With the closure of the neural tube, which occurs between E8.5 and E10 in mice and between E19 and E29 in humans [16,17], a compartmentalization process is initiated that restricts the access to the enclosed space and the neural progenitor cells lining it. This, in turn, necessitates the establishment of transport mechanisms to enable the flux of substances critical for later stages of brain development. TH represents such a factor that needs to be taken up from the circulation and transported to the inner compartment across forming barriers. In the developing murine brain, T4 (thyroxine; 3,3′,5,5′-tetraiodothyronine) has been detected as early as E16.5, while T3 was observed from at least E18.5 onwards [1,18,19,20]. However, whether this rather late-stage detection simply reflects technical challenges, such as a limited detection sensitivity, remains to be seen. Indeed, the expression of deiodinases and TH receptors as a proxy for TH tissue sensitivity was noted at much earlier stages. Consequently, the TH supply to the developing brain across barriers has to be guaranteed from these early stages of brain development onwards.

One barrier of utmost importance is the blood–brain barrier (BBB), which, in its mature state, is formed by endothelial cells and pericytes and covered by astrocytic endfeet [21]. Brain vascularization starts when neural tube closure is still in progress, and the most precocious performant vessels sprouting into the developing murine CNS tissue have been described as early as E9.5 to E10.5 [22,23]. These newly formed blood vessels do not form a tight barrier but are initially rather leaky, and a primitive BBB can be observed around E15 in mice [21]. Paralleling the process of BBB formation and, thus, a reduction in blood vessel leakiness, the T4 transporter Oatp1c1 (organic anion transporting polypeptide 1c1) becomes expressed in murine blood-vessel-like structures as early as E14.5 [24]. Thereafter, prominent Oatp1c1 expression has been observed throughout development and in adulthood in rodent brain endothelial cells at both the luminal and abluminal membrane [25,26,27,28,29]. In contrast, in the developing human brain, only very low expression levels of OATP1C1 were found in BBB endothelial cells from GW32 onwards [30]. A reciprocal species-specific difference was noticed for Mct8 (monocarboxylate transporter 8) that is not present in murine blood vessels before birth but was observed in human BBB endothelial cells from GW14 to GW38 [24,25,30,31]. In the adult CNS, however, pronounced endothelial Mct8 staining was observed in both humans and mice [25,27,32]. Among other known TH transporters, a prominent capillary staining in the adult brain has also been noticed for Lat1 (L-type amino acid transporter 1) [33,34,35]. In in situ hybridization experiments, Lat1 mRNA was observed already in the murine neural tube from at least E8.5 onwards, as well as in arising fore- and hindbrain structures, while in the developing chicken cerebellum, Lat1 transcripts could be mapped to brain capillaries [36,37].

Like the BBB, the blood–cerebrospinal fluid barrier (BCSFB) formed by the choroid plexus (ChP) represents a second barrier that restricts the access to the CNS. In its mature state, the ChP consists of fenestrated blood vessels, while the barrier proper is established by ChP epithelial cells that arise from the neuroepithelium [38]. There are, in total, four choroid plexi: two in the lateral ventricles, one in the third, and a last one in the fourth ventricle [39]. During neurodevelopment, the choroid plexus may play a critical yet understudied role as it generates cerebrospinal fluid (CSF) and, thus, the pressure needed for brain organization, as well as for its transport function, that downstream affects neural stem cells lining the ventricle walls and that are, therefore, directly exposed to compounds transported across the BCSFB [39]. Choroid plexus precursor cells are defined following neural tube closure and begin to emerge in the lateral ventricles from E11 onwards in mice [40]. To this end, a part of the dorsal midline invaginates, giving rise to the choroid plexus that expresses the TH transporter MCT8 already at very early stages, from at least E8 in chicken, E12.5 in mice, and GW14 onwards in humans [24,30,41]. Likewise, OATP1C1 is present at the BCSFB from at least E8 in chicken, E14.5 in mice, and GW32 in human embryos onwards. The spatiotemporal expression pattern of other TH transporters is less well defined. Using in situ hybridization, the L-type amino acid transporter 2 (Lat2) and the monocarboxylate transporter 10 (Mct10) have been observed in the murine ChP at postnatal stages, though their distribution in the embryonic mouse brain remains elusive [33]. The ChP also expresses high levels of the TH binding protein Transthyretin (Ttr), which can be detected from E11 onwards in mice and serves as a marker for the ChP [39]. Ttr represents the only TH binding protein in the CSF and constitutes up to 20% of its entire protein content [42].

In addition to these classical barriers, recent analyses on human embryonic tissue derived from GW14-38 identified two additional barriers: the outer cerebrospinal fluid–brain barrier between pial cells and basal end-feet of radial glia cells and the inner cerebrospinal fluid–brain barrier formed by neuroepithelial cells lining the ventricular system [30]. This latter barrier disappears later when neuroepithelial cells differentiate into radial glia cells. Hence, these observations put a new focus on radial glia cells in the regulation of local TH availability, which is further emphasized by the expression of MCT8 and OATP1C1 in this cell type, peaking at GW20 in the human fetal brain.

Although the prohormone T4 can bind to TRs, it does so with less affinity than the more receptor-active form T3. In the brain, the T4-to-T3 conversion is catalyzed by type 2 deiodinase (Dio2), and early studies in rats have shown that roughly 80% of the total T3 pool in the CNS is generated this way, emphasizing the outstanding importance of Dio2 [43]. Dio2 activity is already detectable during early stages of brain development and even before the onset of fetal thyroid gland function [44,45,46,47]. Thereby, a pronounced ontogenic increase in fetal brain Dio2 levels can be observed with a four-fold elevation between E17 and E22 in rats [47]. Likewise, in human brain development, an increase in Dio2 activity that correlated with a surge in brain T3 content was seen until gestational week 20 [46].

In situ hybridization studies on P15 rat brain tissue mapped Dio2 expression to astrocytes and tanycytes lining the third ventricle [2,48]. Thus, as astrocytes are part of the BBB, astrocytic Dio2 is well-positioned to directly sense and convert T4 taken up across the BBB. In tanycytes, Dio2 is thought to contribute to the T3 supply to hypothalamic nuclei and may even be involved in the feedback regulation to hypothalamic paraventricular nucleus-residing TRH-expressing neurons, though the final proof for the latter concept is still pending [49].

The cellular distribution in the postnatal brain, however, does not necessarily reflect the picture seen during embryonic development. As an example, although Dio2 in the chicken brain is expressed in adult tanycytes, it was not detected there during embryonic stages [45]. In more recent studies, Dio2 mRNA expression was further noted in blood vessels in chicken embryonic brain sections derived from E8 to E18 [41]. In addition, Dio2 transcript was unraveled at the meninges, ependymal layer of the lateral ventricle, and choroid plexus of perinatal mice [18]. Together, these latest studies point to a yet understudied participation of Dio2 in the local regulation of TH availability and activity directly at brain barriers during early development.

Similarly, type 3 deiodinase (Dio3) harbors an important role for central TH metabolism by inactivating T4 and T3 to rT3 and T2, respectively [8,50]. In the postnatal brain, Dio3 presents with a prominent neuronal expression pattern [51,52]. Though Dio3 activity is relatively high at this stage, an even more pronounced activity is observed in the embryonic CNS when Dio3 is critical to protect the brain from the deleterious effects of excessive TH concentrations [50]. This further aligns with a broader expression pattern, as Dio3 mRNA was detected in the chicken ChP from E8 to the young-hatchling stage [41]. Over this time course, an incremental decline in brain Dio3 activity was reported [53]. Recent RNA sequencing studies in mice at E13.5 and E18.5 further advanced the idea that Dio3 is critical for the proper chronology of developmental events [54].

TH executes a multitude of genomic actions by binding to thyroid hormone receptors that essentially are ligand-dependent transcription factors bound constitutively to TH-responsive elements of T3 target genes [55]. TRs are encoded by two genes, THRA/TRα and THRB/TRβ, from which a number of different protein isoforms are generated by alternative transcription start sites or alternative splicing [56,57]. Thereby, TRα1, TRβ1, and TRβ2 are the main TH-binding isoforms. More recently, non-canonical actions of TH either by binding to cytosolic TRs and activating the phosphatidylinositol-3-kinase pathway or through the binding of T4 to the integrin αvβ3 membrane receptor and downstream activation of MAPK signaling have gained increased attention [42,58].

In the embryonic mouse brain, T3 binding was detected at E15.5 and, thus, earlier than the onset of fetal thyroid gland function [59]. However, technical challenges may mask an even earlier impact of TR activity on brain development. Along this line, TRα1 transcript was already found in the rat neural tube at E11.5, increased in the telencephalon around E13.5, and has surged in the developing cortex and hippocampus by E15.5 [60,61]. In comparison, TRβ1 mRNA was observed later and more distinctly from E17.5 onwards in the developing striatum, hippocampus, and neocortex in a pattern that indicated expression in proliferating neuroblasts. Both TRα1 and TRβ1 mRNA were also detected in the developing human brain from GW8 onwards and, thus, at a time when the fetus depends on maternal TH supply [62].

Postnatally, TRα1 expression peaked in the cerebellum, cerebral cortex, hippocampus, striatum, and olfactory bulb in the first three weeks of life and declined in adulthood [6,61,63]. Analysis of TRα1-GFP reporter mice further confirmed the presence of TRα1 in virtually all neurons in the adult brain [64]. TRβ1 expression peaked in the early postnatal cerebral cortex, hippocampus, striatum, and olfactory bulb and plateaued in adulthood at a lower expression level [6,61,63,65]. TRβ2, in contrast, is present in the paraventricular hypothalamic nucleus and the pituitary and, thus, is implicated in the regulation of the hypothalamus–pituitary–thyroid axis [66]. Importantly, while TRα1 is the main TR isoform in murine neurons, human neurons harbor significantly more TRβ; however, the consequences of this species-specific difference need to be further investigated [67].

All these investigations on the spatiotemporal expression profile of TH transporters, deiodinases, and TRs point to an early impact of TH signaling during critical phases of fetal brain development.