Distinct Subsets of Lateral Hypothalamic Neurotensin Neurons are Activated by Leptin or Dehydration

The lateral hypothalamic area (LHA) is essential for ingestive behavior but it remains unclear how LHA neurons coordinate feeding vs. drinking. Most LHA populations promote food and water consumption but LHA neurotensin (Nts) neurons preferentially induce water intake while suppressing feeding. We identified two molecularly and projection-specified subpopulations of LHA Nts neurons that are positioned to coordinate either feeding or drinking. One subpopulation co-expresses the long form of the leptin receptor (LepRb) and is activated by the anorectic hormone leptin (NtsLepRb neurons). A separate subpopulation lacks LepRb and is activated by dehydration (NtsDehy neurons). These molecularly distinct LHA Nts subpopulations also differ in connectivity: NtsLepRb neurons project to the ventral tegmental area and substantia nigra compacta but NtsDehy neurons do not. Intriguingly, the LHA Nts subpopulations cannot be discriminated via their classical neurotransmitter content, as we found that all LHA Nts neurons are GABAergic. Collectively, our data identify two molecularly- and projection-specified subpopulations of LHA Nts neurons that intercept either leptin or dehydration cues, and which conceivably could regulate feeding vs. drinking behavior. Selective regulation of these LHA Nts subpopulations might be useful to specialize treatment for ingestive disorders such as polydipsia or obesity.


Results
Methods to Identify LHA Nts Neurons. In situ hybridization (ISH) identifies many Nts-expressing cells within the LHA (Fig. 1A,B) 28 but does not easily permit determination of co-expressed transcripts or circuit tracing. Nts immunofluorescence (Nts-IF) permits such analyses, but only labels fibers, not cell bodies (Fig. 1C) unless mice were pre-treated with the axonal transport inhibitor, colchicine ( Fig. 1D) 33,34 . Since colchicine induces neuronal dysfunction and lethality its use prohibits study of Nts contributions to normal physiology. Nonetheless, Nts-ISH and -IF confirm a large population of LHA neurons that actively express Nts (Fig. 1B,D). We reasoned that Nts Cre mice would be useful to examine the molecular expression, projections, activation responses and neurochemistry of this large population of LHA Nts neurons without physiology-disrupting colchicine treatment. First, we verified the fidelity of this model for identifying Nts neurons by colchicine-treating Nts Cre mice on a Creinducible GFP reporter background 35 (Nts Cre ; GFP mice). We examined two brain regions that have been shown via Nts-ISH to contain numerous Nts neurons: the subthalamic nucleus (STN) and the LHA 36 . Similarly, Nts Cre ; GFP mice have dense populations of GFP-labeled neurons within the STN and LHA that co-label with Nts-IF ( Fig. 1E-K). Thus, Nts Cre ; GFP mice reliably identify Nts neurons and can be used to characterize features of LHA Nts neurons. Nts Cre ; GFP Mice Confirm that Some LHA Nts Neurons Project to the Midbrain. LHA Nts neurons project to two midbrain regions, the ventral tegmental area (VTA) and the substantia nigra compacta (SNc) 26,37 , which regulate motivated behaviors and motor function, respectively 38,39 . We therefore hypothesized that different subsets of LHA Nts neurons might project to the VTA or SNc. First, we identified LHA Nts projecting neurons by injecting Nts Cre ; GFP mice with FluoroGold (FG) in the VTA ( Fig. 2A-D) or the SNc (Fig. 2E-H), which is taken up by terminals and transported retrogradely to the soma of origin. Examination of VTA-injected mice revealed many LHA cell bodies that accumulated FG, some of which also contained GFP and hence identify LHA Nts neurons that project to the VTA (Fig. 2B-D, yellow arrows). Yet, many LHA Nts neurons did not accumulate FG from the VTA (Fig. 2B-D, green arrows). This could be because some LHA Nts neurons do not project to the VTA, or if there was insufficient FG throughout the VTA subregions containing LHA Nts terminals. However, we also found adjacent FG-labeled cells lacking GFP, indicating non-Nts containing LHA cells that provide input to the VTA and that FG coverage was sufficient to label various LHA soma ( Fig. 2B-D, red arrows). Analysis of SNc-injected mice revealed similar groups of labeled neurons (Fig. 2F-H). Together, these data verify that that some LHA Nts neurons project to the VTA and SNc, but nearly twice as many project to the VTA vs. the SNc. (Fig. 2I). Our methodology prevented determination of whether any LHA Nts neurons provide collateral projections to these areas. However, the differing number of projections hint that there may be some separate subpopulations of LHA Nts neurons that project to the VTA, SNc, or sites other than the midbrain. LepRb-expressing neurons. Some GFP-labeled LHA Nts neurons contained leptin-induced pSTAT3 and these are Nts LepRb neurons ( Fig. 3D-F, cyan-outlined magenta arrows). Yet, many GFP-labeled LHA Nts neurons did not contain pSTAT3 ( Fig. 3D-F, cyan arrows). These data confirmed that there are at least two molecularly distinct populations of LHA Nts neurons: one population expresses LepRb and can be revealed by leptin-induced pSTAT3 (the Nts LepRb neurons), but another LHA Nts population lacks LepRb. Next, we asked whether the molecularly-specified Nts LepRb neurons are also projection-specified. Nts Cre ; GFP mice were injected with FG to label LHA Nts neurons that project to the VTA or SNc, and also treated with leptin to permit pSTAT3-mediated identification of Nts LepRb neurons. This paradigm identified Nts LepRb neurons that accumulated modest amounts of FG from the VTA (Fig. 3G-J, white arrows) and from the SNc (Fig. 3K-N, white arrows). In each case we also observed Nts LepRb neurons that did not accumulate FG; these may be Nts LepRb neurons that project to the midbrain region that was not injected with FG or to a yet undetermined site outside of the midbrain (Fig. 3G-N, cyan-outlined magenta arrows). As expected, we also observed LHA Nts neurons lacking LepRb, some of which project to the midbrain ( Fig. 3G-P, cyan-outlined yellow arrows), but other LHA Nts neurons did not accumulate FG and hence do not project to either midbrain region ( Fig. 3G-P, cyan arrows). Together, these data signify that Nts LepRb neurons are a molecularly-specified subset of LHA Nts neurons, some of which project to the VTA and SNc.

Nts LepRb
Nts Dehy Neurons Are a Subset of LHA Nts Neurons that Do Not Project to the Midbrain. Since dehydration disrupts bodily osmolality and upregulates Nts mRNA in the LHA 28 , we postulated that it might also modify the activity of LHA Nts neurons. To test this, we provided Nts Cre ; GFP mice with ad lib water (euhydration) or removed water overnight (dehydration), then examined the LHA for GFP and cFos (a marker of recent neuronal depolarization). During euhydration we observed few LHA Nts neurons with cFos, suggesting that LHA Nts neurons are not activated during normal fluid balance ( Fig. 4A-C). In contrast, overnight dehydration increased cFos within some (but not all) LHA Nts neurons ( Fig. 4D-F, cyan outlined magenta arrows). Thus, dehydration activates a subset of LHA Nts neurons, termed Nts Dehy neurons. We then assessed midbrain projections of Nts Dehy neurons in dehydrated Nts Cre ; GFP mice previously injected with FG in the VTA or SNc. Although we observed Nts Dehy neurons in the LHA (Fig. 4G-N  Together these data confirm that Nts Cre ; GFP mice correctly identify Nts-expressing cells, and can be used to visualize them.  and cyan-outlined yellow arrows). Together with Fig. 3

Verification of Cre-Dependent Reagents to Determine Classical Neurotransmitter
Content. Given the heterogeneity of LHA Nts neurons at the molecular and circuit level, we hypothesized that they might differ in other ways, such as in classical neurotransmitter content. Neuropeptide-expressing neurons may contain GABA or glutamate, release of which determines whether synaptic targets are inhibited or activated, respectively 25 . LHA Nts neurons have been reported as GABAergic 24 or glutamatergic 40 , thus, we sought to define whether subpopulations might be neurochemically distinguishable. Since GABA and glutamate cell bodies cannot be labeled using immunoreagents, we crossed vGat Cre and vGlut2 Cre mice with Cre-inducible GFP reporter mice to label these cells with GFP (vGat Cre ; GFP and vGlut2 Cre ; GFP mice). A limitation of this model is that neurons can alter neurotransmitter expression over lifespan 41 , but recombination during development produces permanent GFP-labeling that may not reflect the neurochemistry of the mature neuron. Thus, we also injected adult vGat Cre and vGlut2 Cre mice with an AAV Cre-Lox-red fluorescent protein (RFP) so that only actively-expressing GABA and glutamate-expressing neurons undergo recombination to express RFP. Finally, we compared these reporter models with ISH data in brain regions known to primarily contain glutamatergic neurons (the STN) or GABAergic neurons (the zona incerta, ZI) to verify the fidelity of each method for identifying neurotransmitter-expressing neurons in the adult brain. We observed similar distributions of vgat-ISH, and GFPor AAV-Lox-RFP-labeled vGAT neurons in the ZI, but no labeled neurons in the adjacent STN (

Determination of Classical Neurotransmitter Content Within LHA Nts Neurons Using the Dual
Recombinase system. We generated dual recombinase mice to simultaneously label Nts cells and vGat or vGlut2-expressing cells by crossing Nts FlpO mice to vGat Cre or vGlut2 Cre mice, producing Nts FlpO ; vGat Cre and Nts FlpO ; vGlut Cre mice respectively. To test the fidelity of these dual recombinase mice, we injected them in the ZI with AAV-Frt-GFP (to permit FlpO-mediated expression of GFP that identifies Nts neurons) and AAV-Lox-RFP LepRb; magenta-outlined yellow arrows = VTA/SNc projecting LepRb neurons that do no express Nts; white arrows = VTA/SNc-projecting Nts LepRb neurons. These data demonstrate that at least some Nts LepRb neurons project to the VTA and the SNc. VTA-injected vehicle-treated n = 6; female VTA-injected leptin-treated n = 10; SNc-injected vehicle-treated n = 5; SNc-injected leptin-treated n = 9. GFP mice were given ad libitum water (Euhydrated) or were dehydrated overnight. Brains were assessed for GFP-labeled Nts neurons (green) and cFos, a marker of recent neuronal depolarization (blue). Cyan arrows label Nts-only neurons and magenta arrows identify dehydration activated neurons that do not express Nts. Cyan-outlined magenta arrows identify Nts-GFP neurons that co-express cFos (Nts Dehy neurons). (G-N) Nts Cre ; GFP mice were injected with FG into the VTA or SNc (to identify midbrain projecting neurons) and dehydrated overnight (to identify dehydration-activated neurons via cFos). Assessment of the LHA revealed some Nts Dehy neurons (Cyan-outlined magenta arrows) that did not accumulate FG. Similarly, (K-N) Nts Dehy neurons were found within the LHA of SNc-injected mice but none of these contained FG. Key for arrows: cyan arrows = Nts-only neurons; magenta arrows = cFos-only (dehydration-activated) neurons; cyan-outlined magenta arrows = Nts Dehy neurons that do not project to the VTA/SNc; yellow arrows = FG-only neurons that project to the VTA/SNc; cyan-outlined yellow arrows = Nts neurons that project to the VTA/SNc but are not activated by dehydration; magentaoutlined yellow arrows = dehydration-activated VTA/SNc projecting neurons that do no express Nts. No arrows are present to label VTA/SNc-projecting Nts Dehy neurons because no such neurons were found. VTA-injected, euhydrated n = 5; VTA-injected, dehydrated n = 9; SNc-injected, euhydrated n = 5, SNc-injected, dehydrated n = 9. (to permit Cre-mediated expression of RFP for detection of vGat or vGlut2 neurons). Since the ZI contains primarily GABAergic but not glutamatergic neurons as well as some Nts neurons (Figs 1 and 6), the dual recombinase system should only result in GFP and RFP co-labeling of ZI cells in Nts FlpO ; vGat Cre mice, but not in Nts FlpO ; vGlut Cre mice. As anticipated, dual AAV injection into the ZI of Nts FlpO ; vGlut Cre mice yielded many GFP-labeled Nts neurons, none of which contain RFP-vGlut2 (Fig. 7A,B", cyan arrows). The GABAergic ZI is in fact devoid of RFP-vGlut2, despite the robust induction of RFP in surrounding regions known to contain glutamate (Fig. 7A). By contrast, in dual AAV-injected Nts FlpO ; vGat Cre mice, we observed that all of the GFP-labeled Nts cells co-label with RFP-vGat, indicating that ZI Nts neurons are GABAergic (Fig. 7C,D"; white arrows indicate co-labeled cells, magenta arrows identify RFP-vGat cells that do not contain GFP-Nts). Taken together, these data confirm that the ZI Nts cells are GABAergic, but not glutamatergic, as would be expected of this primarily GABAergic brain region. Our findings also confirmed the reliability of the dual recombinase system to distinguish classical neurotransmitter content, so we next used it to determine whether subsets of LHA Nts neurons can be discriminated via their classical neurotransmitter expression. Dual AAV injection into the LHA of Nts FlpO ; vGlut Cre mice identified GFP-Nts neurons confined within the perifornical LHA (Fig. 7E,F", cyan arrows) and RFP-vGlut2 neurons (Fig. 7E,F", magenta arrows), but we did not observe any LHA cells that co-expressed both labels. By contrast, dual AAV injection into the LHA of Nts FlpO ; vGat Cre mice identified GFP-Nts neurons within the perifornical LHA, most of which co-labeled with RFP-vGat (Fig. 7G,H", white arrows). We also observed many RFP-vGAT neurons that did not co-label with GFP-Nts (Fig. 7G,H" magenta arrows). Together these data suggest that LHA Nts neurons are predominantly GABAergic, and they comprise a subset within the larger population of LHA GABA neurons.

Determination of Classical Neurotransmitter Content of LHA Nts Neurons Using Colchicine-Mediated Nts-IF.
The dual recombinase method suggests that LHA Nts neurons do not contain glutamate based on the absence of neurons co-expressing both Nts-GFP and RFP-vGlut2. This negative result could also be an artifact, perhaps if there were inefficient AAV-Lox-RFP infection within Nts FlpO ; vGlut Cre mice that resulted in under-detection of LHA glutamate neurons. We therefore sought to validate the classical neurotransmitter content of LHA Nts neurons using an independent strategy that did not depend on AAV-mediated recombination. We attempted to generate dual-reporter mice, but commercially available FlpO reporter lines proved ineffective for simultaneous labeling of Nts and vGat or vGlut2 neurons. This also raised concern that limited efficiency of FlpO-mediated recombination in Nts FlpO mice might under-report LHA Nts neurons and diminish the likelihood of detecting a small population of glutamatergic LHA Nts neurons. Alternately, we treated vGat Cre ; GFP and vGlut2 IRESCre ; GFP mice (validated in Fig. 6) with colchicine, allowing for simultaneous visualization of GFP-labeled vGat and vGlut2 and Nts-IF (as in Fig. 1). Similar to findings using the dual recombinase strategy, we observed numerous Nts-IF cell bodies within the LHA and ZI of colchicine-treated mice (Fig. 7I-N"). While the ZI from vGlut2 Cre ; GFP mice was devoid of GFP-vGlut2 neurons (Fig. 7I,K-K"), vGat Cre ; GFP mice had numerous GFP-vGat neurons in the ZI, many of which also contained Nts-IF (Fig. 7L,N-N", white arrows). These data are consistent with the GABAergic phenotype of the ZI and our findings using the dual recombinase system (Fig. 7G,H) that ZI Nts neurons are GABAergic but not glutamatergic. We observed many Nts-IF labeled cell bodies within the LHA, as well as GFP-labeled LHA glutamate neurons  ( Figure I') and GFP-labeled GABA neurons (Fig. 7L). Despite the robust GFP-labeling induced in both lines, we did not observe any Nts-IF cell bodies within the LHA that co-localized with GFP-vGlut2 (Fig. 7J-J"). By contrast, essentially all of the LHA Nts-IF cell bodies overlapped with GFP-vGat ( Fig. 7M-M", white arrows), but many GFP-vGat neurons did not contain Nts-IR (magenta arrows). In sum, colchicine-mediated Nts-IR recapitulated our findings using the dual-recombinase system: that LHA Nts neurons are GABAergic but comprise a subset of the larger population of LHA GABA neurons.

Discussion
The LHA is essential for the motivation to eat and drink but the neural mediators of these behaviors have yet to be fully understood. While most LHA populations promote food and liquid intake, LHA Nts neurons divergently regulate ingestive behavior by suppressing feeding and promoting drinking 22,24 . We therefore hypothesized that separate subpopulations of LHA Nts neurons might exist to coordinate opposing feeding and drinking behavior.
Here we characterized two separate subpopulations of LHA Nts neurons that are differentially activated by leptin (Nts LepRb neurons) or dehydration (Nts Dehy neurons). While all LHA Nts neurons are GABAergic, the Nts LepRb and Nts Dehy subpopulations differ in molecular expression of LepRb and at the circuit level, and hence can be distinguished via these criteria (Fig. 8). These data demonstrate the heterogeneity of LHA Nts neurons and their specific responsiveness to either energy or fluid balance cues suggest that they may coordinate different ingestive behaviors (feeding vs. drinking).
Here we used Nts Cre and Nts FlpO mice to non-invasively and reliably identify Nts neurons, which permitted their study under normal, physiologic conditions. Using Nts Cre mice, we found that dehydration treatment increased activation from ~6% at baseline to ~18%, demonstrating that ~12% of the LHA Nts neurons are Nts Dehy neurons (Fig. 5C). Additionally, Nts LepRb neurons make up a separate 15% of LHA Nts neurons 22 . While these are modestly sized populations, they can significantly influence homeostasis; for example, mice lacking leptin-regulation via Nts LepRb neurons have impaired response to energy balance cues and diminished dopamine signaling that causes overweight 22,26 . Characterization of the remaining 70% of LHA Nts neurons at the molecular and circuit level may provide insights about their function. For example, some LHA Nts neurons are activated by LPS-mediated inflammation and inhibit local orexin/hypocretin neurons, and these may contribute to illness-behavior 43 . Nts signaling is also implicated in regulation of analgesia, thermoregulation, stress and addiction 44 , so it will be important to determine if/how the remaining LHA Nts neurons contribute to these diverse aspects of physiology. Characterizing the heterogeneity of LHA Nts neurons may also suggest intersectional or pharmacological strategies to target specific subpopulations of LHA Nts neurons, and hence selective physiological outputs.
LHA neurons project widely throughout the brain and differentially modify behavior depending on their targets 44 . Our finding that Nts LepRb neurons, and not Nts Dehy neurons, project to the midbrain suggests that there are distinct LHA Nts neural mechanisms for leptin-mediated suppression of feeding vs. regulation of drinking and fluid balance. Experimental activation of all LHA Nts neurons causes release of Nts to the VTA, and dopamine release into the nucleus accumbens 27 that can modify motivated intake behavior 27,45,46 . Thus, at least some portion of anorectic leptin regulation via Nts LepRb neurons could occur via direct projections to, and modulation of, mesolimbic dopamine signaling. This is consistent with the requirement of leptin action via Nts LepRb neurons for regulating body weight and the integrity of the mesolimbic dopamine system, which are due in part to Nts signaling via VTA neurons expressing neurotensin receptor-1 (NtsR1) 26,37 . In contrast, Nts Dehy neurons must act via other yet-to-be determined projection targets, and do not directly modulate dopamine signaling to modify physiology. While the function of Nts Dehy neurons and their projection sites remains to be established, the activation of Nts Dehy neurons in response to dehydration suggests that they may coordinate fluid need with the motivation to drink. The discovery of specific subsets of LHA Nts neurons also hints at why experimental activation of all LHA Nts neurons results in diverging ingestive behaviors. Such activation simultaneously induces Nts LepRb neurons that act partially via the VTA (and may be anorectic) as well as the Nts Dehy neurons that regulate separate targets, and it is possible that these populations suppress feeding and promote drinking, respectively 24 . Since Nts LepRb and Nts Dehy neurons are induced by separate physiological cues (leptin or dehydration), it remains to be determined whether there are any physiological situations in which these subpopulations are concurrently activated. In any case, our data confirm that Nts LepRb and Nts Dehy neurons have distinct circuitry, thus projection-specific modulation may be a useful strategy to discern their respective contributions to ingestive behavior.
Despite the molecular and circuit heterogeneity of LHA Nts neurons, they all contain the same classical neurotransmitter, GABA. LHA Nts neurons presumably inhibit synaptic targets via release of GABA, as well as regulating postsynaptic and adjacent neurons via release of Nts. It remains to be determined if GABA and Nts are always co-released from LHA Nts neurons. Indeed, different physiological stimuli bias the release of glutamate. Nts FlpO ; vGlut2 cre n = 5, Nts FlpO ; vGat Cre n = 6). (I,N") vGlut2 Cre ; GFP mice and vGat Cre ; GFP mice were treated with colchicine to permit detection of Nts-IF (red) and vGat-or vGlut2-GFP (green). (J,J") In vGlut2 Cre ; GFP mice, many GFP-vGlut2 cell bodies are found in the LHA (magenta arrows) along with Nts-IF cell bodies (cyan arrows), but no overlapping cells were found. (K-K") The ZI from vGlut2 Cre ; GFP mice contained a few Nts-IF neurons (cyan arrows) but no vGlut2-GFP cells, consistent with the GABAergic neurochemistry of the ZI. (M-M") The LHA of vGat Cre ; GFP mice contained many GFP-vGat cell bodies (magenta arrows) and Nts-IF cells that all co-labeled with GFP-vGat (white arrows). (N-N") Similarly, co-labeling was observed in the GABAergic ZI, (N-N") but not in vGlut2 Cre ; GFP mice. These data confirm that LHA Nts neurons contain vGat and are GABAergic, but do not contain vGlut/glutamate. vGlut2 Cre ; GFP mice n = 4; vGat Cre ; GFP mice n = 5. Abbreviations: mt = mammillothalamic tract; f = fornix; LHA = lateral hypothalamic area, STN = Sub-thalamic Nucleus, ZI = Zona Incerta.
Scientific RepoRts | (2019) 9:1873 | https://doi.org/10.1038/s41598-018-38143-9 neurotransmitter vs. neuropeptide signals in some LHA neurons, and the receipt of these messages depends on the repertoire of receptors expressed on target neurons, which can also vary [47][48][49] . Our finding that LHA Nts neurons are GABAergic is consistent with other reports of overlapping LHA Nts and GABAergic neurons 27,50 but contrasts with a report of glutamatergic LHA Nts neurons that directly project to the VTA 40 . This discrepancy may be because AAV-Frt-GFP induced GFP expression was confined to the perifornical LHA (~Bregma −1.34 to 1.70 51 ), and limited our characterization to this area only. However, the glutamatergic LHA Nts neurons were identified around the "rostral lateral hypothalamus" corresponding to Bregma −0.40, where the rostral LHA merges into the preoptic area, and well beyond the perifornical LHA. Hence, the GABAergic perifornical LHA Nts neurons studied here could be anatomically and neurochemically distinct from LHA Nts neurons of the hypothalamus-preoptic continuum. While all LHA Nts neurons contain GABA, they are a subset of the vast population of LHA GABA neurons. This may account for the strikingly different behaviors observed after experimental activation of LHA Nts neurons (suppression of feeding, increased drinking) vs. activation of all LHA GABA neurons (increased feeding, drinking and gnawing directed at non-biological objects) 19,20,24,27 . Since activation of LHA Nts and LHA GABA neurons promotes drinking, LHA Nts neurons contribute to at least some of the polydipsic effect. The orexigenic effect observed with activation of all LHA GABA neurons likely masks anorectic effects mediated by the subset of LHA Nts neurons encompassed within them. Our findings agree with reports of functionally-distinct subpopulations of LHA GABA neurons 18 , and LHA Nts neurons are a functionally unique subset within the larger population of all GABA neurons that suppress feeding instead of promoting it. This arrangement may also explain differences in VTA regulation that have been ascribed to these populations. Some LHA GABA neurons project to the VTA, where they disinhibit VTA GABA neurons that in turn releases inhibition of DA neurons to facilitate DA release and feeding 19,52 . Some GABA-containing LHA Nts neurons, including Nts LepRb neurons, also project to the VTA, and their precise synaptic targets are yet to be defined.
Taken together, our data reveal the heterogeneity of LHA Nts neurons. Since LHA Nts neurons are differentially regulated by energy status (leptin) vs. fluid status (dehydration), and comprise separate subpopulations, there may be separate neural mechanisms to coordinate feeding and drinking necessary for survival. If true, then these data may suggest strategies to selectively modify the LHA Nts neurons that control feeding vs. those that modify drinking.

Materials and Methods
Animals. Adult male and female mice were used for studies. Some Nts Cre ; GFP and LepRb Cre ; GFP mice were generated and treated with euhydration or dehydration at the University of Michigan, under the supervision of the Unit for Laboratory Animal Medicine (ULAM). These procedures were approved by the University of Michigan Institutional Animal Care and Use Committee (IACUC). All other mice were generated from a breeding colony at Michigan State University. MSU mice were cared for by Campus Animal Resources (CAR) and all animal protocols were approved by the Institutional Animal Care and Use Committee (IACUC) at Michigan State University. All mouse experiments were performed in accordance with Association for Assessment and Accreditation of Laboratory Animal Care and National Institutes of Health guidelines. In all cases mice were housed in a 12 h light/12 h dark cycle and had ad libitum access to water and chow diet unless otherwise noted.  Leptin or Dehydration treatment. Some FG-injected Nts cre ; GFP mice were treated with PBS or recombinant mouse leptin (5 mg/kg, i.p.) purchased from the National Hormone and Peptide Program (Los Angeles Biomedical Research Institute, Los Angeles, CA) then were perfused 2-4 hours later to enable detection of leptin-induced pSTAT3. In our hands, both of these leptin treatment times produce comparable pSTAT3. Males and females were studied but no notable differences between sexes were observed so they were pooled for analysis: VTA-injected vehicle-treated n = 6; female VTA-injected leptin-treated n = 10; SNc-injected vehicle-treated n = 5; SNc-injected leptin-treated n = 9.

Generation of Nts FlpO Knock-In
Other FG-injected Nts cre ; GFP mice underwent a dehydration-activation paradigm in which they were either had ad libitum access to water (euhydration) or the water bottle was removed for 12 hr during the dark cycle (when mice drink most of their daily water). Mice were perfused the following morning and brain sections were stained for cFos, GFP and FG (see below). Due to the lack of observable differences between sexes, males and females were pooled for analysis: VTA-injected, euhydrated n = 5; VTA-injected, dehydrated n = 9; SNc-injected, euhydrated n = 5, SNc-injected, dehydrated n = 9.
Adult male 8-12 wk old Nts Cre ; GFP and LepRb Cre ; GFP mice were also treated via euhydration or had water bottles removed for 24 hr (including during the dark cycle) to induce dehydration then were perfused (Euhydrated Nts Cre ; GFP n = 7, LepRb Cre ; GFP n = 5; Dehydrated Nts Cre ; GFP n = 4, LepRb Cre ; GFP n = 3). Brains were analyzed for cFos and GFP, and 3 representative LHA sections spanning the same Bregma sections from each mouse were analyzed using Photoshop to count the number of GFP-only labeled neurons and GFP neurons containing cFos. Graphed data represent the average percentage of GFP neurons containing cFos out of the total number of GFP-labeled neurons ±SEM. Significant differences between genotypes and treatments were determined via 2-way ANOVA with Bonferroni posttests.
Immunohistochemistry and Immunofluorescence (IF). Mice were treated with a lethal dose of i.p.
pentobarbital and perfused transcardially with either 10% formalin or 4% paraformaldehyde (Sigma-Aldrich 158127) containing 0.4% picric acid (Sigma-Aldrich 197378). Brains were removed, post-fixed in the same fixative overnight at 4 °C, dehydrated with 30% sucrose/PBS for 2-3 days, then sectioned into 30 µm slices using a sliding microtome (Leica). Brain sections were analyzed by immunohistochemistry and/or IF as previously described 22 . For activation studies, brain sections first were exposed to either rabbit-anti pSTAT3 (1:500, Cell Signaling) or goat-anti cFos (1:500, Santa Cruz) followed by incubation with species specific Alexa-488 conjugated (Jackson ImmunoResearch, 1:200) or Alexa-568 conjugated antibodies (LifeTech, 1:200) and visualization with DAB (Sigma). IF labeling was performed exposing sections to primary antibodies, including chicken anti-GFP (1:2000, Abcam), rabbit anti-FG (1:500, Fluorochrome), rabbit-anti Nts (1:500, Phoenix) and/or anti-dsRed (1:1000, Clontech), followed by incubation with species-specific secondary antibodies conjugated to AlexaFluor 488 or 568 fluorophores (1:200, Life Technologies or Jackson ImmunoResearch). Immunolabeled brain sections were analyzed using an Olympus BX53 fluorescence microscope outfitted with transmitted light to analyze DAB-labeling as well as FITC and Texas Red filters for IF. Microscope images were collected using Cell Sens software and a Qi-Click 12 Bit cooled camera, and images were analyzed using Photoshop software (Adobe). We collected images from all mice within an experiment and compared them side-by-side to select the representative images included in the manuscript.

Quantification of Nts Neuronal LHA -> VTA and LHA -> SNc Projections. FG-injected Nts cre ;
GFP mice were perfused and brains were stained as described above. An intact section between Bregma −1.3 and −1.5 was taken from each sample for quantification. The area medial and dorsal to the Fornix and ventral to the MT was counted to determine the number of FG (VTA/SNc-projecting) and GFP (Nts) labelled neurons in each. Males and females were studied but no notable differences between sexes were observed so they were pooled for analysis: VTA-injected n = 6; female; SNc-injected vehicle-treated n = 5. Graphed data represent the average number of GFP neurons containing FG. Significant differences between brain areas were determined via Student's t-test and were calculated using GraphPad Prism (GraphPad Software Inc., San Diego, CA). Error bars depict ± standard error of the mean (SEM). **p < 0.01.