Pancreatic cancer cell-derived IGFBP-3 contributes to muscle wasting

Background Progressive loss of skeletal muscle, termed muscle wasting, is a hallmark of cancer cachexia and contributes to weakness, reduced quality of life, as well as poor response to therapy. Previous studies have indicated that systemic host inflammatory response regarding tumor development results in muscle wasting. However, how tumor directly regulates muscle wasting via tumor-derived secreted proteins is still largely unknown. Methods In this study, we performed bioinformatics analysis in two datasets of pancreatic ductal adenocarcinoma, which causes cancer cachexia and muscle wasting with the highest prevalence, and uncovered that IGFBP3, which encodes IGF-binding protein-3 (IGFBP-3), is dramatically up-regulated in pancreatic tumor samples. We also verified the wasting effect of IGFBP-3 on C2C12 muscle cells with biochemical and genetic assays. Results IGFBP-3 potently leads to impaired myogenesis and enhanced muscle protein degradation, the major features of muscle wasting, via IGF signaling inhibition. Moreover, conditioned medium from Capan-1 pancreatic cancer cells, which contains abundant IGFBP-3, significantly induces muscle cell wasting. This wasting effect is potently alleviated by IGFBP3 knockdown in Capan-1 cells or IGFBP-3 antibody neutralization. Strikingly, compared to muscle cells, IGF signaling and proliferation rate of Capan-1 cells were rarely affected by IGFBP-3 treatment. Conclusions Our results demonstrated that pancreatic cancer cells induce muscle wasting via IGFBP-3 production. Electronic supplementary material The online version of this article (doi:10.1186/s13046-016-0317-z) contains supplementary material, which is available to authorized users.


Background
Patients with pancreatic cancer often develop the most severe degrees of cachexia that is highly associated with cancer death [1]. Clinically, cancer cachexia is defined as an unintentional 10 % loss of body weight over 12 months [2]. Previous studies have indicated that the progressive loss of skeletal muscle, termed muscle wasting, is a key phenotype of cancer cachexia and results in weakness, reduced ambulation, diminished quality of life, poor response to therapy, as well as death due to respiratory failure or infection [3]. However, approved effective treatments for muscle wasting in pancreatic cancer patients are still missing. Thus, understanding the molecular mechanisms of muscle wasting will provide novel insight into developing targeted therapies and improving the quality of life for pancreatic cancer patients and, possibly, for other malignancies.
There are increasing evidences that both impaired myogenesis and increased muscle protein degradation contribute to muscle wasting during cancer cachexia [4][5][6]. Systemic hormones have been shown to regulate these biological processes. For example, TGFβ superfamily members, including activin A, GDF15, as well as Myostatin, can cause muscle loss through SMAD signaling [4,7,8]. Systemic inflammatory cytokines, including TNFα, IL-1α, IL-1β, IL-6 and related ligands haven been shown to cause muscle wasting in both mouse models and human samples [9]. Growing studies across different species indicated that tumor-derived hormones also play essential roles for muscle wasting. For example, conditioned medium from pancreatic cancer cells that contains numerous cancer-derived peptides, including Myostatin and activin A, is sufficient to cause muscle wasting [4,10,11]. In addition, tumor-derived parathyroid-hormone-related protein (PTHrP) has been shown to induce muscle wasting and lipid depletion in a mouse model [12]. An insulin-like binding protein, ImpL2, is secreted from tumor-like cells and impairs muscle function and systemic tissue growth via inhibition of IGF-like signaling in Drosophila [13][14][15]. Thus, revealing how tumor-derived secreted proteins cause muscle wasting will shed the light on novel mechanisms of tumor-host interaction regarding cancer cachexia.
Mammalian insulin-like growth factor binding protein (IGFBP) 1-7 and Drosophila ImpL2 share high homology in structures or functions. Classically, IGFBPs bind to insulin-like growth factors (IGFs) to stabilize the complex and enhance the half-life and distribution of IGFs to target tissues. On the other hand, excess IGFBPs restrain the bio-ability of IGFs to their receptors and suppress intracellular IGF signaling that is required for myogenesis and myotube atrophy [16][17][18]. The notion is further supported by the evidence that endogenous IGFBP-5 has been shown to promote myogenesis via activation of IGF-2/AKT/FoxO signaling, whereas, IGFBP-5 overexpression tremendously causes retarded muscle development [19,20]. In addition to IGF signaling, IGFBPs also regulate cell biological processes via other signaling pathways, including NF-κB, TGF-β, JAK/STAT, and heat shock protein signalings [21,22]. Notably, injection of IGF-1/IGFBP-3 complex improves weight loss in tumor-bearing mice [23]. However, whether excess IGFBPs are secreted from tumors to regulate muscle wasting is far less established. In this study we analyzed the gene expression profile and identified that IGFBP3 is dramatically induced in pancreatic tumors. We further demonstrated that IGFBP-3, which is abundantly produced in pancreatic cancer cells, causes muscle wasting through both impaired myogenesis and enhanced myotube protein degradation via, at least, inhibition of IGF/PI3K/AKT signaling. Thus we propose that pancreatic tumors result in muscle wasting via secretion of IGFBP-3.

Secreted protein genes are induced in pancreatic tumors
In order to study whether secreted proteins that likely regulate tumor-host crosstalk are up-regulated in pancreatic tumors, we analyzed the gene expression profile in both pancreatic tumors and normal pancreatic tissues. Here two different datasets (GSE15471, 36 normal-tumor sample pairs, and GSE16515, 36 tumor samples and 16 normal samples) were used for bioinformatics analysis. Subsequently, we found that 756 genes were up-regulated and 160 genes were down-regulated in both pancreatic tumor samples compared to normal tissues (fold change > = 2 and FDR < 0.05) (Fig. 1a, Additional file 1: Figure S1A-S1C, Additional file 2: Table S1 & Additional file 3: Table S2). Gene ontology (GO) enrichment analysis of signaling pathways revealed that overlapping up-regulated genes involved in TGF-β, integrin, PDGF, as well as p53 signalings, are significantly enriched (Additional file 1: Figure S1D). Further, significant enrichment of multiple tumorigenic biological processes, including inflammatory response, cell proliferation, angiogenesis, extracellular matrix, cell migration, immune response, and cell adhesion, was observed in overlapping up-regulated genes (Fig. 1b, up). Interestingly, GO enrichment analysis regarding cellular component also revealed significant enrichment of "plasma membrane" and "extracellular region" where secreted protein are processed and secreted (Fig. 1b, down). However, GO enrichment analysis of overlapping down-regulated genes uncovered that only "amino acid metabolism" biological process was significantly enriched (Additional file 1: Figure S1E).

IGFBP-3 is transcriptionally increased in pancreatic tumor
Previous studies have indicated that Hippo-induced cancer-like cell over-proliferation leads to muscle dysfunction and systemic tissue wasting via production of ImpL2, an insulin-like binding protein, and inhibition of insulinlike signaling in fly model [14]. To address whether the mechanism is conserved in pancreatic cancer cachexia, we measured the expression levels of human ImpL2 homologs in pancreatic tumors. Six structural homologs of ImpL2, IGFBP 1-6, and one functional homolog, IGFBP-7, exist in human [16,30]. Interestingly, both IGFBP3 (two probes, more than eight folds) and IGFBP5 (three probes, more than 3 folds) are dramatically increased in both pancreatic tumor datasets, whereas, IGFBP2 is significantly decreased in both datasets (Fig. 1e). IGFBP1 (more than five folds) is induced only in GSE16515 and IGFBP 4 & 7 (more than three & five folds) are induced only in GSE15471 dataset (Fig. 1e). As the IGFBP3 exhibited the highest induction in both pancreatic tumor datasets, we proposed that pancreatic tumors probably communicate to muscle cells via production of IGFBP-3.

IGFBP-3 impairs C2C12 myogenesis
Increasing evidence indicated that impaired myogenesis, including myoblast proliferation and myotube differentiation, contributes to muscle wasting [4][5][6]31]. Excess IGFBP5 has been shown to regulate myotube differentiation probably via modulation of IGF-2 signaling [19,20], however, whether IGFBP-3, which predominantly binds to IGF-1 and exhibits the greatest induction in pancreatic tumors, affects myogenesis is largely unknown. To address this question, we studied whether IGFBP-3 regulates myoblast proliferation and myotube Significantly enriched (EASE score < 0.05) GO terms were identified in overlapping up-regulated gene list (Up, biological process. Down, cellular component). c Venn diagram indicated the overlap in Human Atlas secreted-protein-gene database, GeneCards secreted-protein-gene database, and up-regulated gene list in both pancreatic tumor datasets. d Overlapping up-regulated genes in both datasets were indicated with basal expression level (probe signal) and fold change. e The differential expression patterns of cachexia-related and IGFBP family genes were indicated in heatmaps. The intensity reflected fold change differentiation. We incubated C2C12 myoblasts with different doses of synthetic IGFBP-3 and measured the myoblast proliferation rate. Interestingly, IGFBP-3 significantly inhibited C2C12 myoblast proliferation, which was indicated by reduced myoblast cell number, in a dose-dependent manner ( Fig. 2a-b). Then we incubated C2C12 myoblasts with different doses of synthetic IGFBP-3 from day 0 (D0) of differentiation and monitored myotube differentiation rate. Impressively, IGFBP-3 treatment potently reduced the number of MHC-positive differentiated myotubes in a dosedependent manner on day 4 (D4) of differentiation as compared with untreated controls (Fig. 2c). The differentiation rate indicated by cell morphology and the fusion rate indicated by nuclei number per myotube were also dramatically decreased by IGFBP-3 treatment ( Fig. 2d-e). We also measured the expression of muscle regulatory factors MyoD (myogenic differentiation 1) and Myogenin. On the D4 of differentiation, IGFBP-3 treatment significantly decreased mRNA levels of both MyoD and Myogenin (Fig. 2f). Collectively, our results indicated that IGFBP-3 impairs C2C12 myogenesis.

IGFBP-3 promotes C2C12 myotube protein degradation
The atrophy of skeletal muscle due to the elevated activity of proteolytic pathways and protein degradation is another important feature of cancer-induced muscle wasting [32]. To examine whether IGFBP-3 induces muscle wasting in myotube, we treated C2C12 myotubes on day 4 (D4) of differentiation with different doses of IGFBP-3 for 48 h. Impressively, IGFBP-3 treatment resulted in a distinct atrophic myotube phenotype (Fig. 3a) with a decrease in myotube diameter (Fig. 3b). IGFBP-3-induced atrophy was also evidenced by the loss of myotube protein in a dose-dependent manner (Fig. 3c). Moreover, the amount of protein synthesis that was quantified with [ 3 H] tyrosine incorporation was potently reduced after 24 h IGFBP-3 treatment (Fig. 3d). In contrast, the level of protein degradation indicated by [ 3 H] tyrosine loss after 24 h IGFBP-3 treatment was significantly elevated as compared to untreated controls (Fig. 3e), indicating that the proteolysis highly associated with myotube wasting is enhanced by IGFBP-3 treatment. The enhanced proteolysis could be caused by elevated activity of ubiquitin-proteasome pathway [4,33]; to assess this possibility, the amount of ubiquitinated protein after IGFBP-3 treatment was measured. Consistent to our hypothesis, immunoblot analysis using ubiquitin antibody revealed that IGFBP-3 increased protein ubiquitination in myotube in a dose-dependent manner (Fig. 3f). Thus, our results indicated that IGFBP-3 enhances ubiquitination associated proteolysis and reduces C2C12 myotube mass.

IGFBP-3 is required for Capan-1 pancreatic cancer cells to induce C2C12 myotube wasting
Treating muscle cells or adipocytes with conditioned medium from pancreatic cancer cells was used to mimic tumor-host interaction and identify potential tumorderived secret factors. Capan-1 pancreatic cancer cells have been established to cause muscle wasting [10,11]. We next assessed whether pancreatic cancer cells induce myotube wasting via IGFBP-3 production. We first measured the IGFBP-3 level in Capan-1 pancreatic cancer cells. Immunoblot analysis with a specific antibody against IGFBP-3 revealed that, compared to C2C12 myoblasts, Capan-1 cells produced abundant IGFBP-3 (Fig. 5a, up). Similarly, abundant IGFBP-3 was detected in conditioned medium from Capan-1 cells, but not in normal growth medium (Fig. 5a, down). Subsequently, compared to normal growth medium (Con), conditioned growth medium from Capan-1 cells (CM, see Methods) dramatically decreased IGF signaling in myoblasts (Fig. 5b) and inhibited myoblast proliferation ( Fig. 5c and Additional file 9: Figure S5B). In addition, conditioned differentiation medium from Capan-1 cells (CM) inhibited myotube differentiation (Fig. 5d, Additional file 9: Figure S5C-S5E and Additional file 5: Figure S6C) and promoted myotube protein degradation and atrophy ( Fig. 5e-g, Additional file 9: Figure S5F and  (Up). IGFBP-3 levels in normal growth medium (Con) and conditioned medium of Capan-1 cells infected with si-GFP (CM) or si-IGFBP-3 (CM-I) lentivirus were indicated by immunoblot (Down, non-specific band of IGFBP-3 antibody were used as loading control). b IGF signaling, which is indicated with pAKT levels, in C2C12 myoblasts that were treated with normal DMEM growth medium (Con) or conditioned growth medium from si-GFP (CM) or si-IGFBP3 (CM-I) Capan-1 cells for 24 h (UP); or with Con, CM, or CM that was pre-treated with 5 μg/mL or 10 μg/mL IGFBP-3 antibody for 30 min (CM + Ab) for 24 h. c Myoblasts were seeded at density of 20,000 cells/well and were grown with Con, CM, CM-I, or CM + Ab (10 μg/mL IGFBP-3 antibody) growth medium for 96 h. d Myotubes were differentiated in Con, CM, CM-I, or CM + Ab (10 μg/mL IGFBP-3 antibody) differentiation medium for 96 h. e-g C2C12 myotubes were normally differentiated for 96 h and then incubated with Con, CM, CM-I, and CM + Ab (10 μg/mL IGFBP-3 antibody) differentiation medium for 48 h. The protein degradation (f) and ubiquitinated protein levels (g) were measured. Data are presented as means ± SEM. * p < 0.05 Additional file 6: Figure S7C) compared to normal differentiation medium (Con).
Capan-1 cells produce multiple secreted proteins that might contribute to myotube wasting, thus our results indicated that pancreatic cancer cells induce myotube wasting via, at least, IGFBP-3 secretion.

Capan-1 pancreatic cancer cells can escape from IGFBP-3 inhibitory effects
Previous Drosophila studies indicated that cancer-like cells restrain host tissue growth via ImpL2 production, but these cancer-like cells escape from ImpL2 inhibitory effects [14]. We next examined whether Capan-1 pancreatic cancer cells escape from IGFBP-3 inhibition in a similar manner. We treated Capan-1 cells with different doses of IGFBP-3 and monitored IGF signaling and cell proliferation. Strikingly, in contrast to the fact that IGF signaling in C2C12 myoblast was significantly suppressed by IGFBP-3 at 0.1 μg/mL (Fig. 4a), IGF signaling in Capan-1 cells wasn't suppressed by IGFBP-3 unless at a very high dose, namely 10 μg/mL (Fig. 6a), suggesting that the sensitivity to IGFBP-3 in Capan-1 cancer cells is decreased by over 100 folds compared to C2C12 myoblasts. Furthermore, IGFBP-3 below 10 μg/mL failed to restrain ATP production, which is indicative of cancer cell proliferation, and to change the growth rate of Capan-1 cells (at all concentrations tested) (Fig. 6b-d). This result was also in strong contrast with the fact that 0.1 μg/mL IGFBP-3 is sufficient to impair C2C12 myogenesis and induce C2C12 myotube wasting. Thus, our results indicated that Capan-1 pancreatic cancer cells escape from IGFBP-3 inhibitory effects.

Discussion and conclusions
Pancreatic cancer patients have the highest prevalence and often develop the most severe degrees of cachexia including muscle wasting. One of the main pathogenic mechanisms underlying cancer-induced muscle wasting is the tumor-host interaction, thus examining the secreted proteins that are specifically produced in tumor cells and regulate muscle function will provide novel insights in the understanding of pancreatic cancer cachexia. We performed bioinformatics analyses in two different gene expression datasets of pancreatic ductal adenocarcinoma and mined the secreted protein genes that are dramatically induced in pancreatic tumors. Genes, which encode systemic inflammatory factors, including CRP, ILs, as well as TNF [9,24], are not consistently up-regulated in pancreatic tumors. This is consistent with the notion that these inflammatory factors are predominantly produced in host tissues. Similarly, our results also indicated that other cachexia-associated genes, like DCD, AZGP1, GDF15, and MSTN [4,25,28,29], are not predominantly produced in pancreatic tumors. They are mainly produced in liver, adipose tissue, and muscle. TGFB1 and TGFB2, which encode TGF-β1 and TGF-β2, respectively, are moderately elevated in pancreatic tumors (~3 folds, data not shown). Our results suggested that pancreatic tumor-derived TGFβs may contribute to cachexia and muscle wasting, even though a few studies indicated that functional TGF-βs are produced in multiple host tissues [39,40]. Importantly, INHBA that encodes activin A and activin BA is dramatically increased in pancreatic tumors (>10 folds) in both datasets. As activin A is highly associated with cancer cachexia [26], our results indicated that INHBA induction in pancreatic tumors might play a critical role in cancer-associated cachexia. Another tumor-derived secreted protein gene PTHLH, which encodes PTHrelated protein and induces muscle wasting [12], is upregulated only in GSE16515 dataset. Thus, our results demonstrated that the secreted protein genes induced in pancreatic tumors are valuable in uncovering tumorderived factors causing muscle wasting and cancer cachexia.
IGFBPs homolog, ImpL2, has been shown to be produced in cancer-like cells and to impair host tissue growth and muscle function via inhibition of insulinlike signaling in Drosophila model [14,41]. In this study we measured the expression patterns of all IGFBPs in two different datasets to validate the conserved regulation in pancreatic tumor samples. Interestingly, expression levels of IGFBP3 (>8 folds) encoding IGFBP-3 were dramatically induced in both pancreatic tumor datasets. The induction of IGFBP3 in pancreatic tumor samples was also observed in previous studies [42,43]. In addition, we also found that IGFBP-3 is abundantly produced in Capan-1 pancreatic cancer cells and secreted into culture medium. We further demonstrated that either exogenous IGFBP-3 or IGFBP-3enriched Capan-1 cell-conditioned medium potently enhances muscle wasting via both impaired C2C12 myogenesis and increased C2C12 myotube proteolyisis. Strikingly, IGFBP-3 deprivation in Capan-1 cell-conditioned medium, which is achieved by knockdown of IGFBP-3 expression in Capan-1 cells or specific IGFBP3 antibody neutralization, significantly improved the wasting effects in muscle cells. Thus, our results indicated that pancreatic cancer cells directly cause muscle cell wasting via IGFBP-3 production.
IGF-1 signaling stimulates muscle growth and protein synthesis, as well as proliferation and differentiation of satellite cells, and exerts anti-apoptotic effects on muscle cells to suppress proteolysis and inhibit the ubiquitin-proteasome system [17,[44][45][46]. IGFBP-3 has been shown to bind to IGF-1 and IGF-1/IGFBP-3 ratio in serum is essential for IGF-1 bio-ability and IGF-1 signaling [16]. Moderate amount of IGFBP-3 promotes IGF-1 stability in blood and its interaction with IGF1 receptor, as well as intracellular IGF-1 signaling. Injection of IGF-1/IGFBP-3 complex, not IGFBP-3 alone, into tumor-bearing mice attenuates cancer cachexia phenotypes, including weight loss and appetite, probably due to stabilized IGF-1 in blood and enhanced IGF-1 signaling [23]. However, excess IGFBP-3 prevents IGF-1 from binding to its receptor and inhibits IGF-1 signaling. In this study, we observed that both exogenous IGFBP-3 and IGFBP-3-enriched Capan-1 conditioned medium potently decrease IGF-1 signaling in muscle cells. Our results indicated that Capan-1 pancreatic cancer cells produce IGFBP-3 to restrain IGF-1 availability, suppress IGF-1/PI3K/AKT signaling in muscle cells, and induce muscle cell wasting.
Another important finding of our study is that we uncoupled IGFBP-3 mechanistic impacts on myogenesis and myotube protein degradation. Genetic enhancement of PI3K/AKT signaling managed to rescue IGFBP-3 inhibition of myogenesis, including C2C12 myoblast proliferation and C2C12 myotube differentiation; however, it failed to fully rescue IGFBP-3-induced C2C12 myotube proteolysis. It is possible that regulation of myotube proteolysis is much more complicated than myogenesis regarding IGFBP-3 function. IGF signaling controls myotube proteolysis via two important regulators downstream PI3K/ AKT cascade, FoxO1 and mTOR [17,47]. AKT phosphorylates FoxO1 and suppresses its transcriptional activity of ubiquitin ligases expression to inhibit proteolysis [17]. AKT also phosphorylates mTOR to promote protein synthesis via activation of 4E-BP1 and S6K [47]. It is well known that, in addition to IGF signaling, IGFBP-3 also triggers other pathways, like TGF-β, NF-κB, and JAK/ STAT signalings [21,[48][49][50], in a cell-context manner. Thus, we speculated that IGFBP-3 enhances myotube proteolysis via IGF-signaling-independent mechanisms as well. How other pathways are involved in IGFBP-3 regulation of myotube proteolysis will be another interesting question for our future research.
Several cytokines/hormones have been reported to modulate tumor cell growth in an autocrine manner [51][52][53]. However, in this study we indicated that growth rate of Capan1 pancreatic cancer cells is rarely affected by IGFBP-3. Another outstanding question in this field could be: how do pancreatic tumor cells survive while other host tissue growths are suppressed by IGFBPs? Previous studies mentioned that tumors or cancer-like cells survive from IGFBPs suppression 1) overexpressing IGF signaling components, like IGF1R, PI3K, as well as AKT, to potentiate intracellular IGF signaling; 2) overexpressing other growth signaling components, like EGFR, JAK/STAT, as well as TGF-β pathways, to compensate IGF growth signaling [14,16]. A plausible explanation is that, similar to Drosophila Yki-induced cancer-like cells [14], pancreatic cancer cells obtain enhanced IGF signaling independent of extracellular IGF1/IGFBP-3 impact. In addition, our bioinformatics analysis uncovered that genes encoding components of TGF-β, integrin, p53, and PDGF signaling pathways, are dramatically increased and significantly enriched in both pancreatic tumor datasets, suggesting that pancreatic tumor cells increase other growth signalings to counteract IGFBP-3 suppression.

Bioinformatics analyses of microarray results
Microarray data GSE16515 and GSE15471 were obtained from NCBI-GEO (www.ncbi.nlm.nih.gov/geo). Expression levels of probes were normalized using RMA and mapped in Affymetrix Human Genome U133 Plus 2.0. Gene expressions in pancreatic cancer samples with fold change > =2 and false discovery rate (FDR) < 0.05 were considered as significant. All comparisons were made between pancreatic cancer tissues and normal tissues. As for gene ontology enrichment analysis, the significantly up-and down-regulated genes were uploaded separately to DAVID Bioinformatics Resource [54]. The human genome U133 Plus was used as a background for the GO analysis. The GO terms with EASE score < 0.05 were selected for interpretation [55]. After finding overlapping genes/probes of datasets GSE15471 and GSE16515, Pearson's correlation analysis of fold change of each detected gene in each dataset was performed. Predicted secreted protein genes were annotated regarding gene lists that encode secreted proteins from both The Human Protein Atlas and (www.proteinatlas.org) and GeneCards (www.gene cards.org). Heatmaps of selected genes were made using ggplot2 package of software R.

Antibodies and reagents
Anti-MHC (MF20) and α-tubulin antibodies were obtained from the Developmental Studies Hybridoma Bank and Sigma, respectively. Anti-PTEN, phospho-AKT (S473), AKT and ubiquitin antibodies were purchased from Cell Signaling Technology. Anti-IGFBP-3 antibody was purchased from Santa Curz Biotechnology. All secondary antibodies for immunostaining and western blot were from Jackson Laboratory and Thermo Fisher Scientific, respectively. Recombinant human IGFBP-3 and IGF-1 proteins were obtained from R&D Systems. PI3K inhibitor LY294002 was from Abcam. shRNA lentivirus vectors for PTEN (TRCN0000002747) and IGFBP3 (TRCN0000072512) were purchased from Sigma.

Cell culture and treatment
Murine C2C12 myoblasts obtained from American Type Culture Collection (ATCC) were maintained in Dulbecco's modified Eagle's medium (DMEM) supplied with 10 % fetal bovine serum and antibiotics (50 U/ml penicillin and 50 μg/ml streptomycin), referred as growth medium. To induce myogenic differentiation, myoblasts were grown to reach 100 % confluence (day 0) and cultured with DMEM containing antibiotics and 2 % heat-inactivated horse serum, referred as differentiation medium. Capan-1 pancreatic cancer cells were obtained from ATCC and cultured in ATCC-formulated Iscove's Modified Dulbecco's Medium with 20 % fetal bovine serum and antibiotics. For CM (conditioned medium) preparation, Capan-1 cells were plated at a density of 50,000 cells/cm 2 , and after 12 h of seeding, cells were washed twice with PBS and cultured in growth or differentiation medium for C2C12 cells for the next 24 h. Conditioned mediums were centrifuged at 1200 g for 10 min and filtered with a 0.2 μm syringe filter and used immediately or stored at −80°C. For the neutralization of IGFBP-3 in CM, anti-IGFBP-3 antibody was pre-incubated at five and 10 μg/mL in CM for 30 min. For IGFBP-3 treatment, myoblasts or myotubes were treated with indicated amounts of IGFBP-3 in growth medium or differentiation medium.

Cell proliferation assays
C2C12 myoblasts and Capan-1 cells were seeded at a density of 20,000 cells/well with growth medium. Cells were counted at different time points after treatment of indicated reagents. Total ATP level in cells per well was determined by using an ATP assay kit (Roche).

Immunostaining and myotube analysis
For immunofluorescence analysis, cells were seeded onto sterile preprocessed glass coverslips that were pre-coated with 1 % gelatin. After the differentiation to myotubes, cells were washed two times with PBS followed by fixation in 4 % paraformaldehyde for 15 min. After being rehydrated in PBS, cells were blocked for 30 min in 1 % Bovine serum albumin (BSA) in PBST, PBS containing 0.2 % Triton-X. Afterward, cells were incubated with anti-MHC (1:20) in 1 % BSA/PBST overnight in cold room. Cells were next incubated with fluorescence labeled secondary anti-mouse antibody (1:200) and DAPI (1:1000) at room temperature for 1 h. The specimens were examined in a Leica TCS-NT laser scanning confocal microscope. The myotube differentiation index was calculated as the percentage of nuclei in MHC-positive myotubes. The nuclei number in each MHC-positive myotube was also calculated to evaluate differentiation. To assess myotube atrophy, diameters of myotubes were measured for each condition separated by 50 μm along the length of the myotube. Each data point was generated from at least 200 randomly chosen MHCpositive myotubes.

Quantitative RT-PCR
Total RNAs extracted from cells were reverse transcribed using oligo dT primers and quantitative PCR assays of cDNA pools were carried out using the CFX96 Real-time PCR system (Bio-Rad) to evaluate the abundance of target transcripts in relative to house-keeping gene GAPDH. Target cDNAs were amplified using the following probe sets.

Lentivirus-mediated RNAi
shRNAs in the pLKO.1-puromyosin vector were used for knocking down PTEN or IGFBP3 expression. Lentivirus packaging and testing were performed as previously described [56]. C2C12 or Capan-1 cells were infected with lentivirus in medium containing 8 μg/ml polybrene and selected in 3 μg/ml puromycin for 4 days. Knockdown efficiency was further confirmed with western blot or qPCR.

Measurement of protein synthesis and degradation in C2C12 myotubes
Total protein synthesis was assessed quantifying the amount of [ 3 H] tyrosine (PerkinElmer) incorporation into C2C12 myotube cultures. After treatment of indicated reagents for 24 h, the C2C12 myotubes were incubated with differentiation medium containing 5 μCi/ml [ 3 H] tyrosine for 2 h. The medium then was discarded, cells were washed twice with PBS, and 1 mL of 10 % trichloroacetate was added to each well. Total cell lysates were centrifuged and the pellets were washed with 95 % ethanol and dissolved in 0.1 M NaOH. Samples were analyzed for total radioactivity and protein concentration using a scintillation counter (PerkinElmer) and the Bradford assay, respectively. The final radioactivity was normalized to protein level. As for protein degradation, C2C12 myotubes were incubated with 5 μCi/ml [ 3 H] tyrosine for 48 h to label cellular proteins. Afterward, myotubes were incubated with medium containing 2 mM unlabeled tyrosine and indicated reagents control for 24 h. The medium was collected, precipitated with 10 % trichloroacetate and centrifuged. The acid-soluble radioactivity, which reflects degraded protein level, was measured using a scintillation counter.

Statistical analysis
Statistical analysis was performed using Student's t test in Microsoft Excel. The values were presented as means ± SEM and significance was defined as * P < 0.05.