Because of their known anti-inflammatory effects in the vascular endothelium (Diamantis et al., 2017; Mangoni et al., 2017), we explored whether MTX or HMG-CoA reductase inhibitors (statins) activated the expression of the anti-inflammatory MIR181B (Lin et al., 2016; Sun et al., 2014a; MICU Registry et al., 2012; Sun et al., 2016). MTX (10 µM) but not statins significantly upregulated MIR181B expression by 1.35-fold in cultured human umbilical vein endothelial cells (HUVECs) (Figure 1A). The MTX-induced expression of MIR181B was dose- and time-dependent with the highest induction at 10 µM and 4 hr (Figure 1B,C). Previous studies have established that adenosine (Ad) is the major anti-inflammatory effector of MTX, which is a product of adenosine monophosphate, a reaction catalyzed by the enzyme ecto-5’ nucleotidase (Chan and Cronstein, 2013; Gadangi et al., 1996; Tian and Cronstein, 2007). When using a competitive enzyme inhibitor of ecto-5’ nucleotidase (α, β-methylene adenosine-5’-diphosphate (APCP)), the MTX-induced MIR181B expression was completely blocked (Figure 1D). Conversely, Ad alone increased MIR181B expression in a dose- and time-dependent manner with the highest observed increase by 2.3-fold at 50 µM and 4 hr (Figure 1E,F).

MIR181B expression is rapidly reduced in response to pro-inflammatory stimuli such as TNF-α in ECs and overexpression MIR181B suppressed TNF-α-induced EC activation by repressing VCAM-1, E-selectin, and ICAM-1 expression (MICU Registry et al., 2012; Yamasaki et al., 2003). To address whether MTX or Ad may rescue the TNF-α-mediated reduction in MIR181B expression, HUVECs were stimulated with TNF-α in combination with either MTX or Ad. Treatment with MTX or Ad fully rescued the TNF-α repression of MIR181B expression (Figure 1G). As expected, ECs treated with MTX or Ad significantly reduced VCAM-1, ICAM-1, and E-selectin expression stimulated by TNF-α (Figure 1—figure supplement 1A,B). Because two loci for MIR181B (i.e. MIR181B-1 and MIR181B-2) exist (Sun et al., 2014b), we examined which locus is regulated by TNF-α, MTX, or Ad. To this end, primary miRNA transcripts were analyzed in response to stimulation with TNF-α. Only primary transcript of the MIR181B-2 locus, and not primary MIR181B-1, MiR-181A-1, or MiR-181A-2, was rescued by MTX or Ad co-stimulation (Figure 1H,I; Figure 1—figure supplement 1C,D).

To study the relationship between MTX or Ad to Mir181b in mouse tissues, primary lung ECs were isolated from Mir181a2b2^—/— knockout mice (KO), reflecting deficiency of Mir181a-2 and Mir181b-2, or control Mir181a2b2^flox/flox (flox). Because Mir181a-2 and Mir181b-2 are located in close proximity to each other, the knockout strategy deleted both miRNAs (Henao-Mejia et al., 2013). Isolated primary ECs were stimulated with TNF-α in the presence or absence of MTX or Ad. In ECs isolated from wild-type flox mice, MTX or Ad inhibited VCAM-1 protein expression by 40% and 35%, respectively. However, MTX or Ad did not reduce VCAM-1 expression in primary ECs from Mir181a2b2^—/— KO mice (Figure 1J). Taken together, these findings suggest that pri-Mir181b-2 and its mature isoform Mir181b is increased by MTX or Ad, and may mediate the protective anti-inflammatory effects of MTX or Ad on TNF-α-induced EC activation.

Ad is released by cells to the extracellular environment at a low concertation and acts as a local modulator with a generally cytoprotective function through interaction with its four cell surface Ad receptors, ADORA1, ADORA2A, ADORA2B, and ADORA3. Each of these Ad receptors have a unique pharmacological profile, tissue distribution, and effector coupling (Haskó and Cronstein, 2004; Jacobson and Gao, 2006). To examine which Ad receptor mediates the MTX-induced MIR181B expression, individual knockdown for each of the Ad receptors was performed in ECs first in the absence of MTX or Ad (Figure 2—figure supplement 1A–H). Only silencing of ADORA3 reduced MIR181B basal expression by 40%, while knockdown of ADORA1, ADORA2A, and ADORA2B had no effect on MIR181B expression (Figure 2A). This suggests that ADORA3 might mediate the MIR181B expression activated by MTX or Ad. Therefore, we next examined whether AdoRA₃ knockdown blocked the MTX or Ad-mediated increase in MIR181B expression. Indeed, neither MTX nor Ad increased MIR181B expression in ADORA3-deficient HUVECs compared to negative control siRNA (Figure 2B). MIR181B expression is reduced in response to pro-inflammatory stimuli such as TNF-α (Lin et al., 2016; Sun et al., 2014a; MICU Registry et al., 2012; Sun et al., 2016). Treatment of ECs with TNF-α in the presence of MTX or Ad rescued the TNF-α-mediated repression of MIR181B expression in an AdoRA₃-specific manner (Figure 2C and Figure 2—figure supplement 1A). Furthermore, silencing of ADORA3, (and not ADORA1, ADORA2A, or ADORA2B) abrogated the anti-inflammatory effect of MTX or Ad in TNF-α-activated ECs on VCAM-1, ICAM-1, and E-Selection expression (Figure 2D and Figure 2—figure supplement 1B). Finally, to further assess whether the anti-inflammatory effect of MTX or Ad is mediated through an ADORA3-MIR181B signaling cascade, knockdown of ADORA3 was combined with simultaneous overexpression of MIR181B mimic. While the MTX or Ad anti-inflammatory effects were blocked in the presence of ADORA3 silencing, overexpression of MIR181B fully rescued the MTX and Ad-mediated reduction of ECs activation markers (Figure 2E). In summary, our data indicate that ADOR3 plays central role in mediating MTX- or Ad-induced MIR181B expression, and this ADORA3-MIR181B signaling pathway contributes to the inhibitory effects of MTX or Ad on TNF-α-induced EC activation.

To address, how MTX regulates MIR181B-2 expression on the transcriptional level, different lengths of the proximal MIR181A2B2 promoter (containing sequences 5’ to both MIR181A-2 and MIR181B-2) were cloned upstream of a luciferase reporter. MTX and Ad activated the −606 bp MIR181A2B2 promoter reporter construct from 1.47- to 1.57-fold, respectively (Figure 3A–B). Treatment of MTX or Ad similarly activated the −402 bp reporter comparable to the full-length constructs (Figure 3B). However, reducing the promoter length to −301 bp or −150 bp significantly reduced luciferase activity from 30- to 2600-fold, respectively (Figure 3B). At the post-transcriptional level, mRNA stability is an important factor determining mRNA abundance. To investigate whether MTX or Ad would affect pri-MIR181B-1 or pri- MIR181B-2 transcript stability, we used the transcription inhibitor actinomycin D in MTX or Ad-treated HUVECs. We found that neither MTX nor Ad affected the stability of pri-MIR181B-1 transcript (Figure 3—figure supplement 2A). However, MTX significantly increased pri-MIR181B-2 transcript stability, while Ad had a modest opposite effect (Figure 3—figure supplement 2B). These results indicate that the promoter region between −402 and −301 is likely required for MTX and Ad-induced transcriptional activity of the MIR181A2B2 locus and MTX and Ad affects pri-MIR181B-2 expression at both transcriptional and post-transcriptional levels. To further identify specific transcription factor binding sites, in silico prediction tools identified a SMAD2/3/4 binding site in the region −402 to −301 (Figure 3B). This was further investigated by individual knockdown of SMAD2, SMAD3, or SMAD4 (Figure 3—figure supplement 3). Silencing of SMAD3 or SMAD4 reduced luciferase activity of the full length −606 bp construct by 37% and 42%, respectively, in ECs compared to control siRNA (Figure 3C). However, SMAD2 silencing did not impact luciferase activity (Figure 3C). Consistent with this observation, MIR181B expression was significantly reduced only by silencing SMAD3 or SMAD4, and not by SMAD2 knockdown (Figure 3D). Importantly, silencing of SMAD3 or SMAD4 completely blocked the MTX and Ad-mediated increase in MIR181B expression (Figure 3E). Furthermore, knockdown ADORA3 blocked the MIR181A2B2 promoter luciferase activity induced by MTX and Ad (Figure 3F); however, silencing of the other three Ad receptors had no effect on luciferase activity (Figure 3G). Taken together, these data indicate that SMAD3 and SMAD4 are not only required for transcriptional activity of MIR181B expression, but also significantly contribute to MTX and Ad-induced MIR181B expression.

Previous reports suggested that MTX can ameliorate diet-induced insulin resistance and inflammation (DeOliveira et al., 2012; Myers et al., 2017), although the mechanisms remained poorly understood. To evaluate the effective dose of MTX on Mir181b expression in vivo, C57Bl6 mice were injected with 0.5 mg/kg or 1.0 mg/kg of MTX. Low-dose (i.e. 0.25 mg/kg) MTX showed no significant increase in Mir181b expression in plasma, liver, and intima, but 1 mg/kg did (Figure 4—figure supplement 1A). Moreover, C57Bl6 mice treated with adenosine exhibited significantly elevated Mir181b expression in plasma by 168% and was even more pronounced than high-dose MTX (Figure 4—figure supplement 1B). In accordance with this dose-response of MTX on Mir181b expression, we examined whether the effect of MTX (1 mg/kg) on insulin resistance and adipose tissue inflammation is dependent on Mir181b2 expression in vivo. To this end Mir181a2b2^flox/flox (flox) and Mir181a2b2^—/— (KO) mice were placed on a 60% high-fat diet (HFD) for 12 weeks. Both flox and KO mice were injected with vehicle or MTX for 12 weeks (1 mg/kg/week by intraperitoneal injection), and insulin tolerance testing (ITT) and glucose tolerance testing (GTT) were performed (Figure 4A). As shown in Figure 4B, the body weights among the four groups were not significantly different, which suggest that the gain of body weight was independent of Mir181a2b2 expression and MTX treatment. However, MTX treatment significantly improved insulin sensitivity and reduced the area under curves (AUC) for ITT by 23% (Figure 4C) compared with vehicle injected flox mice, but not in glucose tolerance (Figure 4D). In line with our previous report showing a protective effect of Mir181b mimics on insulin resistance (Sun et al., 2016), KO mice developed insulin resistance (IR) as shown by improved insulin sensitivity (increased AUC for ITT by 121%) and glucose tolerance (increased AUC GTT by 122%) (Figure 4C,D). Notably, the beneficial effect of MTX on insulin sensitivity was completely blocked in KO mice (Figure 4C,D). Moreover, flox mice treated with MTX had increased insulin signaling activity in eWAT and liver compared vehicle injected, indicated by an increase in phosphorylation of Akt (phospho-Akt) in eWAT (by 23%) and in liver (by 34%) (Figure 4E). In contrast, KO mice without MTX injections showed a significant reduction of Akt phosphorylation by 24% in eWAT and 44% in liver compared with control flox mice (Figure 4E). Moreover, the MTX-mediated upregulation of phospho-Akt in eWAT and liver was blocked in KO mice (Figure 4E). However, MTX did not increase phospho-Akt levels in skeletal muscle (SM) (Figure 4E). Taken together, these findings from systemic miR-181a2b2 KO mice suggest that MTX increases Akt phosphorylation in eWAT and liver through miR-181a2b2 expression, which contributes to improved insulin sensitivity.

To better characterize the underlying mechanisms how MTX and Mir181a2b2 affects insulin resistance, we performed next-generation sequencing of RNA isolated from eWAT of these HFD-fed Mir181a2b2 flox and KO mice with or without MTX treatment. Gene set enrichment analysis (GSEA) suggested that the most statistically significant [false discovery rate (FDR) < 0.5] altered genes in eWAT from Mir181a2b2 KO mice were linked to cell adhesion (e.g. platelet-endothelium-leukocyte interactions, cell-matrix interactions, and cell junctions), chemotaxis, and immune response (Figure 4—figure supplement 2A-B). In eWAT from MTX-treated Mir181a2b2 flox and KO mice, GSEA found that cell cycle, inflammation, and chemotaxis pathways were more affected by MTX (Figure 4—figure supplement 2C-D). Macrophage accumulation plays a critical role in obesity-associated IR (Chawla et al., 2011; McNelis and Olefsky, 2014) and overexpression of Mir181b did not directly regulate cell-intrinsic monocyte/macrophage migration, proliferation, and activation (Sun et al., 2016). Moreover, we confirmed that MTX administration reduced VCAM-1 mRNA expression by 36% in eWAT in flox mice, but not MTX-injected KO mice (Figure 4F,G). In line with reduced VCAM-1 expression, we observed that macrophage content, as indicated by Mac2 staining, was significantly reduced by 80% in eWAT in MTX-injected flox mice (Figure 4H). In contrast, KO mice treated with MTX showed no inhibitory effect on VCAM-1 mRNA expression and macrophage content in eWAT (Figure 4F,G). Taken together, these results suggest that MTX administration leads to reduced EC activation and macrophage accumulation in white adipose tissue, largely dependent on Mir181a2b2 expression.

It is well-established that some miRNAs may have a cell-type and/or tissue-specific expression and function which can impact the regulation of gene networks (Gao et al., 2011; Sood et al., 2006; Thomou et al., 2017; Zhou et al., 2013). Mir181b plays a dominant role in the vascular endothelium (Lin et al., 2016; Sun et al., 2014a; MICU Registry et al., 2012; Sun et al., 2016); however, its EC-specific genetic deletion has not been previously described. To assess whether endothelial Mir181a2b2 is mediating the observed phenotypes with respect to MTX and insulin sensitivity, we generated inducible EC-specific Mir181a2b2 mice (iEC-KO) by breeding the Mir181a2b2^flox/flox mice to tamoxifen-inducible endothelial-specific Cre (vascular endothelial cadherin promoter [VECad-Cre-ER^T2]) (Figure 5A). Compared to vehicle control injected mice, Mir181b expression in tamoxifen injected iEC-KO mice was reduced by 60% and 80% in aortic intima and ECs isolated from SM, respectively, but not in the aortic media and non-EC fraction of SM (Figure 5—figure supplement 1B–E). To assess whether endothelial Mir181a2b2 is driving the diet-induced IR and mediates MTX’s protective effects, we first examined MTX regulation of Mir181b expression. While Mir181b was reduced in the aortic intima of iEC-KO mice compared to flox-Cre, MTX increased Mir181b in flox-Cre, but not in iEC-KO mice (Figure 5—figure supplement 2A). Consistent with upregulated Mir181b expression, MTX reduced VCAM-1 expression in the intima of flox-Cre mice, but not in iEC-KO mice (Figure 5—figure supplement 2B). Because we identified that SMAD3/4 signaling contributed largely to Mir181b2 expression (Figure 3), we next investigated expression of SMADs in eWAT tissues from these HFD-fed flox-Cre and iEC-KO mice. We found that SMAD2 and SMAD3 expression decreased in MTX-treated iEC-KO mice compared to flox-Cre mice, whereas there were no differences between flox-Cre and iEC-KO mice without MTX treatment, suggesting a feedback mechanism involved in the MTX-SMAD-Mir181b cascade (Figure 5—figure supplement 3).

As shown in Figure 5B, there were no significant differences in body weights between iEC-KO and control mice treated with or without MTX after 12 weeks of HFD. However, MTX administration significantly improved insulin sensitivity by 23% in ITT of control flox-Cre mice (Figure 5C), whereas MTX had no effect on insulin sensitivity in iEC-KO mice (Figure 5C). Similar to Mir181a2b2 systemic KO and control mice, MTX administration did not improve glucose tolerance either in iEC-KO or control mice (Figure 5D). Moreover, the MTX-mediated induction of Akt phosphorylation in eWATs and livers was blocked in iEC-KO mice (Figure 5E). Furthermore, MTX reduced VCAM-1 expression in eWAT in flox-Cre control mice, but not in iEC-KO mice (Figure 5F). Consistent with previous results (Figure 4H), MTX administration significantly reduced Mac2 positive macrophage content in eWAT tissues in flox-Cre mice, whereas no effect was observed in iEC-KO mice (Figure 5G). Finally, iEC-KO mice without MTX treatment showed a significant increase in Mac2 positive macrophages compared to flox-Cre control mice in eWAT tissues (Figure 5G). Taken together, these results identify a crucial role of endothelial Mir181a2b2 for mediating MTX’s anti-inflammatory effects.

HUVECs were purchased from Lonza (cc-2159) and cultured in EC growth medium EGM-2 (cc-3162). HEK 239 T cells (CRL-3216, ATCC) were cultured in DMEM supplemented with 10% FBS and 1% penicillin and streptomycin. Cells were passaged less than five times for all experiments. Cell types were authenticated as follows: HUVECs, morphology; HEK293T, cell size and efficiency in transfection reporter assays. All cell lines were tested mycoplasma-free. Primary lung ECs from mice were isolated and cultured in as described (Sun et al., 2016). Briefly, isolated lungs from miR-181a2b2^flox/flox and miR-181a2b2^-/- mice were grinded with scissors (1 × 2 mm²) and digested with collagenase type II (Worthington) and dispase (Roche) (1 mg/ml each in DMEM/F12). Digestion was neutralized with DMEM/F12 medium containing 10% FBS, followed by centrifugation at 500 g for 10 min at 4°C and re-suspend in incubation buffer (PBS pH 7.2, 0.1% BSA, 2 mM EDTA, 0.5% FBS). ECs were captured using 10 µl anti-PECAM-1 (1:12.5, 557355, BD Pharmingen) antibody-conjugated sheep anti-rat IgG Dynabeads (00412289, Invitrogen) mixed with buffer containing the re-suspended cell pellet and incubated for 20 min at 4°C. After incubation, Dynabeads were mounted using the Magnetic Separation Rack and washed five times using wash buffer (PBS pH 7.2, 0.1% BSA) and cultured in gelatin (0.1%) pre-coated plates in EC medium (M1168, Cell Biologics) with kit (M1168-kit, Cell Biologics). EC identity was verified by both morphology and expression for CD31 by flow cytometry. HUVECs or primary lung ECs were plated on 12-well plates at 60,000/well or 150,000/well and allowed to grow to 80–90% confluency under growing conditions or 70–80% for transfection. Cells were pretreated with MTX (0.1, 1, 10 or 100 µM, M8407, Sigma), adenosine (1, 5, 10, 50 or 100 µM, A4036, Sigma), atorvastatin (1 µM, 1044516, Sigma), mevastatin (1 µM, M2537, Sigma), simvastatin (1 µM, S6169, Sigma), or rosuvastatin (1 µM, SML1264, Sigma) for 3 hr and stimulated with or without 10 ng/ml of recombinant human TNF-α (210-TA/CF, R and D Systems) or 20 ng/ml recombinant mouse TNF-α (410-MT/CF, R and D Systems) for various times, according to the experiment: Western blot, 8 hr; real-time qPCR, 1, 4, 7, or 12 hr. Lipofectamine 2000 transfection reagent (11668019, Invitrogen) was used for transfection, following the manufacturer’s instructions. On-Target plus human control siRNA-Smart poll, On-Target plus human AdoRA1 siRNA-Smart poll (L-005415–00, Dharmacon), On-Target plus human AdoRA2A siRNA-Smart poll (L-005416–00, Dharmacon), On-Target plus human AdoRA2B siRNA-Smart poll (L-005417–00, Dharmacon) or On-Target plus human ADORA3 siRNA-Smart poll (L-005418–00, Dharmacon) were transfected at 100 nM. SiRNA control (AM4636) and validated siRNA including SMAD2 (VHS41106), SMAD3 (VHS41111), SMAD4 (s8403) from Invitrogen were transfected at 20 nM. For co-transfection studies, On-Target plus human control siRNA-smart poll (100 nM), On-Target plus human ADORA3 siRNA-smart poll (100 nM), miRNA negative control (10 nM, AM17110, Ambion), or pre-Mir181b (10 nM, PM12442, Ambion) were transfected as indicated in respective experiments. HUVECs were treated with 10 ng/ml recombinant human TNF-α 36 hr post-transfection at indicated times points.

HEK 293 T cells were plated on 12-well plates at 300,000/well up to 70–80% confluency and transfected with 500 ng of the indicated Mir181a2b2 promoter reporter constructs (Supplementary file 1) and 10 ng Renilla plasmid (E2231, Promega). Adenosine receptor siRNAs and SMAD2/3/4 siRNAs were co-transfected at 100 or 20 nM final concentration and cells were treated with MTX (10 µM) or Adenosine (50 µM) 24 hr post-transfection for 12 hr. Transfected cells were collected in 200 µl reporter lysis buffer (E1910, Promega) and luciferase activity was measured using a Dual-Luciferase reporter assay system (E1910, Promega). Each reading of luciferase activity was normalized to the Renilla activity.

For mRNA stability assays, HUVECs were incubated with 10 µM MTX or 50 µM Ad with 10 µg/ml Actinomycin D (SBR00013, Sigma) to inhibit transcription. At the indicated time points after the addition of Actinomycin D, cells were harvested and total RNA was extracted. The expression levels of pri-Mir181b1 and pri-Mir181b2 at each time point were measured by real-time qPCR as described below and normalized to the according Hprt levels. The remaining mRNA was determined by comparison with the expression level of the relevant gene at the zero time point (designated 100%) when Actinomycin D was added.

Mir181a2b2^flox/flox and Mir181a2b2^—/— mice were used as previously described (Henao-Mejia et al., 2013) Mice expressing tamoxifen-inducible endothelial-specific VE-cadherin (VECad-Cre-ER^T2; also known as Cdh5(PAC)-CreER^T2) was kindly provided by R. Adams (Wang et al., 2010). C57BL/6 mice were purchased from Charles River Laboratories (#027). Inducible EC-specific Mir181a2b2-deficient mice (Mir181a2b2 ^flox/flox; VECad-Cre-ER^T2) was generated by crossbreeding Mir181a2b2^flox/flox mice and VECad-Cre-ER^T2 mice. For induction of Cre activity, four weeks old male Mir181a2b2^flox/flox mice carrying the VECad-Cre-ER^T2 transgene were treated with either 4-hydroxytamoxifen (H6278, Sigma) (10 mg/kg, i.p.) or the same volume of vehicle for 5 consecutive days to generate EC-specific Mir181a2b2-deficient mice (Mir181a2b2^iECKO) and flox-Cre control mice (Mir181a2b2^ECfl/fl). For measurement of Mir181b in plasma, 8 weeks male C57BL/6 mice were tail vein injected with MTX (0.25 mg/kg or 1 mg/kg) or adenosine (4 mg/kg) for 3 hr and blood was collected in EDTA-containing tubes. Plasma samples were generated by centrifugation at 3000 rpm for 10 min and stored at −80°C. Total RNA was isolated from plasma by using the Total RNA Purification Kit from Norgen Biotek Corporation as described in our previous studies (Sun et al., 2014a; MICU Registry et al., 2012; Sun et al., 2016). Male mice were age-matched in all experiments and cage-matched littermates were used for experiments. Analyses of in vivo samples were performed by blinded observers. All mice were maintained under SPF conditions at an American Association for the Accreditation of Laboratory Animal Care-accredited animal facility at the Brigham and Women’s Hospital. All animal protocols were approved by the Institutional Animal Care and Use Committee at Harvard Medical School, Boston, MA and conducted in accordance with the National Institutes of Health Guide for Care and Use of Laboratory Animals.

Mir181a2b2^flox/flox, Mir181a2b2^—/—, Mir181a2b2^ECfl/fl and Mir181a2b2^iECKO mice were injected i.p. with either MTX (M8407, Sigma, 1 mg/kg/week) or the same volume of vehicle control and placed on a high-fat diet (HFD) for 12–15 weeks containing 60 kcal% fat (D12492, Research Diets). Insulin tolerance testing (ITT) and glucose tolerance testing (GTT) were performed after HFD treatment as described in our previous study (Sun et al., 2016). Briefly, for GTT, mice were fasted for 12 hr and then injected i.p. with D-glucose (G7201, Sigma, 1 g/kg). ITT was performed on mice after 6 hr fasting and injected i.p. with recombinant human regular insulin (0.75 U/kg). Blood glucose levels were measured before injection and at 15, 30, 60, 90, and 120 min after glucose or insulin injection using One Touch Ultra glucometer (LifeScan).

IHC was performed as described in our previous study (Sun et al., 2016). Briefly, tissues were fixed with neutral buffered 10% formalin solution (HT501128, Sigma), embedded in paraffin wax, cut into sections at 6 μm, and deparaffinized. Antigen retrieval was performed using Bond ER1 (AR9961, Leica) for 30 min. Sections were incubated with anti-Mac-2 (553322, BD Pharmingen) in 1:900 dilutions for 90 min at room temperature. Primary antibodies binding to tissue sections was visualized using Bond Polymer Refine Detection kit (DS9800, Leica), and counterstained with hematoxylin. Representative images were captured by a digital system, and the Mac-2-postive areas in each section were determined by detecting the staining intensity with computer-assisted image analysis software (Image-Pro Plus, Media Cybernetics) and data were presented as a ratio of positive area to tissue area. Data were analyzed in a blinded fashion by two independent observers.

Isolation of intima RNA from aorta was performed as described in our previous studies (Sun et al., 2014a; Sun et al., 2016). In brief, aortas were carefully flushed with PBS, followed by intima peeling using TRIzol reagent (15596018, Invitrogen). TRIzol was flushed for 10 s −10 s pause - another 10 s flushed and collected in an Eppendorf tube (~300–400 µL total) and snap frozen in liquid nitrogen.

Tissues were homogenized using TissueLyser II (Qiagen) according to manufacturer’s instructions. For RNA isolation TRIzol reagent (Invitrogen) or RNeasy kit (Qiagen) was used based on manufacturers protocol. Subsequent RT-qPCR was performed using High-Capacity cDNA Reverse Transcription kit (4368813, Applied Biosystems). For SyberGreen-based assay GoTaq qPCR Master Mix (A6001, Promega) was used; and for TaqMan Universal Master Mix II, UNG (4440038, Life Technologies) was used. Expression of mRNAs and miRNA expression were normalized to Hprt, β-actin, or U6 (Aglient, AriaMx Real-Time PCR System). Specific primers including Mir181b-5p (#001098), U6 (#001973), human-primary-miR-181a1b1 (#Hs03302966_pri), human-primary-miR-181a2 (#Hs03302889_pri), human-primary-Mir181b1 (#Hs03302963_pri), for TaqMan system were purchased from Life Technologies. Changes in expression were calculated using deltadelta Ct method. Primer sequences are described in (Supplementary file 2).

TRIzol reagent was used for RNA isolation from eWAT tissues based on manufacturers’ protocol. RNA-Seq analysis was performed after ribodepletion and standard library construction using Illumina HiSeq2500 V4 2 × 100 PE (Genewiz, South Plainfield, NJ). All samples were processed using an RNA-seq pipeline implemented in the bcbio-nextgen project (https://bcbio-nextgen.readthedocs.org/en/latest/). Raw reads were examined for quality issues using FastQC (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/) to ensure library generation and sequencing were suitable for further analysis. Trimmed reads were aligned to UCSC build mm10 of the Mouse genome, augmented with transcript information from Ensembl release 79 using STAR. Alignments were checked for evenness of coverage, rRNA content, genomic context of alignments (for example, alignments in known transcripts and introns), complexity, and other quality checks using a combination of FastQC, Qualimap. Counts of reads aligning to known genes were generated by featureCounts (Liao et al., 2014). Differential expression at the gene level were called with DESeq2. The total gene hit counts and CPM values were calculated for each gene and for downstream differential expression analysis between specified groups was performed using DESeq2 and an adapted DESeq2 algorithm, which excludes overlapping reads. Genes with adjusted FDR < 0.1 and log2 fold-change (1.5) were called as differentially expressed genes for each comparison. Mean quality score of all samples was 36.12 within a range of 22,000,000–31,000,000 reads per sample. All samples had at least >70% of mapped fragments over total. MetaCore (v20.2) was used for gene set enrichment analysis. RNA-Seq data are available through the Gene Expression Omnibus (GSE164251).

Tissues were homogenized using TissueLyser II (Qiagen) according to manufacturer’ instructions. Proteins were isolated using RIPA buffer (Boston BioProdcuts, BP-115) with protease inhibitor and phosphatase inhibitors. Protein concentrations were determined using Pierce BCA assay (Thermo Scientific). Twenty μg protein were loaded per lane on a 4–20% Mini-PROTEAN TGX Gel (Bio-Rad, 456–1096). Separated proteins were transferred to PVDF membranes using the Transfer Turbo Blot system (Bio-Rad) and Trans-Blot Turbo RTA Transfer Kit (Bio-Rad, 170–4272). The membrane was blocked with 5% nonfat milk in TBST for 1 hr at room temperature. After blocking, the membrane was incubated overnight at 4°C with antibodies against ADORA1 (1:500, ab82477, abcam), ADORA2A (1:750, ab3461, abcam), ADORA2B (1:750, ab40002, abcam), ADORA3 (1:500, ab203298, abcam), VCAM-1 (1:1000, ab134047, abcam), ICAM-1 (1:3000, BBA3, R and D Systems), E-Selectin (1:1000, BBA16, R and D Systems), p-AKT (1:1000, 4060, Cell Signaling), T-AKT (1:1000, 2920, Cell Signaling), β-actin (1:4000, 4970, Cell Signaling). Quantification of protein bands were performed using a luminescent image analyzer (BioRad, Chemidoc).

Statistical analyses were performed using GraphPad Prism version 7.0 (GraphPad Software, Inc, CA). Unpaired two-tailed Student t test was used to determine statistical significance between two groups. Multiple groups were analyzed by using One-way ANOVA. All data are presented as mean ± SEM. Number of experiment repeats, biological replicates and p values are indicated in figure legends.