Hubs and spokes of the lateral hypothalamus: cell types, circuits and behaviour

Abstract

The hypothalamus is among the most phylogenetically conserved regions in the vertebrate brain, reflecting its critical role in maintaining physiological and behavioural homeostasis. By integrating signals arising from both the brain and periphery, it governs a litany of behaviourally important functions essential for survival. In particular, the lateral hypothalamic area (LHA) is central to the orchestration of sleep–wake states, feeding, energy balance and motivated behaviour. Underlying these diverse functions is a heterogeneous assembly of cell populations typically defined by neurochemical markers, such as the well‐described neuropeptides hypocretin/orexin and melanin‐concentrating hormone. However, anatomical and functional evidence suggests a rich diversity of other cell populations with complex neurochemical profiles that include neuropeptides, receptors and components of fast neurotransmission. Collectively, the LHA acts as a hub for the integration of diverse central and peripheral signals and, through complex local and long‐range output circuits, coordinates adaptive behavioural responses to the environment. Despite tremendous progress in our understanding of the LHA, defining the identity of functionally discrete LHA cell types, and their roles in driving complex behaviour, remain significant challenges in the field. In this review, we discuss advances in our understanding of the neurochemical and cellular heterogeneity of LHA neurons and the recent application of powerful new techniques, such as opto‐ and chemogenetics, in defining the role of LHA circuits in feeding, reward, arousal and stress. From pioneering work to recent developments, we review how the interrogation of LHA cells and circuits is contributing to a mechanistic understanding of how the LHA coordinates complex behaviour.

Abbreviations

Arc

arcuate nucleus

BLA

basolateral amygdala

BNST

bed nucleus of the stria terminalis

BNST^GABA

(GABAergic neurons in the BNST)

CART

cocaine‐ and amphetamine regulated transcript

CeA

central nucleus of the amygdala

ChR2

channelrhodopsin‐2

CRH

corticotropin‐releasing hormone

DA

dopamine

DR

dorsal raphe

Dyn

dynorphin

Enk

enkephalin

Fx

fornix

Gal

galanin

Hcrt/Ox

hypocretin/orexin

HPA

hypothalamic‐pituitary‐adrenal

Iapp

islet amyloid polypeptide

IR

immunoeactive

LC

locus coeruleus

LC^NA

noradrenergic neurons in the LC

LepRb

long‐form of the leptin receptor

LHA

lateral hypothalamic area

LHA^GABA

GABAergic neurons in the LHA

LHA^Glu

glutamatergic neurons in the LHA

LHA^LepRb

LepRb‐expressing neurons in the LHA

LHb

lateral habenula

LS

lateral septal nuclei

MC4R

melanocortin receptor 4

MCH

melanin‐concentrating hormone

MRF

midbrain reticular formation

MSN

medium spiny neuron

MT

mammillothalamic tract

NA

noradrenaline

NAc

nucleus accumbens

NAc^D1R/GABA

(GABAergic D1R‐expressing medium spiny neurons in the NAc)

NTS

nucleus of the solitary tract

Nts

neurotensin

N/OFQ

nociceptin/orphanin FQ

PAG

periaqueductal grey

PB

parabrachial nucleus

PeF

perifornical

PFC

prefrontal cortex

POA^GABA

(GABAergic neurons in the hypothalamic preoptic area)

PV

parvalbumin

PVN

paraventricular nucleus of the hypothalamus

PVT

paraventricular nucleus of the thalamus

RMg

nucleus raphe magnus

TH

tyrosine hydroxylase

TMN

tuberomammillary nucleus

TMN^HA

(histaminergic neurons in the TMN)

TRH

thryotropin‐releasing hormone

TRN^GABA

(GABAergic neurons in the thalamic reticular nucleus)

UNC3

urocortin‐3

VGAT

vesicular GABA transporter

VGLUT2

vesicular glutamate transporter‐2

VTA

ventral tegmental area

VTA^DA

dopaminergic neurons in the VTA

The hypothalamus maintains homeostatic balance in physiology and behaviour by integrating vast amounts of neural and humoral information and conveying appropriate instructions to downstream effector systems. Hypothalamic function relies on the coordinated output of a highly differentiated network of hypothalamic nuclei best understood at the level of cells, circuits and transmitter systems (Saper et al. 2005, 2010; Sternson 2013; Gautron et al. 2015). Through classical neuroanatomical and behavioural work, the lateral hypothalamic area (LHA) emerged as a linchpin in the regulation of key aspects of vertebrate behaviour including arousal, feeding, energy balance, stress, reward and motivated behaviour. Steady progress over decades has been punctuated by seminal discoveries concerning the molecular, cellular and systems‐level substrates of LHA function (Berthoud & Münzberg, 2011; Burdakov et al. 2013; Hurley & Johnson, 2014; Brown et al. 2015; Stuber & Wise, 2016).

Despite significant inroads, the basic mechanisms underlying the function of LHA circuits at the cellular and systems levels are poorly understood, a gap attributable to the daunting heterogeneity, cytoarchitectural complexity and behavioural state‐dependent function of these circuits. Now, all the firepower of modern neuroscience is rapidly advancing our understanding of LHA circuits. New molecular and genetic tools, together with evolving multidisciplinary approaches such as optogenetics, chemogenetics, in vivo electrophysiology and deep brain calcium imaging, are transforming our understanding of LHA circuits and their role in health and disease. Our goal in this review is to overview the cellular and circuit‐level architecture of the LHA, with a focus on recent advances in the field that are beginning to disentangle the multifaceted identity and function of its constituent neuronal cell types and circuits.

Deconstructing the cellular diversity of LHA neurons

Despite its reticular cytoarchitecture, the LHA is a functionally discrete structure within the hypothalamus comprised of an intricate assembly of cell types and circuits, with unique patterns of inputs and outputs that interconnect with numerous physiological systems (Saper et al. 1979; Nieuwenhuys et al. 1982; Luiten et al. 1987; Geeraedts et al. 1990; Bernardis & Bellinger 1993; Bernardis & Bellinger 1996; Swanson et al. 2005; Hahn & Swanson 2010). A central question in attempting to disentangle cell types and circuits in the LHA, and in fact any brain region, is defining what constitutes a discrete cell type. Neuronal cell types may be defined by some confluence of cell type‐specific molecular markers, developmental history, capacity for intercellular signalling, intrinsic electrophysiological properties, unique morphology and patterns of local and long‐range synaptic inputs and outputs (Nelson et al. 2006; Bota & Swanson 2007). LHA neurons are broadly identified by their expression of specific markers, such as the neuropeptides hypocretin/orexin (Hcrt/Ox) (de Lecea et al. 1998; Sakurai et al. 1998) or melanin‐concentrating hormone (MCH) (Bittencourt et al. 1992). However, accumulating evidence suggests that single markers fail to adequately account for LHA cell type diversity in terms of gene expression, connectivity, circuit‐level function and role in behaviour. Another key question is precisely how circuit interactions within the LHA integrate central and peripheral signals and mobilize downstream targets in a concerted, specific and adaptive manner during different behavioural states. What follows is a summary of major LHA cell populations, with an emphasis on classifying neurons according to neurochemical phenotype, and recent advances in our understanding of local circuit interactions among defined cell types.

Two major neuropeptidergic neuron populations in the LHA

Hypocretin/orexin neurons

Neuropeptidergic projection neurons expressing Hcrt/Ox appear to be a distinct cell population, uniquely found within the LHA and particularly concentrated in the perifornical (PeF) region of the LHA (de Lecea et al. 1998; Sakurai et al. 1998). Hcrt/Ox neurons are the best characterized LHA population at the molecular, cellular and systems levels (Sakurai, 2007, 2014; Bonnavion & de Lecea, 2010; Sakurai & Mieda, 2011; Alexandre et al. 2013; Jones & Hassani, 2013; Shan et al. 2015). While defined by their expression of Hcrt/Ox, these neurons overlap significantly with other neuropeptidergic LHA populations and are likely to be capable of co‐release of multiple neuropeptides and transmitters. A majority of Hcrt/Ox neurons co‐express the neuropeptide dynorphin (Dyn) (Chou et al. 2001; Muschamp et al. 2014). Hcrt/Ox neurons are also thought to co‐express nociceptin/orphanin FQ (N/OFQ) (Maolood & Meister, 2010), although other evidence suggests that N/ORQ fibers innervate Hcrt/Ox neurons (Xie et al. 2008) and that these populations are intermingled but non‐overlapping within the LHA (Geraschenko et al. 2011). A small subpopulation of Hcrt/Ox neurons appears to overlap with galanin (Gal) (Hakansson et al. 1999), although this finding was not recapitulated in a Gal‐EGFP transgenic mouse (Laque et al. 2013). Recent evidence also suggests that Hcrt/Ox neurons co‐express neurotensin (Nts) (Furutani et al. 2013); however, these results are inconsistent with data describing no overlap between these populations (Watts & Sanchez‐Watts, 2007; Leinninger et al. 2011). Interestingly, Hcrt/Ox neurons were recently reported to co‐express the pancreatic peptide islet amyloid polypeptide (Iapp), a precursor to amylin (Li et al. 2015) (Fig. 1). Further work will be required to resolve the full extent of neuropeptide co‐expression within Hcrt/Ox neurons.

Figure 1. Neuronal cell type diversity in the LHA .

Schematic representation of numerous overlapping neuronal cell types in the LHA that can be classified on the basis of specific neurochemical markers. Three major subpopulations can be identified based on the expression of neuropeptides: Hcrt/Ox, MCH and an apparently separate, though mixed population that represents a confluence of multiple neuropeptide and receptor markers. The Hcrt/Ox population significantly overlaps with Dyn, N/OFQ and amylin, the MCH population largely overlaps with nesfatin and CART and the third population includes overlapping subpopulatons of LepRb, Nts, MC4R, CART and Gal‐expressing neurons. Also included are poorly characterized cell types found in the LHA; UNC3, TRH, Enk, TH, CRH and PV. Furthermore, these largely neuropeptidergic subpopulations are thought to also express the machinery for the release of fast amino acid transmitters GABA or glutamate. The varying degree to which these three main subpopulations, and others, can be characterized as functionally GABAergic or glutamatergic is still the subject of debate and not represented in this diagram. Note that the representation is meant to be illustrative and not quantitative with regard to cell numbers or degree of co‐expression. The following abbreviations are used: CART, cocaine‐ and amphetamine regulated transcript; CRH, corticotropin‐releasing hormone; Dyn, dynorphin; Enk, enkephalin; Gal, galanin; Hcrt/Ox, hypocretin/orexin; LepRb, long‐form of the leptin receptor; MC4R, melanocortin receptor 4; MCH, melanin‐concentrating hormone; Nts, neurotensin, N/OFQ, nociceptin–orphanin FQ; PV, parvalbumin; TH, tyrosine hydroxylase; TRH, thyrotropin‐releasing hormone; UCN3, urocortin‐3.

Importantly, Hcrt/Ox neurons also appear capable of fast, amino acid neurotransmitter release. The synaptic terminals of Hcrt/Ox neurons co‐localize with markers of glutamatergic neurotransmission, including the vesicular glutamate transporter VGLUT2, suggesting they are capable of rapid, synaptic glutamate release (Abrahamson et al. 2001; Rosin et al. 2003; Torrealba et al. 2003; Henny et al. 2010). Additionally, Hcrt/Ox varicosities in the locus coeruleus (LC) were found to end in excitatory asymmetric synapses (Horvath et al. 1999; Henny et al. 2010). Functional evidence for co‐release of glutamate and Hcrt/Ox was recently demonstrated using optogenetics, whereby low‐frequency photostimulation of channelrhodopsin‐2 (ChR2)‐expressing Hcrt/Ox fibres induced glutamate release onto histaminergic (HA) neurons in the tuberomammillary nucleus (TMN) (Schöne et al. 2012) and high‐frequency stimulation could additionally induce Hcrt/Ox release (Schöne et al. 2014). Co‐transmission among these neurons has been proposed to play a critical role in the stability and time course of postsynaptic excitability (Schöne & Burdakov, 2012; Schöne et al. 2014; Kosse & Burdakov, 2014).

While Hcrt/Ox neurons are generally considered to have a glutamatergic phenotype, some evidence suggests that a subpopulation expresses the machinery for GABA synthesis, based on gene expression (Harthoorn et al. 2005) and the morphology of Hcrt/Ox varicosities (Guan et al. 2002; Balcita‐Pedicino & Sesack, 2007), raising the possibility of some variation in neurotransmitter phenotype. Surprising recent work demonstrated that optogenetic activation of Hcrt/Ox neurons profoundly suppressed firing in nearby MCH neurons through light‐evoked inhibitory postsynaptic currents that were unaffected by glutamate receptor blockade, attenuated by Hcrt/Ox receptor blockade and abolished by gabazine (Apergis‐Schoute et al. 2015). Such inhibition could be explained by local Hcrt/Ox release, leading to the activation of local GABAergic neurons, or the release of GABA by Hcrt/Ox neurons themselves, although this result would need to be evaluated under conditions of exclusively monosynaptic release. Interestingly, immunohistochemical detection of GABA was found in close to 20% of Hcrt/Ox neurons, although these cells were devoid of the vesicular GABA transporter, VGAT (Apergis‐Schoute et al. 2015). Evidence for the regulation of excitability through local feed‐forward inhibitory (Matsuki et al. 2009) and excitatory (Li et al. 2002) inputs has been put forward; however, further work is required to parse the role of local circuits from differential transmitter release with regard to synaptic interactions between Hcrt/Ox and MCH neurons.

Beyond neuropeptides and transmitters, the cellular identity of Hcrt/Ox neurons can further be defined by other key markers such as transcription factors. Lhx9, a LIM‐homeodomain‐containing transcription factor, has emerged as a potential specifier for Hcrt/Ox neurons, as revealed though molecular profiling and developmental analysis in multiple species (Shimogori et al. 2010; Dalal et al. 2013; Liu et al. 2015; Sokolowski et al. 2015). Recent RNA‐sequencing of zebrafish Hcrt neurons confirmed the presence of Lhx9 and identified other key markers enriched in these neurons (Yelin‐Bekerman et al. 2015). Additionally, many orexigenic LHA subpopulations including Hcrt/Ox neurons express Dbx1, another developmentally regulated homeodomain‐containing transcription factor (Sokolowski et al. 2015).

Accumulating evidence suggests that even the small population of Hcrt/Ox neurons may be diverse, evidenced by divergent anatomical and electrophysiological characteristics. For example, Hcrt/Ox neurons exhibit two distinct electrophysiological phenotypes, so‐called D‐type and H‐type (Schöne et al. 2011). Also, Hcrt/Ox cells exhibit differential responses to neurotransmitters such as cholinergic agonists (Sakurai et al. 2005; Ohno et al. 2008). Some investigators have proposed a dichotomy in Hcrt/Ox function based on their topology and projection patterns between medial and lateral zones (Harris & Aston‐Jones, 2006), although others report a rather diffuse pattern of Hcrt/Ox projections (González et al. 2012). In addition, gene expression analysis in both mice and zebrafish suggests cell type heterogeneity (Appelbaum et al. 2007; Dalal et al. 2013; Yelin‐Bekerman et al. 2015). Given the diversity of cellular properties, neurochemical content and the myriad roles that they play in behaviour, Hcrt/Ox neurons are unlikely to be monolithic. Further work is required to put cell type heterogeneity in register with functional diversity.

Melanin‐concentrating hormone

A separate population of LHA projection neurons expresses MCH (Bittencourt et al. 1992), which is well‐described anatomically and physiologically (Broberger et al. 1998; Elias et al. 1998; Peyron et al. 1998; Bayer et al. 2002; Guan et al. 2002; van den Pol et al. 2004). MCH and Hcrt/Ox neurons are intermingled throughout the LHA and surrounding regions, but each population exhibits a distinct pattern of localization (Hahn, 2010; Hahn & Swanson, 2010). Similar to Hcrt/Ox neurons, MCH neurons co‐express a variety of neuropeptides, such as nesfatin‐1 (Oh‐I et al. 2006; Foo et al. 2008; Fort et al. 2008). Additionally, approximately 50% of MCH neurons co‐express cocaine‐ and amphetamine‐regulated transcript (CART) (Broberger, 1999; Vrang et al. 1999; Elias et al. 2001; Hathroon et al. 2005; Hanriot et al. 2007) (Fig. 1). The transcription factor Dbx1 can also be found in MCH‐expressing neurons (Sokolowski et al. 2015), while Lhx9 is absent (Dalal et al. 2013).

MCH neurons are generally considered to be GABAergic owing to their co‐localization with two isoforms of glutamic acid decarboxylase, GAD65 and GAD67, the synthetic enzymes for GABA (Elias et al. 2001, 2008; Harthoorn et al. 2005; Sapin et al. 2010). In rat, the vast majority of MCH immunoreactive (IR) neurons co‐express GAD67 (Elias et al. 2008; Sapin et al. 2010). Approximately 67% of hypothalamic CART‐expressing neurons express GAD67 mRNA and LHA CART neurons exhibit extensive overlap with MCH neurons (Elias et al. 2001). In addition, VGAT was found to be co‐localized in approximately 6% of MCH‐IR varicosities found in the LC (del Cid‐Pellitero & Jones, 2012). Finally, optogenetic stimulation of MCH cells has been shown to result in GABA release onto HA neurons in the TMN (Jego et al. 2013), lending support to the anatomical evidence that MCH neurons are functionally GABAergic.

However, some MCH neurons have been shown to express the vesicular glutamate transporters VGLUT1 and VGLUT2 (Abrahamson et al. 2001; Harthoorn et al. 2005; del Cid‐Pellitero, 2012). In support of this anatomical evidence, recent work shows that MCH neurons projecting to the lateral septum co‐express VGLUT2 and surprisingly exhibit photostimulation‐evoked monosynaptic release of glutamate, not GABA (Chee et al. 2015). Additionally, LHA GABAergic neurons, as revealed using Vgat‐Cre (Vong et al. 2011) or Gad65‐EGFP transgenic mice, do not overlap with MCH‐IR neurons (Chee et al. 2015; Jennings et al. 2015; Karnani et al. 2013). These data raise intriguing questions about neurochemical phenotype diversity and perhaps target‐specific signalling of MCH neurons. MCH and other neuropeptide‐expressing LHA cell populations thus appear more diverse and dynamic than previously appreciated.

Identification of a third, distinct LHA neuron population with multiple phenotypes

Leptin‐sensitive neuropeptidergic neurons in the LHA

A recent body of work has elucidated a novel population of LHA neurons identified through both hormone receptor and neuropeptide expression. The anorexigenic, adipose tissue‐derived hormone leptin, produced by the ob gene, is a major peripheral signal modulating hypothalamic functions including feeding, energy expenditure and endocrine function. Leptin crosses the blood–brain barrier, binds to the long isoform of the leptin receptor (LepRb) and engages the Janus kinase–signal transducer and activator of transcription (JAK–STAT) signal transduction pathway (Leinninger & Myers, 2008). LepRb‐expressing neurons are widely expressed but are enriched in the hypothalamus and brainstem (Elmquist et al. 1998). The generation of strains of Lepr‐Cre driver lines (DeFalco et al. 2001; Leshan et al. 2006) and identification of labelled LepRb‐expressing neurons in the brain (Scott et al. 2009; Louis et al. 2010; Patterson et al. 2011; Leinninger et al. 2011) opened the door to more detailed anatomical analysis and cell type‐specific genetic manipulations of these neurons.

Interestingly, LHA LepRb neurons may comprise a unique cell population that modulates LHA network function and behaviour (Myers et al. 2009; Leinninger et al. 2009, 2011; Louis et al. 2010; Leinninger, 2011; Goforth et al. 2014; Bonnavion et al. 2015). LepRb‐expressing neurons are diverse and, within the LHA, overlap with multiple populations of known neuropeptidergic neurons. Approximately 60% of the LHA LepRb population co‐expresses Nts; conversely, about 30% of LHA Nts neurons co‐express LepRb (Leinninger et al. 2011). Interestingly, 95% of this Nts–LepRb population co‐localizes with Gal, whereas the proportion of LHA LepRb neurons expressing Gal is only around 20–44% (Laque et al. 2013) (Fig. 1).

LepRb neurons in the LHA are also thought to be GABAergic. Studies using Gad67‐EGFP reporter mice revealed leptin‐induced activation of STAT3 signalling in EGFP‐positive LHA neurons (Leinninger et al. 2009). Furthermore, in leptin‐injected Vgat‐Cre mice, phosphorylated STAT3 (pSTAT3)‐IR cells could be detected in putative LHA GABAergic neurons, similar to the arcuate nucleus (Arc), and dorsomedial hypothalamus (DMH) (Vong et al. 2011). This work suggests that LepRb neurons in the LHA act via GABAergic mechanisms to exert inhibitory influence on their targets. Interestingly, given the strong co‐localization between LHA LepRb and Nts neurons (Leinninger et al. 2011), one might predict that specific Nts‐ergic inputs to the VTA are GABAergic. However, in the presence of GABA_AR blockers, optogenetic stimulation induced the release of glutamate from ChR2‐expressing LHA Nts projections onto VTA targets (Kempadoo et al. 2013), underscoring the apparent neurochemical diversity of LHA Nts neurons.

Adding further complexity to our understanding of LHA LepRb neurons is the existence of LHA neurons expressing the melanocortin‐4 receptor (MC4R), a G‐protein‐coupled receptor for α‐melanocyte stimulating hormone (α‐MSH) (Elmquist et al. 1999; Schwartz et al. 2000; Cone, 2005). This MC4R‐positive population is distinct from Hcrt/Ox, MCH and nesfatin‐1‐positive neurons (Liu et al. 2003; Cui et al. 2012). Approximately 75% of MC4R‐positive neurons co‐localize with Nts; this Nts–MC4R‐expressing neuron population appears unique to the LHA. Cui and co‐authors showed that ∼80% of MC4R cells co‐localize with pSTAT3 following leptin injection, indicating that these cells are also leptin sensitive (Cui et al. 2012) and suggesting a strong overlap between neurons defined by LepRb and those defined by MC4R. Interestingly, while Nts–LepRb‐positive cells send strong projections to the dorsal raphe (DR) and ventral tegmental area (VTA) (Leinninger et al. 2011), MC4R–pSTAT3‐positive cells do not send projections to either region (Cui et al. 2012). This latter result suggests that LHA LepRb neurons co‐expressing MC4R are a distinct subpopulation, as only about one‐third of LHA LepRb neurons co‐expresses MC4R (Cui et al. 2012) (Fig. 1). MCR4 neurons are also thought to exhibit a GABAergic phenotype evidenced by their co‐localization with GAD67 (Liu et al. 2003).

Using trans‐synaptic tracing techniques in a Lepr‐Cre knock‐in mouse, Louis and co‐workers showed that local LHA LepRb neurons appear to make direct synaptic contact with a subpopulation of Hcrt/Ox neurons (30%) (Louis et al. 2010). Subsequently, we found that optogenetic photostimulation of LepRb neurons in slices yielded GABA_AR‐mediated synaptic inputs in approximately 27.5% of identified Hcrt/Ox neurons in slices (Bonnavion et al. 2015). Although virally transduced LepRb neurons could also be found in nearby hypothalamic regions outside of the LHA, our data demonstrate that LepRb‐expressing GABAergic neurons make functional synaptic contact with a subpopulation of Hcrt/Ox neurons. This functional connection may contribute to the leptin‐mediated suppression in neuronal activity observed in Hcrt/Ox neurons following both leptin infusion and optogenetic activation of LepRb neurons in vivo (Bonnavion et al. 2015).

While signalling between LepRb‐expressing neurons and Hcrt/Ox neurons may, in part, be mediated by GABAergic mechanisms, recent evidence points to a contribution of GABA‐independent mechanisms. A population of LepRb–Nts neurons within the LHA, described earlier, appears to be in direct synaptic contact with a subpopulation of Hcrt/Ox neurons, not MCH neurons (Leinninger et al. 2011). Goforth and colleagues examined the functional connectivity between LHA Nts neurons and Hcrt/Ox neurons and found that leptin, which activates a large subpopulation of LHA Nts neurons (Leinninger et al. 2011), hyperpolarizes Hcrt/Ox neurons even in the presence of blockers of GABAergic transmission (Goforth et al. 2014). As mentioned, the LHA Nts population overlaps significantly with Gal‐expressing neurons in the LHA (Laque et al. 2013). Accordingly, bath application of Gal inhibits Hcrt/Ox neurons and Gal receptor blockade prevents leptin‐induced inhibition of Hcrt/Ox neurons (Goforth et al. 2014). Interestingly, chemogenetic activation of LHA Nts neurons similarly resulted in hyperpolarization of Hcrt/Ox neurons (Goforth et al. 2014) despite evidence that Nts itself has a strong, excitatory effect on Hcrt/Ox neurons (Tsujino et al. 2005; Furutani et al. 2013). The nature of selective neuropeptide co‐transmission by LHA LepRb/Nts/Gal populations and the relative contribution of neuropeptide and GABA co‐transmission on postsynaptic excitability remain open questions.

A larger population of LHA GABAergic neurons

Neuroanatomical evidence suggests that the LHA contains a large population of GABAergic neurons defined by the presence of components necessary for GABA synthesis and release, including GAD65, GAD67 and VGAT (Rosin et al. 2003; van den Pol et al. 2004; Shin et al. 2007; Hassani et al. 2010; Karnani et al. 2013; Jennings et al. 2015; Chee et al. 2015). How this heterogeneous population of GABAergic neurons tracks with known neuropeptide‐expressing LHA populations, corresponds to specific inputs and outputs of the LHA, and interacts locally with other cell types remains poorly understood. Recently, a population of GAD65‐expressing neurons was identified in the LHA that is distinct from both MCH and Hcrt/Ox populations. These cells exhibit a range of intrinsic firing patterns, sub‐divided into four subtypes based on their distinct electrical signatures (Karnani et al. 2013). Furthermore, a role for local inhibition of LHA projection neurons is suggested by glutamate drop experiments (Huang et al. 2007) and transgenic mice lacking GABA_BRs in Hcrt/Ox neurons (Matsuki et al. 2009). Recent electrophysiological, imaging and behavioural data, discussed in the second half of this review, have begun to illuminate the physiological function and significance of these LHA GABAergic neurons.

Other neurochemically diverse neuron populations in the LHA

The LHA is home to a variety of other neuron populations that do not appear to track with better known cell types. These include neuron populations defined by their expression of neuropeptides such as thyrotropin‐releasing hormone (TRH), enkephalin (Enk) and urocortin‐3 (UCN3) (Fig. 1). For example, LHA TRH‐expressing neurons do not appear to co‐localize with the neuropeptides that define other major cell populations; however, a significant population overlap with Enk‐IR (16%) and UCN3‐IR (42%) cells in the juxtaparaventricular region (Horjales‐Araujo et al. 2014). Other markers appear to specify other unique cell populations in the LHA such as tyrosine hydroxylase (TH) (van den Pol et al. 1984; Abrahamson & Moore, 2001; Horjales‐Araujo et al. 2014), an RF‐amide peptide (QRFP) (Takayasu et al. 2006; Chen et al. 2016) or the calcium‐binding protein parvalbumin (PV), which defines a small population of glutamatergic neurons in the LHA (Mészár et al. 2012).

Furthermore, recent evidence using a Cre driver line to identify neurons expressing VGLUT2 (Vong et al. 2011) has revealed a population of glutamatergic LHA neurons that are functionally important (Jennings et al. 2013; Nieh et al. 2015; Stamatakis et al. 2016), as discussed in the following section. Like their GABAergic counterparts, LHA glutamatergic neurons are likely to be highly diverse and future work will be necessary to parse the neurochemical phenotype of functionally distinct LHA glutamatergic cell populations and their role in behaviour.

Finally, an important caveat in defining LHA cell type heterogeneity is the distinct possibility that the neurochemical identity of LHA neurons may be dynamic and depend on developmental factors, behavioural state and metabolic conditions. This form of neurotransmitter switching or respecification, in response to changing environmental conditions, has been described in a variety of species, systems and preparations (Spitzer, 2015). Thus, the boundaries between distinct cell types in the LHA may be blurred by, for example, behavioural state‐dependent changes in activity and shifts in gene expression. The extent to which this occurs in the LHA is still an open question but future work may require an examination of coordinated patterns of gene expression in order to identify discrete cell populations.

Disentangling long‐range LHA connectivity and its role in complex behaviour

Pioneering early lesion and electrical stimulation experiments established the key role of the LHA in feeding, drinking, motivated behaviour and arousal. Lesions of the LHA induced the ‘lateral hypothalamic syndrome’ characterized by hypophagia, adipsia, hypoactivity and weight loss. LHA lesions were also associated with somnolence, sensory neglect, decrease in sucrose consumption and neuroendocrine and autonomic defects (Anand & Brobeck, 1951; Teitelbaum & Epstein, 1962; Hoebel, 1965; Boyle & Keesey, 1975; Luiten et al. 1987; Bernardis & Bellinger, 1996; Hurley & Johnson, 2014; Brown et al. 2015; Castro et al. 2015; Stuber & Wise, 2016). Conversely, electrical stimulation of the LHA resulted in voracious eating and weight gain (Delgado & Anand, 1953; Hoebel & Teitelbaum, 1962; Margules & Olds, 1962), promoted self‐stimulation and had potent rewarding effects (Olds & Milner, 1954; Gallistel et al. 1981; Bernardis & Bellinger, 1996; Stuber & Wise, 2016). Stimulation‐evoked hyperphagia in rats was also associated with increased motivation and improved learning in food reward tasks, suggesting that the LHA is a ‘pacemaker for food motivation’ (Bernardis & Bellinger, 1993).

Given its central role in a wide range of behaviours, it is unsurprising that the LHA extensively interconnects with multiple brain regions. The first descriptions of LHA efferent projections revealed a broad and diffuse arborization from the telencephalon to the medulla and spinal cord. Ascending fibres project to forebrain targets that have functional implications in cognitive and emotional processes. In contrast, descending fibres project to nuclei involved in wakefulness, feeding, autonomic and neuroendocrine functions, in addition to prominent intrahypothalamic projections (Saper et al. 1979; Luiten & Room, 1980; Berk & Finkelstein, 1982; Ter Horst & Luiten, 1987; Luiten et al. 1987; Bernardis & Bellinger, 1993, 1996; Hahn & Swanson, 2010; Berthoud & Münzberg, 2011; Brown et al. 2015) (Fig. 2). Many of the structures targeted by the LHA, in turn, exert feedback control onto the LHA. Major forebrain afferents to the LHA originate in the medial prefrontal cortex (mPFC), lateral septal nuclei (LS), nucleus accumbens (NAc), bed nucleus of the stria terminalis (BNST), basolateral amygdala (BLA) and lateral habenular nucleus (LHb). Midbrain and brainstem afferents are also substantial, notably from the mesencephalic and central grey, VTA, DR, LC and lateral parabrachial area (LPB) (Saper et al. 1979; Barone et al. 1981; Kita & Oomura, 1982; Luiten et al. 1987; Bernardis & Bellinger, 1993, 1996; Sakurai et al. 2005; Yoshida et al. 2006; Hahn & Swanson, 2010; Berthoud & Münzberg, 2011; Brown et al. 2015; Castro et al. 2015; González et al. 2016) (Fig. 2). The advent of powerful new genetic tools has transformed our ability to specifically probe LHA circuitry and its role in complex behaviours. What follows is a summary of the role of defined LHA cell populations in feeding, reward, arousal and stress. The roles of Hcrt/Ox and MCH neurons in behaviour have been detailed in many excellent reviews, cited below, and we will discuss some highlights of this work. Emphasis, however, is placed on recent advances in our understanding of other LHA cell types, including LHA GABAergic and glutamatergic neurons among others, as revealed through opto‐ and chemogenetic approaches.

Figure 2. Long‐range connectivity of the LHA as a substrate for complex behaviour .

Long‐range anatomical inputs and outputs of the LHA reflect the engagement of multiple physiological systems to regulate fundamental behavioural programs. A, schematic of a parasagittal rodent brain illustrating major long‐range inputs (green lines) to the LHA. B, similar schematic illustrating major long‐range outputs (red lines) from the LHA. The following abbreviations are used: Arc, arcuate nucleus; BLA, basolateral amygdala; BNST, bed nucleus of the stria terminalis; CeA, central nucleus of the amygdala; DR, dorsal raphe; Fx, fornix; LC, locus coeruleus; LHA, lateral hypothalamic area; LHb, lateral habenula; LS, lateral septal nuclei; MRF, midbrain reticular formation; MT, mammillothalamic tract; NAc, nucleus accumbens; NTS, nucleus of the solitary tract; PAG, periaqueductal grey; PB, parabrachial nucleus; PFC, prefrontal cortex; PVN, paraventricular nucleus of the hypothalamus; PVT, paraventricular nucleus of the thalamus; RMg, nucleus raphe magnus; TMN, tuberomammillary nucleus; VTA, ventral tegmental area.

Coordination of feeding and reward

The MCH and Hcrt/Ox systems were first examined for their role in feeding and regulating energy balance using pharmacological and genetic deletion approaches (Pissios et al. 2006; Sakurai, 2007, 2014; Bonnavion & de Lecea, 2010; Sakurai & Mieda, 2011; Alexandre et al. 2013; Shan et al. 2015; Burdakov et al. 2013; Brown et al. 2015). In particular, accumulating evidence suggests that the MCH system critically regulates energy homeostasis and reward. Intracerebral delivery of MCH promotes feeding (Qu et al. 1996) while genetic overexpression of MCH results in a hyperphagic and obese phenotype (Ludwig et al. 2001). Genetic deletion of MCH (Shimada et al. 1998) or loss of MCH neurons (Alon & Friedman, 2006) both result in a hypophagic and lean phenotype. In addition, loss of MCH also leads to perturbed mesolimbic dopamine (DA) function (Pissios et al. 2008). One important pathway through which the MCH system may be acting is on the NAc shell, where local administration of MCH enhanced feeding and an MCHR‐1 antagonist reduced feeding (Georgescu et al. 2005) and reduced cocaine self‐administration (Chung et al. 2009). In animal models of obesity, specific MCH antagonists have consistently shown efficacy in acutely reducing food intake and inhibiting weight gain (Macneil, 2013). However, a recent study dissecting the neural substrates of LHA‐induced feeding and reward excluded the involvement of MCH neurons in the voracious feeding and self‐stimulation resulting from optogenetic activation of LHA GABAergic neurons (LHA^GABA) (Jennings et al. 2015). Furthermore, the activity of MCH neurons has been implicated in the rewarding effects of sugar consumption, through its modulation of striatal DA release (Domingos et al. 2013).

The role of the Hcrt/Ox system in feeding is the subject of debate. Although the Hcrt/Ox system is sensitive to metabolic signals (Burt et al. 2011; Berthoud & Münzberg, 2011; Burdakov et al. 2013), effects of Hcrt/Ox dysfunction on energy homeostasis may rather be attributed to motivational deficits or decreased locomotor activity resulting from diminished arousal (Siegel, 2004; Bonnavion & de Lecea, 2010). Interestingly, one striking phenotype in Hcrt/Ox‐ablated mice is the absence of food‐anticipatory arousal resulting from food restriction (Yamanaka et al. 2003). These data suggest that the Hcrt/Ox system orchestrates adaptive motor and arousal‐related behaviour through the integration of metabolic changes and food cues. Further work is necessary to elucidate the circuitry and specific roles of Hcrt/Ox and MCH neurons in feeding by also focusing on associative learning, which may encode incentive responses to food reward. Evidence for the role of Hcrt/Ox neurons in reward includes the presence of excitatory connections to reward circuitry, notably dopamine (DA) and GABAergic neurons of the VTA (Korotkova et al. 2003; Beier et al. 2015). Further behavioural studies demonstrate the influence of Hcrt/Ox neurons on the rewarding effects of food (Borgland et al. 2009; Sharf et al. 2010), drugs of abuse (Borgland et al. 2006; Boutrel & de Lecea, 2008) and sexual behaviour (Muschamp et al. 2007). Recently, investigation of the LHA^Hcrt/Ox→VTA^DA pathway indicates that both Hcrt/Ox and Dyn are involved in modulation of VTA^DA activity and reward. These peptides are co‐expressed, and may be co‐released, by the same LHA neurons, but appear to have opposing actions on the excitability of VTA^DA neurons in vitro, and on reward in vivo (Muschamp et al. 2014). Finally, recent in vivo single‐unit recordings from Hcrt/Ox neurons revealed enhanced firing and cortical activation upon presentation of cues associated with delivery of appetitive (sucrose) vs. aversive (quinine) solutions, characteristic of neurons that encode both reward and arousal (Hassani et al. 2016).

Recent studies highlighted the important role of LHA glutamatergic and GABAergic neurons in the control of feeding and reward. First, Jennings and co‐workers dissected an inhibitory projection from the BNST to the LHA in the context of food intake using optogenetic techniques. Photostimulation of terminals in the LHA originating from BNST GABAergic neurons (BNST^GABA→LHA) induced both rapid and voracious feeding and self‐stimulation (Jennings et al. 2013). Interestingly, these responses varied with the nutritional status of the animals, similar to early LHA electrical stimulation experiments (Bernardis & Bellinger, 1996). Importantly, BNST^GABA neurons appeared to preferentially contact glutamatergic, VGLUT2‐expressing LHA neurons (LHA^Glu). They went on to show that direct photostimulation of LHA^Glu neurons suppressed feeding behaviour in food‐deprived mice and that their photoinhibition promoted feeding in well‐fed mice (Jennings et al. 2013). These data suggest that inhibitory BNST projections directly suppress the activity of anorexigenic LHA^Glu neurons, while disinhibiting the activity of orexigenic neurons that initiate feeding behaviour (Stuber & Wise, 2016). Subsequent work revealed that LHA^GABA neurons may in fact be orexigenic. Optogenetic stimulation of non‐Hcrt/Ox, non‐MCH, VGAT‐expressing LHA^GABA neurons enhances both food consumption and self‐stimulation. Conversely, their optogenetic inhibition or ablation results in decreased food intake. Through in vivo deep brain calcium imaging, the authors could divide these LHA^GABA neurons into two subgroups based on differential activity during feeding behaviour: one activated during the action to obtain the food reward (appetitive responses), and the other during food consumption itself (consummatory responses) (Jennings et al. 2015) (Fig. 3 A).

Figure 3. Summary of optogenetic interrogation of LHA circuits in feeding, arousal and stress behaviour .

Schematic illustration summarizing recent in vivo optogenetic manipulation of specific LHA cell types and circuits in behaviour using Cre driver lines or promoter‐driven viral expression. In each case, the outputs of Hcrt/Ox, MCH, GABAergic and glutamatergic populations, and putative subpopulations, are illustrated along with some relevant inputs. A, optogenetic investigation of feeding behaviour, overlapping with reward‐related behaviour. Note a putative subpopulation of LHA^GABA neurons defined by Pdx1. B, optogenetic interrogation of sleep and wake behaviour. Note a putative subpopulation of LHA^GABA neurons defined by LepRb. C, optogenetic manipulation of stress‐related behaviour, including LepRb neurons. The following abbreviations are used: BNST^GABA (GABAergic neurons in the bed nucleus of the stria terminalis); LC, locus coeruleus; LHb, lateral habenula; NAc^D1R/GABA (GABAergic D1R‐expressing medium spiny neurons in the nucleus accumbens); POA^GABA (GABAergic neurons in the hypothalamic preoptic area); PVN, paraventricular nucleus of the hypothalamus; TMN^HA (histaminergic neurons in the tuberomamillary nucleus); TRN^GABA (GABAergic neurons in the thalamic reticular nucleus); VTA, ventral tegmental area.

Consistent with the orexigenic effects of LHA^GABA neurons, recent work has highlighted the role of specific NAc→LHA projections in feeding. GABAergic dopamine D1 receptor (D1R)‐expressing medium spiny neurons (MSNs) in the NAc shell were found to preferentially innervate LHA^GABA neurons and, when these projections were optogenetically activated, feeding was rapidly suppressed, even in hungry animals (O'Connor et al. 2015) (Fig. 3 A). Inhibitory inputs from the NAc to the LHA have also been investigated with regard to addictive behaviour. Larson and co‐authors found that stimulation of this pathway in vivo resulted in increased motivation for cocaine self‐administration and a long‐lasting enhancement in cocaine‐seeking behaviour after withdrawal (Larson et al. 2015). NAc MSNs were reported to also make direct contact with an unknown population of glutamatergic neurons in the anterior LHA, distinct from MCH and Hcrt/Ox neurons (Sano & Yokoi, 2007). The differential roles of inhibitory NAc→LHA circuits involved in enhanced drug‐seeking (Larson et al. 2015) and those involved in the cessation of feeding (O'Connor et al. 2015) may depend on the nature of the NAc projections, neurochemistry of their LHA targets and behavioural context.

The nature of glutamatergic and GABAergic projections from LHA to the VTA were further examined in sucrose reward‐seeking behaviour (Nieh et al. 2015) and in the control of feeding (Barbano et al. 2016). The first study confirmed that LHA→VTA projecting neurons target both DA and non‐DA GABAergic neurons, with both excitatory and inhibitory inputs. Within the LHA, in vivo single‐unit recording in mice trained to respond to a cue predicting sucrose reward revealed that at least two patterns of activity could be distinguished during a sucrose‐seeking task: Type 1 cells showed steady, phasic responses to the sucrose dispenser once the conditional training was acquired, and Type 2 cells exhibited phasic responses to both dispenser and stimulus, which were systematically affected by unexpected reward or unexpected absence of reward (Nieh et al. 2015). These data suggest that two different subsets of LHA→VTA projecting neurons differentially encode conditioned responses and sucrose reward expectation. However, the identity of these cells has not yet been identified. Interestingly, activation of Type 1 LHA→VTA neurons is associated with increased feeding and counterbalances stressful foot‐shock effects to obtain further sucrose rewards (Nieh et al. 2015). We can thus question whether LHA stimulation influences and modifies the value of reward (Stuber & Wise, 2016). It has been shown that LHA neurons display different types of firing responses to taste stimuli that differ in palatability, suggesting that encoding of hedonic and incentive responses to food and taste aversion also occur in the LHA (Li et al. 2013). More recently, Barbano and co‐workers showed that eating and rewarding effects of LHA^GABA→VTA neurons were differentially evoked depending on the frequency of stimulation used to synchronize their activity. Low frequency (5‐10 Hz) stimulation of LHA^GABA→VTA terminals preferentially induced feeding, while higher frequencies (40 Hz) preferentially induced rewarding effects (Barbano et al. 2016). The authors hypothesized that these effects may depend on differential neuropeptide co‐release (Barbano et al. 2016; Stuber & Wise, 2016). Furthermore, Stamatakis and co‐workers recently found that optogenetic inhibition of an excitatory projection of LHA^Glu neurons to the LHb enhanced consumption of a palatable reward whereas its activation mediated avoidance behaviour (Stamatakis et al. 2016) (Fig. 3 A). This LHA^Glu→LHb circuit may be a pathway through which the LHA negatively regulates both feeding and reward.

The properties of LHA^GABA and LHA^Glu neurons may be compared with the distinctive features of LHA^LepRb neurons and their functional role in feeding, energy balance and reward (Leinninger et al. 2011; Leinninger, 2011; Laque et al. 2015). LHA^{LepRb/Nts/Gal} neurons directly modulate Hcrt/Ox and mesolimbic DA functions (Leinninger et al. 2009; Louis et al. 2010; Leinninger et al. 2011; Kempadoo et al. 2013; Goforth et al. 2014; Bonnavion et al. 2015; Laque et al. 2015) and convey leptin's inhibitory effects on food intake (Leinninger et al. 2009; Leinninger et al. 2011). Increased concentration of Nts in the VTA, through chemogenetic activation of LHA^Nts cells, facilitates prolonged DA release in the NAc and increased locomotor activity (Patterson et al. 2015). LHA^Nts neurons contact both VTA DA and GABA neurons, but display twice as many synapses onto the former than the latter (Beier et al. 2015). However, recent work suggests that LHA^Gal neurons, do not innervate the VTA (Laque et al, 2015). Optical self‐stimulation, induced by activation of the LHA→VTA pathway, is partly mediated by Nts and glutamatergic neurotransmission (Kempadoo et al. 2013). Although LHA^Nts/Glu→VTA projecting neurons are likely to express LepRb, most LHA^LepRb/Nts→VTA neurons express GABAergic markers (Leinninger et al. 2011). In light of the recent work describing the orexigenic effects of LHA^GABA neurons (Jennings et al. 2015, Wu et al. 2015; Nieh et al. 2015; Barbano et al. 2016), we can question whether those various effects on feeding and reward behaviour can be explained by multiple LHA^GABA/LHA^LepRb/GABA subpopulations, or by the same neurons that exhibit differential neurotransmitter/neuropeptide co‐release driving one behaviour over another. The VTA may be a relay for LHA‐induced feeding and self‐stimulation, but many other downstream targets remain unexplored including key intrahypothalamic feeding centres interconnected with the LHA (Sternson, 2013; Krashes et al. 2014; Sternson et al. 2016). Indeed, stimulation of a GABAergic projection from the LHA to the paraventricular nucleus of the hypothalamus (PVN), LHA^Pdx1/GABA→PVN, was shown to promote feeding behaviour (Wu et al. 2015) (Fig. 3 A). How these inhibitory neurons track with identified LHA subpopulations, respond to circulating signals of metabolic status, and effect postsynaptic signalling in identified targets will no doubt be the focus of intense future investigation.

Control of sleep and wakefulness

Our understanding of how LHA neuron populations and their efferents regulate sleep and wakefulness primarily resulted from work on the MCH and Hcrt/Ox systems. MCH and Hcrt/Ox neurons exhibit in vivo firing patterns that correlate tightly with vigilance state in rodents, i.e. waking (W), rapid‐eye movement (REM) sleep and non‐REM sleep (NREM) stages. In vivo recordings reveal that MCH neurons are silent during waking, fire occasionally during NREM sleep and fire maximally (from 3 to 12 Hz) during REM sleep, suggesting that they have a critical role in promoting and maintaining REM sleep (Hassani et al. 2009). Indeed, optogenetic activation of MCH neurons selectively modulates REM sleep state (Jego et al. 2013; Konadhode et al. 2013; Tsunematsu et al. 2014) (Fig. 3 B). Activation of MCH neurons during NREM sleep promotes the transition to REM sleep, during which their high‐frequency stimulation extends REM sleep duration (Jego et al. 2013). Chronic activation of MCH neurons over 24 h leads to an increase in both NREM and REM sleep durations (Konadhode et al. 2013) while selective ablation of MCH neurons results in decreased NREM sleep without affecting REM (Tsunematsu et al. 2014). These results indicate that MCH neurons can influence all stages of sleep; however, their impact on sleep as it relates to behavioural and cognitive function requires further investigation. MCH control of REM sleep state is thought to result from multiple downstream pathways involving brainstem GABAergic neurons (Clément et al. 2012) and two distinct forebrain pathways, including a projection from GABAergic MCH neurons to HA‐containing neurons in the TMN (LHA^MCH/GABA→TMN^HA) and to neurons in the medial septum (MS) (LHA^MCH/GABA →MS) (Jego et al. 2013). Each of these connections is involved in inducing and stabilizing REM sleep. Activation of the LHA^MCH → brainstem^GABA pathway facilitates REM sleep by shutting off REM suppressor neurons (Clément et al. 2012). Photostimulation of the LHA^MCH/GABA →TMN^HA pathway extends REM sleep duration by maintaining inhibitory tone on HA neurons, which in turn delays the transition to wakefulness (Fig. 3 B). Activation of the LHA^MCH/GABA →MS pathway also extends REM sleep duration and is hypothesized to stabilize the septo‐hippocampal regulation of REM sleep‐associated theta rhythm (Jego et al. 2013).

Elucidating the functions of the Hcrt/Ox system has significantly informed our understanding of the basic biology of sleep and the aetiology of sleep disorders (Scammell, 2003; Burgess & Scammell, 2012). Hcrt/Ox deficiency or depletion results in narcolepsy in humans and animals (Chemelli et al. 1999; Peyron et al. 2000; Thannickal et al. 2000; Hara et al. 2001). In animal models, narcoleptic attacks – characterized by brief periods of wakefulness, frequent transitions to REM sleep, and behaviour resembling cataplexy – can result from Hcrt/Ox cell ablation or genetic mutations that impair normal synthesis of Hcrt/Ox or its receptors (Scammell, 2003). In contrast to MCH neurons, in vivo recordings from rodent Hcrt/Ox neurons revealed that they are primarily active during wakefulness (from 3 to 13 Hz), less active during quiet waking and silent during NREM and REM sleep, but start firing prior to the onset of wakefulness (Lee et al. 2005; Mileykovskiy et al. 2005; Takahashi et al. 2008). Furthermore, the firing rate of Hcrt/Ox neurons can vary depending on behavioural state, showing phasic activity reaching frequencies up to 15 Hz during sleep‐to‐wake transitions or high‐vigilance states associated with increased muscle tone and/or EEG desynchronization (Lee et al. 2005; Mileykovskiy et al. 2005; Takahashi et al. 2008). In addition, recent fibre photometry experiments revealed a rapid increase in the activity of Hcrt/Ox neurons, but not MCH neurons, in response to an air‐puff (González et al. 2016). These findings indicate that the in vivo firing patterns of Hcrt/Ox neurons are dynamic across sleep–wake states, similar to other systems that promote wakefulness and arousal, but with the particularity of being activated prior to the return of cortical activity and muscle tone during REM sleep‐to‐wake transitions.

Optogenetic methods were also used to interrogate Hcrt/Ox neurons and their causal role in sleep‐to‐wake transitions (Adamantidis et al. 2007; Carter et al. 2009, 2012; Rolls et al. 2011; Bonnavion et al. 2015) (Fig. 3 B). These studies demonstrate that directly driving the activity of Hcrt/Ox neurons during sleep disrupts sleep and promotes transitions to wakefulness (Adamantidis et al. 2007), which may substantially contribute to the mechanisms underlying forms of insomnia. Furthermore, prolonged stimulation of the Hcrt/Ox system impairs memory consolidation during sleep (Rolls et al. 2011), and alters hypothalamic–pituitary axis (HPA) endocrine function (Bonnavion et al. 2015). From the known targets of Hcrt/Ox, the LC is a critical downstream effector of Hcrt/Ox‐mediated sleep‐to‐wake transitions (Carter et al. 2012). In addition, optogenetic activation of sleep‐active GABAergic preoptic (POA) neurons were found to suppress firing in Hcrt/Ox neurons (Saito et al. 2013). These findings suggest that Hcrt/Ox neurons act as a gatekeeper of behavioural state stability.

In vivo electrophysiological recordings of single LHA neurons across natural sleep–wake cycles also identified a population of non‐Hcrt/Ox, non‐MCH LHA^GABA neurons that preferentially fire during sleep states (Hassani et al. 2010). Recent work showed that optogenetic stimulation of LHA^GABA neurons, identified in Vgat‐Cre mice, that project to thalamic reticular nucleus (TRN) GABA neurons, evokes rapid arousal when the stimulation occurs specifically during NREM sleep (Herrera et al. 2016) (Fig. 3 B). While 1 Hz stimulation evokes rapid transitioning from NREM sleep to wakefulness, only 20 Hz stimulation also suppressed the in vivo firing rate of TRN^GABA neurons, the transient inhibition of which allows fast cortical arousal during NREM sleep. LHA^GABA → TRN neurons were shown to exhibit spontaneous activity during sleep and increase their firing when the animals switch from sleep to wakefulness (Herrera et al. 2016). The authors note that these LHA^GABA neurons do not overlap with either Hcrt/Ox neurons, which induce a much slower transition to wake (Adamantidis et al. 2007), or MCH neurons. Wake‐promoting LHA^GABA neurons were also shown to evoke arousal through other targets including the LC and the paraventricular nucleus of the thalamus (PVT). Stimulation of LHA^GABA terminals in both targets favours wakefulness (Herrera et al. 2016), suggesting that LHA^GABA projections may act through disinhibition of arousal systems. Interestingly, LHA^LepRb/Gal neurons were found to densely project directly to noradrenergic LC neurons (LC^NA) neurons (Laque et al. 2015), but the functional implications of this pathway for sleep–wake regulation and cortical arousal remain to be investigated. Finally, Herrera and co‐workers found that the selective optogenetic activation of LHA^LepRb neurons, a putative subpopulation of LHA^GABA neurons, delays the transition from sleep to wake, suggesting that these neurons may recruit a circuit that favours sleep promotion and/or maintenance (Herrera et al. 2016). This work suggests functional diversity among LHA^GABA populations, which remains to be fully elucidated.

Regulation of stress and anxiety

An increasing body of literature suggests that the Hcrt/Ox system has a key role in stress, anxiety and other high‐arousal states (Winsky‐Sommerer et al. 2004; España et al. 2003; Furlong et al. 2009; Berridge et al. 2010; Johnson et al. 2011; Giardino & de Lecea, 2014). Collectively, these studies suggest that Hcrt/Ox neurons are capable of integrating multiple stress‐related central and peripheral inputs, and are critical modulators of the neural circuitry of stress. Furthermore, Hcrt/Ox neurons exhibit increased firing rates in vivo in novel environments or when exposed to an unpredictable sound stimulus, which can be associated with stress (Mileykovskiy et al. 2005). In addition, fibre photometry experiments showed a rapid increase in Hcrt/Ox neuron activity during immobilization stress (González et al. 2016). However, the in vivo firing dynamics of single Hcrt/Ox neurons during states of stress and anxiety remain unknown, and a causal relationship between Hcrt/Ox activation and stress initiation had not been established until recently.

Interestingly, elevated Hcrt/Ox neurotransmission has been associated with anxiety/panic attacks in humans and the development of panic‐like responses (Johnson et al. 2010). Additional studies highlighted the involvement of Hcrt/Ox neurons in modulating fear and fear memory (Flores et al. 2015). Mice lacking the Hcrt‐1 receptor showed impaired freezing responses and reduced neuronal activation in the lateral amygdala in fear‐conditioning paradigms (Soya et al. 2013). Interestingly, the restoration of Hcrt‐1 receptor expression in LC^NA neurons normalized cue‐induced fear behaviour and increased lateral amygdala activity. Further investigation using optogenetics showed that Hcrt/Ox activity in the LC enhanced threat memory formation in an amygdala‐dependent process (Sears et al. 2013). These studies suggest that activity of the Hcrt/Ox system can influence the dysregulation of stress responses and emotional memory formation, which may lead to the development of anxiety disorders.

Recently, we demonstrated a causal relationship between prolonged activation of the Hcrt/Ox system and neuroendocrine and behavioural stress responses. We showed that in vivo optogenetic stimulation of Hcrt/Ox neurons resulted in a state of hypercorticosteronaemia corresponding to increased HPA activity, which was associated with enhanced cardiovascular responses, sleep fragmentation and disrupted exploratory behaviour with signs of freezing (Bonnavion et al. 2015) (Fig. 3 C). These effects likely involve the PVN, the PVT and brainstem nuclei including the LC and DR. This study also demonstrated that the Hcrt/Ox‐evoked increase in corticosterone concentrations is influenced by the nutritional status of the mice, notably in a leptin‐dependent manner. Food deprivation enhanced Hcrt/Ox‐dependent HPA axis activation, while local infusion of leptin into the LHA blunted Hcrt/Ox neuronal activity, as did the optogenetic activation of LHA^LepRb neurons. Indeed, optogenetic stimulation of LHA^LepRb neurons reduced both corticosterone release and suppressed Hcrt/Ox neuron activation in response to stress. Interestingly, we found that the activation of LHA^LepRb neurons in obese leptin‐deficient mice exhibiting endogenous hypercorticosteronaemia acutely normalized corticosterone levels, without affecting food intake or body weight (Bonnavion et al. 2015) (Fig. 3 C). How these LHA^LepRb neurons correspond to LHA peptidergic populations is not yet known. In addition, we did not find any direct projections from LHA^LepRb neurons to the PVN, which controls the neuroendocrine stress response through the secretion of corticotropin‐releasing factor (CRF). This observation indicates that the inhibitory influence of LHA^LepRb neurons on stress‐induced activation of the PVN is indirect. We showed that synaptic input from LHA^LepRb/GABA neurons may account for inhibition of Hcrt/Ox neurons, which would in turn weaken PVN activation through diminished excitatory Hcrt/Ox inputs (Bonnavion et al. 2015). Modulation of other targets of LHA^LepRb neurons, such as LC^NA neurons (Laque et al. 2015) may also have a role in modulating the neuroendocrine stress response. Evidence for direct LHA^Pdx1/GABA → PVN projections (Wu et al. 2015) suggests that LHA^Pdx1/GABA neurons may be distinct from LHA^LepRb/GABA neurons; however, their roles in the regulation of HPA activity and stress responses have not yet been investigated.

Further study is required to fully dissect the role of LHA^LepRb neurons in stress and anxiety. For instance, does the in vivo firing rate of LHA^LepRb neurons mirror that of Hcrt/Ox neurons across the sleep–wake cycle? Do they participate in the regulation of sleep, appetite and/or reward (Jennings et al. 2015; Nieh et al. 2015; O'Connor et al. 2015; Herrera et al. 2016; Stuber & Wise, 2016)? How much do they account for the full spectrum of leptin effects in the central nervous system? These LHA^LepRb neurons appear to modulate stress, arousal, reward and feeding in response to changes in metabolic status, thereby positioning these neurons to shape the comorbidity of anxiety disorders and eating disorders, such as anorexia nervosa or binge eating (Kaye et al. 2004). Such questions are fertile ground for future investigation.

Concluding remarks

The application of new technology is rapidly advancing our understanding of the cell types and circuits that comprise the mammalian hypothalamus – a structure that accounts for only ∼0.3% of the volume of the human brain (Hofman & Swaab, 1992), yet is critical to survival. The LHA in particular has long been a target of investigation owing to its pivotal role in regulating arousal, feeding and motivated behaviour. The LHA is uniquely positioned at the intersection of multiple neural and humoral systems that drive essential behavioural programmes and maintain homeostatic balance in physiology and behaviour. However, the cell‐type and circuit‐level heterogeneity of the LHA has made it a challenging structure to study.

An ever‐increasing number of innovative tools and reagents are being employed to deconstruct the constituent circuits of the LHA. These studies are beginning to provide detailed insight into anatomical inputs and outputs of the LHA and the ability to visualize specific classes of neurons for electrophysiological recording and imaging. Opto‐ and chemogenetic interrogation is providing critical complementary insight into the necessity and sufficiency of these circuits in complex behaviour. Emerging from these discoveries is a mechanistic understanding of the LHA and its constituent neural circuits. Nonetheless, fundamental questions remain concerning the nature of LHA circuits. What is the overarching taxonomy of functional cell types within the LHA? What is the synaptic organization of specific local and long‐range LHA circuits? How do the network dynamics of LHA circuits encode and drive specific facets of behaviour? How might cell type identity, synaptic relationships and larger scale circuit organization shift during development and in the pathogenesis of disease states? Furthering our understanding of the basic biology of LHA cell types and circuits holds the promise of identifying potential targets for selective therapeutic interventions in a variety of disease processes, including neuropsychiatric illness, addiction, sleep disorders and obesity.

Additional information

Competing interests

None declared.

Funding

P.B. was supported by an IBRO Research Fellowship, a NARSAD Young Investigator grant and by the Hilda and Preston Davis Foundation. L.dL. was supported by grants from the NIMH (R01MH087592061), NIA (R01AG04767102) and Klarman Family Foundation. A.C.J. was supported by a grant from the NIMH (R00MH097792).

Biographies

Patricia Bonnavion received her PhD training in neuroscience at the Université Pierre et Marie Curie Paris VI, followed by a postdoctoral fellowship in the laboratory of Luis de Lecea, Professor in the Department of Psychiatry and Behavioural Sciences at Stanford University School of Medicine. Currently, she is a Marie Curie fellow researcher in the Laboratory of Neurophysiology, Université Libre de Bruxelles (ULB) and her research focuses on monoamine control of arousal and cognitive processes.

Laura Mickelsen received her undergraduate degree in Integrative Neuroscience at State University of New York at Binghamton. She and Akie Fujita are PhD students in the laboratory of Alexander Jackson, Assistant Professor in the Department of Physiology and Neurobiology at the University of Connecticut. Her research concerns the cellular and synaptic neurophysiology of cell types and circuits in the lateral hypothalamus.