Molecular profile and response to energy deficit of leptin-receptor neurons in the lateral hypothalamus

Leptin exerts its effects on energy balance by inhibiting food intake and increasing energy expenditure via leptin receptors in the hypothalamus. While LepR neurons in the arcuate nucleus of the hypothalamus, the primary target of leptin, have been extensively studied, LepR neurons in other hypothalamic nuclei remain understudied. LepR neurons in the lateral hypothalamus contribute to leptin's effects on food intake and reward, but due to the low abundance of this population it has been difficult to study their molecular profile and responses to energy deficit. We here explore the transcriptome of LepR neurons in the LH and their response to energy deficit. Male LepR-Cre mice were injected in the LH with an AAV carrying Cre-dependent L10:GFP. Few weeks later the hypothalami from fed and food-restricted (24-h) mice were dissected and the TRAP protocol was performed, for the isolation of translating mRNAs from LepR cells in the LH, followed by RNA sequencing. After mapping and normalization, differential expression analysis was performed with DESeq2. We confirm that the isolated mRNA is enriched in LepR transcripts and other known neuropeptide markers of LepRLH neurons, of which we investigate the localization patterns in the LH. We identified novel markers of LepRLH neurons with association to energy balance and metabolic disease, such as Acvr1c, Npy1r, Itgb1, and genes that are differentially regulated by food deprivation, such as Fam46a and Rrad. Our dataset provides a reliable and extensive resource of the molecular makeup of LH LepR neurons and their response to food deprivation.

www.nature.com/scientificreports/ We used LepR-cre mice to study the transcriptome of neurons that express the long isoform (LepRb) of the leptin receptor. These cells will, besides leptin, respond to a variety of hormones and neuropeptides. TRAP is an RNA polysome immunopurification technique which facilitates the isolation of mRNA bound to GFP-tagged ribosomes, from cells that artificially express GFP fused to the ribosomal subunit L10 18 . TRAP-Seq of the total population of hypothalamic LepR cells already revealed a molecular signature 19 . However, by looking at the whole hypothalamic population of LepR cells as bulk, the large population of LepR neurons in the arcuate nucleus dominates the results. To target LepR neurons in the LH, we employ viral TRAP to isolate mRNA exclusively from LepR LH neurons. Therefore, we provide an extensive and reliable resource of LepR LH -enriched genes as well as genes differentially regulated by energy deficit, associated with energy balance and metabolic disease.

Results and discussion
In order to isolate mRNA from LepR LH cells, an AAV carrying a Cre-dependent cDNA for L10:GFP was injected into the LH of LepR-Cre adult male fed and fasted mice and 4 to 5 weeks later hypothalami were dissected and TRAP was performed, a timepoint that allows sufficient viral expression and detection of L10a:GFP in polysome fractions 17 . cDNA libraries were prepared with the CelSeq2 protocol optimized for bulk RNA and the libraries were sequenced with next generation sequencing (NGS). Analysis of the RNA-sequencing reads revealed enrichment of genes in the LepR LH population against the total hypothalamus as well as genes differentially regulated by food deprivation.

Validation of selective mRNA enrichment from LepR LH neurons. A series of pilot experiments
were performed in which different serotypes and virus titers of rAAV-flex-L10:GFP were stereotaxically injected into mouse brains with several coordinates targeting the LH, in order to achieve sufficient GFP expression targeted in the LH. GFP expression was observed in the LH of LepR-Cre (cre + /cre +) mice and it was minimal in wild-type (wt/wt) littermates ( Fig. 1a-left). The TRAP experiment was performed in 7 batches of male mice, both from the fed and fasted condition. In each batch few mice were sacrificed before performing TRAP in order to assess GFP expression (see Methods for details). Data from a total of 12 mice, showed 1256.7 ± 215.7 GFPexpressing cells per mouse, of which 83.5% ± 3.8 were localized in the LH ( Fig. 1a-right, see Methods). Viral spread upon injection in the LH caused a small proportion of labeled cells outside the LH.
Immunoprecipitated RNA from LepR LH cells (LepR LH ) as well as RNA from the whole hypothalamus (hypo) was sequenced. Following normalization of sample by Deseq2 (Median of ratios method) to an average of ~ 3,785,000 reads per sample and exclusion of lowly expressed genes (a cutoff of 5 reads per sample), we obtained a final dataset containing 13,093 of the 17,988 genes detected in the dataset. Differential expression analysis between LepR LH and whole hypothalamus (hypo) samples identified 768 genes with significantly different expression levels (with adjusted p < 0.01) (sup. Table 1). 343 genes (including GFP) were enriched in LepR LH over hypo (log2FoldChange > 1.5) and 425 genes were enriched in hypo over LepR LH (log2FoldChange < − 1.5) (Fig. 1b). As expected, GFP and LepR were both enriched in LepR LH over hypo (38.3-fold and 6.4-fold respectively, p < 0.01, Fig. 1b) with GFP being the most significantly enriched gene of the whole dataset. The well-established markers of LepR LH neurons, Gal and Nts, were also enriched in LepR LH over hypo (3.8-fold and 24.2-fold respectively, p < 0.01, Fig. 1b). We detected a significant enrichment of lateral hypothalamic genes by comparing the 342 genes (excluding GFP) enriched in LepR LH with upregulated genes across brain regions provided by the Allen Brain Atlas using the Enrichr platform (sup. Table 2), confirming effective targeting of the LH. Thus, our dataset identifies the molecular makeup of the LepR LH neurons.
Next, we asked whether specific cell types are enriched in LepR LH neurons compared to the whole hypothalamus. Non-neuronal markers, such as Gfap for astrocytes, Mobp and Olig1 for oligodendrocytes, Cx3cr1 for microglia, and Rax for tanycytes, as well as markers of distinct hypothalamic populations, such as Agrp and Npy (Fig. 1c), and LH-specific Pmch and Hcrt (not shown) were less abundant in IP (IP = immunoprecipetated fraction enriched in LepR LH ) than IN (IN = input RNA from all cells of LH) samples. RNA enriched from LepR LH cells showed enrichment of GABAergic markers Slc32a1, Gad1 and Gad2, and low levels of glutamatergic marker Slc17a6, showing that we enriched for GABAergic LepR LH neurons (Fig. 1d). These results also confirm the literature that has extensively established LepR LH neurons as GABAergic 15 .

Enrichment and localization of established markers of LepR LH neurons. Besides their established
GABAergic nature, literature has shown the expression of a handful of neuropeptides and receptors in LepR LH neurons. LepR neurons in the LH express Neurotensin (Nts), Galanin (Gal), CART peptide (Cartpt), Tachykinin 1 (Tac1) as well as Corticosterone-releasing hormone (Crh) 15,19,20 . In the LH, LepR neurons have been reported to express melanocortin receptors Mc3r and Mc4r as well the receptor for calcitonin Calcr [21][22][23] . We confirmed the enrichment of all of these genes in the IP fraction (immunoprecipetated fraction enriched in LepR LH ) over IN (input RNA from all cells of LH) in our dataset (Figs. 1b and 3). Moreover, we show that neuropeptides Calca, Cbln2, Pomc and Vgf were also enriched in LepR LH neurons compared to the whole LH (Fig. 3).
We next determined the co-localization pattern of the five neuropeptides with LepR in the LH. To this end, we performed fluorescent in situ hybridization (FISH) for Nts, Gal, Cartpt, Crh and Tac1 on tissue from LepR-tdTomato mice stained for tdTomato. We focused on the LH (including the Pernifornical area (PeF), Zona Incerta (ZI), Tuberal nucleus (Tu)) and explored co-localization of markers with tdTomato (LepR). Cartpt, Nts and Tac1 were expressed in almost one third of LepR cells in the LH, while Gal and Crh were expressed in 15% and 12% of the cells respectively (Fig. 2a). Literature shows that more than half (60%) of LepR LH neurons express the anorexigenic peptide Neurotensin (Nts), while 20-44% of pSTAT3 positive cells in the LH express Galanin 15,20 . The difference observed between the two sets of data can be attributed to different detection methods used-for   www.nature.com/scientificreports/ Co-localization of each marker with LepR LH cells was observed in the dorsal part of the LH/Zona Incerta (ZI), Pernifornical area (PeF) and ventral part of the LH/tuberal nucleus (Fig. 2b). Based on literature, Neurotensin is almost fully co-expressed (95%) with Galanin, and CART peptide is expressed in half (56%) of Galanin positive neurons in the LH 20 . In line with this, we observed that Gal co-localizes its expression with LepR in the same subregions as Nts as well as Cartpt. In the case of Crh, co-localization is observed in the rostral part of the LH and PeF. Therefore, it is highly probable that all three neuropeptides could have overlapping expression in the same cell, while we did not observe a distinct pattern of Gal, Nts and Crh expression in LepR LH neurons.
Even though Tac1 and Cartpt co-localized with LepR in the rostral part of the LH/ZI and PeF, we also observed distinct patterns of expression for them. We observed Cartpt/LepR cells in the ventral LH/MTu rostrally (Bregma − 1.58), while there were Tac1/LepR cells also localized in the ventral LH/MTu caudally (Bregmas − 2.06 and − 2.30). These findings suggest that in contrast to the Arc, where LepR is mostly expressed in distinct populations of neurons and has directly opposing functions on Agrp and POMC neurons, LepR in the LH is found in neurons with diverse profiles, with overlapping as well as distinct expression of neuropeptides. Therefore, leptin may act on a variety of distinct neurons with partially overlapping genetic identity.
Novel markers of LepR LH neurons. We next focused on novel markers of LepR LH neurons. Using the Uniprot database, we assigned the TRAP-enriched genes to gene ontology and molecular function terms (sup. Table 3). Where possible, LepR LH -enriched genes were split into main sub-categories based on their molecular function: secreted proteins, enzymes, catalytic receptors, GPCRs, ion channels, transporters and transcription factors (see Column "Type" in sup. Table 3 and Methods for databases used for annotation). We continued with the top 10 significant genes of each category (Fig. 3). The genes were analyzed for their expression in the hypothalamus and particularly the LH, based on ISH data of Allen Brain Atlas (referred to as ABA from now on) and mouse GTEx data from the protein atlas. Moreover, in order to assess relevance, we performed literature search (PubMed) using as search terms the gene name combined with either hypothalamus, diabetes or obesity as keywords.
LepR LH -enriched markers with selective expression in the LH and other hypothalamic nuclei. We identified genes with selective expression in the LH and other hypothalamic nuclei where LepR is expressed, based on Allen Brain Atlas ISH data (ABA).
Catalytic receptors. Itgb1, which codes for β1 integrin associates with integrins α1 and α2 to form integrin complexes which function as collagen receptors. It shows prominent expression in the LH, tubelar nucleus (Tu), PVN and weakly in Arc (ABA), but its function in the hypothalamus remains unexplored. Acvr1c (also known as Alk-7) is expressed in the Arc, VMH and weakly in the LH (ABA) and is downregulated in adipose tissue in obese subjects 24 , while variations in its DNA sequence are associated with protection against diabetes type 2 25 . Besides being enriched in Agrp neurons in the Arc 12 , nothing is known about the function of Acvr1c in the hypothalamus.
Secreted proteins. The secreted protein Stc1 codes for Stanniocalcin 1, involved in lipid synthesis in brown adipose tissue 26 and its mRNA is expressed very weakly in the LH and DMH in the hypothalamus (ABA). Another secreted protein, Cthrc1 (collagen triple-helix repeat-containing 1), shows expression in the LH (ABA) as well as in the PVN and SON and it regulates lipid storage and cellular glycogen levels 27 . Serpina3N, the gene encoding the anti-protease alpha-1-antichymotrypsin (α 1 AC), is expressed most prominently in the Arc and weakly in the LH (ABA), while it has also been localized in the DMH and VMH and its expression in the hypothalamus was upregulated upon exposure to high-fat diet and leptin 28 .
Transcription factors. Besides the LH, the transcription factor Asb4 is also expressed in the DMH and Arc (ABA) and has been identified in a GABAergic cluster of the LH 29 . This cluster also expressed Sst and Col25a1; nevertheless, there was no enrichment of these markers in LepR LH neurons of our dataset, suggesting that LepR/ Asb4 expressing cells form a separate population of neurons. Dlx1, a transcription factor involved in the production of Agrp and Ghrh neurons during development 30 , is expressed in the DMH, Arc, ZI and lightly in the LH (ABA).
GPCRs. Besides the Arc, the GPCR Y1 for Npy (Npy1r) is also expressed in the LH (ABA & 31 ) and its signaling mediates NPY's effects on chow intake 32 . The LH receives input from Npy/Agrp cells in the Arc 33 , but also from Npy-expressing cells in the DMH 34 . It is also highly-probable that Npy1r LH cells receive input directly from Npyexpressing neurons resided in the LH, which are glucose-sensing 35 . Finally, another GPCR, Gpr101, is mainly expressed in the DMH, VMH, PVN as well as the LH (ABA), and its expression in the posterior hypothalamus is increased by fasting but decreased in the ob/ob mouse 36 .
A handful of genes with an association with energy balance were identified to be enriched in LepR LH neurons and have proven expression in the LH and thus can be used as marker genes and be investigated further.
LepR LH -enriched markers with widespread expression in the hypothalamus (ABA). We continued exploring potential markers of LepR LH neurons by looking at which genes have widespread expression in the hypothalamus based on ISH data (ABA) and have relevance with obesity and diabetes. The transcription factor Isl1 has an important role in melanocortin neuron development in the hypothalamus 37 , and has been found to be associated with diabetes type 1 and 2 38 40,41 . In the Arc, Cacna1e regulates leptin-induced excitation of POMC neurons with possible effects on hepatic insulin production and insulin resistance 42 and it could potentially have a similar function in LepR LH neurons. Kdm6b, a selective H3K27me3 demethylase, has widespread expression throughout the body. Its knock-out in RIP-Cre neurons in the Arc is associated with increased obesity 43 , suggesting that it may also regulate the function of LepR LH neurons.
Transcriptional response of TRAP-targeted cells to fasting. In order to determine the transcriptional response of LepR LH neurons to food restriction, we fasted mice for 24 h before TRAP-seq and performed differential expression analysis between fed and fasted conditions. Weight gain as well as food and water intake were monitored before the fasting protocol, showing no differences between fed and fasted mice (Fig. S1a,b), while the 24-h food restriction caused significant weight loss (Fig. S1c). Differential expression (DE) analysis of LepR LH -enriched mRNA between fasted and fed mice revealed differences in gene expression levels induced by energy deficit (sup. Table 4). It is known that 24-h fasting induces Lepr upregulation in the Arc and VMH 44 , while total hypothalamic expression of Nts, Gal, Tac1 and Cartpt is not changed [45][46][47] . Crh is upregulated by fasting in the hypothalamus 47 , a change which is probably driven by the PVN, in which Crh is most prominently expressed. However, it is not known how these neuropeptides are regulated specifically in the LH or in LepR cells. Therefore, we examined whether food restriction would change the expression levels of the five neuropeptides as well as LepR in LepR LH -enriched samples from fasted and fed mice. There was no difference in Lepr, Nts, Gal, Cartpt or Crh expression levels, while Tac1 was slightly upregulated by fasting (log2FoldChange = 0.84, p = 0.002).
When looking at the top regulated genes by fasting (with adjusted p < 0.05), there were 10 significantly DE genes with adjusted p < 0.05 out of which 3 were upregulated (Fam107a, Ubr3a, Ormdl1) and 7 were downregulated by fasting (Fosl2, Rrad, Crem, Fos, Fam46a, Rgs4, Hdac5) (Fig. 4a). Within the downregulated genes, we observed the immediate early gene Fos, confirming previous immunohistochemical data 15 , which show that food deprivation induces a decrease in the percentage of LepR LH neurons that engage in cFos, from 12% in fed mice to 5% in fasted.
Fasting downregulated the expression of transcription factors Fosl2, Crem, Hdac5, all of which are associated with energy balance. Hdac5 (Histone deacetylase 5) is a key epigenetic regulator of leptin's actions in the hypothalamus, by controling leptin signaling via STAT3 deacetylation, while it regulates POMC (and not Agrp) expression and protects against diet-induced obesity 48 . The transcription factor Fosl2 promotes leptin gene expression in human and mouse adipocytes 49 . Crem belongs to the family of transcription factors CREB that promote insulin resistance in obesity in adipocytes 50 . It remains to be determined what role these transcription factors play in LepR LH neurons.
The enzyme Fam46a, a translocator of leptin in adipocytes, was also downregulated by fasting. Fam46a is expressed prominently in the LH, PVN, SON and DMH (ABA) (Fig. 4b) and is associated with Body mass index (BMI) 51 . Rgs4 (regulator of G-protein signaling-4) is a negative regulator of insulin release from pancreatic β-cells in vitro and in vivo (Ruiz de Azul et al., 2010, PNAS) and controls fatty acid and glucose homeostasis 52 , while its expression in the whole hypothalamus is downregulated by fasting 46 . Finally, Rrad (Ras Associated with Diabetes), which is expressed in the LH, DMH and VMH (ABA), is associated with type II diabetes by affecting insulin resistance and glucose intolerance 53 and is overexpressed in the white adipose tissue of obesity-prone rats 54 . Although these proteins have been identified to affect body weight and to act in adipocytes and pancreatic β-cells, their role in LepR LH neurons and link to regulation of energy balance remains to be determined.
Overall, food deprivation caused changes in the expression of many key players in energy balance and metabolic diseases. It would be interesting to characterize these genes as mediators of leptin's action in the LH and if and how they regulate neuropeptide expression as well as to unravel their role in energy homeostasis.

Conclusion
Using viral RNA TRAP, we successfully profiled translated mRNAs of LepR LH neurons. We identified several genes enriched in LepR LH neurons with relevance to energy balance and metabolic diseases providing novel avenues for future studies. While expression of few of these genes (e.g. Acvr1c, Npy1r, Itgb1, Cacna1e) in the LH and in other hypothalamic nuclei has already been demonstrated, further characterization of their localization, function and their relation to leptin would provide insight on their metabolic function.
We also explored the transcriptional response of LepR LH neurons to fasting. Neither Lepr nor four out of five established neuropeptides (Nts, Gal, Cartpt, Crh) were differentially regulated by food deprivation of 24 h, with Tac1 being slightly increased. However, among the top DE genes, the transcription factors Hdac5, Fosl2 and Crem were downregulated, indicating that food deprivation of 24 h induces changes in key transcriptional players. A longer food deprivation could potentially induce more robust transcriptional changes, although a 24 h fast is already a serious metabolic stress for a mouse. Nevertheless, exploring different fasting protocols that can be maintained for longer periods could add further knowledge to how LepR LH neurons respond to energy deficit. Fasting also induced changes in expression levels of genes not previously reported to be affected in the whole hypothalamus with association to energy balance, such as the enzyme Fam46a as well as the GTP-binding protein Rrad, which are expressed prominently in the LH.
Overall, our dataset captures important information of rare cell types. We show that GABAergic LepR LH neurons are diverse, reveal a set of genes that discriminate them from other LH cell types, and identify several obesity-and metabolism-related genes involved in their response to fasting. Our results can be the starting point of further characterization and studies about the involvement of this population, or its subpopulations, in diseases related to metabolism and energy balance. AAV production. Plasmid pAAV-FLEX-EGFPL10a (kind gift from Alexander Nectow 17 ) was packaged in AAV serotype 5 and rAAV1-FLEX-EGFPL10a was generated as described earlier 55 . Virus titer (number of genomic copies) was determined with RT-PCR with primers binding on the wPRE element. TRAP protocol. The protocol previously published was followed 18 and was performed in 7 batches of mice. www.nature.com/scientificreports/ RNA sequencing analysis. RNA-seq reads were aligned to annotated exons using the mm10mouse reference genome and to the L10:GFP sequence using bwa mapping. Reads that did not align or aligned to multiple locations were discarded. Two samples (one from fed and one from fasted groups) were removed: in the former sample the IP fraction had very low number of reads and in the latter sample the IP fraction had extremely high amount of reads for LepR. Mitochondrial genes and gene Rn45s were excluded. DESeq2 57 was used to normalize the reads according to the dispersion profile of the whole dataset and to perform differential expression analysis between LepR LH and hypo fed samples as well as between fed and fasted in IP samples. We calculated fold-enrichment as normalized reads in LepR LH divided by normalized reads in hypo samples. Cloning and DIG probe labeling. PCR with forward and reverse primers (see Table 1 TGT TGC AGA TCG AA  ACA TGC TTC AAT TTG TGT GGCT   Crh  TGC GTG CTT TCT GAA GAG GG  TTT TGG CCA AGC GCA ACA TT   Gal  ATC CAG CCC GCC ACT CTT CA  ACA GCT TCA AAG CAG AGA ACAGA   Nts  AGA AGA AGA TGT GAG AGC CCTG TGC TTT GGG TTA ATA ACG CTCC   Tac1  ATG AAA ATC CTC GTG GCC GT  TCA TAC AAT GAC TGA AGA CCA   with anti-sense RNA probes for Nts, Gal, Cartpt, Tac1 and Crh ("Neuropeptide"-red channel) followed by immunohistochemistry (IHC) for tdTomato ("LepR"-green channel) and imaging. The percentage of Neuropeptide + LepR + cells out of the total LepR + in the LH was quantified, as represented in the barplots (mean, n = 3 hypothalami). (B) Localization of Neuropeptide + LepR + cells was observed under the microscope and drawn on representative graphic images of coronal sections from the mouse Brain Atlas 58  However, it was not possible to detect GFP expression under fluorescent light during the dissection, probably due to the low number of targeted cells and the autofluorescence of the tissue. Therefore, to assess expression of GFP in the batches of mice that were used for TRAP, the brains of 1 or 2 mice per batch were fixed 2 weeks after surgery and hypothalamic sections were processed in order to immunohistochemically detect GFP (see below). At the microscope we confirmed that most of GFP cells were located in the LH. For quantifications of GFP expression, 10 × magnification pictures were taken with an epi-fluorescent microscope (Zeiss Scope A1, ZEISS, Germany). Five sections with an interval of 0,2-0,3 mm ranging from − 1.1 to − 2.4 mm caudal to bregma (as indicated in 58 ), were selected. GFP cells in the LH, ZI and PeF were manually counted using the Cell Counter plugin in ImageJ. For quantifications of co-localized cells regarding FISH for neuropeptides and tdTomato IHC, the same settings were used, except that 20 × magnification of the LH, ZI, PeF, Tu were taken. We quantified the total number of tdTomato cells located in the LH/ZI/PeF/Tu and the number of tdTomato cells co-expressing the corresponding neuropeptide, in order to calculate the percentage of LepR LH cells expressing each neuropeptide.

Statistical analyses.
Regarding body weight and food intake data, statistical analyses were performed with GraphPad Prism 7.0 (Graphpad Software, San Diego, CA, US). Data was checked for normality before doing statistical tests.

Data availability
All data generated and analysed during this study are included in this article and its supplementary information files. Additional information related to this study can be obtained from the corresponding author on reasonable request.