Transcriptome analysis of ciliary-dependent MCH signaling in differentiating 3T3-L1 pre-adipocytes

An understanding of adipocyte responsiveness to G-protein-coupled receptor-(GPCR) derived signals must take into consideration the role of membrane microenvironments; that individual sub-populations of proteins may vary significantly across different regions of the cell, and that cell differentiation alters those microenvironments. 3T3-L1 pre-adipocytes undergo a dramatic phenotypic transformation during differentiation into adipocytes, requiring the development of a transient primary cilium. We demonstrate that melanin-concentrating hormone (MCH) receptor 1, a GPCR that stimulates appetite, translocates to the transient primary cilium during early 3T3-L1 cell adipogenesis. Furthermore, we used RNA-Seq to investigate whether MCH signaling is influenced by its receptor localization and whether MCH can influence the transcriptome of early adipocyte development. We found that MCH signaling is sensitive to receptor localization to cilia, and this alters the adipogenic transcriptional program. Also, novel MCH signaling pathways in 3T3-L1 cells are identified, including those for circadian rhythm, the inflammatory response, and ciliary biogenesis. The presence of active MCH-signaling pathways in pre-adipocytes and the discovery that these pathways intersect with the early adipogenic program, among other newly-identified signaling pathways, suggests that the use of MCH receptor 1 antagonists for clinical interventions may have unintended consequences on adipose tissue development.


Results and discussion
MCH receptor 1 is localized to primary cilia in differentiating 3T3-L1 cells. We hypothesized that MCH receptor 1 would be localized to the plasma membrane of 3T3-L1 cells, but also primary cilia when they appeared on Day 2 of our 10-day differentiation protocol. The experiment in Fig. 1 shows the cellular localization of transiently-transfected eYFP-tagged MCHR1 expressed in 3T3-L1 pre-adipocytes. The localization of MCH receptor 1 depended upon the developmental stage of the cells, moving from the cell surface of pre-adipocytes ( Fig. 1, left panel), to a tiny cellular appendage indicated by the arrows (Fig. 1, right panel) during early differentiation.
To better determine if MCHR1 is actually localizing to primary cilia of 3T3-L1 cells on Day 2, we performed confocal microscopy on cells immunostained with primary antibody to MCH receptor 1 (Fig. 2, left panels) or cells expressing MCH receptor 1 tagged with eYFP ( Fig. 2, right panels) plus antibody towards two primary cilia markers, acetylated tubulin and ARL13B. The overlap between MCH receptor 1 and the ciliary marker patterning indicates that the receptor is indeed localized to the primary cilia in differentiating 3T3-L1 cells. MCH receptor 1 has been previously localized to the primary cilia in neurons 16 , but the significance of this remarkable receptor translocation is unknown. Primary cilia of neurons are quite stable 17 , whereas that of the developing 3T3-L1 preadipocyte are transient, lasting less than 48 hours 5 . Our model system provided a unique environment in which to study the role of primary cilia in regulating MCH receptor 1 signaling. Our samples, taken from non-ciliated 3T3-L1 cells, and Day 2 differentiating and ciliated 3T3-L1 cells provided a tool to capture transcriptomes under different conditions to determine whether MCH signaling is changed when its receptor is localized to primary cilia. It also gave us an opportunity to determine whether MCH could alter the progression of early adipogenesis.
Transcriptome analysis reveals MCH-mediated changes in gene expression that are influenced by primary cilia. Gene expression changes were studied to determine the following: (1) whether the cellular differentiation program toward adipocyte formation was activated, (2) whether MCH treatment influenced this cellular differentiation, and (3) whether primary cilia formation affected MCH signaling. Cells were treated with or without 100 nM MCH for 6 h prior to total RNA extraction. Triplicate RNA-Seq samples were used in the experimental design to minimize the effect of sample variation and increase statistical robustness.

3T3-L1 Pre-adipocytes
Ciliated, Differentiating 3T3-L1 Pre-adipocytes  www.nature.com/scientificreports/ Following RNA sequencing, a large dataset containing sequencing reads was produced from each of the twelve samples. The beginning and tailing part of the 150 bp read were generally having poor quality; these were trimmed, which then focused on bases from 7 to 106. Reads were aligned against the mouse mm10 genome assembly and RefSeq annotation. DeSeq2, an R/Bioconductor package, was used to identify statistically significant changes in transcript levels between the four conditions in pairwise manner: ciliated vs non-ciliated, plus vs minus MCH. Four comparison results were produced that analyzed the transcriptional profiles of: ciliated cells; non-ciliated cells; ciliated cells treated with MCH; and non-ciliated cells treated with MCH. Figure 3 reflects the number of genes that are transcriptionally altered between the conditions stated. False discovery rate (FDR) was used to determine statistical significance; data for FDR < 5% are shown. The number of genes responsive to MCH in 3T3-L1 cells is very small with only 68 genes changing expression in unciliated cells and 92 genes in ciliated cells. Interestingly, there are only 2 genes overlapping between those 2 lists. When the triplicate data set was explored in more detail, we found good agreement across replicates (Fig. 4A) and clear differentiation between clusters using a principal component analysis (Fig. 4B). Select genes passing statistical filtering criteria are shown in Fig. 4C.
It seems that the initiation of the 3T3-L1 cell differentiation program, highlighted by the appearance of a primary cilium, causes activation of a gene expression program elucidating changes in over 2,400 genes (FDR < 5%) in just two days, but when cells are treated with MCH, only about 1,400 genes (FDR < 5%) are differentially expressed as shown in Fig. 3, suggesting that MCH signaling alters the transcriptional program that drives differentiation of preadipocytes. Comparisons of triplicate agreement of ciliated 3T3-L1 cells treated with and without MCH is shown via Fig. 5. There is clearly a cluster of genes on Day 0 and Day 2 that respond to MCH, some genes increasing expression and some genes decreasing expression according to the principal component analysis. This data indicates that MCH signaling alters the transcriptome in both ciliated and non-ciliated 3T3-L1 cells. Although genes for hypoxanthine-guanine phosphoribosyltransferase (hprt), RNA pol II and RNA Pol III were validated as acceptable reference genes for 3T3-L1 cell development 18 , only RNA Pol II was used for comparative calculations throughout this study.

Figure 2.
Co-localization of MCH receptor 1 (MCHR1) with ciliary markers. 3T3-L1 cell adipogenesis was initiated with differentiation media for 2 days. Left panels: Endogenous MCHR1 was detected using anti-MCHR1 antibody. Arl13B-antibody was used to label primary cilia. Right panels: MCHR1-eYFP was transiently transfected into cells prior to differentiation treatment. Anti-Actub102 was used as a ciliary marker. The Overlay panels illustrate colocalization. www.nature.com/scientificreports/  www.nature.com/scientificreports/ Validation of RNA-Seq dataset using known regulators of adipogenesis. The combination of growth arrest with the activation of key adipogenic transcription factors such as c/ebp-α, c/ebp-β, and ppar-γ, change the gene expression profile in these pre-adipocytes resulting in the accumulation of lipid droplets and conversion to an adipocyte 19 . Thus, it would seem like these key genes would be best to validate our dataset. However, elevated c/ebp-β, c/ebp-δ and ppar-δ gene expression occurs just prior to ppar-γ induction, and Gregoire et al. suggest that it is the low levels of these genes that keeps the pre-adipocyte from fully committing to differentiation 20 . It wasn't surprising that these genes didn't stand out as greatly upregulated in ciliated pre-adipocytes when compared to non-ciliated pre-adipocytes during the qPCR validation process as shown in Table 1. However, ppar-γ was found to be highly upregulated (> fivefold) using the qPCR method as opposed to only slightly upregulated (1.3-fold) using RNA-Seq (Fig. 6). We propose that differing sensitivities in the sequence amplifications may be contributing to the varied outcomes. Importantly though, both methods agree that the gene was upregulated in ciliated cells when compared to non-ciliated cells, and the sheer number of significantly  www.nature.com/scientificreports/ changed genes in ciliated 3T3-L1 cells compared to unciliated 3T3-L1 cells supports the claim that the differentiation program has been successfully initiated. Regulation of the adipogenic program by MCH receptor 1 is of great interest. Despite Mul et al.'s findings that sympathetic drive is largely responsible for MCH-mediated effects on adipose tissue, their findings that direct action of MCH on adipose tissue influences cell size, but not the number of cells, leaves open the possibility that circulating MCH agonists or antagonists could positively affect human health 15 . Recently, Hilgendorf and colleagues made a case for ciliary-localized omega-3 fatty acid signaling through FFAR4 coupling to G s as a primary control mechanism for the preadipocyte to adipocyte switch in vivo 21 . 3T3-L1 cells are also known to be responsive to MCH, and the literature includes reference to MCH mediating both G i -and G q -mediated responses 22 . We next selected three of the most significantly expressed genes for further validation of early adipogenesis (non-ciliated compared to ciliated pre-adipocyte) by qPCR. The gene for lipocalin 2, lcn2, codes for an adipokine which has been previously shown to be rapidly upregulated during 3T3-L1 pre-adipocyte differentiation 23 . We confirm this dramatic upregulation in lcn2 expression in ciliated cells in both our RNA-Seq (69-fold) and qPCR (76-fold) experiments. Interestingly, MCH caused a 0.62-fold decrease (p < 0.05) in lcn2 expression as measured by RNA-Seq, suggesting that MCH receptor 1 signaling in primary cilia partially inhibits the adipogenesis program; this is supported by a slight, but statistically significant decrease in ppar-γ expression due to MCH treatment (Fig. 6).
Igfbp4 has been identified as a key regulator of adipogenesis and is also appropriately upregulated in our ciliated pre-adipocytes (6.9-fold via RNA-Seq and 3.7-fold via qPCR), indicating that our adipogenic induction is strong 24 . Growth arrest-specific gene 7 (gas7) regulates mesenchymal stem cell differentiation into osteoblasts and is upregulated when differentiation is induced with dexamethasone 25 . Since osteoblasts and adipocytes have a shared lineage and transdifferentiation has been reported 26 , it isn't too surprising that we also see upregulation of this gene early during 3T3-L1 cell differentiation (6.2-fold via RNASeq and 1.1-fold via qPCR). In conclusion, we are confident that early adipogenesis was initiated in our Day 2 samples. Our data indicates that, at least in early adipogenesis, that MCH signaling in cilia prevents the adipogenic switch from fully activating, and it is highly likely that G i signaling is in part responsible, since G s signaling to cAMP is a necessary step in early adipogenesis 27 .
MCH-mediated transcriptional changes in non-ciliated 3T3-L1 preadipocytes. We initially used a PANTHER Pathway Overrepresentation Test (PANTHER version 14.1; Overrepresentation Test Released 03/12/2019) to determine if certain pathways were statistically overrepresented in our non-ciliated 3T3-L1 cell data set comparing untreated to MCH-treated pre-adipocytes 28 . The Mus musculus reference list was used with a Fisher's Exact Test with FDR multiple test correction. When only results with FDR p < 0.05 were considered, there were no statistically significant hits. We screened the same data using the Reactome Bioinformatics Tool against their human homologues, focusing on hits with p-values < 0.05 29 and obtained a more extensive list of pathways in undifferentiated pre-adipocytes affected by MCH (Supplementary Table 1). Of particular interest to this study are the four clusters of interleukin signaling pathway (Interleukin-4, -13, -10, and -12) genes, sug- www.nature.com/scientificreports/ gesting MCH may be influencing the inflammatory response, and the "cilium assembly" pathway genes (4 genes; cep63, tuba1c, rab8a, hsp90aa13; p-value 3.98E−03), suggesting MCH may be influencing early ciliogenesis. Interleukin-4 (IL-4) and Interleukin-13 (IL-13) expression in 3T3-L1 cells is correlated with lowered lipid storage, while IL-4 has specifically been shown to inhibit adipogenesis and activate lipolysis by stimulating HSL activity. If 3T3-L1 cells can influence the migration of eosinophils then this crosstalk between adipocytes and eosinophils may contribute to chronic inflammation in adipose tissue and adipocyte maturation 30 . We demonstrate that MCH can influence the production of cytokines by differentiating adipocytes. These data, together with reports that MCH knockout-mice have lowered levels of macrophages infiltrating intestinal tissues 13 , strongly support a role for MCH in mediating the inflammatory condition.
Hamamoto et al. have reported that in hRPE1 cells transfected with MCH receptor 1, MCH triggers primary cilia shortening through Akt, so MCH signaling has previously been coupled to ciliary stability 31 . This study, however, is the first time MCH has been implicated in de novo ciliary assembly and this has implications for its function in the central nervous system as well. Not only was altered ciliary assembly gene expression identified as a cluster in non-ciliated cells treated with MCH, but interestingly, the gene for M-Phase Specific PLK1 Interacting Protein (mplkip) was also upregulated 1.68 fold (p < 0.05) with MCH in ciliated cells, and its centrosomal location and putative interaction with Plk1 suggest a role for MCH signaling in regulating centriolar duplication and/or ciliary biogenesis 32 .
Interestingly, when 50 down-regulated genes responding to MCH treatment of ciliated preadipocytes were subjected to PANTHER Pathway Analysis, Purine Metabolism (2 genes, ampd3 & xdh; p-value 1.54E−04; FDR 2.53E−02) and Circadian Clock System (2 genes, per1 & per2; p-value 2.35E−04; FDR 1.93E−02) genes were found to be moderately overrepresented. A follow-up GO-Slim Biological Process Analysis also re-identified per1 and per2, as well as a third Circadian Rhythm gene, nfil3 (79.06-fold enrichment; p-value 1.13E−05; FDR 2.02E−02). We again screened the same data using the Reactome Bioinformatics Tool against their human homologues, focusing on hits with a FDR < 0.05 29 . These genes plus a fourth gene, sik1, appeared in the "circadian rhythm" pathway (p-value 3.54E−07). Period 2 (per2) has been shown to inhibit adipogenesis through its interactions with PPARγ 33 and Period 3 (per3) prevents development of adipocytes from mesenchymal stem cells 34 . Nfil3, the third clock gene identified as a gene sensitive to MCH, was previously identified as a linker between the microbiota, circadian clock proteins and the host cells. Microbiota may act to increase nfil3 expression in epithelial cells of the intestine, which in turn regulates lipid storage and fat deposition 35 .
When we screened the full set of data for significantly changed genes with the Reactome Bioinformatics Tool (p-values < 0.05) 29 , we again obtained a more extensive list of pathways in ciliated pre-adipocytes affected by MCH ( Table 2). Pathways related to interleukin signaling reappeared. It seems that MCH can also elicit mRNA expression changes in pathways governing nuclear receptor signaling. The "nuclear receptor transcription" pathway included 2 genes (nr3c1 and vdr, p-value 6.37E−09) and the "signaling by nuclear receptors" pathway included 5 genes (cav2, mmp3, pdk4, dhrs9, fkbp5; p-value 2.28E−02). Additionally, a fifth gene involved in the "circadian clock" pathway (p-value 4.62E−06), nr3c, was identified. Our finding that the gene for the Vitamin D receptor was the most up-regulated gene in ciliated cells in response to MCH was at first surprising to us, but upon further examination and when placed in the context of the complete transcriptome response in ciliated adipocytes, it becomes clearer. Vitamin D regulates the inflammatory response in adipocytes 36 , and modulation of receptor availability by MCH could be a means by which the inflammatory response is controlled. Expression of vdr is elevated in early adipogenesis and it is hypothesized that 1,25(OH)2D3 could exert an anti-adipogenic effect via Wnt/b-catenin. In fact, bone marrow stromal cells are precursors to the osteoblast/adipocyte lineage and the Vitamin D receptor inhibits adipogenesis in these cells by suppressing inhibitors of the Wnt signaling pathway 37 .

MCH signals to circadian clock genes in ciliated 3T3-L1 cells. Molecular circadian clocks establish
rhythmic patterns in cells and tissues to achieve cyclic signaling. MCH signaling caused a small, but significant downregulation of the expression of all three of these genes only in ciliated pre-adipocytes, suggesting a potential connection between primary cilia and circadian clocks. According to Grimaldi et al., the circadian clock regulates key genes responsible for lipolysis 38 . In an effort to establish a connection between MCH signaling, clock gene inhibition, and lipolysis, we followed MCH-mediated gene expression changes of key lipases (atgl, hsl, mgll and abhd5) expressed in non-ciliated and ciliated 3T3-L1 cells by qPCR. As shown in Fig. 7, adipose triglyceride lipase (atgl) and the hydrolase that activates it (abhd5) are normally upregulated when adipogenesis is initiated. MCH signaling significantly downregulates that response. Perilipins are abundant proteins that serve to protect the lipid droplet from lipases. Perilipin 1 is expressed in both white and brown adipose tissue and acts to sequester ABHD5 from ATGL in unstimulated cells. Upregulation of plin1, the gene for perilipin 1, www.nature.com/scientificreports/ is significantly diminished by MCH treatment of ciliated pre-adipocytes as shown in Fig. 8. Together, these data present a model whereby activation of MCHR1on transient primary cilia of differentiating adipocytes dampens the adipogenic signal. Recently, a 24-h temporal gene expression analysis was completed to elucidate the role of circadian rhythms in human white adipose tissue. Despite the small sample size, rhythmicity in gene expression was quite evident in their study 39 . The propensity for MCH receptor 1 to generate signals in ciliated pre-adipocytes that influence circadian rhythm gene expression should be taken into consideration as clinical antagonists are explored for treatment of obesity and metabolic syndromes.

Conclusion
Both ciliated and non-ciliated 3T3-L1 pre-adipocytes responded to MCH, however those responses were quite different, indicating that translocation of MCH receptor 1 to primary cilia during early adipogenesis significantly alters its signal transduction pathway, with the potential to influence adipose tissue development from primary cilia. We discovered that MCH signaling in ciliated pre-adipocytes inhibits gene expression of well-known drivers of adipogenesis and predict that the unique signaling environment of the primary cilium is what enables MCH receptors to participate in the adipogenic decision. This study reveals the importance of studying endogenous GPCRs in all cellular contexts involved in tissue development.

Methods
Tissue culture. Low passage mouse 3T3-L1 Pre-adipocytes (ATCC CL-173) were cultured as a tissue monolayer using high glucose DMEMmedia (ATCC) containing 10% bovine calf serum (Atlanta Biological) without antibiotics. Cells were fed every three days and passaged when they were subconfluent. Culture conditions were set at 37 °C, 5% CO 2 , and 80% humidity.
Induction of 3T3-L1 adipogenesis. 3T3-L1 pre-adipocytes were initially plated and grown to about 70% confluence in maintenance media. Then cells were transferred to DMEM (ATCC) + 10% fetal bovine serum (Atlanta Biological) with 10 μg/ml insulin (SantaCruz), 115 μg/ml isobutylmethylxanthine (Acros), and 400 ng/ ml dexamethasone (AlfaAesar) and incubated for 2 days in a 37 °C, 5% CO 2 incubator to induce primary cilia formation. For some experiments, induction media was withdrawn on Day 2 and replaced with DMEM + 10% fetal bovine serum with 250 ng/ml insulin for two days.  Confocal and fluorescence microscopy. Confocal Imaging was performed with the structured light imaging OptiGrid confocal microscope with Image ProPlus software. In some cases, fixed cell images were taken using a Zeiss Axiocam Mrm fluorescence microscope with AxioView imaging software.
Total RNA extraction. Prior to collection of RNA, we verified that these low passage 3T3-L1 cells retained the capacity to differentiate. 3T3-L1 preadipocytes were grown on 60 mm tissue culture dishes. Total RNA was harvested on Days 0 and 2 of differentiation. Six hours prior to collection, cells were treated with or without 100 nM MCH in differentiation media. Cells were then harvested on ice, and a Qiagen EZ Shredder was used in conjunction with the RNeasy mini kit (Qiagen) per manufacturer's instructions to homogenize the cells and collect the Total RNA. An on-column DNase digestion (Qiagen) was performed to minimize genomic DNA contamination. Each 60 mm dish of monolayer cells yielded 80-100 μl volume of 50 ng/μl total RNA in nucleasefree water. RNA samples were held at − 70 °C. To confirm the quality and quantity of RNA extracted by this procedure, all samples were analyzed on an Agilent 2100 Bioanalyzer at University of Rochester Functional  Figure S1]. Subsequent quantifications were performed on a Nanodrop2000 (Thermofisher). Quantitative PCR. In preparation for quantitative PCR, cDNA was generated from 3T3-L1 RNA stocks using qScript cDNA Supermix (Quanta Biosciences, Gaithersburg, MD). Each 20 µl reaction consisted of 4 µl cDNA Supermix, 4 µl RNA template, and 12 µl nuclease-free water. The reactions were placed in thermocycler for a single three-step cycle: a 5-min step at 25 °C, a 30-min step at 42 °C, and a 5 min step at 85 °C. Following the cycle, each cDNA sample was diluted in nuclease-free water and utilized for subsequent qPCR analysis. For storage, 5 µl of 10 µM Tris/0.1 µM EDTA buffer was added and samples were held at − 20 °C. Early qPCR analyses were performed using PerfeCTa SYBR Green Supermix for iQ (Quanta Biosciences) chemistry using a BioRad MyiQ Thermocycler. Each 25 µl reaction consisted of 12.5 µl PerfeCTa SYBR Green Supermix (2×), 0.97 μl each forward and reverse primers (300 nM final concentration), 2 µl undiluted cDNA template, and 8.56 μl nuclease-free water. Quantitative PCR cycles consisted of a single denaturation cycle of 2.5 min at 95 °C, followed by 40 cycles of denaturation for 12 s at 95 °C and an annealing/extension step of 30 s at 59-62 °C. Melt curve analysis was performed following qPCR runs and ranged from 55 to 95 °C, increasing by 0.5 °C every 30 s step. C T baseline was determined via automatic baseline cycle calculation and C T values determined within the BioRad iQ5 software. Dynamic well factors were determined from each individual PCR plate. Gene expression data obtained from qPCR was used in conjunction with the −ΔΔC T method described by others 40 . Student's t-test was used for statistical analysis of qPCR data comparisons, n = 3 for each, unless otherwise noted.

One
Later qPCR analyses were performed with PowerUp SYBR Green Master Mix using the Applied Biosystems QuantStudio 3 (Thermo Fisher Scientific). Each 25 µl reaction consisted of 12.5 µl PowerUp SYBR Green Master Mix (2×), 1 μl each forward and reverse primers (400 nM final concentration), 2.5 µl cDNA template (optimal concentration for each primer set was determined using standard dilution series), and 800A0μl nuclease-free water and were prepared in quadruplicate following the same qPCR cycle design as previously described. Data was stored in the Thermo Fisher Connect Cloud and analyzed with the Relative Quantitation qPCR Application, where ΔΔC T was calculated. www.nature.com/scientificreports/ All primers were obtained from Integrated DNA Technologies. Upon receipt, lyophilized DNA primers were briefly centrifuged and resuspended in TE buffer (10 µM Tris, 1 µM EDTA) to a stock concentration of 100 µM. Standard curves were used to validate all qPCR primers and melt curve analyses were used to identify off-target PCR products. Primer sequences are available in Supplementary Table 1. RNA-Seq and bioinformatic analysis. Total RNA from twelve samples were sent to the Center for Genome Research and Biocomputing (CGRB) at Oregon State University for library construction and strandedspecific sequencing on an Illumina HiSeq3000 as previously reported 19 . Library construction was described in detail by Shen et al. 4 . Sequencing libraries were prepared with a WaferGen PrepXTM mRNA Strand-Specific Library Preparation Protocol (WaferGen Biosystem). As a result, all libraries are strand-specific and the direction of sequencing reads were identical the original direction of transcripts or mRNAs. All 12 libraries were barcoded and sequenced in one lane with 151 base-pair single-end.

Plin1 Expression Relative to Reference Gene
Approximately 400 million 151-base reads were obtained for all 12 samples. Each sample had about 33 million reads in average after demultiplex ranging from 27 to 48 million [Supplementary Figure S2]. FastQC 41 was used to check the quality of sequencing reads. Bases at the beginning and the tailing part of reads were trimmed since they had lower quality scores. The remaining bases of 7-106 were kept for each read containing 100 bases. Further analysis followed the procedures as described by others 4 . Briefly, all the reads were mapped to mouse mm10 assembly with RefSeq gene annotation using STAR 19 . HTSeq were employed to count the number of reads assigned to each gene 42 . After alignment, there were about 27.8 million in average mapped to the reference, which were corresponding to a mapping rate averaging 84%, varying from 78 to 88% [Supplementary Table 3]. The genes with very low number of reads or no reads were filtered out based on data distribution using density plotting (data not shown) to ensure the reliable data analysis. Transcript-specific variance estimates were obtained by fitting the negative binomial model implemented in DESeq2 43 and significantly differentially expressed genes were determined with Wald statistics among the biological replicates. Those genes were functionally analyzed using DAVID (The Database for Annotation, Visualization, and Integrated Discovery) 19 , PANTHER 28 , and Reactome 29 . The averages of the data were reported ± the standard error of the mean (SEM). Student's t-test was performed to determine statistical significance with data scoring equal to or greater than the 95th percentile considered significant.

Data availability
The RNA-Seq expression data for Day 0 and Day 2 differentiating 3T3-L1 cell transcriptomes is available at GEO accession number GSE145452. www.nature.com/scientificreports/ Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. This is a U.S. Government work and not under copyright protection in the US; foreign copyright protection may apply 2021