogenicity of bisphenols

Ten of the eleven compounds detected in the literature survey were found to have measured EC50 values on estrogen receptor activity i.e estrogenicity, values for BP-2 were missing. All of the bisphenols with published experimental values were tested for estrogen receptor activity relative to BPA. The resulting values are presented in table 8. Compounds with several values are presented more than once. The bisphenols that showed a stronger binding to ER than BPA are displayed in red within the table. Dependent on which system was used during the study the results vary, and the EC5o values cannot be compared to one another for compounds with several values. Different cells will have different sensitivity to the compound and the in vitro experiments have cell lines from human, yeast, zebra fish and rat. However, the pattern of estrogenicity of the analogue in comparison to BPA is

consistent for compounds with more than one value, with minor deviations for BPB and BPZ. A relative potency with a value of 1.0 would indicate the same activation potency for the estrogen receptor as BPA, higher value than 1.0 means stronger potency. Within the table are information on the risk of consumer contact that the bisphenol has according to the survey of bisphenols published by KemI 20176.

Table 8. Summary of EC50 values for the studied bisphenols in comparison to EC50 values of BPA. Values in red mark those experiments where the analogue had a higher potency than BPA.

EC50 (µM) ERα BPA reference Relative potency

In vitro system Consumer contact6

BPAF 0.05 0.63 12,60 MCF-719 low

1.4 30.1 21,50 Two hybrid yeast45

0.39 3.6 9,23 Yeast39

0.13 1.14 8,77 ERa-Hela990346

BPB 0.07 0.63 9,00 MCF-719 no data

0.12 0.08 0,66 BG1Luc4C247

9mg/l 24mg/l 2,66 Yeast48

BPZ 0.21 0.63 3,00 MCF-719 high

1.16 1.14 0,98 ERa-Hela990346

BPC 0.42 0.63 1,50 MCF-719 no data

11.8 30.1 2,55 Two hybrid yeast45

1.01 1.14 1,13 ERa-Hela990346

TMBPA 0.73 0.63 0,86 MCF-719 no data

BPE 0.91 0.63 0,69 MCF-719 no data

69.9 30.1 0,43 Two hybrid yeast45

0.47 0.08 0,17 BG1Luc4C247

36mg/l 24mg/l 0,66 Yeast48

BPF 1.0 0.63 0,63 MCF-719 high

4.67 3.6 0,77 Yeast39

0.82 0.08 0,96 BG1Luc4C247

80mg/l 24mg/l 0,30 Yeast48

5.8 0.8 0,14 ZELH-zfERs49

4.39 1.14 0,26 ERa-Hela990346

BPS 1.1 0.63 0,57 MCF-719 high

1.17 0.08 0,07 BG1Luc4C247

1.75 1.09 0,62 MCF-750

76mg/l 24mg/l 0,32 Yeast48

4.06 0.8 0,20 ZELH-zfERs49

4.24 2.32 0,55 MELN51

4.93 0.395 0,08 BG1Luc4E251

3.9 1.14 0,29 ERa-Hela990346

TBBPA 19 0.63 0,03 MCF-719 medium


27 The relative potency numbers was summed together and the mean value was used to form the

illustration in figure 16 below. Even though the experimental systems in these studies were different the pattern of estrogenicity in relation to BPA is consistent. It can be compared with the study that included most of the bisphenols, done by Kitamura et al. (2005)19, and the pattern of relative potency is the same.

Figure 16. Diagram of the mean BPA to analogue fraction. Note that the figures represent different in vitro systems. Compounds in red represent those that suggest a higher ER-activity than BPA, and the ones in blue show experimental values lower than those of BPA.

The diagram suggests that the estrogenic potency of BPAF is roughly 13 times higher than is it for BPA, however the risk of consumer contact is low for BPAF, as seen in table 8. BPZ is a compound with high risk for consumer contact and has a higher potency than BPA, suggesting that solely based on

estrogenicity BPZ would be a poor substitutional chemical. Potency of BPF and BPS are roughly half of that of BPA. However it is important to address that any potential endocrine disrupting effect also has to do with the time of exposure to the chemical41. Another interesting angle was provided by Skledar et al. (2020) that recently did an investigation of the cocktail effect on ER activity when different

bisphenols (BPA,BPF,BPAF,BPS,BPC and BPZ) were mixed, and found that toxic effects were reached when levels considered safe were combined. An equal mixture of all the bisphenols gave EC50 values of 0.7µM, and showed higher estrogenicity than BPA alone (1.14µM)46. Such patterns are important to keep in consideration, especially since there is a mixture of bisphenols verified in humans8.

0 2 4 6 8 10 12 14


Relative potency

Estrogenicity in relation to BPA


28 3.5 Estrogenicity of metabolites of BPA, BPS and BPF

The estrogenicity of the parent compounds have been both predicted in silico6 and verified through experimental studies. However to fully evaluate a toxicity it is of interest to see what happens to the estrogenicity after metabolism. The hypothesis is that glucuronidation and sulfation works as detoxification pathways, but as presented in this study there are other biological transformations of bisphenols. Therefore the literature was screened for estrogenic studies on metabolites of bisphenols to asses any possible bioactivation, the findings focused on the compounds where the metabolism was studied in more detail, BPA, BPS and BPF. The result is presented in table 9 below.

Table 9. Values on estrogenicity of metabolites in relation to the parental molecule from the three studied compounds BPA, BPS and BPF. Values where the metabolite is more active than the parental compound are marked in red

EC50 (µM) ERα Parent compound EC50 (µM)

In vitro system

BPA-catechol 1.8 0.63 MCF-719

weak YES-assay18

BPA-quinone weak YES-assay18

BPA-COOH 1.1 0.63 MCF-719

BPA-OL 11 0.63 MCF-719

IPP 36 0.63 MCF-719

Similar to BPA MCF-715

HAC 102 times BPA MCF-715

MBP 0.014 3.6 Yeast39

0.0083 0.14 Yeast two-hybrid18

0.0011 0.52 MCF-718

BPA-Glu Not active 0.71 MCF-710

BPA-Sul Not active MCF-714

BPS-catechol 670 84 Yeast11

BPS-Glu Not active 84 Yeast11

Unknown More active than BPS MVLN40

BPF-catechol Not active 2.39 HepG234

4-(hydroxymethyl) phenol

Not active 4.67 Yeast39

DHB 4 2.39 HepG234

The published data on the estrogenicity of metabolites does not give a clear pattern on bioactivation of bisphenols. There are not enough studies and data to make any such determinations. However it becomes clear that there are metabolic pathways that produce estrogenic metabolites that are more active than the parent compound. In this report no metabolites of BPF has been found to be more active than BPF, but in studies on BPA and BPS there are signs of bioactivation.


29 3.6 Major findings of this report

Points on the major findings:

• The most common phase I metabolite in vitro for bisphenols is the parent-catechol. For all three of the studied compounds this is the predominant phase I product.

• Phase I metabolites are not as commonly seen in vivo as in vitro, suggesting that the metabolism of parent compound by phase II enzymes are faster than phase I enzymes,

alternatively that the phase I products are brief and quickly metabolised by phase II enzymes.

Studies on the kinetics of phase I products would be of interest to further examine this question.

• The predominant phase II enzyme depend on the physicochemical properties of the bisphenol, and where it gets metabolised. Metabolism by UGTs in the intestine could lead to rapid

excretion via faeces, as for BPF, leading to larger abundance of sulfonated metabolites in the circulatory system.

• The in silico programs evaluated in this report are capable of performing predictions of the metabolism of bisphenols. All the major metabolites was found by all programs, with the exception of CTS and BPF-catechol.

• The in silico programs did not manage to find all of the metabolites. Especially difficult was it to find the phase I metabolites of BPA and BPF. None of the programs managed with this task.

• The most advanced program, Meteor Nexus, did not perform best on the prediction. The free online software BioTransformer was the program that did the most accurate predictions.

• The estrogenicity of bisphenols are consistent within the studies included in this report. BPAF is the compound of biggest concern based solely on ER-activity.

• Taking consumer contact into consideration BPZ shows twice the estrogenicity as BPA, and could be considered as a concerning compound.

• BPS and BPF does not induce the same estrogenicity as BPA, experimental data suggest a mean value of about 50% of the potency of BPA, however the systemic clearance of BPS suggests that ER-values are not enough to make a risk assessment of toxicity and endocrine disrupting properties.

• The endocrine disrupting properties of a compound are the result of a combination of different factors, such as the time the parent compound spends in the system, compound specific

metabolites and their estrogenicity and half time within the system, endocrine disrupting mechanism other than ER-activity, ability of the organism to clear the compound/metabolites from the system, the cocktail effect of several compounds and the time of exposure (foetus during development).

• Bioactivation does not follow a specific found pattern for bisphenols. The ipso-substitution pathway that increased estrogenicity for BPA does not do so for BPF. For BPS no such reaction is even seen experimentally.

• Most likely the estrogenicity seen in catechol metabolites is weaker than the parent compound due to increased steric hindrance.

• No metabolic activity is seen from phase II metabolites. Suggesting that glucuronidation and sulfatation infer detoxification pathways.





Relaterade ämnen :