All mouse studies were executed in accordance with the European Union (EU) directive 2010/63/EU on the protection of animals used for scientific purposes and in compliance with regulations by the respective local Animal Welfare Committees LMU; Committee on Animal Health and Care of the local governmental body of the state of Upper Bavaria; Germany; Animal Welfare Committee of the Landesamt für Natur, Umwelt und Verbraucherschutz Nordrhein-Westfalen (LANUV; Recklinghausen, Germany). CD-1 mice were purchased from Charles River Laboratories, housed in individual cages with free access to food and water on a 12:12 light dark cycle at the Biocenter, Ludwig-Maximilians-University of Munich (LMU). Mice were sacrificed by cervical dislocation after IsoFlo (Isofluran, Abbott) narcosis. Mct8-KO mice (Trajkovic et al., 2007), Trhr1-KO mice (Rabeler et al., 2004); and Thrb-KO mice (Forrest et al., 1996), all on C57BL/6 background, were kept at 22°C in IVC cages in the central animal facility of the University Hospital Essen and had access to normal chow and water ad libitum. Adult mutant mice and control littermates were killed by CO₂ inhalation prior to organ collection.

Human thyroid tissue was sampled from a freshly operated struma according to the ethical consent (ethical approval No. 12-5133-BO to DF) of the Department of Endocrinology, University Hospital Essen, Germany. Thyroids from chicken (Gallus gallus domesticus) were collected from white legorn birds. Fertilized eggs of the M11 chicken line were kindly provided by Dr. S. Weigend (Federal Research Institute for Animal Health, Mariensee) and hatched at the Faculty for Veterinary Medicine, Munich. Birds were housed under conventional conditions in aviaries with groups of up to 10 birds and received food and water ad libitum. The animals were treated according to the standard protocol approved by The Committee on Animal Health and Care of the local governmental body of the state of Upper Bavaria, Germany. Thyroids from Xenopus tropicalis (Silurana tropicalis) were collected from four-year-old adult animals. Frogs were purchased at the Centre de Ressources Biologiques de Rennes (CRB) and raised in the PhyMa Laboratory in aquatic housing system (MPAquarien, Rockenhausen, Germany) at 24°C. Generation of transgenic zebrafish (Danio rerio) lines expressing fluorophores in thyrocytes, tg (tg:mCherry) and tg (tg:GFP) is described elsewhere (Opitz et al., 2012; Opitz et al., 2013).

In each case, freshly dissected tissues were washed with PBS and then fixed with 4% paraformaldehyde (Carl Roth) solution in PBS for 12–20 h.

Generation of functional mouse and human thyroid tissues in vitro is described in (Antonica et al., 2012; Romitti et al., 2022). In our experiments we used thyrocyte follicles formed in vitro in 3D matrigel culture and follicles formed after grafting differentiated in vitro thyrocytes into mouse kidney. All grafting experiments were performed in accordance with local Animal Ethics [Commission d’Ethique du Bien-Être Animal (CEBEA) Faculté de Médecine ULB, Project CMMI-2020-01].

For cultivating mouse primary thyrocytes in vitro, the protocol from Jeker et al., 1999 was adapted. Briefly, thyroid glands of two mice, were minced into pieces using micro scissors under binocular and transferred into 2 mL tube containing 200 U/mL collagenase and 1 U/mL dispase in DMEM/F12/GlutaMax medium. Incubation was performed in a shaking thermo-block at 37°C for 2 h and followed by mechanical disruption using a glass Pasteur pipetting. Then cells were centrifuged at 2,000 rpm (358 g) for 5 min and resuspended in DMEM supplemented with 10% FCS and 3% Pen/Strep. Single thyrocytes and follicles were seeded on coated coverslips (pre-incubated with 1 μg/mL polylysine) and cultured for 30 min, 24 h or 72 h before fixation with 4% formaldehyde.

After fixation, thyroids were washed with PBS, cryoprotected in a series of sucrose, and embedded in Tissue-Tek O.C.T. compound freezing medium (Sakura). Blocks were stored at −80°C before cutting into 16–20 µm sections using a cryostat (Leica CM3050S). Cryosections were collected on Superfrost Plus slides (Thermo Scientific) and stored at −80°C before use.

Mouse thyroid glands were fixed with 2% glutaraldehyde in 300 mOsm cacodylate buffer (75 mM cacodylate, 75 mM NaCl, 2 mM MgCl2) for 30 min, postfixed with 1% OsO4 in the same buffer for 1 h at room temperature. After washings in distilled water, samples were incubated in 1% aqueous solution of uranyl acetate (Serva) for 1 h at 4°C, dehydrated in ethanol series and acetone, and embedded in Epon Resin. Thin sections (50–70 nm) were prepared using Reichert Ultracut, stained with Reynolds lead citrate and examined with a transmission electron microscope (JEM 100 SX, JEOL) at 60 kV.

Dissected thyroids were immediately placed in RNAlater solution and total RNA was isolated using the NucleoSpin RNA Kit (Macherey-Nagel) according to the manufacturer’s instructions. RNA integrity was checked by separating RNA fragments on a 1% agarose gel. Only samples with a 28S:18S rRNA ratio of ∼2:1 and without genomic contamination and RNA degradation were used for downstream applications. 1 μg of total RNA was reverse transcribed according to the manufacturer’s instructions using the High capacity cDNA reverse transcription Kit with random primers (Applied Biosystems) or Maxima H Minus Reverse Transcriptase (Thermo Scientific) with gene specific primers. qPCR was performed in technical and biological triplicates in 10 µL reactions using LightCycler 480 SYBR Green Master Mix (Roche) or Luna Universal qPCR Master Mix (New England Biolabs) according to the manufacturer’s instructions.

Freshly dissected thyroids were immediately placed in ice cold TriZol and homogenized using an ultra-turrax dispersing tool. Poly(A)+ transcripts were isolated using magnetic oligoT-beads (Lexogen). The sequencing library was created using the PCR-cDNA Sequencing Kit (PCB111.24, Oxford Nanopore) and sequenced on a PromethION P24 on a R9.4.1 flowcell. The sequencing data were basecalled using Guppy v6.4.6 and mapped to the mouse genome (mm10) using minimap2.

BAC clones encompassing the thyroglobulin genes and flanking regions of different vertebrate species, as well as Tshr, Slc16a2 and Tpo mouse genes, were selected using the UCSC genome browser (see the list of BACs in Table 1) and purchased from BACPAC Resources (Oakland children’s hospital) as agar stabs (https://bacpacresources.org/). BACs were purified via standard alkaline lysis or the NucleoBond Xtra Midi Kit (Macherey-Nagel), followed by amplification with the GenomiPhi Kit (GE Healthcare) according to the manufacturer’s instructions. Amplified BAC DNA was labeled with fluorophores using homemade conjugated fluorophore-dUTPs by nick translation (Cremer et al., 2008). Labeled BAC DNA was ethanol precipitated with 10-fold excess of Cot-1 (1 mg/mL; Invitrogen, 18440-016) and 50-fold excess of salmon sperm DNA (5 μg/μL; Sigma), pellet was dried in a SpeedVak, and dissolved in hybridization mixture containing 50% formamide, 1xSSC and 10% of dextran sulphate. Oligoprobe for the Tg mRNA was generated using SABER-FISH protocol and described in detail previously (Leidescher et al., 2022).

FISH on cryosections was performed as previously described (Solovei, 2010). For DNA-FISH, sections were treated with 50 μg/mL RNaseA at 37°C for 1 h. For DNA-FISH or FISH detecting DNA and RNA simultaneously, denaturation of both probe and sample DNA was carried out on a hot block at 80°C for 3 min. For only RNA-FISH, RNasing and denaturation steps were omitted. Probes were loaded on sections under small glass chambers and sealed with rubber cement [for detail, see (Solovei, 2010)]. Hybridizations were carried out in a water bath at 37°C for 2 days. After hybridization, rubber cement and chambers were removed, slides with sections were washed with 2xSSC at 37°C, 3 × 30 min, and then with 0.1xSSC at 60°C 1 × 7 min. Hybridized SABER probes were detected by incubating with 1 μM fluorescently labeled detection oligonucleotides in PBS for 1 h at 37°C followed by washing with PBS for 10 min.

The primary antibody for cytoplasmic TG detection (1:50; rabbit anti-TG Abcam, ab 156008) and the secondary donkey anti-rabbit conjugated with Alexa 555 (1:250; Invitrogen, Cat# A31570) were diluted in blocking solution (PBS +2% BSA +0.1% Saponin +0.1% TritonX100) and applied under glass chambers covering sections. Incubations with primary and secondary antibodies were carried overnight at RT, in between and after incubations, sections were washed with PBS +0.05% TritonX100 warmed up to 37°C; 3 × 30 min. In all experiments, nuclei were counterstained with 2 μg/mL DAPI in PBS for 30 min and Vectashield (Vector) was used as an antifade mounting medium.

Confocal image stacks were acquired using a TCS SP5 confocal microscope (Leica) using a Plan Apo 63/1.4 NA oil immersion objective and the Leica Application Suite Advanced Fluorescence (LAS AF) Software (Leica). Z step size was adjusted to an axial chromatic shift and typically was either 200 nm or 300 nm. XY pixel size varied from 20 to 120 nm. Axial chromatic shift correction, as well as building single grey-scale stacks, RGB-stacks, montages and maximum intensity projections was performed using ImageJ plugin StackGroom (Walter et al., 2006). The plugin is available upon request.

In agreement with its length (180 kb in mouse) and level of transcription (23,000 TPM), the Tg gene forms very prominent transcription loops (Figure 1B4). To examine other long thyroid-specific genes, we selected thyroid peroxidase (Tpo) with a length of 78 kb, the TH transporting monocarboxylate transporter MCT8 (Slc16a2) that is 125 kb long and the thyroid stimulating hormone receptor (Tshr) with a length of 139 kb. The RNA signals of Slc16a2 and Tshr, which have comparable length and ca. 100-500 times lower expression levels than Tg, are small and only slightly elongated. On the contrary, the much shorter Tpo gene, which is about 80 kb long but expressed at ca. 550 TPM, noticeably expands and forms short but distinctive transcription loops (Figure 1B). This finding confirms that formation of microscopically resolvable transcription loops is dependent not only on the gene length, but also on the intensity of transcription (Leidescher et al., 2022).

Remarkably, every single thyrocyte in mouse thyroid manifests Tg transcription, and the majority of cells possess two TLs, with only 6.7% of thyrocytes exhibiting monoallelic expression (Leidescher et al., 2022). The high transcription of the Tg gene is consistent with high production of the thyroglobulin protein (TG). The Tg transcripts comprise 2.5% of the entire mRNA pool in mouse thyroid, similarly to 2.6% shown for human thyrocytes (Pauws et al., 2000). In accordance, the probe for Tg mRNA highlights not only Tg TLs and intranuclear mRNA, but also the whole thyrocyte cytoplasm (Figure 1C). In agreement with this, the cytoplasm is brightly stained with an antibody against TG (Figure 1D), and exhibits dilated cisternae of the Golgi apparatus, as well as a remarkable hypertrophy of the rough endoplasmic reticulum, which occupies most of the thyrocyte cytoplasm (Figure 1E). The above observations corroborate the statement that the Tg gene is abnormally upregulated and thyrocytes produce immense amounts of Tg mRNAs and TG protein.

In mouse, the onset of folliculogenesis starts at E15.5 and the thyroid gland forms before birth (De Felice and Di Lauro, 2011). We aimed at estimating the Tg gene expression in early postnatal development and performed qPCR on thyroid glands dissected at stages P1 and P14. We showed that Tg expression increases during development, reaching a 3.3-fold rise in the adult mice compared to P1 pups (Figure 1F). RNA-FISH revealed that although Tg TLs in most P1 and P14 thyrocytes are fully developed, ca. 20% and 12% of the cells, respectively, possessed small, undeveloped loops and often only one active allele (Figure 1G, arrows). Apparently, these cells represent freshly differentiated or still differentiating thyrocytes. Although mitotic cells in the follicle epithelium are relatively infrequent, in both developmental stages, we observed several thyrocytes entering or exiting the cell cycle. In particular, we noticed that Tg TLs are manifested when most of the nuclear chromatin is condensed: they appear in early G1, remain visible in thyrocyte nuclei until early-to-mid prophase and are withdrawn only in the very late prophase (Supplementary Figure S1). Thus, we conclude, that the Tg gene is expressed at a high level from the moment cells acquire thyrocyte identity.

Initially, the thyroglobulin TLs have been studied in mouse thyrocytes (Leidescher et al., 2022). However, THs are essential regulatory elements in the development and metabolism of all animals, and the mechanisms of TH generation are evolutionary conserved (Di Jeso and Arvan, 2016; Holzer et al., 2016). Therefore, we anticipated that in other mammals and other vertebrate classes, expression of thyroglobulin gene orthologues is also highly upregulated.

First, we demonstrated that human TG (length of 280 kb) exhibits TLs in thyrocytes of a freshly operated human thyroid (Figure 2). Next, we detected the thyroglobulin gene (tg) in the chicken Gallus domesticus (length of ca.140 kb) and the frog Xenopus tropicals (length of ca.154 kb), verifying that the gene forms TLs in these species as well (Figure 2). Finally, we aimed at visualization of the thyroglobulin gene in fish species, where thyrocytes are assembled into single follicles, which are not gathered into a gland, remain separated and are scattered along the esophagus. Therefore, to overcome the difficulties of finding follicles, we made use of the transgenic zebrafish Danio rerio lines generated previously, in which thyrocytes express fluorophores under the tg promoter, tg (tg:mCherry) or tg (tg:GFP) (Opitz et al., 2012; Opitz et al., 2013). Using cryosections from these transgenic lines, we demonstrated that, despite the relatively small length of the D.rerio tg (68 kb), it forms microscopically resolvable TLs strongly expanding throughout the nucleus (Figure 2).

Importantly, all the characteristic TL features, shown previously only for mouse TLs, are utterly manifested in other species as well. Firstly, we observed separation of TL flanks (Figures 3A, B), which has been previously described for other highly expressed long genes and is presumably explained by increased gene stiffness (Leidescher et al., 2022). Secondly, it has been demonstrated that probes highlighting introns over long gene stretches label TLs sequentially as a result of co-transcriptional splicing (Leidescher et al., 2022). This phenomenon is now also confirmed for human and fish thyroglobulin TLs by using sequential BAC probes hybridizing mostly to introns of nascent RNA transcripts decorating the loops (Figures 3C, D).

These data show a high upregulation of the thyroglobulin gene in the major vertebrate classes confirming the key role of TG in hormone production. This conclusion is reinforced by the recent work demonstrating a highly conserved structure of TG protein among vertebrates with all domains conserved from basal vertebrates to mammals (Di Jeso and Arvan, 2016; Holzer et al., 2016). Moreover, the authors of this paper conclude that since no orthologues of thyroglobulin gene have been found in invertebrates, TG protein and the mechanism for TH formation are vertebrate inventions.

Sabine Costagliola’s research group described the generation of a functional thyroid tissue in vitro in 2012 (Antonica et al., 2012). The authors demonstrated that transient overexpression of the transcription factors Nkx2-1 and Pax8 in mouse embryonic stem cells, followed by 3D culturing, allows the generation of follicles similar in morphology and gene expression to the thyroid follicles in vivo. 10 years later, the same group using similar strategy succeeded in generating and growing human thyroid organoids (Romitti et al., 2022). To confirm the functionality of mouse and human follicles generated in vitro, cultured follicles of both species were grafted into kidneys of athyroid mice resulting in successful rescue of hormone production after 4–5 weeks.

Not surprisingly, RNA-FISH revealed Tg TLs in mouse and human follicles cultured in vitro, although the loop size was noticeably smaller in comparison to thyrocytes in glands (Supplementary Figure S2). To visualize mouse Tg TLs in the grafted thyrocytes, we used RNA- and DNA-FISH simultaneously—to identify injected male mouse cells in the female host tissue by a probe for the Y chromosome (DNA-FISH) and to detect Tg loops (RNA-FISH). Accumulations of small follicles were clearly distinguished from the host kidney tissue by Y chromosome DNA-FISH signals (Figure 4A), and predictably, only cells with this marker exhibited Tg TLs, which were much more extended compared to cultured follicles or even thyrocytes within the thyroid gland (Figures 4A2, 4A3).

Differentiation between host mouse cells and human cells grafted into mouse kidney is an easy task, because in difference to human, mouse nuclei possess multiple chromocenters formed by subcentromeric major satellite repeat (Vissel and Choo, 1989; Erdel et al., 2020) brightly stained with DAPI (Supplementary Figure S3A). In addition, transgenic differentiated thyrocytes express NKX2-GFP and thus could be distinguished from non-thyrocytes (Supplementary Figure S3B). Similarly to mouse, human grafted thyrocytes form follicles (Figure 4B1) and the TG genes become strongly expanded (Figures 4B2, 4B). What is more, expressed TG modifies the harboring chromosomal locus by pushing its flanks away from each other (Supplementary Figure S3C) in a manner described for other resolvable TLs (Leidescher et al., 2022). Interestingly, using probes for the gene body and flanks, we noticed that several thyrocytes had four instead of two TG TLs (Supplementary Figure S3C), indicating that either during culturing in vitro or after grafting, some human thyrocytes became tetraploid. The TG TLs exhibit another typical TL feature, the co-transcriptional splicing: the three consecutive overlapping genomic probes sequentially label the nascent RNAs decorating the gene because the introns are sequentially spliced out (Supplementary Figure S3D). Thus, the work with thyrocytes differentiated from ESCs confirms that thyroglobulin TL formation is an invariable mark of differentiated thyrocytes.

The noticeably stronger extension of both Tg and TG TLs in grafts in comparison to thyrocytes in vitro indicates a higher upregulation of the gene within an organism. Culturing of single thyrocytes isolated from thyroid confirms the importance of the follicle structure for thyrocyte functional activity. Following the protocol by (Jeker et al., 1999) for thyroid disintegration and culturing (see Methods), we generated primary transient cultures of mouse thyrocytes, consisting mostly of single thyrocytes and remnants of follicles. Immediately after disintegration, cells were attached to coverslips, fixed and hybridized with a Tg genomic probe. RNA-FISH showed that thyrocytes exhibit Tg TLs for at least an hour after follicle disintegration (Supplementary Figures S4A, S4B). After 24 h of incubation, however, single thyrocytes became flatter and lost Tg TLs (Supplementary Figure S4C), although strongly reduced TLs were still present in some cells within the remaining flattened follicles. At this stage of the culture, no dividing cells were observed. After 72 h of incubation, all thyrocytes migrated out of follicles, became very flat, exhibited proliferative activity (with a mitotic index of 2.7%) and formed a monolayer, in which not a single cell exhibited Tg TLs (Supplementary Figure S4D). Apparently, thyrocytes lost their identity, possibly de-differentiated and entered the cell cycle as reported earlier for thyrocyte cultures established from other vertebrates (Kimura et al., 2001). These data indicate that results of various analyses conducted on cultured thyrocytes, as well as on other differentiated cells transferred to in vitro conditions, must be treated cautiously.

TH production is tightly regulated by the activity of the hypothalamus-pituitary-thyroid axis and controlled by negative feedback loops involving the TH receptor THRB (Ortiga-Carvalho et al., 2016). Hypothalamic thyrotropin-releasing hormone (TRH) activates its pituitary TRHR1 receptor and stimulates the thyroid by stimulating hormone (TSH) release, which in turn acts on TSHR of thyrocytes and stimulates the production and secretion of THs (Figure 5A). However, data on the regulation of the Tg gene activity are controversial. Earlier works showed that continuous presence of TSH is required to maintain TG production (Van Heuverswyn et al., 1984; Kim and Arvan, 1993). More recent work, however, showed that TSH deprivation or lack of functional TSHR due to early development does not affect the Tg expression but greatly reduces the expression of thyroperoxidase and the sodium/iodide symporter (Postiglione et al., 2002).

To test whether Tg expression is regulated by the thyroidal TH status, we sampled thyroids from mice with increased or decreased thyroidal TH production. For an increased thyroidal production we investigated Mct8-KO mice that exhibit a highly increased thyroidal TH content and reduced thyroidal T4 secretion (Trajkovic et al., 2007; Di Cosmo et al., 2010; Trajkovic-Arsic et al., 2010) (Figures 5A, C) as well as Thrb-KO mice that show highly elevated TSH and TH levels (Forrest et al., 1996) (Figures 5A, D). We also examined Trhr1-KO mice that display central hypothyroidism with decreased thyroidal and serum TH concentrations (Rabeler et al., 2004; Groba et al., 2013) (Figures 5A, B). For each condition, three thyroids were fixed for RNA-FISH and other three used for qPCR.

We anticipated to observe a drop of Tg transcription level and consequently a decrease in Tg TL size under “hyper” conditions, whereas “hypo” condition with decreased thyroidal TH concentration might show elevated Tg transcription level and thus increased Tg TL size. In all three conditions, however, Tg mRNA level did not differ from that of control samples and thyrocytes exhibited Tg TLs of the size and morphology similar to those in control animals (Figures 5B–D). Based on these results, we concluded that the intrathyroidal TH status does not influence Tg expression and that the gene is perpetually upregulated regardless of the activity of the hypothalamus-pituitary-thyroid axis.

The thyrocytes function as both exocrine and endocrine glands. On the apical side, a thyrocyte secretes proteins (e.g., TG, TPO) into the follicle cavity filled with colloid, and on the basolateral site it releases TH into the circulation (Figure 6A). Surprisingly, the question whether these two phases of activity happen simultaneously or during separate time-windows remains open. One possibility to separate the phases would be a regulated oscillation in synthesis and excretion of TG, which is the major component of the follicle colloid, due to a potential Tg gene circadian rhythmicity. Indeed, human TSH exhibits a clear circadian rhythm with a peak between 2 and 4 a.m. (Russell et al., 2008). This fact suggests that the Tg gene might also be subject to circadian activity, separating in this way the two thyrocyte physiological phases.

To check the possible oscillation of the Tg gene, we assessed the abundance of Tg RNAs and the presence of Tg transcription loops in thyroid glands from healthy mice round-the-clock. The mice were entrained to a 12 h: 12 h light: dark schedule and sacrificed at four time points according to one of the standard schemes for circadian activity testing. The time points included 1 hour after the light switched on (07:30), 1 hour after the light switched off (19:30), and two time points in the middle (13:30 and 01:30) (Figure 6B). Thyroids of three animals were sampled for each time point, with one lobe of each thyroid used for qPCR to estimate Tg mRNA level and the other lobe used for microscopy.

In the first part of our analysis, using qPCR, we found that levels of Tg mRNA at the four time points were not significantly different (Figure 6C), indicating continuing high transcription and lack of acute waves of mRNA decay (Garneau et al., 2007). However, since the mRNA was extracted from entire cells (more accurately, from entire thyroid glands), our analysis included mixture of cytoplasmic and nucleoplasmic RNA fractions. Therefore, since Tg mRNA is super-abundant in the cytoplasm and apparently less so in nuclei, from such analysis we could not conclude that Tg transcription is not altered between the time points.

To clarify this question, in the second part of our analysis, we performed RNA-FISH highlighting nascent RNA transcripts and found that thyrocytes exhibit well developed Tg TLs at all four time points (Figure 6D and Supplementary Figure 6). Presence of TLs unlikely can be explained by “frozen” or “slowed down” transcription, because speed of RNAPII machinery is known to be pretty stable (3.8 kb/min(⁻¹); Singh and Padgett, 2009), and therefore, microscopy control demonstrates that Tg transcription is indeed going on round the clock. According to the above data, we conclude that the Tg gene is not rhythmically expressed and remains highly upregulated throughout 24 h.

The presence of extended Tg TLs in thyrocytes round-the-clock still leaves an opportunity for the regulation of gene expression by accumulation of transcripts in the nucleoplasm followed by their concurrent export into the cytoplasm. Such mechanism of fine gene expression tuning is known as the intron retention phenomenon described for various cell types (Braunschweig et al., 2014). Intron retention allows an acute release of mRNA into the cytoplasm by synchronous excision of retained introns in response to a stimulus (Mauger et al., 2016). Thus, despite the perpetual transcription of Tg, intron retention might be a mechanism for Tg mRNA accumulation within nuclei during the endocrine phase and their release during the exocrine phase. To test this hypothesis, we isolated RNA from three mouse thyroid glands and performed Nanopore sequencing. As shown in Figure 6E, we have not detected intron retention in the poly(A) fraction of RNA: all sequenced reads mapped exclusively to the gene exons, which indicates that introns of all Tg mRNAs are excised. Therefore, we can rule out intron retention as a mechanism for restricting the TG production window.