Obesity induces profound transcriptome changes in adipocytes, and recent evidence suggests that long-noncoding RNAs (lncRNAs) play key roles in this process. We performed a comprehensive transcriptome study by RNA sequencing in adipocytes isolated from interscapular brown, inguinal, and epididymal white adipose tissue in diet-induced obese mice. The analysis revealed a set of obesity-dysregulated lncRNAs, many of which exhibit dynamic changes in the fed versus fasted state, potentially serving as novel molecular markers of adipose energy status. Among the most prominent lncRNAs is Lnc-leptin, which is transcribed from an enhancer region upstream of leptin (Lep). Expression of Lnc-leptin is sensitive to insulin and closely correlates to Lep expression across diverse pathophysiological conditions. Functionally, induction of Lnc-leptin is essential for adipogenesis, and its presence is required for the maintenance of Lep expression in vitro and in vivo. Direct interaction was detected between DNA loci of Lnc-leptin and Lep in mature adipocytes, which diminished upon Lnc-leptin knockdown. Our study establishes Lnc-leptin as a new regulator of Lep.
Obesity has reached an epidemic level worldwide (1). Central to the obesity problem are adipocytes, which play a dual role of storing excess energy as triglycerides and secreting adipokines that exert systemic effects on metabolic homeostasis (2). Governing adipose tissue function is a set of expressed transcripts and proteins, many of which are dysregulated upon obesity. To discover novel obesity genes, transcriptome analysis has been extensively carried out in both mouse and human adipose tissues (3–7), but most studies primarily have focused on protein-coding genes. Long-noncoding RNAs (lncRNAs) are relatively new players in the field of gene regulation (8,9). We and others have shown that lncRNAs are essential regulators of adipogenesis, insulin sensitivity, and thermogenesis (10–12). By using RNA sequencing (RNA-Seq) on three types of mouse adipose tissues, namely inguinal white adipose tissue (iWAT), epididymal WAT (eWAT), and interscapular brown adipose tissue (BAT), followed by de novo transcriptome assembly, we have built a catalog of >1,500 mouse adipose lncRNAs (13). Another catalog of lncRNAs that regulate energy metabolism in liver, adipose tissue, and muscle has been built on the basis of microarray data (14).
Mutation in leptin (Lep), a circulating adipokine released from adipocytes, leads to an extreme form of obesity exemplified by ob/ob mice (15). A few reports of obese patients who harbor the Lep mutation exist, and such patients are responsive to recombinant leptin treatment (16,17). Great interest exists in understanding the regulation of the leptin gene. By using leptin-bacterial artificial chromosome (BAC) enhanced green fluorescent protein transgenic mice, a region 4.5 kilobases (kb) upstream of Lep, acts as an adipocyte-specific enhancer, and this region is bound by the transcription factor FOSL2 (18). A similar strategy reveals a completely different element required for Lep expression in vivo: a nuclear factor Y-bound element −16.5 kb upstream of the Lep transcription start site (19).
To evaluate the changes of lncRNA transcriptome systemically upon obesity, we performed RNA-Seq on adipocytes isolated from BAT, iWAT, and eWAT of control and diet-induced obese mice. We identified 68 lncRNAs that are differentially expressed upon obesity, termed obesity-regulated lncRNAs in adipocytes (lnc-ORIAs). Specifically, we focused on one particular lnc-ORIA, Lnc-leptin, which is located in an enhancer region upstream of Lep and highly correlates to the expression of Lep. By using multiple independent loss-of-function approaches, we show that Lnc-leptin regulates the expression of Lep in vitro and in vivo.
Research Design and Methods
Diet-Induced Obesity Models
Male mice on the C57BL/6 background were kept at the Duke-NUS animal facilities. Mice were fed normal chow diet (ND) or high-fat diet (HFD) (#D12492; Research Diets) for 16 weeks commenced upon weaning at age 3 weeks.
Primary Adipocyte Culture and Differentiation
Inguinal fat pads from 3-week-old C57BL/6 pups were excised, minced, and digested in collagenase solution at 37°C for 20 min. The suspension was filtered through 100-μm strainers and spun at 2,000 rpm for 5 min. The pelleted stromal vascular fraction (SVF) was resuspended in 10 mL DMEM supplemented with 10% newborn calf serum (Invitrogen), 100 units/mL penicillin, 100 μg/mL streptomycin, and 10 μg/mL gentamicin (Invitrogen). Cells were grown to confluence, and differentiation was initiated at day 0 with DMEM containing 10% FBS, 0.5 μmol/L dexamethasone, 850 nmol/L insulin, 0.25 mmol/L 3-isobutyl-1-methyxanthine, and 1 μmol/L rosiglitazone for 2 days. Cells were then incubated in DMEM containing 10% FBS and 170 nmol/L insulin for 2 more days. After day 4, cells were maintained for 2 more days in DMEM containing 10% FBS. Experiments were performed on mature adipocytes at day 6.
Adipocytes and SVF Isolation From Adipose Tissue
Adipose tissues were excised from mice and immediately minced in a collagenase solution comprising 0.2% collagenase (C6885; Sigma) and 2% BSA dissolved in Hanks’ balanced salt solution (Gibco). Minced tissues were transferred to a 50-mL tube and incubated at 37°C for 20 min (for eWAT and iWAT) or 40 min (for BAT) at 500 rpm. Subsequently, 10 mL complete DMEM was added. Cell resuspensions were filtered through 100-μm strainers, spun at 2,000 rpm for 5 min, and washed once with PBS. The floating adipocyte layer and the pelleted SVF were collected separately. The SVF was treated with ammonium chloride solution (#07800; STEMCELL Technologies) to lyse red blood cells.
Lnc-Leptin Knockdown by Using Short Hairpin RNAs
Sequences targeting Lnc-leptin were cloned into a retroviral vector pSUPER (oligoengine). Short hairpin RNA (shRNA) sequences are listed in Supplementary Table 4. Retroviral vectors were transfected into packaging cell line 293T cells by using X-tremeGENE 9 (Roche). Virus-containing media were harvested 48 h posttransfection and used to infect primary preadipocytes at ∼60% confluence supplemented with 8 μg/mL polybrene. Media were changed the next day, and cells were induced to differentiate 48 h postinfection.
Lnc-Leptin Knockdown by Using Dicer Substrate Small Interfering RNAs or Antisense Oligos In Vitro
For knocking down Lnc-leptin in preadipocytes to assess its role in adipogenesis, dicer substrate small interfering RNA (DsiRNA) or antisense oligos (ASOs) and their respective controls were transfected into day −2 preadipocytes (>90% confluence) by using lipofectamine (6 μL/mL; Life Technologies). Media were changed the next day, and cells were induced to differentiate 48 h posttransfection. For knocking down Lnc-leptin in primary mature adipocytes, a reverse transfection protocol was used (20). DsiRNA 200 nmol/L (Integrated DNA Technologies) or ASOs 150 nmol/L (GapmeRs; Exiqon) mixed with lipofectamine in Opti-MEM medium (6 μL/mL) were added to each well of a 24-well plate precoated with 0.1% gelatin. Mature primary adipocytes at day 6 were trypsinized and reseeded onto the oligo-lipofectamine mix. Medium was changed the next day, and knockdown efficiency was measured 48 h posttransfection. The sequences of DsiRNAs and ASOs used in this study are listed in Supplementary Tables 5 and 6, respectively.
Lnc-leptin Knockdown by Using ASOs In Vivo
Eight- to 12-week-old C57BL/6 male mice were anesthetized. Hair located at the inguinal area was removed with a trimmer, the underlying skin incised, and the inguinal adipose tissue exposed. Control ASO or ASO Lnc-leptin (20 mg/kg) were injected into the left- and right-side inguinal adipose tissue (∼50 μL/injection), respectively. The surgical wounds were closed with sutures and disinfected with 70% ethanol. Adipose tissues from both sides of the inguinal depot were excised 48 h postinjection, and RNA was extracted and subjected to quantitative RT-PCR.
Preadipocytes or mature adipocytes were trypsinized and resuspended in PBS. A two-step cross-linking protocol was used (21) as follows: Cells were incubated with 1.5 mmol/L ethylene glycol-bis (Sigma) at room temperature for 30 min followed by 1% formaldehyde for 10 min. Cross-linking was stopped by quenching with 0.125 mol/L glycine. The chromatin immunoprecipitation (ChIP) experiment was performed as previously described (22). Five micrograms MED1 antibody (A300-793A; Bethyl Laboratories) were used for immunoprecipitation, and normal rabbit IgG (sc-2027; Santa Cruz Biotechnology) was used as control. ChIP primers used in this study are listed in Supplementary Table 7.
Chromatin Conformation Capture
Chromatin conformation capture (3C) was performed as previously described (23), with modifications. Briefly, mouse adipose cells and tissues were cross-linked with 1% formaldehyde for 10 min, and the reaction was quenched by 125 mmol/L glycine for 5 min. Lysed nuclei were resuspended in 500 μL 1.2× restriction enzyme buffer before incubation at 65°C for 20 min with 22.5 μL 20% SDS followed by an additional 1 h of incubation at 37°C. Next, 150 μL 20% Triton X-100 was added, and samples were incubated at 37°C for another 1 h. Samples were then digested with 800 units XbaI (New England BioLabs) by incubating at 37°C overnight. After restriction enzyme digestion, 40 μL 20% SDS was added to the digested nuclei and incubated at 65°C for 15 min, and 6.125 mL 1.15× ligation buffer and 375 μL 20% Triton X-100 were added to dilute the total DNA to favor intramolecular ligation. The diluted sample was incubated at 37°C for 1 h before the addition of 100 units T4 DNA ligase (New England BioLabs) at 16°C for 4 h followed by 30 min at room temperature. Samples were finally decross-linked at 65°C overnight with an addition of 300 μg proteinase K (Thermo Fisher Scientific) before phenol-chloroform extraction and ethanol precipitation. Samples were further purified by QIAquick Spin columns (QIAGEN) and total DNA concentration quantified using NanoDrop. BAC that spans the whole locus of interest is RP24-369M21. All primers were designed to be within a region of 25–150 base pairs (bp) from the restriction enzyme digestion site and are unidirectional from the 5′ side of the restriction fragment. Primers were designed by using Primer3 software (Supplementary Table 8). Quantitative real-time PCR was carried out with SYBR Green Master Mix on the ABI ViiA 7. Semiquantitative PCR analysis of these primers pairs using the control template reconfirmed that there was only a single PCR product of the correct size when visualized on a 2% agarose gel. The identities of the PCR products also were confirmed through direct sequencing. To obtain data points for normalized relative interaction in the final results, cycle threshold (Ct) values of the 3C template were first normalized with values from an internal primer of control interaction frequencies, which commonly used the Ercc3 locus in mouse (23,24). Each quantitative PCR was carried out in duplicate, and 3C validations were repeated four to six times independently for each condition.
Clustering was done in Cluster software and visualized in TreeView. For each model and for each gene, the gene expression value in fragments per kilobase of transcript per million (fpkm) was log-transformed and mean-centered before clustering.
Western Blot and Real-time PCR
Antibodies used for Western blot analysis were leptin (Ab16227; Abcam), Pparg (sc-7273; Santa Cruz Biotechnology), and β-actin antibody (A1978, 43 kD; Sigma) as loading control. Total RNA was extracted by using RNeasy Mini Kit (QIAGEN). Sequences of quantitative PCR primers are listed in Supplementary Table 3.
RNA-Seq Library Preparation, Sequencing, and Analysis
One microgram total RNA was used for each RNA-Seq library preparation according to the manufacturer’s instructions (New England BioLabs), and sequencing was done on HiSeq 2000 (Illumina). Pair-end reads from each sample were aligned to the mouse genome (mm10 build) using TopHat version 2.0.9. Differential expression between HFD and ND samples was quantified using Cuffdiff 2.1.1. Differentially expressed genes are those that have a log-twofold change of >1 or < −1 and q < 0.05 compared with the control condition. We also required that the differentially expressed genes used for downstream analysis have an fpkm >1 in any of the conditions.
Gene Ontologies and Pathway Analysis
For obesity-induced protein-coding genes, gene ontology (GO) and network analysis was performed using GeneGo (Thomson Reuters). For obesity-induced lncRNAs, GO and motif analysis was done with Genomic Regions Enrichment of Annotations Tools (GREAT) software (25).
All studies involving animals have been approved by the institutional review board of Duke-NUS.
Transcriptome Analysis of Adipocytes From Three Different Adipose Depots Identified a Set of lnc-ORIAs
To evaluate the changes of adipocyte lncRNA transcriptome systemically during obesity, we isolated adipocytes from BAT, iWAT, and eWAT of mice fed an HFD or ND by using a collagenous digestion and fractionation method (Fig. 1A). Because adipose tissue is infiltrated with macrophages and other immune cells upon obesity (26), this step enriches for adipocytes and minimizes the contribution from the other cell types. Lep, an adipocyte-specific gene, was enriched >100-fold in the floating adipocyte layer compared with the pelleted SVF (Fig. 1B). In contrast, expression of the macrophage marker F4/80 (Emr1) was highly enriched in the SVF compared with adipocytes (Fig. 1B). Lineage marker expression, such as Hoxc9, Hoxc10, and Ucp1, both before and after collagenous digestion indicated that the BAT and isolated adipocytes were not contaminated by one another (Supplementary Fig. 1).
We performed RNA-Seq on adipocytes isolated from three different adipose depots (BAT, iWAT, and eWAT) in HFD and ND. More than 200 million paired-end reads in total were aligned (Supplementary Table 1) to Ensembl protein-coding genes (mm10) and a published adipose lncRNA catalog (13). To assess whether obesity affects global mRNA and lncRNA transcriptomes in a similar manner, we performed unsupervised hierarchical clustering on the sets of expressed mRNAs and lncRNAs (fpkm >1), respectively. The main branch of the dendrogram, with the exception of BAT that only has one data set per condition, primarily separates all samples on the basis of diet rather than sites of origin in both mRNA and lncRNA clustering (Fig. 1C). There are 353 protein-coding genes differentially expressed in adipocytes from the three different depots upon obesity (Fig. 1D), including Lep (15), Sfrp5 (27), and Egr1 (28), many of which have previously been identified and studied in the context of obesity. Network analysis of the obesity-upregulated genes using GeneGo, which incorporates curated data from published literature, identified the nuclear factor-κB subunits RelA, Esr1, and Creb1 to be the top three transcription factor hubs (Supplementary Fig. 2A) to mediate the expression changes. GO analysis identified developmental processes (P < 1E-17), response to stress (P < 9E-14), and cell differentiation (P < 3E-11) as the top categories associated with the obesity-upregulated protein-coding genes (Supplementary Fig. 2B). These processes, together with the transcription factors nuclear factor-κB, Esr1, and Creb1, have been implicated in previous studies (29,30), indicating that the current data do reflect biological changes of adipocytes during obesity.
Compared with protein-coding genes, less is known about adipocyte lncRNA changes upon obesity. The data indicate that 68 lncRNAs are significantly differentially expressed between ND and HFD in at least two of the three types of adipocytes we profiled (Fig. 1D). We termed them lnc-ORIAs (Supplementary Table 2). Many lnc-ORIAs display adipocyte-specific expression (Fig. 1E and Supplementary Fig. 3). By using GREAT, which infers function of genomic regions on the basis of the ontology annotations of their neighboring protein-coding genes (25), we found that activation of c-Jun N-terminal kinase (JNK) activity is the only significant (P < 1E-6) GO category associated with the obesity-induced lncRNAs (Fig. 1F). Furthermore, the C/EBPβ motif was found to be enriched from this group of induced lncRNA genes (Supplementary Fig. 1C). JNK has been shown to play a central role in obesity (31), whereas C/EBPβ has been implicated in adipose insulin resistance (32). Together, the current data suggest that meaningful biological insight could be gleaned from transcriptome analysis of lncRNAs and that lncRNAs could be important regulators of obesity.
Many lnc-ORIAs Are Regulated in Various Metabolic Conditions and Bound by PPARG
To confirm the expression changes of the lnc-ORIAs identified from the high-throughput RNA-Seq, we isolated RNA from independent cohorts of ND- and HFD-fed mice, and we confirmed the expression changes of 10 selected lnc-ORIAs on the basis of their expression values, fold changes, and gene structures (Fig. 2A). Of the 10 lnc-ORIAs chosen on the basis of their higher expression level and unambiguous gene structure, 9 were induced upon obesity and 1 was repressed (lnc-ORIA1). The expression changes generally occurred in all three tissues profiled, although tissue-specific differences exist (e.g., lnc-ORIA6). To test whether the changes of these lncRNAs are a general feature in other obesity models, we examined their expression in adipose tissue from ob/ob and control mice and found that the expression of all 10 lnc-ORIAs changes in the same direction as that in diet-induced obesity, with 8 reaching significance (Supplementary Fig. 4).
To investigate whether these selected lnc-ORIAs are responsive to alteration of nutritional status, we measured their expression in adipose tissue of ad libitum mice (fed) and mice that underwent an overnight fast (fasted). Nine of 10 of these targets (except lnc-ORIA7) had significantly decreased expression in the various adipose tissues upon fasting (Fig. 2B). Fasting and diet-induced obesity represent two extremes of the nutritional spectrum, one being an acute nutrient-deprived state and the other a chronic nutrient-excess state. The majority of the lncRNAs examined displayed an inverse correlation pattern of expression in these two conditions (Fig. 2C), suggesting that these lnc-ORIAs could be molecular sensors reflecting the energy status in adipose tissue.
Previous studies have shown that tissue- or condition-specific lncRNAs, similar to protein-coding genes, often are bound and regulated by key transcription factors. By using published PPARG ChIP sequencing data from mouse WAT and BAT (33), we found that one-half of the 10 lnc-ORIAs studied are bound by PPARG at their promoters (3 are shown in Fig. 2D), suggesting that many of these lnc-ORIAs are transcriptionally regulated by PPARG in vivo.
Lnc-Leptin Is an Enhancer lncRNA
We are particularly intrigued by lnc-ORIA9 (hereafter Lnc-leptin), which lies 28 kb upstream of Lep, a satiety hormone secreted by adipocytes that acts centrally to regulate systemic metabolism and immunity (17). Lnc-leptin has two exons, and it overlaps largely with an uncharacterized known transcript Gm30838, which shares the same splice sites as Lnc-leptin (Fig. 3A). The promoter of Lnc-leptin has an open chromatin conformation as shown by published DNase sequencing data. There is positive H3K4 trimethylation (H3K4Me3) signal and RNA polymerase II binding at the transcription start site of Lnc-leptin as shown by published ChIP sequencing data (Fig. 3B and Supplementary Fig. 5A), indicating that this gene is actively transcribed in adipose tissue. Lnc-leptin also harbors positive H3K4 methylation 1 (H3K4Me1) and H3K27 acetylation (H3K27Ac) markers in WAT (Fig. 3B); these histone modifications typically associate with enhancers (H3K4Me1) and active enhancers (H3K27Ac) (34). Similar histone modification architecture in this region also was observed in BAT (Supplementary Fig. 5B). Furthermore, to test whether the Lnc-leptin region is associated with MED1, a component of the mediator complex known to bridge enhancer regions with the general transcription machinery and RNA polymerase II at gene promoters (35), we performed a ChIP experiment in differentiated white primary adipocyte culture (Fig. 3C). Because Lnc-leptin is undetectable in brown adipocyte culture, the ChIP experiment was only performed in differentiated white primary adipocytes. The ChIP results indicate that MED1 is recruited to the promoter regions of Lnc-leptin and Lep (Fig. 3C). Taken together, Lnc-leptin is transcribed from an enhancer near Lep and is an enhancer lncRNA.
Expression of Lnc-Leptin Is Highly Correlated to That of Lep
We next investigated the spatial and temporal expression of Lnc-leptin. By using mouse SVF-derived primary adipocyte culture, we found that expression of Lnc-leptin increases gradually as differentiation progresses in a similar manner as Lep and Pparg (Fig. 4A). To assess whether the expression of Lnc-leptin is specific to adipose tissue, we measured its expression in 20 different mouse tissues and found that it is highest in eWAT followed by iWAT and BAT. Lnc-leptin also is expressed, albeit at a much lower level, in testicle and eye (Fig. 4B). Of note, the tissue-specific expression of Lnc-leptin highly mirrors that of Lep (Fig. 4B). To examine whether the expression of these two genes are correlated, we plotted the expression of Lnc-leptin and Lep across a variety of conditions, including HFD versus ND (n = 45), ob/ob versus wild-type (n = 9), and fasted versus fed (n = 42) mouse adipose tissues. A tight correlation (r > 0.76) was found between the expression of Lnc-leptin and Lep in all examined conditions (Fig. 4C). To also assess whether Lnc-leptin and Lep respond similarly to hormonal signaling, we treated differentiated primary adipocytes with various agents known to alter the expression of Lep. Acute insulin stimulation induced Lep expression (36,37); we found that such an increase was accompanied by an induction of Lnc-leptin (Fig. 4D). Conversely, upon tumor necrosis factor-α (TNF-α) and norepinephrine treatment where Lep was repressed (38,39), Lnc-leptin expression was concomitantly reduced (Fig. 4E and F). Taken together, we have demonstrated a close correlation between the expression of Lnc-leptin and Lep, pointing to a potential causative relationship between them.
Lnc-Leptin Is Required for Adipogenesis
To investigate the role of Lnc-leptin during adipogenesis, we used two independent strategies to knock it down in primary adipocyte cultures: shRNAs and DsiRNAs. First, we infected primary white preadipocytes with retrovirus-harboring control shRNA or shRNA constructs targeting Lnc-leptin. Cells then were differentiated per normal, and RNA was harvested at day 6 postdifferentiation. Lnc-leptin knockdown using two different shRNA constructs led to a >80% reduction of the gene (Fig. 5A). Furthermore, it almost completely blocked adipocyte differentiation. Oil Red O staining showed very little lipid accumulation in the knockdown cells compared with the control (Fig. 5B). This defect was accompanied by a reduction in Lep and the adipocyte-markers Pparg and Adipoq (Fig. 5C). Similar results were obtained when Lnc-leptin was knocked down in preadipocytes by DsiRNA (Fig. 5D), suggesting that Lnc-leptin is required for adipogenesis.
Lnc-Leptin Regulates the Expression of Lep in Mature Adipocytes
Depletion of Lnc-leptin during adipogenesis results in severe inhibition of cell differentiation that can indirectly block Lep expression, so whether Lnc-leptin can directly affect Lep expression is unclear. To test this question, we knocked down Lnc-leptin in mature adipocytes. Primary white preadipocytes differentiated into mature adipocytes, and DsiRNAs or ASOs were transfected into the cells by using a reverse transfection protocol (20). Both methods resulted in a >80% reduction in Lnc-leptin expression (Fig. 5E and F), and both were accompanied by a concomitant reduction of Lep expression. Of note, the expression of two mature adipocyte markers, Pparg and Adipoq, was also significantly reduced upon Lnc-leptin knockdown, but the extent of reduction was less. To test whether knocking down Lnc-leptin would affect Lep expression in vivo, we injected ASO against Lnc-leptin directly into one side of mouse inguinal tissue, with the contralateral side injected with a scrambled control. Tissues were harvested 2 days later for expression analysis. Both Lnc-leptin and Lep expression were significantly reduced upon ASO injection (Fig. 5G), and the decrease was accompanied by a less significant reduction in Lep and Pparg mRNA (Fig. 5G). Lnc-leptin knockdown also led to a decrease in LEP but not PPARG protein expression (Fig. 5H), arguing that the reduction of LEP protein is not due to a decreased PPARG protein level. Taken together, knocking down Lnc-leptin affects Lep expression in mature adipocytes both in vitro and in vivo.
To test whether Lnc-leptin is sufficient to promote Lep expression, we used retroviral vector to overexpress Lnc-leptin in primary white adipocyte culture. Overexpression of Lnc-leptin did not promote the expression of Lep or other adipocyte markers (Supplementary Fig. 6). Thus, Lnc-leptin is necessary but not sufficient to promote Lep expression or adipogenesis.
Lnc-Leptin Mediates a Loop Formation Between Genomic Loci of Lep and Lnc-Leptin
The above studies demonstrate that Lnc-leptin is transcribed from an enhancer region and positively regulates the expression of Lep. A common mechanism many enhancers use is to form a long-distance interaction with the promoter of their target genes to facilitate transcription by recruiting positive regulators. We hypothesize that Lnc-leptin is involved in such an interaction near the Lep promoter. To test long-range chromatin interaction between the genomic loci of Lnc-leptin and Lep, 3C experiments were performed to interrogate the chromatin structure around these two genes (Fig. 6A). By using the promoter of Lep as an anchoring point, the genomic locus that encompasses exon2 of Lnc-leptin was found to interact with Lep promoter (Fig. 6B). The 3C-ligated fragment was sequenced to confirm that it is indeed a hybrid product ligated from the two separate genomic regions. This interaction was attenuated upon knocking down Lnc-leptin (Fig. 6C), indicating that Lnc-leptin is required for the looping event. We propose that Lnc-leptin is a novel enhancer lncRNA that regulates Lep expression by bringing Lep with its upstream enhancer and the transcriptional machinery to proximity (Fig. 6D).
Whole-genome sequencing efforts in the past two decades have revolutionized our understanding of the mammalian genome; now recognized is that the mammalian genome is pervasively transcribed to generate thousands of noncoding RNA species, including lncRNAs (40). Several lncRNAs play key roles in regulating energy metabolism (41). We systemically profiled lncRNAs in three types of adipocytes in diet-induced obese mice and identified 68 regulated lncRNAs, termed lnc-ORIAs. Among the lnc-ORIAs, we focused on Lnc-leptin because of its proximity to Lep. The local genomic structure and histone modification patterns of Lnc-leptin are reminiscent of those of a typical enhancer. However, in contrast to the neuronal enhancer RNAs, which generally are unspliced and lack polyadenylated tails (42), Lnc-leptin was identified through polymerase A tail-enriched RNA-Seq and consists of two exons. Lnc-leptin also differs from the bidirectional enhancer-derived transcripts (43) because our directional RNA-Seq did not detect any transcript coming from the opposite direction (Fig. 2D).
Knockdown experiments using DsiRNA and ASO indicated that knocking down Lnc-leptin leads to a concomitant reduction in Lep expression both in vitro and in vivo (Fig. 5), suggesting that the RNA transcript itself rather than the act of transcription confers the function of the lncRNA. Lnc-Leptin positively regulates the expression of Lep, similar to those lncRNAs that activate the expression of their neighboring protein-coding genes (44–46). We showed by 3C experiment that Lnc-leptin is likely to be required for chromatin interaction between exon 2 of Lnc-leptin and the promoter of Lep (Fig. 6). This putative interaction brings the two genes in proximity in a three-dimensional space that potentially enhances the expression of Lep (Fig. 6). In our proposed model (Fig. 6D), the Lnc-leptin transcript acts as a bridge to enhance the interaction between the Lep promoter and enhancer: It could directly interact with Lep promoter or merely serve as a scaffold to bring transcription factors and histone-modifying proteins together. We acknowledge that many details about the mechanism remain unanswered. For example, we do not know what proteins Lnc-leptin directly interacts with, whether Lnc-leptin forms an RNA-DNA duplex directly with the enhancer or the promoter (or both), or whether Lnc-leptin can interact with other DNA segments. These questions warrant further investigation in future studies.
The regulatory mechanism of Lnc-leptin on Lep may not account for its effects on adipogenesis. Knockdown of Lnc-leptin during adipogenesis resulted in a severe reduction of lipid accumulation and expression of mature adipocytes markers (Fig. 5A–D). Because the formation of mature adipocytes in ob/ob mice is not impaired, we believe that Lnc-leptin may use an LEP-independent mechanism to regulate adipogenesis, which warrants further investigation. In mature adipocytes, knocking down Lnc-leptin seems to reduce the expression of adipocyte markers Pparg and Adipoq in addition to Lep. How such a reduction occurs remains to be answered.
In two previous studies, the genomic region of Lnc-leptin was not identified as a cis-regulatory element that regulates the adipose-specific expression of Lep. Both studies used BAC transgenic reporter mice to identify the cis- and transregulatory elements of Lep in vivo. Wrann et al. (18) identified that a region containing the three Lep exons, both introns, and 5.2 kb of the 5′ flanking sequence was sufficient to drive adipocyte-specific enhancer green fluorescent protein expression, whereas Lu et al. (19) identified −22 to 8.8 kb of Lep as the region required for adipose-specific Lep expression. Both studies excluded the genomic location of Lnc-leptin, which lies ∼28 kb upstream of Lep. It is plausible that Lnc-leptin is not primarily involved in the basal expression of Lep but instead serves as a metabolic sensor to regulate the expression of Lep upon various energy statuses in adipocytes. For a dynamically regulated gene like Lep, its regulation is likely to be controlled by the coordination of multiple regulatory mechanisms. The current study reveals a new layer of regulatory complexity, whereas earlier studies (18,19) have identified several cis- and transregulatory factors. These layers of regulation are not mutually exclusive but are all orchestrated by the cellular energy status. Together, they weave a sophisticated network to rapidly adjust Lep expression to respond to nutritional level alteration.
Funding. This research was supported by the National Medical Research Council (NMRC) under its Cooperative Basic Research Grant (CRBG) (NMRC/CBRG/0070/2014 and NMRC/CBRG/0101/2016) and Open Fund-Individual Research Grant (OFIRG) (NMRC/OFIRG/0062/2017) and by the Ministry of Education–Singapore Tier 2 grant MOE2017-T2-2-015. This work was supported by the RNA Biology Center at Cancer Science Institute Singapore, Duke-NUS, from funding by the Ministry of Education’s Tier 3 grant MOE2014-T3-1-006. S.H. and M.L. are supported by an NMRC-CBRG New Investigator Grant (NMRC/BNIG/2027/2015). This work was supported by National Research Foundation Singapore fellowship NRF-2011NRF-NRFF 001-025 and the Tanoto Initiative in Diabetes Research to L.S.
Duality of Interest. No potential conflicts of interest relevant to this article were reported.
Author Contributions. K.A.L., S.H., A.C.E.W., Z.-c.Z., M.K.-S.L., and M.L., performed the experiments and analyzed the data. K.A.L. and L.S. designed the project, interpreted the data, and wrote the manuscript. L.S. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.