IV.  Functional Anatomy of Hypothalamic Homeostatic Systems

A.      Regulation of Hypophysiotropic Neurons

    The secretion of hypothalamic releasing and inhibitory hormones from axon terminals of tuberoinfundibular neurons into the portal capillary system is dependent upon several layers of control that can be exerted directly on the perikarya and/or processes of these neurons.  For one, neurons of the tuberoinfundibular system can be modulated by substances circulating in the bloodstream that either pass the blood-brain barrier because they are fat soluble steroids or small molecules, or access tuberoinfundibular neurons via the cerebrospinal fluid due to the periventricular location of many tuberoinfundibular neurons and poor development of tight junctions between ependymal cells in these regions (53).  Feedback effects of thyroid hormone, for example, occur directly on TRH-producing neurons within the paraventricular nucleus as demonstrated by the ability of a microcrystalline implant of T3 adjacent to the paraventricular nucleus to prevent the hypothyroid-induced increase in TRH biosynthesis on that side but not the opposite side (181,182).  In addition, tuberoinfundibular neurons receive numerous axosomatic and/or axodendritic contacts from local interneurons and/or other regions in the brain that contain a variety of chemical messengers that contribute to intercommunication between specific neuronal groups or are important in establishing the set point at which the hypophysiotropic substances are secreted in response to hormonal feedback signals.  To demonstrate how the CNS can exert regulatory control over hypophysiotropic neurons, examples of modulation of GH secretion and regulation of the hypothalamic-pituitary-adrenal (HPA), hypothalamic-pituitary-thyroid (HPT), and reproductive axes will be given below.

1.  Modulation of GHRH/SRIF Tuberoinfundibular Neurons
A well-studied example of local afferent influences on the activity of tuberoinfundibular neurons is demonstrated by the hypothalamic regulatory system involved in the control of GH secretion.  The pattern of GH secretion is episodic, showing a regular periodicity of one pulse every 2 to 4 hours and low or undetectable trough values (183).  This rhythm is the result of the control by two separate components of the tuberoinfundibular system, including GHRH-producing neurons (stimulatory) in the basolateral portion of the arcuate nucleus and somatostatin-producing neurons (inhibitory) in the periventricular nucleus, each secreting into the portal capillary plexus.  To coordinate this rhythmic secretion, reciprocal axonal connections between these two populations of neurons may be necessary (Fig. 28).  In this manner, somatostatin neurons receive direct, stimulatory inputs from GHRH neurons while GHRH neurons receive direct, inhibitory inputs from somatostatin-containing neurons, which in addition to hormonal feedback signals from the periphery (GH and IGF-1), contribute to a finely tuned regulatory system (184).  GH secretion can also be modulated by a number of neurotransmitters, peptides and circulating hormones as a result of their action on somatostatin and/or GHRH producing tuberoinfundibular neurons (185).  The rise of GH during sleep, for example, is probably mediated by cholinergic pathways suppressing somatostatin secretion (186).  Stress and sepsis can also be associated with a rise in GH levels mediated by catecholamines by increasing GHRH (185).  The precise origin of the neurons giving rise to these neuromodulators, however, is not known.  Glutamate may also be of importance by increasing the secretion of GHRH as N-methyl-D-aspartate increases GH secretion (187), and this can be attenuated by GHRH antibodies (188).  In addition, the majority of GHRH neurons in the arcuate nucleus are contacted by axons expressing the glutamate transporter 2 (189), a selective marker for glutamatergic elements. 

Fig. 28. Schematic representation of the interactions between somatostatin (SRIH)-producing neurons in the hypothalamic periventricular nucleus (Pev) and growth hormone releasing hormone (GHRH)-producing neurons in the arcuate nucleus.  Note that in addition to projections to the external zone of the median eminence, these neurons may also have reciprocal connections.  Ghrelin (GHR) may also influence GH secretion by acting directly on somatotrophs or through stimulatory effects on GHRH neurons.  AP = anterior pituitary, GH = growth hormone, IGF = insulin-like growth factor, ME = median eminence, PP = posterior pituitary.  R-SRIH and R-GHRH correspond to receptors for the respective peptides.  (Adapted from Epelbaum J: Intrahypothalamic neurohormonal interactions in the control of growth hormone secretion. Functional Anatomy of the Neuroendocrine Hypothalamus, Wiley, Chichester, 1992; pp 54-68.)

    A new and potentially exciting chapter in the understanding of the physiology of GH secretion has been the discovery of ghrelin, the most potent (on a molar basis) GH secretogogue known in man (190).  Ghrelin circulates in the bloodstream, secreted primarily from the stomach, but is also produced by neurons in the hypothalamic arcuate nucleus (191).  Although the anterior pituitary somatotrophs contain receptors for ghrelin, suggesting a direct action, it is more likely that ghrelin exerts its main effects in the hypothalamus by triggering GHRH secretion (192).   Ghrelin levels fall in response to rising levels of GH, providing evidence for a gastro-hypophysial feedback loop (193).

2.  Modulation of CRH Tuberoinfundibular Neurons by Stress
Afferent input to tuberoinfundibular neurons from distant loci in the brain is yet another important regulatory mechanism over hypophysiotropic function and is one way that tuberoinfundibular neurons are integrated with other functions of the brain.  The parvocellular subdivision of the paraventricular nucleus receives direct, dense, afferent input from autonomic centers in the lower brain stem including the nucleus of the tractus solitarius (NTS), dorsal motor nucleus of the vagus (DMNv) and several catecholamine groups in the dorsal and ventral lateral medulla, carrying visceral sensory information primarily from the abdomen and thorax through the vagus and glossopharyngeal nerves (194).  At least part of this projection is noradrenergic but other substances such as neuropeptide Y (195), activin (196), and GLP-1 (197) are also carried in these fibers, some coexisting in catecholaminergic neurons.  After traversing through the medial forebrain bundle in the lateral hypothalamus, axons containing catecholamines have been observed to make numerous synaptic contacts with CRH-producing neurons in the paraventricular nucleus (198,199) and to induce the secretion of CRH primarily via 1 adrenergic receptors (200).  In this manner, sensory information from the periphery (e.g. heart rate, blood pressure as during hemorrhagic shock) has the potential to alter the set point for the secretion of hypophysiotropic CRH using norepinephrine as the central mediator, and thereby increase circulating levels of glucocorticoids.
In a similar fashion, increased secretion of glucocorticoids in response to infection or inflammation is due to the activation of catecholaminergic neurons of the NTS (A2 noradrenergic, C2 adrenergic) and rostral ventrolateral medulla but initiated by endotoxin and proinflammatory cytokines such as interleukin-1 (IL-1) (201).  Under these circumstances, the set point for feedback inhibition of hypophysiotropic CRH secretion is altered to allow the powerful immunosuppressant action of glucocorticoids to limit the severity of the inflammatory response (202).  If the ascending catecholamine pathway to the PVN is transected, the ability of IL-1 to increase CRH mRNA in PVN neurons is reduced (203).  It is proposed that IL-1 exerts its effect on endothelial cells and/or perivascular microglia at the blood-brain interface, resulting in activation of cyclooxygenase-2 (the rate limiting enzyme for the formation of prostaglandins), the release of prostaglandin E2 (PGE2) into the surrounding tissue, and ultimately activation of catecholaminergic neurons through prostaglandin receptors (204).   This hypothesis is supported by the demonstration that focal injection of PGE2 into the medulla reproduces the activating effects of IL-1 on CRH neurons in the PVN (205).  Alternatively, cytokines may exert their effects on vascular cells at the blood-brain-barrier directly within the PVN, itself, or after penetrating the blood-brain barrier at circumventricular organs such as the OVLT and then transmitting the information through neural pathways that interconnect these structures with the PVN (204, see Thermoregulation below).
Neurogenic stress also leads to resetting of the HPA axis and similarly characterized by elevated circulating glucocorticoid levels and increased CRH gene expression in hypophysiotropic neurons (201).  However, the mechanism would appear to be vastly different than that described above as indicated by persistent HPA activation under these conditions despite disruption of the ascending catecholaminergic pathways to the PVN (206).  CRH neurons receive inputs from other portions of the brain such as the forebrain limbic system, and surgical ablation of hippocampal efferents to the hypothalamus (207) or lesions in the bed nucleus of the stria terminalis (208) increase the concentration of CRH mRNA in the paraventricular nucleus.  Thus, while the end result to increase circulating levels of glucocorticoids is similar in all stress paradigms, depending upon the type of stress, different regions of the brain are recruited to allow resetting of the HPA axis.  Nevertheless, recent evidence suggests that the mitogen activated protein (MAP) kinase signaling pathway may be a common final mechanism in hypophysiotropic CRH neurons that links a variety of different stimuli that activate the HPA axis to an increase in CRH gene expression (209-211). 
The circuitry described above that allows resetting of the HPA axis is illustrated in Fig. 29.

3.  Modulation of TRH Tuberoinfundibular Neurons by Fasting and Infection
Elucidation of the mechanisms by which the hypothalamic-pituitary-thyroid (HPT) axis responds to fasting provides another excellent example of how afferent input from neurons arising outside of the PVN can influence the secretion of hypophysiotropic neurons.  Similar to the feedback mechanisms controlling the adrenal axis, maintenance of normal thyroid function is dependent upon a negative feedback control system in which circulating levels of thyroid hormone (T4 and T3) influence the biosynthesis and secretion of TRH in hypophysiotropic neurons in the PVN (Fig. 30A,B) and TSH in the anterior pituitary (182).  In response to fasting or infection, however, this normal homeostatic system is altered in a way that is presumably beneficial for survival.  Under these circumstances, there is a fall in circulating thyroid hormone levels but a seemingly paradoxical reduction in TRH gene expression in the PVN (Fig. 30C,D), reduced secretion of TRH into the portal blood and low or inappropriately normal plasma TSH (212-215), rather than the anticipated increase in all of these parameters as seen in primary hypothyroidism.  Thus, during fasting, the normal feedback mechanism described above is overridden, and a state of central hypothyroidism is transiently induced.  Presumably, by reducing thyroid thermogenesis and preserving nitrogen stores, this mechanism is an important adaptive response to reduce energy expenditure until the adverse stimulus is removed (216).

Fig. 30. High magnification in situ hybridization autoradiographs of proTRH mRNA in the paraventricular nucleus (PVN) of a (A) euthyroid and (B) hypothyroid animal.  Note marked increase in TRH mRNA when circulating levels of thyroid hormone fall.  Low magnification in situ hybridization autoradiographs of proTRH mRNA in the PVN of (C) normal fed and (D) fasting animals. Hybridization signal is markedly reduced in the fasting animals despite low circulating levels of thyroid hormone.

    The HPT axis is primarily modulated by afferent input derived from the hypothalamus, itself.  At least two anatomically distinct populations of neurons in the arcuate nucleus with opposing functions, proopiomelanocortin (POMC)-producing neurons that also co-express cocaine and amphetamine-regulated transcript (CART), and NPY-producing neurons that co-express agouti related peptide (AGRP), appear to be responsible (144,217-219).  Both neuronal populations express receptors for the white adipose tissue-derived circulating hormone, leptin, and project to hypophysiotropic TRH neurons in the PVN through a monosynaptic, arcuate-PVN pathway (220-222).  Alpha-MSH, a translation product of POMC, and CART (originally described as a mRNA induced in the striatum following psychostimulant drug administration) both induce transcription of the TRH gene in hypophysiotropic neurons (136,217), whereas NPY and AGRP are inhibitory (218,219), NPY via direct effects on Y1 and Y5 receptors on TRH neurons (195), and AGRP by antagonizing -MSH at melanocortin receptors (224).  Thus, during fasting, when circulating levels of leptin decline, expression of the genes encoding POMC and CART are reduced simultaneously with a marked increase in the genes encoding NPY and AGRP (225,226), effectively lowering the threshold of feedback inhibition of hypophysiotropic TRH by circulating levels of thyroid hormone.  This important homeostatic system, which is present in all animal species studied including man, is illustrated in Fig. 31.

Fig. 31. Schematic representation of the regulatory system involved in establishing the set point for feedback regulation of TRH mRNA in hypophysiotropic neurons in the hypothalamic paraventricular nucleus in (A) fed and (B) fasted states.  Insets in A and B are in situ hybridization autoradiograms of coronal sections through the rat hypothalamus showing the effect of fasting to reduce TRH mRNA in the PVN.  ARC = arcuate nucleus, PVN = paraventricular nucleus.

    Circulating thyroid hormone levels also fall in association with severe illnesses and infection (226), but use a different set of regulatory controls.  This is based on the observation that both POMC and CART gene expression are increased in the arcuate nucleus (227) and circulating levels of leptin are elevated under these conditions (228).  In addition, norepinephrine secretion is increased in the PVN, and ordinarily would be expected to stimulate the secretion of TRH (229).  The precise anatomical pathways and mediators that override the activating effects of catecholamines, leptin and -MSH on TRH neurons are not yet known.  Recent evidence, however, has demonstrated that type 2 iodothyronine deiodinase (D2), an enzyme that converts thyroxine into the more biologically active thyroid hormone, tri-iodothyronine, is expressed in tanycytes and that D2 expression and enzymatic activity is substantially increased by endotoxin (230).   Given the location of these cells in the median eminence in contact with both the CSF and blood in the portal vascular plexus, and recent evidence that they express the thyroid hormone transporter, monocarboxylate transporter 8 (MCT8) (231), it is conceivable that tanycytes contribute to the fall in circulating thyroid hormone associated with infection by increasing the concentration of T3 in the mediobasal hypothalamus and suppressing the synthesis of TRH in hypophysiotropic neurons by local feedback regulation.  Hence, tanycytes may extract T4 from the bloodstream or the CSF, convert T4 to T3, and then release T3 into the CSF that could diffuse into the PVN by volume transmission (232) after moving between ependymal cells lining the third ventricle (2002d); release T3 directly into the median eminence that could be taken up by TRH axon terminals and then transported retrogradely to the PVN; and/or concentrate in arcuate nucleus neurons that have known projections to TRH neurons in the PVN (217, 221,222,234).  T3 may also be released into the portal capillary system for conveyance to the anterior pituitary and contribute to the mechanism whereby endotoxin inhibits the secretion of TSH.  The hypothesized regulatory mechanism is schematized in Fig. 32.

 
4.  Regulation of GnRH Secretion
Pulses of GnRH initiate the pulsatile release of anterior pituitary gonadotropins, and changes in the GnRH pulse frequency dictate how much LH and FSH ultimately will be released (235,236).  This intermittent signal is of critical importance for pubertal development and necessary for the regulation and maintenance of normal reproductive function throughout the ovulatory cycle.  In the absence of episodic GnRH release, such as with continuous, exogenous administration of GnRH, the synthesis and secretion of gonadotropins are profoundly suppressed as a result of desensitization of GnRH receptors (237).  The central mechanisms governing the pulsatile secretion of GnRH may involve a variety of factors, but the most important would appear to be the recently discovered kisspeptin/G protein-coupled receptor 54 (GPR54) neuroregulatory system (235,239,242).  The kisspeptins derive from a single precursor but comprise a group of peptide molecules ranging from 10-54 amino acids, all capable of binding and activating GPR54 with similar efficacy (243).  In humans, kisspeptin-54 has also been termed metastin on the basis that it was originally recognized to suppress cancer metastasis.
GnRH-producing neurons are located primarily in the preoptic region in rodents, but in all animal species give rise to axons that project caudally to terminate in the external zone of the median eminence (244).  Although GnRH neurons may have intrinsic oscillatory characteristics that might explain the pulsatile secretion of GnRH (245,246), evidence would support the importance of GPR54 and the kisspeptins in modulating their secretory responses and in particular, the resurgence of GnRH pulsatility during puberty.  Namely, both humans and animals with GRP54 deficiency have hypogonadotropic hypogonadism despite normal development of GnRH neurons and normal LH and FSH secretion in response to GnRH (235,247), GPR54 mRNA is expressed by GnRH neurons (247), central and peripheral administration of kisspeptins potently induce gonadotropin secretion ( 248,249), kisspeptin-induced LH secretion can be blocked with GnRH receptor antagonists ( 249,250), kisspeptin depolarizes ~90% of GnRH neurons in adult mice (251), and transgenic mice with targeted disruption of the Kiss1 gene (that gives rise to the kisspeptins) has an identical phenotype as transgenic mice deficient in GPR54 (252).  As brief iv infusions of kisspeptin every hour for 48h induce pulsatile LH discharges similar to those observed during puberty, whereas continuous kisspeptin infusion in mice or monkeys downregulate LH secretion by desentensitizing GRP54 (253,254,255), it has been proposed kisspeptin-producing neurons comprise the pulse generator for GnRH neurons or at the very least, amplify the activity of the pulse generator (255).
Neurons producing the kisspeptins are located in the hypothalamic arcuate nucleus and in some species, the anteroventral paraventricular nucleus (256), and express the alpha estrogen receptor (257).  Curiously, these two populations are regulated inversely to each other.  Thus, kisspeptin gene expression in arcuate nucleus neurons increases following ovariectomy and decreases with estrogen administration, whereas the reverse occurs in anteroventral paraventricular cells (255,256).   It has been proposed, therefore, that these two, kisspeptin neuronal populations may mediate the negative and positive feedback effects of estrogen on GnRH neurons (259), with the anteroventral paraventricular nucleus neurons involved in the estradiol-induced preovulatory GnRH/LH surge.  Indeed, kisspeptin-containing axon terminals have been observed to terminate on hypophysiotropic GnRH neurons, but only the minority of cells and with surprisingly few boutons (255,257).  In contrast, kisspeptin-containing axons heavily inundate the median eminence and extensively intermingle with GnRH-containing axon terminals (255,257).  Although axo-axonal interactions between kisspeptin- and GnRH-containing axon terminals have not been observed (255), these specializations may not be required for physiologic function in the median eminence.  Evidence that peripheral administration of kisspeptin has a similar potent action on LH secretion as central administration (247), and recent physiologic data showing that exogenous administration of kisspeptin to hypothalamic explants deficient in GnRH neurons still potently release LH into the medium (253), strengthen the possibility that kisspeptin‘s actions may be directly on GnRH axon terminals in the median eminence.  Figure 33 summarizes a hypothesized mechanism for the neuroendocrine regulation of GnRH neurons by kisspeptin.

    Glial-neuronal interactions in the median eminence may also be involved in regulating the delivery of GnRH to the portal system.  Two mechanisms have been proposed (261,262).  The first involves the release of glutamate, prostaglandins and growth factors from tanycytes that induce the secretion of GnRH from their axon terminals in the median eminence.  The second involves plastic rearrangements between tanycyte end foot processes and GnRH axon terminals, allowing or disallowing secreted GnRH from entering portal capillaries.  Thus, it is proposed during the preovulatory gonadotrophin surge, estrogen binds to alpha type estrogen receptors on tanycytes and induces tanycyte end feet retraction through PGE2-dependent production of TGF1, allowing GnRH axon terminals to establish better contact with the portal vessels (263).  Estrogen may also affect the synthesis of adhesion molecules such as polysialylated neuronal cell adhesion molecule (PSA-N-CAM) and synaptic cell adhesion molecule (SynCAM), that facilitate glial-neuronal interactions and remodeling (261,262).  The release of nitric oxide from portal vessel endothelial cells may also participate in tanycyte retraction by affecting actin cytoskeleton remodeling (263).

B. Modulation of Vasopressin Secretion and Osmoregulation

    Maintenance of the appropriate solute concentration in plasma (osmotic homeostasis) and plasma volume (volume homeostasis) is dependent upon two major factors, the perception of thirst and the ability to synthesize and secrete the antidiuretic hormone, arginine vasopressin from magnocellular neurons in the hypothalamic PVN (265). These two factors are closely interrelated such that amount of vasopressin circulating in the periphery is proportional to the plasma osmolality (266).  Vasospressin induces cAMP and the translocation of specific aquaporin-2 water channels to the apical plasma membrane of tubular epithelial cells in the kidney, allowing water resorption (267,268).  In addition, the rise in osmolality has independent behavioral effects.  Thus, when plasma osmolality rises above basal levels, there is inducement to drink, shortly following the rise in vasopressin (265).  While some vasopressin neurons in the PVN are intrinsically osmosensitive (269,270), the major mechanism of osmoregulation is via afferent pathways originating from osmoreceptor cells in other neuronal populations.   These include inputs from the OVLT and the median preoptic nucleus, which if damaged, simultaneously abolish vasopressin secretion and thirst responses to hyperosmolality in both experimental animals and man (111,271).  The SFO is also activated by a rise in osmolality and may contribute to vasopressin release through direct afferent projections to the PVN and/or to the OVLT using angiotensin II as a mediator (108-111,272).  As mice with targeted disruption of the transient receptor potential (TRP) ion channels, TRPV1 and TRPV4, have impairment in vasopressin secretion and reduced drinking in response to hypertonic stimuli and show diminished cFos responses in the OVLT, these ion channels may be responsible for osmoreception (273).
Whereas forebrain pathways communicate information about osmolality to the PVN, brainstem projections tend to carry nonosmotic, baroregulatory information, and important for vasopressin secretion, particularly in association with hypovolemia and hypotension (111,265).  This information is carried through the vagus and glossopharyngeal nerves to the NTS and ventral lateral medulla, and then to the PVN through the ascending catecholaminergic pathways (Fig. 34).  Magnocellular neurons in the PVN appear to be primarily innervated by the A1 catecholamine-producing cells in the ventral lateral medulla (194).

Fig. 34.  Schematic drawing of the major pathways involved in regulation of vasopressin secretion.  Information about osmolality is relayed to the hypothalamus largely through osmoreceptor cells in the organum vasculosum of the lamina terminalis (OVLT), subfornical organ (SFO) and medial preoptic nucleus (MnPO).  Baroregulatory information is carried to magnocellular neurons in the paraventricular nucleus (PVN) and supraoptic nucleus (SON) largely by direct afferent projections from the brainstem including the bed nucleus of the stria terminalis (NST) and ventrolateral medulla (VLM) using catecholamines as its neurotransmitter, or indirectly through the parabrachial nucleus (PBN).  Vasopressin secretion can also be stimulated by activation of the chemoreceptor trigger zone in the area postrema (AP).  (Adapted from Stricker and Verbalis, Fundamental Neuroscience, ch 42, pp 1111-1126, 1999.)

    As described for hypothalamic tuberoinfundibular neurons, the threshold for vasopressin secretion by neurons of the magnocellular neurosecretory system can also be modified by their afferent signals as well as circulating factors.  For example, the osmotic threshold for vasopressin release can be altered by glucocorticoids.  Dexamethasone attenuates the vasopressin response to salt loading (274) and hypoadrenalism is commonly associated with the syndrome of inappropriate secretion of antidiuretic hormone (SIADH) that can be corrected by glucocorticoid administration (275).  These effects are exerted directly on vasopressin neurons given the presence of glucocorticoid receptors in these cells (276).  Other causes for SIADH, however, such as pulmonary disease and central nervous system disorders may be mediated through afferent pathways to the PVN.  Hormone mediators of these projections include vasoactive intestinal polypeptide (VIP), acetylcholine, angiotensin II, neuropeptide Y and noradrenaline, among numerous others (277,278).