Fig 1.
Transcriptome landscape associated with browning of white adipose tissue (WAT).
(A) Schematic representation of the 3 treatments (chronic cold exposure at 4°C, treatment with β3-adrenoceptor agonist, CL316243, or swimming exercise) used to induce browning of WAT in mice. (B,C) Hierarchical clustering (Ward’s method, spearman correlation) of treatments based on RNA sequencing–derived gene expression profiles (mRNA and long noncoding RNA [lncRNA]) of adipose tissue undergoing browning. (D-G) Venn analysis of overlap between significantly differentially expressed mRNA and lncRNA (false discovery rate [FDR] ≤ 0.05, absolute log2FC ≥ 1) during adipose tissue browning induced by the 3 treatments. (H,I) Scatterplot depicting the significance of overlap between differentially expressed browning-related mRNAs (H) and lncRNAs (I) and their tissue-specific expression patterns. Significance of overlap between tissue-specific and browning-related gene expression was estimated via FDR based on the hypergeometric test. (J) Scatterplot depicting significance of overlap (FDR) between transcription factor binding sites (from chromatin immunoprecipitation sequencing [ChIP-seq] data) and the promoters of browning-induced genes under the 3 treatments. The dash line indicates the position of FDR 0.25. (K) Overlap of the genomic location of brown adipose tissue–enriched lncRNA 10 (lncBATE10) with Peroxisome proliferator-activated receptor gamma (PPARγ) and PR domain containing 16 (PRDM16) binding sites. The PPARγ and PRDM16 binding sites are indicated by red bars. The 3 tracks at the bottom represent the relative abundance of RNA sequencing reads (corresponding to exons) in this region for the different browning-inducing treatments.
Fig 2.
Gene and network expression in browning and whitening studies.
(A) Schematic representation of cellular transitions in brown adipose tissue (BAT) due to cold (4°C) or warm exposure (30°C). (B,C) Comparison of the overlap between differentially expressed genes due to contrasting thermal shifts in BAT. Genes (mRNAs) up-regulated at 4°C are compared to genes down-regulated at 30°C (B) and vice versa (C). (D) Five-way Venn diagram comparing the overlap among significantly up-regulated long noncoding RNAs (lncRNAs) due to browning-inducing treatments in white adipose tissue (WAT) and cold exposure in BAT, and lncRNAs significantly down-regulated in BAT due to exposure at 30°C. (E) Heatmap summarizing expression patterns of lncRNAs regulated in at least 4 out of 5 conditions. The—log false discovery rate (FDR) was used as input. (F,G) Principal components analysis (PCA) on mRNA and lncRNA expression in response to treatments of WAT and BAT. Genes with a fragments per kilobase of exon per million reads (FPKM) > 5 are included for both plots. Treatments are color coded as per the PCA legend, with squares and circles representing WAT and BAT samples, respectively. The first 2 principal components are plotted, and the percent variation of mRNA/lncRNA expression explained by each component is noted in the axis label. (H) mRNA–lncRNA coexpression network based on expression data from WAT and BAT samples. Included in the analysis were 819 mRNAs and 79 lncRNAs showing differential expression in at least 3 of 5 conditions (FDR ≤ 5%, ≥2-fold absolute change). The partial correlation matrix for each pair of mRNAs/lncRNAs was determined via GeneNet, and a clustered gene coexpression network was constructed using iGraph. The size of the cluster was proportional to the number of mRNAs/lncRNAs contained in it, and the width of the edges connecting the clusters was proportional to the total number of inter-cluster links arising from correlated genes in the different clusters. The major functional categories, and the overrepresentation of Gene Ontology Biological Processes (GOBPs) in each cluster, were determined via PANTHER. GOBPs were categorized into 4 broad groups in each cluster. Statistical overrepresentation of GOBP in each cluster was tested via the binomial test. Highly significant processes (P < 1E−09) are listed beside their relevant clusters.
Fig 3.
Brown adipose tissue–enriched lncRNA 10 (lncBATE10) is required for a brown adipose tissue (BAT)-selective gene program in brown adipocytes.
(A) Northern blot to examine the expression of lncBATE10 in mouse brown-, inguinal-, and epididymal adipose tissues. (B) Real-time PCR result of lncBATE10 across mouse tissues and (C) differentiation time course of primary brown and white adipocyte culture. Error bars represent mean ± SEM, n = 3. *P < 0.05. (D) Expression of lncBATE10 in BAT isolated from animals treated with acute cold exposure (4°C for 6 hours). Error bars represent mean ± SEM, n = 4. *P < 0.05. (E) Hosted at thermoneutrality (30°C for 7 days). Error bars represent mean ± SEM, n = 4. *P < 0.05. (F) Expression of lncBATE10 in inguinal white adipose tissue (iWAT) browning, induced by indicated conditions. Error bars represent mean ± SEM, n ≥ 6. *P < 0.05. (G) Primary brown preadipocytes were infected by retroviral control small hairpin RNA (shRNA) and shRNAs targeting lncBATE10, followed by differentiation for 5 days. Oil-red-O (ORO) staining was conducted to examine the lipid accumulation. (H) Real-time PCR was used to detect the expression of lncBATE10, (I) pan-adipogenic markers, and (J) BAT-selective markers. Error bars represent mean ± SEM, n = 4. *P < 0.05. (K) Western blot to examine the expression of Ucp1, Pgc1α, and Peroxisome proliferator-activated receptor gamma (Pparγ) upon lncBATE10 knockdown. (L) Gene-set enrichment analysis (GSEA) analysis was performed on RNA-seq data from control and lncBATE10 knockdown BAT samples. An enrichment plot for genes involved in respiratory electron transport pathway is shown. (M) Expression of thermogenic gene Ucp1 and Pgc1α in the norepinephrine (NE)-treated control and the shRNA-infected brown adipocytes treated. Error bars represent mean ± SEM, n = 4. *P < 0.05 (Student t test). The individual numerical values that underlie the summary data can be found in S13 Data.
Fig 4.
Brown adipose tissue–enriched lncRNA 10 (lncBATE10) is required for a brown adipose tissue (BAT)-selective gene program in the browning of white fat.
(A) The effect of lncBATE10 knockdown on lipid accumulation in primary white adipocyte culture was examined by Oil-Red-O (ORO) staining. (B) The knockdown efficiency, (C) pan-adipogenic markers, (D) mitochondria genes, and (E) BAT-selective genes were examined by real-time PCR. Error bars represent mean ± SEM, n = 3. *P < 0.05 (1-way ANOVA). (F) Gene-set enrichment analysis (GSEA) analysis was performed on RNA-seq data from control and lncBATE10 knockdown white adipose tissue (WAT) samples. An enrichment plot for genes involved in respiratory electron transport pathway is shown. (G) Heatmap for significantly affected Reactome pathways (false discovery rate [FDR] < 0.05) due to lncBATE10 knockout in brown and white adipocyte cultures. The heatmap is color coded by the pathway normalized enrichment scores (NES) obtained from GSEA, with blue representing down-regulated pathways and purple representing up-regulated pathways in knockout samples. (H) Cidea and (I) Ucp1 expression were examined by real-time PCR to determine the effect of lncBATE10 deception on browning induced by rosiglitazone and norepinephrine (NE). Error bars represent mean ± SEM, n = 4. *P < 0.05. (J) The in vivo function of lncBATE10 depletion on maker expression was examined by real-time PCR on inguinal white adipose tissue (iWAT) injected with adenovirus expressing empty vector or small hairpin RNA (sh)-lncBATE10. (K) Western blot was performed to examine the expression of Ucp1 and Pgc1α. Error bars represent mean ± SEM, n = 8. *P < 0.05 (Student t test). The individual numerical values that underlie the summary data can be found in S13 Data.
Fig 5.
Transcription of brown adipose tissue–enriched lncRNA 10 (lncBATE10) is controlled by cAMP-cAMP response element-binding protein (Creb) signaling pathway.
(A) Real-time PCR analysis of the expression of lncBATE10 in primary brown adipocytes culture treated with cAMP and norepinephrine (NE) for 4 hours. n = 3 (B) Potential transcriptional factor binding sites in the lncBATE10 promoter region. Predicted by online program MatInspector (www.genomatix.de). The arrow indicates the transcriptional orientation. (C) LncBATE10 Promoter reporter assay. Promoter regions upstream of the transcriptional start site of lncBATE10 with different truncations were cloned into pGL3-Basic vector. Reporters were transfected into 293T cells. Thirty-six hours after transfection, cells were further treated with 1 uM forskolin for 2 hours and subjected to luciferase assay. Error bars are mean ± SEM, n = 3, *P < 0.05 (Student t test). (D) Four site-specific mutations were made in the functional Creb binding site to construct the mutant promoter. (E) 293T cells transfected with wide-type or mutant reporter were treated with a different dose of forskolin, followed by luciferase assay. Data were normalized by Renilla activities of a cotransfected pRL-CMV plasmid. Error bars are mean ± SEM, n = 3, *P < 0.05 (1-way ANOVA). (F) Chromatin immunoprecipitation (ChIP)-PCR with primers detecting the CREB binding site and a control region 3,000 bp upstream the promoter before and after forskolin treatment. The individual numerical values that underlie the summary data can be found in S13 Data.
Fig 6.
Brown adipose tissue–enriched lncRNA 10 (lncBATE10) decoys CUGBP Elav-Like Family Member 1 (Celf1) from Pgc1α mRNA.
(A) Western blots to detect HuR, Celf1, Celf2 in the RNA-pulldown assay. (B) We used retrovirus to knock down lncBATE10 in primary brown adipocytes. Expression of lncBATE10 were examined by real-time PCR. (C) We performed RNA immunoprecipitation (RIP) using Celf1-specific antibody in the control and lncBATE10 knockdown cell lysates, and examined lncBATE10 and Pgc1α mRNA binding to Celf1. Error bars are mean ± SEM, n = 3, *P < 0.05 (1-way ANOVA) (D) An illustration of predicted Celf1 protein binding sequence (CBS) within lncBATE10 and Pgc1α mRNA. (E) Western blot to examine Celf1 protein binding to CBS RNA fragments. The RNA-pulldown assay was carried out using short CBS RNA fragments derived from lncBATE10 and Pgc1α mRNA. (F) Real-time PCR examination of lncBATE10 and Pgc1α mRNA in the competitive RIP assay are illustrated in S7D Fig. Error bars are mean ± SEM, n = 3, *P < 0.05 (1-way ANOVA). (G,H) RNA-pulldown and competitive RIP assay similar to (E) and (F), respectively, but use an approximately 1-KB Pgc1a 3′UTR fragment with or without the CBS site. n = 3. (I, J) Celf1 was retrovirally expressed in brown preadipocytes followed by 5 days differentiation. The expression level of Celf1 and Pgc1α were examined by (I) real-time PCR and (J) Western blots. (Error bars are mean ± SEM, n = 3, *P < 0.05 (Student t test). (K) A diagram of the luciferase constructs containing CBS from lncBATE10 and Pgc1α. (L) Examination of luciferase activities of reporters with or without CBS sequences within their 3′UTRs. 293T cells were transfected with 3′UTR reporters, respectively. Cells were collected for luciferase assay after 24 hours of transfection. Error bars are mean ± SEM, n = 3, *P < 0.05 (1-way ANOVA) (M) 293T cells were transfected with Celf1 shRNA, followed by transfection of 3′UTR reporters. Western blot was used to examine CELF1 protein level. (N) Luciferase activities of these reporters were measured. Error bars are mean ± SEM, n = 3, *P < 0.05 (Student t test). (O) The diagram to illustrate how lncBATE10 functions. cAMP-cAMP response element-binding protein (Creb) pathway can stimulate the transcription of lncBATE10 and Pgc1α. LncBATE10 functions as a decoy to titrate Celf1 way from Pgc1α mRNA, which otherwise will be repressed by Celf1. The individual numerical values that underlie the summary data can be found in S13 Data.