The type 2 iodothyronine deiodinase is expressed primarily in glial cells in the neonatal rat brain

Neonatal Wistar rats (15 days old) were used in these studies. The National Institutes of Health rules and the European Union directives for the care and handling of animals were followed. Northern blots were prepared using nylon membranes (Nytran, Schleicher and Schuell), and hybridizations were carried out according to standard protocols (19, 20). In situ hybridization for the detection of D2 and neuronal-specific enolase mRNA was performed on free-floating 25-μm sections using protocols previously described in detail (21). D2 sense and antisense riboprobes were synthesized in the presence of [³⁵S]UTP using a 366-bp DNA template spanning nucleotides 535–901 from the rat D2 cDNA sequence (18). This area contains little sequence homology with the rat type 1 deiodinase (D1) and D3 mRNAs. (Attempts at using digoxigenin-labeled probes in these studies resulted in unacceptable nonspecific staining.) The sections were either counterstained with Richardson’s blue or subjected to immunohistochemistry for glial fibrillary acidic protein (GFAP) using a polyclonal antibody (Dako) at a 1/2,000 dilution. The atlas of Patxinos and Watson (22) was used for the identification of rat brain structures.

In situ hybridization was performed on the brains of euthyroid 15-day-old neonatal rats, the time when D2 activity is maximal in this tissue. The riboprobe used in these experiments specifically detected D2 mRNA, as demonstrated by Northern analysis, by using brown adipose tissue RNA prepared from control and cold-exposed rats (Fig. 1). Cold exposure, which is known to induce a marked increase in brown fat D2 activity (23), was associated with increased expression of a 7-kb major RNA species, as well as other minor bands, as we have previously described (18).

Representative coronal sections from different levels of the neonatal rat brain demonstrated that D2 mRNA is expressed in several regions. These include the olfactory bulb and anterior olfactory nucleus (Fig. 2A), the cerebral cortex (Fig. 2 A–E), the caudate-putamen nucleus (Fig. 2 B and C), the dorsal portion of the lateral septal nucleus (Fig. 2B), the thalamus (Fig. 2 D and E), the hippocampus (Fig. 2 D and E), and the hypothalamus (Fig. 2 E and F). D2 was also expressed in the granular layer of the cerebellum as well as pontine nuclei such as the ventral cochlear nucleus and the olivary nuclei (Fig. 2G). Lower levels of expression were found in the midbrain (Fig. 2 H and I). Areas of white matter, such as the corpus callosum, subcortical white matter, internal capsule, and fimbria of the hippocampus, were not labeled.

The distribution of D2 expression in the cerebral cortex was not uniform; layers 1–4 showed the strongest signal, whereas expression was low in layer 5. In the piriform cortex, expression was noted in layer 1, whereas no signal was detected in the neuronal layer 2. There was also some regionalization in the caudate, with the highest expression localized to the ventrolateral portion of the nucleus. In the hippocampus, both the dentate gyrus and the hippocampus proper (Ammon’s horn) were labeled. D2 mRNA appeared to be absent from the amygdala.

An extremely high signal was present in the hypothalamus (Fig. 2 E and F) in the regions corresponding to the lateral aspects of the median eminence and the lining of the lower portion of the third ventricle (Fig. 2 E and F, arrows). This corresponds to the known distribution of ependymal tanycytes (24, 25), a specialized population of glial cells implicated in the transport of hormones, and other substances in the hypothalamus (26). The relative amounts of D2 mRNA in the hypothalamus and layers 1–4 of the parietal cortex were compared by measuring the optical density of these regions on a low-exposure autoradiograph (Fig. 2F). Signal strength in the region of the third ventricle and median eminence was 4- to 5-fold higher than that in the outer layers of the cortex.

The expression pattern in the cerebral cortex was intriguing because layer 1, which shows high D2 expression, is practically devoid of neuronal bodies and contains mainly glial cells and neuropile. To more precisely localize the signal to specific cell types, we performed emulsion autoradiography and counterstained the sections. Fig. 3 shows corresponding dark-field and bright-field images of several brain regions. In the olfactory bulb (Fig. 3 A and B), D2 was expressed in the external plexiform layer, which contains primarily glial cells and processes. The neuron-containing mitral cell and granular layers immediately above showed little signal. Also, the packed cells in the subventricular layer (Fig. 3B), which was still prominent at this age, completely lacked a hybridization signal. In the dentate gyrus (Fig. 3 C and D), the label was present in the molecular layer, whereas the granular cell layer of neurons was practically devoid of signal. In Ammon’s horn (Fig. 3 C and D), the radioactive signal was present in the stratum radiatum in close association with the pyramidal cell layer. The pattern in the cerebral cortex (Fig. 3 E and F) confirmed the results obtained by macroscopic autoradiography; expression was localized mainly to layers 1–4, with less signal in layers 5 and 6.

In the hypothalamus (Fig. 3 G and H), high expression was present in the lining of the third ventricle (Fig. 3G, arrow) and in a narrow rim at the lateral and basilar regions of the median eminence (Fig. 3G, arrowhead). The inset in Fig. 3H shows a magnification of this area under bright-field illumination. The silver grains were concentrated immediately above the packed cells of the pars tuberalis (arrowheads), which completely lacked hybridization signal.

These results strongly suggested that D2 was expressed mainly by two types of glial cells, astrocytes and tanycytes. This was confirmed by performing in situ hybridization for D2 mRNA and immunohistochemistry with an anti-GFAP antibody on the same sections. Fig. 4A shows astrocytes from layer 1 of the parietal cortex. They have the typical star-like appearance, with many long and delicate processes forming an intricate network of fibers. The majority of silver grains were observed over the cell processes, sometimes at a considerable distance from the cell body. Fig. 4B shows colocalization of silver grains with the GFAP-positive fibers crossing the molecular layer of the dentate gyrus. The distribution of D2 mRNA in the tanycytes lining the third ventricle is shown in Fig. 4C. Silver grains were observed in both the cell bodies (arrows) and the cellular processes (arrowheads) that cross into the adjacent arcuate nucleus. There were also numerous hybridization grains in the apical protrusions (blebs) of the tanycytes. Although the periventricular location of the hypothalamic D2 expressing cells indicated that they were tanycytes, their identity was confirmed in other studies (data not shown) by demonstrating intense staining of these cells for vimentin, a tanycyte protein not found in mature astrocytes (27).

D2 expression in the cerebellum appeared localized to the densely packed layer of granule neurons, suggesting neuronal expression in this region. This was further examined by in situ hybridization using two riboprobes, an ³⁵S-labeled D2 probe and a digoxigenin-labeled neuronal-specific enolase probe, which specifically labels neurons. As shown in Fig. 4D, there is little association of silver grains with the neuronal-specific enolase mRNA (blue spots). In contrast, immunostaining of astrocytes with GFAP antibodies demonstrated the presence of D2 mRNA in these cells (Fig. 4E). Thus, expression of D2 in the cerebellum is also primarily in astrocytes.

The overall pattern of D2 mRNA expression observed in these studies agrees well with prior reports describing the distribution of D2 activity in brain (28–30). Unlike these prior studies, however, which relied exclusively on the determination of 5′-deiodinase activity in homogenates of brain regions isolated by gross dissection techniques, the in situ studies presented here allow for the much more accurate, precise, and specific localization of D2 expression.

Of importance, D2 is expressed in areas of the brain, such as the cerebral cortex, hippocampus, caudate, and hypothalamus, which are known to express nuclear T3 receptors and to be targets of thyroid hormone action during development (3, 4). For example, congenital hypothyroidism in the rat results in marked alterations in the cytoarchitecture of the cerebral cortex (31) as well as alterations in expression of specific genes (8). The latter abnormality is also observed in the caudate and the hippocampus, where thyroid hormone controls expression of the RC3/neurogranin gene (21), whose protein product appears to be important in long-term potentiation and memory (32). Cerebral cortical and basal ganglia abnormalities are also observed in the human cretin, where mental retardation and alterations of posture and movement are observed (32). The expression of D2 in auditory relay nuclei such as the medial geniculate, the olivary, and the ventral cochlear nuclei suggests that these regions are also thyroid hormone responsive, a finding that may be relevant to the deafness observed in congenital hypothyroidism.

Previous studies using immunohistochemistry have reported that T3 receptors are located in neurons and oligodendrocytes but not in astrocytes (5–7). The finding that perinatal hypothyroidism does not affect the expression of astrocytic marker genes provides additional evidence favoring the view that astrocytes are not target cells of thyroid hormone (33). Thus, it was surprising to find that D2 is expressed primarily in specific subpopulations of glial cells, namely the protoplasmic astrocytes and tanycytes, whereas little expression appears to occur in neurons or fibrous astrocytes of the white matter. Several of our observations support this conclusion. For example, cell layers containing primarily astrocytes (e.g., layer 1 of the cerebral cortex) demonstrated strong D2 mRNA expression, whereas layers containing primarily neurons (e.g., the granular layer of the dentate gyrus) were devoid of staining (Fig. 3). Furthermore, colocalization studies demonstrated directly the presence of D2 mRNA in GFAP-positive astrocytes and tanycytes, with the majority of the signal being located in the cellular processes. This latter finding is difficult to demonstrate on some photomicrographs, given that many of the grains are located some distance from the cell bodies in delicate end processes that have been cut in cross-section and lie in different planes. However, when viewed firsthand by microscopy, which allows the processes to be traced through the 25-μm-thick sections by altering the focus, it is readily apparent that the grains are contained within these astrocytic projections. These observations suggest a unique and important role for glial cells in thyroid hormone homeostasis in the brain.

Astrocytes serve important functions in neural biology (34, 35). Astrocytic processes ensheath synapses and thereby influence synapse formation and synaptic transmission. Glial cell processes also surround capillaries and participate in the transfer of nutrients between the blood and neurons. For example, glucose, the main metabolic fuel used by neurons, is actually taken up from the blood by astrocytes and processed to lactate, which is then released for neuronal use (34). Glial cells also play a role in maintaining brain glutamate concentrations (36).

A similar functional coupling between astrocytes and neurons may be operative with regard to T3 production. According to this postulate, T4, which enters the developing brain more easily than T3 (12), would be taken up from the capillaries by astrocytes, deiodinated to T3, and released. This T3 would then be available to enter neurons and interact with T3 receptors to regulate specific neuronal functions.

Anatomic studies have demonstrated that both tanycyte processes and the apical blebs that extend into the third ventricle contain ribosomes and smooth endoplasmic reticulum (24, 37). Thus, the localization of the D2 mRNA to astroglial processes and blebs suggests that this microsomal enzyme is translated and localized within these cellular compartments. Previous studies by Dratman and Crutchfield (38) support the possibility that T4-to-T3 conversion may occur within astrocytic processes, with the T3 transported to the end-feet and released. These investigators demonstrated that T3 derived from T4 in the brain is concentrated rapidly in synaptosomes in which the T3:T4 ratio is 3-fold greater than in cytosol.

The pattern of D2 mRNA in the hypothalamus strongly suggests that this enzyme is expressed primarily in ependymal cells, termed tanycytes. These cells are characterized by extensive processes that extend into all areas and cell groups of the medial hypothalamus and encircle capillaries and abut on neurons (39). Tanycytes have been implicated in the uptake and transport of hormones and other substances into the hypothalamus from the cerebral spinal fluid and the blood (40, 41). For example, the transport of T4 by ependymal cells from the third ventricle into the median eminence and the portal vessels has been demonstrated both in vivo and in organ culture systems (42).

The finding of D2 expression in tanycytes suggests that these cells may not only take up T4, from the cerebrospinal fluid and/or capillaries, but also convert it to the active hormone T3 for use in various regions of the hypothalamus. This would explain the apparent paradox that the paraventricular nucleus, which contains thyrotropin-releasing hormone-secreting neurons responsive to feedback by T4, as well as T3 (43), expresses very little D2 activity (30). The deiodinating and transport capacity of tanycytes thus may represent the “alternative source of T4 monodeiodinating activity” postulated to exist by Lechan and Kakucska (44). Uptake of T4 from the cerebrospinal fluid, with its subsequent deiodination to T3 and transport to the median eminence and the portal system, could also influence pituitary function. Alternatively, T4 could be taken up from the fenestrated vessels of the median eminence, deiodinated, and the T3 transported to the cerebrospinal fluid. Indeed, bidirectional transport of substances by tanycytes between the third ventricle and the vascular compartment has been demonstrated (45).

To date, studies localizing D1 and D3 expression to specific cell types in the brain have not been reported. The D1 is also a 5′-deiodinase that converts T4 to T3 and, if present in neurons, could be an alternative source of T3 for these cells. However, only limited information is available in the literature on the pattern of 5′-deiodinase expression in the developing rat brain (14, 29), and the techniques employed in these studies have not allowed the accurate determination of D1 activity. However, ongoing studies have demonstrated that D1 activity, though detectable, is minimal in the cerebral cortex of neonatal rats (J. Bates and V. A. Galton, personal communication). This observation, along with the known poor efficiency with which D1 utilizes T4 (as compared with reverse T3) as a substrate for 5′-deiodination, makes it unlikely that this isoform contributes significantly to the neuronal pool of T3 in the neonatal brain.

In summary, our observations suggest that astroglial cells play a central role in thyroid hormone homeostasis in the developing central nervous system and may modulate neuronal and oligodendrocyte function through the production and release of T3. This concept raises the intriguing possibility that neuronal processes might influence astroglial D2 activity through local regulatory mechanisms. In this context, the demonstrated effects of adrenergic agents and other substances to stimulate D2 activity in primary cultures of rat neonatal astroglial cells may be of relevance (46–48).

This work was supported in part by grants from Dirección General de Investigación Cientifica y Técnica (PM95-0019), the Ramon Areces Foundation, and the National Institutes of Health (DK-42271 and HD-09020). We thank Fernando Nuñez and Pablo Señor for taking care of the animals and Gloria Chacón for technical help.