Transcriptomic and MicroRNA Analyses of Gene Networks Regulated by Eicosapentaenoic Acid in Brown Adipose Tissue of Diet-Induced Obese Mice

Abstract

Brown adipose tissue (BAT) dissipates chemical energy as heat via thermogenesis and protects against obesity by increasing energy expenditure. However, regulation of BAT by dietary factors remains largely unexplored at the mechanistic level. We investigated the effect of eicosapentaenoic acid (EPA) on BAT metabolism. Male C57BL/6J (B6) mice fed either a high-fat diet (HF, 45% kcal fat) or HF diet supplemented with EPA (HF-EPA, 6.75% kcal EPA) were used for 11 weeks. RNA sequencing (RNA-Seq) and microRNA (miRNA) profiling were performed on RNA from BAT using Illumina HiSeq and miSeq respectively. We conducted pathway analyses using ingenuity pathway analysis software (IPA^®) and validated some genes and miRNAs using qPCR. We identified 479 genes that were differentially expressed (2-fold change, n=3, p ≤ 0.05) in BAT from HF compared to HF-EPA. Genes negatively correlated with thermogenesis such as hypoxia Inducible factor 1 alpha subunit inhibitor (Hif1αn), was downregulated by EPA. Pathways related to thermogenesis such as peroxisome proliferator-activated receptor (PPAR) were upregulated by EPA while pathways associated with obesity and inflammation such as nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB) were downregulated by EPA. MiRNA profiling identified nine and six miRNAs that were upregulated and downregulated by EPA, respectively (log2 fold change > 1.25, n=3, P ≤ 0.05). Key regulatory miRNAs were involved in thermogenesis, such as miR-455–3p and miR-129–5p were validated using qPCR. In conclusion, the depth of transcriptomic and miRNA profiling revealed novel mRNA-miRNA interaction networks in BAT which are involved in thermogenesis which regulated by EPA.

1. Introduction

Obesity is a major metabolic disease that has reached global epidemic proportions. It increases the risk of obesity-associated comorbidities including type 2 diabetes, cardiovascular diseases, and cancer [1]. White adipose tissue (WAT), includes both subcutaneous adipose tissue (SAT) and intra-abdominal visceral adipose tissue (VAT), which are depots for major fat storage in obesity [2]. On the other hand, brown adipose tissue (BAT) which is another adipose depot, acts as an anti-obesity target due to its thermogenic function [3]. BAT is further classified into classical, (uncoupling protein 1 positive; (UCP-1⁺); found in the interscapular and perirenal depots and non-classical, “beige” or “brite” fat which are also UCP-1⁺, but found in WAT depots [4].

BAT induces non-shivering thermogenesis by producing heat which is activated by cold, exercise, therapeutic and pharmacological approaches [5]. Cold temperature stimulates sympathetic nervous system (SNS) to activate β-adrenergic function which in turn increases lipolysis of triglycerides (TG) leading to increased release of fatty acids and fuel substrates for thermogenesis [6]. BAT also consumes circulating glucose and lipids for its thermogenic function making it a valuable anti-obesity target [7]. Recent research indicates that activation of thermogenesis leads to a higher weight loss in individuals with obesity, which has spiked interest in identifying potential pharmacological and dietary approaches to activate BAT [3].

While energy restriction is the main dietary approach to prevent and treat obesity, bioactive compounds such as fish oil which contains long-chain ω−3 polyunsaturated fatty acids (PUFA) such as docosahexaenoic acid (22:6n-3, DHA) and eicosapentaenoic acid (20:5n-3, EPA) have potential anti-obesity effects [8]. These fatty acids have established roles as anti-inflammatory, cardioprotective and triglyceride-lowering agents [9]. Moreover, our laboratory has reported that feeding a high-fat (HF) diet enriched with EPA to mice significantly reduced body weight, fat mass and glucose intolerance while improving insulin sensitivity and inflammation (to levels comparable to low-fat fed mice) [10]. Additionally, we recently reported that EPA increased levels of several classical thermogenic biomarkers at both gene and protein levels in BAT of mice fed EPA and in clonal brown adipocytes [11].

Recent studies also demonstrate the role of microRNAs (miRNAs) to regulate BAT [12]. However, to our knowledge, no studies have conducted a global search for mRNAs and miRNAs pairs which coordinately mediate dietary regulation of BAT. We hypothesize that EPA regulates novel genes and miRNAs to mediate BAT regulation. For these purposes, we employed RNA sequencing (RNA-Seq), which is an unbiased and integrated approach to identify differential regulation of BAT genes, along with small RNA-Seq to detect miRNA profiles. We also identified integrated pathways and validated few targets to gain further insight into mechanisms mediating metabolic effects of EPA.

2. Materia and Methods:

2.1. Animals and Experimental Design:

The experimental groups used in this study have been previously described in detail [10]. Briefly, male B6 mice aged 5–6 weeks were fed a HF (45% kcal from fat) or a HF diet supplemented with 6.57% kcal EPA for 11 weeks (HF-EPA) [10]. BAT from interscapular depot was carefully dissected and weighed for subsequent analyses [10]. These protocols were all approved by the Institutional Animal Care and Use Committee at the University of Tennessee, Knoxville and Texas Tech University.

2.2. RNA Extraction and cDNA Library Preparation

Approximately 70 mg of frozen BAT was homogenized in QIAzol^® (QIAGEN, Redwood City, CA, USA). RNA was isolated according to manufacturer’s protocol using RNeasy Mini Kit (QIAGEN, Redwood City, CA, USA) and quantified using Nanodrop, (Thermo Scientific, Waltham, MA, USA). Quality of the RNA samples was assessed using Agilent 2200 Tape station (Agilent Technologies, Santa Clara, CA, USA). RNA integrity numbers (RIN) for the samples ranged from 7.2 to 8.1. cDNA libraries were made using Illumina’s TruSeq RNA sample preparation kit v2 (Illumina Inc., San Diego, CA, USA) for 2–4 μg of RNA per sample. Briefly, the poly-A containing mRNA was purified, fragmented from total RNA and double stranded cDNA was made. cDNA was recovered using Agencourt, AMPure XP beads (Beckman Coulter, IN, USA). The sample identifier adapters were ligated to the ends of cDNA using 3’ adenylation and adapters ligation. We used specific index for each sample to facilitate multiplexing then adaptor ligated fragments were purified by using AMPure XP beads and amplified with polymerase chain reaction (PCR). Validation and identification of mid insert size of cDNA libraries was performed using Agilent 2200 Tape station. Libraries were quantified using Qubit 2.0 fluorimeter (Invitrogen, Life Technologies Waltham, MA, USA). Prior to loading into HiSeq rapid flow cell, the cDNA libraries were diluted, and denatured.

2.3. mRNA and miRNA Sequencing Strategy

For transcriptomic studies, paired end sequencing was performed using Illumina HiSeq 2500 (Illumina, San Diego, CA, USA) with a 108 bp read length at the Center for Biotechnology and Genomics Core Facility at Texas Tech University to study gene expression. For miRNA profiling, Illumina small RNA protocol (Illumina, Inc., San Diego, CA, USA) was used for libraries preparation and sequencing, then the Illumina Genome Analyzer NextSeq 500 (Illumina, Inc., San Diego, CA, USA) was used to perform sequencing at Department of Molecular and Cell Biology, Baylor College of Medicine, Houston.

2.4. Quality Control (QC) and reads alignment

For gene expression, FastQC v0.10.1 High Throughput Sequence QC report (Version 0.11.2) [13] was used to check the quality of raw sequence reads. The average base call quality Q score based on fastqc data was 38 (Supplementary Figure S1). The reads were mapped to the Mus musculus genome (GRCm38.p4) using Qseq® software Version 12 (DNASTAR Madison, WI). For miRNA expression, the Gunaratne Next Generation pipeline was used to identify miRNA expression profiles [14]. Fastqc data on quality control for small RNA sequencing are provided in supplementary figure S2. In the normalization process, counts of each unique read of miRNA were normalized to total usable reads and then 40 counts were added. Normalized expression values of genes and miRNAs were presented as reads per kilo base per million mapped reads (RPKM).

2.5. Data availability

RNA sequencing data for HF and HF-EPA were submitted in the BioProject at NCBI under PRJNA353387 individual accession numbers for HF Sequence Read Archive (SRA) submission: BAT1 (SAMN05717642), BAT2 (SAMN05717643), and BAT3 (SAMN05717644). For HF-EPA SRA submission: BAT-EPA1 (SAMN05991740), BAT-EPA2 (SAMN05991741), and BAT-EPA3 (SAMN05991742). The small RNA-Seq data were submitted to Gene Expression Omnibus (GEO) data repository (http://www.ncbi.nlm.nih.gov/projects/geo/) under accession number GSE85101 and GSE99506 for BAT HF and BAT HF-EPA respectively.

2.6. Quantitative real-time polymerase chain reaction (q-PCR) validation

RNA was isolated from BAT using RNeasy kit (QIAGEN, Redwood City, CA, USA) followed by cDNA synthesis using the iScript kit (Bio-Rad Laboratories, Inc. CA.USA) and TaqMan® Advanced miRNA cDNA Synthesis Kit (Life Technologies Corporation, Pleasanton, CA, USA) for genes and miRNAs respectively. Quantitative polymerase chain reaction (qPCR) was performed using BioRad CFX-96 real time PCR detection system (Bio-Rad Laboratories, Inc. CA.USA). 18S and miR-191–5p was used as housekeeping control for genes and microRNAs respectively.

2.7. Statistical analyses:

Differentially expressed (DE) genes between HF (control) and HF-EPA groups were identified as those with 2 or more-fold change at 95% confidence using Moderated t-test with false discovery rate (FDR) of 5% (using Benjamini & Hochberg test) with QSeq software. P ≤ 0.05 was considered significantly different based on Student’s t-test. The Ingenuity Pathways Analysis (IPA®, QIAGEN Redwood City, CA, USA; www.qiagen.com/ingenuity) software was used for analysis of target pathways, genes and networks. We used RPKM 0.3 or more in both HF and HF-EPA as the threshold for mRNA expression [15]. In the canonical pathway analyses, Fischer’s exact test p values indicates the significance of enrichment of pathways by DE genes and the Z score to determine whether the canonical pathway is activated or inhibited based on the DE genes in the dataset. Z score ≤ −2 is inhibited and ≥ 2 is activated. For miRNAs, a cut off of log2 fold change > 1.25 was used to determine differentially regulated ones that were DE between HF vs. HF-EPA, RPKM ≥ 3.76. Also, we submitted differentially expressed miRNAs with log2 fold change > 1.25 to canonical pathways using the grow function in IPA®. For gene and miRNA validation, results are presented as means ± SEM and the ΔΔCt method was used to determine mRNA expression and fold changes. Three to five replicates were used for each dietary or treatment groups.

3. Results

We have previously reported significant reductions in body weight, insulin resistance, and body fat in HF-EPA and low fat (LF) group compared to HF fed mice [10]. The mean final body weights of the HF, HF-EPA, and LF fed mice were 40.4 ± 1.2 (g), 35.9 ± 0.9 (g), and 31.7 ± 1.0 (g) respectively (Supplementary Figure S3A), (P ≤ 0.05). Mean BAT weights of the HF, HF-EPA, and LF groups were 0.28 ± 0.02 (g), 0.18 ± 0.02 (g), and 0.2± 0.02 (g) respectively (Supplementary Figure S3B), (P ≤ 0.05). No significant differences were observed in food intake between the HF, HF-EPA, and LF groups (Supplementary Figure S3C) [10]. Furthermore, hematoxylin and eosin stained histological sections of BAT (HF, HF-EPA, and LF) showed reduced lipid content in HF-EPA compared to HF; while HF-EPA was comparable with LF (supplementary Figure 4).

3.1. RNA sequencing data Quality

BAT from HF and HF-EPA groups were used for Illumina paired-end sequencing for whole transcriptome mRNA sequencing. Total sequences per sample were about 24 million paired-end reads of 108 bp length. For small RNA sequencing, an average of 13.8 million sequences reads per sample was mapped to the Mus musculus genome (build mm 10) and a total of 307 miRNAs were identified in BAT.

3.2. Analyses of gene and miRNA differences between two diets

Using Qseq software analysis, genes with RPKM cut off ≥0.3 was filtered to avoid poorly expressed genes. HF and HF-EPA groups expressed a total of 11686 and 12054 mRNAs genes respectively; among these 11667 were common genes that existed between HF and HF-EPA as shown by venn diagram (Figure 1.A). Hence, only 19 and 387 genes, were found to be exclusively expressed in HF or HF-EPA groups respectively. Out of 11667 common genes identified, 479 of them were differentially expressed (DE) between HF and HF-EPA with a criterion of 2-fold change and significance of p ≤0.05 (Supplementary Table 1). Out of the 479 DE genes, 31 were upregulated while 448 were downregulated with EPA respectively (Figure 1.A).

Figure 1. Gene and miRNA profiling and Hierarchical clustering process in B6 mice.

(A) Gene filtering process using venn diagram following RNA sequencing for HF vs. HF-EPA in BAT. Four hundred and seventy nine DE genes were identified using QSeq software of which 31 and 448 genes were upregulated and downregulated by EPA respectively (P ≤ 0.05, fold change ≥2, RPKM ≥0.3). (B) MiRNA filtering process using venn diagram following miRNA profiling for HF vs. HF-EPA in BAT. Fifteen DE miRNAs were identified of which 9 and 6 miRNAs were upregulated and downregulated by EPA respectively (P ≤ 0.05, log 2 fold change > 1.25, RPKM ≥ 3.76). (C) Four hundred and seventy nine DE genes in BAT, which were differentially regulated by EPA. The vertical line represents the different diets (HF vs. HF-EPA) and the horizontal line represents the list of genes. Lower values are represented with blue tones while higher values have more intense yellow tones (P ≤ 0.05, fold change ≥ 2, RPKM ≥ 0.3). (D) Fifteen DE miRNAs in BAT, which were differentially regulated by EPA. The vertical line represents the different diets (HF vs. HF-EPA) and the horizontal line represents the list of miRNAs. Lower values are represented with blue tones while higher values have hotter red tones (P ≤ 0.05, log 2 fold change > 1.25, RPKM ≥ 3.76). HF, high fat; HF-EPA, HF diet supplemented with EPA; DE, differentially expressed.

In the miRNA analyses, a total of 307 miRNAs were expressed in both groups with 267 common miRNAs, out of which, 290 and 284 miRNAs were expressed in HF and HF-EPA respectively. Using an RPKM cut off ≥ 3.76, we identified 23 and 17 miRNAs that were expressed only in HF or HF-EPA respectively. Next, using the Gunaratne Next Generation pipeline, we identified 15 miRNAs that were DE with a log2 fold change > 1.25 (P ≤ 0.05). Out of the 15 miRNAs DE, 9 were upregulated and 6 were downregulated with EPA respectively (Figure 1.B). Based on the DE expression of genes and miRNAs, we used hierarchical clustering (HC) to determine the similarity in expression level in different treatment groups (HF and HF-EPA), which is represented by specific colors (Figure 1.C and 1.D).

3.3. Top up and downregulated mRNAs and miRNAs by EPA

In this study, we identified several DE genes by EPA using Qseq software in BAT (Table 1). Among these, Serine/threonine-protein kinase (Sgk2) gene which is associated with thermogenesis [16], was significantly upregulated (2.11 fold change) by EPA compared to HF diet. Also, significant increases of Paraoxonase 1 (Pon1) gene (2.11 fold change), which has antioxidant properties [17] was observed with EPA. Top downregulated genes by EPA (Table 1) included genes negatively associated with thermogenesis such as translocation-associated Notch protein TAN-1(Notch1) [18], hypoxia Inducible factor 1 alpha subunit inhibitor (Hif1αn) [19], adenosine A1 receptor (Adora1) [20], nuclear receptor corepressor 2 (Ncor2) [21], early growth response 1 (Egr1) [22], insulin like growth factor 2 (Igf2) [23], transforming growth factor receptor 3 (Tgfβr3), and smad family member 3 (Smad3) [24]. Genes associated with obesity, and type 2 diabetes such as CREB binding protein (Crebbp), inhibin beta B (Inhbb), Sp1 transcription factor (Sp1), caspase 7 (Casp7), and forkhead box O1 (Foxo1) were also downregulated [25–27]. Additionally, genes associated with inflammation, Jun proto-oncogene (Jun) [28], and lipid metabolism, Glycerol-3-phosphate dehydrogenase 1 (Gpd1) were downregulated by EPA (P ≤ 0.05, fold change ≥ 2, RPKM ≥0.3) [29].

Table 1.

Top upregulated and downregulated mRNA transcripts in BAT from B6 mice (HF vs. HF-EPA), which were differentially regulated by EPA using QSeq software (P ≤ 0.05, fold change ≥2, RPKM ≥0.3). HF, high fat; HF-EPA, HF diet supplemented with EPA; RPKM, reads per kilo base per million mapped reads.

	Gene Name	RPKM HF	RPKM HF-EPA	Fold Change	P value
Top upregulated genes
Klhl6	Kelch-Like Family Member 6	0.65	1.39	2.14	0.01
Sgk2	Serine/threonine-protein kinase	0.5	1.05	2.11	0
Pon1	Paraoxonase 1	8.43	17.8	2.11	0
Rrl41	Catalyze Protein Synthesis, Ribosomal Protein L41	5.48	11.49	2.1	0.01
Top downregulated genes
Notch1	translocation-associated Notch protein TAN-1	9.43	4.66	0.49	0.01
Casp7	Caspase 7	3.8	1.79	0.47	0.05
Hif1an	Hypoxia Inducible Factor 1 Alpha Subunit Inhibitor	2.44	1.13	0.46	0
Gpd1	Glycerol-3-Phosphate Dehydrogenase 1	3.53	1.63	0.46	0.00
Crebbp	CREB Binding Protein	4.57	2.04	0.45	0
Adora1	Adenosine A1 Receptor	19	8.25	0.43	0
Foxo1	Forkhead Box O1	4.65	1.95	0.42	0
Smad3	Smad Family Member 3	2.51	0.9	0.36	0
Igf2	Insulin Like Growth Factor 2	2.67	0.97	0.36	0.02
JUN	Jun Proto-Oncogene	28.37	9.56	0.34	0
Inhbb	Inhibin Beta B	1.57	0.52	0.33	0.01
Sp1	Sp1 Transcription Factor	2.96	0.95	0.32	0.01
Egr1	Early Growth Response 1	5.41	1.75	0.32	0
Tgfbr3	Transforming Growth Factor Receptor 3	5.52	1.64	0.3	0
Ncor2	Nuclear Receptor Corepressor 2	14.14	5.58	0.2	0.01

We identified 15 miRNAs which were differentially expressed between HF and HF-EPA diets (Table 2). Candidates with high differential expression were further validated; these include miR-455–3p, miR-129–5p, miR-129–2-3p, miR-129–1-3p, miR-181c-3p, miR-199a-3p, and miR-31023p. Of these, miR-455 was previously reported to play a critical role in BAT adipogenesis [19]. Also, miR-129–5p which was identified in this study, was also shown previously to regulate thermogenesis and energy expenditure [30]. However, to our knowledge, no studies have identified regulation of these miRNAs in response to EPA. The rest of the miRNAs we identified are novel and not previously known to be involved in BAT regulation, specifically in response to EPA.

Table 2. List of miRNAs regulated by EPA.

Fifteen miRNAs were differentially expressed between HF and HF-EPA diets in B6 mice. Nine and 6 miRNAs were upregulated and downregulated by EPA respectively. HF, high fat; HF-EPA, HF diet supplemented with EPA; RPKM, reads per kilo base per million mapped reads.

miRNA Name	RPKM HF	RPKM HF-EPA	Fold Change	P value
Upregulated miRNA
miRNA-17–3p	3.74	5.33	1.59	0.03
miRNA-129–2-3p	9.75	11.30	1.55	0.01
miRNA-129–5p	8.70	10.27	1.57	0.01
miRNA-129–1-3p	3.32	4.94	1.61	0.01
miRNA-148a-5p	6.56	7.22	0.67	0.02
miRNA-150–5p	14.52	15.26	0.74	0.02
miRNA-181c-3p	3.85	5.33	1.7	0.04
miRNA-2137	12.48	13.27	0.79	0.03
miRNA-455–3p	12.15	13.52	1.38	0.05
Downregulated miRNA
miRNA-125b-1–3p	7.66	6.44	−1.22	0.03
miRNA-199a-3p	10.3	9	−1.03	0.03
miRNA-199b-3p	10.3	9	−1.03	0.03
miRNA-3102–3p	7.50	4.26	−3.24	0.03
miRNA-503–3p	4.89	3.32	−1.57	0.00
miRNA-674–5p	7.22	5.52	−1.7	0.02

3.4. Validation of gene and miRNA profiling results

qPCR was performed for validation of target genes and miRNAs using BAT from HF vs. HF-EPA groups. As previously reported, we observed significantly higher expression of key thermogenesis genes, peroxisome proliferator-activated receptor gamma coactivator 1 alpha (Pgc1α), PR domain containing 16 (Prdm16), Ppar-γ, and Ucp1 in HF-EPA Vs. HF diet [11]. Additionally, we observed significantly lower expression of genes known to have a negative correlation with thermogenesis, such as Tgfbr3, Smad3, and Ncor2 (P≤ 0.05, Figure 2.A). We further tested 5 miRNAs (miR-455–3p, miR-129–1-3p, miR-1292–3p, miR-129–5p, and miR-181c-3p), which were significantly upregulated by EPA and 2 miRNAs (miR-3102–3p and miR-199a-3p), significantly downregulated by EPA in our miRNA profiling. qPCR validation showed that the expression levels of miR-455–3p and miR-129–5p were significantly higher (P≤ 0.05) in HF-EPA compared to HF groups. Other miRNAs validated showed trends towards significance (P<0.1) (Figure 2.B). The validation of gene and miRNA (HF vs. HF-EPA) compared to low LF diet is included in supplementary figure 5.

Figure 2. Validation of genes and miRNAs in B6 mice:

Genes and miRNAs were validated using RT-qPCR (A) Gene expression of thermogenesis markers Tgfbr3, Smad3, Ncor2, and Hif1an. Data are expressed as mean ± SEM, P ≤ 0.05, n=5. (B) MiR-455–3p, miR-199a3p, miR-129–1-3p, miR-129–2-3p, miR-129–5p, miR-181c-3p, and miR-3102–3p were validated. Data are expressed as mean ± SEM, P ≤ 0.05, n=5. HF, high fat; HF-EPA, HF diet supplemented with EPA.

We conducted further analysis of the DE miRNAs in canonical pathways to create a proposed mRNA/ miRNA network model in IPA (Figure 3). We previously reported that mRNAs such as Ppar-γ, Ppar-α, Prdm16, Pgc1α, and Ucp1 mainly involved in energy expenditure and fatty acid oxidation were upregulated by EPA [11], while some genes negatively associated with thermogenesis such as Hif1αn [19], Tgfβr3 [24], Notch1 [18], Smad3 [24], Ncor2 [21], Igf2 [23], and Egr1 [22], were all downregulated by EPA. Also, in our main network analysis, miR-455–3p and miR-129–5p, were upregulated by EPA, consistent with their role in BAT thermogenic function, and were therefore incorporated from our miRNA profiling analysis [19, 30].

Fig 3. mRNAs and miRNA networks leading to activation of BAT in B6 mice.

Solid lines around boxes represent the genes, which were identified by our study (mean ± SEM, P ≤ 0.05, n=5), while dashed lines represent genes from our previously published work or published literature.

3.5. IPA analysis including canonical pathways and networks

Using IPA, we identified that EPA upregulated pathways such as peroxisome proliferator-activated receptor (PPAR), which is involved in lipid metabolism and fatty acid oxidation and pathways involved in energy balance such as PPAR and retinoid X receptors (PPAR/RXR) [31]. Additionally, energy expenditure pathways such as phosphatase and tensin homolog (PTEN) [32], Rho GDP-dissociation inhibitor (Rho-GDI) signaling pathway which is involved in multi-molecular functions such as shuttling capability for the particular membrane microenvironment were also upregulated by EPA (Table 3) [33]. As expected, inflammatory pathways were downregulated by EPA. They include transforming growth factor beta 1 (TGFβ1) [34], signal transducer and activator of transcription 3 (STAT3) [35], nuclear factor kappa-light-chain-enhancer of activated B cells (NFκB) [36], high mobility group box 1 (HMGB1) [37], integrin linked kinase (ILK) [38], interleukin 6 (IL6) [39], tumor necrosis factor receptor 1 (TNFR1) [40], and toll like receptor signaling (TLR) [41] (Table 3).

Table 3.

Top upregulated and downregulated ingenuity canonical pathways, which were differentially regulated by EPA in BAT from B6 (HF vs. HF-EPA) using IPA. HF, high fat; HFEPA, HF diet supplemented with EPA.

Ingenuity Canonical Pathway	Gene Name	-log (p-value)	Ratio
Top upregulated
PPAR Signaling	Crebbp, Fos, Insr, Jun, Ncor2, Ppard, Rxra	2.47E	7.53E-02
PTEN Signaling	Bmpr2, Cbl, Foxo1, Foxo3, Foxo4, Igf1r, Insr, Pdpk1, Tgfbr3	3.07E	7.63E-02
PPAR/RXR Activation	Bmpr2, Crebbp, Gpd1, Insr, Jun, Map2k6, Ncor2, Rxra, Smad3, Tgfbr3	2.88E	6.21E-02
RHO-GDI Signaling	Arhgap35, Arhgef17, Crebbp, Esr1, Fnbp1, Gnao1, Pip4k2b, Rhob, Wasf2	1.96E	5.20E-02
Top downregulated
TGFB1 Signaling	Bmpr2, Crebbp, Fos, Inhbb, \Jun, Map2k6, Pmepa1, Smad3	3.32E	9.20E-02
STAT3 Pathway	Bmpr2, Cish, Igf1r, Insr, Tgfbr3	1.73E	6.85E-02
NF-κB Signaling	Bmpr2, Crebbp, Igf1r, Insr, Map2k6, Map3k1, Pik3c2b, Tgfbr3, Tnfaip3	1.98E	5.23E-02
HMGB1 Signaling	Fnbp1, Fos, Jun, Kat6a, Map2k6, Pik3c2b, Rhob, Sp1	2.42E	6.67E-02
IL-k Signaling	Crebbp, Fnbp1, Fos, Irs1, Irs2, Jun, Map2k6, Myh9, Pdpk1, Pik3c2b, Rhob	2.73E	5.95E-02
IL-6 Signaling	Fos, IL-6r, Jun, Map2k6, Pik3c2b, Ptpn11	1.45E	5.17E-02
CD 40 Signaling	Fos, Jon, Map2k6, Pik3c2b, Tnfaip3	1.93E	7.69E-02
TNFR1 Signaling	Fos, Jun, Map3k1, Tnfaip3, Casp7	2.45E	1.02E-01
Toll-like Receptor Signaling	Map2k6, Fos, Jun, Map3k1, Tnfaip3	1.71E	6.76E-02

4. Discussion

Our study is the first report of transcriptomic and miRNA profiling in BAT from mice fed a HF compared to HF diet supplemented with EPA. The depth of transcriptomic and miRNA profiling performed in this study uncovered many genes and miRNAs involved in pathways for thermogenesis, energy expenditure, brown fat development, which may link to lipid homeostasis, and fatty acid oxidation. Our findings suggest that EPA increased markers of thermogenesis, similar to cold exposure or diet induced thermogenesis.

In the previous report [11], we investigated the effects of both EPA and DHA on cultured brown adipocytes. We observed more mitochondrial content with EPA than DHA. Therefore, for this study we only used EPA to detect its effect on BAT metabolism related to thermogenesis. While, DHA is a key component of membranes specifically in the central nervous system and contribute to brain development [42], studies have shown that EPA suppresses inflammation and major coronary events better than DHA [43].

Out of four canonical pathways activated by EPA, we focused on PTEN and PPAR pathways, which are related to thermogenesis. PTEN is a tumor suppressor protein which inhibits the activity of phosphatidylinositol 3-kinase (PI3K). Moreover, transgenic mice with higher copies of Pten displayed increased energy expenditure and higher levels of Ucp1 in BAT [32]. Corroborating with this, deficiency of Pten in mice is linked to obesity and insulin resistance suggesting Pten activation is beneficial [44]. These reports and our transcriptomic profiling together suggests that EPA could activate PTEN in BAT, and this in turn may be involved in increasing thermogenesis and energy expenditure.

Another pathway that is modulated by EPA in BAT includes PPAR. This is consistent with previous studies which show that omega-3 fatty acids are ligands for PPARs, namely Ppar-α and Ppar-γ, leading to activation of adipogenesis, thermogenesis, fat catabolism, mitochondrial fatty acid oxidation rates and energy expenditure in BAT and WAT [45]. Ppar-α is highly expressed in BAT and increases levels of thermogenic marker including Pgc1α and Ucp1 [46], all of which are suitable targets for BAT activation and metabolic disorders. Moreover, some studies have already reported that n-3 PUFAs induce thermogenesis in BAT by increasing the level of Ppar-α and Pparγ [47–49]. Beside the activation of BAT, omega 3 fatty acids act as natural ligands and activators of Ppar-α, which blocks NF-KB activity in WAT to prevent inflammation [50]. Furthermore, the free fatty acid receptor 4 (Ffar4), also known as G-protein-coupled receptor 120 (Gpr120), was also significantly upregulated by EPA in our study. However, it was not listed in the differentially expressed genes in the tables above, due to the higher cut off used in our data analyses. These findings are consistent with a recent report demonstrating that Ffar4 was a functional receptor for omega 3 fatty acids in mouse BAT [51].

Our results show that EPA significantly increased miR-455 expression (3). This miRNA targets key brown adipogenic signaling molecules including Hif1αn, Ppar-γ, and Tgfβr3 [19]. Mir-455 inhibits Hif1an, which in turn induces Ucp1. Indeed, higher BAT Ucp1 levels were reported in Hif1αn whole-body knockout mice [52]. Furthermore, Hif1αn is upstream of the AMPK-PGC1α regulatory pathway [19], and miR-455 also activates protein kinase, AMP-activated, alpha1 (AMPKa1), an enzyme which catalytic subunit of AMP-activated protein kinase (AMPK) by targeting Hif1an [19]. AMPK1 activation in turn induces Pgc1α and Ucp1 expression to induce mitochondrial biogenesis and thermogenesis [53]. Thus, our findings are consistent with these reports and validate the importance of this pathway in BAT regulation.

Moreover, the main network pathway in figure 3 shows that Hif1αn serves as an intermediate target between miR-455 and another gene, Notch1. Mir-455 reduces the expression of Notch1, which in turn promotes browning of white adipose tissue to prevent obesity [54, 55]. Additionally, mRNA levels of beige cell markers such as CD137 and T-Box1 (Tbx1), and mitochondria markers such as carnitine palmitoyltransferase 1A (Cpt1a) and Cpt2, were also higher, suggesting beneficial effects of Notch depletion [54]. This has been also confirmed in mouse studies, where Notch specific deletion in inguinal WAT induced Ucp1, cell death-inducing DFFA-like effector A (Cidea), Pgc1α and Prdm16 expression [56]. Consistent with these reports, our data predict an inhibitory effect of EPA on Hif1αn and Notch1 genes, which may increase thermogenesis and energy expenditure by increasing levels of Prdm16, Pgc1α, and Ucp1 as the key regulators of thermogenesis.

Our results further demonstrated that miR-455 also downregulated Tgfβr3 level and its downstream member, Smad2/3. TGF-β signals through dual serine/threonine kinase receptors and its transcription factors named Smads, with Smad3 serving as the major facilitator of TGF-β signal [57]. Blockade of TGF-β/Smad3 signaling leads to enhanced metabolic profile and energy expenditure [24]. Interestingly, Smad3 global knockout mouse is resistant to diet induced obesity and has elevated thermogenic markers and increased basal rate of oxygen consumption in the WAT compared to controls [24]. These results confirm that loss of Tgfβ/Smad3 triggers browning of WAT. In agreement with these studies, our study validates these findings and identified a link between miR-455 and its Tgfβ/Smad3 targets.

PPAR-γ is a nuclear receptor highly expressed in BAT and WAT, which serves as a master transcriptional regulator of brown adipocyte differentiation and it is crucial for tissue development, function, and survival [58]. In our study, EPA induced miR-455 and PPAR-γ expression, inhibiting Ncor expression, known for its role in skeletal muscle and BAT energy homeostasis [59]. Ncor1 is a direct corepressor of PPAR-γ. Hence, clearance of Ncor from the PPAR-γ complex is necessary for recruitment of the brown cofactor Prdm16 [21]. Furthermore, inhibition of Ncor leads to higher levels of PPAR-α, which also stimulates mitochondrial activity, fatty acid β-oxidation via Ucp1 [60]. Thus, our results demonstrate that EPA increased thermogenic markers, in part via regulation of miR-455, a novel EPA target, upstream of PPAR-γ and PPAR-α, which inhibit Ncor to induce key thermogenic biomarkers. However, other studies have also reported potential roles of miR-455 in cancer, specifically as a tumor suppressor [61–63]. Taken together these studies suggest potential dual benefits for miR-455 in both cancer and energy balance.

Another miRNA, miR-129–5p was also significantly upregulated by EPA. Its reported targets are Igf2 and Egr1 [30, 64]. Interestingly, mouse mutant for Igf2 gene demonstrated massive BAT hypertrophy, less lean mass and higher levels of Ucp1 and Prdm16 compared to wildtype mice [23]. Additionally, inactivation of Igf2 in preadipocytes increased Ucp1 and Prdm16 expression consistent with in vivo studies. The other target for miR-129–5p, Egr1 when knocked down in the WAT induced expression of BAT Ucp1 and Pgc1α [22]. These data indicate that miR-129–5p regulates important genes involved in thermogenesis. Although the fold differences in the expression of miRNAs were small in the current study, previous studies have shown that even a small (2-fold) change will have a significant impact on target protein levels [65].

In the present study, we used a diet containing 36 g/kg EPA, which is equivalent to 6.75% of energy intake in diet. In comparison, the intakes of EPA and DHA in the United States are low for omega 3s ~0.1–0.2 g/d [66] and the current recommendations vary from 1 g/d to 4 g/d [67] with upper levels necessary to reduce hypertriglyceridemia. Higher doses have been used in human studies from 4.2 g/d [68] to doses up to 15 g/d described in a recent meta-analysis [69, 70]. However, in human studies, the effect of fish oil in improving fat oxidation and whole-body metabolism is inconsistent. Some factors responsible for theses inconclusive results include usage of lower doses in human studies compared to animal studies, genetic variabilities in humans compared to selected animals for research, and different ratio of EPA/DHA which were used for clinical studies [71, 72]

5. Conclusion

We report several previously identified as well as new genes, miRNAs, and pathways involved in thermogenesis, energy expenditure, and brown fat development. Some of the identified targets of EPA in BAT include miRNAs 455, 129 and PPAR/NCOR, and SMAD3/TGF pathways. These all may lead to improved metabolic homeostasis, suggesting that omega-3 fatty acids, and potentially other nutritional interventions are viable therapies for preventing and/or treating obesity and related metabolic disorders, in part via regulation of BAT. While our study has many strengths and novel findings, it has some limitations. We did not measure the energy expenditure directly in our mice, thus we did not validate whether increased expression of thermogenic biomarkers by EPA translated into increased energy expenditure in the EPA fed mice. Moreover, additional molecular and physiological studies are necessary to further validate the mRNA-miRNA regulatory pairs that we identified in our unbiased screens, their direct role in thermogenesis and mechanism of their regulation by EPA both in vitro and in vivo. Such studies include using miRNA mimics and inhibitors in cells and animals, to validate target genes which are beyond the scope of the current report. Lastly, it will be worthwhile in future to determine if lower doses of EPA compared to DHA or whole fish oil recapitulate the responses reported here for EPA in animal models; and whether these findings can be translated in clinical studies.

Highlights:

• EPA upregulated thermogenic markers similar to cold exposure or exercise.

• MiR-455 and −129 are novel miRNAs involved in regulation of thermogenesis by EPA.

• Smad3, Tgfbr3, and Notch are novel RNAs down-regulated by EPA for thermogenesis.

• Modulation of BAT function using EPA is a potential novel target for obesity

Abbreviations used:

Adra1

adenosine A1 receptor

AMPK

AMP-activated protein kinase

Arhgap35

Rho GTPase activating protein 35

Arhgef17

Rho guanine nucleotide exchange factor 17

Atg1

autophagy related 1

BAT

brown adipose tissue

BMPR

bone morphogenetic protein receptor

Casp7

caspase 7

Cbl

Cbl proto-oncogene

Cidea

cell death-inducing DFFA-like effector A

Cish

cytokine inducible SH2 containing protein

Cpt-1

carnitine palmitoyl transferase I

Crebbp

CREB binding protein

DE

differentially expressed

DHA

docosahexaenoic acid

DMEM

dulbecco’s modified eagle medium

Egr1

early growth response 1

EPA

eicosapentaenoic Acid

Esr1

estrogen receptor 1

FDR

false discovery rate

Ffar4

Free fatty acid receptor 4

Fnbp1

formin binding protein 1

Fos

fos proto-oncogene

Foxo1

forkhead box O1

Gapdh

glyceraldehyde 3phosphate dehydrogenase

GEO

gene expression omnibus

Gnao1

G protein subunit alpha O1

Gpd1

glycerol-3-phosphate dehydrogenase 1

(Gpr120)

G-protein-coupled receptor 120

HC

hierarchical clustering

HF

high-fat diet

HF-EPA

HF diet supplemented with EPA

Hif1an

hypoxia inducible factor 1 alpha subunit inhibitor

Hi-Seq

high sequencing

HMGB

high mobility group box

HS

hematopoietic lineage cell-specific protein

Igf2

insulin like growth factor 2

IL

interleukin

ILK

integrin linked kinase

Inhbb

inhibin beta B

Insr

insulin receptor

IPA

ingenuity pathway analysis

Jun

Jun proto-oncogene

Kat6a

lysine acetyltransferase 6A

Klhl6

kelch-like family member 6

LF

low fat

miRNA and miR

microRNA

mRNA

messenger RNA

Myh9

myosin heavy chain 9

Ncor

nuclear receptor corepressor

NF-κB

nuclear factor kappa-light-chainenhancer of activated B cells

Notch1

Notch homolog 1, translocation-associated (Drosophila)

Pdpk1

3-phosphoinositide dependent protein kinase 1

Pgc1α

peroxisome proliferator-activated receptor gamma coactivator 1 alpha

Pik3c2b

phosphatidylinositol-4-phosphate 3-kinase catalytic subunit type 2 beta

Pmepa1

prostate transmembrane protein, androgen induced 1

Pon

Paraoxonase

Ppar

peroxisome proliferator-activated receptor

Prdm16

PR domain containing 16

Prl41

catalyze protein synthesis, ribosomal protein L41

PTEN

phosphatase and tensin homolog

Ptpn11

protein tyrosine phosphatase, Non-Receptor Type 11

PUFA

polyunsaturated fatty acid

Rhob

ras homolog family member B

RHO-GDI

Rho GDP-dissociation inhibitor

PI3K

phosphatidylinositol 3-kinase

RIN

RNA integrity numbers

RNA-Seq

RNA sequencing

RPKM

reads per kilo base per million mapped reads

RT-qPCR

quantitative reverse transcription

RXR

retinoid X receptor

SAT

subcutaneous adipose tissue

SEM

standard error of mean

Sgk

serine/threonine-protein kinase

Smad

SMAD family member

SNS

sympathetic nervous system

Sp1

Sp1 transcription factor

SRA

sequence read archive

STAT

signal transcription and activator of transcription

Tbx

T-box transcription factor

TG

triglycerides

TGFβ

transforming growth factor –β

TGFβR

transforming growth factor beta receptor

TLR

toll-like receptor

Tnfaip3

TNF Alpha Induced Protein 3

TNFR

tumor necrosis factor receptor

UCP

uncoupling protein

VAT

visceral adipose tissue

Wasf2

WAS protein family member 2

WAT

white adipose tissue