The Chip3 circuits were filled with medium at 1 day prior to adding the organoids in the chip. On the day of the Chip3 assembly, the organoids were transferred to the compartments of the Chip3: 40 liver spheroids in one compartment, one Phenion FT skin model in a Millicell culture insert was integrated into a second culture compartment and two to four Matrigel beads containing thyroid follicles were transferred to the third compartment. The circuit was filled with 1 mL fresh medium. The Chip3 circuits were connected to HUMIMIC Starter, the pump control units, operating at a pressure of 350 mbar and vacuum of 300 mbar with 0.5 Hz as the pump frequency. The chips were only removed from the Starter during medium exchange, in which 0.5 mL medium was removed and replaced with the same volume of appropriate fresh medium.

In the first pilot Chip3 experiment using thyrocyte-derived thyroid follicles, the follicles were generated using Method 1 (and thyrocytes from Donor 1) and precultured in static culture for 8 days before being transferred to the Chip3. The Chip3 medium was a ratio of 50:50 ALI medium to h7H medium containing 0.5% BSA, with and without 1 mU/mL TSH. Four days after the Chip3 organoid assembly, circuits were treated with a single dose of 10 µM MMI (Day 12, final circuit concentration) or solvent control (0.1% DMSO). The medium was fully replaced on Day 14 without MMI and the incubation continued until Day 16. All media from the static incubations and Chip3 medium changes were analyzed for T₃ and T₄. All treatments were run in five replicate chips.

The aim of the final Chip3 experiments was to measure T₄ and T₃ levels in the Chip3 after topical doses of daidzein and genistein and compare these with T₄ and T₃ levels after systemic doses of daidzein and genistein which were expected to alter T₄ and T₃ levels. Systemic and topical doses of genistein and daidzein were based on their lowest observed effect (plasma) concentrations (LOEC), estimated from the preliminary static incubations with thyroid follicles described above and results from the thyroid peroxidase inhibition assay (See Supplementary Material S3). The LOECs from the thyroid peroxidase inhibition assay were 19 ± 8.9 µM for genistein and 28 ± 13 µM for daidzein. To ensure systemic doses approximating the LOECs in the Chip3, the metabolism of the test chemicals by liver organoids (i.e., first-pass metabolism in the Chip) was taken into account. This was achieved by adjusting the nominal doses according to the correlation between the nominal concentration added and the concentration remaining after 24 h incubation with liver spheroids (see Supplementary Material S2 for the detailed calculation). The adjustment factors for genistein and daidzein were 1.39 and 2.1, respectively, resulting in doses of 25 µM genistein and 58 µM daidzein being applied.

Following the rationale of applying safety assessment factors to derive safe concentrations in conventional cosmetic safety assessment (Yang et al., 2017), the LOECs were divided by three to obtain a “no observed effect concentration” (NOEC). The NOECs were divided by 100 as a Margin of Safety (MoS) typically considered safe by regulatory authorities. After application of the safety factors, the corresponding “safe” steady state plasma concentrations were estimated to be 63 and 93 nM for genistein and daidzein, respectively. Subsequently, reverse physiologically based pharmacokinetic (PBPK) modelling was applied to calculate a topical dose for each chemical that would translate to the respective safe plasma concentration in the human population i.e., 63 and 93 nM for genistein and daidzein, respectively. The main parameters for the simulation were chosen by combining parameters for a body lotion exposure scenario (60 kg body weight, 15,760 cm² body surface recommended by the Scientific Committee on Consumer Safety (SCCS) (Scientific Committee on Consumer Safety, 2021)) and experimental results for bioavailability: 76% for daidzein and 91% for genistein (mean+1SD, based on skin absorption in native frozen human skin and Phenion FT models (data not shown)). The parameters in the SCCS notes of guidance were slightly modified to meet experimental requirements: Instead of 2.28 applications per day (the mean frequency of body lotion use), we applied a single dose per day and, instead of the recommended application of 0.5 mg/cm² for body lotions, we applied 5 mg/cm² as this amount resulted in an even spread of lotion across the entire skin model. Based on these factors, doses for the first Chip3 experiment were determined to be 1.55 and 2.35 μg/cm² for genistein and daidzein, respectively, and were expected to be “no observed effect level” (NOEL) doses. Since this was not the case (see Section 3.4.1), a second experiment was performed. The topical doses of daidzein for the second experiment were 2.35 μg/cm² and a 10-fold lower dose of 0.235 μg/cm² to derive a “safe dose” that would not be expected to alter hormone levels. The standard conversion of 3-fold from a LOEL to a NOEL, according to Yang et al. (2017) was considered to have insufficient confidence with respect to the resulting hormone inhibition; therefore, the overall safety factor was increased to 10.

The medium used for the final two Chip3 experiments was a ratio of 30:70 ALI to h7H medium, containing the full supplement mix, 0.5% BSA and either 1 mU/mL TSH or 0.1 mU/mL TSH. One day after inoculation of the Chip3 with skin, liver and thyroid organoids (Day 6), the exposure to test chemicals, solvent control or MMI was started. For systemic application of the test chemicals, genistein and daidzein (both dissolved in 100% ethanol in the first experiment and in 100% DMSO in the second experiment) or MMI (dissolved in 100% DMSO in both experiments), the working solutions (or solvent controls) were mixed with 500 µL medium, which was then added to the liver compartment after removal of 500 µL medium. For genistein and daidzein in the first experiment, the final concentration of ethanol in the Chip3 was 2% (v/v) and in the second experiment, the final concentration of DMSO in the Chip3 was 0.1% (v/v). For MMI, the final concentration of DMSO in the Chip3 was 0.1% (v/v) in both experiments. For topical application, 3 mg of the lotion formulation containing the nominal doses of daidzein or genistein was applied to the top of the Phenion FT model (5 mg/cm² to 0.6 cm² surface area). The test chemicals, solvent controls and MMI were applied daily for 4 or 5 days. Treatments were run in 3–5 chips. After every 24 h, a volume of 0.5 mL medium was removed from the circuits and replaced with 0.5 mL of fresh medium (with or without test chemical depending on the treatment scenario).

For the analysis of test chemicals and their metabolites and T₃ and T₄, a volume of 80 µL of the sampled medium was aliquoted into 1.5 mL tubes or 96-well plates and stored at −80°C until analysis by LC-MS/MS. For metabolite analysis only, samples of medium were removed after 4 h (from separate circuits per timepoint such that only 80 µL was removed from each). All remaining supernatants were transferred into 96-deep well plates and stored at 4°C until lactate dehydrogenase (LDH), glucose, lactate and albumin analyses were performed. Tissues were harvested at the end of the incubation and processed further for histochemical analysis or analysis by LC-MS/MS.

The first time thyrocyte-derived thyroid follicles were tested in the Chip3, it was assumed that more cells in the beads would lead to higher hormone production. Therefore, the method of (Kühnlenz et al., 2022) was modified to entrap 150,000 thyrocytes from Donor 1 (i.e., Method 1) instead of freshly isolated thyroid follicles. When this method was used, basal and TSH-stimulated hormone production in the Chip3 decreased over time (Figure 2). This was in contrast to the basal and TSH-stimulated T₃ levels in static cultures of thyroid follicles from the same donor, which increased consistently over time (Supplementary Figure S4). In addition, T₃ production in the Chip3 was much lower than that from thyrocyte-derived thyroid follicles from the same donor but cultured according to Method 2 under static conditions (Supplementary Figure S4). It was not possible to measure T₄ in static cultures prior to adding the thyroid follicles to the Chip3 due to high background levels in the FCS used (from Sigma), despite it being charcoal-stripped. All further experiments with FCS were from a different supplier (Gibco) to avoid this issue. Despite the lower hormone levels in the Chip3, there was a statistically significant stimulation of T₃ by 1 mU/mL TSH (Figure 2A). Similarly, T₄ production by thyroid follicles in the Chip3 was also stimulated by TSH (which were measurable since BSA was used in the Chip3 instead of FCS) (Figure 2B). A mean T₄:T₃ ratio of 4.2-fold was observed in the Chip3 in the presence of 1 mU/mL TSH (Table 1).

Possible reasons for the observed decrease in hormone levels over time in the Chip3 were investigated. These included 1) metabolism by the liver spheroids, 2) binding to proteins in the medium, 3) loss of viability of the organoids and 4) sub-optimal culture of the thyroid follicles. The first three reasons were ruled out by 1) testing the medium for T₄ and T₃ metabolites—none were detected (i.e., below the LOD), 2) incubating T₄ and T₃ with BSA or thyroglobulin for 2 h—there was no loss in the mass balance, and 3) analyzing the media for glucose, lactate, LDH and albumin—none indicated a loss of viability over time (data not shown). The fourth reason was investigated in a separate experiment in which different culture methods were compared—as described in the next section.

In static cultures, after the application of a single dose of MMI on Day 10, T₃ production by non-stimulated and TSH-stimulated thyroid follicles was decreased by 21%–22% on Day 12, and by 48% on Day 14. By contrast, in the Chip3 model, a single application of MMI on Day 12 did not statistically significantly affect T₃ production by non-stimulated or TSH-stimulated thyroid follicles (data not shown).

Two notable differences between the freshly isolated thyroid follicles and the thyrocyte-derived thyroid follicles using Method 1 were observed. Firstly, freshly isolated thyroid follicles were much larger than follicles formed from isolated thyrocytes (Supplementary Figure S5). Despite this difference, the thyrocyte-derived thyroid follicles formed aggregates of cells with the clear presence of lumen, typical for thyroid follicles. Secondly, static cultures of thyrocyte-derived thyroid follicles produced ∼10-fold lower concentrations of hormones compared to freshly isolated thyroid follicles (Table 1). Although the levels of hormones were different in freshly isolated thyroid follicles and thyrocyte-derived thyroid follicles both produced similar ratios of T₄:T₃.

The effect of the thyrocyte culture method on the T₄ and T₃ production by thyroid follicles on Day 7 of culture is shown in Figures 3A,B. Since the number of cells were different, the values here (and elsewhere) were normalized to the number of cells seeded. Method 3 resulted in the highest production of both hormones from one or more donors. The lowest T₄ and T₃ production was observed when single cells were entrapped in Matrigel (Method 1), even though the cell number was 10-fold higher than for Methods 2 and 3. The highest donor-variation was observed for T₄ production, with Donor 2 producing the highest levels when cultured in all three methods, whereas Donor 3 produced negligible amounts of T₄ when cultured using Method 3. Based on these results, Donor 3 was ruled out as a source of cells for the Chip3 experiment, since Method 3 was optimal for the practical application in the chip circuit. Figure 3C shows the TSH-stimulated T₃ and T₄ production in static incubations with thyroid follicles from Donor 2 using Method 3 between Days 7 and 14. T₃ levels were relatively stable over time; however, T₄ production decreased in the last 2 days from 35.8 ± 1.5 to 26.2 ± 4.8 pmoles/million cells. Donors 1 and 3 were only incubated up to Day 11 but levels at this timepoint were similar to those on Day 7 (data not shown). Based on these results, the duration of the culture of thyroid follicles for the first main Chip3 experiment was limited to 10 days.

The effects of repeated topical applications of 1.55 μg/cm² genistein and 2.35 μg/cm² daidzein in a lotion formulation on T₃ and T₄ in the Chip3 over 4 days were investigated (Figure 5). These topical doses were expected not to have any effects on T₄ or T₃ compared to systemic applications of LOECs. None of the treatments tested caused any changes in viability parameters compared to solvent control conditions (see Supplementary Figure S8). T₄ and T₃ concentrations in solvent control treated Chip3 circuits peaked on Day 2 and still exceeded initial values at the end of the experiment (Day 4; 139% and 120% of the initial concentrations, respectively). Repeated topical applications of genistein and daidzein did not statistically significantly alter T₄ and T₃ concentrations compared to solvent control values; however, unlike solvent control-treated circuits, they did decrease over time to 64% (T₄) and 57% (T₃) of their initial values after genistein treatment (which was statistically significant) and to 88% (T₄) and 79% (T₃) of the initial concentration after daidzein treatment (Supplementary Figure S9). In addition, the T₄:T₃ ratio was generally lower after genistein treatment compared to solvent control ratios (Figure 5C). Both these observations suggested that the doses selected did cause a change in the hormone production and could be considered to be the lowest observed effect level (LOEL) doses.

Unfortunately, in this first experiment, the solvent used for the systemic application of genistein and daidzein (ethanol at a final concentration of 2% v/v) was toxic to the thyroid organoids, evident as a complete loss of T₄ and T₃ production, making it impossible to interpret this application scenario. This loss of viability of the thyroid follicles was not evident in the glucose, lactate or LDH measurements, which were all comparable to solvent control values (Supplementary Figure S8).

Since the first Chip3 experiment could not confirm a lack of an effect of genistein and daidzein after topical application and the unexpected toxicity to thyroid follicles by systemically applied ethanol, a second experiment was conducted. The main changes compared to the first experiment were: 1) the solvent for systemic application of genistein and daidzein (100% DMSO was used at a final concentration of 0.1% v/v instead of ethanol), 2) the TSH concentration was reduced to 0.1 mU/mL based on additional static incubations indicating 1 mU/mL was too high (data not shown), 3) a longer test chemical incubation period of 5 days (to capture potentially latent effects) 4) collection of all media and organoids on Day 5 (i.e., the end of the incubation in the Chip3), for the measurement of the distribution of test chemicals and their metabolites, and 5) excluding the topical application of genistein and testing two doses of daidzein, one at the LOEL and a second at an estimated safe dose. While it would have been preferable to test multiple doses of both chemicals, it was not practically possible—indeed the maximum number of chips were used in both experiments (35 chips). Therefore, we focused on the effects of topical doses of daidzein compared with the systemic application of the positive control chemical, MMI, and LOECs of genistein and daidzein. None of the treatments tested caused any changes in viability parameters compared to solvent control conditions (see Supplementary Figure S10). The morphology of the skin, liver and thyroid follicle organoids after selected dosing scenarios are shown in Supplementary Figure S11.

In this experiment, notably lower concentrations of T₄ were observed in solvent control samples compared to those in the first experiment, whereas the concentrations of T₃ were similar in both experiments (Table 1; Figure 6A). The concentration of both hormones decreased in solvent controls over time, such that by Day 5, T₄ and T₃ were 37% and 18% of the initial concentrations, respectively. To validate the results from the second experiment, the effect of the positive control, MMI, was expressed as a percentage of the solvent control at each timepoint and then compared between the two experiments (Figure 6B). In both experiments, MMI resulted in a time-dependent decrease of T₄, and T₃ - down to 19% and 37% of the solvent control, respectively, in the first experiment (by Day 4) and by 5%–14% and 32%–55% of the solvent control, respectively, in the second experiment (by Days 4–5). Importantly, the profile of hormone inhibition was similar; therefore, for this second experiment, all data were expressed as a % of the solvent control at each timepoint.

Repeated systemic application of 25 µM genistein resulted in a time-dependent decrease of T₄ down to 16% of the solvent control values by Day 5, the profile of which was similar to that after MMI treatment (Figure 7A). Like MMI, genistein also decreased T₃ concentrations but to a lesser extent than T₄ (wn to 63%–80% of solvent control treated circuits (Figure 7B)). This lowered the T₄:T₃ ratio from 1.4 on Day 0 to 0.47 on Day 5. Repeated systemic application of 58 µM daidzein increased both the T₄ and T₃ concentrations by different extents (Figures 7A, B), resulting in a lowered the T₄:T₃ ratio comparable to that caused by genistein and MMI over the first 3 days. However, the T₄:T₃ ratio appeared to recover after treatment with daidzein over the remaining 2 days. This recovery was not observed after exposure to MMI or genistein (Figure 7C).

The topical LOEL dose caused both hormones to increase compared to solvent control concentrations - up to 187% (T₄) and 291% (T₃) of solvent control values by Day 5. Despite these increases, the T₄:T₃ ratio in LOEL dose-treated incubations was unaltered over time. Importantly, repeated topical application of the estimated “safe dose” (0.235 μg/cm²) of daidzein affected neither T₃ nor T₄ concentrations over 5 days (Figures 7D, E). As a result, the T₄ and T₃ ratio was also unaltered (Figure 7F).

The concentrations of the test chemicals in the medium at each timepoint are shown in Figure 8. It should be noted that the medium concentration was halved at each timepoint due to replacing half of the medium to sustain the organoid culture; therefore, the symbols denote “trough” concentrations and are connected using dotted lines. Medium samples for the single early timepoint (at 4 h) were collected for systemic applications of genistein and daidzein, which showed initially high concentrations and that ∼40% of genistein (initial was 25 μM, denoted by a star in Figure 8) and ∼57% of daidzein (initial was 58 μM, denoted by an asterisk in Figure 8) had already been metabolized by this time. The metabolites of genistein and daidzein were also monitored and detected but since their formation was associated with decreased T₄ and T₃ inhibition in the static incubations (i.e., they are detoxification pathways), they are not shown. The concentrations of genistein increased over time and were comparable to those of MMI (which was not metabolized, data not shown). The concentrations of daidzein also increased over time but, unlike genistein or MMI, there was noticeable visible precipitation on the liver spheroids after systemic application (Supplementary Figure S11).

Topical application of both the LOEL dose and the “safe dose” resulted in detectable daidzein concentrations in the circuits (0.45 ± 0.13 and 0.01 ± 0.00 µM, respectively, on Day 1, Figure 8B). These concentrations were lower than those expected after systemic administration of the same doses to 1 mL circuit volume.

LOEL dose = 2.35 μg/cm² = 1.41 µg/skin model = 5.55 nmoles/skin model → 5.55 nmol/mL circuit = 5.55 µM

After repeated topical application, the concentrations increased to 2.70 ± 0.47 µM and 0.30 ± 0.02 µM for the LOEL dose and “safe dose”, respectively.

At the end of the incubation, the presence of test chemical and metabolites was analyzed in the different organoids and Chip3 compartments to show their distribution and to confirm that the thyroid follicles were exposed to them. On Day 5, the majority (43%–55% of the total nmoles) of test chemical was in the medium (Figure 9). Due to some precipitation of daidzein on liver spheroids after systemic application, ∼10% of this chemical was recovered from the liver compartment (Figure 9B). Importantly, 14%–20% of the test chemicals were recovered from the thyroid follicles, regardless of the route of application or dose applied. Only 2% of the topical “safe dose” of daidzein remained on the skin surface 24 h after its last application, indicating that most had entered the skin and the systemic circulation. Approximately 20% of the LOEL dose (which was 10-fold higher than the “safe dose”) remained on the skin surface, suggesting that absorption of this dose was not complete. Analysis of the metabolites showed a similar distribution, with the exception that no metabolites were recovered in the skin washes or chip washes (data not shown).

An immediate hurdle in this study was the lack of a reliable supply of fresh human thyroid follicles. This was overcome by generating thyroid follicles using cryopreserved primary human thyrocytes, which had been isolated in two of our laboratories for exploratory use. The use of cryopreserved cells allowed easy scheduling of the experiments and the pre-qualification of thyrocytes from healthy donors before using them in the Chip3, which is not possible using freshly isolated thyroid follicles. Thyrocyte batches that were qualified for use produced similar concentrations of hormones, indicating a low donor-variability; however, we tried to overcome potential donor variation by using thyrocytes from two donors in the final Chip3 experiment. Primary cells were preferred over the use of immortalized cell lines to avoid their associated disadvantages e.g., becoming dedifferentiated and non-responsive to TSH over extensive passaging (van Staveren et al., 2007). The formation of thyroid follicles in culture requires careful 3D culture techniques to be able to express key proteins such as thyroglobulin and exhibit a polarized form (Kraiem et al., 1991; Mauchamp et al., 1998; Kusunoki et al., 2001; Bernier-Valentin et al., 2006; Toda et al., 2011; Saito et al., 2018; Deisenroth et al., 2020). We based the culture of thyrocytes in our studies on the method of Deisenroth et al. (2020), in which thyrocytes were able to self-organize into follicle-like structures when placed on top of Matrigel. Deisenroth et al. reported mean ranges of 8,741–9,311 pg/mL of T₄ and 2,700–3,792 pg/mL of T₃ being produced by thyrocyte-derived thyroid follicles after stimulation with 1 mU/mL TSH, which equate to 75–80 pmoles/million cells T₄ and 27.6–38.9 pmoles/million cells T₃. Using the same method (Method 2), the production of thyroid hormones in this study were somewhat lower although in the same order of magnitude, with 10.7–25.2 pmoles/million cells T₄ and 7.6–18.8 pmoles/million cells T₃. Although not directly comparable, the concentrations of T₄ and T₃ produced by freshly isolated thyroid follicles were ∼10-fold higher than thyrocyte-derived thyroid follicles, which could be attributed to the smaller size and number of fully functional follicles present in the latter. Despite the lower thyroid hormone production by the thyrocyte-derived thyroid follicles, they were still considered to be high enough for the purposes of the case study, especially since the ratio of T₄ to T₃ known to occur in vivo (Quinlan et al., 2020) was also exhibited by these thyroid follicles. While the method of Deisenroth et al. (2020) was shown to be suitable for medium-throughput assays, it was not suitable for use in the Chip3 since it was designed for multi-well plates, whereas the Chip3 compartments need spheroids or beads of cells/tissues which cannot detach and move around the circuit. Therefore, a method in which thyroid follicles could be retained in beads was developed. Interestingly, increasing the cell number and entrapping thyrocytes in Matrigel did not increase the production of T₄ and T₃, moreover, there was lower production per million cells than cultures with 10-fold lower cell numbers. TSH-stimulated thyroid hormone production by thyroid follicles from two of the three donors tested was increased to ∼30 pmoles/million cells by adding a Matrigel overlay to the self-formed thyroid follicles seeded on top of Matrigel, which was maintained in static culture over 10–12 days. Since this method (Method 3) was also more practically applicable to the Chip3 model (the beads of thyroid follicle-containing Matrigel were easily scooped out of the round-bottomed wells and placed into the Chip3 compartment), this was adopted for the main Chip3 experiments.

Having identified a method for thyroid follicles formation, the Chip3 medium was successfully modified to exchange FCS with BSA, which was more suited to the Phenion FT models and liver spheroids. Another advantage of BSA was that this resulted in lower background levels of T₄ and T₃ in the medium and lower LOQs for the analytical method. While the Chip3 medium was modified to optimize thyroid follicles function, this should not be at the expense of the viability of other organoids in the Chip3. The h7H medium contains phenol red, which is less suited to HepaRG cells, therefore alternative phenol red-free thyroid medium and different ratios of the thyroid and ALI medium could be tested in future studies using these cells.

Another change to the medium which was instigated in the final Chip3 experiment was the lowering of the TSH concentration from 1 to 0.1 mU/mL. The reason for this was because concentration response curves in separate static thyrocyte-derived thyroid follicle experiments exhibited a bell-shaped curve with concentrations greater than 0.1 mU/mL resulting in lower T₄ values (data not shown). However, although the lower TSH concentration had minimal effects on T₃ levels in the Chip3, T₄ production was 1.5- to 2-fold lower, and levels gradually decreased after Day 2. In addition to the differing results from static and dynamic incubations regarding the optimal TSH, there was also a difference in the inhibition of hormone synthesis by MMI. T₄ and T₃ production in static incubations were inhibited by a single application of MMI, whereas sustained inhibition by MMI in the Chip3 required repeated dosing (possibly due to recovery of hormone production in the latter). Therefore, future optimization steps for the Chip3 will be conducted in the chips, rather than basing the experimental design on findings from static incubations.

Future studies could explore the use of primary human hepatocytes to form liver spheroids. These better represent the metabolism of thyroid hormones than HepaRG cells (which did not metabolize T₄ or T₃ in this study) and are not sensitive to the presence of phenol red nor high concentrations of ethanol (Easterbrook et al., 2001). Pooled batches of human hepatocytes could also be used to address donor variation or, alternatively, incubations can be carried out using thyrocytes and hepatocytes from the same donor.

In the first Chip3 experiment comparing topical and systemic applications, ethanol was used as the solvent for the systemic application of genistein and daidzein at a final concentration of 2% v/v. This was toxic to the thyroid organoids, evident as a complete loss of T₄ and T₃ production, but not to the skin or liver organoids (since there was no alteration in glucose, lactate or LDH measurements compared to solvent control values). In retrospect, it may have been expected that this concentration of ethanol would cause a loss of hormone production, since it is reported to directly suppress thyroid function in vivo due to cellular toxicity (Balhara and Deb, 2013). Therefore, future studies will evaluate which solvents can be used for in vitro assays using thyrocyte-derived thyroid follicles and, importantly, the highest concentrations which can be used. The endpoint best suited to detecting a loss of thyroid follicle viability in the Chip3 is the production of T₄ and T₃, while other markers measured in the Chip3 are more reflective of skin and liver spheroid viability.

The topical doses of genistein and daidzein for the main experiments were set using a whole body PBPK model to predict a dose that would result in a specific target plasma concentration. When applied to the Chip3 model, the resulting circuit concentrations after repeated topical applications of daidzein were much higher (2.70 ± 0.47 µM) compared to the intended target plasma concentration of 93 nM. This indicated that the compartments considered in the whole-body PBPK model are not representative of the Chip3 circuit and were not optimized to the scaled-down size of the Chip3 circuit (and distribution in the body). To implement a PBPK model for selecting a safe dose in the Chip3 experiments, further refinements are needed to reflect the Chip3 conditions and the associated parameters used for building the models. For the current study, more empirical estimations based on the amount applied, dermal penetration and the circuit volume (1 mL in this case) may have been more suitable for dose setting in the absence of a Chip3-specific PBK model.

Before testing chemicals in the Chip3, confirmation was needed that they could indeed inhibit T₄ and T₃. Therefore, static incubations of thyrocyte-derived thyroid follicles treated daily with genistein, daidzein and MMI confirmed that T₄ and T₃ were both inhibited by all three chemicals and that the biological activity of daidzein was lower than genistein. Moreover, the IC₅₀ values for genistein and daidzein inhibition of T₄ (54 µM and 152 μM, respectively) were comparable to those derived from the thyroid peroxidase inhibition assays (54 µM and 73 μM, respectively), thus confirming the suitability of the thyroid follicle model. This experiment also indicated that metabolism of the test chemicals decreases their inhibition potency in inhibiting T₄ and T₃ production since preincubation with liver organoids decreased the inhibition. The latter is an important consideration when incorporating a metabolizing tissue into any bioassay since the bioactivity of the test chemicals are often impacted.

There are several ways to calculate the effect of a test chemical on T₄ and T₃ levels: nM, pmoles/million cells, % initial levels, % of solvent control values and the T₄:T₃ ratio. Each of these have advantages, for example, the use of nM units enables a direct comparison of concentrations in the Chip3 with in vivo plasma concentrations (which were comparable when freshly isolated thyroid follicles were used but 10-fold lower when thyrocyte-derived thyroid follicles were used). However, if values in solvent controls decrease over time, the impact of the test chemicals is best visualized as a % of the initial levels or of solvent control values at each timepoint. The T₄:T₃ ratio has the advantage that it compares the relative concentrations of T₄ and T₃ in the same sample which is independent of the number of functional thyroid follicles in the circuit and therefore provides a more robust reflection of the impact of the test chemical. The T₄:T₃ ratio is also used as a clinical marker for patients with thyrotoxicosis (Mortoglou and Candiloros, 2004; Narkar et al., 2021), making this an especially relevant marker for the Chip3 experiments. Regardless of the unit, the values for hormone levels were more variable than measurements of test chemical concentrations in the same medium samples. For example, the mean %CVs for T₄ and T₃ expressed as nM across all circuits (solvent control and test chemical samples) and timepoints were 47% and 45%, respectively, whereas the mean %CV for the measurement of the concentrations of test chemicals was only 12%. The variability in hormone production across circuits is likely to be due to the different number of functional thyroid follicles present in the beads, indicating that the culture method needs further optimization to standardize the follicle size and thus production of hormones.

It is important to confirm the biokinetics of test chemicals in any in vitro assay. A lack of bioactivity could be due to a lack of intracellular exposure or, indeed of the tissue itself. In the final Chip3 experiment, the distributions of the test chemicals were measured in all three organoids, the medium and residual chip wash at the end of the 5-day incubation. Importantly, in each dosing scenario, the thyroid contained 15%–20% of the total chemical present, indicating that this target organoid was exposed to genistein and daidzein. Interestingly, very little of the test chemicals were present in the liver spheroids, apart from after repeated systemic application of 58 µM daidzein due to some precipitation of daidzein on liver spheroids.

Only 2% of the topical “safe dose” of daidzein remained on the skin surface 24 h after its last application, indicating that most had entered the skin and the systemic circulation. Approximately 20% of the LOEL dose (which was 10-fold higher than the “safe dose”) remained on the skin surface, suggesting that absorption of this dose was not complete. Analysis of the metabolites showed a similar distribution, with the exception that no metabolites were recovered in the skin washes or chip washes (data not shown).