Leptin regulates exon-specific transcription of the Bdnf gene via epigenetic modifications mediated by an AKT/p300 HAT cascade

Leptin is an adipocyte-derived hormone with pleiotropic functions affecting appetite and mood. While leptin’s role in the regulation of appetite has been extensively studied in hypothalamic neurons, its function in the hippocampus, where it regulates mood-related behaviors, is poorly understood. Here, we show that the leptin receptor (LepRb) colocalizes with brain-derived neurotrophic factor (BDNF), a key player in the pathophysiology of major depression and the action of antidepressants, in the dentate gyrus of the hippocampus. Leptin treatment increases, whereas deficiency of leptin or leptin receptors decreases, total Bdnf mRNA levels, with distinct expression profiles of specific exons, in the hippocampus. Epigenetic analyses reveal that histone modifications, but not DNA methylation, underlie exon-specific transcription of the Bdnf gene induced by leptin. This is mediated by stimulation of AKT signaling, which in turn activates histone acetyltransferase p300 (p300 HAT), leading to changes in histone H3 acetylation and methylation at specific Bdnf promoters. Furthermore, deletion of Bdnf in the dentate gyrus, or specifically in LepRb-expressing neurons, abolishes the antidepressant-like effects of leptin. These findings indicate that leptin, acting via an AKT-p300 HAT epigenetic cascade, induces exon-specific Bdnf expression, which in turn is indispensable for leptin-induced antidepressant-like effects.


Introduction
Leptin is produced and secreted from adipocytes [1], circulates in the blood [2], and is transported across the blood-brain barrier [3]. Leptin's target neurons are distributed widely throughout the brain [4][5][6], through which leptin exerts pleiotropic effects by activating the signaling form of the leptin receptor (LepRb) [7][8][9]. While extensive studies have focused on leptin's actions in the hypothalamus in the regulation of feeding and energy metabolism [10,11], LepRb is also highly expressed in extra-hypothalamic brain regions including the hippocampus, which is implicated in mood regulation [4,[12][13][14]. We and others have shown that circulating leptin levels are reduced in chronic stress animal models of depression [15][16][17], whereas systemic and intracerebroventricular injections of leptin produce antidepressant-like behavioral effects [15,[18][19][20]. In humans, leptin levels were found to correlate negatively with the severity of depression symptoms [21], Multiple lines of evidence suggest that leptin targets hippocampal neurons to regulate depression-related behaviors. First, direct infusion of leptin into the dentate gyrus of the hippocampus induces an antidepressant-like effect, similar to the effect observed after systemic injection [15]. Second, deletion of LepRb in the dentate gyrus results in depression-like behaviors and attenuates leptin's antidepressant-like effects [13,22]. Third, blockade of leptin signaling in the dentate gyrus attenuates the antidepressant-like effects of leptin [14]. These findings support an important role of the hippocampus in mediating leptin's actions on mood-related behavior. However, the precise molecular mechanisms remain unknown.
Brain-derived neurotrophic factor (BDNF) is a neurotrophin that has been implicated in the pathophysiology of depression and the mechanisms of action of antidepressant drugs [23]. Hippocampal mRNA or protein levels of BDNF are increased by treatment with different classes of antidepressants [24,25] in rodents and in depressed patients [26]. Antidepressant drugs activate the BDNF receptor, tropomycin receptor kinase B (TrkB), in the hippocampus, suggesting enhanced BDNF release [27]. Furthermore, infusion of BDNF directly into the hippocampus is sufficient to induce antidepressant-like effects [28,29]. Importantly, the antidepressant-like response to classical antidepressants is attenuated in transgenic mice with disrupted BDNF signaling [27]. In addition, the rapid onset antidepressant effects of ketamine require expression of BDNF [30,31]. Together, these data indicate that increased BDNF expression and signaling are both necessary and sufficient to produce the behavioral effects of antidepressants.
In this study, we investigated (1) whether, and to what extent, LepRb is colocalized with BDNF in the hippocampus; (2) whether and how leptin regulates exon-specific Bdnf expression; (3) whether leptin and deficiencies of leptin or its receptor have opposite effects; and (4) whether Bdnf expression in the hippocampus is required for leptin's antidepressant-like behavioral effects. To address these questions, we employed a genetic approach to investigate colocalization of LepRb and BDNF expression in the brain. Furthermore, we examined exon-specific Bdnf expression in response to leptin treatment and impaired leptin signaling and further explored the epigenetic mechanisms underlying leptin-induced regulation of Bdnf gene expression. Finally, we determined the effects of loss of BDNF in the hippocampus on behavioral responses to leptin. 005628) were purchased from Jackson Laboratory (Bar Harbor, ME, USA) and maintained as breeding colonies. The ob/ob and db/db mice and their littermates were obtained by intercrossing ob/+ or db/+ mice (ob/+, Stock No. 000632; db/+, Stock No. 000642; Jackson Laboratory). LepRb-ires-Cre mice were generated with an IRES-NLS-Cre cassette "knocked in" to the region immediately 3′ to the LepRb stop codon [48] to drive Cre expression in all Lepr-expressing cells. Ai14 mice have a loxP-flanked STOP cassette preventing transcription of a CAG promoter-driven tdTomato protein in all cells. Ai14 mice express robust tdTomato fluorescence following Cre-mediated recombination [49]. Bdnf flox/flox mice possess loxP sites on either side of the Bdnf coding region [50]. Bdnf klox/klox mice possess loxP sites flanking the Bdnf encoding exon 9, with a polyadenylation sequence upstream of the 3′ loxP site, and lacZ downstream of the 3′ loxP site [51]. All mice were maintained in the C57BL/6J background. Mice were housed in a group of 5 and maintained on a 12 h light/12 h dark cycle with ad libitum access to food and water. The animal protocols used in this study were approved by the Institutional Animal Care and Use Committees of the University of Texas Health Science Center at San Antonio, Binzhou Medical University Hospital, and Augusta University.

Immunohistochemistry
Lepr-ires-cre tdTomato and Bdnf klox/+;Lepr-ires-cre mice were transcardially perfused with 4% paraformaldehyde. The brains were removed, postfixed overnight and then cryoprotected in 30% sucrose and cut into 20-μm coronal sections. From serial coronal sections of the entire dorsalventral axis of dentate gyrus (from −1.34 to −3.52 mm posterior to bregma), every sixth section was selected from each animal and processed for immunohistochemical staining. The sections were rinsed three times in phosphatebuffered saline (PBS), and incubated in blocking buffer (1% bovine serum albumin, 3% goat serum, 0.3% Triton X-100 in PBS) for 1 h. The sections were then incubated with mouse anti-NeuN antibody (#ab104224, 1:500; Abcam, Cambridge, UK) and rabbit anti-β-galactosidase antibody (#ab4761, 1:1000, Abcam, Cambridge, UK) overnight at 4°C. After washing in PBS, sections were incubated for 4 h with fluorescent secondary antibodies: Alexa Fluor® 488 donkey antimouse IgG (#A-21202, 1:400, Invitrogen, Carlsbad, CA, USA) and Alexa Fluor® 555 donkey antirabbit IgG (#A-31572, 1:400, Invitrogen, Carlsbad, CA, USA). Finally, the sections were washed in PBS, mounted onto poly-lysine-coated glass slides, cover-slipped using fluorescence mounting medium and visualized via an Olympus FV1000 confocal microscope (Olympus, Shinjuku, Tokyo, Japan). Cell numbers were quantified on every sixth section throughout the dorsal-ventral axis of hippocampus using unbiased stereology. LepRb-tdTomato-, βgalactosidase-, and NeuN-positive cells within both sides of the hippocampus were counted. Fluorescent cells that intersected the exclusion boundaries of the unbiased sampling frame were excluded from counting. Cells that met the counting criteria were counted bilaterally throughout the dentate gyrus. The colocalization of LepRb-tdTomato or β-galactosidase with NeuN was confirmed with z-stack of the respective cell soma using line-based sequential scan (4-μm interval). The percentage of NeuNpositive neurons that were also double-labeled for LepRb-tdTomato and β-galactosidase within the granule cell layer was calculated.

Intra-dentate gyrus microinjection
For the deletion of the Bdnf gene selectively in the dentate gyrus, anesthetized adult Bdnf flox/flox mice underwent bilateral stereotaxic injections of AAV-Cre-GFP or AAV-GFP (0.5 μl /side; Vector Biolabs, Malvern, PA) into the dentate gyrus (coordinates: AP = −2.1 mm, ML = ±1.5 mm, DV = −2.3 mm from Bregma), at a rate of 0.1 μL/min with a 33gauge stainless steel injector connected to a UMP3 micro syringe pump (World Precision Instruments, Sarasota, FL). Additional 5 min were allowed for diffusion and prevention of backflow. Behavioral experiments were conducted 21 days after AAV injection. For intra-dentate gyrus microinjection of inhibitors of AKT and p300 HAT, the mice were anesthetized with 4% chloral hydrate (400 mg/ kg, i.p.). AKTi and C646 (0.5 μl/side) were injected bilaterally into the dentate gyrus of adult C57BL/6J mice at a rate of 0.5 μl/min with a 33-gauge stainless steel injector. Additional 5 min were allowed for diffusion and prevention of backflow, and 30 min later, mice were injected with leptin (5 mg/kg, i.p.).

Estrous cycle
The stages of the estrous cycle were monitored by analysis of cell types in vaginal lavages. Vaginal lavages were collected daily between 7:00 and 8:00 a.m. as described previously [53]. Briefly, vaginal fluid was placed on slides, and the slides were stained with crystal violet. The types of cells in vaginal smears were examined under a light microscope. Diestrus was defined by the presence of small leukocytes, and proestrus was defined by the presence of clumps of large, round, nucleated epithelial cells.

Forced swim test
Mice were placed into a clear Plexiglas cylinder (25 cm in height and 10 cm in diameter) filled with water (24°C) to a depth of 15 cm. A 6-min swim session was videotaped by a camera mounted above the cylinder. The duration of immobility was measured for the last 4 min. Immobility was defined as the absence of all movements except those required for respiration.

Tail suspension test
The apparatus consisted of a box (30 × 30 × 30 cm) with an open front and a bar placed horizontally 1 cm from the top with an attached vertical bar hanging down in the center. Mice were individually suspended by the tail on the vertical bar with adhesive tape affixed 2 cm from the tip of the tail. A camera positioned in front of the box was used to record the animals' behavior for a 6-min test session. Immobility in this test was defined as the absence of any limb or body movements, except those caused by respiration.
The behaviors of each mouse were scored by experimenters who were blinded to the genotypes or treatment conditions.

Statistical analysis
All statistical analyses were performed using the statistical software GraphPad Prism 8. Shapiro-Wilk test and F-test were used to test normality and equal variance assumptions, respectively. For normally distributed data, two-tailed t-tests were used to assess differences between two experimental groups with equal variance. For normally distributed data with unequal variances, two-tailed t-tests with Welch's correction were used. One-way analyses of variance (ANOVAs) followed by Sidak post hoc tests were used for analysis of three or more groups. For nonnormally distributed data, Mann-Whitney U tests were performed to compare two groups. For analysis of three or more groups with nonnormally distributed data, the Kruskal-Wallis test followed by Dunn's multiple comparisons test was used. Two-way or three-way ANOVAs followed by Bonferroni tests were used where appropriate. P < 0.05 was considered significant.

Results
Colocalization of LepRb and BDNF in the dentate gyrus of the hippocampus While alternative splicing produces multiple isoforms of the LepR, the long isoform, LepRb, has an intracellular domain that is essential for signal transduction required to elicit the physiological effects of leptin [54,55]. We examined the distribution of LepRb-expressing neurons in the brain using LepRb-ires-Cre × Ai14-tdTomato (LepRb-tdTomato) reporter mice, in which tdTomato marks LepRbexpressing cells [4,14]. As shown previously [4,14,56], leptin target neurons were found in the hypothalamus, prefrontal cortex, hippocampus, and ventral tegmental area (Fig. 1a). To examine if these neurons express BDNF, we crossed LepR-ires-Cre mice with Bdnf klox/+ mice, in which BDNF neurons express β-galactosidase once the floxed Bdnf allele is deleted by Cre-mediated recombination [51,57]. In Bdnf klox/+;LepR-ires-Cre mice, very few βgalactosidase-expressing neurons were detected in the hypothalamus, prefrontal cortex, or ventral tegmental area, indicating lack of co-expression between LepRb and BDNF in these brain regions. However, dense populations of granule neurons in the dentate gyrus of the hippocampus expressed β-galactosidase (Fig. 1a), suggesting the coexpression of LepRb and BDNF.
The dentate gyrus is composed of the suprapyramidal and infrapyramidal blades, and the dorsal and ventral regions are functionally distinct [58]. Therefore, we analyzed the distribution of neurons expressing LepRb alone and coexpressing LepRb/BDNF in these subcompartments of the dentate gyrus. In LepRb-tdTomato reporter mice, LepRb neurons were labeled with tdTomato. We quantified the numbers of LepRb neurons and total NeuN-positive cells in the suprapyramidal versus infrapyramidal blade and in the dorsal versus ventral region. Data analysis revealed that a higher proportion of neurons in the ventral region expressed LepRb (blade: F (1,12) = 1.8970, P = 0.1936; region: F (1,12) = 28.4700, P < 0.001; total number of LepRb neurons counted, 2380; total number of NeuN-positive neurons counted, 27,155 from 4 mice) (Fig. 1b). Furthermore, we quantified the numbers of β-galactosidase-positive cells, indicative of colocalization of LepRb and BDNF, and total NeuN-positive cells in Bdnf klox/+;LepR-ires-Cre mice. Cell counting results indicated a region-specific distribution with a higher percentage of neurons in the dorsal region expressing both LepRb and BDNF (blade: F (1,12) = 0.1587, P = 0.6973; region: F (1,12) = 144.5000, P < 0.001; total number of LepRb/BDNF neurons counted, 2029; total number of NeuN-positive neurons counted, 37,318 from 4 mice) (Fig. 1c). By comparing the numbers and the distribution of LepRb-expressing neurons with LepRb/BDNFcolocalized neurons in the dentate gyrus, we determined that a high proportion of LepRb-expressing neurons (up to 79%) contain detectable BDNF especially in the dorsal region, which provide an anatomical basis to study functional interactions between LepRb signaling and BDNF.
Regulation of exon-specific Bdnf mRNA expression by leptin treatment, leptin deficiency and leptin receptor deficiency Leptin-induced changes in total (exon IX) Bdnf mRNA levels could be related to the changes in noncoding exons I-VIII spliced to exon IX (Fig. 3a). To determine which noncoding exons may directly contribute to the modulation of total Bdnf expression by leptin, we analyzed exons I, II, III, IV, and VI that are highly expressed in the hippocampus. Exons V, VII, and VIII were not included due to their very low expression levels (data not shown). While mRNA levels of Bdnf exons II and III remained unchanged, exons I, IV, and VI transcripts were found to increase 2 h after leptin treatment (5 mg/kg) (treatment: F (1,155) = 52.8300, P < 0.001; exon: F (4,155) = 2.8020, P = 0.0278; treatment × exon interaction: F (4,155) = 2.8020, P = 0.0278; exons I: P < 0.001; exons II: P = 0.0511; exons III: P > 0.9999; exons IV: P < 0.001; exons VI: P < 0.001) (Fig. 3b). These results suggest that Bdnf transcripts containing exons I, IV, and VI are responsible for leptin-induced upregulation of total Bdnf gene expression in the hippocampus.
All experiments described above were conducted in male mice. We also examined the effects of leptin and deficiency in leptin or LepRb on Bdnf gene expression in female mice. As we have previously demonstrated that leptin induces AKT phosphorylation and antidepressant-like effects in the proestrus phase but not in the diestrus phase [14], we therefore tested the effects of leptin on Bdnf gene expression during these two stages of the estrous cycle.  (Fig. 4a). Furthermore, mRNA levels for total Bdnf and exons IV and VI varied across the estrous cycle with higher levels in the proestrus phase (total Bdnf, P = 0.0301; exons I: P = 0.8960; exons II: P = 0.9686; exons III: P = 0.2856; exon IV, P = 0.0467; exon VI, P = 0.0454) (Fig. 4a). These results suggest that basal expression levels of total Bdnf and Fig. 3 Regulation of exon-specific Bdnf mRNA expression by leptin signaling. a The structure of the mouse Bdnf gene. b Wild-type mice. Exon-specific Bdnf mRNA expression levels. Saline, n = 16; leptin (5 mg/kg, i.p.), n = 17. c ob/ob mice. Upper panel: left, schematic diagram of breeding strategy; middle, total Bdnf mRNA; right, Bdnf exon-specific mRNA. WT, n = 6, ob/ob, n = 6. Lower panel: left, schematic diagram of breeding strategy and leptin treatment; middle, total Bdnf mRNA; right, Bdnf exon-specific mRNA. n = 5 per group. d db/db mice. Left, schematic diagram of breeding strategy; middle, total Bdnf mRNA; right, Bdnf exon-specific mRNA. WT, n = 5; db/db, n = 5. e Lepr conditional knockout mice. Left, schematic diagram of breeding strategy; middle, total Bdnf mRNA; right, Bdnf exon-specific mRNA. Lepr flox/flox , n = 7; Lepr flox/flox;Emx1-Cre , n = 7. *P < 0.05; **P < 0.01; ***P < 0.001 compared with wild-type littermates or salineinjected controls. specific exons and their responses to leptin treatment are estrous cycle-dependent, which could be partially due to a dynamic change in endogenous leptin and LepRb across the estrous cycle [63].
Both ob/ob and db/db female mice have impaired ovarian development [64]. They are infertile and exhibit no evidence of estrous cycles [65]. As seen in male mice, total Bdnf mRNA expression in the hippocampus was decreased Fig. 4 Regulation of exonspecific Bdnf mRNA expression by leptin in female mice. a Wild-type female mice. Upper panel, schematic diagram depicting the identification of the estrous cycle and leptin treatment. Middle and lower panels, total Bdnf and exonspecific mRNA expression levels in proestrus versus diestrus in response to leptin injection (5 mg/kg, i.p.). n = 11 per group. b ob/ob female mice. Left, schematic diagram of breeding strategy; middle, total Bdnf mRNA; right, Bdnf exonspecific mRNA. n = 6 per group. c db/db female mice. Left, schematic diagram of breeding strategy; middle, total Bdnf mRNA; right, Bdnf exonspecific mRNA. n = 6 per group. *P < 0.05; **P < 0.01; ***P < 0.001 compared with WT or saline-injected controls.
Alternatively, DNA methylation changes in the Bdnf gene region might mediate leptin-induced regulation of Bdnf gene expression in the hippocampus [66,67]. Thus, we chose the CpG island of the Bdnf gene upstream of exon I (CpG1) and CpG islands within the promoters II, IV, and VI and the exonic regions of the Bdnf gene after the transcriptional start site of exons II, IV, and VI (CpG2-6) as targets for methylation analysis. Using bisulfite pyrosequencing, the methylation levels of CpG sites at Bdnf promoters/exon regions were quantified. No significant changes in methylated DNA levels associated with exons I, IV, and IV were observed after leptin treatment compared with vehicle controls (Fig. 5b), except for one out of 13 CpG sites within the promoter II CpG island, i.e., CpG island 2 (CpG1: treatment: Leptin-induced regulation of Bdnf mRNA expression is mediated by stimulation of an AKT/p300 HAT/H3 modification cascade The binding of leptin to LepRb activates three major signaling pathways, i.e., JAK2/STAT3, PI3K/AKT, and ERK pathways [68] (Fig. 6a). To explore which signaling pathway(s) is responsible for mediating leptin's effects on Bdnf gene expression, we assessed activation of STAT3, AKT, and ERK in the hippocampus by leptin. Leptin treatment at a dose used for gene expression (i.e., 5 mg/kg, i.p.) failed to induce a change in phosphorylation levels of STAT3 or ERK (t (7) = 0.5803, P = 0.5782), but significantly increased AKT phosphorylation (t (7) = 3.8020, P = 0.0067), suggesting that leptin action in the hippocampus may be mediated by AKT signaling (Fig. 6a). Indeed, we have previously shown that AKT activation in the hippocampus mediates leptin's antidepressant-like behavioral effects [14]. To examine whether AKT signaling is required for leptin-induced Bdnf gene expression, an AKT selective inhibitor, AKTi, was infused into the dentate gyrus to block AKT signaling prior to leptin injection. We found that inhibition of AKT activation eliminated the effects on total Bdnf mRNA expression induced by leptin (Kruskal-Wallis test, P < 0.001) (Fig. 6b). In addition, pretreatment with AKTi blocked the effects of leptin on acetylation of H3 at promoters I, IV, and VI and methylation of H3K4 at promoters I, IV, and VI and H3K9 at promoters II, IV, and VI ( AKT can phosphorylate p300 HAT at Ser-1834, thereby increasing its promoter recruitment and histone acetylation [69]. We then examined whether AKT activation is involved in p300 HAT binding activity at Bdnf promoters. Leptin was found to increase p300 HAT binding to Bdnf promoters I, IV, and VI, and this effect was abolished by pretreatment with AKTi in the hippocampus (treatment: F (2,138) = 14.2600, P < 0.001; promoter: F (5,138) = 4.0820, P = 0.0017; treatment × promoter interaction: F (10,138) = 2.5300, P = 0.0079) (Fig. 6d). Next, we tested whether p300 HAT activity is necessary for leptin-induced changes in Bdnf gene expression. C646, a competitive p300 HAT inhibitor [70], was injected into the dentate gyrus prior to leptin treatment. Pretreatment with C646 was sufficient to block leptin-induced increase in total Bdnf mRNA expression (F (2,15) = 6.9240, P = 0.0074) (Fig. 6e) and subsequent acetylation of histone H3 at promoters I, IV, and VI was abolished (treatment: F (2,90) = 9.8950, P < 0.001; promoter: F (5,90) = 2.2800, P = 0.0532; treatment × promoter interaction: F (10,90) = 2.1570, P = 0.0277) (Fig. 6f), suggesting that p300 HAT mediates histone acetylation at Bdnf promoters. p300 HAT-mediated histone acetylation events have been shown to modulate H3 methylation [71]. We therefore examined whether methylation of H3K4 and H3K9 at Bdnf promoters is affected by p300 HAT activity. Pretreatment with C646 was found to abolish leptin-induced increase in H3K4 methylation and decrease in H3K9 methylation (H3K4, treatment: F (2,90) = 12.1900, P < 0.001; promoter: F (5,90) = 0.9871, P = 0.4302; treatment × promoter interaction: F (10,90) = 1.1900, P = 0.3083; H3K9, treatment: F (2,90) = 10.7300, P < 0.001; promoter: F (5,90) = 0.8255, P = 0.5348; treatment × promoter interaction: F (10,90) = 1.1440, P = 0.3393) (Fig. 6g), suggesting p300 HAT activity is also indirectly involved in histone H3 lysine methylation. Together, these results suggest that leptin induces Bdnf gene expression through an AKT/p300 HAT/ H3 modification cascade, leading to chromatin remodeling and transcriptional activation.

Discussion
The hippocampus is a highly vulnerable, dynamic, and responsive brain structure in that it demonstrates rapid plasticity at the molecular, cellular, circuit, and functional levels [72,73]. BDNF, which is expressed at the highest levels in the hippocampus, plays a vital role in hippocampal plasticity and function [74,75]. In this study, our finding that dentate gyrus neurons expressing the functional leptin receptor isoform LepRb also contained BDNF suggests that BDNF-expressing neurons are targets of leptin action. Further studies showed that leptin upregulated total Bdnf gene expression and protein synthesis, causing activation of TrkB in the hippocampus. Bdnf transcription is uniquely controlled by nine individual promoters, which drive expression of multiple transcripts encoding the same protein [34][35][36]. Such complex gene structure suggests that Bdnf gene expression is finely tuned. We found that leptin treatment increased transcriptional activity of Bdnf promoters I, IV, and VI in the hippocampus. However, leptin deficiency or loss of its receptor decreased Bdnf gene expression, albeit via suppression of transcriptional activity at different promoters. Furthermore, leptin activated AKT signaling, which in turn regulated p300 HAT recruitment and subsequently promoted the transcriptional activity of Bdnf promoters via an epigenetic mechanism. Importantly, the loss of Bdnf gene expression in the dentate gyrus eliminated the antidepressant-like behavioral effects of leptin, suggesting that upregulation of BDNF underlies leptin's action on mood-related behavior.
BDNF in the hippocampus has been implicated in the pathophysiology of depression and the therapeutic mechanisms of antidepressant treatments [23,47,76]. In animal studies, BDNF levels in the hippocampus decrease in response to a variety of stressors and increase following various antidepressant treatments [76]. Postmortem studies demonstrate that levels of BDNF in the hippocampus (both mRNA and protein) are reduced in depressed patients [77,78], whereas BDNF levels are elevated in patients receiving antidepressants [26,[77][78][79][80]. Moreover, overexpressing BDNF in excitatory neurons of the forebrain (including hippocampus) caused an antidepressant-like behavioral phenotype [81], while a single infusion of BDNF directly into the hippocampus was sufficient to induce an antidepressant-like effect [29]. Importantly, loss of forebrain BDNF or reduced BDNF signaling (overexpression of truncated TrkB) attenuates the antidepressantlike response to antidepressants [27,[82][83][84]. The rapid acting antidepressant ketamine is also dependent on increased BDNF signaling [30,85]. Furthermore, mice with reduced BDNF signaling caused by overexpressing truncated TrkB are resistant to the effects of antidepressants [27]. We have shown that leptin has antidepressant-like effects mediated by LepRb in the hippocampus [13-15, 22, 53]. Our findings that leptin increased Bdnf gene expression, while the loss of BDNF abolished leptin's antidepressant-like effects, support the idea that hippocampal BDNF is required for leptin action on mood-related behavior. Furthermore, this effect is likely to be mediated by direct interactions between LepRb and BDNF, as deletion of Bdnf from LepRb neurons in the dentate gyrus eliminated the antidepressant-like effects of leptin.
Bdnf gene expression is regulated by diverse stimuli and physiological and pathological conditions through differential recruitment of individual exon-specific Bdnf promoters. Bdnf transcripts containing exons I, II, III, IV, and VI are present in the hippocampus, and they make differential contributions to total BDNF [86]. Many antidepressants have been shown to result in distinct exonspecific Bdnf transcripts [44,[87][88][89]. For example, tranylcypromine and desipramine increase exon I, II, and IV (original III) mRNAs, fluoxetine treatment enhances activity of promoter II [88], whereas reboxetine increases activation of promoter IV [87]. Leptin, similar to classic antidepressants, increased total Bdnf gene expression while inducing sexually dimorphic profiles of Bdnf promoter activation, with male mice exhibiting increased expression of exons I, IV, and VI and female mice exhibiting increased expression of exons I, II, III, and VI. In contrast, leptin deficiency or leptin receptor deficiency decreased total Bdnf gene expression. However, three distinct lines of male mice deficient in leptin or leptin signaling, i.e., ob/ob, db/db and Lepr flox/flox;Emx1-Cre mice, exhibited different patterns of downregulation of exon-specific Bdnf gene expression, with ob/ob associated with decreased expression of exons II, III, IV, and VI, db/db associated with decreased expression of exons I, III, and IV and Lepr flox/flox;Emx1-Cre associated with decreased expression of exons I, II, IV, and VI. In ob/ob mice, replenishing leptin restored the levels of transcriptional activity at promoters IV and VI but failed to reverse the downregulation of exons II and III. These results suggest that leptin deficiency is responsible for the downregulation of exons IV and VI expression, but not exons II and III. We considered the possibility that the downregulated expression of exons II and III in ob/ob mice was caused by obesity, which cannot be normalized by acute leptin treatment. This seems to be the case for exon III but not for exon II, as expression of this exon was also decreased in Lepr flox/flox;Emx1-Cre mice with normal body weight. Bdnf regulation appears to be even more complicated in female mice. First, leptin increases the transcription of the Bdnf gene in proestrus but not diestrus, with specific changes in expression of exons I, II, III, and VI. Second, while leptin deficiency caused a reduction of expression of exons I, III, and VI in female ob/ob mice, female db/db mice exhibited decreased expression levels of exons I, III, IV, and VI. Bdnf promoters show differential transcription efficiency with promoters IV and VI contributing more substantially to total BDNF in the hippocampus [86]. How the different patterns of Bdnf exon-specific upregulation or downregulation induced by leptin, leptin deficiency and leptin receptor deficiency contribute to total BDNF are dependent on the relative abundance of individual Bdnf transcripts and the fold changes of their relative abundance following different manipulations. Our results demonstrate that transcription of the Bdnf gene is tightly regulated by leptin, displaying distinct exon-specific, sex-dependent expression profiles in response to elevated leptin levels, defective leptin signaling, and leptin resistance. These findings suggest that Bdnf expression in the hippocampus is sensitive to fluctuations in leptin availability and leptin signaling, serving as a finely tuned neural component of the adipose-brain axis.
One intriguing and important finding of this study is that leptin activates the PI3K/AKT pathway to recruit the p300 HAT, leading to histone modifications and subsequently transcriptional activation of the Bdnf gene (Fig. 8). This finding constitutes strong evidence for an intersection between cell signaling and epigenetic regulation. First, among three major signaling pathways downstream of LepRb, the AKT pathway was selectively activated in the hippocampus by leptin. Moreover, we have previously shown that AKT signaling in the hippocampus is required for leptin's antidepressant-like behavioral effects [14]. Second, AKT can phosphorylate p300 HAT at Ser-1834 [69], which could induce its recruitment to the Bdnf promoter, leading to the acetylation of histones and activation of the transcriptional machinery. Indeed, leptin treatment increased p300 HAT binding to Bdnf promoters I, IV, and VI, and this effect was attenuated by inhibition of AKT. Furthermore, C646, a p300 HAT inhibitor, attenuated leptin-induced acetylation of histone H3. AKT can also phosphorylate other chromatin-modifying enzymes, such as the histone methyltransferase EZH2 (H3K27 methylation) and DNA methyltransferase DNMT1, resulting in changes in function of these enzymes [90,91]. Unlike histone acetylation, which occurs rapidly in response to stimuli, histone methylation, and DNA methylation are relatively static [90]. In this study, we analyzed the methylation level Fig. 8 Schematic diagram illustrating the mechanism by which leptin induces epigenetic regulation of Bdnf gene transcription. Leptin binds to the long leptin receptor isoform, LepRb, and activates AKT signaling, which in turn phosphorylates p300 HAT, resulting in histone modifications at Bdnf exonspecific promoters and thereby promoting chromatin remodeling and gene transcription. of 76 CpG sites (6 CpG islands) within the promoters I, II, IV, and VI and the exonic regions following leptin treatment. The methylation at 75 CpG sites was unaltered by acute leptin treatment, suggesting that DNA methylation may not play a significant role in leptin-induced Bdnf gene expression. However, we cannot rule out the possibility that other CpG sites we did not analyze may be sensitive to leptin treatment. Methylation of histones at different residues results in different transcriptional outcomes. For example, methylation of H3K27 and H3K9 is associated with transcriptional repression, whereas methylation on H3K4 usually increases gene expression [92][93][94]. We observed increased H3K4 methylation coupled with decreased H3K9 methylation with no changes in H3K27 methylation at the Bdnf promoter following leptin treatment. Interestingly, blockade of p300 HAT with C646 also abolished leptin-induced changes in methylation of H3K4 and H3K9. The involvement of p300 HAT in the regulation of histone H3 methylation at the Bdnf promoter may be secondary to its effect on histone H3 acetylation. In support of this, one study reported that H3K4 methylation is dependent upon p300 HAT-mediated H3 acetylation [71]. The methyltransferase SET1C, which methylates histone H3 at lysine 4, was demonstrated to act synergistically with p300 HAT through direct interactions and coupled histone modifications [71]. Our results support that direct communications between the AKT signaling pathway and the chromatin-modifying machinery mediates Bdnf gene expression in response to leptin. This may represent a generalized epigenetic mechanism that permits cells to respond dynamically to environmental signals.
Our findings support the idea that leptin interacts with BDNF via LepRb in the hippocampus to exert its behavioral effects, beyond its role as an adiposity signal. Although the present study only investigated leptin's antidepressant-like effects, the interactions between leptin signaling and BDNF hypothetically could extend to other hippocampal functions, such as learning and memory [18,95,96]. Moreover, regulation of Bdnf gene expression by leptin signaling is complex, with differential regulation of exon-specific expression in a sex-dependent manner. BDNF, in turn, is required for leptin action in the hippocampus. Thus, dysregulation of BDNF presumably could lead to resistance to leptin, contributing to the pathogenesis of depression and Alzheimer's disease.