We have developed three 3D in vitro cellular models, able to mimic the microenvironment of a thyroid gland at different levels of anatomical complexity:
• Beyond purified thyroid cells/progenitors from murine origin coaxed into thyroid follicles, we have established the derivation of thyroid follicles from human embryonic (hESC) and induced pluripotent stem cells (iPSC). Our results demonstrated that we can generate structures capable to organise three-dimensionally in follicles, produce thyroid hormone in vitro, and rescue the levels of TH when transplanted in thyroid-ablated mice.
• Recellularization of decellularized scaffolds with these 3D organoids has been accomplished and allows to reproduce the biological composition of a native thyroid. We have also developed two simpler models based on the repopulation of a collagen scaffold and of a cell-laden hydrogel “cookie” configuration. Both allowed to recreate functional rat thyroid function.
• Bioprinted constructs comprising the 3D organoids have shown successful maintenance of murine and human thyroid follicles derived from PSC. We also developed a vascularized bioprinted thyroid model that showed the successful formation of a primitive vascular network, which resulted in enhanced thyroid hormone production.
“Magnetic cells” were developed using endothelial cells and a rat thyroid cell line, demonstrating maintained cell viability and morphology. Proof of concept was achieved for cell guiding in vitro using a static magnetic field in both 2D standard condition and a 3D collagen-based scaffold.
In parallel, a modular microbioreactor was developed to host the 3D in vitro assays. It is compatible with organoid culture, sensing technology and plug-and-play microfluidics. The initial microfluidic bioreactor platform has been upscaled to host up to 54 culture chambers for 3D thyroid tissue constructs simultaneously under flow conditions, suited for high throughput screening. It has been successfully used with mESC- and hESC-derived thyroid follicles. Proof-of-concept validation of the microfluidic bioreactor platform, designed to be compatible with industry demands and translatable for commercialization, has thus been achieved.
Using this platform, EDC screening experiments elucidated key thyroid responses to EDC exposure. Comparative studies between the 3D in vitro models have been performed on selected representatives per class of screened chemicals. Such studies showed that bioprinted assays were more sensitive, indicating the need for complex biological biomimicry. Sex-specific experiments revealed significant differences at the single-cell level when simulating a sex-specific environment in the 3D in vitro models.
To help elucidate the interference of EDCs with thyroid development and function, two mechanisms of action (MoA) have been identified:
• activation of the aromatic hydrocarbon receptor by PAHs (a class of chemicals that occur naturally in, and from the burning of, coal, crude oil and petroleum products) and planar PCBs (industrially-generated fat-soluble substances that persist in the environment and living organisms);
• inhibition of succinate dehydrogenase by phthalates and phthalate esters (a group of chemicals used to make plastics more durable or to help dissolve other materials).
Three adverse outcome pathways (AOPs) initiated by these MoAs have been created and documented in the AOP-Wiki. Additionally, mathematical models linking changes in thyroid function to observed changes in circulating thyroid hormones in vivo were developed.
Several biomarkers reflecting the interference of a chemical with thyroid function were identified using LC-MS/MS proteomics. This approach revealed a protein candidate biomarker signature for EDCs affecting the thyroid. By integrating proteomics with transcriptomic data from EDC-exposed samples, differentially expressed genes and proteins were identified as potential biomarkers for the classification of samples based on their exposure to different EDC classes. Machine learning, particularly using a Random Forest model, highlighted the top 20 biomarkers for each class, providing high classification accuracy and detailing the importance and direction of change for each biomarker.
Regular discussions with regulatory and industry representatives through the SCREENED Stakeholder Group and the EURION cluster have clarified the requirements for our assays in regulatory contexts. Efforts have been devoted to ensure that methods are well documented and standardised. Since in vitro assays are typically part of a battery of assays rather than used alone, we are exploring their combination with other assays within EURION.
