Hormonal responses to acute stress rely on the rapid induction of corticotropin-releasing hormone (CRH) production in the mammalian hypothalamus, with subsequent instructive steps culminating in corticosterone release at the periphery. Hypothalamic CRH neurons in the paraventricular nucleus of the hypothalamus are therefore considered as ‘stress neurons’. However, significant morphological and functional diversity among neurons that can transiently produce CRH in other hypothalamic nuclei has been proposed, particularly as histochemical and molecular biology evidence associates CRH to both GABA and glutamate neurotransmission. Here, we review recent advances through single-cell RNA sequencing and circuit mapping to suggest that CRH production reflects a state switch in hypothalamic neurons and thus confers functional competence rather than being an identity mark of phenotypically segregated neurons. We show that CRH mRNA transcripts can therefore be seen in GABAergic, glutamatergic and dopaminergic neuronal contingents in the hypothalamus. We then distinguish ‘stress neurons’ of the paraventricular nucleus that constitutively express secretagogin, a Ca2+ sensor critical for the stimulus-driven assembly of the molecular machinery underpinning the fast regulated exocytosis of CRH at the median eminence. Cumulatively, we infer that CRH neurons are functionally and molecularly more diverse than previously thought.
The introduction of molecular methods in the neurobiologists’ tool kit has changed many long-held views on the structural and functional organization of the nervous system. One of the most recent additions is the use of single-cell RNA-sequencing (scRNA-seq), which can produce quantitative mRNA counts to predict the origin, lineage segregation, functional diversification and disease susceptibility of any cell type in the nervous system. Taking the power of ‘forward’ (or predictive) transcriptomics to the level of neuroanatomy, connectivity mapping by electrophysiology and genetic models, one can expect a sea-change of organizational rules that have hitherto been applied on neuronal identity and function (Shapiro et al. 2013, Romanov et al. 2015, Zeisel et al. 2015). Alternatively, the use of ‘reverse’ transcriptomics can validate many datasets that were generated using alternative tools. This is a particularly exciting prospect in neurobiology where sometimes mm-to-cm-scale distances exist between synaptic release sites and the somata of relevant projection neurons. Thus, functional transcriptomics at projection sites might be validated by scRNA-seq of somatic compartments (Romanov et al. 2015), which contain all mRNAs for the protein domain sustaining the regulated release of any neurotransmitter or neuroactive modulator. Here, we will review recent data on corticotropin-releasing hormone (CRH) mRNA-containing neurons in the mammalian hypothalamus to describe their unexpected molecular heterogeneity. By relying on a combination of novel experimental tools, we suggest that Crh expression can occur in glutamatergic, GABAergic and dopaminergic neurons and is a descriptor of specific ‘working modes’ and that the co-localization of CRH with secretagogin is a defining hallmark of ‘stress on’ CRH neurons in the paraventricular nucleus of the hypothalamus (PVN).
Organization of the hypothalamus–pituitary–adrenal axis: from systems to molecules
The body’s reaction to environmental stressors, including physical and psychosocial factors (McEwen 2007), is orchestrated by sequential hormone release in the hypothalamus–pituitary–adrenal (HPA) axis (Herman et al. 2003). HPA activation triggers corticosteroid release from the cortex of the adrenal gland to mobilize energy utilization, thus leading to integrative-protective responses to noxious conditions (Sapolsky et al. 2000, McEwen 2007). The first hierarchical level in the HPA axis is the release of CRH (Spiess et al. 1981, Vale et al. 1981) from the terminals of parvocellular hypothalamic neurons into hypophyseal portal vessels (Green & Harris 1947) at the median eminence (Swanson et al. 1983). According to the classical definition, the peripheral stress response then progresses through the secretion of adrenocorticotropic hormone (ACTH) (Mains et al. 1977, Roberts & Herbert 1977, Nakanishi et al. 1979) from the anterior lobe of the pituitary. Finally, ACTH induces steroidogenesis in, and corticosteroid release from, the adrenal cortex (Munck et al. 1984, McEwen & Stellar 1993). Corticosteroid action is then executed through glucocorticoid receptors (Gustafsson et al. 1987), encoded by the Nr3c1 gene (McEwen 2012), with ensuing changes in cellular states (‘cell-state switch’) tuned to overcome metabolic compromise.
Parvocellular neurons are broadly viewed as the origins of CRH for stress-induced, fast release and are situated in the PVN (Merchenthaler et al. 1983, Swanson et al. 1983, 1986, Alonso et al. 1986, Ceccatelli et al. 1989b, Swanson 1991). The PVN itself showcases the complexity of neuronal organization in the hypothalamus: it contains at least eight subdivisions (Swanson & Kuypers 1980, Simmons & Swanson 2009) with the tight spatial concentration of magnocellular and parvocellular neuroendocrine secretory motor neurons (that is, cells that release bioactive peptides into the portal circulation through release sites located outside the blood–brain barrier), neurons projecting to the brainstem and locally projecting hypothalamic neurons (Daftary et al. 1998, Krashes et al. 2014). Oxytocin and vasopressin (Du Vigneaud 1954) are produced in magnocellular neurons projecting directly to the posterior pituitary (Swaab et al. 1975, Rhodes et al. 1981). In contrast, parvocellular neurons are a kaleidoscope of neuroendocrine modalities, often co-releasing several neuropeptides and fast neurotransmitters in an ‘on-demand’ manner (Swanson et al. 1986, Ceccatelli et al. 1989a,b). Among parvocellular PVN neurons, CRH cells, besides projecting to the median eminence, also form axon collaterals toward other intrahypothalamic nuclei (e.g., lateral hypothalamus), thus increasing integrative capacity through the coincident control of neuronal circuitries and hormonal responses at the pituitary (Fuzesi et al. 2016).
Besides CRH at its ‘stress center’ (Alonso et al. 1986), parvocellular neuroendocrine cells can contain ‘releasing’ and ‘release-inhibiting’ hormones (e.g., growth hormone-releasing hormone and somatostatin (growth hormone-inhibiting hormone), thyrotropin-releasing hormone, gonadotropin-releasing hormone and dopamine (as prolactin-inhibiting hormone)) to regulate hormone release from the anterior pituitary (classed as growth hormone, thyroid-stimulated hormone, luteinizing and follicle-stimulating hormones and prolactin) (Guillemin 1978, Schally 1978, Ben-Jonathan & Hnasko 2001). Of particular interest in relation to stress, vasopressin potentiates the effect of CRH (Gillies et al. 1982, Turkelson et al. 1982, Rivier & Vale 1983) and can undergo induced expression in CRH neurons after adrenalectomy (Tramu et al. 1983, Kiss et al. 1984, Sawchenko et al. 1984). As such, these peptides co-exist in the same storage granules in boutons residing in the external layer of the median eminence (Whitnall et al. 1985).
The observation that basal ACTH and corticosterone levels can always be detected in blood (rather than being ‘on’/‘off’ signals) (Girotti et al. 2009, Romanov et al. 2015) indirectly suggests that CRH is constitutively released in minute amounts to continuously stimulate the production and release of downstream hormones. This concept has broad physiological significance as, for example, pancreatic β cells require glucocorticoids for their survival. Consequently, hypothalamic CRH hypofunction leads to Addison’s disease with diabetes being its primary comorbidity (Miller & O’Callaghan 2002, Napier & Pearce 2014, Kahaly & Hansen 2016). There also is a well-established relation of CRH (and corticosteroids) to mood disorders (Gold et al. 1995, Reul & Holsboer 2002, Lloyd & Nemeroff 2011, Waters et al. 2015): CRH release follows the diurnal cycle with autonomous peaks observed during the early morning hours and late afternoon/evening (Salata et al. 1988, Owens et al. 1990, Watts et al. 2004, Girotti et al. 2009). These data suggest that CRH might have more widespread biological roles than previously thought through cell- or tissue-specific expression of its receptors (see below). As such, and beyond regulating the stress response, CRH is increasingly implicated in the regulation of appetite (Drescher et al. 1994, Uehara et al. 1998).
CRH production in the brain
Beyond its canonical expression site identified as the parvocellular neurons of the PVN (Simmons & Swanson 2009), CRH production is not restricted to the PVN. In the hypothalamus, CRH is also detected in magnocellular cells of the lateral hypothalamus and preoptic area (Alon et al. 2009). A critical difficulty to identify the cellular sites of CRH production is its unusually low levels at rest and fast induction during stressful conditions or in particular metabolic states. Therefore, histochemical solutions so far relied on agents blocking the axonal transport machinery to elevate CRH levels in neuronal somata (Merchenthaler et al. 1982, Alonso et al. 1986) or by experimentally boosting its expression (Ceccatelli et al. 1991). In accord with the expansion of amenable molecular genetic tools, e.g. the Cre/lox system, we recently used Crh-IRES-Cre-driven green fluorescent protein (GFP) expression to allow the accumulation of GFP in response to episodic Crh expression during a cell’s life time (Fig. 1A). These results document many GFP+ neurons scattered within the hypothalamus, including its anterior and ventromedial areas. This genetic model suggests, in accord with earlier histochemical studies (Merchenthaler et al. 1982, Swanson et al. 1983, Keegan et al. 1994, Potter et al. 1994, Alon et al. 2009), that episodic Crh mRNA expression might be intrinsic to many hypothalamic neurons. Likewise, Crh gene transcription and protein expression were documented in many extrahypothalamic brain areas, including thalamic nuclei, bed nucleus of the stria terminalis, amygdala, medulla oblongata (especially inferior olive), piriform cortex and many neocortical and hippocampal subregions (Merchenthaler et al. 1982, Potter et al. 1994, Lantos et al. 1995, Alon et al. 2009, Zeisel et al. 2015, Hooper & Maguire 2016). Thus, an essential heterogeneity of CRH neurons emerges with CRH likely being a molecular determinant of functional cellular competence under particular behavioral and metabolic conditions.
Exquisite co-localization studies suggest that CRH-positive(+) neurons could mediate either excitatory or inhibitory neurotransmission in a regionally defined manner. For excitatory CRH+ neurons, those of the bed nucleus of stria terminalis are considered as prime candidates (Lee & Davis 1997, McNally & Akil 2002, Hrabovszky et al. 2005). Likewise, CRH+ cells of the central nucleus of the amygdala participate in extended excitatory circuits in which these neurons might potentiate differential responses to tonic/unpredictable vs phasic/predictable stressors is currently under debate (Lee & Davis 1997, Palkovits 2000, McNally & Akil 2002), even including peripheral responses such as the regulation of heart rate (Nijsen et al. 2001). In turn, cortical and hippocampal interneurons might also express CRH, which use GABA as inhibitory neurotransmitter (Zeisel et al. 2015, Hooper & Maguire 2016). Nevertheless, CRH actions are chiefly diversified at the subcellular level as its two G-protein coupled receptors (with Gs/Gq coupling), CRHR1 (Crhr1) and CRHR2 (Crhr2) can differentially signal in response to this neuropeptide (Dautzenberg et al. 2001, Dautzenberg & Hauger 2002, Aguilera et al. 2004, Gutknecht et al. 2010).
ScRNA-seq-based subclassification of Crh mRNA+ hypothalamic neurons
The classical doctrine of hypothalamic stress induction posits CRH neurons being parvocellular neurosecretory cells situated in the PVN in mammals (Swanson et al. 1986, Simmons & Swanson 2009). However, ‘why’, ‘when’ and ‘for how long’ other neurons (GFP+; Fig. 1A) beyond those in the PVN produce CRH and at which biophysical thresholds and synaptic sites release their CRH contents are much less unequivocal. Among the many reasons for this ambiguity at the molecular regulatory levels, we emphasize the rapid induction of CRH production (Watts et al. 1999) under certain metabolic demands. This would allow for CRH being re-classified as a permissive cellular feature underscoring functional competence and circuit recruitment rather than a cell-identity mark (we define a cell-identity mark as a molecular feature that determines the cell’s function, sets it apart from neighboring cells and whose loss might render the cell functionally incompetent and misplaces it in the molecular taxonomy of cells in a given brain volume. In contrast, functional modality refers to a molecular feature (gene, gene set and combinatorial code) that is only present under certain (patho)physiological conditions yet does not define either its neurotransmitter phenotype or connectivity pattern over long time scales. To test if CRH is an identity mark or a cell-state modality, we analyzed the level of heterogeneity of gene expression defining CRH signaling (that is, Crh, Crhr1, Crhr2 and Crhbp) in hypothalamic neurons (Fig. 2A)). If so, a combination of novel methods including scRNA-seq and in situ gene expression (Stahl et al. 2016) profiling could decode the heterogeneity of hypothalamic CRH+ neurons and allow for their subclassification based on the correlated analysis of fast neurotransmitter contents. In addition, one would expect these data be supportive of more conventional approaches, such as the intracerebroventricular injection of colchicine, an axonal transport inhibitor (Ceccatelli et al. 1991), which is widely used for the improved detection of neuropeptide expression (and also represents a substantial stress stimulus).
As such, many neuropeptides (beyond CRH in the PVN) are used as specific hallmarks of neuronal identity in the hypothalamus (e.g. neuropeptide Y, oxytocin, agouti-related peptide (AgRP)) (Meister et al. 1990, Lantos et al. 1995, Aponte et al. 2011, Dietrich et al. 2015), asserting that expression of a single gene (or at most their circumscribed subset) typifies a neuronal subtype. This knowledge can thus be used to justify recent developments in quantitative mRNA profiling tools, as any single-cell analysis, if correct, would lean toward neuronal identity marks that are stably expressed in a given contingent of neurons. Accordingly, the recent development of mathematical algorithms (t-distributed stochastic neighbor embedding (tSNE) and BackSpin clustering (van der Maaten & Hinton 2008, Islam et al. 2014, Macosko et al. 2015, Zeisel et al. 2015)) allows for the refined analysis of scRNA-seq data on hypothalamic neurons (Romanov et al. 2015, 2016). This approach confirmed that many neuropeptide mRNAs (Hcrt, Agrp, Npy and Trh) are indeed neuronal identity marks (Romanov et al. 2016) and once ensuing peptides are produced, their ‘use-dependent’ release will impinge upon a neuronal circuit to determine metabolic output (e.g. AgRP invariably present in a neuronal subset in the arcuate nucleus that controls body weight and metabolic rate (Mizuno et al. 2003)). In contrast, Crh mRNA expression fails to meet these criteria: t-SNE projections demonstrate widespread Crh expression in hypothalamic neurons, which lack subtype definition. In accord, BackSpin clustering characterized several neuronal subtypes (including clusters #14, 15, 44 and 45 as per Romanov et al. (2016)) that contain fractions of Crh mRNA-expressing neurons (Fig. 1B and C). Thus, we conclude that a neuron’s ability to produce CRH is a functional modality rather than an identity-defining feature.
If CRH expression is a functionally induced modality, then the diversity of fast neurotransmitter contents in hypothalamic neurons expressing this neuropeptide can also be expected (Fig. 1C). Our detailed analysis of all hypothalamic Crh neurons showed their segregation into either GABAergic (that is, Gad1, Gad2 and Slc32a1 co-expression) or glutamatergic (Slc17a6) neurotransmitter phenotypes. As neurotransmitter coexistence appears as a unique and relatively common feature of hypothalamic neurons (in 10–20% of cells as coincidently revealed by both Fluidigm C1 and Drop-seq transcriptomics) (Romanov et al. 2016), dual GABA/glutamate, dopamine/GABA and dopamine/glutamate neurons can also contain Crh mRNA (Fig. 1E). In sum, Crh+ neurons encompass several subtypes of GABA and glutamate neurons, reinforcing the concept that episodic CRH production can be associated with multiple network-driven, metabolic or disease-related contexts rather than being a feature of a developmentally coded subset of hypothalamic neurons (that is, phylogeny defined at the transcriptional level).
ScRNA-seq uncovered that CRH neurons residing in the PVN are glutamatergic. This cell cluster can be distinguished by the co-existence of several mRNA marks, including Tmem59l, Fuca1, Npr3 and Arnt2, and whose existence in the PVN is validated by open-source histochemical databases (see in situ hybridization, Allen Brain Atlas/brain-map.org). Meanwhile, GABAergic neurons that can contain CRH co-express Pgr15l, a histochemical mark mapped onto the anterior part of the PVN, cells spread around the histochemical boundaries of the PVN and those residing in the dorsomedial hypothalamus (Fig. 1C). Alternatively, GABAergic CRH neurons can co-express LIM homeobox 6 (Lhx6) transcription factor and map onto the preoptic area (Fig. 1D). Moreover, a continuum of Lhx6+/Crh+ neurons extends toward the bed nucleus of stria terminalis, which is an extrahypothalamic area with innate projections to, among other, the PVN where it forms local circuits regulating the stress axis (Choi et al. 2007). Based on their transcriptional code, we hypothesize that GABAergic CRH+ neurons, despite their spatial heterogeneity, can originate from the same precursors, migrate toward different (extra)hypothalamic subregions and ultimately be a unified neuronal subtype. An outstanding question based on novel scRNA-seq and localization data is the determination of when (context and mode of operation) and for how long (induction) any neuronal subclass can express physiologically meaningful levels of CRH protein and if their postsynaptic targets are endowed with CRH receptors for the efficient translation of CRH signals into functional outcome.
CRH receptors in the brain
Two genes (Crhr1 and Crhr2) encode receptors to mediate CRH action with their mRNAs detected in many brain regions (Wynn et al. 1984, Aguilera et al. 2004). Crhr1 is widely expressed in the neocortex, piriform cortex, olfactory system and cerebellum (Van Pett et al. 2000). Most recently, cortical and hippocampal Crhr1 mRNA expression was also validated by scRNA-seq (Zeisel et al. 2015), showing pronounced Crhr1 gene expression in pyramidal cells in cortical layers 2/3, 4 and 5. These findings contrast those in interneurons, which generally lack Crhr1 mRNA transcripts. Given that interneurons can instead express CRH itself, the combination of cell-type-specific transcriptomics with circuit mapping allows predictions on local CRH signaling between CRH+ interneurons and Crhr1-laden pyramidal cells. In situ hybridization demonstrates a more restricted distribution pattern for Crhr2 mRNA, with its primary concentration in lateral and triangular nuclei of the septum and the amygdaloid complex. Thus, CRHR1 and CRHR2 functions might mediate vastly different biological responses and involve a combination of short- and long-range modes of intercellular signaling.
When analyzing CRH receptor distribution in the hypothalamus, expression levels are relatively low for both Crhr1 and Crhr2 with the former being present in arcuate, suprachiasmatic, anterior and dorsomedial hypothalamic nuclei (Van Pett et al. 2000), whereas the latter is distributed across anterior periventricular, ventromedial, arcuate and anterior hypothalamic nuclei (Van Pett et al. 2000). Notably, stress induces Crhr1 expression in the PVN, which is a significant feedback station for sustained CRH production and release (Makino et al. 1995, Aguilera et al. 2004).
Recently, we sought to address the precise cellular distribution of neurons that can respond to CRH as a neuromodulator locally in the hypothalamus (Romanov et al. 2016). Although ~10% of hypothalamic neurons express Crh mRNA (>1 mRNA copy), Crhr1 and Crhr2 were only sparsely present in hypothalamic neurons (4.5% and 2.6% of 898 neurons, respectively) at unexpectedly low copy numbers. The following steps of caution are warranted when considering these data: (i) low levels of Crhr1 mRNA detection (Hata et al. 1993, Sollner et al. 1993, Sutton et al. 1998, Pang & Sudhof 2010) might lead to false negatives (missed fraction of neurons), (ii) mRNA-to-protein translation and the stability of ensuing receptor complexes and their effectors (Fig. 2A) might diversify the efficacy of signal transduction and (iii) induced expression (Aguilera et al. 2004) could grossly affect CRH-responsive hypothalamic neuronal networks. Yet, our comparison of scRNA-seq data from hypothalamic and cortical datasets (Romanov et al. 2015, Zeisel et al. 2015) produced on identical platforms suggests that CRH+ efferents commonly form synapses outside the hypothalamus and/or CRH is mainly released into the ventricular and hypophyseal portal systems. Conversely, the likelihood of local-circuit neuronal ensembles using CRH as neuromodulator at rest is relatively modest. An exception to the above rules might be an autocrine role for CRH in the PVN as these neurons express both Crh and Crhbp (CRH binding protein) mRNAs with Crhbp inactivating CRH action (Dautzenberg & Hauger 2002, Seasholtz et al. 2002). Cumulatively, these data suggest a dichotomy of CRH content and CRH-responsive neuronal substrates within the hypothalamus and at extrahypothalamic sites.
Secretagogin controls CRH release at the median eminence
The regulated release of any bioactive peptide within the brain or at the median eminence relies on the coordinated assembly of a soluble N-ethylmaleimide-sensitive-factor attachment receptor (SNARE) complex, allowing for the priming, docking and fusion of neuropeptide-filled large dense-core vesicles (LDCVs) (Hata et al. 1993, Sollner et al. 1993, Sutton et al. 1998, Pang & Sudhof 2010). In the hypothalamus, the exocytosis machinery includes many region- and cell-type-specific proteins, such as CAPS-1 and secretogranins (Miyazaki et al. 2011, Tobin et al. 2012). Until recently, our understanding of the molecular underpinnings of CRH release was limited. Recently, a focused scRNA-seq study (Romanov et al. 2015) in the mouse PVN identified the coincident presence of Crh, Nr3c1 (glucocorticoid receptor, subfamily 3, group C member 1) (van Rossum et al. 2004) and secretagogin (Scgn) (Romanov et al. 2015) (Fig. 2B and C). These data were validated by light and electron microscopy in both the PVN and the median eminence, localizing secretagogin to membrane surfaces at terminals releasing CRH into the hypophyseal portal system (Romanov et al. 2015) (Fig. 3B and C), anatomically tying secretagogin to the machinery priming CRH release.
Secretagogin is an appealing subject of study as it is a member of the EF-hand superfamily of Ca2+-binding proteins, expressed in all organ systems typically associated with neuroendocrine cells (Wagner et al. 2000, Adolf et al. 2007, Alpar et al. 2012), including the pituitary, adrenal gland, gut and pancreas (Wagner et al. 2000, Hasegawa et al. 2013). In the nervous system, secretagogin is localized to neurons populating the neocortex, hippocampus, cerebellum, basal ganglia, rostral migratory stream and olfactory bulb (Attems et al. 2008, Mulder et al. 2009, 2010, Gati et al. 2014). Secretagogin is a Ca2+ sensor protein ([Ca2+]0.5 is of ~25 µM for secretagogin in physiological salt buffers), which by definition induces protein–protein interactions through conformational change upon Ca2+ binding followed by downstream signaling to control discrete cellular functions (Schwaller 2010). Secretagogin’s in vitro interactome initially included synaptosomal-associated protein 25 kDa (SNAP-25) (Rogstam et al. 2007) with the more recent discovery of vesicle cargo, traffic and docking/release proteins (Bauer et al. 2011a,b, Romanov et al. 2015). Particularly, the 5th EF-hand domain can interact with cytoskeletal components (microtubules) (Maj et al. 2010, Yang et al. 2016). Thus, secretagogin possesses basic properties of an integrative sensor that can orchestrate the release of neuropeptide-containing LDCVs.
In the hypothalamus, secretagogin is chiefly expressed in the PVN and arcuate nuclei (Mulder et al. 2009, Romanov et al. 2015). Even though secretagogin can partially co-exist with vasopressin and oxytocin in some mammals (Fig. 2C), the majority of secretagogin+ neurons lacks any appreciable neuropeptide/hormone co-labeling in resting mice (Romanov et al. 2015). This constellation led to the detailed study of secretagogin+ neurons and their molecular interrogation by correlated transcriptomics and proteomics. Notably, converging genetic and histochemical evidence suggests that secretagogin is a constitutive mark of stress-activated CRH+ neurons that gate ACTH release from the pituitary (Romanov et al. 2015). Acute silencing of secretagogin expression (by siRNA) in vivo and in vitro provided critical insights into secretagogin’s function (Hasegawa et al. 2013, Romanov et al. 2015): (i) secretagogin knockdown induces CRH accumulation in the PVN. This suggests that secretagogin deficiency might limit the translocation of CRH to release sites and its Ca2+-dependent release into the portal circulation and (ii) concordantly, ACTH (Cam & Bassett 1983) and corticosterone production are blunted after secretagogin knockdown in animals exposed to acute formalin stress (Romanov et al. 2015) (Fig. 3D). These data also provide genetic evidence on a selective parvocellular Ca2+-dependent gatekeeper to the HPA axis whose function underpins a peripheral corticosterone surge from the adrenal cortex (Makara et al 1981).
Even though secretagogin is broadly recognized to scale hormone release (Wagner et al. 2000, Hasegawa et al. 2013, Romanov et al. 2015) (Fig. 3A), its exact interacting partners in neurons that might prime axonal transport and docking secretory vesicles remained unknown until recently. Using subtractive proteomics comparing Ca2+-saturated and Ca2+-free conditions and ‘reverse’ transcriptomics to verify the expression of co-purifying proteins in parvocellular neurons, we have defined protein subfamilies essential for the formation, intracellular traffic, priming and docking of release vesicles (examples: α-SNAP, Rab family members, syntaxins, synaptophysin, TMED10 and CAPS1), and some of them are specific to hypothalamic neurons (Fig. 3E) and release mechanisms (e.g. CAPS-1) (Tobin et al. 2012). In accord, ultrastructural data localized secretagogin to large axon terminals at the median eminence and substantiated its association to the readily releasable pool of vesicles, as well as plasmalemmal compartments. Because CRH, as many other peptide hormones released at the median eminence, is neither reused nor recycled after entering the blood stream, we hypothesize that impaired CRH secretion into the portal blood upon genetic secretagogin ablation impairs not one but many regulated Ca2+-sensitive intracellular checkpoints in parvocellular neurons (Fig. 3E). Thus, secretagogin qualifies as a molecular determinant rate-limiting CRH release.
Here, we collated recent molecular evidence highlighting the unprecedented molecular heterogeneity of hypothalamic neurons. The use of multiparametric experimental paradigms on the backdrop of array technologies provides unique precision and analytical power to distinguish neuronal modalities, let these be graded identity differences or temporal recruitment to neuronal circuits. Besides CRH neurons, novel levels of evidence emerge on the molecular heterogeneity of those affecting food intake (Henry et al. 2015), hedonism (amygdala), multisensory processing (e.g. association of vision and smell) (Macosko et al. 2015, Saraiva et al. 2015) and conscious representation of metabolic demands (cortex) (Zeisel et al. 2015). Thus, an unexpected and rapid expansion of concepts substantially refining and sometimes revising decade-long theories might be expected, allowing multimodal integration of information flow and refinement at extra- and intrahypothalamic circuits. The advent of deciphering the molecular heterogeneity of CRH neurons already produced novel insights into permissive checkpoints of the release of this hormone, as well as the parsing of probable subclasses of CRH neurons. Accordingly, CRH belongs primarily to ‘functional modality’ rather than ‘cell-identity’ marks. Future studies dissecting the upstream control of CRH neurons classically not considered ‘stress effectors’ in the hypothalamus, the metabolic contexts prompting their CRH production and if CRH receptors are coincidently expressed at synaptically wired circuit components and if receptor-level switches in decoding the physiological gain of CRH signaling exists will undoubtedly fuel the establishment of novel organizing principles of hypothalamic neuronal diversity.
Declaration of interest
The authors declare that there is no conflict of interest that could be perceived as prejudicing the impartiality of this review.
This work was supported by the Swedish Research Council (T Ha, T Hö), Hjärnfonden (T Ha), the Novo Nordisk Foundation (T Ha, T Hö), the European Research Council (Secret-Cells; T Ha), the National Brain Research Program of Hungary (MTA-SE NAP B, KTIA_NAP_13-2014-0013; A A) and intramural funds of the Medical University of Vienna (T Ha). R A R was supported by the European Molecular Biology Organization (ALTF reference 596-2014) and EU FP7 program (Marie Curie Actions EMBOCOFUND2012, grant GA-2012-600394).
The authors thank Dr A Zeisel, Dr S Linnarsson (Karolinska Institutet), Dr M Farlik and Dr C Bock (CeMM, Austrian Academy of Sciences) for their initial contributions to single-cell RNA-sequencing and Dr J S Bains (The Hotchkiss Brain Institute, University of Calgary, Calgary, Canada) for CRH-Ires-Cre mice.
AlonsoGSzafarczykAAssenmacherI1986Immunoreactivity of hypothalamo-neurohypophysial neurons which secrete corticotropin-releasing hormone (CRH) and vasopressin (Vp): immunocytochemical evidence for a correlation with their functional state in colchicine-treated rats. Experimental Brain Research61497–505. (doi:10.1007/bf00237574)
AttemsJPreusserMGrosinger-QuassMWagnerLLintnerFJellingerK2008Calcium-binding protein secretagogin-expressing neurones in the human hippocampus are largely resistant to neurodegeneration in Alzheimer’s disease. Neuropathology and Applied Neurobiology3423–32. (doi:10.1111/j.1365-2990.2007.00854.x)
CeccatelliSErikssonMHokfeltT1989aDistribution and coexistence of corticotropin-releasing factor-, neurotensin-, enkephalin-, cholecystokinin-, galanin- and vasoactive intestinal polypeptide/peptide histidine isoleucine-like peptides in the parvocellular part of the paraventricular nucleus. Neuroendocrinology49309–323. (doi:10.1159/000125133)
ChoiDCFurayAREvansonNKOstranderMMUlrich-LaiYMHermanJP2007Bed nucleus of the stria terminalis subregions differentially regulate hypothalamic-pituitary-adrenal axis activity: implications for the integration of limbic inputs. Journal of Neuroscience272025–2034. (doi:10.1523/JNEUROSCI.4301-06.2007)
DautzenbergFMPy-LangGHigelinJFischerCWrightMBHuberG2001Different binding modes of amphibian and human corticotropin-releasing factor type 1 and type 2 receptors: evidence for evolutionary differences. Journal of Pharmacology and Experimental Therapeutics296113–120.
GirottiMWeinbergMSSpencerRL2009Diurnal expression of functional and clock-related genes throughout the rat HPA axis: system-wide shifts in response to a restricted feeding schedule. American Journal of Physiology: Endocrinology and Metabolism296E888–E897. (doi:10.1152/ajpendo.90946.2008)
GoldPWLicinioJWongMLChrousosGP1995Corticotropin releasing hormone in the pathophysiology of melancholic and atypical depression and in the mechanism of action of antidepressant drugs. Annals of the New York Academy of Sciences771716–729. (doi:10.1111/j.1749-6632.1995.tb44723.x)
HasegawaKWakinoSKimotoMMinakuchiHFujimuraKHosoyaKKomatsuMKanekoYKandaTTokuyamaH2013The hydrolase DDAH2 enhances pancreatic insulin secretion by transcriptional regulation of secretagogin through a Sirt1-dependent mechanism in mice. FASEB Journal272301–2315. (doi:10.1096/fj.12-226092)
HermanJPFigueiredoHMuellerNKUlrich-LaiYOstranderMMChoiDCCullinanWE2003Central mechanisms of stress integration: hierarchical circuitry controlling hypothalamo-pituitary-adrenocortical responsiveness. Frontiers in Neuroendocrinology24151–180. (doi:10.1016/j.yfrne.2003.07.001)
MeisterBCortesRVillarMJSchallingMHokfeltT1990Peptides and transmitter enzymes in hypothalamic magnocellular neurons after administration of hyperosmotic stimuli: comparison between messenger RNA and peptide/protein levels. Cell and Tissue Research260279–297. (doi:10.1007/BF00318631)
MerchenthalerIVighSPetruszPSchallyAV1983The paraventriculo-infundibular corticotropin releasing factor (CRF) pathway as revealed by immunocytochemistry in long-term hypophysectomized or adrenalectomized rats. Regulatory Peptides5295–305. (doi:10.1016/0167-0115(83)90287-2)
MulderJSpenceLTortorielloGDinieriJAUhlenMShuiBKotlikoffMIYanagawaYAujardFHokfeltT2010Secretagogin is a Ca2+-binding protein identifying prospective extended amygdala neurons in the developing mammalian telencephalon. European Journal of Neuroscience312166–2177. (doi:10.1111/j.1460-9568.2010.07275.x)
OwensMJBartolomeJSchanbergSMNemeroffCB1990Corticotropin-releasing factor concentrations exhibit an apparent diurnal rhythm in hypothalamic and extrahypothalamic brain regions: differential sensitivity to corticosterone. Neuroendocrinology52626–631. (doi:10.1159/000125655)
SalataRAJarrettDBVerbalisJGRobinsonAG1988Vasopressin stimulation of adrenocorticotropin hormone (ACTH) in humans. In vivo bioassay of corticotropin-releasing factor (CRF) which provides evidence for CRF mediation of the diurnal rhythm of ACTH. Journal of Clinical Investigation81766–774. (doi:10.1172/JCI113382)
SwansonLWKuypersHG1980The paraventricular nucleus of the hypothalamus: cytoarchitectonic subdivisions and organization of projections to the pituitary, dorsal vagal complex, and spinal cord as demonstrated by retrograde fluorescence double-labeling methods. Journal of Comparative Neurology194555–570. (doi:10.1002/cne.901940306)
TramuGCroixCPillezA1983Ability of the CRF immunoreactive neurons of the paraventricular nucleus to produce a vasopressin-like material. Immunohistochemical demonstration in adrenalectomized guinea pigs and rats. Neuroendocrinology37467–469. (doi:10.1159/000123595)
Van PettKViauVBittencourtJCChanRKLiHYAriasCPrinsGSPerrinMValeWSawchenkoPE2000Distribution of mRNAs encoding CRF receptors in brain and pituitary of rat and mouse. Journal of Comparative Neurology428191–212. (doi:10.1002/1096-9861(20001211)428:2<191::AID-CNE1>3.0.CO;2-U)
van RossumEFVoorhoevePGte VeldeSJKoperJWDelemarre-van de WaalHAKemperHCLambertsSW2004The ER22/23EK polymorphism in the glucocorticoid receptor gene is associated with a beneficial body composition and muscle strength in young adults. Journal of Clinical Endocrinology and Metabolism894004–4009. (doi:10.1210/jc.2003-031422)
WagnerLOliyarnykOGartnerWNowotnyPGroegerMKasererKWaldhauslWPasternackMS2000Cloning and expression of secretagogin, a novel neuroendocrine- and pancreatic islet of Langerhans-specific Ca2+-binding protein. Journal of Biological Chemistry27524740–24751. (doi:10.1074/jbc.M001974200)
WattsAGTanimuraSSanchez-WattsG2004Corticotropin-releasing hormone and arginine vasopressin gene transcription in the hypothalamic paraventricular nucleus of unstressed rats: daily rhythms and their interactions with corticosterone. Endocrinology145529–540. (doi:10.1210/en.2003-0394)