iTRAQ-Based Quantitative Proteomic Comparison of 2D and 3D Adipocyte Cell Models Co-cultured with Macrophages Using Online 2D-nanoLC-ESI-MS/MS

The demand for novel three-dimensional (3D) cell culture models of adipose tissue has been increasing, and proteomic investigations are important for determining the underlying causes of obesity, type II diabetes, and metabolic disorders. In this study, we performed global quantitative proteomic profiling of three 3D-cultured 3T3-L1 cells (preadipocytes, adipocytes and co-cultured adipocytes with macrophages) and their 2D-cultured counterparts using 2D-nanoLC-ESI-MS/MS with iTRAQ labelling. A total of 2,885 shared proteins from six types of adipose cells were identified and quantified in four replicates. Among them, 48 proteins involved in carbohydrate metabolism (e.g., PDHα, MDH1/2, FH) and the mitochondrial fatty acid beta oxidation pathway (e.g., VLCAD, ACADM, ECHDC1, ALDH6A1) were relatively up-regulated in the 3D co-culture model compared to those in 2D and 3D mono-cultured cells. Conversely, 12 proteins implicated in cellular component organisation (e.g., ANXA1, ANXA2) and the cell cycle (e.g., MCM family proteins) were down-regulated. These quantitative assessments showed that the 3D co-culture system of adipocytes and macrophages led to the development of insulin resistance, thereby providing a promising in vitro obesity model that is more equivalent to the in vivo conditions with respect to the mechanisms underpinning metabolic syndromes and the effect of new medical treatments for metabolic disorders.


Results & Discussion
In vitro insulin resistance-induced 3D co-culture system. In vitro 2D mono-culture studies are considered to be a fundamental, accessible, effective, and promising means to identify the cellular mechanisms and key effectors related to metabolic syndromes. However, the 2D mono-culture approach fails to model the influence of the surrounding tissue architecture on adipocytes. Thus, attempts to develop in vitro 3D cell culture systems have been undertaken. In our previous study 30 , we demonstrated that the co-culture of adipocytes and macrophages in a 3D cell culture system results in changes in lipid and glucose metabolism, which is similar to that of the effect of GW9662 on insulin resistance in adipose tissue in diabetic mice. However, there is no clear reason why the insulin sensitivity in adipocytes induced by 3D co-culturing models with macrophages was shown to be consistent with the response observed in vivo in adipose tissue. In this study, we performed a quantitative proteomic analysis of six different 3T3-L1 cells (preadipocytes, adipocytes and each cell type co-cultured with macrophages in 2D or 3D cell culture conditions) under three different culture conditions (2D, 3D, and 3D co-culture of macrophages). This allowed us to quantitatively profile the adipocyte proteome, thereby identifying candidates that are directly or indirectly linked to the occurrence of insulin resistance.
We first prepared 3D-cultured 3T3-L1 preadipocytes that were fabricated in an alginate scaffold at a concentration of 1 × 10 6 cells/mL. Based on a live/dead cell assay, preadipocytes in the 3D scaffold displayed good viability (Fig. 1A). A functional analysis of the 2D and 3D cell culture systems was performed by assessing the expression of adipogenesis-related markers (CEBPα, PPARγ, ACC, FAS, PLIN, FABP4 and GLUT4) during differentiation under mono-/co-culture conditions. Western blotting revealed that 2D mono-cultured preadipocytes were fully differentiated into adipocytes, but the differentiation of 2D co-cultured preadipocytes with macrophages in the trans-well plate was minimal (Fig. 1B). Interestingly, although the expression of proteins related to lipogenesis in both the 3D mono-and co-cultured cells culture was maintained, as in the 2D mono-cultured adipocytes, GLUT4 expression was markedly decreased (Fig. 1C). These findings are consistent with those of previous studies 44, 45 . The distribution of lipid droplets in the 3D co-/mono-culture model as determined by BODIPY lipid staining indicated that 3T3-L1 preadipocytes were fully differentiated into adipocytes to an almost equivalent extent compared to the 3D mono-cultured adipocytes (Figs 1D and S4 in Supporting Information for 3D co-culture and 3D mono-culture model, respectively). In addition, to confirm the ratios of macrophages to the total number of cells, we performed a fluorescence-activated cell sorting (FACS) analysis. There were 243 macrophages in the 3D co-culture model (about 10% of the total number of cells), and there were an estimated 2,399 adipocytes ( Fig. S5 in Supporting Information). These results of FACS-based cell counting are also consistent with that of BODIPY lipid staining (Figs S5-1 in Supporting Information). A glucose uptake assay demonstrated that glucose uptake in the 3D co-cultured adipocytes significantly decreased relative to that of the 3D mono-cultured adipocytes (Fig. 1E). These results are consistent with the observation that adipose tissue in diabetic mice increases insulin resistance because of macrophage infiltration, and also suggests that the 3D co-culture system has the potential to overcome the limitations of the 2D cell culture system, which merely differentiates cells into adipocytes 21,46 . Proteomic analysis of six different 3T3-L1 cells in 2D and 3D mono-/co-culture conditions using online 2D-nanoLC-ESI-MS/MS. To profile the changes in the expression levels of adipocyte proteomes between 2D-and 3D-cultured 3T3-L1 models, we performed a global quantitative proteomic analysis of six different 3T3-L1 cells (2P, 2A, 2C, 3P, 3A and 3C) using iTRAQ-based online 2D-nanoLC-ESI-MS/MS and illustrated in Fig. 2. We included data quality controls in the proteomic analysis and determined the reproducibility of quantified protein groups from six different 3T3-L1 cells in four replicates (two technical and two biological replicates). As shown in Fig. 3A -2]). The box plot shows the correlation of proteins among the six different 3T3-L1 cell lines in various ratios (Fig. 3B). The dotted red line indicates 1.5-fold increase or decrease in each ratio as the threshold value. Using the iTRAQ-based quantitative analysis, the levels of tens or hundreds of proteins expressed in 3T3-L1 cells changed because of differences in culture conditions. The quantified proteins are listed in the Supplementary Dataset (Supporting Table 1).

Carbohydrate metabolism.
To understand how glucose homeostasis and adipocyte differentiation differ between the 2D and 3D cell-culture conditions, we confirmed the differential expression of six proteins in the 3T3-L1 cells that are involved in carbohydrate metabolism. The differential expression of 18 enzymes from six different 3T3-L1 cells (2P, 2A, 2C, 3P, 3A and 3C) were compared to those of 2P, and the results are summarised (Fig. 4). The expression levels of three proteins, acetyl-CoA carboxylase 1 (ACACA), succinate dehydrogenase Figure 1. Fabrication of the insulin resistance-induced 3D co-culture model. For the 3D co-culture system, 3D scaffolds were fabricated using 3T3-L1 preadipocytes with or without 1% macrophages. (A) Fluorescent images of 3D mono-cultured 3T3-L1 preadipocytes at 7 days using a live/dead cell assay (magnification: 100×). Live cells stained green and dead cells stained red. (B) The protein concentrations of adipogenesis-related markers, such as CCAAT/enhancer binding protein alpha (C/EBPα), peroxisome proliferator-activated receptor gamma (PPARγ), acetyl-CoA carboxylase (ACC), fatty acid synthase (FAS), fatty acid binding protein 4 (FABP4), perilipin (PLIN), glucose transporter type 4 (GLUT4) and beta actin (β-actin), were assessed by Western blotting in the 2D and (C) 3D mono-/co-cultured model. (D) Fluorescent image of 3D mono-/co-cultured adipocytes by BODIPY lipid staining (magnification: 100×). (E) Level of glucose uptake in the 3D culture system. Results are expressed as the means ± S.E.M. (n = 4/group) for duplicate experiments. ### p < 0.001 vs insulin-treated 3D preadipocytes, ***p < 0.001 vs insulin-treated 3D adipocytes. Statistical significance was determined using the Student's t-test or a one-way analysis of variance followed by Tukey's multiple-comparison test; p < 0.05 was considered to be statistcally significant. Full-length blots are presented in Figs S1 and S2 in the Supporting Information.
iron-sulphur subunit (SDHβ), and succinate-CoA ligase subunit beta (SCSB), were not significantly different between the 2D and 3D cell culture systems. In the 2D cell culture system, the expression levels of four proteins (pyruvate carboxylase [PCX, 1. 18-and 2.14-fold change for 3A and 3C, respectively], and citrate synthase [CISY, 1.88-fold change for 3A]) and displayed at least a 1.5-fold change in expression levels. In particular, PCX plays a crucial role in the tricarboxylic acid (TCA) cycle. Pyruvate is introduced to the TCA cycle, and ATP citrate synthase (ACLY) allows acetyl-CoA derived from pyruvate to enter the TCA cycle. CISY catalyses the conversion of oxaloacetate into citrate. These enzymes are crucial for proper operation of the TCA cycle and they were significantly up-regulated, probably because of the production of citrate in the cytoplasm during differentiation and cell growth 10,35 . Furthermore, ACSL1 is required for both carbohydrate and fatty acid metabolism and was also up-regulated in 3A and 3C 39 . We demonstrated that the changes in expression levels of three  Fatty acid metabolism with β-oxidation. Because lipids and their metabolites are implicated in regulating metabolic pathways leading to insulin resistance, the investigation of free fatty acids in adipocytes is important 3 . Fig. 5 shows a map of the key proteins involved in fatty acid metabolism and the expression dynamics of six different 3T3-L1 cells compared to the expression dynamics of 2D mono-cultured preadipocytes. Free fatty acids were transported across the plasma membrane and into the mitochondria using platelet glycoprotein 4 (CD36), FABP4, and fatty acid-binding protein 5 (FABP5) 7,35 . The expression levels of CD36 (1.99-, 3.72-, and 2.45-fold change for 2A, 3A, and 3C, respectively) and FABP4 (1.55-, 4.12-, and 3.21-fold change for 2A, 3A, and 3C, respectively) in 2D mono-cultured adipocytes and 3D-cultured adipocytes with/without macrophages were significantly up-regulated. The diverse enzymes involved in lipolysis and fatty acid metabolism, including FABP5 (3.51-and 2.82-fold change for 3A and 3C, respectively), diacylglycerol O-acyltransferase 1 (DGAT1, 1.98-and 1.51-fold change for 3A and 3C, respectively), patatin-like phospholipase domain-containing protein 2 (ATGL, 1.63-fold change for 3A), hormone-sensitive lipase (HSL, 1.62-and 1.70-fold change for 3A and 3C, respectively), stearoyl-CoA desaturase 1 (SCD1, 4.29-and 2.36-fold change for 3A and 3C, respectively), acyl-CoA synthetase family member 2, mitochondrial (ACSF2, 1.67-fold change for 3C), ACSL1 (5.19-and 3.15-fold change for 3A and 3C, respectively), mitochondrial medium-chain specific acyl-CoA dehydrogenase (ACADM, 2.40-and 2.28-fold change for 3A and 3C, respectively), trifunctional enzyme subunit alpha (HADHα, 1.87-and 1.57-fold change for 3A and 3C, respectively), trifunctional enzyme subunit beta (HADHβ, 1.83-and 1.56-fold change for 3A and 3C, respectively), and 3-ketoacyl-CoA thiolase (ACAA2, 1.83-fold change for 3A) varied considerably in 3D mono-cultured adipocytes and/or adipocyte 3D co-cultured with macrophages when compared to 2D mono-cultured preadipocytes. Enzymes that are activated in the fatty acid oxidation pathway (β-oxidation) were catalysed using acetyl-CoA from free fatty acids and they served to introduce acetyl-CoA into the TCA cycle. Proteins up-regulated during fatty acid metabolism might trigger excess acetyl-CoA production for the maintenance of intracellular homeostasis, which could affect mitochondrial proteins in the electron transport   47 . Western blotting subsequently confirmed these changes in expression of both FABP4 and ACADM, verifying the results using iTRAQ (Fig. S3B in Supporting Information).
Other aspects of metabolism. The metabolism of adipose cells associated with insulin resistance, as well as glucose and fatty acid metabolism, are important for understanding obesity and metabolic syndromes. Furthermore, the adipokines or cytokines induced during adipogenesis and the operation of the mitochondrial electron transport chain (ETC) are also key contributors to obesity 10,23,39 . To assess protein expression dynamics among the six different 3T3-L1 cells, we quantified the expression levels of the enzymes in relation to adipogenesis and the ETC pathways. As shown in Fig. S6 in the Supporting Information, four proteins related to the adipogenesis pathway were quantified in six different 3T3-L1 cells. The measured perilipin-1 (PLIN1, 2.44and 2.21-fold change for 3A and 3C, respectively), lipoprotein lipase (LPL, 2.80-and 2.06-fold change for 3A and 3C, respectively), transmembrane protein 120A (TMEM120A, 3.00-and 1.93-fold change for 3A and 3C, respectively), and acyl-CoA-binding protein (ACBP, 1.83-and 1.60-fold change for 3A and 3C, respectively) were significantly over-expressed in 3D mono-cultured adipocytes and adipocytes 3D co-cultured with macrophages. Among these differentially expressed proteins, PLIN1 and LPL correlate with insulin sensitivity 48,49 and TMEM120A are required for adipocyte differentiation 50 .
To determine the influence of mitochondria and ROS, we analysed the expression dynamics of mitochondrial proteins involved in the ETC and ATP synthesis. Four proteins involved in the function of ETC complexes were examined in six different 3T3-L1 cells and compared to the expression levels in 2D mono-cultured preadipocytes (Fig. S7 in the Supporting Information). Electron transfer flavoprotein subunit alpha (ETFA, 1.80-and 1.80-fold change for 3A and 3C, respectively), which transfers electrons from NADH to CoQ across the mitochondrial membrane in ETC complex I 10,51 , was significantly up-regulated (greater than 1.5-fold) in adipocytes that were 3D mono-or co-cultured with macrophages. Additionally, cytochrome c (CYC, 1.80-fold change for 3A) and ATP synthase mitochondrial F1 complex assembly factor 2 (ATPF2, 1.86-and 1.64-fold change for 3A and 3C, respectively) were significantly up-regulated in 3D-cultured adipocytes with/without macrophages, which was similar to the dynamics that were observed for other metabolic pathways. The expression of ADP/ATP translocase 2 (ADT2, 1.77-and 1.96-fold change for 2C and 3A, respectively) was increased not only in 3D mono-cultured adipocytes, but also in 2D co-cultured adipocytes with macrophages. High expression of the enzymes involved

Correlational network analysis of up-and down-regulated proteins in 3D co-cultured adipocytes with macrophages versus other cell models.
To confirm the protein-protein interactions that occur in adipocytes that are in 3D co-culture with macrophages, the quantified proteins were analysed using STRING. Briefly, 48 up-regulated and 12 down-regulated proteins (which were among the top 124 up-regulated and 66 down-regulated proteins) were clustered (Tables S1 and S2 in the Supporting Information). As shown in Fig. 6A, the cluster indicated by the pink circle correlates with glycolysis, gluconeogenesis, and the TCA cycle. The components (including PDHα, MDH1/2, and FH) contribute to carbohydrate metabolism, which supports the results shown in Fig. 4. The other cluster indicated by the yellow circle includes nine proteins that are involved in mitochondrial fatty acid metabolism, including ETFA, very long-chain specific acyl-CoA dehydrogenase (VLCAD), ACADM, ethylmalonyl-CoA decarboxylase (ECHDC1), methylmalonate-semialdehyde dehydrogenase (ALDH6A1), hydroxyacyl-coenzyme A dehydrogenase (HADH), 3-hydroxyisobutyrate dehydrogenase (HIBADH), short-chain specific acyl-CoA dehydrogenase (SCAD), and 3-hydroxyisobutyryl-CoA hydrolase (HIBCH). These up-regulated proteins are involved in fatty acid beta-oxidation in the mitochondria 53 , and correlate with the proteomic data, as shown in Fig. 5. In Fig. 7, 48 up-regulated proteins are indicated by red dots. The abbreviations of proteins in a volcano plot (Fig. 7A), along with MS/MS spectra of MDH1 (FVEGLPINDFSR, +3, m/z 566.64), FABP5 (ELGVGLALR, +3, m/z 411.26), and FH (VLLPGLQK, +3, m/z 492.66) are described (Fig. 7B). Two clusters were formed by 12 down-regulated proteins in adipocytes that were 3D co-cultured with macrophages (Fig. 6B). Six down-regulated proteins denoted by the blue circle are involved in cellular component organisation. Among these proteins, ANXA1 and ANXA2 are involved in glucose and lipid metabolism and are implicated in insulin sensitivity 54 . Moreover, the development of insulin resistance in ANXA1-deficient mice has been reported in a previous study 55 . The cluster denoted by the green circle consisting of MCM family proteins (MCM2, 3, 4, 5, 6) are required to initiate eukaryotic genome replication 56

Functional annotation of differentially expressed proteins in 3D co-cultured adipocytes with macrophages versus 2D mono-cultured adipocytes.
To determine the effects of the 3D adipocyte culturing technique on the physiology of the adipocyte proteome, we performed a bioinformatic-based functional analysis of the differentially expressed proteins from adipocytes that were 3D co-cultured with macrophages versus the typical cell culture system (2D mono-cultured adipocytes). Among the proteins analysed, 124 and 67 proteins were up-and down-regulated, respectively, in 3D-cultured adipocytes with macrophages compared to the expression levels in 2D mono-cultured adipocytes. The molecular functions of those up-/down-regulated proteins were implicated in the catalytic activity, binding, and other activities related to structural molecule, transporter, and molecular transducer (Fig. 8A). Furthermore, many up-regulated and down-regulated proteins were found to be involved in a metabolic process, cellular process, and biological regulation (Fig. 8B). The major cellular components were assigned to the parts of the cell, organelle, and protein-containing complex region (Fig. 8C). The major functions of the up-regulated and down-regulated proteins in adipocytes that are 3D co-cultured with macrophages relate to the metabolism of glucose and fatty acids in adipose cells during differentiation 34,35 . Based on a KEGG pathway analysis of up-regulated and down-regulated proteins (Fig. 8D), the up-regulated pathways in adipocytes that were 3D co-cultured with macrophages relate to metabolism during adipocyte differentiation. This includes such metabolic pathways as carbon metabolism, glycolysis/gluconeogenesis, the TCA cycle, and biosynthesis of amino acids. These results are similar to those of previous studies 57,58 . Conversely, the down-regulated pathways in adipocytes 3D co-cultured with macrophages were associated with cell proliferation, such as DNA replication, cell cycle regulation and tight junctions. The KEGG pathways for the up-regulated and down-regulated proteins are listed in Supporting Table 2. Based on these results, it is possible that adipocytes 3D co-cultured with macrophages can provide a suitable in vitro obesity model.

Functional annotation of 3D versus 2D co-cultured adipocytes with macrophages.
To understand the diverse functions of the up-regulated and down-regulated proteins in adipocytes that were 3D co-cultured with macrophages compared to adipocytes that were 2D co-cultured with macrophages, we submitted 134 up-regulated and 220 down-regulated proteins to the online tool PANTHER classification system and STRING. With regard to molecular function, many up-regulated and down-regulated proteins in adipocytes 3D co-cultured with macrophages are involved in catalytic activity, binding, structural molecular activity, and molecular function regulation (Fig. S9A in the Supporting Information). A biological process analysis revealed that up-regulated and down-regulated proteins in adipocytes 3D co-cultured with macrophages included metabolic processes, cellular processes, and localisation (Fig. S9B in the Supporting Information). Cellular components were related to the parts of the cell, organelle, and protein-containing complex region (Fig. S9C in the Supporting Information). The KEGG pathways for the up-regulated proteins in adipocytes 3D-cultured with macrophages were different than those of the 2D co-cultured adipocytes (Fig. S9D in the Supporting Information). The KEGG pathways that were enriched for the up-regulated proteins included carbon metabolism, metabolic pathways, www.nature.com/scientificreports www.nature.com/scientificreports/ propanoate metabolism, PPAR signalling pathways, and biosynthesis of amino acids, while down-regulated proteins in adipocytes 3D co-cultured with macrophages were mainly involved in oxidative phosphorylation and thermogenesis. From these results, we propose that the adipocytes that were 3D co-cultured with macrophages represent a useful in vitro obesity model.

Functional annotation of adipocytes 3D co-cultured with macrophages versus 3D mono-cultured adipocytes.
To understand the changes in the microenvironment when adipocytes are co-cultured with macrophages, we performed a functional annotation analysis by comparing adipocytes 3D co-cultured with macrophages to 3D mono-cultured adipocytes. Among the 2,885 proteins analysed, 11 up-regulated and 22 down-regulated proteins in adipocytes 3D co-cultured with macrophages were identified based upon a comparison to the expression profile of 3D mono-cultured adipocytes using the PANTHER classification system and STRING. The molecular functions of the differentially expressed proteins mainly related to binding, catalytic activity, and transporter activity (Fig. S10A in the Supporting Information). For the biological processes, the majority of differentially expressed proteins were involved in cellular and metabolic processes (Fig. S10B in the Supporting Information), and the localisation of the proteins was predicted to be in the cell, extracellular region, and organelles (Fig. S10C in the Supporting Information). These functional annotations supported our previous results that were obtained for adipocytes that were 3D versus 2D co-cultured with macrophages or for adipocytes that were 3D co-cultured with macrophages versus 2D mono-cultured adipocytes. Interestingly, fatty acid metabolism, biosynthesis of unsaturated fatty acids and other functions in the KEGG pathway were down-regulated in adipocytes 3D co-cultured with macrophages (Fig. S10D in the Supporting Information and Supporting Table 2). These findings suggest that macrophage infiltration might be affected by metabolic processes and insulin resistance and that the enzymes involved in fatty acid metabolism are downregulated. We suggest that the co-culture of adipocytes and macrophages during differentiation and cell growth reduces lipolysis of adipocytes by altering cytokine production in macrophages.

Conclusions
This is the first report of a comprehensive quantitative proteomic analysis of a 3D in vitro co-culture model of 3T3-L1 cells and RAW264.7 macrophages. This study has focused primarily on metabolic mechanisms. We performed a high-throughput proteomic analysis of six different 3T3-L1 cells and explored differential protein expression between the 2D and 3D co-cultured adipocyte platform, in addition to preforming a functional annotation. Our analysis identified differentially expressed proteins that are involved in glucose metabolism, fatty acid oxidation, adipogenesis, and the ETC pathway. In particular, the expression levels of proteins involved in diverse metabolic processes were up-regulated in adipocytes 3D co-cultured with macrophages compared to the levels in the 2D-cultured adipocytes. Furthermore, the activation of major cellular metabolism pathways leads to insulin resistance caused by the adipocyte hypertrophy, energy accumulation, and mitochondrial dysfunction. The 3D  Table 2.
in vitro co-culture model consisting of 3T3-L1 cells and RAW 264.7 macrophages along with comprehensive proteomic datasets could widen the potential applicability of mechanical studies and drug screening associated with insulin resistance. To better understand the effects of 3D cell culturing on the adipocyte proteome, however, the development of a fluorescence-activated cell sorting (FACS)-based method for isolating adipocyte cells from macrophages in 3D co-culture systems is required for future study, and an in-depth quantitative proteomic analysis may provide insights into the biological alterations of the adipocyte proteome in 3D cell culture systems.

Materials and Methods
Materials and chemicals. Critical materials and chemicals are listed in Method S1, Supporting Information.
3D scaffold fabrication using the cell-plotting system. In our previous study, we developed a 3D cell printing system to assemble 3D-dispensing structures 30 . Briefly, this system consists of an x-y-z stage, a dispenser, a syringe nozzle, a compression controller and a computer system. The dispenser is the tank reservoir used to hold the cell-hydrogel mixture and it has a 10 × 10 × 10 cm space to fabricate the 3D architecture. The computer system controls the pressure, feeding speed, strand size and shape of the scaffold. The pressure and speed of the cell-plotting system were controlled at 650 kPa and 150 mm/s, respectively. Constant air pressure was applied to the dispenser, and a 3D scaffold was plotted layer-by-layer on the x-y-z stage. The scaffold pattern was designed with an orthogonal orientation. Alginate mixed with collagen, gelatin and cells was placed in a dispenser with a 100-400-μm nozzle for cell printing. The resulting alginate constructs were composed of 10 layers Cell culture and differentiation in the 3D culture system. The seeding densities (1 × 10 6 cells/mL) of 3T3-L1 preadipocytes and RAW264.7 macrophages (1 × 10 4 cells/mL) were adjusted using a disposable haemocytometer-based cell counter (SKC Co. Ltd., Seoul, Korea) with the aid of an inverted microscope (Eclipse TE2000-U, Nikon, Tokyo, Japan). Cells/hydrogel mixtures were plotted using the cell plotting system. The cells in the plotted hydrogel were observed by a live/dead cell assay kit (Invitrogen, Carlsbad, CA) using fluorescence microscopy (Eclipse TE2000-U, Nikon, Tokyo, Japan). Fabricated scaffolds were differentiated using the differentiation medium containing 20 μg/mL insulin, 0.5 mM isobutylmethylxanthine and 1 μM dexamethasone for 3 days, which was then replaced with a medium containing 20 μg/mL insulin for 2 days. Cells were fully differentiated, and the medium was exchanged with an insulin-free medium containing 10% FBS.
Lipid droplet fluorescent staining. To confirm the differential rates of adipocytes from preadipocytes on 3D mono-/co-culture system, we performed a Lipid droplet fluorescent staining and methodological details are provided in Method S2, Supporting Information.

Fluorescence-activated cell sorting (FACS) analysis.
We performed a FACS analysis to measure the ratios of macrophages to the total number of cells in 3D co-culture model and detailed in Method S3, Supporting Information.
Sample preparation for proteomic analysis. Each 3T3-L1 in 2D/3D cell culture system was collected into individual tubes and lysed by adding 500 μL of lysis buffer composed of 50 mM HEPES, 70 mM potassium acetate, 5 mM magnesium acetate, one tablet of protease inhibitor cocktail and 0.2% n-dodecyl-beta-D-maltoside in water. For protein extraction from six of 2D-and 3D-cultured 3T3-L1 cells, each mixture of cells and lysis buffer were vortexed for 1 min and then incubated on ice for 4 min. This process was repeated twice. The remaining unbroken cells and debris were removed by centrifugation at 4 °C and 12,000 × g for 15 min. The protein Protein digestion and iTRAQ labelling. One-hundred micrograms of extracted proteins from the 2Dand 3D-cultured 3T3-L1 cells (six types) were transferred into a new tube. The six different cell lysates were reduced utilizing 50 mM ammonium bicarbonate (ABC) solution with 10 mM DL-dithiothreitol (DTT) for 2 h at 37 °C and alkylated by adding iodoacetamide (IAA) to a final concentration of 20 mM for 30 min at room temperature (RT) in the dark. To remove the remaining IAA in solution, L-cysteine was added at a final concentration of 40 mM for 30 min at RT. Protein samples were proteolytically digested with trypsin at a ratio of 1:20 (enzyme:protein) at 37 °C for 18 h. Each peptide solution was purified using HLB cartridges (1 cc, 10 mg) and then dried using a concentrator (Eppendorf, Concentrator plus) under a vacuum. Each of the dried samples were reconstituted with 30 μL of 0.5 M triethylammonium bicarbonate (TEAB) and 50 μL of isopropanol-suspended iTRAQ reagent containing the reporter ion from 114 to 119 (8-plex iTRAQ reagents). Six samples were labelled with the respective tags as follows: 2D mono-cultured preadipocytes (2P)-114 reagent, 2D mono-cultured adipocytes (2A)-115 reagent, 2D co-cultured adipocytes with macrophage (2C)-116 reagent, 3D mono-cultured preadipocytes (3P)-117 reagent, 3D mono-cultured adipocytes (3A)-118 reagent and 3D co-cultured adipocytes with macrophage (3C)-119 reagent. All samples were incubated for 2 h at RT before adding an equal amount of each sample into a new tube and drying using a concentrator under a vacuum. The iTRAQ-labelled samples reconstituted with 1 mL of water were purified using an HLB cartridge (1 cc, 30 mg) and dried using a concentrator under a vacuum. The iTRAQ-labelled peptides were reconstituted with 0.1% formic acid (FA) solution and analysed using 2D-nLC-ESI-MS/MS.

Online 2D-nanoLC-ESI-MS/MS Analysis.
We analysed the resulting digests using online 2D-nanoLC-ESI-MS/MS 59,60 , which is shown in Fig. 2. Additional methodological details are provided in Method S4, Supporting Information. Data analysis. iTRAQ-labelled peptides were identified and quantified by comparing MS/MS spectra against the Mus musculus database (mouse, UP000000589, updated on 01/18/2019) from UniProt (http://www.uniprot.org/) with a false discovery rate (FDR) of 0.01 using Proteome Discoverer 1.4.1.14 (Thermo Scientific, Bremen, Germany). To perform a data dependence analysis, the MS parameters were set to 20 ppm of a precursor peptide ion mass tolerance and 0.02 Da of a fragment peptide ion mass tolerance. Two missed cleavages were allowed. iTRAQ-8plex labelling of the N-terminus and carbamidomethylation of cysteine residues were set as the fixed modifications. The iTRAQ-8plex labelling of lysine residues, oxidation of methionine and phosphorylation of serine, threonine and tyrosine were all set as variable modifications. The quantified proteins obtained from six different 3T3-L1 cells were normalized against 20 median proteins using an experimental bias of the quantification method and were identified with at least two unique peptides per protein in Proteome Discoverer 1.4.1.14. Data processing and statistical analysis were carried out using Perseus 1.5.8.5. The iTRAQ ratios of proteins were transformed by log 2 (x) and filtered by a minimum of three valid values. The missing values in ratios of reporter ions corresponding to each 3T3-L1 cell cultured in the 2D/3D system were imputed using a normal distribution. Subsequently, a one-sample t-test was carried out to obtain the statistical significance among biological and technical replicates. The significance threshold used was a ±1.5-fold change and p < 0.05. The resulting data were exported to Microsoft Excel, and a Venn diagram was generated using the online tool Venny 2.1 (http://bioinfogp.cnb.csic.es/tools/venny/index.html). Protein classification was performed using the PANTHER classification system (http://www.pantherdb.org/tools/) based on functional annotations for cellular component, biological process, protein class and molecular function. The search tool for protein-protein interactions and the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis were used to identify interacting genes and proteins (STRING, https://string-db.org/).

Data availability
The datasets generated during the current study are available in PRIDE, accession number: PXD014437. https:// www.ebi.ac.uk/pride/archive/login.