The hypothalamic–pituitary–thyroid (HPT) axis maintains circulating thyroid hormone levels in a narrow physiological range. As axons containing thyrotropin-releasing hormone (TRH) terminate on hypothalamic tanycytes, these specialized glial cells have been suggested to influence the activity of the HPT axis, but their exact role remained enigmatic. Here, we demonstrate that stimulation of the TRH receptor 1 increases intracellular calcium in tanycytes of the median eminence via Gαq/11 proteins. Activation of Gαq/11 pathways increases the size of tanycyte endfeet that shield pituitary vessels and induces the activity of the TRH-degrading ectoenzyme. Both mechanisms may limit the TRH release to the pituitary. Indeed, blocking TRH signaling in tanycytes by deleting Gαq/11 proteins in vivo enhances the response of the HPT axis to the chemogenetic activation of TRH neurons. In conclusion, we identify new TRH- and Gαq/11-dependent mechanisms in the median eminence by which tanycytes control the activity of the HPT axis.
The hypothalamic–pituitary–thyroid (HPT) axis modulates key physiological processes including brain development, cardiovascular function, basal energy metabolism, and the regulation of body temperature1. Disturbances of the axis are frequent, affecting almost 6% of the US population and 10.5% of Europeans, and can have a severe impact on health2, 3. To ensure a tight control of thyroid hormones, the HPT axis is regulated at multiple points.
Thyrotropin-releasing hormone (TRH) is produced by hypophysiotropic neurons that are located in the paraventricular nucleus (PVN)4, 5 and project into the median eminence (ME) of the hypothalamus6. In the ME, TRH is released into portal blood vessels and stimulates thyroid-stimulating hormone (TSH) secretion from the pituitary. As a circumventricular organ, the ME is characterized by an open blood–brain barrier that serves as an interface between the neural and peripheral endocrine systems. In the ME a specialized cell population, the tanycytes, has been suggested to control the crossing of blood-borne substances from the periphery into the brain7 and to function as chemosensors8,9,10. The cell bodies of tanycytes, which are connected by tight junctions, are located in the ependymal layer of the 3rd ventricle and contact the cerebrospinal fluid7. Tanycytes send long processes into the parenchyma and are classified as α- and β-subtypes according to their location in the wall of the 3rd ventricle and the direction of their projections11. While α-tanycytes reside dorsally, β-tanycytes occupy the ventral sidewall of the 3rd ventricle and line the floor of the 3rd ventricle in the ME11. Processes of the latter reach to the portal blood vessels in the ME, where their perivascular endfeet are closely associated with axon terminals containing releasing hormones, such as TRH12. TRH-containing neurons even form synaptoid contacts on tanycyte processes13.
The strategic localization of tanycytes has suggested their involvement in neuroendocrine circuits. Indeed, tanycytes express the thyroid hormone transporters monocarboxylate transporter 8 (Mct8, Slc16a2) and organic anion transporting polypeptide family member 1C1 (Oatp1c1, Slco1c1)14 as well as iodothyronine deiodinase type 215 (Dio2) that converts thyroxine (T4) into 3,3′,5-triiodothyronine (T3). In addition, tanycytes express the highly specific TRH-degrading ectoenzyme (Trhde, pyroglutamyl peptidase II)12. However, whether tanycytes regulate the HPT axis in vivo is unknown.
Here we report that tanycytes control the HPT axis in vivo. Activation of the TRH receptor 1 (TRHR1) elevates the intracellular calcium concentration ([Ca2+]i) in β-tanycytes of the ME through a Gαq/11-coupled pathway. Gαq/11 signaling in tanycytes leads to an increase in the size of tanycyte endfeet and an upregulation of TRH-DE that might block hormone secretion into the portal blood vessels. Interrupting the tanycytic response to TRH by deleting Gαq/11 proteins specifically in tanycytes of mice (Gα q/11 tanKO) reduces the activity of the hypothalamic TRH-DE as well as the TRH receptor-mediated change in tanycyte endfoot size. In the absence of TRH signaling in tanycytes, TSH release is increased after activating TRH neurons. Overall, our data demonstrate an important role of tanycytes in the fast regulation of the HPT axis.
TRH receptor agonists stimulate [Ca2+]i in β-tanycytes
To target tanycytes, we injected DNA vectors based on recombinant adeno associated virus (rAAV) 1/2 into the lateral ventricle of mice. The serotypes 1 and 2 are known to hardly diffuse into the parenchyma16, 17. After injecting the rAAV vector expressing the Cre recombinase into the lateral ventricle of Ai14 reporter mice, we observed expression of the reporter tdTomato in the ependymal layer suggesting that the rAAV1/2-based vector was trapped in the ependymal layer of the ventricular system (Supplementary Fig. 1a–d).
Using this approach we measured [Ca2+]i in acute brain slices. We expressed the calcium sensor GCamP6s18 in tanycytes by injecting AAV-CAG-GCamP6s into the lateral ventricle of mice. Stimulation of acute brain slices with TRH selectively increased [Ca2+]i in β-tanycytes of the ME but not in α-tanycytes (Fig. 1a). Taltirelin, a TRH receptor agonist, had the same effect (Fig. 1b; Supplementary Movie 1). Imaging by confocal microscopy showed that in response to taltirelin [Ca2+]i rose in cell bodies and in the projections of individual β-tanycytes but not in α-tanycytes (Fig. 1c, d). In contrast, ATP elevated [Ca2+]i in both α- and β-tanycytes (Fig. 1c, d).
Cotreatment with the TRH receptor antagonist midazolam19 inhibited the [Ca2+]i increase after activation of TRH receptors in a reversible manner (Fig. 2a; Supplementary Movies 2, 3) but had no effect on the ATP-induced response (Fig. 2b). Because taltirelin and midazolam act on both TRHR1 and TRH receptor 2 (TRHR2)20, we investigated knockout mouse lines for both subtypes. In Trhr1−/− mice taltirelin lost its stimulatory effect on [Ca2+]i (Fig. 2a, c), whereas ATP as positive control8, 9 was still effective in increasing [Ca2+]i in α- and β-tanycytes (Fig. 2b, c). In contrast, in Trhr2−/− mice taltirelin stimulated [Ca2+]i increase as in wild-type animals (Fig. 2a). After laser microdissection of α- and β-tanycytes, we detected Trhr1 messenger RNA (mRNA) in β-tanycytes but not in α-tanycytes (Supplementary Fig. 2) in accordance with in situ hybridization data of the Allan Brain Institute (experiment: 79591633)21.
TRHR1 is coupled to Gαq/11 proteins and activates phospholipase C (PLC) and inositol-1,4,5-trisphosphate (IP3) production22. To investigate the involvement of this pathway in the increase of [Ca2+]i we deleted Gαq and Gα11 proteins in glial cells and tanycytes with the help of the Cre-driver mouse line GlastCreERT2 (Gα q/11 gliaKO mice; for an overview about mouse models, see Supplementary Table 1). Double deficiency of Gαq and Gα11 is usually required to investigate the physiological role of these two closely related G proteins as they are able to compensate for each other23. In Gα q/11 gliaKO mice, the [Ca2+]i response after stimulation of the TRH receptors was reduced (Fig. 2d). Moreover, pretreatment of acute brain slices with the PLC inhibitor U73122 or with the IP3-receptor antagonist 2-APB blocked the taltirelin effect (Fig. 2d–f). Thus, TRH and its analog taltirelin are able to stimulate β-tanycytes through TRHR1- and Gαq/11-mediated signal transduction.
Gαq/11 signaling modulates the morphology of tanycytes
To evaluate the cellular consequences of Gαq/11 signaling, we adopted a technique that allows both to chemogenetically activate Gαq/11 signaling and to image tanycytes. For this purpose, we expressed a fusion protein of the mutant human muscarinergic receptor 3 (hM3D) and mCherry24 in tancycytes. Our previous data confirmed that rAAV 1/2-based vectors get trapped in the ventricle wall (Supplementary Fig. 1). To further enhance selectivity and to limit expression in the ventricle wall to tanycytes, we put the vector under control of the astrocyte and tanycyte-specific GlastCreERT2 driver gene25, 26. Therefore, we injected the Cre-dependent AAV-CAG-flex(hM3D-mCherry) vector into the lateral ventricle of GlastCreERT2 mice (gTan3D). After inducing recombination by injecting tamoxifen, hM3D-mCherry expression could be detected in tanycytes of the hypothalamus but in hardly any other brain area (Fig. 3a, b; Supplementary Fig. 3; Supplementary Movie 4). This technique allowed us to visualize the processes and endfoot structure of tanycytes in the ME by high-magnification confocal microscopy (Fig. 3b). Costainings of collagen IV as a vessel marker (Fig. 3a, b), of aquaporin 4 as an astrocytic marker (Fig. 3c), and of neurofilament 200 as a neuronal marker (Fig. 3d) clearly showed that tanycytes in the ventricular wall send processes through the ME and terminate on blood vessels on the ventral surface of the ME (Fig. 3c–e). There was no overlap of the transduced cells with astrocytes or neurons in the ME (Fig. 3c, d).
hM3D can be activated by clozapine-N-oxide (CNO) treatment and is coupled to Gαq/11 signaling24. After loading acute brain slices with Fura-2 to measure tanycytic [Ca2+]i, we confirmed that activation of the fusion protein hM3D-mCherry by CNO and subsequently of Gαq/11 signaling enhanced [Ca2+]i in tanycytes of gTan3D mice (Fig. 3f). As expected, CNO had no effect on tanycytes that did not express hM3D (gTanCon; Fig. 3f). When we activated Gαq/11 signaling in tanycytes by treating gTan3D mice with CNO, the diameter of tanycytic endfeet was increased (Fig. 3g–i).
Tanycytic Gα q/11 knockout mice
To explore the in vivo consequences of the Gαq/11-mediated increase in endfoot size, we aimed to delete Gαq/11 proteins selectively in tanycytes. For a conditional knockout approach, the GlastCreERT2 mouse line would not have been suitable because of its Cre activity in astrocytes and tanycytes. Instead, we used the promoter of the Dio2 gene that is mainly expressed in tanycytes27 to drive expression from a rAAV-based vector. After injecting AAV-Dio2-iCre-2A-GFP into Ai14 reporter mice, tdTomato was expressed in cells lining the ventral 3rd ventricle and extending processes into the parenchyma, the typical morphology of tanycytes (Fig. 4a). Costainings revealed that tdTomato-positive cells expressed vimentin, a tanycytic marker28 and contained the Cre recombinase in the nucleus (Fig. 4b). Apart from tanycytes in the ventral wall of the 3rd ventricle, tdTomato-positive cells were also detectable in the subfornical organ and subcommissural organ (SCO, Supplementary Fig. 4a, e). Only a small number of transduced cells were found in the walls of the lateral or the dorsal 3rd ventricle and the choroid plexus, when we used the tanycyte-specific vector AAV-Dio2-iCre-2A-GFP (Supplementary Fig. 4b, c). When we injected AAV-Dio2-iCre-2A-GFP together with AAV-CAG-flex(hM3D-mCherry) as reporter and counted the cells showing signs of recombination, we found that mainly tanycytes were affected (Supplementary Fig. 4f and g). Thus, injecting AAV-Dio2 based vectors into the lateral ventricle allowed us to manipulate tanycytes in vivo.
By injecting AAV-Dio2-iCre-2A-GFP into the lateral ventricle of Gα q fl/fl::Gα 11 −/− mice, we deleted Gαq/11 proteins in tanycytes (Fig. 4c; Gα q/11 tanKO). As controls we used littermates (Gα q fl/fl::Gα 11 +/−) that received AAV-Dio2-GFP (Gα q/11 Con). Four weeks after injecting the vectors, plasma concentrations of TSH, T4, and T3 were determined and revealed no changes in Gα q/11 tanKO compared to Gα q/11 Con mice (Fig. 4d–f). Tanycytes express TRH-DE that degrades TRH12. When TRH signaling was blocked in tanycytes of Gα q/11 tanKO mice, TRH-DE activity was decreased in the mediobasal hypothalamus (Fig. 4g). Down regulation of TRH-DE would impair the degradation of TRH in the ME. However, basal TSH plasma levels were normal in Gα q/11 tanKO mice suggesting that a chronic deficit in TRH degradation by tanycytes can be compensated by multiple feedback mechanisms in the HPT axis.
Acutely stimulating TRH receptors by treating control mice with taltirelin enhanced the enzymatic activity of TRH-DE in the mediobasal hypothalamus (Supplementary Fig. 5a). Furthermore, when we labeled tanycytes with hM3D-mCherry by injecting AAV-Dio2-iCre-2A-GFP plus AAV-CAG-flex(hM3D-mCherry) into the lateral ventricle of C57Bl/6 mice (dTan3D-Bl6, Fig. 5), taltirelin administration increased the endfoot size very much as activation of Gαq/11 signaling by CNO did (Figs. 3g–i, 5). However, endfoot expansion in response to taltirelin was lost, when Gαq/11 signaling was interrupted in tanycytes (dTan3D-Gα q/11 tanKO Fig. 5) confirming that TRH receptor activation and Gαq/11 signaling increase the size of tanycyte endfeet. These data suggest that tanycytic degradation of TRH and blocking of its release from the ME by endfoot expansion could be relevant during acute hypothalamic stimulation of the HPT axis.
Chemogenetic stimulation of TRH release
To investigate the consequences of an acute endogenous TRH pulse, we aimed at a selective chemogenetic stimulation of hypophysiotropic TRH neurons that terminate on tanycytes13. For this purpose, we injected the vector AAV-TRH-hM3D-mCherry, in which the expression of the fusion protein hM3D-mCherry was under transcriptional control of the rat TRH promoter29, bilaterally into the PVN of C57Bl/6 mice (PVN3D-Bl6). After 2 weeks, hM3D-mCherry could be detected in Trh mRNA expressing neurons of the PVN (Fig. 6a–d). Stimulating hM3D by CNO increased cFOS expression in the PVN (Supplementary Fig. 6) confirming the activation of TRH neurons. In parallel, plasma concentrations of TSH were elevated (Fig. 6e). As expected, activation of TRH neurons did not elevate TSH levels in Trhr1−/− mice (Fig. 6e). Chemogenetic activation of TRH neurons by CNO administration also enhanced pituitary Fos mRNA expression in wild-type but not in Trhr1−/− mice (Fig. 6f).
While most TRH neurons were expressing hM3D-mCherry, some transduced cells were not positive for Trh mRNA (Fig. 6a–d). This may be due to an incomplete labeling of TRH neurons by in situ hybridization or off-target activity of the TRH promoter fragment in non-TRH neurons. To evaluate possible off-target effects of our chemogenetic strategy on other neuroendocrine axes mediated by the PVN, we analyzed the localization of the transduced hM3D-mCherry-positive fibers in the ME. hM3D-mCherry-positive terminals overlapped with preproTRH (ppTRH; Fig. 6g) and corticotropin-releasing hormone (CRH; Fig. 6h) in the ventral layer of the ME. This is in accordance with reports showing TRH- and CRH-positive fibers in close proximity to the fenestrated capillaries of the portal venous system12, 30. However, CNO administration only elevated TSH (Fig. 6i) but not adrenocorticotropic hormone (ACTH, Fig. 6j) plasma concentrations after 2 h. As described previously vasopressin-31, oxytocin-, and gonadotropin-releasing hormone (GnRH)-positive32 fibers terminated in a more dorsal part of the ME and were not colocalized with hM3D-mCherry-positive fibers (Fig. 6k–m).
Tanycytic Gα q/11 knockout influences the HPT axis activity
To test whether an acute stimulation of the HPT axis is influenced by the tanycyte-specific deletion of Gαq/11 signaling, we activated hypophysiotropic TRH neurons in mice with tanycytic Gα q/11 knockout. For this purpose, we transduced the PVN of Gα q/11 tanKO and Gα q/11 Con animals with AAV-TRH-hM3D-mCherry to generate PVN3D-Gα q/11 tanKO and PVN3D-Gα q/11 Con mice. Gα q/11 Con mice without transduction of the PVN were treated with CNO as control group (PVNCon-Gα q/11 Con; Fig. 7a). As shown above (Fig. 6i), activation of TRH neurons by administering CNO to PVN3D-Gα q/11 Con mice increased TSH plasma levels compared to CNO-treated PVNCon-Gα q/11 Con mice (Fig. 7b). A further rise in plasma TSH was observed after activation of TRH neurons in PVN3D-Gα q/11 tanKO mice, in which TRH signaling in tanycytes was blocked (Fig. 7b). In parallel, we found slightly higher T4 and T3 plasma concentrations in PVN3D-Gα q/11 tanKO animals than in CNO-treated PVNCon-Gα q/11 Con mice (Fig. 7c, d). Activation of TRH neurons also resulted in elevated pituitary Tshb mRNA expression in PVN3D-Gα q/11 tanKO mice (Fig. 7e). Thus, acute activation of TRH neurons was only effective on T4 plasma levels and Tshb mRNA expression in the pituitary, when tanycyte signaling was inhibited. To evaluate whether the increase in circulating TSH protein is accompanied by a rise in hormone activity, we measured TSH activity by determining cAMP production in JP26-CHO cells stably expressing TSH receptors33. When stimulating TRH neurons, the absence of TRH signaling in tanycytes (PVN3D-Gα q/11 tanKO mice) significantly increased plasma TSH activity in comparison to animals with intact tancyte signaling (PVN3D-Gα q/11 Con mice, Fig. 7f). The TSH plasma concentrations and activities correlated (Fig. 7g). Importantly, activation of TRH neurons had a more pronounced effect on the expression of the T3-dependent genes Thrsp and Fasn in the liver, when TRH signaling in tanycytes was blocked (PVN3D-Gαq/11tanKO; Fig. 7h, i) than under control conditions (PVN3D-Gαq/11Con), supporting the metabolic significance of tanycyte function.
To evaluate whether the tanycytic defect of Gαq/11 signaling would affect the pituitary response to TRH stimulation, we treated Gα q/11 tanKO mice with the TRH analog taltirelin and measured TSH plasma levels. TSH plasma levels were not different between mice with and without Gαq/11 proteins in tanycytes (Supplementary Fig. 5b) arguing against an altered TRH responsiveness of the pituitary. Thus, TRH signaling in tanycytes inhibits the HPT axis, probably at the level of TRH release from the ME.
In the HPT axis, negative feedback mechanisms are important to stabilize thyroid hormone concentrations within the physiological range. According to previous concepts, thyroid hormones and TSH exert a negative feedback at the level of the PVN34 and the pituitary35 by inhibiting the expression of Trh and Tshb. Tanycytes in the ME now emerge as new players in the fast regulation of the HPT axis. They represent specialized glial cells in the ependymal layer of the medial basal hypothalamus36 and in other circumventricular organs28. Hypothalamic tanycytes can be divided in subgroups: α-tanycytes, which project into hypothalamic nuclei, and β-tanycytes, which project into the ME36. The projections of both subtypes end mainly on blood vessels37.
Recent evidence suggests that tanycytes play a role as biosensors for the fast detection of hormones and transmitters. It has been shown that tanycytes respond with a fast increase in [Ca2+]i to ATP, glucose8, 9, histamine, acetylcholine8, and sweeteners10. So far, no other stimuli have been reported to increase [Ca2+]i in tanycytes. Here, we show that TRH and the TRH receptor agonist taltirelin selectively increase [Ca2+]i in β-tanycytes of the ME. This effect is mediated by TRHR1, as we detected Trhr1 mRNA expression in β-tanycytes of the ME and the response was lost in Trhr1−/− but not in Trhr2−/− mice. TRHR1 is known to signal through a Gαq/11-coupled pathway22. Blocking this pathway either by pharmacological tools or by deleting Gαq/11 proteins inhibited the elevation of [Ca2+]i in response to TRHR1 stimulation. These experiments revealed a selective and fast response of β-tanycytes to TRH and led to the hypothesis that tanycytes could modulate the HPT axis via a TRH-mediated mechanism that controls the release of TRH from the hypothalamus.
To test this concept in vivo, we had to establish a tool that would allow selective activation of hypophysiotropic TRH neurons in the PVN of the hypothalamus. By activating the HPT axis at the level of the PVN, we avoided pituitary effects of exogenous TRH agonists that could override hypothalamic mechanisms. For chemogenetic activation of TRH neurons, we expressed the hM3D receptor fused to mCherry24 in the PVN under control of the TRH promoter29. In situ hybridization identified transduced hM3D-mCherry-positive cells as TRH neurons. Projections of the transduced neurons were visible as hM3D-mCherry-positive fibers in the ventral part of the ME where TRH-6 and CRH-positive fibers terminate. The specific activation of TRH neurons is supported by the observation that CNO treatment elevated TSH plasma levels but had no effect on ACTH concentrations. Activation of TRH neurons also led to a fast TRHR1-dependent Fos mRNA expression in the pituitary, as has been observed after TRH stimulation in vivo38.
To investigate the physiological function of tanycyte activation in vivo, we developed techniques that allowed for a selective and efficient genetic manipulation of tanycytes. Our approach of targeting tanycytes relied on two factors. First, after intraventricular injection, rAAV1/2-based vectors were trapped in the ventricle wall and did not reach the parenchyma to an appreciable extent. Second, we controlled expression by regulatory gene sequences of Dio2 or Glast that show selectivity for tanycytes and astrocytes. However, due to the administration route and serotype of vectors, only tanycytes in the ventricle wall but hardly any astrocytes in the parenchyma were transduced. For Dio2, we identified a short promoter fragment that mediated tanycyte specificity when inserted in rAAV vectors. Alternatively, we used the Cre-driver GlastCreERT2 mice25 together with a rAAV-based Flex system24, 39. This combination has the additional advantage that recombination is tamoxifen-inducible. In previous studies, GlastCreERT2 mice were reported to selectively modify α-tanycytes without obvious recombination in β-tanycytes40. However, Glast is expressed in all subtypes of tanycytes26 and we were able to recombine the Flex system and to express hM3D-mCherry also in β-tanycytes. Overall, the toolbox that we have developed will be helpful to analyze tanycyte functions in vivo. In the literature, a similar strategy has been reported that uses the protein tat-Cre41 to label tanycytes42 or induce tanycytic knockouts32. The advantage of our approach is that we were able to overexpress genes, e.g., hM3D-mCherry, specifically in tanycytes.
With the tanycyte-specific rAAV-based Cre expression, we selectively deleted Gαq/11 proteins in tanycytes to inhibit their TRH-induced activation. The tanycyte-specific deletion of Gαq/11 proteins had no effects on basal plasma levels of TSH and thyroid hormones. However, the acute activation of the hypophysotropic TRH neurons had a stronger effect on plasma TSH levels when the Gαq/11 pathway was inactivated in tanycytes. The fact that we did not see a basal elevation of thyroid hormones, when TRH signaling was inhibited in tanycytes, may partially be explained by compensatory mechanisms, which seem to be overridden by acute activation of the HPT axis.
Two mechanisms may explain the increase in TSH release after acute stimulation of TRH neurons. First, tanycytes in the ME are in close proximity to TRH-positive nerve terminals and express TRH-DE12, 13, 43. This membrane-bound enzyme degrades TRH and seems to be an important modulator of the TRH release into the portal vessels of the pituitary12, 43. TRH-DE activity depends on Gαq/11 signaling in tanycytes as the TRH-DE activity was reduced in Gα q/11 tanKO mice. The stimulation of TRH-DE activity by Gαq/11 signaling in tanycytes is in contrast to in vitro experiments in pituitary cells, where TRH reduces the TRH-DE activity44 suggesting cell-specific regulation. The reduced TRH-DE activity in Gα q/11 tanKO mice is a potential candidate to explain the increase of TSH levels after activation of TRH neurons.
A second possible mechanism relates to morphological changes in the ME. We observed a change in morphology of the tanycytic endfeet when we activated Gαq/11 signaling in tanycytes. By expressing the membrane-bound hM3D-mCherry fusion protein in tanycytes, we were able to visualize the endfoot morphology. One hour after activating the Gαq/11 pathway in tanycytes the tanycytic endfeet were larger, an effect that was abolished in Gα q/11 tanKO mice. Fast morphological changes are described in astrocytes in response to increased intracellular calcium45. In astrocytes vimentin and GFAP are phosphorylated via calcium-dependent kinases46. Both proteins are expressed in tanycytes36. Phosphorylation of these proteins leads to a change in the cytoskeleton and morphological changes in astrocytes46,47,48. Increasing of intracellular calcium waves also promotes movement of radial glia49. Thus, a change in tanycytic morphology by Gαq/11 signaling and elevated [Ca2+]i is not without precedent from glial cell biology. As for tanycytes, Parkash et al.32 showed a morphological alteration in the ME via a semaphorin 7a-dependent mechanism. This leads to a tightening of the endfoot structure of the tanycytes covering the fenestrated vessels of the ME as well as a retraction in GnRH neuronal terminals. Thereby, tanycytes in the ME seem to regulate GnRH secretion32.
Our data show that tanycytes limit the TRH release from the hypothalamus and thus control the response of the HPT axis to an acute activation of TRH neurons. Interfering with their function led to higher T4 and T3 plasma concentrations but also induced the expression of T3-dependent genes in the liver. This tanycytic mechanism may play a homeostatic function during chronic hypo- or hyperactivity of the HPT axis. Loosening the tanycytic brake during hypothyroidism or tightening it during hyperthyroidism will stabilize the axis. In addition tanycytes may play a role in shaping the pulsatile pattern of TSH secretion50. Until now it is unknown how exactly the pulsatile release of TSH is regulated, although there is evidence that the pulsatile TSH secretion pattern is mediated by TRH51. In rats, TRH is released into the portal vein system in a pulsatile manner52. A combination of morphological changes of tanycytes and regulation of the TRH-DE activity could be responsible for pulsatile release.
In summary we have identified a new mechanism by which tanycytes control the HPT axis (Fig. 8). After release from neuronal terminals in the ME, TRH activates TRHR1 and the coupled Gαq/11 pathway in tanycytes. This pathway induces structural changes in tanycytic endfeet, which increase their size. In addition, the Gαq/11 pathway in tanycytes modulates the TRH-DE activity. We suggest that by changing their morphology, tanycytes impede the vascular outflow of TRH and instead favor its degradation via TRH-DE. Thereby, tanycytes are able to modulate the HPT axis at the hypothalamic level.
All mouse lines were established on a C57Bl/6 background. We used male littermate mice that were age-matched between experimental groups. Mice were between 8 and 14 weeks of age. All animal experiments were approved by the local animal ethics committee (Ministerium für Landwirtschaft, Umwelt und ländliche Räume, Kiel, Germany). Mice were kept at constant temperature (22 °C) on a 12-h light/dark cycle and were provided with standard laboratory chow (2.98 kcal g−1; Altromin, Hannover, Germany) and water ad libitum.
Mice that carried a floxed Gα q allele (Gα q fl/fl) and were Gα 11 -deficient (Gα 11 −/−)23 (Gα q fl/fl::Gα 11 −/−) were used to knockout the Gαq/11 pathway in tanycytes by transducing tanycytes with the cell-specific AAV-Dio2-iCre-2A-GFP vector of serotype 1/2 (Gα q/11 tanKO). Gα 11 +/− (Gα q fl/fl::Gα 11 +/−) littermates that received an injection of AAV-Dio2-GFP of serotype 1/2 (Gα q/11 Con) were used as controls. To delete Gαq/11 signaling in glial cells, GlastCreERT2 mice25 were crossed with animals that carried a floxed Gα q allele (Gα q fl/fl) on a Gα 11 -deficient (Gα 11 −/−) background23 (Gα q/11 gliaKO). GlastCreERT2 mice25 were also used to express the hM3D-mCherry specifically in tanycytes. For this purpose, we injected the Cre-dependent AAV-CAG-flex(hM3D-mCherry) vector into the lateral ventricle of mice heterozygous for the Cre allele (gTan3D). To induce CreERT2 activity in transgenic animals at an age of 8–10 weeks, they were treated i.p. with 1 mg tamoxifen dissolved in 90% miglyol 812, 10% ethanol every 12 h for 5 consecutive days. Generation and phenotyping of Trhr1 knockout mice (Trhr1−/−) have been reported previously53, 54. Trhr2 knockout mice (Trhr2−/−) were obtained from Deltagen. For chemogenetic stimulation of hypophysiotropic neurons in the PVN, AAV-TRH-hM3D-mCherry was injected bilaterally into the PVN (PVN3D). To test the Cre activity of rAAV vectors, we used the reporter mouse line Ai1455. All mice were randomly allocated to treatment groups. Investigators were blinded for treatment or genotype of mice or both in all experiments and analyses. Mice were only excluded from analysis if they did not survive during surgical procedures or if no blood samples could be obtained.
Cloning of viral vectors
rAAV gene transfer vectors were generated by inserting specific gene cassettes into rAAV2-based expression vectors, containing inverted terminal repeats of serotype 256, a woodchuck posttranscriptional regulatory element (WPRE), and a bovine growth hormone polyadenylation site (bGHpA). The plasmid pAAV-Dio2-iCre-2A-GFP was generated by PCR amplifying the small upstream fragment of the human DIO2 gene (−1757 to +157, GenID: 1734) from human complimentary DNA (cDNA) and inserting it in the Xho1/Age1 site of pAAV-CMV-iCre-2A-GFP57 to replace the CMV promoter. The plasmid pAAV-TRH-hM3D-mCherry was generated by PCR amplifying of the hM3D-mCherry sequence from the plasmid pAAV-Syn-flex(hM3D-mCherry)24 and inserting it in frame in pAAV-CAG-Venus58 in order to replace Venus. In a second step, the CAG promoter was exchanged for a short TRH promoter fragment (−776 to +84; GenID: 25569)29 amplified from rat cDNA. The plasmid pAAV-CAG-flex(hM3D-mCherry) was generated by inserting the CMV enhancer chicken β-actin (CAG) promoter from pAAV-CAG-BMP2-2A-Tomato58 into the Mlu1/Sal1 site of pAAV-Syn-flex(hM3D-mCherry)24 to replace the promoter. The plasmid pAAV-CAG-GCamP6s18 was generated by inserting the CGaMP6s sequence of pGP-CMV-GCaMP6s (Plasmid #40753, Addgene) into the Nhe1/Not1 site of pAAV-CAG-Venus58 to replace the Venus sequence. After construction, all plasmids were sequenced.
Recombinant AAV production
rAAV with a mosaic capsid of serotype 1 and 2 (1:1) were generated as described57 and purified by AVB Sepharose affinity chromatography59. For each vector, the genomic titer was determined by quantitative PCR (qPCR) using primers against WPRE (WPRE forward primer: 5′-TGCCCGCTGCTGGAC-3′; WPRE reverse primer: 5′-CCGACAACACCACGGAATTG-3′) as described previously58.
Stereotaxic rAAV injections were performed according to a protocol described previously60. The following coordinates relative to bregma were used for injections: lateral ventricle, anteroposterior −0.1 mm, mediolateral −0.9 mm, dorsoventral from the skull surface −2.3 mm; PVN, anteroposterior −0.7 mm, mediolateral ±0.25 mm, dorsoventral from the skull surface −4.9 mm. Mice were anesthetized with an intraperitoneal (i.p.) injection of a mixture of ketamine hydrochloride (65 µg g−1 body weight) and xylazine (15 µg g−1 body weight) in NaCl (0.9%). After loss of reflexes, animals were fixed in a stereotaxic frame (David Kopf instruments; Nr: 1900). At the defined positions, small holes were made into the skull using a dental drill (freedom; K.1070 Micromotor Kit). Vector solution was injected into the PVN (120 nl per site) and into the lateral ventricle (2–2.5 µl) at a rate of 100 nl min−1. After injection, the micropipette was kept in place for 5 min to avoid backflow of the injected vector during micropipette retraction. The scalp was sutured, and the animal was placed on a heating pad until full recovery from surgery and then returned to its home cage. Carprofen (5 mg kg−1 body weight; s.c.) was applied once a day for 2 days after rAAV injection.
To detect changes of [Ca2+]i, tanycytes were transduced with the calcium sensor GCamP6s18 by injecting AAV-CAG-GCamP6s (1.5 × 1010 genomic particles) into the lateral ventricle of mice 2 weeks prior to the measurements. Brains were dissected from deeply anesthetized mice (ketamine hydrochloride, 65 µg g−1 body weight; xylazine, 15 µg g−1 body weight) and placed in ice-cold artificial cerebrospinal fluid (aCSF; 124 mM NaCl; 26 mM NaHCO3; 1.25 mM NaH2PO4; 3 mM KCl; 2 mM CaCl2; 1 mM MgSO4; 10 mM d-glucose). Brain slices (150–200 µm in thickness) were prepared under controlled conditions (aCSF + 10 mM MgSO4, 5% CO2, 95% O2; pH 7.4; 4 °C) using a vibrating blade microtome (Leica VT1200 S). Slices were placed in normal aCSF at 4 °C for 45 min before being transferred to an ice-cold aCSF with low glucose (1 mM glucose + 9 mM sucrose). Slices could be stored up to 5 h at 4 °C without a reduction in reactivity.
To measure [Ca2+]i in response to CNO stimulation in gTan3D or gTanCon mice, acute brain slices were loaded with Fura-2AM. Slices were incubated with Fura-2AM (12.5 μg ml−1 with 0.5% DMSO and 0.05% pluronic 127, Invitrogen) in low-glucose aCSF for 30 min at 37 °C. Then, they were incubated in aCSF containing probenecid (1.25 mM, 30 min 37 °C) and stored at 4 °C before measurement.
[Ca2+]i was measured using a high-speed calcium imaging setup (Till Photonics) mounted on the Axio Examiner D1 upright fluorescent microscope (Zeiss) coupled to the polychrome V monochromator and a high-speed CCD camera (Retiga EXi-blue, QImaging). Data acquisition and quantification was done using life acquisition and offline analysis software (FEI GmbH Munich, formerly Till Photonics). Brain slices were placed in a flow chamber. For measuring GCamP6s, an excitation wavelength of 488 nm and an emission filter (> 498 nm) were used. Changes in fluorescence (F) were normalized to basal mean values of the first 30 s (F0) of each region of interest (ROI) and were shown as (F/F0). ROIs covered the whole tanycytic layer of the ME.
For single-cell calcium imaging, we used a confocal microscope (TCS SP5, Leica) with a 488 nm laser using a × 20 objective (HCX PL APO CS 20X). Changes in fluorescence (F) were normalized to basal mean values of the first 30 s (F0) of each ROI and were shown as (F/F0). ROIs covered the cell bodies of single tanycytes of the ME or tanycytes of the ependymal layer of the 3rd ventricle. F and F0 were analyzed with Fiji.
For the ratiometric measurement of Fura-2, we used excitation wavelengths 340 and 380 nm. Changes in the fluorescence ratio of F340/F380 (ΔF) were normalized to basal mean values of the first 30 s (F0) of each ROI and were shown as (ΔF/F0). ROIs covered single tancytic cell bodies. hM3D-mCherry-positive tanycytes were identified by the red fluorescence of mCherry.
After 1 h of recovery in the measurement buffer (aCSF; CO2, 5%; O2, 95%; pH, 7.4; flow rate, 2 ml min−1), stimuli were administered into the perfusion flow via a dedicated perfusion system (TSE-systems). One minute after starting the measurement, TRH (33 µM) or taltirelin (33 µM, 150 µM for single-cell response) were administered for 15 s. ATP (33 µM, 150 µM for single-cell response, 15 s) was used as positive control.
For blocking the Gαq/11 signaling pathway, slices were stimulated once with taltirelin (33 µM, 15 s) followed by incubation with 2-APB (100 µM, Tocris) or U73122 (100 µM, Tocris) in aCSF for 30 min and afterwards a second taltirelin stimulation. As 2-APB and U73122 were dissolved in 0.5% DMSO, we treated control slices with aCSF containing 0.5% DMSO. For TRH receptor inhibition, after a first taltirelin stimulation midazolam (500 µM) was added for 20 min directly to the aCSF flow with a perfusion pump (TSE) followed by a second taltirelin stimulation (33 µM, 15 s) in addition to midazolam. After 20 min washout with aCSF (2 ml min−1), a third taltirelin stimulation was performed. For quantification the maximal F/F0 was used. For heat maps we used the Look-Up Table royal from Fiji, representing a gray sale from 0 to 255.
Chemogenetic stimulation of TRH neurons and hormone measurements
AAV-TRH-hM3D-mCherry (1.2 × 108 genomic particles) was bilaterally injected into the PVN of mice (PVN3D). Mice without injection of the PVN (PVNCon) were used as control. For the parallel knockout of the Gαq/11 pathway in tanycytes, AAV-Dio2-iCre-2A-GFP (2 × 1010 genomic partials, 2 µl) was injected into the lateral ventricle of Gα q fl/fl::Gα 11 −/− mice (PVN3D-Gα q/11 tanKO) during the same surgery. Gα q fl/fl::Gα 11 +/− mice that received an injection of AAV-Dio2-GFP virus in the lateral ventricle (PVN3D-Gα q/11 Con and PVNCon-Gα q/11 Con) were used as controls. Three weeks after rAAV injection, all three animal groups received CNO (0.5 mg kg−1 body weight; i.p.). After CNO stimulation, plasma was collected for determining TSH, T4, and T3 concentrations using a MILLIPLEX mouse kit (Millipore Corp.) and a BioPlex-system (Bio-Rad) according to the manufacturers’ instructions, while pituitary mRNA of Tshb and Fos was quantified by qPCR.
PVN3D-Trhr1−/− and PVN3D-Bl6 as well as control animals (PVNCon) were treated with CNO (0.5 mg kg−1 body weight, i.p.) 3 weeks after rAAV injection. Plasma TSH was measured 45 min or 2 h after CNO stimulation. Pituitary mRNA of Tshb and Fos was quantified by qPCR.
TRH-DE activity assay
Gα q/11 tanKO and Gα q/11 Con mice were killed by decapitation and the medial basal hypothalamus (right and left arcuate nuclei plus ME) was microdissected with a stereomicroscope using small spring scissors (FST). Tissue was homogenized in a microcentrifuge tube with an ultra-turrax in 50 mM sodium phosphate buffer (pH 7.5) followed by centrifugation for 15 min at 14,000×g. Pellets were rinsed with 1 M NaCl and resuspended in 50 mM sodium phosphate buffer (pH 7.5). Membrane TRH-DE activity was determined using the substrate TRH-β-naphthylamide (TRH-βNA, Bachem) in a coupled enzyme assay in the presence of excess dipeptidyl aminopeptidase IV (ACRO biosystems; activity: 4 nmol Gly-Pro-βNA (Bachem) hydrolyzed per min) containing 0.2 mM N-ethylmaleimide (Sigma Aldrich), 0.2 mM bacitracin (Sigma Aldrich). The enzymatic reaction was initiated by addition of 800 μM TRH-βNA to the reaction mixture (total volume 300 μl) and carried out in duplicates at 37 °C. About 50 µl was withdrawn every 30 min over 120 min. The reaction was stopped by the addition of methanol (50 μl, 100%). Aliquots were diluted 1:20 with 50% sodium phosphate buffer (50 mM, pH 7.5, containing 50% methanol). β-NA was measured in a fluorometer (excitation 335 nm; emission 410 nm). An aliquot was kept for protein quantification using the Lowry method. The β-NA concentration was calculated by a standard curve diluted in 50% sodium phosphate buffer (50 mM, pH 7.5, containing 50% methanol). The activity was linear within the first 2 h and was analyzed relative to the membrane protein content for each individual sample.
Bioassay of TSH activity
TSH activity in plasma samples was measured with a bioassay, using CHO cells stably transfected with cDNA for a human TSH receptor (JP26-CHO, a gift of Gilbert Vassart)33. Cells (1 × 104 per well in a 96-well plate) were starved for 30 min with Krebs Ringer HEPES buffer (KBH: 124 mM NaCl; 1.25 mM KH2PO4; 1.45 mM CaCl2; 1.25 mM MgSO4; 5 mM KCl; 25 mM Hepes; 8 mM d-glucose; 0.05% bovine serum albumin; pH 7.4) before they were stimulated with plasma samples (100 µl diluted 1:2 in KBH) for 1 h. At the end of the incubation time, cells were lysed by adding 100 µl 0.1 N HCl containing 0.5% Tween 20 and cAMP was measured with cAMP ELISA (Enzo) as described61.
Immunohistochemistry and analysis of endfoot morphology
To visualize tanycytic endfeet, AAV-CAG-flex(hM3D-mCherry) was injected into the lateral ventricle of GlastCreERT2 mice (Tan3D; 1 × 109 genomic particles/mouse). One week after the injection, mice were treated with tamoxifen as described above. Two weeks after vector transduction, mice were stimulated with CNO (0.5 mg kg−1 body weight; i.p.) and after 60 min perfused with 4% paraformaldehyde (PFA) during deep anesthesia (ketamine hydrochloride, 65 µg g−1 body weight; xylazine, 15 µg g−1 body weight). Brains were postfixed in 4% PFA at 4 °C overnight. Slices (50 µm) were prepared with a vibratome. Free-floating sections were washed twice in Tris-buffered saline (TBS), permeabilized with 0.3% Triton X-100 in TBS (TBS-TX) for 30 min, and incubated in bovine serum albumin (5% in TBS-TX) for 2 h. Then, primary antibodies were added followed by overnight incubation with gently shaking. Sections were washed twice in TBS-TX for 10 min and incubated for 2 h with secondary antibodies. After washing twice in TBS, sections were mounted on glass slides and covered with Mowiol. Endfoot sizes were analyzed using a confocal microscope (SP5, Leica; objective, HCX PL APO CS 63X oil UV corrected; aperture, 1.4; microscopic zoom, 5 to ×13; scanning frequency, 100 Hz; average: 4 times and pinhole, 0.5 AU with a z-stack over 20 µm with a step size of 0.5 µm). Pictures were analyzed by Fiji. Per animal 15–32 endfeet in five areas of three slices were analyzed. For costainings with anti-vimentin mice were transcardially perfused with 4% PFA. Subsequently, the brains were snap-frozen in 2-methylbutane on dry ice and were sectioned with a cryostat. Staining was performed using the protocol described above. Primary antibodies: mCherry (1:1,000, Acris #AP32117-PU-N), vimentin (1:250, Thermo Fisher #PA1-16759), 2A (1:1000, Millipore #ABS31), collagen IV (1:1000, Abcam #ab6586), ppTRH160–169 (1:2000, gift from M. Wessendorf), GnRH (1:500, immunoStar #20075), vasopressin (1:500, immunoStar #20069), corticotropin-releasing hormone (1:500, immunoStar #20084, CRH), oxytocin (1:500, immunoStar #20064), aquaporin 4 (1:1000, Millipore #AB3594), cFos (1:500, Abcam #ab7963), and neurofilament 200 (1:500, Sigma Aldrich #N-4142). Secondary antibodies: anti-rabbit alexa 488 (1:500, Invitrogen #A-21206) anti-chicken Cy5 (1:500, Abcam #ab97147), anti-goat Cy3 (1:500, Jackson ImmunoResearch #705-165-147). DAPI was used in a concentration of 1 µg ml−1.
In other experiments, we injected AAV-CAG-flex(hM3D-mCherry) and AAV-Dio2-iCre-2A-GFP (2 × 1010 genomic particles of each vector, 2.5 µl) into the lateral ventricle of Gα q fl/fl::Gα 11 −/− mice to generate dTan3D-Gα q/11 tanKO and of C57Bl/6 mice to generate dTan3D-Bl6 mice. dTan3D-Bl6 mice were used as controls. Two weeks later, mice were treated with taltirelin (1 mg kg−1 body weight; i.p.), CNO (0.5 mg kg−1 body weight; i.p.), or saline (5 µl g−1 body weight; i.p) and brains were dissected and analyzed as described above.
In situ hybridization
To determine the localization of Trh mRNA expressing neurons and hM3D-mCherry in PVN3D mice, free-floating in situ hybridization was performed. Animals were killed in deep anesthesia (ketamine hydrochloride, 65 µg g−1 body weight; xylazine, 15 µg g−1 body weight) and brains were postfixed in 4% PFA at 4 °C overnight. Sections (50 µm) were prepared with a vibratome and hybridized with a digoxigenin-labeled anti-Trh RNA probe62 (530 bp, bases 202–733 of NM_009426) overnight. Hybridization was detected with anti-digoxigenin-POD Fab-fragment antibodies (Roche) at a dilution of 1:100 overnight. The POD signal was increased with the Tyramin Cy5 amplification kit (PerkinElmer, 1:100). After in situ hybridization, hM3D-mCherry was detected by immunohistochemistry with anti-mCherry antibodies (1:500, Acris #AP32117-PU-N).
RNA was isolated from pituitary glands and liver by using the Qiagen Nucleospin 96 kit according to the manufacturer’s instructions. First strand cDNA was synthesized using oligo-(dT)15 primer and AMV Reverse Transcriptase (RT; Invitrogen). cDNA was stored at −20 °C until further analysis. The following primers were used for quantitative real-time PCR (qPCR): Tshb forward, 5′-CCGCACCATGTTACTCCTTA-3′, Tshb reverse, 5′-GTTCTGACAGCCTCGTGTAT-3′, PCR product 104 bp; Fos forward, 5′-CTTTCCCCAAACTTCGACCA-3′, Fos reverse, 5′-TCGTAGTCGGCGTTGAAACC-3′, PCR product 51 bp; Gapdh forward, 5′-ATGTGTCCGTCGTGGATCTGA-3′, Gapdh reverse, 5′-TGAAGTCGCAGGAGACAACCT-3′, PCR product 144 bp; Thrsp forward, 5′-CTTACCCACCTGACCCAGAA-3′, Thrsp reverse 5′-CATCGTCTTCCCTCTCGTGT-3′ PCR product 120 bp; Fasn forward, 5′-GGAGGTGGTGATAGCCGGTAT-3′, Fasn reverse 5′-TGGGTAATCCATAGAGCCCAG-3′, PCR product 140 bp. qPCR was performed according to the following protocol: 2 min at 50 °C, 2 min at 95 °C, 15 s at 95 °C, and 1 min at 60 °C (40 cycles). Amplification was quantified using Platinum SYBR Green qPCR SuperMix (Invitrogen). Quantified results were normalized to Gapdh using the ΔΔCt method.
Cryosections of C57Bl/6 mice (20 μm in thickness) were mounted on glass slides covered with a membrane of polyethylene naphthalate (1.35 μm in thickness, Zeiss, Germany) and stored at −80 °C. Alpha- and β-tanycytes were microdissected using a system with a pulsed 337-nm UV laser (PALM MicroBeam; PALM Microlaser Technologies) and the laser-pressure catapulting mode. RNA was isolated with Absolutely RNA Nanoprep Kit (Agilent). cDNA was amplified as described for qPCR. A two-step PCR was performed (20 cycles followed by 30 cycles) with the following primers: Trhr1 forward, 5′-CAGCACCTACAAAAACGCTG-3′; R-Trhr1 reverse, 5′-CTTTATGGCCTTCTTTAGTTCTCAG-3′; fragment size: 693 bp; Actb forward, 5′-ATGGAATCCTGTGGCATCCAT-3′; Actb reverse, 5′-TTCTGCATCCTGTCAGCAATG-3′; fragment size: 140 bp.
All analyses were performed using Prism 5 (GraphPad Software). Data were assessed for normality (Shapiro–Wilk test) and equality of variances (Bartlett’s test or Levene test). If the raw data did not satisfy these conditions, a non-parametric method was used. Data of two groups were compared by a two-tailed, unpaired, parametric Student’s t-test. For parametric multiple comparisons, we used one- or two-way ANOVA with Bonferroni post hoc test. Kruskal–Wallis with post hoc Dunn’s test was used for non-parametric multiple comparisons analysis. The significance level was set at p < 0.05. All data are shown as means ± S.E.M. G-Power (Version 3.1.8) was used to analyze power and group size before the experiments.
All the relevant data are available from the authors upon request.
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
We thank Ines Stölting, Andreas Makurat for technical assistance, Magdalena Goetz for providing the GlastCreERT2 mouse line, Martin Wessendorf for providing us the ppTRH-Antibody, Gilbert Vassart for the JP26-CHO cell line. We also thank Rolf Sprengel for providing the plasmid pAAV-CMV-iCre-2A-GFP and pAAV-CAG-BMP2-2A-Tomato as well as Bryan Roth for the pAAV-Syn-hM3D-mCherry plasmid. This work was supported by grants from the Deutsche Forschungsgemeinschaft (SPP1629 “Thyroid Trans Act”, MU 3743/1-1 to HMF; GRK1957 “Adipocyte-Brain Crosstalk”, and TRR134 to MSch).
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.