History and future of leptin: Discovery, regulation and signaling

Abstract

The cloning of leptin 30 years ago in 1994 was an important milestone in obesity research. Prior to the discovery of leptin, obesity was stigmatized as a condition caused by lack of character and self-control. Mutations in either leptin or its receptor were the first single gene mutations found to cause severe obesity, and it is now recognized that obesity is caused mostly by a dysregulation of central neuronal circuits. Since the discovery of the leptin-deficient obese mouse (ob/ob) the cloning of leptin (ob aka lep) and leptin receptor (db aka lepr) genes, we have learned much about leptin and its action in the central nervous system. The first hope that leptin would cure obesity was quickly dampened because humans with obesity have increased leptin levels and develop leptin resistance. Nevertheless, leptin target sites in the brain represent an excellent blueprint to understand how neuronal circuits control energy homeostasis. Our expanding understanding of leptin function, interconnection of leptin signaling with other systems and impact on distinct physiological functions continues to guide and improve the development of safe and effective interventions to treat metabolic illnesses. This review highlights past concepts and current emerging concepts of the hormone leptin, leptin receptor signaling pathways and central targets to mediate distinct physiological functions.

Introduction - The history of leptin discovery

The term leptin (from Greek leptos, meaning thin) was coined with the cloning of the leptin gene in 1994 [1]. However, the discovery of leptin started much earlier in 1949 at the well-known mouse breeding facility - Jackson Laboratories – where the first mouse line with a natural occurring recessive monogenic mutation was observed. These mice were noticeable due to their early onset and severe obesity, as well as the profound diabetes phenotype. The gene locus was named obese (ob) and mice carrying the homozygous mutation were named ob/ob mice (and would be later identified as leptin-deficient mice) [2, 3]. A second recessive monogenic mutant mouse line was found in 1966 with remarkably similar phenotypes as ob/ob mice, but with a more pronounced diabetes phenotype. The gene locus was termed diabetes (db) and mice carrying the homozygous mutation were named db/db mice (and were later defined as leptin receptor-deficient mice) [4]. These ob/ob and db/db mice started an exciting discovery journey for the genetic location of obesity and diabetes. It was soon clear that ob and db are two different genes, but homozygous animals shared profound similarities, including a variety of metabolic and endocrine disorders. This early work led to the conclusion that both gene products likely acted in the same metabolic pathway [5], while any distinction with regards to the severity of their diabetes development could be traced back to their genetic background strain [6]. An elegant line of work by Douglas Coleman and co-workers used parabiosis to join the circulation of ob/ob, db/db and control mice, and demonstrated that a circulating factor that is ineffective in db/db mice causes severe weight loss, suppression of hunger and even death by starvation in ob/ob and control mice [7, 8]. Thus 20 years after the first discovery of ob/ob mice, Coleman and colleagues developed the hypothesis that a circulating factor was missing in ob/ob mice but was plentiful in db/db mice. This circulating factor could cure obesity in ob/ob mice, while db/db mice were unresponsive to it [8]. These studies started a hunt for this bloodborne factor that can regulate body weight and is today known as the hormone leptin which acts via leptin receptors (Lepr) in the brain to maintain energy homeostasis.

It took another 20+ years for the discovery of the gene responsible for the observed effects in ob/ob and db/db parabiosis studies, at a time when positional cloning was still in its infancy. The discovery of the hormone leptin by cloning was initially hailed as a cure for human obesity. After that, further milestones were quickly reached with the cloning of the leptin receptor lepr [9] and the production of recombinant leptin [10]. As a proof of concept, daily injections of recombinant leptin fully corrected obesity and other associated neuroendocrine abnormalities in rare cases of leptin-deficient humans and rodents [11, 12]. Notably, despite the initial findings from parabiosis studies where death by starvation was observed [7, 8], recombinant leptin in rodents will not cause starvation even though it will completely deplete body fat stores [10].

Contrary to the weight loss achieved with leptin in ob/ob mice, it was discovered that obesity causes a robust induction of circulating leptin levels that positively correlate with adiposity [13–15], and additional leptin injections had little effect on body weight and food intake in diet-induced-obese mice compared to lean controls; a condition that is known as leptin resistance [13, 16]. Indeed, the vast majority of people with overweight and obesity are unresponsive to leptin [13, 17, 18] and even in lean and healthy humans large leptin doses had no substantial effect on body weight [18] . Yet, despite its clinical ineffectiveness for most people with obesity, the importance of leptin signaling for the maintenance of normal energy homeostasis is well-accepted. Today, leptin remains an approved treatment for people with rare genetic leptin deficiency, or deficiency of the adipose tissue (e.g., select cases of lipodystrophy) [19–21]. Also, people with secondary leptin deficiency due to excessive weight loss from eating disorders and prolonged relative energy deficiency often seen in athletes could benefit from leptin treatment, [22, 23], even though this is currently not an approved application.

The leptin gene, peptide and expression

Leptin is derived from the lep gene, and is found on chromosome 7, which transcribes a 167 amino acid peptide with a molecular weight of 16kD. The lep gene sequence is highly preserved across mammals, and leptin orthologs exist in amphibians, reptiles and fish with considerable divergence in primary amino acid sequences. The function of leptin is highly conserved in all mammalians and non-mammalians due to the preservation of key secondary and tertiary structures allowing the formation of disulfide bonds [24]. Based on its crystal structure, leptin belongs to the family of long-chain helical cytokines, which includes leukemia inhibitory factor, ciliary neurotrophic factor and human growth hormone [25].

Leptin is produced and secreted from adipose tissue into the circulation, although other tissues may contribute to circulating leptin, e.g. placenta [26, 27] and stomach [28, 29]. Specifically, stomach-produced leptin is responsive to food intake and to the hormone cholecystokinin (CCK, a satiety promoting hormone produced in the gut), which contributes to leptin’s anorexigenic effects by vagal sensory routes [29–31]. Long-term circulating leptin levels positively reflect adipose tissue size and communicate energy storage status to the brain [13, 14]. However, leptin expression and circulating levels also fluctuate in response to acute physiological changes that are not immediately causative for body weight changes such as circadian rhythm [32, 33], temperature [34], fasting/refeeding [35], torpor [36], sleep restriction [37, 38], hypoxia [39, 40], type 1 diabetes [41, 42] and methionine/protein restriction [43–46]. Nutritional state is the best studied condition for changes in leptin levels [47], with fasting robustly decreasing circulating leptin levels and refeeding or obesity increasing circulating leptin levels, often more rapidly than detectable changes in body adiposity [48, 49]. Notably, leptin expression in the stomach also responds to nutritional state and it cannot be ruled out that the stomach contributes to fluctuations of circulating leptin levels [50].

The leptin promoter is the general regulatory element responsible for proper leptin gene expression [51, 52] . Compared to studies on leptin function, surprisingly little effort has been invested to define the regulatory promoter elements for restricted leptin expression to adipose tissue (or placenta, stomach), as well as long-term and short-term fluctuations of leptin gene expression and secretion. One recent study [53] clarified that the originally described leptin promoter is insufficient to explain adipose tissue-specific leptin expression, and leptin enhancer sites (LE1, LE2) are required to interact with the leptin promoter for fat-specific leptin expression. The study further suggests that non-coding RNAs (e.g. IncOb) interact with the promoter/enhancer complex to modulate the level of leptin gene expression and is down and upregulated in states of low and high leptin level, respectively [53] (Figure 1). Even though this study fell short of unraveling how different physiological states (fasting, refeeding, cold exposure) acutely modulate leptin expression, this study greatly highlights our knowledge gap for this important part of leptin function. Several factors can stimulate leptin levels, e.g. inflammatory cytokines, glucocorticoids and insulin [25, 54], but possible sites of interaction with the leptin locus remain unknown. Also, sympathetic norepinephrine release, β3-adrenergic receptor activation and G_s-coupled signaling in adipose tissue are sufficient to decrease leptin gene expression [55–58]. Conversely, intact sympathetic signaling is required to mediate the reduction in circulating leptin levels during fasting [36]. Circulating leptin levels also reflect the physiological potency of leptin, such that low leptin levels (ob/ob mice) are associated with enhanced sensitivity to injected leptin, while hyperleptinemia results in diminished leptin response (aka leptin resistance) [59, 60].

Figure 1: Leptin promoter and environmental modulator of leptin gene expression.

A schematic of the leptin gene locus is shown to illustrate that beyond the classic promoter and leptin gene components, more recent findings identified leptin enhancer sites (LE1 and LE2) and a non-coding sequence (IncOb) upstream of the leptin promoter sites that forms a hairpin RNA structure and interacts with the leptin promoter. The leptin promoter is the key to understanding how leptin production and release is restricted to adipose tissue, reflects body adiposity in the long-term, while allowing acute physiological adaptations (fasting). This is a surprisingly understudied area, despite its importance to understand the role of fluctuating leptin levels in energy homeostasis.

Overall, there is compelling evidence that hyperleptinemia induces leptin resistance, but the importance of leptin resistance for whole-body energy homeostasis and obesity development remains to be conclusively resolved [61]. Preventing the drop in leptin levels during fasting reverses common physiological and endocrine adaptations to fasting and is beneficial for weight maintenance [49, 62, 63]. This highlights the requirement of falling leptin levels to enable the physiological and endocrine adaptation to negative energy balance. Nevertheless, while a lack of leptin clearly causes severe obesity, the beneficial or harmful effects of fluctuating high or low leptin levels are still controversial in the literature. Diet-induced-obesity and therefore increased fat mass is a well-accepted cause of high leptin levels and metabolic disease, suggesting that hyperleptinemia drives negative outcomes. In line with this concept, experimental reduction of leptin levels by a heterozygous deletion of the leptin gene improved glucose tolerance in diet-induced-obese mice, even though it had no impact in chow fed, lean and healthy mice [64]. Another study found that a IncOb deletion lowered hyperleptinemia (only reaching around 20mg/ml, compared to >30mg/ml in typical obese mice, and 3mg/ml in normal mice [65]), caused enhanced weight gain, despite the maintenance of leptin sensitivity [53]. Also, hyperleptinemia overall properly induces enhanced leptin signaling, and leptin resistance rather depicts a reduced maximal signaling capacity due to increased negative feedback signals [66, 67]. We do not know the mechanisms to explain the paradox that very low leptin levels (ob/ob mice) and high leptin levels (diet-induced-obesity) are both associated with obesity and metabolic disease. Similarly, we are unable to explain why bariatric surgery and glucagon-like-peptide-1-receptor (GLP1R) agonist based treatments cause robust weight loss that effectively reduce leptin levels, but do not cause the associated adaptations typical during weight loss (increased hunger, low metabolism) [68].

The hypothalamic-pituitary-adrenal (HPA) axis is an excellent example of leptin interaction with other signaling pathways. Adrenalectomy in Lepr-deficient rats profoundly reversed their obesity [69, 70]. Recent work showed that circulating leptin and glucocorticoid levels depend on each other, the drop in leptin level during fasting enables glucocorticoid levels to increase [71] . Other acutely low leptin level states such as sleep deprivation and cold exposure are indeed associated with increased glucocorticoid levels [37, 72], in line with an interdependency of both systems. These data overall support the idea that increased glucocorticoids mediate or at least contribute to the increased hunger during fasting [71] and the existence of other signaling systems that are enabled by low or disabled by high leptin levels could be the cause for paradox and diverse data on leptin action.

In summary, hypo- and hyperleptinemia mediate important interoceptive signals for central control of energy homeostasis. A cohesive model that explains these diverse experimental outcomes may involve enabling and/or suppressing interactions with other signaling pathways and is the most pressing roadblock to further progress in the treatment of metabolic diseases.

Central leptin access

Leptin is too large to passively cross the blood brain barrier (BBB) and is instead transported across the BBB by a regulated, saturable transporter system. Even though the molecular identity of this leptin transporter system is still unclear, it acts independent of Lepr [73]. However, Lepr expressing pericytes have been described that connect the vasculature with Lepr neurons and may contribute to leptin transport across the BBB [74]. The integrity of a functional BBB in the arcuate nucleus (ARC) is well established and indicated by a lack of fenestrated capillaries [75]. Fenestrated capillaries are found in select brain regions in close proximity to the ventricular space, these regions are collectively termed circumventricular organs (CVO’s) [76]. The border between fenestrated capillaries in the median eminence (ME) and ARC is lined by tanycytes, which are highly specialized glial cells connected via tight junctions that shield the ARC from the circulation and the adjacent median eminence [77, 78].

The BBB is often seen as a rigid barrier and its disruption is associated with illness and unphysiological conditions [79]. However, the BBB may also actively convey metabolic changes to the ARC. Fasting extends fenestrated capillaries to proximal parts of the ARC, possibly enhancing leptin access to ARC neurons [80]. Furthermore, tanycytes may express Lepr-b and contribute to transport leptin into the cerebrospinal fluid (CSF) from where leptin can reach other Lepr target cells [81], however, the existence of Lepr-b on tanycytes is controversial [82]. The proximal ARC also connects to the perivascular space of the median eminence (Virchow-Robin space), which allows blood-derived substances to reach proximal ARC neurons by perivascular routes [75]. Finally, at least some ARC Lepr neurons have fibers extending into the ME and thereby gain direct access to circulating leptin levels [83]. Leptin likely reaches most Lepr neurons via a saturable transport across the BBB, but ARC Lepr neurons are uniquely positioned to detect changing leptin levels independent of a BBB and thereby respond to these changes with increased time- and dose-dependent sensitivity [83]. This unique anatomy could explain why ARC Lepr neurons are prone to develop leptin resistance while other leptin target sites remain leptin sensitive [84]. In addition, tanycyte-mediated leptin transport is sensitive to leptin resistance and select improvement of tanycyte transport reverses DIO and reinstates leptin sensitivity [81]. Similarly, NG2 glia cells (that may include, but are not restricted to tanycytes) interact with Lepr expressing neuronal processes in the median eminence and are important to maintain their integrity, conversely ablation of NG2 glia causes leptin resistance in ARC neurons and weight gain [85]. The different routes of leptin access to the brain that have been proposed in the literature are summarized in Figure 2.

Figure 2: Mechanisms of central leptin access.

Schematic drawing depicting the border at the level of the median eminence (ME) and arcuate nucleus (ARC) as an example to show different mechanisms of central leptin access. A. Saturable transport of leptin across the blood brain barrier (BBB). B. Direct leptin access to the circulation via projections into circumventricular organs (CVO’s, e.g., median eminence, area postrema, organum vasculosum) that lack a BBB and show fenestrated blood vessels for open exchange into and from the circulation. The neuronal processes that contact fenestrated blood vessels in the CVO are stabilized by NG2 glia cells at least in the median eminence. C. Leptin is also transported by tanycytes (specialized glia-like cells) into the cerebrospinal fluid (CSF). Tanycytes span from the median eminence into the CSF of the third ventricle (3V).

Leptin receptors and signaling

Six leptin receptor isoforms (Lepr a-f) are generated by alternative splicing of the lepr (aka db) gene and orthologs of the Lepr gene have been documented in a variety of vertebrate species. The function of leptin signaling is largely conserved across species with the exception of birds [86]. These isoforms share a common leptin binding domain but differ in their intracellular domains [87]. Lepr-a, b, c, d and f are transmembrane receptors that all possess the box 1 motif required to bind janus kinase 2 (JAK2). Lepr-e uniquely lacks a transmembrane domain and is a soluble Lepr isoform, which binds circulating leptin and inhibits central leptin transport [88]. Lepr-b features an extended intracellular signaling domain that is phosphorylated at three distinct tyrosine residues by activated JAK2, and this long Lepr-b isoform is responsible for the main effects of leptin on energy homeostasis and other neuroendocrine functions [9].

Lepr-b is a typical class I cytokine receptor without intrinsic kinase activity; instead leptin binding to Lepr-b allows the recruitment and activation of JAK2, which propagates phosphorylation of JAK2 itself and three tyrosine residues on Lepr-b (Y985, Y1077and Y1138) [9] (Figure 3). Each of these phosphorylation sites induces a specific signaling pathway with distinct physiological leptin functions. Y985 activates src-homology-2 domain protein (SHP-2) and mitogene-activated-protein-kinase (MAPK) signaling and mediates negative feedback signaling of the leptin signaling pathway [89]. Y1077 enables the phosphorylation and activation of signal-transducer-and-activator-of-transcription-5 (STAT5), and STAT5 signaling events mediates reproductive effects of leptin [90]. Finally, Y1138 enables STAT3 phosphorylation (pSTAT3) to activate STAT3 signaling and mediates the main effects of leptin on energy homeostasis and neuroendocrine functions but has little effect on reproduction [91]. Leptin-induced pSTAT3 also reflects relative changes in leptin sensitivity and the physiological state of leptin resistance is nicely recapitulated by decreased leptin-activated STAT3 [84].

Figure 3: Leptin signaling pathways and cellular leptin resistance.

Schematic drawing of leptin signaling pathways via the long form leptin receptor (Lepr-b). The right side depicts phosphorylation sites (P) on tyrosine residues (Y) and their main signaling axis through STAT signaling and transcriptional regulation, that is also the source of negative feedback proteins (SOCS3, PTP1B). Lepr signaling levels via PY1138 and pSTAT3 nicely reflect the leptin sensitivity. In addition, Lepr acetylation also profoundly sensitized leptin signaling by stabilizing Lepr-phosphorylation. On the left side a variety of other signaling mechanisms are depicted that either directly interact with Lepr signaling via serin phosphorylation of JAK2 and subsequent IRS and PI3K signaling or other mechanisms that interact with leptin induced transcription. PI3K = phosphatidylinositol-3-kinase; IRS = insulin receptor substrate; JAK2 = janus kinase-2; ER = endoplasmic reticulum, STAT = signal-transducer-and-activator-of-transcription; SOCS-3 = suppressor-of-cytokine-signaling-3; PTP1B = phosphotyrosine phosphatase 1B; TSC1/2 = tuberous-sclerosis1/2; mTOR = mammalian-target-of-rapamycin; pS6 = phosphorylated ribosomal protein S6; AMPK = AMP-activated protein kinase; ACC = acetyl-CoA carboxylase; SHP-2 = src-homology-2 containing phosphotyrosine phosphatase 2; MAPK = mitogene-activated-protein-kinase, HDAC6 =histone deacetylase 6

Lepr-b signaling, and leptin sensitivity are linked to the adaptor molecules suppressor-of-cytokine-signaling-3 (SOCS-3) and phosphotyrosine phosphatase-1B (PTP1B). Their gene expression is driven by leptin signaling as well and thus act as a classic negative feedback regulators of leptin signaling [89, 92, 93]. Selective mutation of Y985, to prevent leptin-induced Y985 phosphorylation, results in lean mice with enhanced leptin signaling and demonstrates the role of Y985 in negative feedback signaling [94]. Most importantly, leptin resistant mice have increased hypothalamic PTP1B and/or SOCS-3 mRNA expression, which contribute to their leptin resistance [92, 95, 96]. Conversely, leptin sensitizing agents generally show robust reversal of diet induced obesity [97–99], even though the variety of proposed or involved mechanisms rather suggests that leptin resistance and sensitization are general physiological mechanisms that interact with a variety of inputs [100]. Most recently the acetylation of Lepr-b and its ability to enhance leptin signaling (aka leptin sensitivity) was demonstrated. Inhibition of the cytoplasmatic deacetylase HDAC6 was particularly noted for its profound effect to reverse diet-induced-obesity [101]. However, the mechanisms are not entirely clear and may involve a circulating factor from adipose tissue [102] or may act centrally by directly preventing Lepr-b acetylation [101].

JAK2 is also highly regulated by phosphorylation of serine and tyrosine sites that modulate JAK2 activity and signal transduction of Lepr-b; e.g. signaling cascades of insulin-receptor-substrates (IRS) and activate phosphoinositol-3-kinase [103–105], induction of PI3K signaling and mammalian-target-of-rapamycin (mTOR) signaling [106, 107], phosphorylation of 5’-adenosine monophosphate-activated protein kinase (AMPK). AMPK is a metabolic master switch that regulates energy fluxes, and leptin stimulates AMPK activity in the periphery to promote catabolic pathways (fatty acid oxidation, glucose transport) [108], but leptin inhibits AMPK activity in the brain, which specifically regulates food intake via regulation of hypothalamic neuropeptides [109]. In addition to feedback signals, Lepr-b signaling events induce transcription of several hypothalamic neuropeptides that modulate energy homeostasis. Gene expression for food intake suppressing neuropeptides like pro-opiomelanocortin (POMC) and cocaine-and-amphetamine-regulated-transcript (CART) are decreased in states with low leptin levels (e.g. during fasting) but can be increased with leptin injections. Conversely, gene expression for food intake increasing neuropeptides like agouti-related-protein (AgRP) and neuropeptide Y (NPY) are increased with low leptin levels but can be suppressed with leptin injections [110].

Leptin and central control of energy homeostasis

After the discovery that leptin acts in the brain to regulate energy homeostasis, the distribution of Lepr-b expressing neurons revealed the hypothalamus and brainstem as main target for central leptin action, even though several populations have not been studied yet [111, 112]. Subsequent neuron-specific genetic Lepr-b deletion or neuron-specific reactivation of Lepr-b in otherwise Lepr-deficient mice demonstrated that leptin’s effects on energy and glucose homeostasis are predominantly mediated by neurons in the brain, while deletion of Lepr-b from peripheral tissues had no effects on energy homeostasis [113–115]. For this reason, significant research effort has focused on outlining brain areas and neuronal populations that mediate leptin action within the brain.

We now appreciate that diverse hypothalamic and extra-hypothalamic Lepr-b populations collectively contribute to leptin’s overall effect on energy homeostasis [116, 117] (Figure 4A–E). It is also important to realize that these circuits do not act in isolation, such that circuits modulating feeding, energy expenditure and glycemia also interact with other systems such as circadian rhythm, sleep, arousal, stress [118]. Thus, while it is helpful to highlight function-specific Lepr circuits individually, their potential for interconnectivity should be noted as potential route to enable or suppress distinct physiological circuits and adaptations.

Figure 4: Central Lepr-b expression sites and related neuronal circuits to regulate energy homeostasis.

A-E. Depiction of Lepr circuits that contribute directly to distinct aspects of energy homeostasis. These are examples of central circuits that have been studied in more detail for leptin function. Other Lepr populations (grey in top panel) remain to be studied and integrated into a holistic picture of energy homeostasis. Specifically higher, cortical brain structures and descending effector pathway have not been well integrated into leptin regulated energy homeostasis. Note, that the depicted pathways are interconnected at several levels, some of these interconnections were omitted to highlight distinct circuitries. Thus, the schematic does not intend to show a complete map of possible interconnections.

PVN = paraventricular nucleus; ARC = arcuate nucleus; NTS = nucleus of the solitary tract; AgRP = agouti-related-peptide; POMC = proopiomelanocortin; CeA = amygdala; PB =parabrachial nucleus; GABA = ɣ-aminobutyric acid; Glu = glutamate; NAc = nucleus accumbens; LHA = lateral hypothalamic area; VTA = ventral tegmental area; POA = preoptic area; DMH = dorso-medial hypothalamus; RPa = raphe pallidus; SNS = sympathetic nervous system, DMV = dorso-motor complex of vagus; PAG = periaqueductal grey; SCx = sensory cortex; HPC = hippocampus

Homeostatic feeding

Early studies of leptin signaling in the brain focused primarily on the hypothalamus, particularly the ARC where Lepr-b is strongly expressed. This work led to the now classic ‘two-neuron’ model in which leptin regulates food intake by inhibiting a population of orexigenic neurons expressing the neuropeptides NPY and AgRP, while leptin simultaneously stimulates a population of anorexigenic neurons expressing POMC [110]. This work predated modern genetic tools for selective neural manipulation and was based largely on fasting or leptin-induced changes in Npy, Pomc, and Agrp gene expression, as well as the effects of brain-specific injection of their protein products. More recent studies using modern genetic tools to manipulate (pharmacogenetic/optogenetic) or monitor the activity (fiber photometry) of these two neuronal populations have extended this early model. Acute activation of AgRP neurons robustly increases feeding behavior while also altering energy expenditure [119, 120]. Furthermore, in vivo calcium dynamics allows to monitor acute activation changes in POMC and AgRP neurons and confirmed a differential activation state and switch in fasted mice that are acutely refed [121, 122]. However, this work also discovered that AgRP neurons are rapidly inhibited by the sight of food, although food consumption was required to produce larger or more sustained decreases. Finally, gut nutrient infusion is also sufficient to relatively rapidly inhibit AgRP neuron activity [123]. Collectively, while leptin regulates AgRP neuron activity, more recent data indicate that AgRP neurons do not simply track long-term energy status (leptin levels) but are more dynamically regulated to predict food ingestion [121]. Similar data exists for POMC neurons, where food availability rapidly stimulates POMC neuron activity [124]. However, acute stimulation of POMC neurons was less impressive and required a 24h stimulation to detect a significant suppression of food intake [120]. The general concept for differential regulation of POMC and AgRP neurons by leptin is well supported, but differences in the timeline and sustained action of these neurons remain unclear. However, possible mechanisms may involve developmental and acute leptin-dependent plasticity to rewire ɣ-amino-butyric-acid (GABA)/glutamatergic or other select synaptic inputs e.g. to ARC or paraventricular nucleus (PVN) neurons [125–127] and long-term potentiation of ARC to PVN circuits for long-lasting effects on food intake and body weight [128]. Finally, the melanocortin system is a particularly important mediator of ARC leptin action, with POMC neurons activating melanocortin-4-recepotors (MC4R) while AgRP neurons inhibit MC4R function [110]. Notably, not all POMC (and possibly AgRP) neurons express Lepr, and leptin signals may be propagated by direct or indirect pathways [42, 129–131] . Indeed, Lepr- and Glp1r-expressing POMC populations are distinct and their ability to suppress food intake and respond to interoceptive signals varies considerably [130]. These emerging concepts may be important to understand why lifestyle changes are less successful compared to Glp1r agonist medication and will be exciting avenues to explore. The elegant and simple “two-neuron” pathway of ARC leptin signaling is found in all newer physiology textbooks and remains an important pathway for homeostatic regulation of feeding (Figure 4A).

Despite the importance of ARC leptin signaling, conditional deletion or reactivation of Lepr-b (Cre/loxP and derivatives) from POMC and/or AgRP neurons had only a mild impact on body weight and food intake specifically when compared to the severe obesity and hyperphagia observed in db/db mice [132–134]. Subsequent studies targeted Lepr neurons in other hypothalamic and extra-hypothalamic sites and demonstrated that many neuronal populations contribute to overall leptin effects on energy homeostasis [116]. In addition, we now appreciate that distinct Lepr-b populations perform different aspects of physiological leptin function (Figure 4A–E) as further highlighted below.

Satiety

Several studies (but not all [135]) indicate that ablation of AgRP neurons in adult mice results in severe hypophagia from starvation [136, 137]. NPY/AgRP neurons are inhibitory neurons and co-express the inhibitory neurotransmitter GABA. Unexpectedly, the starvation behavior in ablated mice was independent of melanocortin signaling, but GABA fully prevented starvation when microinjected into the parabrachial nucleus (PB) [138]. The PB is activated by visceral malaise, e.g. induced by toxins, lithium chloride, but also by an excessive meal [139, 140]. It was subsequently found that GABAergic AgRP inputs to the PB oppose excitatory brainstem inputs to the PB that regulate nausea and satiety [141]. As noted above, leptin inhibits AgRP neurons and might mediate GABA release to the PB to enhance meal-induced satiety, even though Lepr-b expression in AgRP neurons that innervate the PB has not been confirmed yet. Leptin also directly acts on Lepr neurons within the nucleus of the solitary tract (NTS) in the brainstem to promote anorexia independent of hypothalamic inputs [142, 143] and independent of glucagon-like-peptid-1 (GLP1) expressing NTS neurons [144, 145]. Indeed, satiety is induced by many distinct NTS neurons, and leptin induced satiety can be enhanced when more NTS populations are activated together. These satiety signals and their amplification is mediated by the PB and is a major satiety node [146]. Thus, leptin-induced satiety may involve ARC and NTS Lepr neurons with a common integration site of their signal in the PB (Figure 4B).

Hedonic feeding

An important aspect of food intake is the motivation to seek out food despite competing motivations such as flight from danger (predators, cold/hot environment). The organism’s nutritional state greatly influences the motivation to eat, and a drop in leptin levels during fasting contributes to the increase in motivated behavior, and can be reversed by leptin injections, thus reducing the rewarding value of food and motivation to eat [147, 148].

The mesolimbic pathway is well-described to mediated motivated behavior where midbrain dopamine (DA) expressing neurons in the ventral tegmental area (VTA) release DA into the nucleus accumbens (NAc). The release of DA into the NAc encodes motivated events, which may be either positive (rewarding) or negative (aversive) [149]. DA release thus modulates the motivation to work for reward, or how much the reward is “wanted”. Thus, increased dopamine (e.g. in hyperdopaminergic mice) results in higher “wanting” and increased motivation to work for a food [150]. Leptin decreases the rewarding value of food [147, 151], which may involve Lepr neurons in several brain sites.

Leptin can directly inhibit VTA Lepr-b neurons [152] and prevents sucrose-induced DA release into the NAc [153] as well as other motivated behavior [154], even though the requirement and sufficiency of Lepr signaling in VTA DA neurons to prevent food-reward and the overall impact on body weight homeostasis remains unexplored. Leptin deficiency in ob/ob mice significantly decreased NAc DA content and is recovered by leptin injections (98;104). This is consistent with the idea that dopamine deficiency in the NAc of ob/ob mice diminishes the rewarding effect of food, and the hyperphagia in ob/ob mice could be explained at least in part by their attempt to enhance reinforcement (107;108).

Furthermore, leptin acts also on LHA Lepr-b neurons via a complex network that interacts with VTA neurons [155]. Finally, recent data demonstrated that the above mentioned homeostatic and anorexic aspects of feeding substantially interact with hedonic feeding paradigms [156, 157], and complex behaviors like feeding should be understood as an integration of distinct aspects of feeding behaviors [117](13)(Figure 4C).

Energy expenditure

Energy expenditure counteracts energy intake and together they are critical parts of body weight homeostasis as well as thermoregulation. Acute leptin injections in lean normal mice have no impact on energy expenditure per se, but fasting-induced hypometabolism is fully reversed by leptin treatment [63, 158, 159]. Also, leptin-deficient ob/ob mice have lower body temperature and energy expenditure, and can be corrected by exogenous leptin treatment [11, 160, 161], supporting the idea that leptin signaling interacts with neuronal circuits that are involved in thermoregulation and energy expenditure (even though this has been debated [162]). Leptin-deficient mice are also characterized by brown adipose tissue (BAT) atrophy and innervation [125], both key components for adaptive thermogenesis. In addition, leptin also induces weight loss independent of food intake via mechanisms that require BAT function [55, 163], suggesting that leptin induced energy expenditure and BAT thermogenesis contributes to body weight control.

Thermogenic brain circuits in the hypothalamus that multisynaptically connect to BAT have consistently highlighted the preoptic area (POA) and dorsomedial hypothalamus (DMH) to impact energy expenditure and body temperature [164]. The preoptic area (POA) acts as a temperature sensor and integrates local temperature with afferent information from peripheral and deep-body thermoreceptors [165]. The POA connects with the DMH to regulate further downstream effector neurons in the raphe pallidus (RPa) that control sympathetic BAT activity [164]. Lepr-b is expressed in BAT-related POA and DMH neurons and recapitulate these known thermoregulatory projections to the RPa [166]. Chemogenetic activation of DMH Lepr neurons robustly increase energy expenditure, BAT temperature and decreases body weight [167]. Similarly, DMH leptin injection is sufficient to increase sympathetic BAT input and rescue hypothermia in ob/ob mice [167, 168]. Conversely, silencing of DMH Lepr neurons or deletion of DMH Lepr robustly increases body weight, transiently by increasing food intake, but largely by decreasing energy expenditure [167, 169]. The DMH also contains NPY expressing neurons that impact whole body energy expenditure and thermogenesis that requires sympathetic innervation of adipose tissue, even though these effects are largely independent of leptin function [170].

In contrast, stimulation of POA Lepr neurons decreases energy expenditure and mediates temperature-dependent adaptations to ambient temperature change [171]. Nevertheless, Lepr deletion from POA neurons had no effect on temperature-dependent energy expenditure adaptations or body weight. Instead, changes in energy status were impacted by POA Lepr deficiency and caused increased weight gain when fed a high fat diet and prevented fasting induced hypometabolism [159], suggesting interactions and integration with homeostatic circuits [118].

Thermal sensory information from the skin is processed via the spinal cord and parabrachial nucleus PB to the POA [172]. Even though Lepr neurons are found in the PB, thermosensing PB neurons are likely not Lepr (see below for Lepr PB function). However, the close proximity of anorexia, hypoglycemia and thermoregulatory populations in the PB that integrate vagal and spinal sensory inputs is interesting and could allow unexplored crosstalk and integration of diverse sensory inputs [173]. In line with this, the brainstem can mediate thermoregulatory leptin functions independent of the hypothalamus as observed in decerebrate rats [174] and co-injection of leptin with thyroid releasing hormone (TRH) robustly enhanced the thermogenic capacities of TRH and is mediated by direct leptin effects on NTS Lepr neurons [175]. NTS Lepr neurons innervate the PB, but also innervate the RPa, which may both allow changes in energy expenditure and BAT thermogenesis.

Changes in energy expenditure and thermogenesis always require the simultaneous adaptation of cardiovascular outputs, which are similarly regulated by the sympathetic nervous system. Circulating leptin levels correlate with blood pressure [176] (126) and leptin induces both sympathetic outputs to BAT and kidney [177]. The effect of leptin on sympathetic output involves the ARC and NTS [178, 179], however, it remains unclear how thermosensing Lepr systems coordinate with distinct autonomic systems. These circuits are summarized in Figure 4D.

Glucose homeostasis

Leptin also exerts a powerful effect on blood glucose homeostasis, as indicated by experiments in leptin- or leptin receptor-deficient mice. Both ob/ob and db/db mice show hyperglycemia, hyperinsulinemia, and glucose intolerance, which is independent of their excessive body weight (56), and is effectively normalized with leptin injections in ob/ob mice [161]. There is consensus that leptin’s effects on blood glucose are primarily mediated by the central nervous system, because central leptin infusions are sufficient to improve glucose homeostasis [180, 181].

There is much support for the ARC as a site for leptin-dependent regulation of glucose homeostasis, because restoration of Lepr in the ARC of Lepr-deficient mice is sufficient to normalize blood glucose levels [107, 133]. POMC and AgRP neurons both contribute to leptin-dependent regulation of blood glucose and reactivation of Lepr in either of these neurons in diabetic Lepr-deficient mice will improve glucose homeostasis [182, 183]. However, inconsistency exists in the literature regarding the role of leptin signaling in POMC or AgRP neurons to induce diabetes or improve glycemia [42, 182–184].

Lepr expressing neurons in the PB also contribute to glucose homeostasis and are activated specifically by hypoglycemia. Deletion of Lepr from PB neurons enhances hypoglycemia (e.g., induced by insulin or 2-deoxyglucose) [185, 186]. These PB-Lepr neurons innervate the ventromedial hypothalamus (VMH), which is well-known for its glucose sensing neurons [187], even though the exact circuit how VMH neurons change blood glucose, e.g. via the pancreas, liver or HPA axis remains unexplored (Figure 4E).

Additional work has focused on cellular Lepr-b signaling events and collectively indicate that PI3K, but not STAT3 signaling, mediates the effects of leptin on glucose homeostasis, because lack of leptin-induced Stat3 signaling does not cause hyperglycemia in contrast to db/db mice [91]. Instead, inhibition of PI3K signaling prevents leptin-enhanced insulin sensitivity and glucose tolerance [107]. Also, POMC-specific ablation of PI3K signaling is sufficient to alters glucose homeostasis [188], even though ablation of PI3K signaling in all Lepr neurons does not affect glycemia [189] . Leptin also alters glucose and lipid metabolism by affecting AMPK signaling in the hypothalamus [190]. Thus, it remains unclear which of these signaling pathways involve direct or indirect leptin effects and how they act at a circuit level.

Finally, the glucose lowering effect of central leptin action involves insulin-dependent and insulin-independent mechanisms. Considering insulin’s established role in the regulation of blood glucose homeostasis, several studies established that leptin improves insulin sensitivity [107]. Yet in addition to this insulin sensitizing effect, leptin also lowers glucose levels independent of insulin. Severe hyperglycemia in insulin-deficient rodents (e.g. pancreatic β-cell ablation with streptozotocin treatment) can be normalized with centrally applied leptin [191, 192]. This brain centered glucoregulatory system seems to contribute significantly to the regulation of blood glucose via mechanisms that are independent of, but synergize with, the classic insulin-dependent models of blood glucose homeostasis [193] .

Vagal functions

The NTS contributes to all metabolic aspects described in the circuits highlighted above and the NTS is indeed an important linchpin to receive and propagate incoming vagal afferent signals from the periphery particularly along the entire food passage (alimentary canal). Lepr expressing neurons are found directly in the NTS and may directly sense circulating leptin levels via the nearby area postrema, a circumventricular structure with vascular fenestration. Another major sensory input to the NTS is from vagal afferent nerve terminals, whose neuronal cell bodies reside in the nodose ganglion. Lepr mRNA may be expressed in vagal afferent neurons and several studies convincingly demonstrated vagal afferent activation in vitro and in vivo by leptin alone or in synergism with CCK [30, 194]. Interestingly, these actions may instead engage local leptin release from the stomach, rather than depend on circulating leptin levels, and seem to contribute significantly to the food intake suppressing effect of leptin [30]. Also, conditional deletion of Lepr from Nav1.8 expressing cells, which targets a subpopulation of vagal afferents (but also trigeminal and dorsal root neurons, [195], supports a vagal afferent mediated suppression of food intake and body weight by direct leptin action on vagal afferents Lepr [196]. Vice versa, hypothalamic neurons innervate the brainstem and can interact with vagal efferents in the DMV that are able to induce rapid leptin release from the stomach [197]. Recent studies further highlight that hypothalamic leptin targets, such as POMC and AgRP neurons, are acutely regulated by gut hormones like CCK, while systemic leptin only provided slow modulation of neuronal activity in these ARC neurons [124]. Indeed, low doses of leptin infusions targeted to the stomach innervation were sufficient to suppress food intake, while those leptin levels were ineffective systemically [30]. This suggests that fast stomach/vagal afferent leptin signaling acts independent of slow adipose/humoral systemic leptin signaling and may need to be reinvestigated for a more holistic understanding how each route (Figure 5A) contributes to the suppression of food intake and what possible interactions they have.

Figure 5: Leptin and integration into autonomic circuits.

A. Parasympathetic axis with Lepr expressing neurons in the NTS that receive vagal sensory information, but Lepr expressing neurons are also directly found in vagal-sensory neurons in the nodose ganglion. Those vagal-sensory Lepr neurons may directly sense leptin released from the stomach mucosa or circulating leptin may reach Lepr-expressing NTS neurons by transport or by projections outside the BBB (area postrema). These signals are further mediated to the hypothalamus and hypothalamic signals can modulate parasympathetic outputs in the DMV. B. The sympathetic nervous system affects leptin production in adipocytes by adrenergic stimulation of β₃-adrenergic receptors. Leptin acts directly as interoceptive signal to communicate directly with hypothalamic neurons. These Lepr neurons are often sensitive to a variety of environmental signals and thus take part in the integration of these signals. There is no consensus on how leptin levels are reflective of the whole-body system and is an important roadblock to understanding dynamic changes in homeostatic levels (body weight, glucose levels, blood pressure, body temperature).

Hyp= hypothalamus; Glp1= glucagon like peptide 1; CCK = cholecystokinin; NTS= nucleus of the solitary tract; DMV = dorso-motor complex of vagus; DRG=dorsal root ganglia; StG = stellate ganglion; T1–12=thoracic sympathetic ganglia 1–12; L1 = lumbar sympathetic ganglia 1

Sympathetic and spinal sensory functions

As noted earlier, the sympathetic nervous system is an important descending effector pathway of hypothalamic leptin signaling. Several hypothalamic (PVN, ARC, VMH) and brainstem sites (RPa, NTS) send projections to the spinal cord and preganglionic cholinergic neurons in the IML. Select peripheral organs receive sympathetic innervation according to their location and Figure 5B depicts the distinct sympathetic and spinal sensory innervation of interscapular BAT vs. inguinal WAT [198]. Peripheral and central leptin injections increase circulating catecholamines which may involve norepinephrine release from select tissues, e.g. adipose tissue, heart or vasculature, or may be released from the adrenal gland (norepinephrine and epinephrine) [199]. Similarly, central and peripheral leptin enhance sympathetic nerve activity in adipose tissue and kidney to impact thermogenesis and blood pressure [200, 201]. However, Lepr-b expression on post-ganglionic sympathetic neurons or spinal sensory neurons in dorsal root ganglia has not been reported, thus suggesting that all sympathetic effects are mediated by circulating leptin.

Several data suggest that leptin expression and release is suppressed by overall sympathetic signaling [36, 58, 202] and specifically by β₃-adrenergic receptors in adipose tissue [203]. However, the close interaction of leptin and adrenal responses is noticeable and adrenal signals are likely to at least contribute to the sympathetic leptin circuit and thermogenesis [204]. The sympathetic and spinal sensory circuits are not well studied for their role in metabolic disease, even though the general organization of the spinal sensory and sympathetic circuits [173] would allow comprehensive interactions with central Lepr circuits at several levels.

Conclusion and perspective

Over the last 30 years the discovery of leptin taught us much about distinct physiological leptin functions that are relevant for obesity but did not allow us to cure obesity. Diverse Lepr populations in the CNS propagate distinct aspects of physiological leptin functions. Our understanding of the complex network that underlies leptin regulated feeding behavior and other homeostatic functions continues to grow. However, individual Lepr populations often contribute to similar physiological functions, but we lack a refined model to integrate their function and their ability to enhance or suppress physiological functions. The state of leptin resistance has plagued the treatment of obesity, and lifestyle changes were unable to overcome counteracting physiological adaptation (low metabolism and increased hunger) to preserve energy following weight loss and low leptin levels. Yet, the astoundingly strong and sustainable effects of bariatric surgery and GLP1R agonist-based drugs with impressive weight loss and improved glucose homeostasis remarkably override or suppress these physiological adaptations despite negative energy balance and low leptin levels. There is an increasing interest to understand homeostatic regulation and physiological adaptation in a more holistic way that appreciates the interaction, integration, enhancement and suppression capabilities of physiological adaptations that allows the body to defend a new homeostatic levels. Thus, in the coming 10 years, leptin will likely continue to guide this next exciting chapter to understand the flexibility of the diverse energy homeostatic systems and how they work together in the ever-changing environment. It is hoped that this will allow us to refine our strategies to treat metabolic diseases more efficiently and safely in the future.

Highlights:

• Physiological and obesity related changes in leptin gene expression.

• Mechanisms of CNS leptin access and known impacts for obesity.

• Leptin receptor signaling events and impact in leptin resistance for whole body energy homeostasis.

• Function-specific circuits of Lepr-expressing neurons and their role in energy and glucose homeostasis.

Abbreviations:

ob/ob mice

leptin-deficient mice

db/db mice

leptin receptor-deficient mice

Lepr

leptin receptor

CCK

cholecystokinin

LE1/2

leptin enhancer 1/2

IncOb

leptin promoter interacting non-coding RNA

GLP1R

glucagon-like-peptide-1-receptor

HPA

hypothalamic-pituitary-adrenal axis

BBB

blood brain barrier

ARC

arcuate nucleus

CVO

circumventricular organ

ME

median eminence

CSF

cerebrospinal fluid

NG2

Nerve/glial antigen 2

JAK2

janus-kinase-2

Y985/1077/1138

tyrosine residues 985/1077/1138

SHP-2

src homology-2 domain protein

MAPK

mitogen-activated-protein-kinase

STAT3/5

signal-transducer-and-activator-of transcription-3/5

pSTAT3

phosphorylated STAT3

SOCS-3

suppressor-of-cytokine-signalig-3

PTP1B

phosphotyrosinephosphatase-1B

HDAC6

histone deacetylase 6

IRS

insulin-receptor-substrates

PI3K

phosphoinositol-3-kinase

mTOR

mammalian-target-of-rapamycin

AMPK

5’-adenosine monophosphate-activated protein kinase

POMC

pro-opiomelanocortin

CART

cocaine-and-amphetamine-regulated-transcript

AgRP

agouti-related-protein

NPY

neuropeptide Y

PVN

paraventricular nucleus

LHA

lateral hypothalamic area

MC4R

melanocortin-4-recepotors

GABA

ɣ-amino-butyric-acid

PB

parabrachial nucleus

NTS

nucleus of the solitary tract

GLP-1

glucagone-like-peptide-1

DA

dopamine

VTA

ventral tegmental area

NAc

nucleus accumbens

BAT

brown adipose tissue

UCP1

uncoupling protein-1

DMH

dorsomedial hypothalamus

POA

preoptic area

RPa

raphe pallidus

TRH

thyroid releasing hormone

Nav1.8

genetic marker for sensory neurons in the nodose, trigeminal and dorsal root neurons

DMV

dorso-motor complex of vagus

VMH

ventromedial hypothalamus