miR-375 is cold exposure sensitive and drives thermogenesis in visceral adipose tissue derived stem cells

Activation of brown adipose tissue may increase energy expenditure by non-shivering thermogenesis. Cold exposure is one of the options to activate brown adipocytes. To link changes in energy metabolism with microRNA expression (miRNAs), we analyzed 158 miRNAs in serum of 169 healthy individuals before and after cold exposure. Validating the results of a miRNA array, a significant down-regulation of miR-375 after cold exposure (P < 0.0001) was detected. These changes went along with a significant negative correlation between miR-375 and visceral adipose tissue (VAT) mass (P < 0.0001), implicating a specific function of miR-375 in this depot. Significantly higher expression levels of miR-375 were found in VAT in comparison to subcutaneous fat (SAT). Using in silico prediction, we identified putative miR-375 target genes involved in the thermogenesis pathway. Cold-stimulation of subcutaneous and visceral pre-adipocytes (PACs) led to significantly higher expression levels of FABP4, FGF21, PPARGC1A and PRDM16 in VC-PACs. Analyzing miR-375 knock down and cold stimulated VC-PACs revealed a significant up-regulation of thermogenesis associated genes PPARGC1A, ELOVL3 and PRDM16. In summary, our findings identified miR-375 as a potential adipogenic and thermogenesis-associated miRNA exclusively acting in visceral adipose tissue.


Results
Identification of differentially expressed miRNAs in serum of individuals undergoing cold exposure using a miRNA array. We profiled the miRNA spectra via the human miRCURY LNA miRNA Focus PCR Panels before and after cold exposure in pooled samples from 12 male participants to identify miR-NAs affected by cold in serum. In a second array screening, we used pooled samples derived from 12 female individuals. (demographic data Table S1). We identified changes of 158 miRNAs in response to cold exposure in all samples, while miRNAs in males displayed stronger changes after cold exposure (FOC males up to 5.5, FOC females up to 1.8). Among the most abundant miRNAs in serum, five miRNAs including miR-22-3p, miR-99a-5p, miR-185-5p, miR-361-5p, and miR-375 were significantly changed after cold exposure in male samples (Fig. 1A). miR-361-5p displayed the most significant changes in the combined 12 sample by a 5.5-fold difference compared to the samples before CIT. Raw data of the array are shown in Table S4.

Specific circulating miRNAs under CIT.
To validate the five miRNAs identified in the discovery array, as well as miR-92a 14 and miR-125 19,29 which have been previously implicated in brown/beige adipocyte function, we used qPCR on an independent validation sample set comprising paired serum samples from 169 individuals before and after CE (demographic data Table S2). Among the selected miRNAs, miR-375 was predominantly down-regulated in the validation cohort (mean change − 0.922 to − 1.042, p ≤ 0.0001) (Fig. 1B). miR-375 is predominantly expressed in visceral adipose tissue and related to visceral adipose tissue mass. Correlating multiple clinical characteristics (see Table S5) with the miR-375 expression level, a moderate negative correlation of miR-375 levels under thermoneutral conditions with the visceral adipose tissue (VAT) volume was detected (r = − 0.47, P < 0.0001) (Fig. 1C). Analyzing miR-375 in human visceral and subcutaneous adipose tissue revealed a significantly higher expression level in VAT (Fig. 1D). Since SAT and VAT differ in cellular composition, expression differences of miR-375 between the two depots could also stem from nonadipogenic cells. Therefore, miR-375 expression was evaluated using isolated and in-vitro differentiated preadipocytes (PACs). Comparing miR-375 expression levels of VAT derived PACs in comparison to subcutaneous PACs a slightly higher level was detectable in visceral PACs, but did not reach significance (p = 0.2063) (Fig. 1E).
MiR-375 target genes are involved in cold-induced thermogenesis. Using in silico sequencebased target gene prediction, we identified numerous putative miR-375 target genes. Matching these target genes with those of the cold-induced thermogenesis pathway (GO:0106106) we were able to identify 144 overlapping genes. By using string.db and focusing on direct target genes related to uncoupling protein 1 (UCP1) we were able to create a network of direct interaction genes (interaction score = high confidence 0.7) ( Fig. 2A). Namely, adiponectin (ADIPOQ), adrenoreceptor beta 3 (ADRB3), cell death activator (CIDEA), iodothyronine deiodinase 2 (DIO2), fatty acid elongase 3 (ELOVL3), fatty acid binding protein 4 (FABP4), fibroblast growth factor 21 (FGF21), leptin (LEP), peroxisome proliferator-activated receptor gamma coactivator 1-alpha (PPARGC1A) and PR/SET domain 16 (PRDM16) could be identified. The expression levels of the computationally derived genes were assessed for all following cell experiments. www.nature.com/scientificreports/ MiR-375 target genes are differentially expressed between VAT and SAT. To assess the overall gene expression differences between SAT and VAT we used data originating from the "Genotype-Tissue Expression (GTEx)" project. Briefly, GTEx is a public resource to study gene expression in various tissues and has been described elsewhere in detail 33,34 . Samples were harvested post-mortem with SAT originating from the leg and VAT from the greater omentum. From the identified miRNA-375 related genes and direct targets of UCP1 violin plots were generated using the GTEx portal's "Multi Gene Query" (N-paired = 99, Fig. 2B). Gene expressions of ADRB3, CIDEA, ELOVL3, FABP4 and PPARGC1a was higher in VAT comparied to SAT. Detailed fold changes are listed in Table S6. An opposite regulation was seen for ADIPOQ, DIO2, LEP and PRDM16. In addition, the thermogenesis-related genes FGF21 and UCP1 showed higher expression levels in VAT compared to SAT but overall expression remained at a low level.
Adipocytes are the predominant cell type in WAT expressing miR-375 target genes. Using single-cell/nucleus RNA sequencing data publicly available on the Single Cell Portal (study no. SCP137) gene expression levels of the network genes in relation to WAT-resident cell types were displayed (Fig. 2C). The genes ADIPOQ, CIDEA, FABP4, LEP and PPARGC1A are predominantly found in adipocytes, as well as in adipocyte stem and precursor cells (Fig. 2C). Expression of ADRB3, ELOVL3 and UCP1 is solely detected in adipocytes but not in other cell types. The thyroid hormone activating gene DIO2 is predominantly expressed in endometrium cells. FABP4 as an ubiquitous fatty acid carrier protein is relevant in different cell types. The same applies for FGF21, which is important for cell survival and involved in a variety of biological processes. PRDM16 was seen predominantly in smooth muscle cells (SMC), where it promotes muscle cell differentiation 35 . In white adipocytes, as used in this single cell data approach, only low expression of PRDM16 could be expected because PRDM16 specifies the brown fat lineage and is not involved in white adipogenesis. Regarding UCP1, only adipocytes did show detectable expression levels in comparison to the other cell types. www.nature.com/scientificreports/ Visceral pre-adipocytes respond to CE. To evaluate the cold response of paired SC and VC preadipocytes we exposed PACs to cold (30 °C) for several hours (Fig. 2D). Apoptosis as well as cell stress markers such as p53 and lactatde hydrogenases (LDH) were measured over the time to ensure cellular integrity (Fig. S1). Both parameters showed no significant changes in comparison to control conditions. ADRB3 expression was under the detection level in cells from both depots. ADIPOQ, CIDEA, DIO2, ELOVL3, LEP and UCP1 gene expression did not show significant differences independent of the depot origin. However, gene expression of FABP4 was significantly up-regulated in visceral PACs after 6 h of cold stimulation in comparison to subcutaneous PACs (p < 0.05). FGF21 did show a significantly higher expression level (p < 0.05) for visceral PACs after 24 h of CE. Additionally, PPARGC1A and PRDM16 showed a significantly stronger gene expression response for visceral PACs after 24 h of CE in comparison to subcutaneous PACs (p < 0.05). In general, the differentiated PACs of VAT origin showed a stronger response to cold stimulation. As our work indicates that miR-375 itself as well as its thermogenesis-related target genes act primarily in VAT, visceral PACs were used for all following up experiments.

miR-375 knock down in visceral PACs elevates important adipogenesis and thermogenesis genes.
To understand the influence of miR-375 on the differentiation process of adipocytes we first investigated its expression level on day 0, 3, 7, 10 and 14 of differentiation in visceral PACs (Fig. 3A). In parallel, the differentiation state of the cells was tested by oil red staining. During adipogenic differentiation, we observed increasing levels of miR-375 gene expression in in vitro cultured human pre-adipocytes until day 7 followed by a decrease until day 14 (Fig. 3A). Transfection of the cells with siRNA-miR-375 three days before the induction for differentiation revealed a high knock-down efficiency of up to 89% in comparison to the mock transfection (NK) siRNA (Fig. 3B). The knock down remained stable up to d14 of the differentiation process (61-89% efficiency). In parallel with the knockdown of miR-375, we measured the adipogenesis marker genes ADIPOQ and www.nature.com/scientificreports/  www.nature.com/scientificreports/ ADIPOR2. Both genes were significantly up-regulated in fully differentiated VC-PACs on d14 when miR-375 was repressed (p = 0.0286, p = 0.001, Fig. 3C,D). Because of the strong response to cold exposure between 6 and 24 h of visceral PACs in the first experiment (Fig. 2D), we decided to harvest transfected and differentiated VC-PACs after 6 h, 12 h, 24 h and 48 h at 30 °C, respectively. Analyzing gene expression after 12 h of cold exposure revealed significantly higher expression levels of PPARGC1A, ELOVL3, and PRDM16 in VC-PACs (Fig. 3E). Additionally, higher levels of FABP4 and FGF21 were detected but did not reach significance.
Summarizing, cold exposure of volunteers had a suppressing effect on miR-375 in serum. As miR-375 is predominantly found in VAT, this finding implicated a specific function of this miRNA for this adipose tissue depot. Cold stimulation (in vitro) affected gene expression primarily in PACs derived from the visceral depot. Therefore, visceral PACs were used in all follow up experiments to study miR-375 related molecular thermogenesis questions. miR-375 inhibition in visceral PACs induced important thermogenesis genes, indicating that miR-375 is involved in browning/beigeing as well as adipogenic processes in cells of visceral origin (Fig. 3F). The up-regulation of important adipogenesis and thermogenesis genes raises the possibility that suppression of miR-375 is related to cell fate processes and supports thermogenesis.

Discussion
Cold-induced non-shivering thermogenesis received growing interest, as studies in humans suggested that the underlying activation of BAT could play a significant role in energy homeostasis 13 . This study aimed at identifying miRNAs that are modulated by cold exposure. Out of the seven miRNAs screened, miR-375 expression was negatively associated with the cold response in the participants. Since their discovery in 1993 36 , miRNAs have been widely studied concerning their potential use for diagnostic and prognostic purposes. Specific circulating miRNA profiles have been reported to play a role in various diseases including cancer 37 , osteoporosis, skeletalassociated diseases 38,39 , coronary heart disease 40,41 , and metabolic diseases such as obesity 42 and type 2 diabetes 43 . Here, we report for the first time a serum-based miRNA profiling following a defined cold exposure as part of a highly standardized human intervention study.
We detected a downregulation of miR-375 under cold exposure in serum and an inverse correlation to VAT volume, suggesting a strong involvement of this miRNA in visceral adipose tissue homeostasis. These findings were underlined by the high expression of miR-375 in visceral adipose tissue and higher expression in visceral preadipocytes in comparison to subcutaneous PACs. Hu, et al. identified miR-375 as a "marker" to distinguish the different fat depot-derived stem cells and showed that miR-375 was most abundantly expressed in visceral PACs 44 . Our target enrichment analysis provided a specific link between miR-375 and cold-induced thermogenesis, suggesting that miR-375 may have a generic role on diverse cold-induced processes in human adipose tissue. In-silico predicted target genes of the thermogenesis pathway were predominantly regulated in visceral adipose tissue and mainly connected to cells of adipose tissue origin 15,34,45 . VAT is known to have a higher secretory capacity than SAT, and this difference is an intrinsic feature of its cellular components 46 . On a cellular level, cold exposure experiments revealed a stronger reaction of visceral PACs on key genes involved in the miR-375 thermogenesis pathway, including FABP4, FGF21, PPARGC1A and PRDM16.
Compared to subcutaneous adipocytes, visceral adipocytes show higher fatty acid turnover and lipolysis rates. Furthermore, mature visceral adipocytes are less responsive to the antilipolytic effect of insulin 47 . Additionally, recent omics studies suggest a higher content of cells with the potential to perform thermogenesis in VAT compared to SAT 14,15,48 . Both, the greater lipolytic capacity and involvement in browning-related processes indicate a high relevance of VAT for CIT.
To date, some studies have indicated that specific miRNAs are involved in BAT activation upon cold exposure 49,50 . However, these studies mainly focused on identifying miRNAs in one cell type, such as immortalized brown adipocytes or murine mesenchymal stem cells. Also, a regulatory effect of certain miRNAs in specific animal models such as Ercc1 −/− , Dgcr8flox/flox, adiponectin-Cre transgenic mice, or Pparα −/− was described 51 . For example, Oliverio et al. could identify miR-328 as a controller of brown adipose tissue differentiation and BAT function in mice 25 . In human individuals undergoing a mild cold exposure, only exosomal expression changes of miR-92a and miR-122-5p were reported 19,52 . Both groups found negative associations of these exo-miRNAs and BAT activation in mice and humans. In contrast, we analyzed circulating miRNAs because exosomal miRNAs are mainly detected upon malignant conditions. Further, the majority of exo-miRs is not conserved in exosomes but rather bound to argonaut proteins and is therefore not captured 53 . Further, by analyzing exo-miRs, the spectrum of detectable miRNAs is lower 54 . In the present study, miR-375 was the only circulating miRNA that significantly changed after cold exposure in both the exploratory screening and validation cohort. In a study in individuals with or without diabetes, an up-regulation of miR-375 was detected in the diabetic group 55 . miR-375 has been shown to promote adipogenic differentiation by increasing mRNA levels of C/EBP/α and PPARG2 as well as by inducing adipocyte fatty acid-binding protein and triglyceride accumulation 56 . In addition, miR-375 is upregulated during the osteogenic differentiation of human adipose-derived mesenchymal stem cells 57,58 highlighting its impact on mesenchymal cell differentiation. Kraus et al. identified miR-375 as an androgen-regulated miRNA, showing an androgen-mediated inhibition of miR-375 and the associated regulation of ADIPOR2 in differentiating human adipocytes 59 . We could confirm these findings by miR-375 knock out revealing an up-regulation of ADIPOR2 and ADIPOQ in visceral PACs. Cold exposure experiments in visceral PACs revealed significantly higher expression levels of the thermogenesis genes PPARGC1A, ELOVL3, and PRDM16. An up-regulation of PPARGC1A is known to promote thermogenesis in adipose tissue and skeletal muscle 60 . In brown adipocytes, the transcriptional coactivator PPARGC1A is a key regulator of UCP1 expression, mitochondrial biogenesis and oxidative metabolism 61 . ELOVL3 elongates very long chain fatty acid (VLCFA) in adipose tissue. The suggested function in cold-stimulated BAT is to replenish intracellular pools of specific VLCFAs when the fatty www.nature.com/scientificreports/ acid turnover rate is high 62 . The gene PRDM16 is a powerful driver of brown and beige adipocyte identity 63,64 . PRDM16 co-activates PPARγ and PPARα in adipocytes to activate the expression of thermogenic genes, like UCP1. miRNA-133 was found to be markedly downregulated in BAT and SAT of C57Bl/6N mice and negatively regulates PRDM16 as well 65 . In mice exposed to cold, levels of miR-33 in BAT were rapidly downregulated, consistent with a role for miR-33 in repressing adaptive thermogenesis 66 . Mainly, miRNAs have been shown to be potent repressors of brown fat differentiation by targeting different transcriptional coactivators. It is widely accepted that miRNAs have multiple-sometimes hundreds-of targets. This supports our findings of the depletion of miR-375, which might have multi-effects on adipocyte maturation as well as on thermogenic pathways. miR-375 and several other miRNAs have been identified to exert prominent roles in regulating the beiging of inguinal white AT and activation of BAT 67 .
In conclusion, our findings raise interesting questions about the role of miRNA-375 in the beiging/browning process of adipocytes and underlines the need to understand the biological origins and functions of miRNAs in this context. We could show a decreased expression of miR-375 after cold stimulation in humans. Tissue and cell comparison identified a prominent role of miR-375 in the visceral fat depot. Depletion of miR-375 revealed interactions with cell fate and thermogenic response in visceral adipose tissue.

Methods
Study approval. The ethical review committee of the Faculty of Medicine of the Technical University of Munich approved the FREECE (Effect of the FTO-Genotype on Resting Energy Expenditure after defined Cold Exposure) study protocol (project number 236/16) as well as the MOBB study (Munich obesity biobank, project number 5716/13). All participants provided written informed consent before any study procedures. All procedures were conducted according to the principles of the Declaration of Helsinki. The demographic data of the included individuals are presented in the supplemental material (Tables S1, S2 and S3).
Short-term cold exposure. All participants underwent the procedure in the morning and in the fasted state. First, anthropometric measurements were taken and body composition was assessed through bio impedance analysis (TANITA Body Composition Analyzer Type BC-418 MA, Tanita Europe GmbH, Sindelfingen, Germany). Blood pressure and heart rate were measured and skin temperature was recorded using iButtons (Thermochron, Wisconsin, United States). Eight sensors were placed on defined sites of the body to assess the overall skin temperature according to ISO 988620 68 . One additional iButton was placed at the site of the supraclavicular area. The participants were placed in a supine position and allowed to rest for 15 min, followed by 30 min of measurement of resting energy expenditure by indirect calorimetry (Cosmed Quark RMR 1.0, Fridolfing, Germany). Blood was drawn from each participant, followed by the subsequent non-shivering cold exposure over two hours using water-containing thermic blankets (Maxi-Therm Lite, Cincinnati Sub-Zero Products, LLC). In the last 30 min of cold exposure, resting metabolic rate was measured again. Thereafter, the second blood drawing took place. A detailed description of the cooling protocol was recently published 31 .
Blood sample analysis. Serum samples were analyzed by a certified laboratory (Synlab, Munich, Germany) for free triiodothyronine (T 3 ), C-reactive protein, and triglycerides. Non-esterified fatty acids (Wako Chemicals, Neuss, Germany), insulin (DRG Instruments, Marburg, Germany), leptin, and total adiponectin (R&D, Wiesbaden, Germany) were analyzed in plasma using commercially available ELISA kits according to the instructions of the manufacturers.
GTEx and single cell data. Violin plots displaying log10 transcripts per million + 1 (TPM + 1) of miR-375 related genes were generated using the Genotype-Tissue Expression (GTEx) project's multi gene query function (Release V8) 33,34 . Dotplots of white adipose tissue single-cell RNA expression were retrieved from the Single Cell Portal (study no. SCP1376) 15 .
Isolation, culture and differentiation of human preadipocytes (PAC). All methods for the isolation, culture and differentiation of human PACs derived from adipose tissue material are described in detail, elsewhere 69,70 . Briefly, PACs were isolated from adipose tissue material based on collagenase-digestion, grown until confluency in T-75 flasks (Falcon, Corning, NY) and cryopreserved. For differentiation experiments cells were thawed, grown until confluency in T-25 flasks (Falcon) and splitted on to 6-well plates (Falcon). At confluency, differentiation was induced and cells were cultured for 14 days to allow for differentiation and accumulation of lipids. Cells were cold stimulated up to 48 h at 30 °C. H&E and Oil red O staining, LDH activity measurement and RNA isolation were performed using established methods, respectively. www.nature.com/scientificreports/ tions of the manufacturer (Qiagen, Hilden, Germany). RNA was precipitated with 900µI ethanol, triple washed with wash solution, followed by RNA elution in 20 μl nuclease-free water and stored at − 80 °C. Regarding cells, total RNA including microRNAs was isolated at several time points using TRIzol™ Reagent (Thermo Fisher Scientific, Massachusetts, USA). The amount and integrity of isolated miRNA was evaluated via gel electrophoresis using a Bioanalyzer device (Small Chip, Agilent Technologies, California, USA). Serum samples were analyzed regarding their hemolytic status. Therefore, miR-451, a miRNA highly abundant in red blood cells was used. The concentration of miR-451 was significantly lower in all samples compared to our internal positive control (Cutoff Cq value = 15). qPCR analysis. The isolated miRNA as well as the spike-in control UniSp6 to assess enzyme efficiency were reversely transcribed into cDNA using the miRCURY LNA RT Kit (Qiagen, Hilden, Germany). The reaction was incubated at 42 °C for 60 min and then heat-inactivated at 95 °C for 5 min. 20-fold diluted cDNA samples were stored at − 20 °C. Real-Time quantitative PCR (RT-qPCR) was carried out using custom 384 well panels (4titude, Wotton, UK). For RT-qPCR analysis, each reaction contained 4 µl of cDNA and 6 µl of master mix using miR-CURY Sybr Green Kit and LNA-enhanced miRNA primer assays (Qiagen, Hilden, Germany). PCR conditions were 95 °C for 2 min, 50 cycles of denaturation (95 °C, 10 s) and annealing (56 °C, 60 s), and a melting curve analysis to complete the run on a LC480 Real-Time PCR system (Roche, Basel, Switzerland). SNORD38b and SNORD48b served as internal control for tissue and cells, let-7i-5p and miR-30e-5p for serum. For analyzing the expression levels of miRNA-375 related genes, cDNA was synthesized from 1 mg total RNA using High capacity cDNA RT Kit (Applied Biosystems, Germany). RT-PCR was performed with 20 ng cDNA with the Maxima SYBR Green Master Mix (ThermoScientific, Germany) using specific primer pairs (Table S6). For normalization, GAPDH and IPO8 were used. To calculate the C q -values, the second derivative method was used. miRNA array analysis. We profiled miRNA spectra from two pooled serum groups, including one pool of 12 males and another of 12 female samples, to identify regulated miRNAs during cold-induced thermogenesis. In total, 169 different miRNAs were profiled by the human miRCURY LNA miRNA Focus PCR Panels YAHS-106YG-2 (Qiagen, Hilden, Germany). The cycle number via RT-qPCR determined the expression levels. Levels were normalized to the internal reference genes using the 2 −∆∆Ct method 71 . Afterwards, the fold-change before and after cold exposure regarding the expression of specific miRNAs was calculated.
Cell staining. Cells were fixed with 4% paraformaldehyde for 20 min at room temperature. After one washing step with PBS, Oil Red O solution was added for 60 min and rinsed off with PBS until clear. For a better contrast, cells were counterstained with HE. After one washing step with PBS, hematoxylin solution (VWR, Heamalum (Mayer´s) Gurr for microscopy) was added for 4 min and rinsed off with PBS until clear. Eosin solution (2% in acetic EtOH) were added for 2 min. After rinsing with PBS stained cells were assessed under a light microscope (VHX series, Keyence, Osaka, Japan).

Measurement of lactate dehydrogenase activity.
Lactate dehydrogenase (LDH) release in the incubation medium was measured as an index for cytotoxicity. As control stimulated medium at 30 °C for up to 48 h were used. Aliquots of the incubation medium were added to an assay mixture containing 81 nM Tris, 203 nM NaCl, and 0.24 nM NADH, pH 7.2. The reactions were started by the addition of 1.55 mM pyruvate, and the conversion of NADH into NAD + was measured spectrophotometrically at 340 nm at 30 °C for 30 min. Results were compared to the control (medium only).

Bioinformatics predictions.
To identify human miR-375-regulated genes, we first employed the miRNA search tool MultiMiR in R, compiling nearly 50 million records from 14 different databases, including miRTar-Base, TarBase, miRanda, miRDB and Targetscan 72,73 . We then investigated the functional interaction between the lists of predicted miR-375 target genes with a comprehensive list of genes associated with cold exposure (gene ontology term 0106106). To generate a map of miR-375 target genes involved in CIT the String DB tool and the program cytoscape were used 74 . Afterwards we filtered genes, which were direct targets of UCP1, as this is the central gene of the thermogenesis pathway.
Statistical analyses. Results are given (after age and BMI adjustment as well as logarithmic transformation) as violin plots, mean with standard error of the mean (± SEM). All data were non-Gaussian distributed (Kolmogorov-Smirnov test P < 0.05). Two-tailed unpaired Mann Whitney test were used to determine the significance of group-wise differences. GraphPad Prism version 9 was used (Graph Pad Software, San Diego, USA). P-values < 0.05 were considered significant. www.nature.com/scientificreports/