An in vitro model for hypertrophic adipocytes: Time‐dependent adipocyte proteome and secretome changes under high glucose and high insulin conditions

Abstract Obesity is the consequence of a positive energy balance and characterized by enlargement of the adipose tissue, which in part is due to hyperplasia and hypertrophy of the adipocytes. Not much is known about the transition of normal mature adipocytes to the hypertrophic state, which in vivo is very hard to study. Here, we have maintained mature human SGBS cells as a surrogate for adipocytes, changes of morphological and molecular metabolism of the adipocytes were monitored over the first 4 days and the last 4 days. In total, 393 cellular proteins and 246 secreted proteins were identified for further analysis. During the first 4 days of high glucose and insulin, the adipocytes seemed to prefer pyruvate as energy source, whereas beta‐oxidation was down‐regulated supporting lipid loading. Over time, lipid droplet fusion instead of lipid uptake became relatively important for growth of lipid droplets during the last 4 days. Moreover, ECM production shifted towards ECM turnover by the up‐regulation of proteases over eight days. The present in vitro system provides insight into the metabolic changes of adipocytes under conditions of high glucose and insulin, which may help to understand the process of in vivo adipocyte hypertrophy during the development of obesity.

to the development of obesity. 13 Gene expression analysis indicated that after 7 or 28 days OF, changes in the expression profile of AT already can be observed. 14,16 Recently, Alligier et al showed that on OF subcutaneous AT shows different responses over time. Changes of genes expression after 14 days OF indicated a significant impact on the lipid metabolism and storage pathways, whereas after 56 days of OF changes related more to pathways of extracellular matrix (ECM) and inflammation. 15 Another approach to study changes in cellular and metabolic behaviour of AT during OF is to use an in vitro culture system. Although such a model system does not reflect the in vivo situation directly, valuable clues to biological processes could be obtained, which specifically pertain to the development of hypertrophic adipocytes.

Simpson Golabi Behmel Syndrome (SGBS) cells have been well ac-
cepted as an excellent in vitro surrogate for human white subcutaneous adipocytes with similarity in morphology, physiology and biochemistry. [17][18][19][20] Here, we subjected mature SGBS adipocytes to a high glucose and high insulin condition for 4 and 8 days and studied the changes of the cellular and secreted proteome. Comparing the changes of the first 4 days with those of the second 4 days OF showed a clear shift in the cellular processes.
Since day 14, the medium of the mature adipocytes was changed

| Morphology monitoring and Oil Red O staining
The morphology changes from day 14 onwards were closely recorded using a Nikon Eclipse TS100 microscope equipped with a Digital Sight microscope camera control unit (DS-L3, Nikon). The mean diameter of the five biggest fat droplets was recorded and Oil Red O (ORO) staining was performed as parameters to monitor the turnover of the stored fat as previously described. 21

| Protein sample collection
For secreted protein isolation, the medium was collected at T14, T18 and T22 from each well separately. The collected medium (4 mL per well) was centrifuged at 2660 g for 10 minutes (Universal 30 RF, Hettich Benelux BV, the Netherlands). Thereafter, the supernatant was gently transferred to a new tube, snap-frozen in liquid nitrogen and stored at −80°C for further analysis.
For cellular protein collection, wells with cultured cells were washed twice with PBS buffer and lysed with SDT buffer (2% sodium dodecyl sulphate/50 mmol/L dithiothreitol/100 mmol/L Tris-HCl pH = 7.6), 300 µL per well. Cells were scraped off with scraper (Corning) and the lysate was collected in tubes, then heated at 95°C for 5 minutes. After heating, samples were sonicated in three 20-seconds cycles and centrifuged at 16 000 g for 5 minutes at 20°C, and then, the supernatant was carefully transferred to another tube. All samples were stored at −80°C for protein digestion and LC-MS/MS quantification. The entire experiment was performed three times and for each experiment triplicates were available for each protein isolation.

| Cellular sample preparation for LC-MS/MS
Amicon Ultra 0.5-mL centrifugal filter devices (Sigma-Aldrich) were pre-treated by soaking overnight in 5% Tween 20, washing with Milli-Q for 10 minutes with 600 rpm shaking, repeat the washing by centrifuge at 14 000 g at 20°C for 25 minutes before use.
Cellular protein sample digestion has been described in detail previously. 21 In short, after induction and alkylation, 42 µg protein of time point T14, T18 and T22 was supplemented with 1 µg trypsin/Lys-C Mix (Thermo Fisher Scientific) and incubated for 9-14 hours at 37°C. After overnight digestion, the peptide samples were cleaned from residual sodium deoxycholate and SDS by precipitation with an equivalent volume of 4 mol/L potassium chloride, acidified to pH = 1-2 with 100% formic acid (FA). Then, peptide samples were desalted with a column made by stacking three layers of a 3 mol/L Empore C18 column (Thermo Fisher Scientific) in a P20 pipet tip. After the column was pre-rinsed with 50 µL 70% acetonitrile (ACN) and equilibrated with 50 µL 100% FA by air pressure, the cellular samples were loaded on the column and eluted with 30 µL 70% ACN/5% FA, and the desalted sample was collected in a clean LoBind tube (Eppendorf, Sigma-Aldrich).
Peptides were dried under vacuum and labelled with TMT 10plex Mass Tagging Kits (Thermo Fisher Scientific; 90111) according to the manufacturer's protocol. In short, 42 µg cellular peptides diluted into 84 µL of 50 mmol/L triethyl ammonium bicarbonate were transferred into the reaction tube, which contained the labelling reagents dissolved in 41 µL anhydrous acetonitrile per tube. The labelling reaction was incubated for 1 hour at room temperature and quenched 15 minutes by adding 8 µL of 5% hydroxylamine. Equal amounts of combined samples were transferred into a new micro-centrifuge tube for LC-MS/MS with a final concentration of 0.33 µg/µL. The entire experiment was performed three times and as such generated nine samples per time point.

| Secretome samples preparation for LC-MS/MS
For secreted proteins, the digestion was on the filter. 22 In general, the medium of each vial was added to the pre-rinsed filter device, Therefore, totally seven samples were generated per time point from the three independent experiments. Compared with cellular proteins, the peptide concentration of the secretome samples was much lower, which made the secretome quantification more challenging and more sensitive method was needed. As TMT labelling yields lower peptide identification rates 23 and lowered accuracy when quantified on MS 2 level, 24 therefore, TMT labelling was omitted and label-free quantification was used for secretome samples.

| Protein identification using LC-MS/MS
A nanoflow HPLC instrument (Ultimate 3000, Dionex) was coupled online to a Q Exactive mass-spectrometer (Thermo Fisher Scientific) with a nano-electrospray Flex ion source (Proxeon). For cellular samples, 5 µL of the TMT labelled cellular peptide samples were loaded. For secreted samples, an equal amount of Pierce Digestion Indicator peptides was added to all peptide samples as internal standard, and then, 5 µL of this mixture were loaded as well. Samples were loaded onto a C18-reversed phase column (Acclaim PepMap C18 column, 75-μm inner diameter × 15 cm, 2-μm particle size). The peptides were separated with a 120 min- The MS data were searched using Proteome Discoverer 2.2 Sequest HT search engine (Thermo Fisher Scientific) against the UniProt human database. The false discovery rate was set to 0.01 for proteins and peptides, which had to have a minimum length of six amino acids. The precursor mass tolerance was set at 10 ppm, the fragment tolerance at 0.02 Da and one miss-cleavage was allowed.
For secreted and cellular samples, oxidation of methionine was set as a dynamic modification and carbamidomethylation of cysteines as fixed. TMT reagent adducts (+229.162932 Da) on lysine and peptide amino termini were set as fixed modifications for cellular samples.

| Protein quantification
For cellular proteins, quantification followed relative comparison of the TMT-specific peaks in the MS 2 spectrum. Label-free quantitation was conducted for secretome samples using the Minora Feature Detector node in the processing step and the Feature Mapper node combined with the Precursor Ions Quantifier node in the consensus step with default settings within Proteome Discoverer 2.2 (Thermo Fisher Scientific, XCALI-97808).

| Data normalization
For cellular proteins, the data of each run were normalized to the total peptide amount in each channel and to compare between the runs scaled to time point T18.
For secreted proteins, the LC-MS analysis was done in seven runs, each run containing a sample from each time point (T14, T18 and T22). The total number of proteins that was identified in the medium was 1264. Data normalization was performed in two steps. First, to correct data for possible differences between runs, we chose the 476 proteins, which were present in all of the analysed samples. We calculated the mean abundance of those 476 proteins in all 7 runs (M) and also mean abundance per run (m x for run x). Normalization factor 1 for run x (f1 x ) = M ÷ m x . Data were corrected (D1) as follows: D1 = f1 x × original protein abundance in run x. Also, the Pierce Indicator added to each sample was normalized by f1. The second normalization was then performed to stratify the protein abundances according to the Pierce Indicator. Normalization factor 2 for sample y (f2 y ) = Pierce's mean abundance from all samples ÷ Pierce abundance in sample y. In general, the second normalization step was: D2 = f2 × D1.

| Validation of secreted proteins
To verify the secreted nature of the identified proteins, their amino acid sequences were obtained from UniProt and analysed with SignalP 25,26 and Deeploc. 27 Proteins identified to contain a signal peptide by SignalP or validated in the extracellular space by Deeploc were picked up as secreted proteins. Accordingly, around 26.58% of the identified proteins from the medium was finally confirmed as secreted proteins.

| Missing value handling
Performing LC-MS analysis of proteins, values could be missing for various reasons. 28 For cellular proteins, only those having no more than four missing values per time point were selected. As for secreted proteins, only proteins recognized as secreted and with no more than three missing values were selected. The Multiple Imputation routine of SPSS was used to impute those selected protein's missing values, and further analysis was subsequently performed.

| Western blotting
The protein concentration was determined as described before by using the BCA kit (Pierce, Thermo Fisher Scientific; 23252). Fifteen microgram of extracted proteins was run on a 12% SDS-PAGE gel, then electro-transferred onto nitrocellulose membranes. After blocking with 4% non-fat milk for 2 hours, the nitrocellulose membranes were incubated overnight at 4°C with primary antibodies against Akt and p-Akt (AKT #9272, p-AKT#9271, all 1:1000, Cell Signaling Technology). After washing three times with Tris-buffered saline with 0.1% Tween 20 (TBST), each time for 10 minutes, the membranes were incubated with horseradish peroxidase (HRP) conjugated secondary antibodies (anti-Rabbit DAKO cat# P0399) for 1 hour at room temperature. Then, after 3 × 10 minutes washing with TBST and 1 × 10 minutes with TBS alone, the protein blots were visualized with ECL detection reagent (Pierce; SuperSignal TM west Dura femto max sensitivity; Thermo Fisher Scientific; 34095). The density of protein bands was determined and quantified with local background correction by using the ChemiDoc XRS system (Bio-Rad).

| Statistical analyses
Data were described as mean ± SEM (the standard error of the mean, SEM), and both cellular and secreted protein abundances were log 2transformed. Proteome changes were analysed by using two-tailed dependent t test with a cut-off for significance of P < .05. Statistical analyses were conducted using SPSS version 22.0.

| Functional analysis
Clustering of protein interactions were visualized by STRING. 29 Enriched pathways were analysed by DAVID, 30 and significantly changed cellular proteins and secreted proteins were pooled together to detect potentially affected processes by early feeding or late feeding.

| Morphologic characteristics of SGBS cells during eight days feeding
After 14 days differentiation, approximately 85%-90% SGBS preadipocytes had differentiated into mature adipocytes, which were largely occupied by fat droplets. The morphologic changes are recorded in Figure 1. During the early feeding period (T14-T18), the adipocytes did not change in size, while the diameter of fat droplets increased. During the late feeding period (T18-T22), the size of the bigger fat droplets continued to increase, while the number of visible fat droplets per cell decreased. Figure 1D shows the mean diameter of the five biggest fat droplets per cell over time. 21,31 The diameter of biggest fat droplets increased 0.30 µm after the early feeding period (P = .004) and further increased 0.35 µm after late feeding (P = .078). It suggests that these fat droplets have more than doubled their fat content between T18 and T22. On the other hand, the fat content per adipocyte measured by ORO staining showed an increase in OD value 0.7 after early feeding (P < .001), which was limited to a 0.4 OD increase after the late feeding (P = .007). Fat droplets can grow by uptake of triglycerides into the cells and by fusion with other droplets. 32 The lower number of fat droplets, the increase in diameter of the biggest droplets and the lower increase in fat content during the late phase seems in line with an increased contribution of fusion to the enlargement of fat droplets during the late feeding period.

| Proteome and secretome changes during the early and late feeding period
In total, 1124 proteins were identified from cell lysates and 1264 proteins from collection medium, of which 393 cellular proteins and 246 secreted proteins were finally selected due to the lower missing value rate as well as secretome validation (Table S1). We then characterized the alterations of the proteome and secretome in the early and late feeding phase. After the early 4-day feeding (T14-T18), 82 cellular proteins (Table S2) and 62 secreted proteins were differentially expressed (Table S3). When it comes to the late 4-day feeding period (T18-T22), 47 cellular proteins (Table S2) and 63 secreted proteins (Table S3) significantly changed.

| Functional analysis of proteins changed during T14-T18
Out of the 82 differential cellular proteins, 61 were unique for the early feeding period (T14-T18 (Figure 2 Table 1).

| Functional analysis of proteins changed during T18-T22
Regarding the late feeding period (T18-T22), 26 out of 47 cellular differential proteins were unique for the prolonged feeding period (Figure 2), of which 18 proteins were up-regulated and 8 proteins were down-regulated. Of the 63 secreted differential proteins, 39 proteins merely showed up during the late feeding period, of which 23 proteins were up-regulated and 16 proteins were down-regulated. Notably, according to the subcellular localization recorded in UniProt, 33  Using STRING, the 47 cellular and 63 secreted proteins were arranged into functional clusters. As can be seen in Figure 4A, only a small cluster of cytoskeletal proteins linked to focal adhesion was observed  (Table 1).

| Comparing metabolic processes between early and late feeding
Of the differentially secreted proteins of T14-T18, 92% was up-regulated (35/38), whereas for T18-T22 this was 56% (23/39). However, of the 24 overlapping proteins, the vast majority was up-regulated during both F I G U R E 2 Subgroups of differentially changed proteins during early feeding (T14-T18) and late feeding (T18-T22) phase. For the cellular proteins, there were 82 proteins differentially expressed during early feeding and 47 during late feeing, of which 21 cellular proteins were overlapping. For the secretome, there were 62 proteins differentially expressed during early feeding and 63 during late feeding, of which 24 secreted proteins were overlapping F I G U R E 3 Functional clusters of differential proteins during early feeding (T14-T18). A, The 82 significantly changed cellular proteins during the early feeding period. B, The 62 significantly changed secreted proteins during the early feeding period. Potential clusters are indicated by a dashed line 82 cellular proteins 62 secreted proteins

A B
periods. Notably, 20 of the 24 proteins had a significant but lower fold change (FC) during T18-T22 than during T14-T18 ( Figure 5). The four proteins that behaved differently were ADAM9, C4B, SEMA7A and PDIA4.
It suggests that there is a levelling off of protein secretion over time.
Proteins involved in the metabolism of glucose and fatty acids which significantly changed during the early feeding or late feeding period are given in Table 2 Note: Pathways were analysed by DAVID. Pathways in bold are overlapping during the early and late feeding period.

| D ISCUSS I ON
In the present study, we investigated the changes over time of the cellular proteome and of the secretome of human SGBS adipocytes under conditions of high glucose and high insulin. We identified 393 cellular proteins and 246 secreted proteins for further analysis.
Pathway analysis, functional clustering analysis, metabolic proteome changes and morphologic characterization of the adipocytes allowed us to determine time-dependent changes in the molecular and metabolic processes of the adipocytes.
Early feeding of mature adipocytes with high glucose in the medium seemed to promote the production of pyruvate  energy consumption towards the production of ECM proteins essential for survival of the adipocytes. 35,36 Because several important collagen-modifying enzymes (PCOLCE, PCOLCE2, P3H1, P4HA4, PLOD1, PLOD3) do not significantly change during the early nor the late feeding, a funnelling of the translational energy consumption seems more likely, but this has to be further investigated.
The which is an inhibitor of cysteine proteases, and with PSAP which can ameliorate the inhibitory activity of CST3. 38 In the late stage, CTSB, CTSD and PSAP are further up-regulated but the inhibitor CST3 is not. Therefore, it is no longer part of the cluster, but another protease CTSA becomes part of it. It suggests that during the late feeding phase the ECM does not anymore have to grow, but needs to be maintained.
Since adipocyte overgrowth in vivo may be accompanied by the development of insulin resistance, 12,39,40 we checked the insulin sensitivity of our cells by determining the phospho-AKT/AKT ratio at T14, T18 and T22. The ratio at T22 was reduced by 35% as compared to T14. Although this was not significant (P = .14; Figure   S1), it suggests that the cells are developing insulin resistance over time. Notably, the above-mentioned measurement does not provide the real cell ability to respond to insulin as we did not compare it to insulin-starved cells. Therefore, it remains possible that already in the first four days there was a reduction of insulin sensitivity. This could explain why two insulin-stimulated enzymes, ACLY and HK1, are down-regulated during T14-T18. Partial insulin resistance could reduce the production of cytoplasmic acetyl-CoA from citrate released by the mitochondria, where it is produced from pyruvate. As such, lipid production is already reduced in the early feeding stage, which could be regarded as a cellular response to limit overgrowth.
Notably, during both periods of high glucose and high insulin feeding various factors of the complement system were significantly altered in abundance. During T14-T18, we found C1R, C1S, C3, CFD, C4B and CLU strongly up-regulated (FC = 1.9-4.6), whereas during T18-T22 up-regulation of C3, C4B and CLU (FC = 1.6-3.7) continued together with up-regulation of CFB (FC = 3.2) and down-regulation of CFH (FC = −1.7). For the moment, we do not know the biological relevance of these changes, but it is tempting to suggest that it somehow relates to the increased inflammatory nature of hypertrophic AT. 41,42 It should be noticed that the culture conditions that we used here with high glucose and high insulin are not directly comparable with in vivo conditions observed in cases of glucose intolerance or diabetes. However, our in vitro system could shed light on the changes of adipocytes as they move from the mature state (T14) to a state with maximum lipid load (T22). As such, the observed changes could mimic what happens in vivo during the development of adipocyte hypertrophy.
In summary, in the early stage the adipocytes seem to prefer pyruvate as energy source, whereas beta-oxidation is down-regulated supporting lipid loading. Also, glycolysis is being limited which is accompanied by reduction of protein translation. Over time, lipid loading of the cells reduces paralleled by a reduction of the triglyceride synthesis capacity. Consequently, fusion becomes relatively more important for growth of lipid droplets during the late stage.
Nevertheless, ECM formation is promoted probably to protect the lipid-loaded cells against mechanical rupture.

| CON CLUS IONS
In conclusion, the present in vitro system provides insight into the molecular and metabolic changes of mature adipocytes under conditions of high glucose and insulin, which may help to understand the process of in vivo adipocyte hypertrophy during the development of obesity. Here, we have used SGBS cells, but similar studies can now be performed in primary adipocytes or induced adipose tissue-derived stem cells (iASCs) of both subcutaneous and visceral adipose tissue to link depot-specific proteome and secretome changes of overgrowing adipocytes to metabolic consequences in humans.

ACK N OWLED G EM ENTS
We would like to thank Prof. Dr Med. Martin Wabitsch (University of Ulm in Germany) for kindly donating the human SGBS cell line.

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.