Specific subpopulations of hypothalamic leptin receptor-expressing neurons mediate the effects of early developmental leptin receptor deletion on energy balance

Objective To date, early developmental ablation of leptin receptor (LepRb) expression from circumscribed populations of hypothalamic neurons (e.g., arcuate nucleus (ARC) Pomc- or Agrp-expressing cells) has only minimally affected energy balance. In contrast, removal of LepRb from at least two large populations (expressing vGat or Nos1) spanning multiple hypothalamic regions produced profound obesity and metabolic dysfunction. Thus, we tested the notion that the total number of leptin-responsive hypothalamic neurons (rather than specific subsets of cells with a particular molecular or anatomical signature) subjected to early LepRb deletion might determine energy balance. Methods We generated new mouse lines deleted for LepRb in ARC GhrhCre neurons or in Htr2cCre neurons (representing roughly half of all hypothalamic LepRb neurons, distributed across many nuclei). We compared the phenotypes of these mice to previously-reported models lacking LepRb in Pomc, Agrp, vGat or Nos1 cells. Results The early developmental deletion of LepRb from vGat or Nos1 neurons produced dramatic obesity, but deletion of LepRb from Pomc, Agrp, Ghrh, or Htr2c neurons minimally altered energy balance. Conclusions Although early developmental deletion of LepRb from known populations of ARC neurons fails to substantially alter body weight, the minimal phenotype of mice lacking LepRb in Htr2c cells suggests that the phenotype that results from early developmental LepRb deficiency depends not simply upon the total number of leptin-responsive hypothalamic LepRb cells. Rather, specific populations of LepRb neurons must play particularly important roles in body energy homeostasis; these as yet unidentified LepRb cells likely reside in the DMH.


INTRODUCTION
Obesity, which affects more than 1/3 of people in developed countries, predisposes to diabetes, cardiovascular disease, and other serious comorbidities [1]. To design effective treatments for obesity, we must first understand the systems that control energy balance and which represent potential therapeutic targets. The hormone leptin, which is produced by adipose tissue to signal the repletion of fat stores, acts via its receptor (LepRb) on hypothalamic neurons to suppress food intake and permit energy expenditure [2]. Leptin-or LepRb-deficient humans and rodent models display dramatic hyperphagia and reduced energy expenditure, leading to severe obesity [3e5]. Thus, the hypothalamic neurons by which leptin mediates the control of energy balance represent important controllers of energy balance.

Hypothalamic LepRb neurons
Within the hypothalamus, the arcuate nucleus (ARC), ventromedial hypothalamic nucleus (VMN), dorsomedial hypothalamus (DMH), lateral hypothalamic area (LHA), and ventral premammillary nucleus (PMv) contain substantial numbers of LepRb neurons [6]. While roles for many molecularly-defined and anatomically-circumscribed populations of LepRb have been examined, the early developmental deletion of LepRb from these previously-studies populations has not resulted in obesity similar to the severe obesity observed in entirely LepRb-deficient db/db mice [7e11]. Indeed, while orexigenic ARC neurons that contain agouti-related peptide (AgRP), neuropeptide Y (NPY) and gamma amino butyric acid (GABA) (AgRP neurons) and anorexigenic ARC proopiomelanocortin (POMC)-containing neurons each express LepRb and play crucial roles in energy balance, early developmental ablation of LepRb from AgRP and/or POMC neurons minimally alters energy balance [7,8]. Similarly, manipulation of LepRb expression in other circumscribed sets of LepRb neurons examined to date (e.g., neurotensin (Nts) neurons of the LHA, steroidogenic factor-1 (Sf1) neurons in the VMN) minimally impacts energy balance [10,11]. In contrast, LepRb deletion from some large, widely-distributed hypothalamic populations results in profound obesity and hyperphagia. Indeed, deletion of LepRb in vGat (Slc32a1)-expressing GABAergic neurons (representing w50% of hypothalamic LepRb neurons, including the majority of LepRb neurons in the DMH and LHA, along with AgRP cells and other ARC neurons) or Nos1 neurons (representing w25% of hypothalamic LepRb neurons, including the majority of PMv neurons plus smaller numbers of LepRb cells in the ARC, DMH, VMN and LHA) promotes dramatic obesity [12,13].
1.2. Populations of LepRb neurons that are or are not subject to compensation following early developmental deletion Early alterations in AgRP neurons (e.g., neuron ablation or LepRb deletion) are compensated during postnatal development, while alterations in mature AgRP neurons in adults have profound effects on leptin action and energy balance [7,14e16]. Indeed, while the deletion of LepRb neurons from adult AgRP neurons provokes dramatic obesity, the early developmental deletion of LepRb from AgRP neurons produces little metabolic derangement. Thus, while leptin action on AgRP neurons in adults plays important roles in energy balance, the lack of direct leptin action on AgRP neurons is unlikely to contribute substantially to the phenotype of entirely LepRb-null db/db mice. Hence, additional (non-AgRP) LepRb neurons that are not subject to developmental compensation must underlie the majority of the db/db phenotype and play important roles in leptin action. In this manuscript, we employ multiple models that mediate the early developmental deletion of LepRb in subpopulations of hypothalamic neurons to define sets of LepRb neurons that mediate leptin action and in which the loss of LepRb is not developmentally compensated (unlike AgRP neurons).

Potential types of LepRb neurons that control energy balance
This non-developmentally compensated control of energy balance might be distributed across multiple hypothalamic nuclei and cell types, with many types of LepRb neurons contributing similarly to the control of energy balance, such that the ablation of leptin action in a threshold number of LepRb neurons (without regard to their location or identity) disrupts the control of energy balance. It is also possible that small, but currently unidentified, populations of LepRb neurons mediate the main effect of leptin/LepRb on energy balance that is not subject to developmental compensation. Here, we test these possibilities by studying mice deleted for LepRb in previously unexamined sets of growth hormone-releasing hormone-Cre (Ghrh Cre ) LepRb neurons of the ARC (LepRb Ghrh cells) as well as serotonin receptor 2c (Htr2c)-expressing LepRb (LepRb Htr2c ) neurons that lie mainly in the PMv, VMH, and LHA (a few also lie in the ARC and DMH). We examined these mouse strains together with mice subjected to early developmental deletion of LepRb in Agrp, Pomc, vGat, and Nos1 neurons. Our results and analysis suggest that small (and as yet unidentified) populations of DMH LepRb neurons that are not subject to developmental compensation likely play crucial roles in the control of energy balance by leptin.

The role for Ghrh Cre neurons in leptin action
To identify novel subpopulations of hypothalamic LepRb neurons, we previously performed translational profiling of their transcriptome using translating ribosome affinity purification (TRAP) followed by RNA-seq (TRAP-seq) [17]. This study revealed the enrichment of Ghrh mRNA in hypothalamic LepRb neurons [17]. Leptin modulates food intake, glucose homeostasis, and linear growth [2], and Ghrh neurons also likely participate in the control of these parameters [18e20]. We therefore postulated that direct leptin action on LepRb Ghrh neurons might mediate these effects. To test this notion, we generated a knock-in mouse line to cotranslationally express Cre recombinase with Ghrh mRNA (Ghrh Cre mice) ( Figure 1A). Breeding Ghrh Cre onto the Cre-dependent Rosa26 eGFP-L10a reporter background (Ghrh eGFPÀL10a mice) revealed the presence of Cre-expressing neurons in the expected areas of the hypothalamus, including the ARC (Figure 1BeD) [21], as well as in a few regions in the midbrain and hindbrain (Fig. S1). While the hypothalamic distribution of eGFP-L10a neurons mirrored the known adult expression pattern of Ghrh [21], the adult midbrain and hindbrain express little detectable Ghrh, suggesting either early developmental Ghrh expression in these regions or low-level expression in adults. Single-cell sequencing of ARC neurons identified Ghrh neurons as distinct from AgRP and POMC neurons [22]. Similarly, Ghrh Cre neurons are more lateral to the third ventricle than AgRP neurons [23] and do not colocalize with the laterally localized POMC neurons (Figs. S1E and F). Thus, our manipulation of Ghrh Cre neurons will not directly impact AgRP or POMC neurons. To identify LepRb Ghrh neurons, we examined the detection of leptinstimulated phosphorylated STAT3 (pSTAT3)-immunoreactivity (IR) and eGFP in Ghrh eGFPÀL10a mice by immunohistochemistry (IHC) (Figure 1CeD). The IHC detection of pSTAT3-IR reveals cellautonomous leptin action on LepRb-expressing neurons [24]. As previously reported, no pSTAT3-immunoreactivity is detected in the brain in the absence of leptin action in ob/ob or db/db mice, and exogenous leptin promotes pSTAT3 in many hypothalamic cells in a distribution consistent with LepRb neurons (Fig. S2); previous results have demonstrated the colocalization of leptin-stimulated pSTAT3 with LepRb neurons [17]. Our analysis in Ghrh eGFPÀL10a mice revealed extensive colocalization of pSTAT3-IR and eGFP in the ARC (w45% of ARC eGFP neurons were pSTAT3 positive). LepRb Ghrh neurons represent a minority of total pSTAT3 neurons in the ARC, however. We also observed sparse colocalization in the DMH, but none in other areas, including the midbrain and hindbrain (data not shown).
To determine the importance of LepRb Ghrh neurons to leptin action, we crossed Ghrh Cre onto the Lepr flox background to generate Ghrh Cre/þ ;Lepr fl/fl (LepRb Ghrh KO) and littermate control (Lepr fl/fl ) mice for study. Leptin treatment failed to promote the accumulation of pSTAT3 in eGFP-containing neurons in LepRb Ghrh KO mice on the Rosa26 eGFP-L10a background ( Figure 1E,F), consistent with the ablation of LepRb from Ghrh Cre neurons in these animals. We examined the body weight and composition of LepRb Ghrh KO mice, as well as of mice with LepRb deleted in ARC POMC or AgRP neurons (LepRb Pomc KO and LepRb Agrp KO mice, respectively) ( Figure 2). This analysis recapitulated the small (2e3 g) increase in body weight previously observed [7,8] in both male and female LepRb Pomc KO and LepRb Agrp KO mice ( Figure 2C,E). The increase in body weight in these animals reflected a tendency toward increased adiposity, except in the case of male LepRb Pomc KO mice, in which the increase in body weight reflected increased lean mass ( Figure 2D,F). In contrast to the LepRb Pomc KO and LepRb Agrp KO mice, however, LepRb Ghrh KO mice exhibited no difference in body weight or body composition compared to controls (Figure 2A,B). Serum leptin and insulin concentrations were indistinguishable from controls for all three lines ( Figure 3). Furthermore, we observed no alteration in food intake, blood glucose, or body length in male or female LepRb Ghrh KO mice compared to their controls (Fig. S2). Thus, direct leptin action via LepRb Ghrh neurons is not required for the control of energy balance, glucose homeostasis, or linear growth in male or female mice, and leptin must control these parameters via other LepRb neurons. Thus, the early developmental ablation of LepRb from all previously-examined populations of ARC LepRb cells fails to substantially alter energy balance. Hence, the LepRb neurons not subject to developmental compensation that are most crucial for the control of energy balance either lie elsewhere in the hypothalamus or are redundant and distributed across multiple hypothalamic cell populations.

The role for widely-distributed LepRb Htr2c neurons in leptin action
We previously generated knock-in mice to express Cre recombinase from the Htr2c locus (Htr2c Cre mice) [25]; breeding these to the Rosa26 eGFP-L10a reporter strain (Htr2c eGFPÀL10a mice) ( Figure 4A) revealed the distribution of Htr2c neurons across many hypothalamic regions (Figure 4BeD). Furthermore, the detection of leptin-stimulated pSTAT3 in Htr2c eGFPÀL10a mice revealed substantial colocalization of pSTAT3 with eGFP in the PMv, VMH, LHA, ARC, and DMH; we observed no colocalization outside of the hypothalamus, however (data not shown). Thus, LepRb Htr2c neurons represent a broadly-distributed population of hypothalamic LepRb neurons. To test whether the total number of hypothalamic LepRb neurons subjected to early developmental LepRb ablation dictates the disruption of energy balance, we bred Htr2c Cre onto the Lepr fl background to generate Htr2c Cre ;Lepr fl/fl (LepRb Htr2c KO) and littermate control (Lepr fl/ fl ) mice. Comparing the detection of leptin-stimulated pSTAT3 in the hypothalamus of LepRb Htr2c KO and control mice revealed the almost complete ablation of pSTAT3/LepRb from the PMv, VMH, and LHA (all three areas p < 0.05), with more modest reductions of pSTAT3 in the ARC and DMH (Figure 4EeG). While Htr2c is expressed in POMC neurons [25], the POMC subpopulation that expresses Htr2c is distinct from that which expresses Lepr [26,27]. Thus, the few LepRb Htr2c neurons in the ARC do not include POMC cells. Overall, LepRb Htr2c KO mice display w50% loss of hypothalamic pSTAT3. We compared the body weight phenotype of LepRb Htr2c KO mice to that of mice ablated for LepRb in vGat neurons (LepRb vGat KO mice), in which LepRb is ablated from the majority of LepRb neurons in the ARC, DMH and LHA [12]. While LepRb Htr2c KO mice of both sexes were slightly heavier than littermate controls ( Figure 5A), their increased body weight was small relative to the dramatic reduction in their number of hypothalamic LepRb cells and arose mainly from increased lean mass, not fat mass ( Figure 5B). In contrast and as previously reported [12], LepRb vGat KO mice of both sexes displayed dramatically increased body weight and adiposity ( Figure 5C,D; Fig. S3A). Furthermore, LepRb Htr2c KO mice exhibited neither hyperglycemia (Fig. S3B) nor alteration in serum insulin or leptin concentrations ( Figure 3). Thus, ablating w50% of hypothalamic LepRb neurons in LepRb vGat KO mice dramatically produced obesity, while ablating similar number of hypothalamic LepRb neurons in LepRb Htr2c KO mice failed to increase adiposity, demonstrating that the specific LepRb cell type affected matters more than the total number of hypothalamic neurons from which LepRb is ablated. We previously discovered a role for LepRb Nos1 neurons in the control of energy balance by leptin [13]. Due to the high frequency of Nos1 Cremediated excision of Lepr fl (producing Lepr D ) in the female germline, we studied Nos1 Cre/þ ;Lepr D/fl (LepRb Nos1/D KO) compared to littermate Lepr D/fl or Lepr D/þ (LepRb D/? ) control animals. It is theoretically possible that the ablation of 50% of LepRb expression in all neurons sensitized the phenotype of LepRb Nos1/D KO; indeed, Lepr D/þ mice exhibit a mild increase in body weight and adiposity compared to Lepr þ/þ or Lepr fl/fl animals (see, for example leptin levels in Figure 3). We thus generated Htr2c Cre/þ ;Lepr D/fl (LepRb Htr2c/D KO) and littermate Lepr D/þ or Lepr D/fl controls for study; we also generated LepRb Nos1/ D KO and littermate controls for comparison ( Figure 6). While LepRb Htr2c/D KO mice of both sexes tended to weigh more than their controls, the difference was not statistically significant ( Figure 6). Furthermore, as for LepRb Htr2c KO mice without a Lepr D allele, male mice exhibited an increase in lean mass, rather than adiposity, and leptin and insulin concentrations did not differ from controls ( Figure 3). In contrast, LepRb Nos1/D KO mice of both sexes exhibited increased body weight and adiposity ( Figure 6C,D), along with increased circulating leptin and insulin (Figure 3), as previously reported. Thus, the  presence of a Lepr D allele failed to unmask an energy balance phenotype due to the ablation of LepRb in Htr2c neurons, confirming that the LepRb Htr2c cells play little role in the control of energy balance by leptin.

DISCUSSION
Despite the passing of more than two decades since the discovery of leptin [4], the cellular mediators (i.e., LepRb neurons) that mediate the largest component of the dramatic obesity phenotype of db/db mice remain undefined. To date, over a dozen anatomically-circumscribed populations of cells have been identified as leptin-responsive, but all of those studied by early developmental ablation of LepRb play only small or negligible roles in the control of energy balance by leptin [7e11,17]. Here, we demonstrate that ARC Ghrh cre neurons are leptin responsive, but are also dispensable for leptin action. It is important to note that our analysis (like most others) of the phenotypes derived from LepRb ablation in specific cell types rests on the analysis of mice in which the LepRb ablation occurs at an early developmental stage [16], rendering developmental compensation/ remodeling of neural circuits to mask potential phenotypes. Indeed, while manipulation of AgRP neurons in adults (including LepRb deletion) promotes dramatic changes in energy balance [14e16], early developmental alterations in AgRP neurons produce minimal phenotypes as a consequence of developmental compensation [7,15]. Thus, our analysis focuses on the circuits whose dysfunction cannot be compensated in this manner.
The failure to alter energy balance upon early developmental LepRb ablation in most cell types might be explained by the distribution of leptin action across multiple populations, such that each population alone is dispensable. Indeed, the findings that LepRb deletion from broadly distributed hypothalamic populations (vGat or Nos1) dramatically increases food intake, body weight and fat mass [12,13] supports this distributed model. Here, we explicitly tested this model by removing leptin action from the large broadly-distributed population of hypothalamic LepRb Htr2c neurons. Ablation of LepRb from Htr2c cells failed to substantially alter energy balance or parameters of glucose homeostasis e even when bred onto a potentially sensitizing Lepr D background. Indeed, if anything, the slight phenotype of Lepr D decreases the ability to detect alterations in weight gain by neuronspecific LepRb ablation. Thus, LepRb Htr2c neurons play a minimal role in the control of energy balance by leptin, and deletion of LepRb from approximately half of hypothalamic LepRb neurons does not necessitate obesity if the important cells are not impacted. Additionally, some set of non-Htr2c, non-AgRP, non-POMC, non-Ghrh cells must be especially important for the db/db phenotype.

Potentially crucial role for DMH LepRb neurons in the control of energy balance
Htr2c Cre deletes LepRb mainly in the PMv, VMH, and LHA (but minimally impacts the ARC and DMH) and promotes no adiposity phenotype, while Nos1 Cre (which deletes LepRb in most of the PMv plus 5e 15% of cells in the ARC, DMH, VMH and LHA) and vGat Cre (which deletes LepRb in the ARC, DMH, and LHA) both produce substantial obesity [12,13]. These data suggest that direct leptin action via LepRb neurons in the PMv, VMH, and LHA play minimal roles in the db/db phenotype, consistent with the minimal phenotypes produced by modulation of LepRb in the PMv [28], in the VMN [11], in vGlut2 cells [12] (PMv plus VMH and many brainstem neurons), and in Ntsexpressing LHA LepRb neurons [10]. Thus, leptin action via LepRb neurons in the ARC and/or DMH must play a predominant role in the phenotype of db/db mice. Since Nos1 Cre -mediated LepRb ablation results in substantial obesity [13],  but relatively few LepRb Nos1 neurons lie in the ARC and DMH, the crucial populations of LepRb neurons must be relatively small. Furthermore, because ablation of LepRb in ARC AgRP, POMC, or Ghrh Cre cells minimally alters energy balance, DMH LepRb neurons likely play an especially important role in the db/db phenotype [29]. Consistent with the small number of LepRb Nos1 neurons that lie in the DMH (but presumably contribute to the phenotype of LepRb Nos1/D KO mice), not all DMH LepRb neurons appear to play a substantial role in the control of energy balance. For instance, ablation of LepRb from DMH prolactin-releasing hormone (Prlh) neurons produces a small phenotype that results from altered energy expenditure, rather than dysregulated food intake [9]. Similarly, ablation of LepRb in prodynorphin neurons (which ablates 43% of DMH LepRb) produces no detectable body weight phenotype [17]. Thus, the small number of DMH LepRb Nos1 cells (presumably those that overlap with DMH Lep-Rb vGat neurons) likely represent crucial controllers of body energy homeostasis. More clearly defining the molecular phenotype of DMH LepRb cells that contribute to the phenotype of early developmental LepRb deficiency and that are crucial for the control of energy balance represents an important goal of future research.

Animals
All procedures performed on animals were approved by the University of Michigan Institutional Animal Care and Use Committee and in accordance with AAALAC and NIH guidelines. All mice were bred in our colony in the Unit for Laboratory Animal Management at the University of Michigan. All mice were provided with water ad libitum and housed in temperature-controlled rooms on a 12-hour/12-hour lightedark cycle. All mice were provided ad libitum access to standard chow diet (Purina Lab Diet 5001). Rosa26 eGFP-L10a [17], Lepr flox [30], Htr2c Cre [25], Agrp Cre (Jax 012899) [23], Pomc-Cre (Jax 005965) [8], Nos1 Cre (Jax 017526) [13], and vGat Cre (Jax 016962) [12] mice have been previously described. All mice were weaned at 21 of age and group-housed with littermates of the same sex unless otherwise stated.

Ghrh Cre generation
To generate Ghrh Cre mice, a selection cassette containing the porcine teschoviral 2 A cleavage sequence linked to Cre recombinase and a FRT-flanked neomycin resistance gene was targeted to replace the stop codon of the Ghrh gene in a bacterial artificial chromosome (Children's Hospital Oakland Research Institute). A targeting plasmid containing the Cre coding sequences plus the selection cassette and w4 kb genomic sequence upstream and downstream of the Ghrh stop codon was isolated and used for embryonic stem cell targeting by the University of Michigan Transgenic Core. Correctly targeted clones were identified by loss of native allele quantitative PCR from ES clone DNA. Chimeric animals generated from blastocyst implantation were then bred for germline transmission of the Ghrh Cre allele. Flp-deleter mice were then used to remove the neomycin selection cassette. Genotyping was by allele-specific PCR.

Longitudinal study
All Cre mouse lines were crossed several times to Lepr fl/fl mice to obtain breeders to generate study mice. For most lines, Cre/þ;Lepr fl/fl mice were bred to Lepr fl/fl mice such that roughly half of all animals would be mutants and the other half would be littermate controls. Htr2c resides on the Y chromosome, so only female Htr2c Cre breeders were used, and all study males were Htr2c Cre/y ;Lepr fl/fl hemizygotes, while all females were Htr2c Cre/þ ;Lepr fl/fl . For vGat Cre , however, vGat Cre/þ ;Lepr fl/þ mice were bred to Lepr fl/fl to generate Lepr fl/fl and vGat Cre/þ ;Lepr fl/fl mice. For Nos1 Cre , Nos1 Cre ;Lepr D/þ males were crossed to Lepr fl/fl females to generate Lepr D/þ and Nos1 Cre ;Lepr D/fl mice for study; we generated Htr2c Cre ;Lepr D/fl similarly. LepRb Ghrh KO mice were single-housed with enrichment at 4 weeks old to measure continuous food intake. Body weight and food weight were measured weekly; blood glucose (from tail vein bleeds; OneTouch Ultra 2) and snout-anus length were measured biweekly. For snout-anus length, mice were briefly anesthetized with isoflurane and gently stretched on their back while calipers (Scienceware) measured the snout-anus distance. At 12-weeks of age, animals were subjected to NMR-based (Minispec LF90ll, Bruker Optics) body composition analysis. Prior to euthanasia, serum was obtained from some animals for the determination of leptin and insulin by commercial ELISA (Crystal Chem). For all other mouse lines (Agrp Cre , Pomc-Cre, vGat Cre , Nos1 Cre , Htr2c Cre ), animals were group-housed until age 8 weeks, at which point they were weighed and subjected to body composition analysis with subsequent serum collection. Additionally, we subjected a separate cohort of LepRb Htr2c KO mice and their littermate controls to weekly body weight and biweekly blood glucose measurements.

Statistics
Data are reported as mean AE SEM; additionally, the values of all replicates are shown when feasible. Data analysis was performed in R 3.4.3. Body weight comparisons, plus insulin and leptin ELISA data, were analyzed with a one-way ANOVA with Sidak's multiple comparisons correction. Body composition analysis was conducted using a linear mixed model with fixed effects of genotype, sex, and component (fat, lean, fluid) and random effects of mouse. Longitudinal studies were conducted using a linear mixed model with fixed effects of age, sex, and, genotype and random effects of mouse. All p values from estimated marginal means calculated from linear mixed models were corrected using Tukey's multiple comparisons correction. Linear mixed models were conducted with lme4 1.1e15, lmerTest 2.0e36, and emmeans 1.1.2. Star code on graphs: *p < 0.05, **p < 0.01, ***p < 0.001.

AUTHOR CONTRIBUTIONS
ACR, MBA, JCJ, CMP, CLF, LKH, DPO, and MGM designed experiments; ACR, MBA, JCJ, CMP, CLF, NB, and DPO researched data. ACR and MGM prepared figures and wrote the initial manuscript draft. All authors edited the manuscript.