We identified a total of 5022 peptides (
Supplementary Data Table S1) mapped to 1050 proteins in at least one of the 12 samples (3 replicates for each of the 4 condition). We filtered these proteins to remove those that have no quantifiable MS1 peak intensity, and 0 or 1 MS/MS count. We then only considered proteins that were identified in at least 2 of the 3 biological replicates for downstream statistical analysis. After applying these filtering steps, we have a final list of 693 proteins (
Supplementary Data Table S2). Functional classification of these proteins showed that the list was diverse including kinases, transcription and translational factors, transporters, carbohydrate and sugar metabolism, structural proteins, and enzymes (
Supplementary Data Table S2). These proteins belong to diverse cellular compartments, including mitochondrion, nucleus, cytosol, extracellular exosomes, integral component of membrane, plasma membrane, and ER (
Supplementary Data Table S2). Using a threshold of
p value ≤ 0.05, we identified 78 proteins that were significantly different in cells treated with Ins only, TNF only, or both, compared to the control (
Table 1). Most quantitative proteomics methods use fold changes to determine differentially expressed proteins, and a 2-fold increase or decrease is the most commonly applied threshold. However, a fold change to determine altered proteins is confounded by biases in protein abundances [
40]. For example, a 1.5 or 2-fold increase can be highly stringent for highly abundant proteins, but for low abundance proteins, this level of change might simply represent technical noise. Therefore, rather than relying on fold changes, we used the one-way ANOVA test to determine differentially expressed proteins and used
p values of ≤ 0.05 as a cut-off to determine significantly altered protein expression between the treatments. We found that many significantly different proteins across experimental groups have fold changes ≤1.5 (
Supplementary Data Table S3). Of the 78 significant proteins, the expression of 18 proteins was at least 2-fold higher and the expression of 28 proteins was at least 2-fold lower in at least one of the 3 treatment groups (
Supplementary Data Table S3). Changes in expression of 32 significantly different proteins was less than 2-fold and 24 proteins were uniquely identified in Ins + TNF (
Figure 2B and
Supplementary Data Table S3). Among the 78 differentially expressed proteins, the expression of 20 proteins was up and 30 proteins was down in all the treatments compared to the control (
Table 1). Eight (8) proteins were up and 4 proteins were down in Ins only treatment, 3 proteins were up and 5 proteins were down in TNF only, and 4 proteins were up and 4 proteins were down in Ins + TNF (
Table 1). The expression of 7 proteins was detected only in the control, 5 proteins only in Ins, 4 proteins only in TNF, and 15 proteins only in Ins + TNF (
Figure 2B). The four proteins elevated in TNF are Sigma non-opioid intracellular receptor 1 (SIGMAR1, OPRS1), 60S ribosomal protein L23 (RPL23), 40S ribosomal protein S28 (RPS28) and ATP dependent RNA helicase (DDX5). Their direct or indirect role or response to TNF is currently unknown. The SIGMAR1 is an endoplasmic reticulum (ER)-resident transmembrane protein which functions in lipid transport from the ER and is involved in a wide variety of disorders, including depression, drug addiction, and pain [
41]. The elevated expression of RPL23 and RPS28 might suggest that cells treated with TNF might require a more efficient translational machinery by regulating ribosome biogenesis and global protein synthesis [
42]. The increased expression of DDX5 is quite interesting. DDX5 is known to participate in all aspects of RNA metabolism ranging from transcription to translation, RNA decay, and mRNA processing [
43]. Its role in cell cycle regulation, tumorigenesis, apoptosis, cancer development, and adipogenesis has been well established [
43]. Understanding how elevated expression of DDX5 is linked to TNF treatment will provide new information about TNF-induced cellular outcomes of adipose tissues.
Figure 3A shows the heat map of the significant proteins in each group, which showed a 2-fold change in abundances in at least one of the 3 treatments. The PCA analysis was performed to project LFQ based proteome measurements into a two-dimensional data space (
Figure 3B). We applied PCA to 78 significant proteins that were quantified in each of the condition. The component 1 of the PCA (PC1) accounts for 42.6% of the variability, and clearly discriminates proteins among different treatment groups. The PC2 accounted for 20.3% of the total variations, and altogether, all 2 components accounted for 62.8% of the total variation. Proteins expressed in control segregated from the others, and proteins expressed in Ins only, TNF only, and Ins + TNF were also clearly discrete from each other.
Figure 3B also shows that the distance between replicates within each group is much smaller than the separation between the groups, supporting reproducible LC-MS/MS analysis and peptide quantification, a prerequisite for accurate label-free protein quantitation
Figure 4 shows the top 5 GO biological processes and molecular functions of the 78 significantly up- or down-regulated proteins (
p ≤ 0.05) in one or more of the 3 treatments compared to the control. For biological processes, down-regulated proteins were enriched for metabolic process, oxidation-reduction process, and TCA cycle in Ins and TNF, and ion transport and regulation and ATPase activity proteins were enriched in Ins + TNF (
Figure 4A). With regard to molecular functions, down-regulated proteins were enriched for oxidoreductase activity, electron carrier activity and fatty-acyl-CoA binding activity in Ins only and TNF only treatments, and ATPase activity, voltage gated anion channel and prion activity were enriched in Ins + TNF (
Figure 4C). Among the up-regulated proteins, those involved in biological processes, such as platelet degradation, translation, glycolysis, and the response to amino acids were enriched in one or more of the treatments (
Figure 4B). For molecular functions, proteins involved in homodimerization, calcium ion binding, structural constituents of ribosomes, and the extracellular matrix were enriched among the up-regulated proteins (
Figure 4D). The complete list of all the GO molecular functions, biological processes, and cellular components for all 693 proteins and 78 differentially expressed proteins are shown in the
Supplementary Data Table S2 and S3. The analysis of cellular components of 693 proteins by GO annotation showed that ~30% of these proteins were resident in mitochondria (
Supplementary Data Table S2). The same analysis showed >50% of the differentially expressed proteins (
p ≤ 0.05) resident in mitochondria including mitochondrial matrix and mitochondrial inner membrane (
Supplementary Data Table S3), suggesting that relatively higher mitochondrial resident proteins were affected by the treatments. Other major cellular components of differentially expressed proteins included extracellular exosomes, focal adhesion, cytoplasm/cytosol, lysosome, and cell-cell adhesion junction.