Epigenomic and metabolic responses of hypothalamic POMC neurons to gestational nicotine exposure in adult offspring

Background Epidemiological and animal studies have reported that prenatal nicotine exposure (PNE) leads to obesity and type-2 diabetes in offspring. Central leptin-melanocortin signaling via hypothalamic arcuate proopiomelanocortin (POMC) neurons is crucial for the regulation of energy and glucose balance. Furthermore, hypothalamic POMC neurons were recently found to mediate the anorectic effects of nicotine through activation of acetylcholine receptors. Here, we hypothesized that PNE impairs leptin-melanocortinergic regulation of energy balance in first-generation offspring by altering expression of long non-coding RNAs (lncRNAs) putatively regulating development and/or function of hypothalamic POMC neurons. Methods C57BL/6J females were exposed ad libitum to nicotine through drinking water and crossed with C57BL/6J males. Nicotine exposure was sustained during pregnancy and discontinued at parturition. Offspring development was monitored from birth into adulthood. From the age of 8 weeks, central leptin-melanocortin signaling, diabetes, and obesity susceptibility were assessed in male offspring fed a low-fat or high-fat diet for 16 weeks. Nicotine-exposed and non-exposed C57BL/6J females were also crossed with C57BL/6J males expressing the enhanced green fluorescent protein specifically in POMC neurons. Transgenic male offspring were subjected to laser microdissections and RNA sequencing (RNA-seq) analysis of POMC neurons for determination of nicotine-induced gene expression changes and regulatory lncRNA/protein-coding gene interactions. Results Contrary to expectation based on previous studies, PNE did not impair but rather enhanced leptin-melanocortinergic regulation of energy and glucose balance via POMC neurons in offspring. RNA-seq of laser microdissected POMC neurons revealed only one consistent change, upregulation of Gm15851, a lncRNA of yet unidentified function, in nicotine-exposed offspring. RNA-seq further suggested 82 cis-regulatory lncRNA/protein-coding gene interactions, 19 of which involved coding genes regulating neural development and/or function, and revealed expression of several previously unidentified metabolic, neuroendocrine, and neurodevelopment pathways in POMC neurons. Conclusions PNE does not result in obesity and type 2 diabetes but instead enhances leptin-melanocortinergic feeding and body weight regulation via POMC neurons in adult offspring. PNE leads to selective upregulation of Gm15851, a lncRNA, in adult offspring POMC neurons. POMC neurons express several lncRNAs and pathways possibly regulating POMC neuronal development and/or function. Electronic supplementary material The online version of this article (doi:10.1186/s13073-016-0348-2) contains supplementary material, which is available to authorized users.


Background
The adipocyte-secreted hormone leptin has critical anorectic and body weight-lowering actions that are mediated by the central melanocortin system [1,2]. The melanocortin system consists of first order signaling neurons in the hypothalamic arcuate nucleus comprising neuropeptide Y (NPY) and Agouti-related protein (AgRP) co-expressing neurons and of proopiomelanocortin (POMC)-expressing neurons. These first order signaling neurons project to second order signaling neurons expressing melanocortin 3 and 4 receptors (MC3/ 4-Rs) located in several hypothalamic and extrahypothalamic areas, where they exert anorectic actions, including the periventricular nucleus (PVN), a satiety center, and the lateral hypothalamic area (LHA), a feeding center. POMC is a precursor polypeptide for α-melanocytestimulating hormone (α-MSH). α-MSH acts as an agonist to MC3/4-Rs of second order signaling neurons in the PVN to inhibit food intake and increase energy expenditure. AgRP is an antagonist of α-MSH at MC3/4-Rs and potent activator of food intake and body weight gain. Leptin regulates transcription of the genes encoding POMC and AgRP. Leptin binds to leptin receptors expressed in POMC and NPY/AgRP neurons. This triggers phosphorylation and activation of Janus kinase 2 (JAK2), which in turn phosphorylates signal transducer and activator of transcription 3 (P-STAT3). P-STAT3 homodimerizes and translocates to the nucleus to promote transcription of the POMC gene and inhibit transcription of the AgRP gene [3]. Leptin also stimulates release of α-MSH from POMC neurons [4] and inhibits release of AgRP from NPY/AgRP neurons [5]. Independently of decreasing food consumption and body weight, leptin signaling via hypothalamic POMC neurons potently decreases glycemia and increases locomotor activity, thereby preventing hyperglycemia and normalizing physical activity of morbidly obese, severely diabetic, and hypoactive, leptinreceptor deficient Lepr db/db mice [6].
Mutations of the leptin, leptin receptor, POMC, and MC4-R genes lead to severe obesity in rodents and in humans, underscoring the importance of central leptinmelanocortin signaling for the regulation of energy balance [7][8][9]. Furthermore, common human obesity is frequently associated with leptin resistance characterized by an inability of leptin to decrease body weight.
Nicotine has anorectic effects. One underlying mechanism is the activation of hypothalamic POMC neurons by nicotinic α3β4 acetylcholine receptors and the subsequent activation of MC4-R expressing target neurons in the PVN [10]. The metabolic consequences of gestational nicotine exposure in first-generation offspring have been investigated in human epidemiological studies [11][12][13][14] and rodent models [15][16][17][18]. These studies concluded that gestational nicotine exposure increases the risk for obesity and type 2 diabetes in offspring. Furthermore, it was reported that in utero or early postnatal nicotine exposure upregulates POMC mRNA in neonate rhesus macaque [19,20]. These observations led us to hypothesize that gestational nicotine exposure impacts leptin-melanocortinergic regulation of energy balance in first-generation offspring.
More than 70 % of the mammalian genome is transcribed as non-coding RNAs (ncRNAs) of various sizes ranging from 20 nucleotides to over 100 kb [21,22]. NcRNAs are subdivided into short and long ncRNAs (lncRNAs), which are shorter and longer than 200 nucleotides, respectively [23], as well as processed transcripts, which do not contain an open reading frame, do not contain retained introns, and cannot be placed in the short and long ncRNA group. Similar to protein-coding mRNAs, lncRNAs can be spliced, polyadenylated, and capped [23]. LncRNAs are subdivided into: (1) antisense RNAs, which are transcribed from the opposite DNA strand of a protein-coding gene overlapping its intronic and exonic sequences; (2) long intergenic non-coding RNAs (lincRNA) transcribed from intergenic DNA regions; (3) sense-intronic RNAs transcribed from introns of a protein-coding gene on the same DNA strand with no overlap of exonic sequence; (4) sense-overlapping RNAs containing a protein-coding gene in one of its introns on the same DNA strand with no overlap of exonic sequence; and (5) 3′-overlapping ncRNAs, which are transcribed from the 3′-untranslated region of a larger gene. LncRNAs are predominantly localized in the nucleus at usually much lower expression levels than mRNAs.
While lncRNAs were initially believed to represent transcriptional noise, studies in recent years have revealed developmental-, tissue-and cell-type-specific expression of lncRNAs [24,25] and regulatory roles in important biological processes such as X-chromosome inactivation in females, silencing of tumor suppressor genes, mediation of DNA damage and cellular stress responses, gene imprinting, heterochromatin spreading across DNA insulator sequences, regulation of stem cell pluripotency, cell fate specification, and neural development [26][27][28][29][30]. However, although many different biological functions have been assigned to lncRNAs, most lncRNAs lack functional annotations. LncRNA genes show poorer sequence conservation than protein-coding genes although well-conserved lncRNAs with biologically important functions exist [31]. The number of lncRNA genes has dramatically increased in late evolution and appears to scale with genome size. Meanwhile, the number of protein-coding genes has not markedly increased during late evolution. Therefore, one emergent hypothesis is that the increase in lncRNA genes contributes to regulatory and organismal complexity [32]. LncRNAs can regulate expression of flanking or overlapping coding genes ("cis-regulation") or far distant coding genes located on the same or another chromosome ("trans-regulation") in a positive (concordant) or negative (discordant) manner [25,33,34]. LncRNA-guided gene regulation involves epigenetic, transcriptional, and post-transcriptional mechanisms. LncRNAs are thought to provide scaffolds for histone methyltransferases, histone acetylases, and DNA methyltransferases [26,27], and can regulate splicing, editing, or degradation of protein-coding RNAs [26].
Here we interrogated whether PNE alters expression of lncRNAs with putative roles in hypothalamic POMC neuronal development/function and leptinmelanocortinergic regulation of energy balance in first-generation offspring in a mouse model of human maternal nicotine exposure from adolescence until parturition. We found that PNE does not impair but rather enhances leptin-melanocortinergic regulation of energy balance in first-generation offspring and selectively upregulates expression of Gm15851, a lncRNA of yet unidentified function, in hypothalamic POMC neurons. Furthermore, we report expression of several signaling pathways and lncRNAs that might regulate development and function of hypothalamic POMC neurons.

Animals
Animals used in the study were from the C57BL/6J mouse strain (RRID:IMSR_JAX:000664, Jackson Laboratories, Bar Harbor, ME, USA) or transgenic mice expressing the enhanced green fluorescent protein (EGFP) in hypothalamic POMC neurons (C57BL/6J-Tg(Pomc-EGFP)1Low/J, RRID:IMSR_JAX:009593, Jackson Laboratories, Bar Harbor, ME, USA). Metabolic, hormonal, hypothalamic gene expression and leptin-signaling studies were conducted in C57BL/6J male offspring. RNA sequencing (RNA-seq) studies of laser microdissected POMC neurons were conducted in POMC-EGFP transgenic male offspring. POMC-EGFP transgenic offspring were tail biopsied at postnatal day 17 and genotyped using established PCR protocols provided by the Jackson Laboratories. Offspring were weaned at postnatal day 21 and kept in groups of 3-4 mice per cage. Male offspring were singly housed with environmental enrichment from the age of 8 weeks onwards until the end of the study. All mice were housed in a pathogen-free barrier animal facility and kept in a temperature-controlled (22 ± 0.5°C) and humidity-controlled (50 %) animal room on a 12 h light/dark schedule with light on at 07:00 and free access to food and water.

Prenatal nicotine exposure
Six-week-old females were treated for 4 weeks ad libitum with drinking water containing 200 μg/mL nicotine hydrogen tartrate sweetened with 2 % saccharin (w/v), pH 7.4 [35]. Controls were administered pH-matched drinking water containing the equivalent amount of tartaric acid and 2 % saccharin (w/v). The inclusion of saccharin was required because nicotine in the drinking water causes taste-aversion [36,37] and reduces maternal fluid intake [35]. Drinking solutions were changed twice a week. At the age of 10 weeks, females were mated to unexposed males. Nicotine administration continued throughout mating and pregnancy and was discontinued at parturition. Mothers were allowed only one pregnancy.

Determination of plasma cotinine
Plasma cotinine levels were determined using a Cotinine (Mouse/Rat) enzyme linked immunosorbent assay kit (Abnova, Walnut, CA, USA) following the manufacturer's directions. All samples were assayed in duplicates. The intra-assay coefficient of variation (CV) of the cotinine ELISA measurements was 3.9 %.

Diets
All animals were fed a standard (low-fat) diet (STD) with a metabolizable energy density of 3.1 kcal/g derived from 28.7 % kcal protein, 58.53 % kcal carbohydrates, and 12.73 % kcal fat (5010, LabDiet, St Louis, MO, USA). A subset of male offspring was subjected to a high-fat diet (HFD) with a metabolizable energy density of 4.7 kcal/g derived from 17 % kcal protein, 43 % kcal carbohydrates, and 41 % kcal fat, containing 0.21 % (w/w) Cholesterol (Western Diet, Research Diets, Inc., New Brunswick, NJ, USA) starting from the age of 8 weeks.

Determination of energy balance
Body composition (body fat content, lean mass, and body fluid) was determined by nuclear magnetic resonance (minispec TD-NMR Analyzer, Bruker Optics, Inc., Billerica, MA, USA). Food consumption of singly housed animals was measured twice a week. Fresh food pellets were provided twice a week to avoid temperaturedependent spoilage. Any residual bits of food in the bedding were included in the measurements. A comprehensive laboratory animal monitoring system (Columbus Instruments, Columbus, OH, USA) determined oxygen consumption, carbon dioxide production, heat production, locomotor activity, and food intake simultaneously in individual animals. Measurements were taken in metabolic cages placed in a temperature-controlled enclosure set at 23.5°C. Mice were acclimated to the metabolic cages for 72 h prior to measurements for an additional 72 h.

Determination of glucose balance
Blood glucose measurements were carried out with a OneTouch glucometer (LifeScan, Inc.). Blood glucose and insulin concentrations were determined by tail venipuncture and submandibular bleeding, respectively, between 09:00 and 10:00 following 14 h of fasting. Insulin was measured by enzyme linked immunosorbent assay (Ultra Sensitive Mouse Insulin ELISA Kit, Crystal Chem, Inc., Chicago, IL, USA). All samples were assayed in duplicates. The intra-assay CV of the insulin ELISA measurements was 9.9 %. Glucose tolerance tests (GTTs) were conducted in 14 h fasted mice by injecting i.p. glucose 1 g/kg body weight. Insulin tolerance tests (ITTs) were conducted in 6 h fasted mice by injecting i.p. Insulin (Humulin R, Eli Lilly, Indianapolis, IN, USA) 0.5 IU/kg body weight. Blood glucose concentrations for the GTT and ITT were measured from the tail tip. Calculating the area under the curve (AUC) of glucose concentration-time point graphs using Prism 5 (RRID:SCR_002798, GraphPad Software, La Jolla, CA, USA) quantitatively assessed glucose and insulin tolerance.

Determination of plasma lipids
All plasma lipids were determined using VITROS® Chemistry Products and analyzer (Ortho-Clinical Diagnostics, Inc., Rochester, NY, USA). Total cholesterol (Tot Chol) was measured by the VITROS CHOL Slide method. High-density lipoprotein cholesterol (HDL-C) was determined by the VITROS direct HDL-C (dHDL) slide assay. Triglycerides were measured by the VITROS TRIG Slide method. Low density lipoprotein cholesterol (LDL-C) and Very low density lipoprotein cholesterol (VLDL-C) concentrations were calculated as follows: VLDL-C = TG/5; LDL-C = Tot Chol -HDL-C -VLDL-C.

Refeeding experiments and determination of Leptin and MTII sensitivity
Refeeding experiments were conducted as described previously [38]. Mice were fasted overnight for 12 h and then injected i.p. leptin 5 mg/kg (A.F. Parlow National Hormone and Peptide Program, Torrance, CA, USA), Melanotan II (MTII) 5 mg/kg (Tocris, Ellisville, MO, USA), or the equivalent volume of vehicle (phosphate buffered saline (PBS)). Body weight was determined prior to injection, as well as 1 h, 2 h, 4 h, 8 h, and 24 h post injection. A known amount of food was added to each cage immediately after injection and food intake was measured 1 h, 2 h, 4 h, 8 h, and 24 h post injection. Mice had unlimited access to water during the entire experiment. Plasma leptin concentrations were determined by ELISA (Mouse Leptin ELISA Kit, Crystal Chem, Inc., Chicago, IL, USA). All samples were assayed in duplicates. The intra-assay CV for the leptin ELISA measurements was 5.1 %.

Determination of hypothalamic POMC mRNA expression
Total RNA was isolated from microdissected whole hypothalami using the RNeasy Mini kit (Qiagen). Total RNA (500 ng) was reverse transcribed (Superscript III first strand cDNA synthesis kit; Invitrogen, Thermo-Fisher Scientific). Real-time PCR reactions were carried out on an ABI 7900HT light cycler using the Taqman Universal Master Mix and validated gene-specific Taqman expression assays for proopiomelanocortin and βactin (Applied Biosystems, ThermoFisher Scientific). POMC mRNA was normalized to β-actin mRNA by the ΔΔCt method.

Laser micro-dissections and RNA extractions of hypothalamic POMC neurons
Prenatally nicotine-exposed and non-exposed transgenic adult male mice expressing the enhanced green fluorescent protein in hypothalamic POMC neurons [40] were euthanized in week 10 of the STD by CO 2 inhalation followed by cervical dislocation and decapitation. Brains were quickly removed, fresh-frozen in dry ice, and stored at −80°C. Brains were sectioned on a cryostat Leica CM 3050 S at 15 μm. Sections were attached onto Director® slides (Expression Pathology, Inc., Rockville, MD, USA), dehydrated in acetone for 1 min at room temperature, and desiccated for 2 min in a closed petri-dish containing dessicant (Drierite). Slides were immediately placed upside-down on the stage of a laser microdissection microscope Leica AS LMD equipped with a HCX PL Fluotar 20×/0.4 NA lens and a Hitachi HV-C20A camera. POMC-GFP neurons in the mediobasal hypothalamus were visualized by the green fluorescent appearance of their cell bodies, which were collected directly in 20 μL lysis buffer of the RNAequous microextraction kit (Life Technologies, ThermoFisher Scientific). Collections from individual sections were not allowed to continue for more than 30 min. Lysates were immediately frozen on dry ice and stored at −80°C. Between 400 and 600 POMC neurons were isolated from the entire hypothalamus of each mouse brain. RNA was isolated following the directions provided in the RNAequous microextraction kit. RNA samples were run on an electropherogram (2100 Bioanalyzer, Agilent Technologies) for determination of RNA quality, concentration, and yield. Samples with RNA integrity numbers (RIN) above 6 were used for generation of RNA-seq libraries.

RNA sequencing
A range of 1-2 offspring coming from five different nicotine-exposed dams (n = 6 PNE offspring total) and 1-2 offspring coming from four different non-exposed dams (n = 5 control offspring total) were subjected to RNA-seq. Directional RNA-seq libraries were generated from a range of 200-400 pg total RNA using the Ovation Single Cell RNA-seq kit (NuGen, San Carlos, CA, USA) following the manufacturer's directions. All RNA-seq libraries were run on an electropherogram (2100 Bioanalyzer, Agilent Technologies) to confirm the expected fragment size distribution. Libraries were paired-end sequenced on Illumina's HiSeq2000 sequencer. Of all read-pairs, 90-95 % passed Illumina's quality filters. For subsequent alignments, the first eight nucleotides of the forward read were trimmed according to the manufacturer's directions for analysis of Ovation Single Cell RNA-seq libraries. Pass-filter read-pairs were aligned to the mouse reference mm10 genome using STAR [41]. Read-pairs aligning to the genome with more than two mismatches or aligning to more than one site of the genome were discarded from further analyses. The aligned read-pairs were run through HTSeq (RRID:SCR_005514) for transcript quantification against the GENCODE reference gtf file vM4 for the mouse mm10 genome [42]. After all features had been quantified the data was run through three differential expression calculators in edgeR (RRID:SCR_012802) [43], DESeq (RRID:SCR_000154) [44], and baySeq (RRID:SCR_012795) [45]. The intersection of the three methods was taken and transformed into a list of the final differentially expressed features. Differentially expressed features were determined by cutoff adjusted p values of 0.05 across all three methods. For heatmap generation, the default parameters of the heatmap.2 function within the R package gplots v2. 16.0 (https://cran.r-project.org/web/packages/gplots/index. html) were used. Function and expression annotations for coding and non-coding genes were retrieved from the Protein ANalysis THrough Evolutionary Relationships (PANTHER) Classification System (RRID:SCR_004869) [46]. Expression differences of Gm15851 were confirmed by strand-specific qRT-PCR in the same RNA samples used for RNA-seq (n = 6 PNE offspring; n = 5 control offspring). Following quantitation of RNA concentrations by the RNA 6000 pico assay on an Agilent 2100 Bioanalyzer (Agilent Technologies), 500 pg of total RNA was reverse transcribed using a strand-specific primer (sequence: TCCTGGATCTGCAGCACAATCG) annealing to the Gm15851 transcript and the superscript IV first strand synthesis system (ThermoFisher Scientific). Reaction conditions followed the manufacturer's directions. After cDNA synthesis, the RNA template was digested with RNAse H. Real-time PCR reactions were run in triplicate using a custom-built Taqman gene expression assay for Gm15851 (forward primer sequence: CCGGCACGTT GCTGATC; reverse primer sequence: CTCCTTCAAC ATCTCCAACTTGCT; Taqman reporter sequence: CCACCTGTCTCACAACAA) (ThermoFisher Scientific) on a QuantStudio 7 Flex real time PCR system (Thermo-Fisher Scientific). Ct values were determined with Quant-Studio software (ThermoFisher Scientific). Dilutions of an expression plasmid containing the cDNA sequence of Gm15851 were assayed in parallel to generate a reference standard curve and quantitate the relative amounts of Gm15851 in each sample.

Data analysis
Study groups consisted of n = 7-9 male offspring unless otherwise indicated. Male offspring were selected from n = 6 nicotine-exposed and n = 7 non-exposed litters. Male offspring in each study group originated from n = 3-4 different dams. Statistical analyses were conducted with GraphPad Prism 5 (RRID:SCR_002798, GraphPad Software, La Jolla, CA, USA). Data were subjected to various normality tests (Kolmogorov-Smirnov, D' Agostino-Pearson Omnibus, and Shapiro-Wilk) prior to performing the following parametric or non-parametric statistical tests: two-tailed unpaired t test, Welch's test for uneven variances, and Mann-Whitney test for twogroup comparisons; Fisher's exact test for contingency analyses of postnatal survival; two-way ANOVA for determination of the effect of PNE and diet on metabolic, hormonal, and gene expression parameters, and the combined effect of PNE and leptin/MTII on body weight gain and food intake in refeeding experiments; two-way repeated measures (RM) ANOVA for determination of the effect of nicotine on maternal fluid intake, food intake, and body weight before and during gestation and the effect of PNE on glucose and insulin tolerance; Bonferroni post-tests following ANOVA for two-group comparisons. The α-value was set at 0.05 for each statistical test. Data were expressed as mean ± SEM.
Prenatal nicotine exposure does not cause obesity or type 2 diabetes in adult first-generation offspring Male offspring were subjected to a 16-week STD or HFD. PNE moderately decreased cumulative food intake of STD mice (PNE: 1268 ± 9.5 kcal, n = 7; control: 1363 ± 37.6 kcal, n = 8) and HFD mice (PNE: 1416 ± 13.48 kcal, n = 8; control: 1450 ± 27.6 kcal, n = 8) (Fig. 2a). Body weight measurements revealed no differences between PNE and control offspring fed the same diet (Fig. 2b). Moreover, PNE had no impact on body composition as determined by NMR measurements of relative fat, lean, and fluid mass ( Fig. 2c-e). Plasma lipid levels were profiled in week 16 of the diets and were the same in nicotine-exposed and non-exposed offspring (Table 1).
To more accurately characterize PNE-induced changes in energy balance, we subjected individual male offspring in week 10 of the STD and HFD to simultaneous determinations of food intake, locomotor activity, and energy expenditure by indirect calorimetry in metabolic cages. PNE moderately reduced 24-h caloric intake of STD mice (PNE: 12.11 ± 0.38 kcal/24 h, n = 7; control: 13.94 ± 0.52 kcal/24 h, n = 7; t 12 = 2.85 and p = 0.015 by t test) and HFD mice (PNE: 14.21 ± 0.34 kcal/24 h, n = 8; control: 17.16 ± 1.16 kcal/24 h, n = 7; F 1,25 = 12.9 and p = 0.0014 by two-way ANOVA; p < 0.01 by Bonferroni posttest) (Fig. 2f ). PNE had no impact on 24-h energy expenditure, 24-h oxygen consumption, 24-h carbon dioxide production, or the respiratory exchange rate of offspring on the STD or HFD (Fig. 2g-j). However, we noted a moderate enhancement in 24-h vertical activity of PNE offspring relative to control offspring under HFD conditions, possibly reflecting increased rearing ( Pre-gestation Gestation Fig. 1 Ad libitum nicotine intake led to high plasma cotinine levels, moderately reduced fluid intake, and did not alter body weight or food consumption of dams. a Twenty-four-hour fluid intake was determined in group-housed females (n = 4 cages of 3 females/group) before pregnancy and in individually housed females (n = 10/group) during gestation. b Nicotine intake before (n = 4) and during pregnancy (n = 10) was calculated based on 24-h fluid intake. c Plasma cotinine levels were determined by enzyme-linked immunosorbent assay (ELISA) after 4 weeks of nicotine ingestion. Plasma cotinine was detected in all nicotine-exposed females (n = 6) and was undetectable in non-exposed females (n = 6). d Twenty-fourhour food intake was determined in group-housed females (n = 4 cages of 3 females/group) before pregnancy. e Body weight was determined weekly before pregnancy (n = 12/group). Data were analyzed by two-way RM ANOVA followed by Bonferroni post-test (a) or by Mann-Whitney test (c) (**, p < 0.01). All data are expressed as mean ± SEM control animals relative to STD feeding ( Fig. 2k-l) as we have reported earlier [6,49]. PNE offspring displayed a transient decrease in fasting glycemia in week 8 of the STD (PNE: 80.3 ± 1.6 mg/dl, n = 7; control: 106.3 ± 5.3 mg/dl, n = 9; p < 0.001 by Bonferroni post-test) (Fig. 2m). Subsequent fasting glucose measurements in weeks 12 and 16 of the STD and in weeks 8, 12, and 16 of the HFD revealed no differences in fasting glycemia (Fig. 2m). A GTT conducted in week 8 of the diets revealed increased glucose tolerance in PNE offspring relative to control offspring under STD conditions (area under the curve (AUC): PNE, 19,348 ± 751 mg/dl, n = 7; controls, 23,069 ± 593 mg/dl, n = 8; p < 0.05 by Bonferroni post-test) but not under HFD conditions (Fig. 2o, q). An ITT conducted in week 10 of the diets showed no difference in insulin sensitivity between PNE and control groups (Fig. 2p, r). Fasting plasma insulin concentrations determined in week 16 of the diets were similar in PNE and control offspring (Fig. 2n). We conclude that PNE did not render offspring more susceptible to obesity or diabetes development but instead moderately decreased food intake and fasting glycemia.

Prenatal nicotine exposure enhances leptin-melanocortin signaling in adult first-generation offspring
We measured the anorectic and body weight lowering actions of leptin and the MC3/4-R agonist MTII in prenatally nicotine-exposed and non-exposed adult firstgeneration offspring. Mice were fasted overnight, injected i.p. leptin, MTII, or vehicle (PBS), and immediately offered food, followed by determination of body weight and cumulative food consumption at defined time points [38].
In summary, the refeeding experiments showed that PNE decreased body weight gain of offspring and rendered them moderately more sensitive to the acute body weight lowering effects of leptin and MTII.
Most human obesity and animal models of dietinduced obesity are associated with elevated plasma leptin levels [50][51][52], and the development of resistance to leptin's body weight lowering actions [5]. Therefore, we measured plasma leptin concentrations as an indicator for hypothalamic leptin resistance. Plasma leptin concentrations were the same in PNE and control offspring on the STD (PNE: 0.9 ± 0.09 ng/ml, n = 7; control: 0.98 ± 0.18 ng/ml, n = 9) but they were decreased in PNE offspring under HFD conditions (PNE, 9.3 ± 1.07 ng/ml, n = 9; control, 14.85 ± 2.21 ng/ml, n = 7; p < 0.01 by Bonferroni post-test) (Fig. 4a). These results further suggest that PNE rendered offspring moderately more sensitive to leptin.
Leptin stimulates P-STAT3 mediated transcription of POMC. To further determine the long-term effects of PNE on leptin signaling, we determined the levels of hypothalamic POMC mRNA at the end of the 16-week diet. PNE moderately increased POMC mRNA of adult male offspring under HFD but not under STD conditions (F 1,28 = 6.7, p = 0.015 by two-way ANOVA; p < 0.05 by Bonferroni post-test) (Fig. 4b) consistent with enhanced leptin-melanocortin signaling.

Prenatal nicotine upregulates Gm15851 in adult offspring hypothalamic POMC neurons
The intersection of the results of three differential expression calculators, edgeR, baySeq, and DESeq, revealed only one consistent expression difference between nicotine-exposed and non-exposed offspring, an 80-fold upregulation of Gm15851 in PNE offspring (FDR-adjusted p value: 0.006 by edgeR, 0.002 by baySeq and DESeq; average expression level: 4.5 CPM). Two-way clustering revealed overlapping gene expression signatures of prenatally nicotine-exposed and non-exposed hypothalamic POMC neurons (Fig. 7a). Quantitative reverse transcription polymerase chain reaction (qRT-PCR) confirmed upregulation of Gm15851 in POMC neurons of nicotine-exposed offspring (t 5 = 3.17, p = 0.025 by t test with Welch's correction) (Fig. 7b). Gm15851 is a lncRNA of 662 nucleotides encoded by three exons on the forward strand of chromosome 1 (Fig. 7c). Supporting evidence for the existence of this transcript comes from one expressed sequence tag in the adult mouse male epididymis cDNA RIKEN full-length enriched library (RIKEN clone 9230112 L10) [21]. Gm15851 lacks a significant open reading frame. It is an antisense RNA overlapping two protein-coding genes, opticin (Optc) and proline arginine-rich end leucine-rich repeat (Prelp), on the reverse strand. The third exon of Gm15851 overlaps the second coding exon of Prelp whereas the first exon of Gm15851 overlaps the first intron of Optc (Fig. 7c). Although Optc and Prelp are potential targets of cis-regulation by Gm15851, both genes were not co-expressed with Gm15851 in POMC neurons.

Discussion
Contrary to expectations based on previous studies of gestational and neonatal nicotine exposure and energy balance, we report here that gestational nicotine exposure does not render offspring more susceptible to dietinduced obesity or type 2 diabetes. Instead, we found that gestational nicotine exposure moderately decreased food intake, body weight gain, and glycemia, and mildly increased sensitivity to leptin and melanocortin receptor stimulation in first-generation adult offspring. Consistent with a moderate enhancement in leptin sensitivity, PNE offspring were more sensitive to leptin-induced STAT3 phosphorylation, and-under HFD conditions-had increased POMC mRNA and decreased plasma leptin concentrations. Consistent with previous reports [35,69], PNE offspring displayed increased voluntary locomotor activity under HFD conditions, which could reflect augmented exploratory behavior [69] or might result from enhanced central leptin-melanocortinergic signaling via POMC neurons as others and we have reported [6,70]. Interestingly, although food intake was moderately decreased and voluntary locomotion moderately increased under HFD feeding conditions, overall energy balance was unaffected in PNE offspring since neither body weights nor body fat content nor energy expenditure were measurably changed. It is possible that altered central regulation of body temperature in PNE offspring accounts for this outcome. Impaired central activation of brown fat thermogenesis could lead to compensatory hyperlocomotion to increase heat expenditure and preserve body temperature. Yet another possibility is that metabolic efficiency-the degree to which ingested food is being metabolized to perform work and generate heat-could be increased in PNE offspring. These alternative mechanisms require further investigation.     [71]. The authors re-expressed POMC selectively in 5-HT2CR containing POMC neurons of POMC-deficient obese and insulin-resistant mice. Reexpression of POMC in this neuronal subpopulation was sufficient to normalize food intake, physical activity, brown fat heat expenditure, body weight, adiposity, and insulin sensitivity in males. Meanwhile, re-expression of POMC in this neuronal subpopulation in females corrected only food intake and insulin sensitivity but did neither increase physical activity, energy expenditure nor block the development of obesity [71].
Previous human epidemiological and meta-analysis studies [11][12][13][14] reported a positive correlation between PNE and the risk for obesity and type 2 diabetes in children and young adults. This correlation was further supported by rat studies of gestational and lactational nicotine exposure [15][16][17][18]. The onset, length, and route of nicotine administration likely determine the cumulative dose and developmental stage at which hypothalamic POMC neuronal progenitor cells are exposed to nicotine. Maternal plasma cotinine levels in the present and in previous rodent studies reporting metabolic disturbances in the offspring were of similar magnitude [15][16][17], suggesting that dams were exposed to comparable doses of nicotine. However, previous studies reporting diabetes in the offspring administered nicotine subcutaneously to dams [15][16][17]72]. Nicotine that is subcutaneously administered to dams may reach the fetus at a higher concentration than maternally ingested nicotine because the former does not undergo first-pass hepatic conversion to cotinine contrary to the latter. In addition to differences in the route of administration, gestational nicotine exposure was continued throughout lactation in two studies reporting diabetes and obesity in the offspring [15,17]. Bruin et al. [72] showed that disturbances in glucose homeostasis in the offspring required nicotine exposure to occur both during gestation and lactation. This would be in line with the timing of development of murine hypothalamic arcuate neurons, which differentiate between embryonic days 10 and 16 [73] and form projections in the first three postnatal weeks to other hypothalamic regions involved in feeding regulation [74,75]. Our observations suggest dosedependent and time-dependent effects of nicotine on developing hypothalamic POMC neurons. Exposure to nicotine at a low dose during gestation might stimulate proliferation, differentiation, neurite outgrowth, and synaptogenesis of developing hypothalamic POMC neurons, thereby sensitizing leptin-melanocortinergic signaling in adult offspring, whereas exposure to nicotine at a high dose during gestation and lactation might impair those processes, leading to the reported obesity and diabetes phenotypes in adult offspring.
RNA-seq of hypothalamic POMC neurons isolated from adult PNE and control offspring followed by differential expression analyses using three expression calculators detected only one consistent change, a nicotineinduced upregulation of Gm15851, an antisense RNA to the Prelp and Optc genes. Although Gm15851 was not co-expressed with Prelp or Optc in terminally differentiated POMC neurons, we cannot exclude that cis-regulatory interactions between Gm15851 and Prelp or Optc exist in POMC neuronal progenitor cells. LncRNAs may serve as scaffolds for gene-specific recruitment of transcription factors, transcriptional complexes, and/or chromatin-modifying complexes during specific developmental time windows. These additional regulators may  not be expressed in terminally differentiated POMC neurons thus preventing regulatory interactions of Gm15851 with Optc and/or Prelp to be seen. It is possible that, in addition to regulation of transcription or chromatin state, Gm15851 might regulate post-transcriptional processes such as splicing, editing, translation, or localization of the Prelp or Optc RNA. Interestingly, Prelp and Optc both encode extracellular matrix proteins of the family of leucinerich repeat (LRR) domain-containing proteins. LRR domain containing proteins have been documented to organize axon guidance, target selection, and synapse formation [76]. Hence, it might be possible that Gm15851 could modulate the establishment of synaptic connections of POMC neuronal progenitor cells by regulating expression of Prelp and/or Optc.
Whole transcriptome sequencing of hypothalamic POMC neurons confirmed the expected expression of the Opioid proopiomelanocortin release pathways, known G-protein coupled receptor pathways regulating feeding and body weight, and the insulin receptor and leptin receptor signaling pathways. Furthermore, we confirmed expression of several nicotinic acetylcholine receptor subunits and acetylcholine metabolizing enzymes in POMC neurons [10]. We also detected expression of specific transport proteins (vGlut2 and vGat) required for the uptake of glutamate and GABA/glycine into synaptic vesicles, supporting the notion that POMC neurons release glutamate and GABA [77,78].
Several neuroendocrine signaling pathways including those mediated by the GnRH, TRH, and CRF receptors were also expressed in POMC neurons. While POMC neurons regulate reproduction, energy expenditure and stress responses through activation of hypothalamic GnRH-producing, TRH-producing, and CRF-producing neurons, respectively, the reciprocal regulation of POMC neurons by these hormonal signaling pathways remains to be explored.
POMC neurons further expressed the netrin, slit, and semaphorin axon guidance pathways. Netrin, slit, and semaphorin are secreted or membrane-bound chemotropic proteins that attract or repulse growing axons and migrating neural progenitor cells by binding, respectively, to the unc, neuropilins, and plexin transmembrane receptors [79]. These proteins guide migration and neuronal outgrowth of hypothalamic oxytocin, antidiuretic hormone, and GnRH-producing neurons [80][81][82]. Furthermore, the slit/robo and semaphorin/plexin/neuropilin pathways are expressed during development of the hypothalamic PVN [83], an important projection site of POMC neurons. These cell migration and axon guidance pathways could therefore regulate migration of POMC neurons and formation of connections to relevant target sites.
In addition, we found expression of the wnt, integrin, cadherin, PDGF, EGF, and Notch signaling pathways in POMC neurons, all of which have been linked to the development of hypothalamic neurons. The wnt and integrin signaling pathways regulate differentiation of hypothalamic neuronal progenitor cells [84] and migration of hypothalamic GnRH producing neurons [85], respectively. Cadherin signaling regulates neuronal connectivity and wiring of hypothalamic POMC neurons [86,87]. The Notch signaling pathway regulates differentiation of hypothalamic arcuate neural progenitor cells into POMC and AgRP neurons [88].
Furthermore, hypothalamic POMC neurons expressed an Alzheimer amyloid secretase and TGF-β signaling pathway, which included expression of several TGF-β encoding genes (TGF-β1, TGF-β2, TGF-β3). Both pathways have been proposed to be disease-relevant. Excessive production of TGF-β by POMC neurons was found to promote hypothalamic inflammation and type 2 diabetes in obesity and during aging [89]. Expression of an Alzheimer amyloid secretase pathway suggests participation of POMC neurons in the pathogenesis of diabetic comorbidities in Alzheimer's disease [90].
We attempted to gain insight into probable biological functions of expressed lncRNAs based on predicted cisregulatory interactions with protein-coding genes. This approach revealed 82 co-expressed lncRNA/coding gene pairs, 19 of which involved coding genes regulating neural development and/or function. Most co-expressed lncRNAs and coding genes overlapped each other. In around 35 % of co-expressed lncRNA/coding gene pairs, the lncRNA also overlapped the coding gene promoter. These observations suggest that some of these lncRNAs could regulate chromatin state and/or promoter activity of the coding gene. Furthermore, we found that only one of the co-expressed lncRNAs was transcribed from an annotated conserved mouse enhancer. Such eRNAs are (See figure on previous page.) Fig. 7 PNE upregulated an antisense RNA, Gm15851, in hypothalamic POMC neurons of adult offspring. a Two-way clustering of RNA-seq expression data of POMC neurons reveal overlapping gene expression signatures in PNE offspring (N1-N6) and control offspring (C1-C5). The color code represents the gene expression value in log 2 CPM. Only genes with average expression levels > 1 CPM across all offspring were considered. b Strand-specific qRT-PCR confirms overexpression of Gm15851 in POMC neurons of PNE offspring (*, p < 0.05, n = 5-6, two-tailed t test with Welch correction). c Schematic of genomic localization of Gm15851 within a 28.29 kb segment spanning the chromosomal coordinates 133.9 Mb and 133.92 Mb of chromosome 1. Gm15851 is a spliced antisense transcript to Prelp and Optc on the forward strand of mouse chromosome 1. Boxes denote exons. Arrowheads denote gene orientation known to mediate gene-specific transcription by facilitating enhancer-promoter interactions, recruiting transcription factors, and blocking enhancer-binding of gene repressive factors [66,67]. Several co-expressed coding genes had annotated functions in nervous system development such as Irx3, Msi2 and Ppp1r9a. Irx3 is a homeobox transcription factor that is expressed in the prospective neural plate in a subset of neural precursor cells and possibly regulates specification of neural progenitor cells [91]. Msi1 is a neural RNA-binding protein that is highly enriched in neural precursor cells and drives proliferation of neurons and glial cells in the CNS during embryonic development [92], postnatally and in adults [93]. Ppp1r9a, also referred to as neurabin, is selectively expressed in neural tissues, where it induces F-actin crosslinking activity in the synapse and lamellipodia of the growth cone to regulate neurite formation [94].