ARTICLE pubs.acs.org/jpr
Proteomic Analysis of Tumor Necrosis Factor-Alpha (TNF-α)-Induced L6 Myotube Secretome Reveals Novel TNF-α-Dependent Myokines in Diabetic Skeletal Muscle Jong Hyuk Yoon,† Parkyong Song,† Jin-Hyeok Jang,‡ Dae-Kyum Kim,† Sunkyu Choi,† Jaeyoon Kim,† Jaewang Ghim,† Dayea Kim,† Sehoon Park,† Hyeongji Lee,† Dongoh Kwak,† Kyungmoo Yea,† Daehee Hwang,‡,§,# Pann-Ghill Suh,†,‡ and Sung Ho Ryu*,†,‡,§ †
Division of Molecular and Life Sciences, ‡School of Interdisciplinary Bioscience and Bioengineering, §Division of Integrative Biosciences and Biotechnology, and #Department of Chemical Engineering, Pohang University of Science and Technology (POSTECH), Pohang, Kyungbuk, Republic of Korea
bS Supporting Information ABSTRACT: There is a strong possibility that skeletal muscle can respond to irregular metabolic states by secreting specific cytokines. Obesity-related chronic inflammation, mediated by pro-inflammatory cytokines, is believed to be one of the causes of insulin resistance that results in type 2 diabetes. Here, we attempted to identify and characterize the members of the skeletal muscle secretome in response to tumor necrosis factor-alpha (TNF-α)-induced insulin resistance. To conduct this study, we comparatively analyzed the media levels of proteins released from L6 skeletal muscle cells. We found 28 TNF-α modulated secretory proteins by using separate filtering methods: Gene Ontology, SignalP, and SecretomeP, as well as the normalized Spectral Index for label-free quantification. Ten of these secretory proteins were increased and 18 secretory proteins were decreased by TNF-α treatment. Using microarray analysis of Zuker diabetic rat skeletal muscle combined with bioinformatics and Q-PCR, we found a correlation between TNF-α-mediated insulin resistance and type 2 diabetes. This novel approach combining analysis of the conditioned secretome and transcriptome has identified several previously unknown, TNF-α-dependent secretory proteins, which establish a foothold for research on the different causes of insulin resistance and their relationships with each other. KEYWORDS: proteomics, insulin resistance, chronic inflammation, diabetes, secretory proteins, TNF-α
’ INTRODUCTION Skeletal muscle is now believed to be a tissue that produces and releases cytokines, which are named “myokines”.1 In addition to its traditional role as an insulin-dependent plasma glucose regulator via GLUT4, it is well-known that contracting skeletal muscle secretes several proteins such as interleukin (IL)-6 and IL-8 to protect the body from the damages of chronic disorders such as type 2 diabetes and cardiovascular disease.2 5 IL-15 has recently been identified as a myokine that functions in metabolic organs: it has been reported to increase glucose uptake and lipid oxidation in skeletal muscle and decrease lipogenesis in liver and white adipose tissue.6 Likewise, there is an increasing body of evidence that implies skeletal muscle derived cytokines may play a critical role in body homeostasis, but there is less systematic analysis for conditioned muscle secretomes. Furthermore, proteomic analysis of the skeletal muscle secretome, which has been limited to primary cell culture, requires the employment of additional criteria to clarify the results and relate them to in vivo conditions. Obesity-related inflammation is now believed to be a major cause of insulin resistance.7,8 It is still unclear that how obesity r 2011 American Chemical Society
promotes insulin resistance, but several results have converged to highlight the role of inflammation.9 Furthermore, there are reports of the association of increased proinflammatory cytokines with obesity, which demonstrate the strong inflammatory bases of obesity and associated metabolic diseases.10 Inflammatory regulators and mediators including tumor necrosis factoralpha (TNF-α), IL-6, IL-1b, and monocyte chemotactic protein-1 (MCP-1) are elevated in experimental murine models of obesity and in obese humans.9,10 It is now well-accepted that chronic low grade obesity-induced inflammatory responses that lead to activation of protein kinases, such as IkB kinases and Jun kinases, play an important role in the etiology of insulin resistance.11 To date, it has been clearly reported that TNF-α is an obesity related proinflammatory cytokine. TNF-α, produced by both adipocytes and macrophages, is known to be upregulated in a variety of obesity models and in obese humans.12 18 Furthermore, mice lacking TNF-α or TNF receptors showed improved insulin sensitivity in models of obesity.19 In this regard, TNF-α is believed Received: June 14, 2011 Published: October 25, 2011 5315
dx.doi.org/10.1021/pr200573b | J. Proteome Res. 2011, 10, 5315–5325
Journal of Proteome Research to be a major upregulated and functioning cytokine under obesity-related inflammatory conditions, which induces insulin resistance in skeletal muscle tissue. We have studied the secretion profile of proteins from skeletal muscle and found that secretion may change in response to various stimuli, depending on which physiological responses are being conveyed, such as hyperinsulinemia.20 In particular, the intimacy with which skeletal muscle tissue pairs with insulin to regulate plasma glucose levels suggests that the protein secretion pattern from skeletal muscle may be different for different insulin resistance states. After considering the limited body of known information, we performed a proteomic analysis of proteins in media conditioned by myotube cultures that were either left untreated or treated with TNF-α to induce insulin resistance. With this experimental approach, we sought to better characterize the skeletal muscle secretome and to identify skeletal muscle-derived proteins whose secretions are modulated by TNF-α. Furthermore, the transcriptome of the skeletal muscle of the Zuker diabetic fatty rat was analyzed to find novel correlating points with the data from the secretome. This work represents the first report to not only characterize the secretory proteins produced by myotubes in response to insulin resistance but is also the first report that combines proteomic analysis between a secretome and a transcriptome.
’ EXPERIMENTAL SECTION Cell Culture and Differentiation
Cell culture and differentiation were carried out as described previously.20 In brief, Rat L6 GLUT4myc skeletal myoblast cells were cultured in the alpha formulation of minimum essential medium supplemented with 10% FBS, 2 mM glutamine, 50 μg/mL streptomycin, and 50 μg/mL penicillin. To induce differentiation, cells (4 104 cells/mL) were placed into α-MEM supplemented with 2% FBS for 5 days, with media refreshed every 48 h. Differentiation status was monitored under a microscope (LSM 510 Meta; Zeiss; Germany). Establishment of Insulin Resistance Condition
To induce TNF-α-mediated insulin resistance, L6 myotubes were incubated with 2 ng/mL of TNF-α (R&D Systems, Minneapolis, MN) for 4 days. Media were changed every 24 h. On the fourth day, cells were washed three times with PBS and incubated in α-MEM medium supplemented with 0.1% BSA in the absence of TNF-α for 90 min. GLUT4 Translocation Measurement
After treatment with TNF-α, 100 nM insulin was added to the cells for 20 min before harvest or analysis of GLUT4 translocation. The amount of GLUT4myc on cell surfaces was determined using an antibody-coupled colorimetric absorbance assay, as was previously described.22 Briefly, after stimulation, L6 cells were treated with antimyc antibody (1:100) for 60 min, fixed with 4% paraformaldehyde for 10 min, and then incubated with HRPconjugated goat anti-rabbit IgG (1:1000) for 1 h. Cells were then washed three times, and 1 mL of ophenylenediamine reagent (0.4 mg/mL of o-phenylenediamine dihydrochloride and 0.4 mg/mL of urea hydrogen peroxide) was added for 30 min at room temperature. The reaction was stopped with 3 N HCl, and optical absorbance of the supernatant was measured at 492 nm. Data are expressed as mean ( SEM. Differences between two groups were assessed using the unpaired two-tailed t test and among more than two groups by analysis of variance (ANOVA).
ARTICLE
Preparation of Conditioned Media
After treatment with TNF-α, L6 myotubes were washed three times with completely unsupplemented α-MEM (no serum, phenol red, or antibiotics). Conditioned media were prepared by adding the unsupplemented α-MEM to the myotubes and incubating them for 5 h in a 37 °C CO2 incubator. Conditioned media were then collected and centrifuged at 3500 rpm for 10 min (Combi-514R; Hanil Science Industry; Korea). Supernatants were collected for the next process. Sample Preparation for Mass Analysis
The resulting supernatants were dried and applied to a hydrophilic lipophilic balance (HLB) extraction column (Waters, MA) for desalting and decontaminating.20,21 Proteins were eluted with 1% TFA in acetonitrile (ACN). For tryptic digestion, dried samples eluted from HLB were reduced using 10 mM DTT in 50 mM ammonium bicarbonate (ABC), and then alkylated by 100 mM iodoacetamide in 50 mM ammonium bicarbonate. Finally, each sample was treated with trypsin (Promega; Madison, WI) for 12 h at 37 °C and dried. Sample Preparation of Skeletal Muscle of ob/ob Mice
Sixteen weeks old male ob/+ and ob/ob mice (Korea Research Institute of Bioscience and Biotechnology, Korea) were killed by vertebral fracture. Skeletal muscle was dissected immediately and washed with ice-cold 1x PBS, then frozen in liquid N2. After the tissue was ground with a mortar and pestle, the resulting powders were put in microtubes. To make protein extracts, each powder was dissolved in an appropriate volume of 10 mM TrisHCl, pH 7.4, which also contained: 150 mM NaCl, 1% Tx-100, 1 mM EDTA, and 2 mM PMSF. Those powders were sonicated 10 6 times at 30% amplitude for 5 s on ice. Protein concentration was determined by Bradford reagent assay. LC ESI MS/MS
Tryptic digests (secretome peptides) were analyzed by a linear trap quadrupole (LTQ) XL mass spectrometer (Thermo Fisher Scientific, San Jose, CA, USA) equipped with a nano-HPLC system (Eksigent, Dublin, CA, USA) for nanoflow chromatography. For preconcentration and desalting, tryptic digests dissolved by solvent A (0.1% formic acid in 2% ACN) were injected on the trapping column (3 cm 75 μm i.d.) packed with C18 (5 μm) and washed for 9 min with 1 μL/min flow rate of solvent A. After the analytical column was placed in line with the trapping column, the peptides loaded in the trapping column were eluted using a mobile phase gradient with solvent A and B (0.1% formic acid in 98% ACN) for 150 min. The peptides were separated by using a reversed phase analytical column (10 cm 75 μm i.d.) packed with C18 resin (5 μm). The peptides were subjected to a 0 35% solvent B gradient, with a flow rate of 260 nL/min, for 80 min. The eluent was then introduced into the LTQ mass spectrometer by a nanoion source with 1.9 kV of electrospray voltage. The analysis method consisted of a full MS scan with a range of 400 1800 m/z, and data-dependent MS/MS (MS2) on the five most intense ions from the full MS scan. Mass spectrometer calibration was performed with the proposed calibration solution following manufacturer’s manual. This study used three biological and technical replicates for each condition in proteomic analysis. SEQUEST Database Searching
We created the peak lists from original RAW files with Bioworks Browser Rev.3.1 SR1 (Thermo Fisher Scientific) with the mininum peak intensity of 1000. We performed peptide identification
5316
dx.doi.org/10.1021/pr200573b |J. Proteome Res. 2011, 10, 5315–5325
Journal of Proteome Research from each experiment using SEQUEST version v.27 (Bioworks Browser Rev.3.1 SR1, Thermo Fisher Scientific), matched to a rat database (IPI version 3.73; 39711 sequences) with decoy sequences (reverse of target database). Carboxyamido-methylation of cysteine was chosen as a fixed modification, and oxidation of methionine was chosen as a variable one. The mass tolerance was set at 3.0 amu for the precursor ions and at 1.0 amu for the fragment ones with tryptic specificity. Two missed cleavages were allowed. Peptide identification was done using DTAselect 2.0.3924 with the default parameters and false discovery rate (FDR) = 0.01 for peptide level. Multiple or ambiguous IDs were not allowed, and the decoy database hits were removed from the results. We also removed frequently observed contaminants such as porcine trypsin and human keratins. Statistics for Selection of TNF-α Modulated Secretory Proteins
The normalized spectral index (SIN) was calculated to estimate the abundance of each protein using the Perl script published in the previous study.25 The spectral index of each protein was normalized with the sum of spectral indexes of all proteins identified in the sample and the amino acid length of each protein, as described in the previous study.25 By comparing the normalized spectral indexes between the TNF-α stimulated and control samples, we then identified differentially secreted proteins as the ones with FDR < 0.05 using a previously reported integrative statistical hypothesis testing,26 which combines FDRs obtained from two-tailed t test and median ratio test using Stouffer’s method.27
ARTICLE
reverse transcribed into cDNA. Quantitative real-time PCR analysis was carried out using HotStart-IT SYBR Green and Bio-Rad iCycler iQ. Target gene primers are presented in Table S3, Supporting Information. The relative quantitation of mRNA was calculated by the comparative Ct method after normalization to glyceraldehyde-3-phosphate dehydrogenase. Nonspecific amplification was excluded by confirming single melting curve patterns and ethidium bromide staining on 2% agarose gels. Samples were analyzed pairwise by Student’s t test. Western Blotting
Western blots were performed as described previously.22,23 After separation by SDS-PAGE, the proteins were transferred to nitrocellulose (NC) membranes using the Hoefer wet transfer system and blocked by TTBS containing 5% skimmed milk for 30 min and incubated with the antibodies for 4 h. After the membranes were washed three times with TTBS, the blots were incubated with horseradish peroxidase-conjugated anti-mouse or anti-rabbit antibody for 1 h. The membranes were then washed with TTBS and developed using ECL. We performed Western blots with the antibodies against p-IRS (Tyr989, homemade), p-Akt (SAB4300334, Sigma-Aldrich Co., MO), p-AMPK (#2531, Cell Signaling Technology, MA), p-ACC (07-303, Millipore, MA), ACC (#3662, Cell Signaling Technology, MA), Clusterin (sc-8354, Santa Cruz Biotechnology, CA), IGFBP-4 (sc-6005, Santa Cruz Biotechnology, CA), Nucleobindin-2 (sc-65160, Santa Cruz Biotechnology, CA) DJ-1 (ab18257, Abcam plc., MA) and Actin (cat# 691001, MP Biomedicals, Solon, OH).
Microarray Data Analysis
A published microarray data set in the GEO database (GSE1080) was chosen for comparison with our proteomics data set. This microarray data were collected using a Rat 5K cDNA microarray with 5184 probe sets that can be mapped to 3380 Entrez gene IDs.28 The microarray data compared the samples obtained from Zucker diabetic fatty (ZDF) and Zucker lean control (ZLC) rats. Among all microarray data in the GSE1080, we only used the four muscle data sets collected from rats at 12 weeks of age in our study. We first normalized the expression values measured from ZDF and ZLC muscle samples using the quantile normalization method.29 We then identified 351 differentially expressed genes (DEGs) between ZDF and ZLC using statistical analysis. The DEGs were selected as the genes with p-values less than 0.01 for the two-tailed t test and an absolute fold change larger than 1.5. For the comparison between the proteomic and microarray data, we converted the IPI of our proteomics data and the probe set IDs in the microarray data into Entrez Gene IDs. Bioinformatic Analysis
Using DAVID software,30 the enrichment analysis of Gene Ontology biological processes (GOBP) was performed for the individual clusters: MS-Secretome, Microarray-Secretome, and Microarray-Nonsecretome. For the comparison among the three groups, we first identified those GO terms that were enriched in at least two of the groups with p < 0.1. Among these GO terms, we removed the ones from high GO levels (1 4) from our analysis. Finally, we reconstructed a network model using the secretory proteins using Ingenuity pathway analysis (http:// www.ingenuity.com/) as described previously.20 Q-PCR Analysis
Medium was removed for secretome analysis, and total RNA was isolated using TRIzol reagent according to the manufacturer’s protocol. Three micrograms of the resulting RNA was
’ RESULTS Condition Establishment of Insulin Resistance
Our strategy for the research is outlined in Figure 1. To establish the insulin-resistant condition, we subjected myotubes to treatment with varying amounts of TNF-α for 4 days. After pretreating with TNF-α, the myotubes were exposed to insulin stimulation to induce GLUT4 translocation to the plasma membrane. TNF-α treatment clearly inhibited GLUT4 translocation (Figure 2A). We selected a TNF-α dose of 2 ng/mL, which inhibited GLUT4 translocation to
