All chemicals were obtained from Sigma Aldrich (Munich, Germany) unless otherwise stated.

Human adipose-derived stromal cells (hASCs) were obtained from stromal-vascular fractions of subcutaneous neck (SC) and DN tissues of volunteers, aged between 35–75 years, undergoing planned surgical treatment. A pair of biopsies from SC and DN areas was obtained from the same donor, to avoid inter-individual variations (Sárvári et al., 2015; Kristóf et al., 2019; Tóth et al., 2020). Patients with known diabetes, body mass index > 30, malignant tumor, infection or with abnormal thyroid hormone levels at the time of surgery were excluded from the study. Written informed consent was obtained from all participants before the surgery. Data of the donors included in RNA-sequencing are listed in Supplementary Table 1.

hASCs were isolated and cultivated as previously described (Sárvári et al., 2015; Kristóf et al., 2019; Tóth et al., 2020). The absence of mycoplasma was confirmed by PCR analysis (PCR Mycoplasma Test Kit I/C, Promocell, Heidelberg, Germany). Cells were differentiated following a previously described white adipogenic differentiation protocol, with or without the addition of human recombinant irisin (Cayman Chemicals, MI, United States) (provided in 50 mM Tris pH 8.0, 150 mM sodium chloride, and 20% glycerol stocks) at 250 ng/mL (20 nM) concentration (the stock was diluted 1:6,500) (Fischer-Posovszky et al., 2008; Raschke et al., 2013; Kristóf et al., 2019). Media were changed every 4 days and cells were collected after 14 days of differentiation. In every repetition, untreated and irisin treated samples were obtained from the same donor. Cells were incubated at 5% CO₂ and 37°C. Where indicated, cells were treated with RGDS peptide (10 μg/mL, R&D systems, MN, United States) (Kim et al., 2018) or SN50 (50 μg/mL, Med Chem Express, NJ, United States) (Sárvári et al., 2015).

Cells were collected in Trizol reagent (Thermo Fisher Scientific, MA, United States) and RNA was isolated manually by chloroform extraction and isopropanol precipitation. To obtain global transcriptome data, high throughput mRNA sequencing was performed on Illumina Sequencing platform (Tóth et al., 2020). Total RNA sample quality was checked by Agilent Bioanalyzer using Eukaryotic Total RNA Nano Kit; samples with RNA integrity number >7 were used to prepare the library. Libraries were prepared by NEBNext^® Ultra^TM II RNA Library Prep for Illumina (New England BioLabs, Ipswich, MA, United States). Sequencing runs were executed on Illumina NextSeq500 using single-end 75 cycles sequencing. The reads were aligned to the GRCh38 reference genome (with EnsEMBL 95 annotation) using STAR aligner (Dobin et al., 2013). To quantify the reads, featureCounts was used (Liao et al., 2014). Gene expression analysis was performed using R. Genes with very low expression and with outlier values were removed from further analysis. To further remove outlier genes, Cook’s distance was calculated and genes with Cook’s distance higher than 1 were filtered out. PCA analysis did not show any batch effect considering sequencing date and the donor origin, sex or tissue origin (data not shown). DESeq2 algorithm was used to detect the differentially expressed genes based on adjusted p values < 0.05 and log2 fold change threshold >0.85. Grouping was performed based on Panther Reactome pathways¹. Heatmap visualization was performed on the Morpheus web tool² using Pearson correlation of rows and complete linkage based on calculated z-score of DESeq normalized data after log₂ transformation (Tóth et al., 2020). The interaction networks were determined using STRING³ and constructed using Gephi 0.9.2⁴. The size of the nodes was determined based on fold change (Tóth et al., 2020).

For RT-PCR, RNA quality was evaluated by spectrophotometry and cDNA was generated by TaqMan reverse transcription reagents kit (Thermo Fisher Scientific) followed by qPCR analysis (Szatmári-Tóth et al., 2020). LightCycler 480 (Roche Diagnostics, IN, United States) was used to determine the normalized gene expression using the probes (Applied Biosystems, MA, United States) which are listed in Supplementary Table 2. Human GAPDH was used as an endogenous control. Samples were run in triplicate and gene expression values were calculated by the comparative cycle threshold (Ct) method. ΔCt represents the Ct of target after deducting the GAPDH. Normalized gene expression levels were calculated by 2^–ΔCt.

Samples were collected, separated by SDS-PAGE, and transferred to PVDF Immobilon-P transfer membrane (Merck-Millipore, Darmstadt, Germany) as previously described (Szatmári-Tóth et al., 2020). The following primary antibodies were used overnight in 1% skimmed milk solution: anti-p50 (1:1,000, 13755, Cayman Chemicals), anti- IκBα (1:1,000, 4812, Cell Signaling Technology, MA, United States), and anti-β-actin (1:5,000, A2066, Novus Biologicals, CO, United States). HRP-conjugated goat anti-rabbit (1:10,000, Advansta, CA, United States, R-05072-500) or anti-mouse (1:5,000, Advansta, R-05071-500) IgG were used as secondary antibodies, respectively. Immobilion western chemiluminescence substrate (Merck-Millipore) was used to visualize the immunoreactive proteins. FIJI was used for densitometry.

hASCs from SC and DN areas were plated and differentiated in eight well Ibidi μ-chambers (Ibidi GmbH, Gräfelfing, Germany). Cells were treated with Brefeldin A (100 ng/mL), an inhibitor of intracellular protein transport, 24 h prior collection to sequester the released CXCL1 (Sárvári et al., 2015; Kristóf et al., 2019). After that, cells were washed with PBS, fixed by 4% paraformaldehyde, permeabilized with 0.1% saponin and blocked by 5% milk as per described protocols (Szatmári-Tóth et al., 2020). The cells were incubated subsequently with anti-CXCL1 primary antibody (1:100, 712317, Thermo Fisher Scientific) and Alexa 488 goat anti-rabbit IgG (1:1,000, A11034, Thermo Fischer Scientific) secondary antibody for 12 and 3 h at room temperature, respectively. Propidium iodide (1.5 μg/mL, 1 h) was used to label the nuclei. A secondary antibody test was also performed where the cells were incubated only with the respective secondary antibodies. Images were acquired with Olympus FluoView 1000 confocal microscope and analyzed by FIJI (Szatmári-Tóth et al., 2020). Boundaries of preadipocytes and differentiated adipocytes were identified manually based on brightfield (BF) images and nuclear staining, followed by quantification of immunostaining intensity. Adipogenic differentiation rate was quantified as described previously (Doan-Xuan et al., 2013; Kristóf et al., 2015).

Supernatants of samples from cell culture experiments were collected at regular replacement of media, on days 4, 12, 18, 21 of differentiation, wherever indicated. For SC and DN, supernatants were collected and stored at −20°C from the differentiated cells of the same donor and considered as one repetition, followed by repetition with subsequent donors. For tissues, 10–20 mg of SC and DN tissue samples from the same donor were floated for 24 h in DMEM-F12-HAM medium with or without the presence of 250 ng/mL irisin (Ballak et al., 2013; Kristóf et al., 2019). The release of CXCL1, CX3CL1, IL-32, TNFα, and IL1-β were analyzed from the stored samples using ELISA Kits (R&D systems, MN, United States).

A human umbilical vein endothelial cell (HUVEC) cell line was generated from endothelial cells isolated from the human umbilical cord vein of a healthy newborn by collagenase digestion as described earlier (Palatka et al., 2006). Cells were cultured in M199 medium (Biosera, Nuaille, France) containing 10% FBS (Thermo Fisher Scientific), 10% EGM2 Endothelial Growth Medium (Lonza, Basel, Switzerland), 20 mM HEPES (Biosera), 100 U/mL Penicillin, 100 μg/mL Streptomycin and 2.5 μg/mL Amphotericin B (Biosera), and immortalized by the viral delivery of telomerase gene using pBABE-neo-hTERT (Counter et al., 1998) (gift from Bob Weinberg, 1774, Addgene). The virus packaging was performed in HEK293FT cells (Thermo Fisher Scientific) based on a calcium precipitation method using pUMVC and pCMV-VSV-G vectors (Stewart et al., 2003) (gift from Bob Weinberg, 8449 and 8454, Addgene). The pseudovirion containing supernatant was used for infection, and selection was started 72 h later using 300 μg/mL G418 (Merck-Millipore). Immortalized cells completely retain the morphological properties of primary endothelial cells.

Prior to the adhesion assay, EGM2 was omitted from the standard medium of HUVEC cells and FBS content was decreased to 1% (in which condition cell proliferation is unlikely) for 24 h. 96-well plates (Thermo Fisher Scientific) were precoated with fibronectin (Merck-Millipore) at 1.25 μg/mL concentration in PBS, for 1 h at 37°C and then washed twice with PBS. After centrifugation, trypsinized HUVEC samples were diluted for coating based on counting with three parallels using KOVA Glasstic Slide with Counting Grids (KOVA International, Netherlands). Then cells were plated at 1,000 cells/well density and left to adhere for 2 h in the CO₂-incubator in the mixture (1:1 ratio) of starvation and conditioned media (incubation period from day 8–12 of differentiation) from SC and DN adipocytes, differentiated in the presence or absence of 250 ng/mL irisin, respectively. Where indicated, recombinant human CXCL1 (275-GR, R&D Systems) was used at 2,500 pg/mL, at the highest observed concentration in media of irisin treated ex vivo differentiated adipocytes, in starvation media. Unattached cells were removed by once washing with PBS and adhered cells were incubated with starvation media containing CellTiter-Blue Cell Viability reagent (resazurin; Promega, WI, United States; 36 times dilution). To determine the ratio of attached cells in various conditions, the fluorescent intensity change of each well (Ex:530 nm/Em:590 nm), due to the conversion of resazurin to resorufin by cellular metabolism, was measured using Synergy H1 (BioTek, Hungary) plate reader 2, 4, 6, 18, and 24 h after adding resazurin. Fluorescent intensity values were plotted with respect to time, followed by calculation of slope, which gave the relative adhesion values, after subtraction of values for only starvation media without cells. A linear slope was obtained, which suggests that there could be only negligible cell proliferation, and the gained values represent endothelial cell adhesion measuring the attached viable endothelial cells during the treatments. The final value of adhesion was represented in RFU/hr units and taken from the mean of technical parallels with a minimum of three independent repetitions.

Results are expressed as mean ± SD for the number of independent repetitions indicated. For multiple comparisons of groups, statistical significance was determined by one- or two-way analysis of variance followed by Tukey post hoc test. In comparison of two groups, two-tailed unpaired Student’s t-test was used. For the design of graphs and evaluation of statistics, Graphpad Prism 9 was used.

Primary hASCs from nine independent donors were isolated and cultivated from SC and DN area of human neck, as described (Tóth et al., 2020). Adipogenic differentiation was driven by a white adipocyte differentiation medium with or without the presence of irisin for 14 days. Then, the global gene expression pattern of differentiated adipocytes and undifferentiated hASCs were determined by global RNA-sequencing (Tóth et al., 2020). Gene expression of general adipocyte markers (e.g., FABP4, ADIPOQ) was higher in all differentiated adipocytes as compared to preadipocytes (Figure 1A). Quantification of the adipogenic differentiation rate by laser-scanning cytometry (Kristóf et al., 2015) revealed that more than 50% of the cells were differentiated following our 14-days long differentiation protocol (Figure 1B). The presence of irisin did not affect the differentiation and gene expression of general adipocyte markers (Figures 1A,B). A recent publication proposed the receptors for irisin to be integrins, Integrin subunit alpha V (ITGAV) and Integrin subunit beta (ITGB) 1/3/5 (ITGB1/3/5) (Kim et al., 2018). Hence the expression of ITGAV was analyzed from RNA-sequencing data (Figure 1C), which revealed that it is expressed in both the preadipocytes and differentiated adipocytes. Upon RT-qPCR validation, a significant increase of ITGAV expression was observed in DN adipocytes in response to irisin (Figure 1D). RNA-sequencing data showed that ITGB1, 3, and 5 were also expressed at a high extent in preadipocytes and in differentiated adipocytes irrespective of the presence of irisin (Supplementary Figure 1).

RNA-Sequencing analysis identified 79 genes to be higher expressed upon irisin treatment that are visualized by a Volcano plot (Figure 2A). 50 and 66 genes were significantly upregulated in SC and DN area adipocytes, respectively, each of which are listed in Supplementary Table 3. 37 genes, including CXCL1, CX3CL1, IL32, IL34, IL6, and CCL2 were found to be commonly upregulated in adipocytes of both depots (Figures 2A,B and Supplementary Table 3). Surprisingly, thermogenic marker genes did not appear among these. Panther enrichment analysis of genes upregulated in both SC and DN adipocytes by irisin treatment revealed pathways such as cytokine signaling (NFKB2, CXCL1, CXCL2, IL32, IL34, IL6, CCL2), interleukin-4 and 13 signaling (IL6, CCL2, JUNB, ICAM1), and class A/1 rhodopsin like receptors (CXCL3, CXCL5, CX3CL1, CXCL2, CCL2, CXCL1), which were commonly upregulated in both SC and DN adipocytes (Table 1). Gephi diagrams illustrate the interaction of upregulated genes that belong to several pathways (Figures 2C,D). Interleukin-10 signaling were amongst the upregulated pathways in SC adipocytes (Figure 2C), while in DN, G-alpha-I and response to metal ions were upregulated (Figure 2D). Cluster analyses and heatmap illustration of the gene expression values of the 79 higher expressed genes upon irisin treatment identified two main clusters: a cluster of 25 genes that were uniquely expressed in irisin treated mature adipocytes, and another group of genes that were expressed highly in preadipocytes, but suppressed in differentiated adipocytes without irisin treatment (Supplementary Figure 2). The higher expression of IL6, CCL2, CX3CL1, and IL32, cytokine encoding genes was observed by both RNA Sequencing and RT-qPCR analysis (Supplementary Figure 3). Next, we investigated if fractalkine (encoded by CX3CL1 gene) and IL-32 were released into the conditioned media collected during the differentiation on days numbers 4 and 12; however, we were unable to detect these factors (data not shown).

Irisin upregulated CXCL1 gene expression at the largest extent in both SC and DN area adipocytes (Figures 2A, 3A and Supplementary Table 3). This observation was verified by RT-qPCR (Figure 3B). As a next step, release of CXCL1 from irisin treated and untreated adipocytes was investigated into the conditioned differentiation media collected on the fourth and twelfth days of differentiation. Irisin treatment resulted in significant increase in CXCL1 secretion at the intervals of days 0–4 and 8–12 in both types of adipocytes (Figure 3C).

We aimed to further investigate the dependence of CXCL1 release on the presence of irisin. Therefore, we differentiated hASCs for 21 days, with three sets of samples, each from SC and DN derived adipocytes. Two sets of hASCs were differentiated as previously described, and for the third set, irisin treatment was discontinued after 14 days. Conditioned media were collected on days number 4, 12, 18, 21 and measured for the release of CXCL1. Large amounts of CXCL1 were secreted throughout the differentiation period in the presence of irisin; however, discontinuation of irisin administration led to gradual and significant reduction of the released chemokine (Figure 3D).

A recent publication indicated that RGDS peptide, an integrin receptor inhibitor, can potentially inhibit the effect of irisin (Kim et al., 2018). Hence, we checked the effect of this peptide on the release of CXCL1 on top of irisin treatment. RGDS partially reduced the irisin-stimulated release of CXCL1 by DN adipocytes at day 12 of the differentiation period (Figure 3E).

Release of CXCL1 throughout the whole differentiation period raised a possibility that both undifferentiated preadipocytes and differentiated adipocytes are able to release the chemokine. To investigate this, the secretion machinery of the mixed cell population was inhibited by Brefeldin A, followed by CXCL1 immunostaining and image acquisition by confocal microscopy. Irisin treatment significantly increased CXCL1 immunostaining intensity in both SC (Figure 4A) and DN adipocytes (Figure 4B). Irisin treated adipocytes accumulated significantly more CXCL1 compared to their preadipocyte counterparts in both SC (Figure 4A) and DN areas (Figure 4B). A test for the secondary antibody alone confirmed that the applied secondary antibody did not produce a labeling on its own by unspecifically binding to the cells (Supplementary Figure 4). Our data suggests that irisin stimulates the release of CXCL1 from differentiating and mature adipocytes which is strongly dependent on the presence of irisin but not prominently on its presumed integrin receptor.

Next, we aimed to investigate the molecular mechanisms underlying the irisin-induced CXCL1 release. According to our RNA Sequencing data, irisin treatment resulted in a significant upregulation of NFKB2 and a very modest trend for an increase in NFKB1 and RELA (Supplementary Figures 5A–C) genes. RT-qPCR validation indicated significant upregulation of NFKB1 (p50 subunit) and RELA (p65 subunit) in DN, while an increasing trend was observed in SC adipocytes (Figures 5A,B). p50 protein expression was significantly increased in DN and a slightly increasing trend was found in the case of SC adipocytes (Figure 5C). Protein expression of IκBα, the inhibitor of NFκB transcription factor, decreased significantly upon irisin treatment in SC and a decreasing trend was observed in DN adipocytes (Figure 5D), indicating the upregulation of NFκB pathway.

To prove the direct involvement of the NFκB pathway in adipocyte response to irisin, we applied a cell permeable inhibitor of NFκB nuclear translocation, SN50 (Sárvári et al., 2015), which significantly reduced the release of the chemokine from both types of adipocytes, when it was applied on top of irisin on both the fourth and twelfth days of differentiation, as compared to cells stimulated only by irisin (Figure 5E).

The observed effects of irisin are not likely to be caused by any contamination of endotoxins, which is proved by the negligible expression of TNFα or CCL3 genes (Supplementary Figures 5D,E), and the decreasing trend of IL1β gene expression (Supplementary Figure 5F) in irisin treated adipocytes. Furthermore, we did not detect secreted TNFα or IL-1β in the conditioned media of either untreated or irisin treated SC and DN derived adipocytes (data not shown).

Finally, SC and DN paired tissue biopsies were floated in the presence or absence of irisin dissolved in empty media, followed by quantification of CXCL1 release. The secretion of the chemokine was significantly stimulated from DN tissue biopsies upon irisin treatment (Figure 6A).

Secretion of CXCL1 plays an important role in wound repair and angiogenesis (Gillitzer and Goebeler, 2001). Angiogenesis is crucial for the thermogenic function of BAT (Cannon and Nedergaard, 2004). Therefore, we intended to detect whether the released chemokine can contribute to increased adhesion ability of endothelial cells. Conditioned media collected on the twelfth day of ex vivo differentiation, from untreated and irisin treated SC and DN area adipocytes, were added to HUVECs followed by a resorufin based adhesion assay. The conditioned medium from irisin treated adipocytes, which contains various released factors (including CXCL1) was able to significantly increase the number of attached viable HUVECs, compared to the conditioned medium of untreated adipocytes (Figures 6B,C). When HUVECs were treated with recombinant CXCL1, at the highest observed concentration in media of irisin treated ex vivo differentiated adipocytes, their adhesion property was enhanced significantly (Figure 6D). This suggests a potential beneficial role of the released CXCL1 in promoting endothelial functions and adipose tissue remodeling to support efficient thermogenesis indirectly by enhancing vascularization.