Evaluating the Endocrine‐Disrupting and Oxidative Stress Potential of a 50‐Component Human‐Relevant Complex Chemical Mixture Using In Vitro Tests

ABSTRACT

Humans are chronically exposed to mixtures of environmental contaminants. Exposure to endocrine‐disrupting chemicals (EDCs) contributes to increased health impairment observed globally. This study aimed to evaluate the endocrine‐disruptive and oxidative stress potential of a human‐relevant, complex chemical mixture in vitro. By testing chemical class subgroup mixtures, the identity of toxicological drivers and mixture additivity could be investigated. A 50‐component mixture was compiled based on Swedish human blood concentrations (xHBC), consisting of six subgroup mixtures: polychlorinated biphenyls (PCBs) and 2,3,7,8‐tetrachlorodibenzo‐p‐dioxin (PCB mixture), brominated flame retardants (BFR mixture), per‐ and polyfluoroalkyl substances (PFAS mixture), pesticide mixture, synthetic phenolic contaminants (phenol mixture), and phthalate mixture. These were tested in four chemically activated luciferase gene expression (CALUX) assays: dioxin responsive (DR‐), estrogen receptor α (ERα‐), androgen receptor. (AR‐), and nuclear factor erythroid 2‐related factor 2 (Nrf2)‐CALUX, along with an adipocyte cell assay. The total mixture caused significant agonistic activity in DR‐ and ER‐, and antagonistic activity in AR‐CALUX at 0.1–15 xHBC, depending on the assay. Mixture additivity was assessed in ERα‐, DR‐, and anti‐AR‐CALUX using subgroup mixtures and the concentration addition (CA) model. The total mixture followed the CA model in ERα‐, anti‐AR‐ and DR‐CALUX. The toxicological drivers of these activities were mainly the PCB and phenol mixture. A significant increase in differentiated adipocytes was observed at 100 xHBC. These results raise concerns regarding potential health effects on the endocrine system. The additive effects at human‐relevant concentrations observed in this study motivate considering mixtures in regulatory contexts to protect the well‐being of future generations.

Short abstract

Humans are exposed to complex mixtures of environmental contaminants that could cause endocrine‐disruptive effects. A 50‐component mixture was compiled based on Swedish human blood concentrations (xHBC), comprising environmental contaminants from various compound classes. The endocrine‐disruptive potential of the 50‐component mixture and the chemical subgroup mixtures was tested using different in vitro assays. Significant activity at 0.1–15 xHBC raises concern about potential health effects.

Abbreviations

2,4‐DBP

2,4‐di‐tert‐butylphenol

AhR

aryl hydrocarbon receptor

AO246

2,4,6‐tri‐tert‐butylphenol

AR

androgen receptor

BBzP

butyl benzyl phthalate

BDE47

2,2',4,4'–tetrabromodiphenyl ether

BFR

brominated flame retardant

BHT

2,6‐di‐tert‐butyl‐4‐methylphenol

BPA

bisphenol A

CA

concentration addition

CALUX

chemically activated luciferase gene expression

DBP

dibutyl phthalate

DCC‐FBS

charcoal‐stripped fetal bovine serum

DEHP

di‐2‐ethylhexyl phthalate

DHT

5α‐androstan‐17β‐ol‐3‐one/dihydrotestosterone

DMEM

Dulbecco's Modified Eagle Medium

DMEM/F‐12

DMEM nutrient mixture F12

DMSO

dimethyl sulfoxide

DR

dioxin responsive

E2

17β‐estradiol

EC

effective concentration

EC50

half maximal effective concentration

EDC

endocrine‐disrupting chemical

ERα

estrogen receptor α

FBS

fetal bovine serum

H4IIE‐luc

rat hepatoma cell line

hMSC

human mesenchymal stem cells

IC

inhibitory concentration

IF

induction factor

LOQ

limit of quantification

MEMα

α‐Minimum Essential Medium

Nrf2

nuclear factor erythroid 2‐related factor 2

PBDE

polybrominated diphenyl ether

PBS

phosphate‐buffered saline, pH 7.4

PCB

polychlorinated biphenyl

PFAS

per‐ and polyfluoroalkyl substances

POP

persistent organic pollutant

PPARγ

peroxisome proliferator‐activated receptor γ

REP _mix

relative potency for mixture

ROSI

rosiglitazone

RXR

9‐cis retinoic acid receptor

SEDB

Swedish exposure database

SPA

synthetic phenolic antioxidant

TCDD

2,3,7,8‐tetrachlorodibenzo‐p‐dioxin

U2OS

Human Bone Osteosarcoma Epithelial Cells

xHBC

times human blood concentration

1. Background

Over the last century, global chemical production and use have increased significantly. As a result, humans are chronically exposed to mixtures of anthropogenic organic contaminants from various indoor and outdoor sources (Drakvik et al. 2020; Rappaport et al. 2014; World Health Organization et al. 2013). Globally, 350,000 chemicals are registered for production and use (Wang et al. 2020). Since 2004, the production and use of persistent organic pollutants (POPs) have been banned by implementing the Stockholm Convention, and new POPs are continuously added (Secretariat of the Stockholm Convention 2023). Due to their persistence and bioaccumulation, POPs are still detected in humans, even as their levels decrease (Engelhardt et al. 2022; Pineda et al. 2023, 2024). Other groups of chemicals found in humans are plastic additives and chemicals in personal care products (Gyllenhammar et al. 2017; Pineda et al. 2024).

Chemical exposure is a risk factor for human health impairments (Karlsson et al. 2020). Some environmental contaminants interact with the systems of the human body and can cause acute or subtle effects. Due to the complexity of the endocrine system, endocrine‐disrupting chemicals (EDCs) affect multiple endpoints. Exposure to EDCs has been linked to effects related to the female and male reproductive systems, thyroid disorders, neurodevelopment, hormone‐related cancers, adrenal disorders, metabolic disorders, immune system disorders, and an increasing number of deaths (Berntsen et al. 2017; Caporale et al. 2022; Hamers et al. 2020; Strand et al. 2024). Furthermore, mixtures of EDCs have been linked to a decreased IQ in children, raising concerns for future generations (Tanner et al. 2020).

In 2022, one in eight people worldwide was obese (World Health Organization 2024). Food intake, genetic susceptibility, and lifestyle cannot alone account for the obesity epidemic (Gupta et al. 2020). Exposure to EDCs has been identified as a potential risk factor for obesity and metabolic diseases later in life (Biemann et al. 2021; Boekelheide et al. 2012; Gupta et al. 2020; Neel and Sargis 2011).

Exposure to POPs, such as polychlorinated biphenyls (PCBs), polybrominated diphenyl ethers (PBDEs), organochlorine pesticides, and per‐ and polyfluoroalkyl substances (PFAS) have been linked to neurotoxicity, immunotoxicity, reproductive and developmental toxicity, carcinogenicity, oxidative stress, and endocrine disruption (Fenton et al. 2021; Jayaraj et al. 2016; Montano et al. 2022; Wu et al. 2020). Plastic‐related chemicals, such as bisphenols, phthalates, and synthetic phenolic antioxidants (SPAs), have also been linked to endocrine‐disrupting effects (Kitamura et al. 2005; Wang and Qian 2021; Xu et al. 2021).

Historically, the risks associated with exposure to environmental contaminants have been assessed individually, but during the last decades, chemical mixtures have been gaining more focus. Since environmental contaminants are present in complex combinations in the environment, using toxicity data based on artificial mixtures is an important step toward a more comprehensive risk assessment (Drakvik et al. 2020). A recent review summarized mixture studies reported in the literature between 2007 and 2017. More than 1000 mixture studies were included, and none had investigated human‐relevant chemical mixtures containing more than 30 compounds (Martin et al. 2021). However, a few in vitro studies have focused on human‐relevant mixtures with more than 20 chemicals (Berntsen et al. 2017; Caporale et al. 2022; Hamers et al. 2020; Strand et al. 2024). Combination effects, such as additive effects, are considered by testing mixtures of compounds. The concentration addition (CA) model assumes that all chemicals, including those present at concentrations below detection limits, are contributing to the overall toxic effect. The Funnel hypothesis states that the likelihood of a mixture exhibiting additive toxicity increases with the number of chemicals in the mixture (Warne and Hawker 1995). Concentration additivity is traditionally tested by measuring individual compounds in a bioassay followed by testing mixtures composed of the compounds that exhibit measurable activity. The predicted mixture effect is calculated based on the concentration‐response curves of the individual compounds. Thereafter, the prediction is compared to the observed effect of the mixture (Faust et al. 2001). When using the CA model, it is assumed that the compounds in the mixture all exert their toxic activity through the same mode of action (Altenburger et al. 2000).

This study aimed to evaluate potentially endocrine‐related effects resulting from the exposure to a human‐relevant complex chemical mixture by using five in vitro assays, i.e., four CALUX (chemically activated luciferase gene expression) and one adipocyte cell assay. The complex chemical mixture was prepared in concentrations representative of those found in the Swedish general population. Along with the complex mixture, six subgroup mixtures were tested individually to identify the toxicologically driving chemical group(s) within the overall mixture. In addition, the aim was to confirm concentration additivity by predicting mixture activities based on subgroup combinations using the CA model.

2. Method

2.1. Chemicals and Reagents

All chemicals used in the mixtures (more info in Table S1), dimethyl sulfoxide (DMSO anhydrous ≥ 99.9%) and toluene (anhydrous, 99.8%), were purchased from Sigma–Aldrich. The H4IIE‐luc (rat hepatoma cells) and U2OS (Human Bone Osteosarcoma Epithelial Cells) cell lines used for the CALUX bioassays were from Bio Detection Systems (Amsterdam, The Netherlands). The solutions and chemicals used in the CALUX assays are the same as described by Eriksson and coworkers (Eriksson et al. 2022; van der Linden et al. 2014). The purchase information of the substances and medium used for the adipocyte assay is the same as listed by Norgren et al. (Norgren et al. 2022).

2.2. Chemical Mixture Compilation

A previous study conducted a literature review to comprehensively understand which anthropogenic organic contaminants could be found in human blood (Engelhardt et al. 2022). Focusing on the general Swedish population, blood concentrations of 166 environmental contaminants reported were compiled into a database (the Swedish Exposure Database [SEDB]) (Engelhardt et al. 2022). The majority (68%) of the environmental contaminants reported in SEDB were POPs, such as PCBs, dioxins and furans, PBDEs, organochlorine pesticides, and PFAS. Other compounds less frequently analyzed in Swedish samples but with a potentially high human exposure due to current use are plastic additives, such as phthalates and bisphenols.

The 166 environmental contaminants found in Swedish blood (SEDB) were searched for availability in‐house and at chemical vendors. From this, 68 environmental contaminants were available, out of which 54 environmental contaminants could be delivered within the time and budget of the project (Figure 1). Four of the purchased environmental contaminants could not dissolve at the desired concentrations in the solvent (10⁶ x human blood concentrations [xHBC]) and were excluded. In addition, three SPAs were added to the complex mixture based on new findings of widespread human exposure in Sweden (Engelhardt et al. 2025). The total number of environmental contaminants included in the mixture was 50 (Table S1).

FIGURE 1.

Selection process for the environmental contaminants from the Swedish Exposure Database (SEDB) in the human‐relevant chemical mixture.

The concentrations in the mixture were based on the average blood concentration of all individuals analyzed in each study (children, adolescents, adults, and elderly) reported between 1987 and 2020 (the time range of the data in SEDB). The SPA blood concentrations were based on blood donors in the year 2020 from Stockholm, Sweden (Engelhardt et al. 2025). By including three decades of reported blood analysis data, the maximum number of environmental contaminants was included in the mixtures, representing potential lifetime exposure in the Swedish general population, including different subpopulations (from children to elderly). The average of the median, arithmetic mean, or geometric mean concentrations was used to derive the mixture concentration (Table S2).

The environmental contaminants were categorized into six subgroup mixtures. The subgroup mixtures were PCBs and 2,3,7,8‐tetrachlorodibenzo‐p‐dioxin (TCDD) (henceforth named PCB mixture), brominated flame retardant (BFR), pesticide, phthalate, phenol, and PFAS mixtures (Figure 2 and Table S2). The stock solutions (1, 10, or 100 mmol/L, depending on blood level) of the individual environmental contaminants were dissolved in toluene, except the PFAS and phthalates, which were dissolved in DMSO. The stock solutions were visually inspected for any crystal formation, which would indicate solubility issues. Some environmental contaminants were further diluted before compiling the mixtures. The subgroup mixtures were combined into the total mixture by pooling equal volumes of the subgroup mixtures and volume correction using DMSO. Before use in the bioassays, the toluene was changed to DMSO.

FIGURE 2.

Mixtures used in the experiments. The six subgroup mixtures in the top row: polychlorinated biphenyls and 2,3,7,8‐tetrachlorodibenzo‐p‐dioxin (PCB mixture, red), brominated flame retardants (BFR mixture, yellow), organochloride pesticides (pesticide mixture, green), phthalates (phthalate mixture, light blue), synthetic phenolic antioxidants (SPAs) and bisphenol A (phenol mixture, dark blue), and per‐ and polyfluoroalkyl substances (PFAS, purple). The three combined mixtures at the bottom are a 5‐component mixture (brown, bottom left), a 3‐component mixture (light brown, bottom center), and the total mixture (brown, bottom right), including all 50 contaminants.

In the estrogen receptor α (ERα‐) CALUX assay, an additional mixture was tested containing only subgroup mixtures showing significant ERα activity at the highest concentration tested (i.e., the 3‐component mixture, consisting of pesticides, phenols, and phthalates). In the androgen receptor‐ (AR) CALUX antagonistic assay, an additional mixture was tested, containing all subgroup mixtures except for the PFAS mixture (5‐component mixture).

2.3. Chemical Mixture Exposure

Five cell assays were used to assess the endocrine‐disruptive effects and oxidative stress of the mixtures. The ER agonistic and AR antagonistic activities were evaluated by the activation of ERα (ERα‐CALUX) and the inhibition of AR activity (anti‐AR‐CALUX). Due to practical limitations, both agonistic and antagonistic activity could not be tested for both assays. Regarding AR activating potency, environmental pollutants are more frequently active as antagonists rather than agonists, which supports the use of the anti‐AR‐CALUX assay (Czernych et al. 2017; Nováková et al. 2022; Suzuki et al. 2013). There seems to be no apparent discrepancy between ER activities of environmental contaminants when comparing the agonistic and antagonistic endpoints (Czernych et al. 2017; Nováková et al. 2022; Suzuki et al. 2013; Zhang et al. 2014). However, the agonistic endpoint for ER‐CALUX was selected based on prior experience with the method.

Oxidative stress was assessed by the expression of the transcription factor nuclear factor erythroid 2‐related factor 2 (Nrf2‐CALUX). Toxicity typically associated with compounds such as dioxins and dioxin‐like PCBs was assessed through the activation of the aryl hydrocarbon receptor (AhR) using the dioxin responsive (DR‐) CALUX assay. The metabolism‐disruptive potential was assessed using the adipogenesis endpoint (only exposed to the total mixture). The highest concentration tested in the ERα‐CALUX and anti‐AR‐CALUX assays was 1000 xHBC. In DR‐CALUX, up to 4000 xHBC was tested. In Nrf2‐CALUX, the highest tested concentrations were 10,000 xHBC. During sample preparation, high concentrations (≥ 100 xHBC) were visually checked under the microscope for mixture solubility (especially for nonpolar mixtures). The PCB, BFR, and pesticide mixtures were not tested above 1000 xHBC due to the risk of precipitation. Stock dilution series were prepared in DMSO, using a 1:10 serial dilution. When needed, additional concentrations were tested.

2.4. CALUX Assays

ERα‐CALUX, DR‐CALUX, and AR‐CALUX were tested as previously described (Eriksson et al. 2022; Murk et al. 1996). Nrf2‐CALUX was tested as described by van der Linden and coworkers (van der Linden et al. 2014). In short, for all CALUX bioassays, cells were seeded in 96‐well plates, followed by incubation for 24 h (37°C, 5% CO₂, and 100% humidity). MEMα (α‐Minimum Essential Medium) with 10% FBS (fetal bovine serum) was used for the H4IIE‐luc cells, and DMEM/F‐12 (Dulbecco's Modified Eagle Medium Nutrient Mixture F12, without phenol red) with 5% DCC‐FBS (charcoal‐stripped FBS) was used for the U2OS cells. PBS (phosphate‐buffered saline, pH 7.4) was added to the outer wells of the plates.

During exposure, two compound mixtures and one standard reference concentration series were tested in triplicate on all plates. A dilution series of compound mixtures (1:10) was prepared in DMSO and tested in six concentrations in DR‐CALUX and five in the other three assays. The reference standards used were E2 (17β‐estradiol, 10 concentrations, 0.1–100 pmol/L) in ERα‐CALUX, TCDD (eight concentrations, 0.4–300 pmol/L) in DR‐CALUX, and curcumin (10 concentrations, 0.03–100 μmol/L) in Nrf2‐CALUX. In anti‐AR‐CALUX, flutamide (nine concentrations, 0.001–10 μmol/L) was used as an antagonist, and DHT (dihydrotestosterone) at a constant concentration of 300 pmol/L was used as an agonist and added to all exposure media. A negative control (DMSO) was added in triplicate to each plate. The final DMSO concentration was 0.1%, 0.2%, 0.4%, and 1% for ERα‐, anti‐AR‐, DR‐, and Nrf2‐CALUX, respectively.

After 24 h of incubation (37°C, 5% CO₂, and 100% humidity), the media was removed from the plates and cells were washed twice with PBS. In each well, 25 μL of SteadyLite and 25 μL of PBS (with Mg²⁺ and Ca²⁺) were added before storing the plates at room temperature in darkness for 15–20 min. Thereafter, 30 μL of the cell lysate was transferred to a 96‐well white plate and placed in a luminometer (Fluostar Omega) for measurement of luciferase activity.

2.5. AlamarBlue Viability Test

The viability test was performed as previously described (Alijagic et al. 2024), but with some adjustments. Seeding and exposure of cells were performed as stated in the CALUX assay section of this study. After the incubation, the medium was discarded and replaced with 100 μL of DMEM/F‐12 spiked with 10% of AlamarBlue reagent. In addition to a DMSO control (triplicate), a negative control (triplicate with only medium) was added to each plate, to measure background levels. After 3 h of incubation (37°C, 5% CO₂, and 100% humidity), the fluorescence from the wells was measured using a luminometer (Fluostar Omega) at wavelengths 540/590 nm excitation and emission.

2.6. Adipocyte Assay

Human mesenchymal stem cells (hMSCs) were cultured, differentiated, and stained for high‐content cell imaging as described by Norgren and coworkers (Norgren et al. 2022). Briefly, to test mixture effects on adipogenesis, hMSCs (in passages 5–8) were seeded in 96‐well plates at a density of 4200 cells in 100 μL of plating medium per well. After 3 days, the cells were exposed to the total chemical mixture at concentrations of 0.01–100 xHBC in 180 μL of differentiation medium per well. The differentiation medium consisted of DMEM (Dulbecco's Modified Eagle Medium) supplemented with 10% CSS, 1 μg/mL insulin, 0.25 μmol/L dexamethasone, and 0.5 mmol/L IBMX. Bisphenol A (BPA), a known obesogen, was also tested in parallel (30 μmol/L equal to 2000 xHBC). Medium, containing test chemical mixtures, was replaced with a fresh one (containing test exposures) after 3 and 7 days of differentiation. Cells were stained for nuclei and lipid droplets and fixed on day 10.

Using the high content analysis system ImageXpress Pico (Molecular Devices), 12 images were taken per well. The image analysis was performed in CellReporterXpress software using a module for cell scoring. Positive cells (adipocytes) were defined as cells with BODIPY493/503 signal in the lipid droplets in the cytoplasm. The total number of cells was obtained by using the Hoechst 33342‐stained nuclei. The percentage of positive cells (adipocytes) was calculated for each well.

2.7. Quality Assurance and Quality Control

For the agonistic assays, the limit of quantification (LOQ) was set to the effective concentration inducing 15% of the maximum induction [EC15] of the reference standard. Accordingly, the lowest significant mixture response was defined as 15% of the maximum induction of the reference compound. LOQ for the antagonistic assay was set to 80% induction, corresponding to an AR inhibitory standard concentration of 20% (IC20). Thus, the lowest significant inhibition was defined as a 20% reduction relative to the maximum response of the standard compound. For Nrf2‐CALUX, a positive response was defined as > IF1.5 (induction factor that gives a 50% higher response compared to the control).

In the CALUX assays, all mixtures were tested in triplicate, with two separate plates (representing two independent experiments), and at least five concentrations were evaluated. Furthermore, mixtures inducing a positive response (> LOQ) were tested in at least one more separate plate for effect confirmation (in total, at least three independent experiments). For the adipocyte assay, triplicate wells were prepared on each plate in five independent experiments.

For the CALUX assays, the acceptable quality criterium was set to a standard deviation of less than 15% within the triplicates of both the standard compounds and tested mixtures on each plate. The half maximal effective concentration (EC50) values of the reference standards TCDD, E2, and flutamide should be between 8 and 18 pmol/L, 1.9–19 nmol/L, and 1–10 μmol/L, respectively. The concentration of curcumin at IR1.5 was required to be between 0.77 and 7.7 μmol/L.

A toluene blank was prepared in triplicate using the same procedure as the mixture components were combined, to account for any potential effects from contamination originating from the mixture preparation. The toluene blank and control (DMSO) were added as negative controls. In the Nrf2‐CALUX assay, 0.1 mmol/L curcumin was used as a positive control. In the adipogenesis assay, the positive control consisted of 1 μmol/L rosiglitazone (ROSI).

Cell viability was checked for all mixtures in the CALUX assays using the AlamarBlue viability test. The background (medium without cells) was subtracted from the fluorescence values before normalizing the values against the DMSO control. The result from these tests is found in Table S3 and Figure S1. In addition, cells were visually inspected for signs of cytotoxicity utilizing a microscope.

2.8. Statistical Analysis

The data was normalized against the response of the highest reference standard concentration for the agonistic assays (ERα and DR‐CALUX) and the response of the lowest reference standard concentration for the antagonistic assay (anti‐AR‐CALUX). In the Nrf2‐CALUX, the IFs were normalized by the DMSO control on each plate. For the adipocyte assay, the mean value of the triplicate wells was calculated and expressed as a fold change over the negative control sample (Control). The mean DMSO control response was subtracted from both the standard and mixture responses in ER‐ and DR‐CALUX before conversion to obtain responses scaled from 0% to 100% of the maximum induction of the standard.

A two‐tailed unpaired t‐test was used to evaluate significant differences in activities between two concentration‐response curves in the CALUX assays. In the AlamarBlue viability test, a multiple unpaired t‐test was performed to see statistically significant differences between the tested mixture concentrations and the DMSO control.

One plate consisted of a set of triplicates and is considered one independent experiment. To calculate the standard deviation, three plates were used. One‐way ANOVA and Dunnett's multiple comparison tests were used for the statistical analysis of the adipocyte data from five independent experiments. A probability (p) value < 0.05 was considered a significant result. The statistical analysis and concentration‐response curves were generated using GraphPad Prism 10 for Windows (GraphPad Software, La Jolla, California, USA, www.graphpad.com).

2.9. Concentration Additivity

The total, 3‐component, and 5‐component mixtures were hypothesized to follow concentration additivity in this study; consequently, the CA model was applied. The concentration additivity of the subgroup mixtures was tested by comparing the observed and predicted mixture activities. Predicted mixture activity was calculated based on concentration‐response curves of bioassay active subgroup mixtures. In this study, all mixtures exhibiting significant bioassay activity were included in the prediction, including those lacking full concentration‐response curves. Data from at least three independent experiments were used to calculate the predicted activity of mixtures.

Because the composition of the mixtures is known, the concentrations of each subgroup mixture can be expressed as a fraction of the total concentration. The calculation for the CA model is expressed mathematically as in Equation (1) (Berenbaum 1985).

$${\mathit{EC}}_{x,\mathit{mix}}={\left({\sum }_{i=1}^n\frac{p_i}{E{C}_{x,i}}\right)}^{-1}$$ (1)

For each data point in the predicted concentration‐response curve (5%–100%, with increments of 5%), the subgroup fraction (p _i ) was divided by the effect concentration at a certain percentage (x) provoked by the i ^th mixture (EC _{x, i} ). This was done for all bioactive subgroup mixtures (n) before combining the fractions.

When predicting concentration additivity, single environmental contaminants are commonly used and tested, and all environmental contaminants with a significant response are combined into a mixture. In this study, the individual environmental contaminants were replaced by subgroup mixtures and combined into the total mixture (Figure 2). The CA model generally includes environmental contaminants with a complete concentration‐response curve. Mixtures showing only a partial concentration‐response curve could be included but then they should be mathematically modified before applying the CA model (Scholze et al. 2014).

3. Results

In this section, the results from all five in vitro assays exposed to the total mixture are first addressed, followed by the results from the exposure to the subgroup mixtures tested in the CALUX assays. Thereafter, the results from the CA model predictions are presented. Relative potency factors for all mixtures (REP_mix) are provided in Table S4. No activity was seen for the toluene blank, representing background contamination during mixture preparations or from the control. Information on any observed cytotoxicity in the assays can be found in the Supporting Information S1: Section 1.1.

3.1. The Total Mixture Tested in All Assays

The total mixture induced a significant response (15%) in ERα‐CALUX (Figure 3A) at 10 xHBC and in the anti‐AR‐CALUX (IC20) (Figure 3B) at 15 xHBC. In the DR‐CALUX (Figure 3C), a significant response of the mixture (EC15) was observed at 0.1 xHBC. In the Nrf2‐CALUX, the total mixture was tested up to the maximum concentration of 1000 xHBC and did not exceed the LOQ (IF1.5) at any concentrations tested (Figure 3D).

FIGURE 3.

Concentration‐response curves (A, B, C) and graphs (D, E, F) for the total mixture in (A) ERα‐CALUX, the effect is normalized against the maximum 17β‐estradiol (E2) induction; (B) anti‐AR‐CALUX, the effect is normalized against the lowest flutamide induction; (C) DR‐CALUX/AhR, the effect is normalized against the maximum TCDD induction; (D) Nrf2‐CALUX, the induction factor normalized to the Control. Curcumin was the positive control; (E) number of adipocytes in the adipocyte assay; (F) total cell count in the adipocyte assay. ROSI was the positive control (Pos. control) in the adipocyte assay. The black dashed lines correspond to the LOQ of EC15 (A, C), IC20 (B), and IF1.5 (D). The error bars are the standard deviation calculated from at least three independent experiments. The asterisk in graph D‐F denotes significance at *p value ≤ 0.05, **p value ≤ 0.01, and ****p value ≤ 0.0001 between the control and the treated groups.

The number of adipocytes after differentiation showed a statistically significant increase compared to the negative control at 100 xHBC of the mixture and with BPA 2000 xHBC (Figure 3E). The total number of cells was constant for all tested concentrations (Figure 3F), indicating that the cells did not proliferate to make new adipocytes, but rather the existing cells were recruited to differentiate into adipocytes.

3.2. Subgroup Mixtures Tested in the CALUX Assays

The subgroup mixtures PCB, BFR, pesticide, phthalate, phenol, and PFAS mixtures were tested in the CALUX assays to identify the chemical group(s) driving the toxicological effects observed in the total mixture.

3.2.1. ERα‐CALUX

Out of the six subgroup mixtures tested, the phthalate (Figure 4A) and phenol (Figure 4B) mixtures caused a significant induction of ERα activity (15%) at 840 xHBC and 10 xHBC, respectively. The total and phenol mixtures showed super‐induction, i.e., the maximum induction caused by these mixtures was higher than the maximum induction of the reference compound (E2). The other subgroup mixtures did not induce any ERα activity above LOQ (EC15) (Figure S2). However, a statistically significant positive response was observed at 1000 xHBC for the pesticide mixture compared to the lowest concentration tested (Figure S2C).

FIGURE 4.

ERα‐CALUX concentration‐response curves for the subgroup mixtures (A) phthalate mixture, (B) phenol mixture. The response is normalized against the maximum 17β‐estradiol (E2) induction. The error bars are the standard deviation calculated from the mean of a set of triplicates from at least three independent experiments. The black dashed lines correspond to the LOQ of EC15.

All subgroup mixtures with a positive response at 1000 xHBC were combined (3‐component mixture: pesticide, phthalate, and phenol subgroup mixtures) for comparison with the total mixture. No significant difference in ERα activating potency was observed between the total and 3‐component mixtures within the tested concentration range (Figure S3). The 3‐component mixture caused a significant induction (15%) of ERα activity at 10 xHBC, comparable to the total mixture.

3.2.2. Anti‐AR‐CALUX

Three subgroup mixtures showed significant inhibition. The PCB, pesticide, and phenol mixtures caused inhibition (20%) of the AR activity at 120 (Figure 5A), 50 (Figure 5B), and 25 xHBC (Figure 5C), respectively. No clear trend was observed for the BFR and phthalate mixtures (Figure S4). The PFAS mixture concentration‐response curve was inverse (significantly higher in 1000 xHBC compared to 0.1 xHBC), suggesting that this mixture has an agonistic effect in the presence of DHT (Figure 5D). An additional test was conducted to test the AR agonistic activity of PFAS (Figure S5). All tested concentrations were < LOQ, indicating that the PFAS mixture alone does not induce AR activity. When combining all subgroup mixtures except PFAS into a mixture (5‐component mixture), it caused inhibition (20%) at 15 xHBC. The concentration‐response curves of the 5‐component and total mixture were not significantly different (Figure S6). Thus, the PFAS mixture does not affect the results within the tested concentration range.

FIGURE 5.

Anti‐AR‐CALUX concentration‐response curves for the subgroup mixtures: (A) PCB mixture, (B) pesticide mixture, (C) phenol mixture, and (D) PFAS mixture. The error bars are the standard deviation calculated from the mean of a set of triplicates from at least three independent experiments. The black dashed lines correspond to the LOQ which is set at 80% induction of the mix, i.e., 20% inhibition of DHT. The data is normalized against the lowest flutamide standard.

3.2.3. DR‐CALUX

The subgroup mixtures were first tested at 4000 xHBC. Mixtures showing a response, cytotoxicity or were expected to precipitate in the medium (the PCB, pesticide, and phenol mixtures) were diluted and tested in a concentration range of 0.04–400 xHBC. Only the PCB and pesticide mixtures showed induction of the AhR activity above LOQ (15%) at 0.07 xHBC and 1000 xHBC, respectively (Figure 6A,B). As previously stated, the total mixture showed induction at 0.12 xHBC (Figure 3C). There was a statistically significant difference in AhR activating potency between the total mixture and the PCB mixture (Figure S7), indicating that the PCB mixture is more potent than the total mixture. At noncytotoxic concentrations, the pesticide mixture induced a weak response (Figure 6B), and the phenol mixture showed no response (Figure S8). The BFR, phthalate, and PFAS mixtures did not elicit responses above LOQ. These findings indicate that PCBs and TCDD primarily cause the observed AhR activity in the total mixture.

FIGURE 6.

DR‐CALUX concentration‐response curves for the subgroup mixtures: (A) PCB mixture and (B) pesticide mixture. Error bars represent the standard deviation (%) of each data point, calculated from at least three independent experiments. The black dashed lines correspond to the LOQ (EC15).

3.2.4. Nrf2‐CALUX

The total and the six subgroup mixtures were tested in Nrf2‐CALUX for their oxidative stress potential (Figures 7 and S9). Only phenols and the PFAS mixture exceeded the IF1.5 (LOQ), but outside the range of human relevance (1500 and 2500 xHBC, respectively). Due to solubility issues, the PCB, BFR, and pesticide mixtures could only be tested up to 1000 xHBC.

FIGURE 7.

Nrf2‐CALUX induction factors (IFs) for the subgroup mixtures with a significant effect (A) phenol mixture and (B) PFAS mixture. Induction factors are calculated by dividing each response (luciferase induction) by the DMSO blank. The error bars are the standard deviation calculated from the mean of a set of triplicates from at least three independent experiments. The black dashed lines correspond to the LOQ (IF_1.5).

3.3. Concentration Additivity

The observed and predicted activities were compared to see if the activity of the mixture follows the CA model when using subgroup mixtures. In the present study, some subgroup mixtures did not reach full concentration‐response curves at the maximum concentration tested. These mixtures were included in the CA model calculations without mathematical extrapolations as their effect concentrations were close to the LOQ. In previous graphs, the concentrations tested of each mixture were presented with the fold change compared to human blood concentrations. In the coming section, the total concentrations of the compounds in the mixtures are presented. When comparing the results, the predicted activities of the individual subgroup mixtures did not differ significantly from the observed activities of the total mixture in ERα‐, anti‐AR‐ and DR‐CALUX (Figure 8). Likewise, the observed activity from the combined 3‐ and 5‐component mixture was not significantly different from the predicted activity in ERα‐CALUX and anti‐AR‐CALUX, respectively. Conclusively, the mixture activities follow concentration additivity in ERα, anti‐AR‐ and DR‐CALUX.

FIGURE 8.

Testing concentration additivity by comparing the predicted (determined by the CA model) and the observed effects in (A) ERα‐CALUX of the total mixture and (B) the 3‐component mixture. (C) anti‐AR‐CALUX of the total mixture, (D) 5‐component mixture, and (E) DR‐CALUX of the total mixture. The effect is normalized against the maximum 17β‐estradiol (E2) induction (A, B), the minimum flutamide induction (C, D), and the maximum TCDD induction (E). The horizontal black dashed line is the LOQ of EC15 (A, B, E) and LOQ of the concentration that gives rise to 20% inhibition [IC20] (C, D). The error bars are the standard deviation calculated from at least three independent experiments.

4. Discussion

The mixture consisting of 50 chemical components demonstrated activity in multiple receptor‐mediated assays targeting mechanism‐specific endocrine pathways (ERα, AR) and the DR pathway (AhR) at concentrations within the exposure range found in the blood of the Swedish population. The phenol and PCB mixtures were the primary toxicological drivers of the observed effects, whereas certain subgroup mixtures did not elicit any measurable activity in the conducted assays. Using subgroup mixtures to predict mixture activities worked well for the mechanism‐specific assays (ERα‐, anti‐AR‐ and DR‐CALUX), saving time and labor compared to testing each chemical separately. The study identified that the activities of the mixtures followed the CA model for both agonistic assays, as well as for the antagonistic assay (anti‐AR‐CALUX). Although a few ecotoxicological studies have focused on mixtures combining 50 chemicals or more (Könemann 1980; Villa et al. 2012), to the authors' knowledge, this is the first study exposing human cells to up to 50 environmental contaminants in human‐relevant concentrations.

The total mixture induced transcriptional activity of the ERα‐mediated reporter gene at 10 xHBC, with the phenol mixture identified as the primary driver of this activity. In previous studies, BPA has been shown to exhibit agonistic estrogen activity (Kitamura et al. 2005; Li et al. 2010). The SPAs included in the phenol mixture, 2,4‐di‐tert‐butylphenol (2,4‐DBP), 2,6‐di‐tert‐butyl‐4‐methylphenol (BHT), and 2,4,6‐tri‐tert‐butylphenol (AO246), have been previously tested individually in different studies for their estrogenic potential (Akahori et al. 2008; Pop et al. 2018; Wang et al. 2018). All three compounds showed no estrogenic activity, although 2,4‐DBP exhibited significant antiestrogenic activity when co‐incubated with E2 (Wang et al. 2018).

The ERα antagonistic activation potential would be interesting to test in the future. However, 2,4‐DBP and phthalates are ER antagonists (Czernych et al. 2017; Wang et al. 2018). Furthermore, the total mixture inhibits the transcriptional activity of the AR‐mediated reporter gene at 15 xHBC, with the PCB, pesticide, and phenol mixture identified as the driver of the toxicity. At least four PCBs (CB52, CB101, CB126, and CB153), some of the organochloride pesticides (p,p′‐DDE, p,p′‐DDT, HCB, HCH, and pentachlorophenol), as well as phenols and SPAs (BPA, 2,4‐DBP, and BHT) included in the mixture have previously shown AR antagonistic activity (Lemaire et al. 2004; Li et al. 2008; Li et al. 2010; Orton et al. 2009; Pop et al. 2016; Sun et al. 2006; Takeuchi et al. 2017). The observed concentration additivity in ERα, anti‐AR, and DR‐CALUX aligns with previous studies that observed additive mixture effects for complex low‐concentration mixtures (Buha Djordjevic et al. 2020; Tam et al. 2022). No synergistic effects were observed. For synergistic effects, more than one activated pathway is usually involved (Crofton et al. 2005; Escher and Leusch 2012).

In DR‐CALUX, the PCB mixture was identified as the major driver of toxicity. The dioxin‐like CB126 and TCDD present in the PCB mixture are known to activate the AhR (Brenerová et al. 2016). The PCB mixture was statistically significantly more potent than the total mixture in DR‐CALUX, as indicated by both their REP_mix factors and concentration‐response curves (Table S4 and Figure S7). The lower potency of the total mixture compared to the PCB mixture could be explained by the presence of non‐dioxin‐like PCBs, such as CB138, CB153, and CB180, that are antagonists in the same assay (Brenerová et al. 2016). These congeners occur in higher concentrations in human blood than CB126 and TCDD (Table S2), and their antagonist REP factors are relatively high (Brenerová et al. 2016). Furthermore, the discrepancy between the activities of the PCB and total mixtures could be due to the presence of 2,2′,4,4′‐tetrabromodiphenyl ether (BDE47), a PBDE congener known to act as an AhR antagonist (Brenerová et al. 2016). Thus, the predicted AhR‐activating potency of a mixture of chemicals containing both agonists and antagonists would be overestimated if calculated solely based on individual PCBs with AhR‐activating potency, without accounting for the antagonistic effects of certain PCBs and PBDEs.

Only the phenol and PFAS mixtures activated the Nrf2 pathway, showing that these compounds can cause oxidative stress. However, the concentrations had to exceed 1000 xHBC and 1500 xHBC for phenols and PFAS, respectively, to cause a significant effect. While some individuals in the Swedish population have higher blood concentrations of the environmental contaminants than the average, levels exceeding 1000 xHBC are not considered realistic. Thus, the oxidative stress response seen at 1000 and 1500 xHBC is not relevant in the context of human health risks. Another study that tested the oxidative stress potential of seven PFAS compounds (also included in our study) observed that six PFAS caused a statistically significant increase at levels corresponding to 10,000 xHBC in the formation of reactive oxygen species (Wielsøe et al. 2015).

PFAS showed a significant inverse AR antagonistic activity at 1000 xHBC in the subgroup mixture test (Figure 5D). However, the PFAS mixture exerted no AR agonistic activity (Figure S5). Similar results have been seen in another study, where eight PFAS were tested individually with or without coculture with DHT (Behr et al. 2018; McComb et al. 2020). The AR activity observed for PFAS only in the presence of DHT could indicate that PFAS facilitates the transport of DHT or enhances cellular uptake, and without coculturing with DHT, no agonistic effect is observed. When excluding PFAS from the mixture, there was no difference in response between the total and 5‐component mixtures in anti‐AR‐CALUX (Figure S6). However, it is possible that the two curves would deviate from each other at concentrations above 100 xHBC, which was the highest concentration tested.

From the comparison between predicted effects (derived by applying the CA model) and observed effects, it could be concluded that the total mixture followed concentration additivity in ERα‐, anti‐AR‐ and DR‐CALUX (Figure 8). Likewise, both the 3‐ and 5‐component mixtures had no significant difference between the predicted and observed activities in their respective assays. This was expected since large and complex mixtures have been seen to follow concentration additivity. The CA calculations were made including some subgroup mixtures that did not reach full concentration response curves from the tested concentrations (pesticide mixture in DR‐ and phthalate mixture in ERα‐CALUX). Comparisons were made where these low activity‐inducing subgroup mixtures were excluded and included, with statistically similar CA model curves.

In the adipocyte assay, it was indicated that the total mixture contains environmental contaminants that exert metabolism‐disruptive activities and may contribute to an increase in white adipose tissue, characteristic of the development of obesity. Obesogens (a subset of EDCs associated with metabolic disruption) can affect the metabolic functions through several different mechanisms, e.g., through interacting with the peroxisome proliferator‐activated receptor γ (PPARγ) or the 9‐cis retinoic acid receptor (RXR) (Jaskulak et al. 2025). Through these pathways, both the size and number of adipocytes can be altered, potentially leading to obesity. Even though BPA showed evident adipogenic activity at 30 μmol/L in our study, it is not likely the driver of the effect observed for the mixture, as the concentration of BPA in the mixture was relatively low (1.5 μmol/L). For comparison, the previously reported benchmark concentration of BPA for adipogenic effects, as measured by this assay, was 9.3 μmol/L (Norgren et al. 2022). Examples of other obesogens are the phthalates di‐2‐ethylhexyl phthalate (DEHP), butyl benzyl phthalate (BBzP), and dibutyl phthalate (DBP), which have been shown to act as agonists to the PPARγ (using a mouse fibroblast cell line) and thus promote the differentiation of stem cells into adipocytes (Pereira‐Fernandes et al. 2013). Exposure to obesogens is seen as one of the contributing factors to the increased incidence of obesity in young children (Celik and Yesildemir 2025). Apart from BPA and phthalates, BFRs, PFAS, and organochlorine pesticides have been shown to be associated with obesity (Gupta et al. 2020).

This study used internal human blood levels of environmental contaminants to evaluate potential endocrine activity. Thus, no extrapolation from external exposure sources to internal blood levels is needed, eliminating the uncertainties coming with physiologically‐based kinetic modeling. Additionally, assays with human receptors and stem cells eliminate the need for intraspecies extrapolations (such as rats to humans). However, the lower effect concentrations observed in the DR‐CALUX, at 0.1 xHBC, may be due to the use of a rat cell line that is more sensitive compared to human cell lines (Zeiger et al. 2001). A limitation of this study is the usage of some of the immortalized cell lines (U2OS) and human primary stem cells (hMSC) which have limited metabolizing capacity. This could be relevant for the endocrine activity of PCBs and some BFRs, as their metabolites can be more endocrine‐disrupting than the parent compounds (Buha Djordjevic et al. 2020; Tam et al. 2022). Additionally, various kinetic processes influence the distribution of the compound between the blood and target tissue, potentially leading to differences in exposure between in vitro and in vivo systems.

The concentrations used in the mixtures are the average concentration of the Swedish population over 30 years, including both children and the elderly (Engelhardt et al. 2022). Thus, it is not very likely that one individual would have the exact composition of chemicals tested in this study. Some studies within the SEDB reported the same average blood levels as the observed effect concentrations in the assays used here. However, for each individual the chemical exposome varies over time, both qualitatively (in terms of the number and type of environmental contaminants) and quantitatively (in terms of the concentration of environmental contaminants in the blood). Using a population‐based chemical mixture will maximize the relevance for the general population, as it represents the cumulative exposure. The chemicals tested here are chemicals found in the majority of the population. However, there are environmental contaminants present in human blood that are not included in this study, such as metals, novel PFAS, synthetic phenolic contaminants, environmental contaminant metabolites, and organophosphorus flame retardants (Duan et al. 2020; Liao et al. 2023; Sdougkou et al. 2023).

5. Conclusions

Although our study employs cell‐based in vitro tests to evaluate mixtures that neither consider population variability nor outliers with high concentrations, it provides a population‐based representation of the potential endocrine‐disruptive effects that the general population may be at risk of experiencing. Children are a sensitive subpopulation that requires protection. Exposure to EDCs during critical developmental stages can have detrimental effects that persist throughout life. The additive endocrine activity observed at human‐relevant concentrations here motivates consideration of mixtures in regulatory contexts to ensure the protection of the well‐being of future generations.

Author Contributions

Josefin Engelhardt: conceptualization, investigation, formal analysis, data curation, writing – original draft, review and editing, visualization, project administration, and funding acquisition. Nathalie Struwe: conceptualization, investigation, formal analysis, data curation, writing – original draft, review and editing, visualization. Annika Jansson: data curation, writing – review and editing. Vesna Munic Kos: supervision, writing – review and editing. Maria Larsson: conceptualization, supervision, writing – review and editing, funding acquisition. Jana M. Weiss: conceptualization, supervision, writing – review and editing, funding acquisition. All authors read and approved the final manuscript.

Funding

This work was supported by the Swedish Research Council FORMAS (RiskMix No. 2018–02264 and No. 2019–00375), the August Emil Wilhelm Smitts stipendie‐ och understödsstiftelse, and the KK Foundation (Knowledge Foundation, No. 20190098).

Ethics Statement

The authors have nothing to report.

Consent

The authors have nothing to report.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Table S1: Chemical identification and purchase information of the included contaminants.

Table S2: Concentration of mixtures in 1 times human blood concentration (xHBC), and lowest and highest concentrations found in human blood taken from the Swedish Exposure Database (SEDB), (Engelhardt et al. 2022) and BHT, 2,4‐DBP, AO246 concentrations are taken from Engelhardt and coworkers (Engelhardt et al. 2025), along with number of included individuals when calculating the average used for the mixture concentrations.

Table S3:. Results from AlamarBlue viability test in anti‐AR, Nrf2 and DR‐CALUX. Only cytotoxic concentrations (statistically significant [p < 0.05] viability deviations to dimethyl sulfoxide [DMSO] control) are presented.

Table S4: Average REP factors of the mixtures that showed a significant activity, over limit of quantification (>LOQ) in DR‐, ERα, Nrf2‐, and antagonistic AR‐CALUX. REPs calculated at EC or inhibition concentration (IC) 25, 50, and induction factor (IF) 1.5.

Figure S1: Results from testing the total and PCB mixtures in the AlamarBlue viability assay, using H4IIE‐luc cells (DR‐CALUX, 0.4% DMSO). The values have been normalized against the DMSO control. Error bars represent the standard deviation (%) of each data point, resulting from two biological replicates.

Figure S2: ERα concentration‐response curves for the subgroup mixtures (A) PCB mixture, (B) Brominated flame retardant (BFR) mixture, (C) pesticide mixture, and (D) Per‐ and polyfluoroalkyl substance (PFAS) mixture. The effect is normalized against the maximum 17β‐estradiol (E2) induction. The dashed lines correspond to the LOQ of EC15. Error bars represent the standard deviation (%) of each data point, resulting from two biological replicates.

Figure S3: ERα concentration‐response curve for the 3‐component mixture (light brown) and the total mixture (dark brown). The effect is normalized against the maximum 17β‐estradiol (E2) induction. The dashed line corresponds to the LOQ of EC15. Error bars show the standard deviation (%) of each data point, which is a result of ≥ three biological replicates. The error bands show the 95% confidence intervals.

Figure S4: Anti‐AR concentration‐response curves for the non‐significant subgroup mixtures; (A) BFR mixture, and (B) phthalate mixture. Error bars represent the standard deviation (%) of each data point, resulting from two biological replicates. The dashed line marks 80% induction of the mix, i.e., 20% inhibition of dihydrotestosterone (DHT). The data is normalized against the lowest flutamide standard.

Figure S5: AR concentration‐response curve for the PFAS mixture. The effect is normalized against the maximum DHT induction. The dashed line corresponds to the LOQ of EC15. Error bars represent the standard deviation (%) of each data point, resulting from two biological replicates.

Figure S6: Anti‐AR concentration‐response curves for the 5‐component mixture (total mixture without PFAS) and the total mixture. The dashed line marks 80% induction of the mix, i.e., 20% inhibition of DHT, which is determined as the method LOQ for anti‐AR‐CALUX. Error bars show the standard deviation (%) of each data point, which is a result of≥three biological replicates. The error bands show the 95% confidence intervals.

Figure S7: DR concentration‐response curves for the PCB mixture (red) and the total mixture (dark brown). The effect is normalized against the maximum 2,3,7,8‐tetrachlorodibenzo‐p‐dioxin (TCDD) induction. The dashed line corresponds to the LOQ of EC15. Error bars show the standard deviation (%) of each data point, which is a result of≥three biological replicates. The error bands show the 95% confidence intervals.

Figure S8:. DR‐CALUX concentration‐response curve for the phenol mixture. Error bars represent the standard deviation (%) of each data point, resulting from two biological replicates. The dashed line marks 15% induction of the mixture which is determined as the method LOQ for DR‐CALUX.

Figure S9: Nrf2‐CALUX induction factors for the subgroup mixtures not showing a significant effect: (A) PCB mixture, (B) BFR mixture, (C) pesticide mixture, and (D) phthalate mixture. IFs are calculated by dividing each response (luciferase induction) by the DMSO blank. Error bars show the standard deviation of each data point, which is a result of≥two biological replicates. The dashed line marks IF_1.5, which is the LOQ for Nrf2‐CALUX.

Engelhardt, J. , Struwe N., Jansson A., Kos V. M., Larsson M., and Weiss J. M.. 2026. “Evaluating the Endocrine‐Disrupting and Oxidative Stress Potential of a 50‐Component Human‐Relevant Complex Chemical Mixture Using In Vitro Tests.” Journal of Applied Toxicology 46, no. 6: 2003–2018. 10.1002/jat.70011.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.