Introduction

The homoeostasis of blood glucose levels is a precisely coordinated process1 and involves a tightly regulated balance between food intake, glucose production by the liver and glucose utilization by insulin-dependent tissues (mainly skeletal muscle and adipose tissue) and insulin-independent tissues (such as brain and kidney).2 Glucose homoeostasis is regulated by secreted proteins and hormones that act on a range of tissues to influence glucose disposal or production and long-term energy balance. For example, leptin (from adipose tissue) and insulin (from the pancreas) are both long-term adiposity regulators.3 Glucagon is also secreted from the pancreas, and is important in regulating glucose metabolism, particularly in the liver.4 Dysregulation of a range of secreted proteins has been associated with obesity, insulin resistance and type 2 diabetes.

Skeletal muscle comprises almost 40% of total body weight and is responsible for approximately 80% of insulin-dependent glucose disposal during glucose and insulin infusion.5 It has been suggested that skeletal muscle, like adipose tissue, produces secretory factors termed myokines that act in an autocrine, paracrine or endocrine fashion.6, 7, 8 Examples of muscle-derived secreted proteins include myostatin, transforming growth factor-β and IL-4; however, these proteins are involved in skeletal muscle growth and differentiation.9, 10, 11 Only a limited number of muscle-derived secreted proteins or hormones involved in regulating energy metabolism have been identified. Recent research has demonstrated that cytokines, such as IL-15 and IL-6, which are secreted by muscle may have a role in glucose metabolism in skeletal muscle in vitro and in vivo.12, 13, 14, 15, 16, 17 In addition, Nishizawa et al (2004) identified a novel skeletal muscle-derived secretory factor, musclin, whose expression was tightly regulated by nutritional changes as well as being regulated by insulin.6

The aim of this study was to identify secreted proteins that were differentially expressed in skeletal muscle during the development of obesity and type 2 diabetes. Red gastrocnemius muscle was selected because it has been extensively studied in relation to insulin resistance and type 2 diabetes (T2D), providing a wealth of background information to facilitate interpretation of results. As a model system we have used Psammomys obesus, a unique polygenic model of obesity, insulin resistance and type 2 diabetes,18, 19, 20 which display a range of phenotypic characteristics such as obesity and insulin resistance when given free access to a standard laboratory diet.20 The P. obesus genome is incompletely characterized, which precluded utilizing a bioinformatics approach to identify muscle-derived secreted proteins associated with obesity and type 2 diabetes in this model system. Therefore, a functional bioassay, the Signal Sequence Trap (SST), was developed to identify proteins secreted from red gastrocnemius muscle that were differentially expressed between lean, normal glucose tolerant (NGT), overweight and insulin resistant (IGT) and obese T2D P. obesus.

Materials and methods

Experimental animals

A colony of P. obesus was maintained at Deakin University, Geelong, Australia. Animals were fed ad libitum a standard rodent diet from which 63% of energy was derived from carbohydrate, 25% from protein and 12% from fat (Barastoc, Pakenham, Australia). Animals were maintained in a temperature-controlled room (22±1 °C) with a 12–12 h light–dark cycle (0600–1800 light). At 16 weeks of age, animals were classified into three groups according to their blood glucose and plasma insulin concentrations as previously described.18 The three groups were classified as follows: lean, NGT, overweight and IGT and obese, T2D. Whole blood glucose was measured using an enzymatic glucose analyzer (Model 27, Yellow Springs Instruments, Yellow Springs, OH, USA) and plasma insulin concentrations were determined using a double antibody solid phase radioimmunoassay (Linco, Billerica, MA, USA).

At 18 weeks of age, male P. obesus in each of the three groups were separated into two treatment groups, either ‘fed’ in which animals had ad libitum access to chow or ‘fasted’ in which animals were fasted for 24 h. The phenotypic characteristics of the animals are previously described.21 Animals were killed by anaesthetic overdose (pentobarbitone, 120 mg/kg, Sigma, St Louis, MO, USA) and the red gastrocnemius tissue for SST library construction was rapidly excised and placed in TRIzol (Invitrogen, Carlsbad, CA, USA) for immediate RNA extraction and further purification using RNeasy Midi Columns (Qiagen, Hilden, Germany).

All experiments were conducted according to National Health and Medical Research Council guidelines, and were approved by the Deakin University Animal Welfare Committee.

Cell culture

X63 cells22 and Factor-Dependent Cell Progenitor 1 cells (FDCP1)23 were a kind gift from Dr David Huang, Walter & Elisa Hall Institute of Medical Research, Melbourne, Australia. X63 cells were engineered to secrete high levels of IL-3, and the supernatant from these cells was used as a source of IL-3 for growth of IL-3-dependent FDCP1 cells. X63 cells were maintained in high glucose (25 mM) DMEM (Gibco, Invitrogen) supplemented with 10% FBS (Gibco), 50 μM β-mercaptoethanol (Sigma) and 100 μM L-asparagine (Sigma) at 37 °C in 5% CO2. The supernatant was collected and filtered through a 0.2 μM membrane (Pall Life Sciences, Ann Arbor, MI, USA), aliquoted and stored −20 °C. FDCP1 cells were maintained in high glucose (25 mM) DMEM supplemented with 10% FBS, 50 μM β-mercaptoethanol, 100 μM L-asparagine and 4% X63 cell supernatant, at 37 °C in 5% CO2. Plat-E24 cells were maintained in high glucose (25 mM) DMEM supplemented with 10% FBS, 10 μg ml−1 blasticidin (Sigma), 1 μg ml−1 puromycin (Invitrogen), 50 U ml−1 penicillin (Invitrogen) and 50 μg ml−1 streptomycin (Invitrogen), at 37 °C in 10% CO2.

Construction of murine IL-3 reporter gene construct for SST

RNA was extracted from X63 cells using RNeasy Mini Columns (Qiagen) and reverse transcribed. The cDNA was used as a template to PCR amplify full-length murine IL-3 (mIL-3) with primers (forward 5′ CCTTGGAGGACCAGAACGAGACAAT 3′ and reverse 5′ GCACTGCCTGCTGTTTTAACATTC 3′). The full-length mIL-3 construct was sequence verified and determined to be in frame.

The PCR product was subsequently used as a template to amplify the pro mIL-3 sequence lacking its signal peptide (amino acids 26–166, NP_034686). Primers were designed to contain NotI and XhoI restriction sites (underlined, forward 5′ TGACACGCCTGCGGCCGCAAGCTTCAATCAGTGGCCGGGATA 3′, and reverse 5′ ACGCCTCTCGAGTTAACATTCCACGGTTCCACGGTTA 3′). PCR products were purified and sequenced.

Construction of retrovirus expression vector containing mIL-3 reporter construct for SST

The purified pro mIL-3 DNA was digested with NotI and XhoI (New England Biolabs, Ipswich, MA, USA), treated with calf intestinal phosphatase (New England Biolabs) and ligated into pLNCX2 (Clonetech, Mountain View, CA, USA) vector DNA previously digested with NotI and SalI. The ligated products were transformed into ElectroMAX DH10β cells (Invitrogen) by electroporation. Recombinant clones were identified by a colony PCR using pLNCX2 vector-specific primers (forward 5′ AGCTCGTTTAGTGAACCGTCAGATC 3′ and reverse 5′ AGCTAGCTTGCCAAACCTACAG 3′). Large-scale preparation of plasmid DNA was performed on sequence verified positive clones using an Endofree Plasmid Maxi kit (Qiagen). The resulting vector was termed pLNCX2PRO.

Construction of cDNA library, cloning and amplification for SST

The SST methodology is illustrated in Figure 1. Total RNA was extracted from the red gastrocnemius muscle of fed and fasted NGT, IGT and T2D P. obesus. RNA from each group was pooled and polyA mRNA twice purified using the FastTrack 2.0 kit (Invitrogen). cDNA was synthesized from polyA mRNA using a modified Superscript Plasmid System with Gateway Technology for cDNA Synthesis and Cloning kit (Invitrogen), with random nonamers designed to contain a 5′NotI restriction site (underlined) (5′pTCTAGATCGCGAGCGGCCGCCCN9 3′). After second strand synthesis, SalI adaptor addition and NotI digestion, the 300–800 bp products were excised from an agarose gel and purified. The cDNA was subsequently inserted into NotI, XhoI sites of pLNCX2PRO and the ligated DNA electroporated into Electromax DH10β cells (Gibco). Plasmid DNA was purified using an Endofree Plasmid Maxi kit (Qiagen).

Figure 1
figure 1

Signal Sequence Trap (SST) methodology. Schematic representation of the SST methodology. SST, Signal Sequence Trap; FDCP1, factor-dependent cell progenitor 1; mIL-3, murine Interleukin-3; RT-PCR, real time PCR.

Full size image

Transfection of Plat-E cells

Plat-E cells were transfected with 14 μg plasmid DNA using Lipofectamine Plus (Invitrogen). Cells were incubated for 48 h and the supernatant was subsequently collected and processed for retrovirus infection of FDCP1 cells.

SPINfection of FDCP1 cells

Retrovirus supernatant from Plat-E cells transfected with skeletal muscle library plasmid was filtered using a 0.45 μM. polysulphonic filter (Pall Lifesciences) and 8 μg ml−1 polybrene (Sigma) and 4% X63 cell supernatant were added. FDCP1 cells (6 × 106) were resuspended in this solution and plated out. The plate was centrifuged at 1000 g for 1 h then the cells were incubated at 37 °C for 4 h. FDCP1 cells were subsequently collected, washed once and resuspended in FDCP1 maintenance media containing 4% X63 cell supernatant. The cells were plated out and incubated overnight before being washed three times and resuspended in FDCP1 maintenance media lacking X63 supernatant. Cells were plated at 103 cells per well in a 96-well round bottom plate and grown for 3 days.

Positive-clone selection and gDNA extraction

Cell media was refreshed with FDCP1 maintenance media (lacking IL-3) every 3 days and cells grown for 5 weeks. Genomic DNA was extracted from clones in wells containing a minimum of 10 000 cells.

PCR amplification of insert and sequence verification

A nested PCR protocol was utilized to amplify cDNA using the genomic DNA template and primary pLNCX2 primers (forward 5′ CTGGTTTAGTGAACCGTCAGATC 3′ and reverse 5′ CTCCTTGACAATAGAGCTGCAA 3′) and secondary primers (forward 5′ TAGCGCTACCGGACTCAGAT 3′ and reverse 5′ CGGCCACTGATTGAAGCTT 3′). A subset of 163 clones were randomly selected for sequencing.

cDNA microarray analysis of gene expression

PCR products were purified using an ArrayIt PCR Purification kit (TeleChem International, Sunnyvale, CA, USA). Samples were resuspended in 1 × Spotting Solution Plus (ArrayIt, TeleChem International) and printed onto SuperAmine microarray glass slides (TeleChem International) with ArrayIt Stealth SMP3 quill-tipped microarray pins (TeleChem International) using a CWP robotic arrayer (Biorad Laboratories, Hercules, CA, USA) at 50–60% humidity. Slides were subsequently stored at 35% humidity to dry for 24 h before being washed and blocked as per the manufacturer's instructions.

A pool of RNA from red gastrocnemius muscle of P. obesus groups (NGT, IGT and T2D in fed state) was reverse transcribed and labelled with the fluorescent dye Cy3 (Amersham Pharmacia Biotech, Little Chalfont, UK) and used as the reference. RNA from individual animals was labelled with Cy5. Hybridization was performed with combined target and reference fluorescently labelled cDNAs on slides in hybridization chambers at 42 °C for 16 h. Slides were scanned using a GenePix 4000B scanner (Axon Instruments, Sunnyvale, CA, USA) and microarray data analyzed using GenePix Pro 5.1 and Acuity 4.0 software (Axon Instruments). Linear normalization was performed on each array using only features that passed the quality control filtering criteria. The filtering criteria defined an acceptable expression value as one that was two s.d. above the background noise, was not derived from a saturated signal and its total fluorescent intensity was greater than 200 U. Standard t-tests were used to identify differential expression between NGT, IGT and T2D P. obesus red gastrocnemius muscle on the microarrays. Gene expression was considered significantly different at a nominal P<0.05 level.

Sequencing of differentially expressed clones and preliminary bioinformatics

Clones identified as being differentially expressed were reamplified by PCR using the primary PCR product, sequenced and analyzed using BLAST (http://www.ncbi.nlm.nih.gov/blast), SignalP (http://www.cbs.dtu.dk/services/SignalP), SecretomeP (www.cbs.dtu.dk/services/SecretomeP-1.0/), TMHMM (http://www.cbs.dtu.dk/services/TMHMM) and Psort II (http://psort.nibb.ac.jp/).

Confirmation of differential expression by real-time PCR

For each gene of interest, forward and reverse real-time PCR (RT-PCR) primers were designed using Primer Express (Version 1.5, PE Applied Biosystems, Foster City, CA, USA). The sequences were as follows: C1r forward 5′ GCCCAACTCCGTGGAAGAG 3′ and reverse 5′ AGGAGAGCCCGTCTTTCTGAA 3′, Susd2 forward 5′ TCATAGCCCCAAAGCTCAATG 3′ and reverse 5′ TCACCTGGAACATCTCCATGC 3′, Postn forward 5′ CCCGCAGATAGCACCTTGAT 3′ and reverse 5′ GTGCCCTCCAGCAGATCCT 3′, Ccbe1 forward 5′ CTCGCCCGAAGACTTCAGAC 3′ and reverse 5′ CGCGACAGAGAAGTGTGCTC 3. A separate group of P. obesus was classified as NGT, IGT or T2D (as described above), killed, tissues were excised and RNA extracted. The concentration of RNA was determined using the RNA 6000 Nano Assay on the Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA). First-strand cDNA was generated using SuperScript First-Strand Synthesis System for RT-PCR (Invitrogen). Differential gene expression was confirmed by RT-PCR using SYBR Green PCR Master Mix (Applied Biosystems) on an ABI PRISM 7700 Sequencing Detector (Perkin Elmer, Waltham, MA, USA).

Adipose tissue fractionation

Visceral (mesenteric) adipose tissue was fractionated as previously described.21 Analysis of gene expression in each fraction was performed by RT-PCR using gene-specific primers as above. RT-PCR primers designed to leptin (Lep) and delta-like 1 homologue (Dlk1) confirmed efficient fractionation. The primer sequences were as follows: Lep forward 5′ TCGCGGTCTACCAACACATC 3′ and reverse 5′ TCTCTAGGTCGTTGGATATTTGCA 3′, Dlk1 forward 5′ GCTGCCCAACTGGATTCATC 3′ and reverse 5′ TGCTGGCACAGGCATTCA 3′.

Statistical analysis

Statistical analysis was performed using SPSS version 14.0 (SPSS Inc., Chicago, IL, USA). Levene's test for homogeneity of variance was used to determine if variance among the six animal groups (NGT fed, NGT fasted, IGT fed, IGT fasted, T2D fed, T2D fasted) was equal. One way ANOVA was used to test for differences between group means with LSD (if the variance was equal) and Games–Howell post hoc analysis (if the variance was not equal). Associations between gene expression levels and phenotypic measures were determined using Pearson's correlation for normally distributed data, or Spearman's correlation for data not normally distributed. Differences and correlations were considered significant at P<0.05.

Results

Development of SST in conjunction with microarray

To identify secreted proteins, an SST was developed in which P. obesus skeletal muscle cDNA was ligated into a retroviral expression vector upstream of a signal sequence-less IL-3. Virus containing supernatant collected from cells transfected with the vector was then used to infect IL-3-dependent FDCP1 cells. Clones which restored the ability of the signal sequence-less IL-3 to be secreted from the cells were detected by the presence of healthy, replicating FDCP1 cells. The cDNA inserts of these clones were amplified by PCR, spotted onto microarray slides and hybridized with cDNA from skeletal muscle of NGT, IGT and T2D P. obesus.

Initial screening of the P. obesus red gastrocnemius library

A total of 1600 clones were generated using the SST. Initial sequencing of 163 randomly selected clones identified 71 transcripts with homology to known genes (Table 1). Of these, 41 were determined by SignalP to contain a signal peptide. An additional 10 proteins were predicted by SecretomeP to be secreted, possibly through non-classical secretory pathways. The 71 genes consisted of four cytokines (6%), 11 secreted proteins (15%), eight receptors (11%), 21 intracellular vesicle membrane proteins (30%), 22 cytosolic/miscellaneous proteins (31%) and five mitochondrial proteins (7%). It is interesting to note that the SST technology identifies genes encoding proteins that may be either secreted or membrane-bound, as both types of proteins can result in the presentation of IL-3 to the FDCP1 cells and facilitate survival of the cells. In this project, however, we focused on identifying secreted proteins. The range of proteins listed above provides evidence that the technology was successful in identifying a range of proteins that are secreted by skeletal muscle in P. obesus.

Table 1 Genes identified from initial sequencing of the positive clones produced from P. obesus skeletal muscle SST
Full size table

Identification of differentially expressed secreted proteins using cDNA microarrays

To identify secreted skeletal muscle proteins associated with obesity and diabetes, a cDNA microarray was constructed using the 1600-positive SST cDNA clones. Cy5-labelled P. obesus red gastrocnemius skeletal muscle cDNA from individual NGT, IGT and T2D animals (n=6–7 per group) were hybridized to the microarrays in combination with Cy3-labelled reference cDNA. Analysis of gene expression identified 91 clones that were differentially expressed between groups. These clones were sequenced and a total of 51 genes were identified (Table 2). Examples of differentially expressed genes previously associated with obesity and T2D included complement component 1, r component (C1r), calsequestrin (Casq1) and secreted protein, acidic, cysteine-rich protein (Sparc).25, 26, 27, 28

Table 2 Differentially expressed clones identified from P. obesus skeletal muscle using SST in conjunction with cDNA microarray technology
Full size table

Additional bioinformatics identified four candidate genes not previously associated with metabolic disease for further study. These genes were sushi domain containing 2 (Susd2, NM_019601), collagen and calcium-binding EGF domains 1 (Ccbe1, NM_133459), periostin (Postn, NM_006475) and decorin (Dcn, NM_133503). All selected genes contained a recognizable signal peptide.

Further gene expression analysis by RT-PCR

To confirm the differential expression observed in the cDNA microarrays and extend the analysis to include fasted animals, RT-PCR was performed on C1r, Susd2, Ccbe1 and Postn genes in P. obesus red gastrocnemius skeletal muscle.

C1r gene expression was significantly increased (by 195%) in obese T2D compared with lean NGT P. obesus in the fed state (P=0.020), and there was a significant correlation between C1r gene expression and body weight (r2=0.20, P=0.035) (Figure 2a). This is consistent with the microarray result. In the fasted state, C1r gene expression was significantly increased (by 314 and 199%, respectively) in obese T2D compared with lean NGT and IGT P. obesus (P=0.005, P=0.027, respectively). Furthermore, skeletal muscle expression of C1r in fasted animals was significantly correlated with body weight (r2=0.30, P=0.010) and plasma insulin concentration (r2=0.48, P=0.001).

Figure 2
figure 2

Analysis of gene expression in P. obesus red gastrocnemius muscle by RT-PCR. Red gastrocnemius skeletal muscle RNA was extracted from fed and fasted NGT, IGT, T2D P. obesus and gene expression determined by RT-PCR. Data are mean±s.e.m. (n=6–8). (a) C1r expression was increased in T2D compared with NGT (*P=0.020) in the fed state, and increased in T2D compared with NGT and IGT (†P=0.005 and P=0.027, respectively) in the fasted state. (b) Susd2 mRNA expression was increased in fasted T2D compared with NGT and IGT (*P=0.004 and P=0.001, respectively) and decreased in fed T2D compared with NGT (†P=0.014). (c) Ccbe1 mRNA expression was increased in fasted T2D compared with NGT and IGT (*P=0.013 and P=0.031, respectively). (d) Postn mRNA expression was increased in fed T2D compared with NGT (*P=0.011).

Full size image

As seen on the microarrays, Susd2 gene expression was significantly lower in T2D compared with IGT fed P. obesus (P=0.014) (Figure 2b). However, in the fasted state, Susd2 gene expression was significantly higher in T2D compared with NGT and IGT animals (P=0.004 and P=0.001, respectively) and was significantly correlated with body weight (r2=0.18, P=0.018) and plasma insulin concentration (r2=0.33, P=0.004).

There were no significant differences in Ccbe1 gene expression between fed P. obesus groups by RT-PCR. The magnitude of the differences between the groups according to the microarray data were smaller than for C1r or Susd2. In the fasted state Ccbe1 gene expression was significantly higher in T2D compared with NGT and IGT animals (P=0.013 and P=0.031, respectively) and was significantly correlated with body weight (r2=0.24, P=0.020), and plasma insulin concentration (r2=0.32, P=0.006) (Figure 2c).

Postn gene expression was significantly higher in fed T2D animals compared with NGT animals (P=0.011), consistent with the microarray result (Figure 2d). In the fasted state, Postn gene expression was significantly correlated with plasma insulin concentration (r2=0.24, P=0.018).

Dcn gene expression was associated with obesity, and has been described in detail elsewhere.21

Periostin: a secreted protein not previously associated with obesity and type 2 diabetes

This study has shown for the first time that Postn was associated with obesity and type 2 diabetes in P. obesus, as indicated by the elevated gene expression in red gastrocnemius skeletal muscle from T2D compared with NGT animals. Therefore, Postn gene expression was characterized further in a variety of P. obesus tissues by RT-PCR (Figure 3). Postn gene expression was detected in most P. obesus tissues examined, with high expression observed in adipose tissue, lung and ovary. Although the original aim of this study was to identify secreted proteins from skeletal muscle, the striking expression of Postn within adipose tissue led us to characterize Postn expression further within this tissue.

Figure 3
figure 3

Postn was highly expressed in adipose tissue of obese, T2D P. obesus. RNA was extracted from a number of P. obesus tissues and Postn gene expression analyzed by RT-PCR. Gene expression is represented relative to hypothalamus expression levels. Postn gene expression was higher in adipose tissue, lung and ovaries compared with all other tissues. IM fat, intramuscular fat; EDL, extensor digitorum longus; white gastroc, white gastrocnemius; red gastroc, red gastrocnemius.

Full size image

Postn gene expression was analyzed in visceral (mesenteric) and subcutaneous (subscapular) adipose tissue from NGT, IGT and T2D fed P. obesus. Overall, Postn mRNA expression was higher in visceral compared with subcutaneous adipose tissue (P=0.007). This was also evident in the separate groups, reaching significance in the NGT animals (P=0.004) (Figure 4). Postn gene expression was significantly higher in T2D P. obesus than NGT animals in both visceral and subcutaneous adipose tissue (P=0.026 and P<0.001, respectively). In visceral adipose tissue, Postn gene expression was significantly correlated with blood glucose concentration (r2=0.42, P=0.006). In subcutaneous adipose tissue, Postn gene expression was significantly correlated with body weight (r2=0.45, P=0.005), blood glucose levels (r2=0.30, P=0.028), plasma insulin concentration (r2=0.28, P=0.037) and percent body fat (r2=0.35, P=0.021). These results demonstrate that Postn gene expression was elevated in both subcutaneous and visceral adipose tissue in obese T2D P. obesus, with elevated expression in visceral adipose tissue compared with subcutaneous adipose tissue.

Figure 4
figure 4

Analysis of Postn gene expression in P. obesus visceral (mesenteric) and subcutaneous (subscapular) adipose tissue by RT-PCR. RNA was extracted from visceral (mesenteric) and subcutaneous (subscapular) adipose tissue of NGT, IGT and T2D P. obesus in the fed state and gene expression measured by RT-PCR. Data are mean±s.e.m. (n=6 per group). Postn gene expression was increased in visceral NGT compared with subcutaneous NGT (*P=0.004), in visceral T2D compared with visceral NGT (†P=0.026), and in subcutaneous T2D compared with subcutaneous NGT (#P<0.001).

Full size image

To determine which cells within adipose tissue were responsible for Postn expression, visceral (mesenteric) adipose depots from P. obesus were fractionated into adipocytes and stromal/vascular cells. RT-PCR of leptin (Lep) and Dlk1 was performed on cDNA from the fractions and confirmed efficient fractionation. Lep gene expression was substantially higher in adipocytes compared with stromal/vascular cells (P<0.001), and Dlk1 gene expression was higher in stromal/vascular cells compared with adipocytes (P=0.022) (data not shown). Postn gene expression was analyzed by RT-PCR in these fractions and was found to be significantly higher in the adipocyte fraction compared with the stromal/vascular fraction (P=0.004) (Figure 5).

Figure 5
figure 5

Postn gene expression in fractionated P. obesus visceral adipose tissue. Visceral (mesenteric) adipose tissue from P. obesus was fractionated into adipocytes and stromal/vascular cells and gene expression measured by RT-PCR. Data are mean±s.e.m. (n=11 per group). Postn gene expression was significantly higher in the adipocyte fraction compared with the stromal/vascular fraction (*P=0.004).

Full size image

Discussion

During the development of insulin resistance and type 2 diabetes glucose, homoeostasis becomes dysregulated, and adipose tissue and skeletal muscle produce secretory factors that act in an autocrine or paracrine fashion to help regulate blood glucose levels. 6 In this study, we sought to identify differentially expressed skeletal muscle genes encoding secreted proteins in P. obesus. Although P. obesus is an excellent animal model to investigate obesity and type 2 diabetes,18, 19, 20 a conventional bioinformatics approach to identify secreted proteins in P. obesus was not possible as it relies on detailed genomic sequence which is currently unavailable for this animal model. We therefore developed a novel methodology that combined an SST with cDNA microarray technology to identify genes encoding secreted proteins that were differentially expressed in the skeletal muscle of normal glucose tolerant, overweight and impaired glucose tolerant and obese type 2 diabetic groups of P. obesus.

The SST library was constructed from red gastrocnemius skeletal muscle. This metabolically important tissue has been extensively studied in relation to processes and pathways contributing to the development of insulin resistance and type 2 diabetes. It must be noted that tissue type, and more significant to our study, muscle and fibre type, would likely substantively alter the SST profile. In addition, we wished to analyze any potential roles of genes in a variety of metabolic processes, for example, nutritional regulation. Therefore, skeletal muscle from fed and fasted P. obesus was utilized for SST library construction and gene expression analysis as there are vast differences in metabolic homoeostasis that occur during fed (absorptive) and fasted (postabsorptive) states.

Sequencing of random clones produced from the skeletal muscle SST library resulted in the identification of 71 genes with homology to known genes. Bioinformatics analysis of the closest orthologues of these genes identified 41 genes with known signal peptides and a further 10 were predicted by the SignalP algorithm to be secreted or membrane-bound. These data show that the majority of SST clones were likely to be secreted or membrane-bound, and validated the decision to proceed with screening these genes for differential expression in skeletal muscle by microarray analysis. It should be acknowledged that ‘validation’ by bioinformatics analysis is not as conclusive as demonstration of secretion by experimentation; however, practical considerations prevented doing this for all 71 genes in this study. Of the 51 SST genes found to be differentially expressed between T2D and NGT P. obesus, 24 of these genes were not previously known to encode secreted proteins. A traditional bioinformatics approach would not have identified approximately 50% of these differentially expressed genes.

A number of false positives were generated by the SST. Previous SST studies have found a 25 to 50% false-positive rate,29, 30, 31, 32 which results from incomplete enrichment of cDNA libraries for secreted proteins, as well as inclusion of transcription factors and oncogenes that may permit mIL-3 secretion or enable the cells to grow in the absence of mIL-3. Some genes identified, such as those encoding mitochondrial proteins, contain an N-terminal hydrophobic domain that may act as a signal peptide. We are unable to explain the identification of structural proteins, for example profilin and alpha actin; however, these may be due to cDNAs being cloned out of frame or in antisense orientation revealing cryptic open-reading frames encoding hydrophobic amino-acid sequences that may act as signal peptides. Generation of false positives is a limitation of SST that needs to be considered when designing studies incorporating this technology.

Microarray analysis of the 1600 SST clones allowed comparison of gene expression between different P. obesus groups. The gene expression changes were not large, especially between the IGT animals and other groups, however this is consistent with previous reports which found gene expression changes in skeletal muscle of subjects with diabetes to be small in magnitude.33, 34, 35 Moreover, the identification of genes previously associated with diabetes and obesity, such as Sparc28 and C1r,25, 26 provided evidence to validate the use of a combined SST and cDNA microarray technology to identify secreted proteins differentially expressed in obesity and type 2 diabetes. This study is based upon gene expression, therefore further experiments are required to demonstrate that the proteins identified are bona fide secreted proteins and investigate their role in obesity and type 2 diabetes.

A number of genes identified using the combined microarray analysis of SST clones have previously been shown to be highly expressed in adipose tissue. These observations raised the possibility that the detected expression of these genes was a result of contamination of skeletal muscle by interstitial adipocytes within the skeletal muscle in tissue samples. To test this possibility, we measured gene expression of adipose-specific fatty acid-binding protein 4 (Fabp4), which is predominantly and highly expressed in adipocytes and highly regulated during adipocyte differentiation36, 37, 38 in the same skeletal muscle samples used to confirm differential expression. The RT-PCR revealed no detectable Fabp4 expression in any of three animal groups (data not shown), suggesting that the genes isolated do originate from skeletal muscle tissue itself, and not from contaminating interstitial adipocytes.

The 51 differentially expressed genes consisted of three cytokines, seven secreted proteins, 11 receptors, cell surface and ion channel activity genes, 15 intracellular membrane vesicle proteins and 15 cytosolic and miscellaneous proteins. These genes are involved in a diverse range of cellular processes and therefore, this technology could be potentially applied to any disease state. In depth bioinformatics and gene ontology analysis of the differentially expressed genes resulted in the identification of four secreted proteins potentially associated with obesity and type 2 diabetes. These genes were Susd2, Ccbe1, Dcn and Postn. We explored the biology of Dcn in detail and found that in addition to it being expressed in muscle, it was highly expressed in adipose tissue by non-adipocytes adjacent to adipose tissue vasculature.21 Currently, the biological functions of Susd2 and Ccbe1 are relatively unknown, therefore these genes represent promising candidates for further functional characterization to determine their potential role in the development of obesity and type 2 diabetes.

Of the genes differentially expressed between the skeletal muscle of diabetic and lean, healthy P. obesus, Postn was the most interesting. Postn is a secreted protein that is involved in cell attachment, adhesion and differentiation.39, 40, 41, 42, 43, 44 It has recently been added to the matricellular class of proteins,45 which are a group of extracellular matrix related molecules that modulate cell–matrix interactions.46 Matricellular proteins are expressed at high levels during development, but are typically limited to wound repair, tissue remodelling or disease in postnatal tissue, where their levels increase substantially.45 Postn has previously been shown to be expressed in skeletal muscle in states of repair,47 and is essential for healing of cardiac muscle after acute myocardial infarction.48 Sparc, another matricellular protein that has been previously found to have increased gene expression in diabetes-related vascular hypertrophy,28 was also found to be upregulated in diabetic P. obesus skeletal muscle by microarray in this study. It is therefore tempting to speculate that Postn may have a homoeostatic role in the response of muscle to type 2 diabetes; however, further studies are clearly required to show such a role.

This study reveals, for the first time, high expression of Postn in adipose tissue of P. obesus. Although the original aim was to identify and subsequently characterize skeletal muscle-derived secreted proteins during the development of obesity and type 2 diabetes, the striking gene expression of Postn within adipose tissue led us to characterize Postn further within this depot. Fractionation studies demonstrated that the expression was in the adipocytes rather than in the stromal or vascular cells. Expression of Postn was higher in obese, diabetic P. obesus than lean, healthy animals in both visceral and subcutaneous depots. This suggests that, as in skeletal muscle, Postn may be involved in repair or be required for expansion of the adipose tissue, having a role in extracellular matrix remodelling, cellular adhesion or adipocyte differentiation. Adipose tissue growth requires extensive vascularization to enable correct tissue functionality,49 and Postn is known to have a role in angiogenesis.50 Of particular interest was the increased Postn expression in visceral compared with subcutaneous adipose tissue in P. obesus as it is well established that visceral adipose tissue accumulation is associated with the development of insulin resistance and type 2 diabetes.51, 52, 53, 54 It should be noted that the correlation between Postn expression and body weight does not provide conclusive evidence that adipose tissue is the major source of circulating periostin, and this remains to be empirically determined.

In conclusion, we have shown that the novel methodology of SST in conjunction with cDNA microarray technology is a powerful tool to identify differentially expressed secreted proteins in red gastrocnemius muscle from NGT, IGT and T2D P. obesus. A number of candidate genes including Postn were identified and provide evidence of the value of this methodology not only for identification of secreted proteins in obesity and type 2 diabetes, but potentially also for a variety of other disease states

Conflict of interest

The authors declare no conflict of interest.