Disruption of the endocrine system can result in impairment of developmental processes, brain function, metabolism, and the reproductive- and immune systems (Encarnação et al. 2019; Gore et al. 2015; van Duursen et al. 2020; Vandenberg et al. 2012). It is therefore important to screen chemicals for their potential endocrine disturbing properties at human relevant concentrations. In this study an downscaled 96-well plate protocol for the OECD TG#456 steroidogenesis assay was established to screen 24 POPs detected in human blood. The results showed that six compounds significantly altered the estradiol synthesis, whereas 12 compounds affected the testosterone synthesis in the H295R cell line, although the effect size was below the OECD threshold of 1.5-fold change compared to control for all but three compounds. Three of the POPs also induced small but statistically significant effects on cell viability as measured by the MTT assay.

While the OECD guideline recommend data analysis by ANOVA on a per-experiment approach is well-established, this approach has some disadvantages. The standard practice in most in vitro cell-based models is to perform the statistical test on the average of minimum three independent experiments, and after standardization of the solvent control to 1, the variation between groups cannot be assumed to be equal. Strictly speaking, the ANOVA also assumes independence between groups which is broken by application of fold-change of solvent control. Additionally, this approach makes it more difficult to compare magnitude of effects between studies, as there is a lack of summarized conclusion of all independent experiments. Furthermore, according to the guidelines, only two experiments are required given that the results are equivalent in both biological replicates. Capturing the extent of biological variation between experiments is difficult with such few biological replicates. Since passage number and cell batch evidently have significant impact on baseline synthesis of estradiol and testosterone, this is particularly true for the H295R steroidogenesis assay. In this study we therefore introduced the LMM approach as an alternative that enable analysis of multiple independent experiments while considering potential batch-dependent variation. The data was log-transformed to adhere to normal distribution, and the LMM analysis was followed by Dunnett’s multiple comparison test using the lme4 package in R to identify statistically significant conditions (Bates et al. 2015). The LMM includes factors that account for random effects and uncertainty, similar to a two-way ANOVA that also considers random effects. The difference lies in the robustness, as the two-way ANOVA, like the one-way, is sensitive to missing values which the LMM is more lenient towards.

Stakeholders from the industry have expressed concern regarding the OECD H295R assay, questioning the frequency of significant but small effect size responses (Tinwell et al. 2023). The argument is that false positives may lead to unnecessary animal testing, which is contradicting to the aim of the 3Rs. The inter-laboratory validation report by Hecker et al., on the other hand, show a very low number of false positives (Hecker et al. 2011). They did however implement a 1.5-fold change threshold below which significant results would be disregarded, which was later integrated into the final OECD guideline (Hecker et al. 2011; OECD 2023; Tinwell et al. 2023). Nevertheless, it is important to know if compounds have the potential to disrupt steroid hormone synthesis, even if the effect size is small. The endocrine system is reliant on low concentrations of circulating sex steroid hormones, and their disruption could result in severe health effects (Encarnação et al. 2019; Gore et al. 2015; van Duursen et al. 2020; Vandenberg et al. 2012). Furthermore, as all 24 tested POPs were selected based on their abundance in human blood, their cumulative mixture effect may be of concern even if the effect of each individual substance is small.

The majority of in vitro studies using the H295R model do not report validation or testing of reference compounds, except for the positive controls prochloraz and forskolin at fixed concentrations as part of the experimental quality control. Here, we tested six reference compounds recommended by the OECD to compare the original set-up to the downscaled screening format applied to allow higher throughput. Both the standard per-experiment analysis of the fold-change using ANOVA and Dunnett’s multiple correction test, and the LMM with Dunnett’s multiple correction test were performed. The standard method correctly identified all six reference compounds when defining LOEC for testosterone synthesis. For estradiol, four out of six compounds were below the LOEC recommended by the guidelines using the same statistical method. The two exceptions were the strong inducer forskolin, and the weak inducer BPA. For forskolin, the LOEC detected using the OECD-standard ANOVA (0.3 µM) was only slightly higher compared to the OECD threshold (0.1 µM). Estradiol-synthesis in response to BPA, a weak inducer, also failed to meet the OECD criteria for LOEC on a per-experiment basis. While an ANOVA applied to the average baseline production of estradiol from three experiments defined LOEC as 10 µM, which would fulfill the criteria, two out of the three independent experiments failed to identify a LOEC. However, for the third inducer (atrazine, moderate inducer) LOEC was defined in accordance with the criteria. The LMM, on the other hand, correctly identified all reference compounds for both testosterone and estradiol. As this was the intended statistical method for this study, allowing analysis of multiple independent experiments, the assay was considered validated for the purposes of this test compound screening.

The 24 assessed POPs were included in the chemical test library based on their abundance in human blood (Donat-Vargas et al. 2018). The eight concentrations tested in this study range between 1 nM and 10 µM, focusing on low human-relevant exposures. For 20 of the 24 tested compounds, at least one person of the participants in the VIP study had a blood concentration of 1 nM or more. The only exceptions were PCB-74, oxychlordane, PBDE-99 and PBDE-153, for which the maximum detected concentration was below 1 nM. The lowest tested concentration in this study is therefore highly relevant for human exposure (Donat-Vargas et al. 2018). The highest tested concentration, 10 µM, was selected to draw parallels to previous studies, where higher concentration than detected in human samples are more frequently tested. The solubility of compounds in DMSO and threshold for DMSO content in media (0.1%) were limiting factors for determining the test range. The Echo 550 acoustic liquid dispenser that was used to generate the treatments at correct concentrations is compatible with stocks solved in pure water, or DMSO. Due to the lipophilic nature of many test compounds, we selected DMSO as a solvent to maintain equal handling strategies for the treatments. Pilot studies using elevated DMSO concentrations above the 0.1% threshold resulted in negative effects on cell adherence to the plate. Here, the nominal concentrations applied were also assumed to be the concentration at which the cells were exposed. This is standard practice in most in vitro-based studies, including the OECD guidelines. However, while bioavailability in the test system differs from factors that are relevant in human exposure, it still remains a variable that may impact effective concentrations in the cells. The chemical of interest might not have reached the cells to the expected extent, by instead adhering to plastics used in the test system, or media constituents. Actual exposure concentrations may therefore be lower than nominal test concentration depending on the physiochemical properties of the compound and materials used (Fischer et al. 2017; Stadnicka-Michalak et al. 2014).

By the OECD guideline definition, none of the tested PFAS, organochlorine pesticides or PBDEs would qualify as positive for endocrine disruption. The induction or inhibition of steroid hormones was too small (below 1.5-fold change compared to control), and/or the effect was only significant at non-consecutive test concentration(s). Among all tested compounds, only PCB congeners -156 and -180 and PFDA fulfill the first criteria of the OECD-defined data analysis approach, as there was >1.5-fold increase of the estradiol production at 10 µM (PCB-156, PCB-180) or 10 nM (PFDA). However, as this effect was not observed in consecutive concentrations, the conclusion of steroidogenic effects of PCB-156, PCB-180 and PFDA remains equivocal. To fulfill the OECD criteria, additional experiments that include concentrations above 10 µM are needed to confirm whether this observation is part of a dose response curve that expands beyond the range tested in this study. In cells treated with PCB-156, a small but statistically significant decrease in viability was observed at the highest concentration, which also increased estradiol synthesis. An increase in viability would motivate further questions on whether the elevation of estradiol was a product of increased proliferation, rather than a steroidogenic mechanism of action. Now, correction for the loss in viability would instead increase the fold-change in estradiol compared to control, from 2.53 to 2.7-fold.

Chemicals can affect steroid hormone levels in several ways by targeting specific key enzymes and pathways involved in their synthesis. Some chemicals may induce testosterone production, while others decrease it, by for example influencing the expression or activity of CYP17A1 and 3β-HSD. Similarly, estrogen levels can be specifically affected through mechanisms involving aromatase (CYP19A1), which converts androgens to estrogens. The results of this study demonstrated statistically significant increases in testosterone synthesis at low concentrations in the nanomolar range for PCB-74, PCB-99, PCB-118, PCB-138, PCB-180, trans-nonachlor, PBDE-99 and PBDE-153. The majority of these effects occurred at low- but not high concentrations, with only two out of twelve compounds demonstrating LOECs in the µM range. In contrast, estradiol synthesis was significantly altered at the highest concentration (10 µM) for five compounds, while only PFNA had a statistically significant LOEC in the nanomolar range. With the lack of concentration response and only singular statistically significant findings for some test compounds, it is not possible to fully rule out artefacts as potential explanations for the findings. However, as many endocrine disruptive compounds display low-concentration effects and non-monotonic dose-response (Vandenberg et al. 2012), it is also not unlikely that the effects are observed exclusively at the lowest tested concentration. The lack of response at higher test concentrations could be attributed to compensatory protective mechanisms by the cell, although additional investigation is needed to clarify this.

Our screening demonstrated that 10 µM PFOA induced a small significant decrease in testosterone synthesis. The highest concentration of PFOA in blood in the VIP cohort was 26.57 nM, though individuals with over 48 nM (20 ng/mL) have been reported in other studies (Emmett et al. 2006). The benchmark dose level, estimated at 0.3 ng/mL or 0.73 nM in serum, is based on immunotoxicity in children, rather than disruption of steroidogenesis (Stockholm Convention on Persistent Organic Pollutants 2016). Exposure to PFOA is reported to inhibit testicular testosterone biosynthesis and decrease serum testosterone, and has been as a potential mechanism of action in the formation of Leydig cell tumors (U.S. Environmental Protection Agency 2024). Previous investigations on the effects of PFOA on steroidogenesis in H295R cells have reported similar decreases in testosterone synthesis after exposure to 30, 100 and 300 nM (Du et al. 2013b) and 100 µM (Kang et al. 2016). Other studies contradict these findings and report no significant effects of PFOA on testosterone synthesis after exposure to PFOA at 1 µM (Running et al. 2022), 1.6 µM (Rosenmai et al. 2016) and 100 µM (Behr et al. 2018; Wang et al. 2015). Another study reported moderately increased testosterone production at 600 nM and 6 µM (Kraugerud et al. 2011), further contradicting the above mentioned studies. While most studies make some efforts to adhere to the OECD test guidelines, there are differences in the experimental procedure that could motivate the observed difference in effects. The steroidogenic enzyme encoded by CYP11A, which is crucial in the conversion of cholesterol into pregnenolone and act as the first rate-limiting step in steroidogenesis, was downregulated in two studies (Du et al. 2013b; Kraugerud et al. 2011). This could play a part in the observed decrease in testosterone synthesis, although our and their studies did not observe a decrease in estradiol levels. Compensatory activity within the steroidogenic pathway might have maintained consistent estradiol synthesis, possibly through an increase in aromatase gene expression (Du et al. 2013b) and activity (Kraugerud et al. 2011). Additional investigation into expression- and activity levels of steroidogenic enzymes would provide additional information on the mechanism of action of PFOA, and other POPs on H295R synthesis of testosterone and estradiol.

With the large number of unique PCB congeners in technical mixtures and the environment, many have not yet been investigated for steroidogenic effects using the H295R assay. Although PCB-180 is commonly detected in human blood, it is poorly studied compared to other congeners. One previous in vitro study found similar positive effects on estradiol levels as shown here when exposing bovine granulosa cells to 7.6 µM of PCB-180 (Mlynarcikova et al. 2014). In our study, PCB-153 induced estradiol synthesis, while no significant effects on testosterone or cell viability were noted. Other studies of PCB-153 in H295R cells have reported similar effects on estradiol, with significantly increased synthesis at 3 µM, 4 µM and 12.7 µM (Kraugerud et al. 2010; Tremoen et al. 2014; van den Dungen et al. 2015). This indicates that the compound affects targets more downstream in the steroid synthesis pathways. Oxychlordane was one of only two tested POPs that reduced testosterone synthesis. This effect was not biased by any effects on cell viability, and no corresponding decrease in estradiol was noted. Oxychlordane has not been commonly studied for its toxicological potential as it is a metabolite of the component chlordane in the pesticide mixture technical chlordane. In a study where female rats were treated with the mixture technical chlordane, lower plasma levels of testosterone compared to controls was noted (Cassidy et al. 1994), which is consistent with our findings. All three tested PBDE congeners induced a small (13-21%), but significant increase of testosterone synthesis in H295R cells. For PBDE-153 and PBDE-99 this effect was observed at concentrations as low as 1 nM and 10 nM, respectively. Similar effects at low concentration have previously been reported in a other in vitro cell model, using ovarian follicles, which synthesized increased levels of testosterone after exposure to 1 nM PBDE-47 or 0.44 nM PBDE-99 for 24h (Karpeta et al. 2011). Neither treatment elicited any effects on estradiol synthesis, which is similar to the findings in our study. The results are, however, contradicted by studies using a co-culture of granulosa- and theca cells from pig ovarian follicles (Gregoraszczuk et al. 2008; Rak et al. 2017).

Comparing studies using H295R cells reveals clear methodological differences. Some studies stimulate baseline cellular steroidogenesis with forskolin or other compounds to increase the dynamic range of the assay, although this is not in the guidelines (Haggard et al. 2018; Källsten et al. 2022). Other studies treat cells for 24h instead of 48h (Running et al. 2022). The use of alternative plate-formats, such as 96-well plates or dishes, is also not covered by the guidelines (OECD 2023). Variations exist in the number of independent experiments, pre-incubation length, sub-culturing vessel, and maximum DMSO concentration used during treatment. Media formulations also have a clear impact on cell behavior. The cortisol levels have for example been shown to be affected by differences in culture medium (Duranova et al. 2022; Kurlbaum et al. 2020). Even minor differences in procedure may alter cell characteristics, which could explain different findings across studies (Behr et al. 2018; Duranova et al. 2022; Kurlbaum et al. 2020). The passage number is another factor that impact the H295R cellular steroidogenesis. OECD TG#456 is therefore clear that the original cell stock received from the provider should be passaged five times before freezing multiple batches of cells, and that these cells only should be used between passage 4 and 10 after thawing. These recommendations were included to reduce variation between experiments and laboratories (Duranova et al. 2022; Hecker et al. 2011; OECD 2023), but is not followed by many studies. Our study confirmed the importance of passage number, and revealed significant differences in base hormone synthesis even between passage 4 to 7. Another noteworthy finding was the significant differences in baseline testosterone synthesis between cryo-preserved batches of cells. This was initially thought to be due to increases in cell viability, but the viability of cells measured by MTT was not dependent on neither cell passage nor batch. The cell confluence in wells, however, was affected by cell batch, but not passage. An increase in covered cell area without correlating increase in cell viability could be indicative of changes in metabolism, or cell morphology, but additional experiments are required to determine this. The difference in sensitivity to batch-effect between testosterone and estradiol was also an unexpected finding. The statistical significance could be lacking due to the higher variability in the estradiol measurements, likely caused by the low concentration in medium which leads to greater fluctuations even with small changes in estradiol synthesis. True variation in estradiol synthesis by batch might thereby be masked by fluctuations caused by non-specific conditions due to technical variation. Testosterone, by contrast, is detected at higher concentration in H295R cell medium, and small changes due to non-specific treatment conditions are thus less likely to impact the base production across experiments. To our knowledge, these batch effects in cell behavior in H295R cells from the same stock has not been reported before.

The results derived from this human cell assay could be used as motivation for additional mechanistic investigations using in vitro techniques to further elucidate in vivo and human relevance. The expanded 96-well microplate format allows for multiplexing with additional plate-based assays, including high content imaging. This technique would allow screening for additional endpoints of interest, including oxidative stress and mitochondrial health, or levels of relevant proteins such as steroidogenic enzymes. The higher throughput format could also enable analysis molecular alterations by omics methods. Another advantage of the increased screening capacity includes the possibility of investigating customized mixtures of interest at higher throughput, which is also compatible with the acoustic liquid handling utilized here. Discoveries of additional affected toxicological endpoints or other indications of endocrine disruption could then warrant further studies in whole organisms.