Gene and protein expression

Amongst the investigated stable cell lines from thyroid [Nthy-ori 3-1, ML-1 (human) and FRTL-5 (rat)] or other tissues [HEK293T (kidney), HeLa9903 (cervix), and HepG2 (liver)], the gene profiling of non-transfected cell lines unveiled detectable consistent endogenous TPO expression only in human ML-1 cells and rat FRTL-5 cells. Notably, the levels of TPO expression in these cell lines were approximately 17 times and 11 times lower compared to HTT and RTT (commercially available standards of total RNA from Human and Rat Total Thyroid), respectively (Fig. 1a, b).

Fig. 1

Comparison of the TPO gene and protein expression among groups of transfected cells, different in vitro models, and reference human (a) and rat (b) thyroid samples. Relative TPO Gene Expression values are related to TPO expression in the reference commercially available standards of total RNA from Human (a HTT) and Rat Total Thyroid (b RTT). a gene expression in human in vitro models (Nthy-ori 3-1, HEK293T, HeLa9903, HepG2, ML-1), human TPO-transfected cell lines (HEK-TPOA2, HEK-TPOA7, HEK-TPOB3, HEK-TPOB12) and reference samples from different tissues normalized to TPO expression in the human total thyroid; b gene expression in rat in vitro model (FRTL-5) compared to reference sample from rat thyroid. c TPO protein expression in the transfected cell lines and different human in vitro models. TPO band is around 103 kDa, and GAPDH is around 36 kDa. HEK-TPOA2, HEK-TPOA7, HEK-TPOB3, and HEK-TPOB12 are hTPO-transfected cell lines. TPO band cannot be found in Nthy-ori 3-1, ML-1, HEK293T, Hela, and HepG2 cell lines. Two bands were found in the HEK-TPOB12 sample (smaller TPO fragment). hTPO standard (842 amino acids) was from LSbio company. We have transfected the full length of TPO, which is 933 amino acids

TPO gene expression was much higher in all assessed single-cell line models of human TPO-transfected HEK cells, whilst it was not detectable in non-transfected HEK293T cells. After the optimization of the vector design, we obtained several TPO-expressing HEK293T clones, in which TPO expression was confirmed both on gene and protein level (Fig. 1). The TPO gene expression was highest in the transfected single-cell culture HEK-TPOA7, exhibiting an approximately 13-fold higher expression compared to the HTT. Also, HEK-TPOA2 and HEK-TPOB12 cell lines showed an approximately eightfold greater TPO expression compared to HTT, followed by HEK-TPOB3 displaying threefold greater expression than the reference HTT (Fig. 1). Furthermore, the TPO gene expression of HEK-TPOA7 was shown to be relatively stable across passages in cultivation (Fig. S1a in Supplementary Materials).

In the protein expression analysis, no TPO-specific band was detectable in either non-transfected HEK293T, HepG2 and HeLa9903 human non-thyroidal cell models or in the human thyroidal cell lines ML-1 and Nthy-ori 3-1 (Fig. 1c). In FRTL-5 cell line, two bands of TPO-antibody reactive proteins were observed, with the smaller size indicating possible TPO fragment (Fig. S2).

Observable TPO-specific bands were detected in all assessed TPO-transfected single-cell line models, including HEK-TPOA2, HEK-TPOA7, HEK-TPOB3, and HEK-TPOB12 at around 103 kDa. However, another truncated human TPO- protein band, appearing at approximately 57 kDa molecular weight, was also detected in HEK-TPOB12 (Fig. 1c). The most robust TPO protein as well as gene expression amongst the novel developed transfected models was observed in HEK-TPOA7 cell line, which was prioritised for further characterization of TPO activity and the development of the assay for TPO inhibition assessment.

ML-1 cells did not show detectable TPO protein expression in the WB analysis despite the detected TPO gene, possibly due to the low protein expression levels. Only FRTL-5 had detectable endogenous levels in both TPO gene and protein expression (Figs. 1b, S2). But the TPO gene expression levels in FRTL-5 was significantly lower than in commercially available rat thyroid reference sample (pooled from 400 male/female Sprague–Dawley rats, age: 8–12 week).

Relationship of protein concentration with peroxidation activity in AUR and Lumi assay

To demonstrate if cells can produce sufficient levels of catalytically active enzyme with activity detectable using AUR and Lumi assay, increasing concentrations of lysates were used in both these assays. In the Lumi assay, the luminescent signal was found to increase with protein concentration across all cell lines regardless of detectable TPO gene or protein expression, with FRTL-5 notably displaying the highest luminescence intensity at the same protein concentration compared to the other cell lines. In the lower concentration range (ranging from 0.015 to 0.18 µg/µL), all cell lines exhibited a nearly linear luminescence-protein relationship (R2 ≥ 0.80), whilst it was less linear at higher protein concentrations (Fig. 2). Importantly, the intensity of the luminol luminescence emitted by HEK-TPOA7 cell lysate with TPO overexpression was low and only slightly greater than that produced by non-transfected HEK293T cells with no TPO at the equivalent protein concentrations (Fig. 2).

Fig. 2

Protein concentration–response relationship of signal intensity fold change in TPO transfected cell line (HEK-TPOA7) and other in vitro models assessed by Luminol assay. Fold change of signal intensity was calculated by dividing the relative luminescence units (RLUs) from cell lysate-containing wells by the RLUs from the background (average of cell lysate-free wells). A fold change value close to 1 indicates minimal or undetectable peroxidation activity. For HEK-TPOA7, HEK293T, and FRTL-5, there were variants with/without treatment with hematin two days before the collection of cell lysates. a The protein concentration range of samples was 0.015–0.6 µg/µL except ML-1 model, where the highest concentration was 0.24 µg/µL. b subset of a chart containing a lower protein concentration range (0.015 to 0.18 µg/µL). c Lower fold change subset of a chart after excluding the FRTL-5 and ML-1 data

The results demonstrate that the luminescence signal in the Lumi assay was not associated specifically with TPO activity, and it could be potentially attributed to the activity of other common peroxidases. One such example is Glutathione Peroxidase (GPx), which plays a pivotal role in safeguarding cells against oxidative damage by reducing hydrogen peroxide and lipid hydroperoxides (Pei et al. 2023). It seems that the Lumi assay can detect unspecific and broad peroxidation in the cellular homogenates of the studied models. However, further research would be needed to characterise the particular peroxidases playing a role in the detected activity across the different cell lines.

In the AUR assay, strongly increasing fluorescence signals with higher protein content were exclusively observed in HEK-TPOA7. The presence of hTPO in HEK-TPOA7 triggered a protein concentration-dependent linear increase in AUR fluorescence signal (R2 ≥ 0.98). FRTL-5 cells exhibited a similar concentration-dependent increase in fluorescence, though much lower TPO activity compared to HEK-TPOA7. ML-1 cells displayed an increase in fluorescence, albeit only at higher protein concentrations and to significantly lesser extent than in case of both HEK-TPOA7 and FRTL-5 lysates. Notably, neither HEK293T nor Nthy-ori 3-1, the cell models without detectable TPO expression, were able to produce elevated fluorescent signal even at high protein concentrations (Fig. 3).

Fig. 3

Protein concentration–response relationship of signal intensity fold change in TPO transfected cell line (HEK-TPOA7) and other in vitro models assessed by AUR assay. Fold change of the signal intensity was calculated by dividing the fluorescence intensity from cell lysate-containing wells by the fluorescence intensity from the background (average of cell lysate-free wells). A fold change value close to 1 indicates minimal or undetectable peroxidation activity. For HEK-TPOA7, HEK293T, and FRTL-5, there were variants with/out treatment with hematin 2 days before the collection of cell lysates. The final protein concentration of samples was 0.015–0.6 µg/µL except of the ML-1 model, where the highest concentration was 0.24 µg/µl. A strong increase of response with increasing protein content was found for the HEK-TPOA7 lysate. b the subset of a after the exclusion of the HEK-TPOA7 data

The presented results demonstrate that the AUR assay is highly specific to the activity of TPO. This was evident from the increase in fluorescence with increasing protein levels observed in HEK-TPOA7 and FRTL-5 cells, which express TPO gene and protein, whilst the cell lines without TPO expression did not show any peroxidation in AUR assay (Fig. 3). Nevertheless, both FRTL-5 and ML-1 models showed only low fold change of fluorescence intensity reflecting very low TPO activity, which renders them unsuitable for sensitive HTS in the AUR assay. The high fold change of fluorescence intensity with increasing protein observed for HEK-TPOA7 lysate and stable TPO activity detected across relatively large span of passages (Figure S1b) document that the transfected hTPO gene is functional. Thus, this novel developed cell line can serve as an efficient source of TPO in the collected cell lysates.

Due to their limitations summarised in Table S3, namely poor growth characteristics, low TPO expression and the expensive media required, the tested stable cell lines (Nthy-ori, FRTL-5, ML-1) derived from thyroid were not suitable for routine HTS purposes. Therefore, transfected models overexpressing TPO were developed to provide enough functional TPO for high throughput testing.

For that purpose, the developed stably transfected HEK293T cells with approximately 30 h doubling time grow much faster than the thyroid-derived cell models and provide for fast protein production. Amongst all assessed cell lines, HEK-TPOA7 exhibits the most robust and consistent TPO gene and protein expression and demonstrates rapid growth (Table S3). This implies its capability to consistently and efficiently supply a substantial amount of TPO protein for the assessment of its inhibition.

Hematin treatment influence

Whilst cultivating the HEK-TPOA7, HEK293T and FRTL-5, the potential impact of hematin addition on peroxidation activity was tested by comparing variants with or without treatment with hematin 2 days before the collection of cell lysates. It has been reported in the literature that haem group is an important cofactor essential for the enzymatic activity of TPO (Ruf and Carayon 2006). Thus, we tried to assess if hematin addition would impact the enzyme activity by running both Lumi and AUR assays with and without hematin addition in concentration described in literature (Schmutzler et al. 2007).

In the AUR assay, the fluorescence signals were comparable between the hematin-treated and non-treated samples (Fig. 3). Moreover, there was no significant difference in IC50 values of the tested compounds between the lysates prepared from cells cultivated with or without hematin (Table S4). This indicates that hematin’s addition is not necessary for the TPO AUR assay for the cell lines cultivated under the conditions used in our study.

However, the Lumi assay indicated a significant increase in signal for FRTL-5 and HEK293T samples after adding hematin (Fig. 2). Although the hematin treatment increased the luminescence signal, it did not have much impact on IC50 values of the model compounds (Table S4). The observed enhancement of signal in the Lumi assay in lysates from cell lines both with and without detectable TPO expression, attributed to hematin, may indicate that hematin had an influence on non-specific peroxidase activity in the cells.

H2O2 influence in the AUR assay

In comparison to the studies by Paul et al. (2014) and Friedman et al. (2016) (300 µM final concentration), we utilised a lower concentration of H2O2 (40 µM final concentration) in the AUR assay due to some artefacts connected with high H2O2 levels in parallel experiments (part of a follow-up study, data not shown here). The results demonstrate that the fold change of fluorescence signal intensity reflecting the TPO activity between these two concentrations of H2O2 in AUR assay was comparable or slightly greater at 40 µM (Table S5). When we compared the IC50 values for various chemicals across these two H2O2 concentrations, we found that there were no significant differences. This suggests that both H2O2 concentration are sufficient for the functionality of the assay with comparable outcomes of IC50 determination for the tested chemicals (Table S6).

Peroxidase activity inhibition by chemical exposure

Figure 4 and Table 1 display the dose–response curves and IC50 or IC20 values and for all tested chemicals, which were obtained from either Lumi assay or AUR assay.

Fig. 4

Comparison of dose–response curves for positive chemicals among lysates from different cells and assays. Relative peroxidation activity rate (%): The RLUs or fluorescence intensity from the well-containing cell lysate and chemical was background-corrected (average RLUs or fluorescence intensity of lysate-free wells), then normalized to the solvent control (100% activity)

In the applied Luminol assay, bisphenol A (BPA), 6-propylthiouracil (PTU), triclosan (TCS), 2,2′-4,4′-tetrahydroxy benzophenone (BP2), resorcinol (RCS), and methimazole (MMI) showed luminescence inhibition across Nthy-ori 3-1, FRTL-5, and HEK-TPOA7 cell lines, allowing for quantification of IC50 concentrations that were comparable across the cell models, except for PFOA, which allowed for quantification of IC20 only with HEK-TPOA7 cell lysate (Table 1).

On the other hand, in the AUR assay, BPA, PTU, TBBPA, TCS, BP2, RSC, and MMI showed strong fluorescence inhibition potency, enabling quantification of IC50 values, whilst PFOA, T3, and SMX exhibited relatively lower inhibition, which enabled to derive only IC20 values in the HEK-TPOA7 cell line (Table 1). There was no significant nonspecific impact of the chemicals on the fluorescence in the AUR assay, except of a minor influence in the case of triclosan, which could be related with the presence of precipitate in the exposure variant without the cell lysate (Table S7 and Fig. S3).

As is obvious from Fig. 4, the dose-–response curves in the Lumi assays were similar for all chemicals across all the tested cell lines regardless of the detectable presence of TPO gene and protein. On the other hand, the AUR assay enabled the measurement of TPO inhibition only in the case of HEK-TPOA7 cell line with high TPO expression. As written above, the other cell lines showed either rather low (FRTL-5; ML-1) or undetectable TPO activity in AUR assay, which corresponds to their level of TPO gene and protein expression.

Interestingly, although as our data show, the Lumi assay is not TPO-specific and probably reflects the general peroxidase activity, it identified the same inhibitors as the TPO-specific AUR assay in our chemical set, though with higher effective concentrations (IC50 levels). There was a complete agreement between Lumi and AUR assays in the identification of nonactive and active inhibitory chemicals across all chemicals tested, with the exception of sulfamethoxazole (SMX). Nonetheless, the IC50 value for SMX in the AUR assay was relatively high (103 ± 14 µM). From the IC50 values presented in Table 1, it is evident that the AUR assay was more sensitive across most compounds. As the effective concentrations in AUR assay on HEK-TPOA7 cells were often more than order of magnitude lower than detected by Lumi assay across the different cell lines, it is not surprising that the low inhibitory effect of SMX was not detected in the Lumi assay. In general, these results suggest that even TPO non-specific assays reflecting general peroxidase activity might provide an indication of potential effect of chemicals on TPO activity or for prioritisation of chemicals for screening with more specific assays.

The AUR assay demonstrated greater specificity to TPO inhibition, but also greater sensitivity demonstrated by lower IC50 values compared to the Lumi assay in all cases with the exception of triclosan (TCS). The higher IC50 observed for TCS in the AUR assay for HEK-TPOA7 results could be attributed to differences in its solubility. Triclosan exhibits a solubility of 10 mg/L (equivalent to 34.5 µM) at 20 °C according to PubChem. Notably, the Lumi assay maintains a higher pH level 8.8 [100 µL total reaction volume contained 50 µL protein dilution (pH = 7.2) and 50 µL GNE buffer (pH = 9)], compared to pH 7.2 in the AUR assay, enhancing the solubility of TCS within this specific assay environment, as TCS is readily soluble in alkaline solutions (Dann and Hontela 2011; Lee et al. 2019). To assess the impact of pH, as the AUR assay for HEK-TPOA7 is conducted at normal pH 7.2, TCS effect was also examined by AUR assay under higher pH conditions (pH 8.8) corresponding to the Lumi assay. The resulting IC50 value for TCS in the AUR assay at higher pH was 72.7 µM (SD = 14), significantly lower than its IC50 value at standard pH 7.2 used in the AUR assay (Fig. S4). However, in the higher pH setting of the AUR assay for HEK-TPOA7, the fold change in signal intensity decreased to the range of 2.7–4.5, compared to more than 10 in the normal pH. These findings underscore the significance of considering the influence of pH on TCS solubility and its potential repercussion on assay outcomes.

We have compared our data to those reported in previous studies for different cell lines using Lumi or AUR assays (Table 1). Nevertheless, for most tested chemicals, including some widespread environmental pollutants, our study brought the first information on their potential to inhibit TPO activity using in vitro human cell-based detection system. In case of Lumi assay, there is only one corresponding study available (Jomaa et al. 2015) and our IC50 levels are in line with it. In the AUR assay, for the few previously tested compounds, IC50 levels of positive chemicals were comparable or in some cases little higher than in studies with other TPO-expressing cell models (Table 1) (Dong et al. 2020; Jomaa et al. 2015). There are various factors that may influence these differences, including the utilization of different TPO sequences in TPO overexpressing models. Notably, Dong et al. (2020) employed the BC095448 sequence, whereas our study utilised newer sequence NM_000547. Whilst they do not differ significanly in the translated region, the newer sequence has significanly longer untranslated region after the terminal codon which might affect the activity of the translated protein. In addition, variations in enzyme conformation and post-translational modifications across different cell lines, such as the use of CHO and LentiX in Dong et al. (2020) versus HEK293T in our transfection, could contribute to these differences. Anyway, our data show good level of agreement with similar studies and in vitro models, documenting the suitability of our HEK-TPOA7 cell model for TPO inhibition screening in combination with the AUR assay since it is TPO-specific and sensitive.

The currently established AUR assay in our and other studies specifically focuses on the peroxidation activity of the TPO. The assessment of TPO’s ability to iodinate and link iodothyronines through ether bonds needs further development of methodology in the future studies.

It must be noted that the AUR data in Jomaa et al. (2015) are based on Nthy-ori 3-1 cells, for which our study did not detect measurable TPO gene or protein expression. Some previous studies (Godlewska et al. 2018; Tuncel et al. 2007; Ząbczyńska et al. 2020) reported TPO expression in this cell line, albeit at very low levels. Specifically, Tuncel et al. (2007) reported a very low TPO expression that would be below the limit of detection in our study (Ct of 35.5). These studies indicate that low TPO expression can be found in Nthy-ori 3-1 in some cases. This may be a reason why Nthy-ori 3-1 worked in the AUR assay in the study by Jomaa et al. (2015).

In addition, the Toxcast (Friedman et al. 2016) employed the AUR assay to identify TPO inhibitors, utilizing microsomes from 120 pooled rat thyroids. Despite the different species and sources of TPO (ex vivo thyroid tissue in ToxCast, in vitro TPO overexpressing cell line in our study), there is a good agreement in the identification of the TPO inhibiting and nonactive chemicals for 12 compounds that were tested both in ToxCast and our study. ETU was the only chemical that showed quantifiable IC50 in the rat thyroid-based assay in ToxCast, but not in our study.

The US EPA CompTox Chemicals Dashboard, referenced in Table S1, provides predictive data on human exposure to the listed chemicals. It includes estimated human exposure prediction data (Table S1) for all effective inhibitors detected in our study. Amongst the tested compounds, the top five chemicals with the highest predicted human exposure rates—Triclosan, Bisphenol A, Resorcinol, Iopanoic acid and Methimazole, cause varying degrees of TPO inhibition. Notably, Resorcinol and Methimazole show strong TPO inhibitory activity with IC50 values at 0.358 (± 0.16) µM and 1.50 (± 0.39) µM in the AUR assay. BPA is also an inhibitor of TPO, though at greater concentration. TCS, although the most prevalent amongst these chemicals, has a relatively low impact on TPO activity, affected also by solubility issues described above. On the other hand, amongst the strongest detected TPO inhibitors belongs commonly occurring benzophenone with relatively high predicted exposure rates and frequent presence in various aquatic ecosystems, as it is often utilised in sunscreens as a UV filter. These results document the ability of chemicals from different use groups, with diverse structures, including some widespread environmental pollutants to inhibit the activity of this critical enzyme controlling the TH synthesis. This highlights the importance and relevance of this mode of action and the need of TD considerations in the hazard and risk assessment.

Cross-species relevance—SeqAPASS analysis

Extrapolation of toxicity data from model species to other species of concern is a considerable challenge in chemical hazard assessment. Differences in chemical sensitivities amongst species can range from a few to more than a thousand times (Doering et al. 2018). TPO inhibition has been connected with the adverse health effects across vertebrates including mammals, birds, amphibians, and fish (Noyes et al. 2019). Based on currently available Adverse Outcome Pathways (in different stages of development), TPO inhibition was identified as highly relevant molecular-initiating event for various vertebrate species linked with diverse adverse outcomes such as impaired neurodevelopment in mammals, swim bladder and eye development in fish and metamorphosis in amphibians (Holbech et al. 2020; Society for Advancement of AOPs 2024). In this work, an in silico tool developed by the US Environmental Protection Agency (EPA) Sequence Alignment to Predict Across Species Susceptibility (SeqAPASS) was used to predict the similarity of sensitivity to chemicals across animal species based on the protein structure (US EPA 2018). The amino acid sequence of the protein in a reference species, human in our case, was compared with the same protein sequence of other species. The higher is the similarity of the proteins of interest, the higher the probability the proteins will be susceptible to the same chemicals and thus cross-species relevance (LaLone et al. 2016). The Level 1 and level 2 analysis of SeqAPASS showed that human TPO is similar to corresponding proteins from other vertebrates, including the commonly used model mammalian, fish and amphibian species (Rat, Mouse, African clawed frog, Rainbow trout, Eurasian carp, Japanese rice fish and Zebrafish) (Table 2, Fig. S5). Furthermore, in the Level 2 functional sequence analysis, it was observed that the similarity scores for these model animal species consistently exceeded 50%. It means that data obtained from studies involving human TPO could inform research and assessment involving other vertebrate species, and vice versa, thereby facilitating cross-species insights and enhancing the translational relevance of experimental findings. The high level of cross-species similarity in susceptibility to TPO inhibition has been shown previously in pig and rat (Paul et al. 2013). The SeqAPASS analyses supports this finding and also demonstrates that the primary structure and functional domain of TPO show high level of similarity not only in mammals but also across other vertebrate classes as well (Table 2, Fig. S5; Haigis et al. 2023). It is likely that the results obtained from our hTPO transfected cell line can be applied qualitatively and indicate relevant TPO inhibitors also for other vertebrate species.

Table 2 SeqAPASS summary analysis report for TPO protein shows significant similarity across candidate animal model species when comparing whole protein (level 1) and the active domain (level 2)

As TPO inhibition has been linked through AOP network to adverse effects including impaired development across species, it is essential to characterise the TPO inhibiting potential of environmental exposure chemicals and their mixtures. The assessment of TPO inhibition by the environmental samples has been scarce so far, namely due to the lack of available high throughput in vitro based methods without the need for ex vivo materials. To our knowledge there is only a single study showing detection of TPO inhibition in rat thyroid microsomes by both surface and treated waste-water samples (Leusch et al. 2018). It is noteworthy that the effect was detected at environmental relevant concentration at least in case of the treated waste-water sample. This indicates that this endpoint could be significant for aquatic organisms exposed to environmental pollutant mixtures. This is supported by a wide range of studies that detected the chemicals that were identified as TPO inhibitors using ex vivo rat thyroid microsomes (Friedman et al. 2016) in different environmental matrices. These include compounds from diverse chemical groups such as polycyclic aromatic compounds, bisphenols, and current use pesticides in matrices such as surface waters (Novák et al. 2018; Sauer et al. 2023), house dust (Nováková et al. 2022; Pinto-Vidal et al. 2024), indoor air (Nováková et al. 2022), and ambient air (Nováková et al. 2020). Even though their detected concentrations can be relatively low, they are mostly part of complex mixtures with many other compounds for which TPO inhibition is unknown. This highlights the necessity of applications of the high throughput methods for the assessment of both individual chemicals, as well as relevant exposure mixtures.