Men and non-pregnant women over 18 years of age were recruited to the study from two separate centers. Written informed consent was obtained from all subjects and ethical approval received from Lothian Research Ethics Committee (UK) and the Regional Ethics Committee in Stockholm (Sweden). Participating patients had a diagnosis of upper gastrointestinal cancer (esophageal, gastric, pancreatic) and were undergoing surgery with the intent of resection of the primary tumor. A small number of weight stable (WS) patients undergoing surgery for benign, non-inflammatory conditions (n = 7) were also included in the analysis. In center 1 (Edinburgh, UK) a fasting venous blood sample was taken and serum C-reactive protein measured as a marker of systemic inflammation (SI). Body mass index (BMI) and mid-arm muscle circumference were calculated. Clinical details and degree of weight loss from self-reported pre-illness stable weight were recorded. A weight loss ≥ 5% identified weight-losing (WL) cancer patients as opposed to WS individuals. A serum C-reactive protein ≥ 5 mg/l was used to define the presence of SI. For patients from center 2 (Stockholm, Sweden) weight and self-reported change in weight over time were recorded. Rate of weight loss was therefore used in these subjects. Due to the small number of controls (otherwise considered as non-cancer patients but with other co-morbidities) and the lack of detailed knowledge of their physical capacity, the primary analysis strategy was chosen to generate molecular changes that varied with the severity of weight loss in patients in center 1 and validate such changes in the independent cohort from center 2 using more than one muscle type. This strategy was devised to provide a stringent test of the molecular changes, as the conclusions are based on a relatively large number of patients with otherwise similar clinical characteristics.
All biopsies were taken at the start of open abdominal surgery. In center 1, the edge of the rectus abdominis was exposed and a 1-cm3 specimen removed using sharp dissection. The biopsy was snap frozen in liquid nitrogen and stored at -80°C until further analysis. In center 2, vastus lateralis muscle biopsies were taken with a Bergstrom needle and diaphragm biopsies were obtained by sharp dissection when possible. Both samples were snap frozen and stored at -80°C for further analysis. Approximately 20 mg of frozen tissue was homogenized in 0.5 ml of lysis buffer (Triton - X100 (1%), NaCl (150 mM), Tris-HCl (50 mM), EDTA (1 mM), PMSF (1 mM), protease inhibitors (Roche Diagnostics, Burgess Hill, UK); 1 tablet per 10 ml), water to 10 ml) using a Powergen 125 (Fisher Scientific, Loughbourgh, UK)) electric homogenizer. Samples were left on ice for 15 minutes prior to centrifuging at 13,000 rpm for 15 minutes. The supernatant was removed, and protein concentration was determined by comparing equal volumes of sample solution to known standards using the Lowry method. Samples were then stored at -80°C.
Approximately 20 mg of muscle was re-suspended in 180 μl of low salt lysis buffer (10 mM HEPES, 10 mM KCl, 1.5 mM MgCl2, 0.1 mM EDTA, 0.1 mM EGTA, 1 mM DTT, 0.5 mM PMSF, protease inhibitors (Roche Diagnostics; 1 tablet per 10 ml)) and ground using a handheld homogenizer. Samples were incubated on ice for 5 minutes before two cycles of freeze-thaw lysis. After a brief vortex, samples were centrifuged at 4,000 rpm for 3 minutes. The supernatant was removed and the pellet (containing the nuclei) re-suspended in 40 μl high salt extraction buffer (20 mM HEPES, 420 mM NaCl, 1 mM EDTA, 1 mM EGTA, 25% glycerol, 1 mM DTT, protease inhibitors (Roche Diagnostics; 1 tablet per 10 ml)). Samples were incubated on ice for 30 minutes with gentle mixing of the tubes every 5 to 10 minutes. Samples were centrifuged at 4,000 rpm for 5 minutes at 4°C. An aliquot of supernatant (containing the nuclear proteins) was stored at -80°C.
Protein from each sample (20 μg) was added to 3 μl of 4 × loading buffer solution (0.5 M Tris-HCl pH 6.8, 20% glycerol, 4% SDS, 0.05% β-mercaptoethanol, 0.004% bromophenol blue) and boiled for 3 minutes. Proteins were resolved using SDS-PAGE at 160 V for 45 minutes. Proteins were transferred to a nitrocellulose membrane (80 mA for 1 hour) using semi-dry transfer (Biorad, Hemel Hempstead, UK). Membranes were blocked with either 3% bovine serum albumen/tris-buffered saline (TBS) with Tween 20 (TBST; 0.05% Tween) overnight at 4°C or with 5% milk/TBST for 1 hour at room temperature. Incubation with primary antibody (1:1,000) was carried out in either 3% bovine serum albumen/TBST or 0.5% milk/TBST solution at room temperature for 2 hours or overnight at 4°C. Membranes were washed with TBST and primary antibody binding detected using horseradish-peroxidase conjugated secondary antibodies (1:2,000 to 1:5,000; anti-mouse, anti-rabbit; Upstate, Dundee, UK). Specific signal was detected using ECL reagent (GE Healthcare, Little Chalfont, UK) and exposure on photographic film (Kodak). Films were scanned and densitometry values estimated using ImageJ (NIH) software. The primary antibodies used in the study were against phos-CaMKII(Thr286), FOXO1 and FOXO3a (New England Biolabs, Hitchin, UK), Lamin A/C (Insight, Wembely, UK), alpha-skeletal actin (Novocaestra, Newcastle, UK) and calcium/calmodulin-dependent protein kinase (CaMK)II (BD Biosciences, Oxford, UK).
Total RNA was extracted from approximately 20 mg of muscle using TRIzol (Invitrogen, Paisley, UK) reagent according to the manufacturer's directions. The RNA pellet was re-suspended in diethylpyrocarbonate-treated water and RNA concentration was determined using a Nanodrop spectrophotometer (LabTech International, Ringmer, UK). RNA quality was assessed using 260/280, 230/260 ratios and the RNA integrity number (RIN) score from the BioAnalyzer 2100 instrument (Agilent Technologies, Stockport, UK). Total RNA (3.5 μg) was reverse transcribed and processed according to the protocol provided by Affymetrix Inc. for the GeneChip Expression 3' Amplification One-Cycle Target Labeling and Control Reagents kit (Affymetrix, High Wycombe, UK). Reverse transcription and second strand cDNA synthesis were followed by in vitro transcription and biotinylation. Biotinylated cRNA products were cleaned up using columns (Affymetrix). The quality of the biotinylated cRNA was assessed by Nanodrop (LabTech International, UK) and BioAnalyzer (Agilent Technologies) instruments and the cRNA was then fragmented according to Affymetrix protocols. Samples were hybridized to the HGU-133plus2 GeneChip array (covering approximately 54,000 sequences). The raw data files can be accessed at the Gene Expression Omnibus using the ID [GEO:GSE18832].
For quantitative real time PCR (qRT-PCR), cDNA was prepared using 1 μg RNA, TaqMan reverse transcription reagents (Applied Biosystems, Warrington, UK) and random hexamer primers (Applied Biosystems). Primers were designed to span introns using Primer Express 3.0 software (Applied Biosystems) and constructed by Invitrogen (Paisley, UK); primer sequences are detailed in Table S1 in Additional data file 1. Samples were run on an ABI 7900HT Fast Real-Time PCR system (Applied Biosystems) in triplicates of 20 μl per well using SYBR Green PCR Master Mix (Applied Biosystems) as per the manufacturer's instructions. Expression levels were normalized to ribosomal 18S RNA and results examined using the ΔCt method . SPSS (SPSS Inc, Chicago, IL, USA) or GraphPad (GraphPad Software, La Jolla, CA, USA) statistical software was utilized. Student's two tailed t-test or one way ANOVA (analysis of variance) was used to compare means between groups. Log transformation was used when appropriate. Mann-Whitney was used for nonparametric analysis. Contingency tables were constructed where relevant and analyzed by Fisher's exact test. Statistical significance was set at P < 0.05.
Microarray data were analyzed using the Microarray Suite software (MAS) version 5.0 (Affymetrix). To improve the accuracy of the gene to probe relationship, a custom chip definition file (CDF)  was used defining the Affymetrix probes by Ensembl transcript ID. Data were normalized using MAS5 and robust multi-array average . Genes called absent on every array by the MAS5 software were filtered from the data and remaining genes analyzed using the quantitative function in significance analysis of microarrays (SAM)  implemented in the Bioconductor suite . Percentage weight loss or SI were the quantitative variables. To test the robustness of the approach, the limma package  in the Bioconductor suite was used to identify genes co-varying with weight loss or SI. Both SAM and limma generate a false discovery rate (FDR) . All genes identified by both procedures with an FDR <10% that covaried with weight loss were further examined. We also carried out a comparative microarray analysis [24, 25] to examine the link between muscle cachexia and other muscle physiological states. The top 20 most regulated genes by eccentric muscle damage , muscle obtained from intensive care unit patients  and in response to exercise training  were obtained from three published articles. The mean values for these highly regulated marker genes for these physiological states were then plotted using the patient values from the present study, where patients had either less than or more than 5% weight loss. Functional annotation of these genes was carried out using Gene Ontology (GO)  utilizing the topGO tool  in the Bioconductor suite along with web-based Ingenuity Pathway Analysis . For analysis of microarray data the Bioconductor suite  and the R language for statistics (R Development Core Team; version 2.7.1) were used.
The gene-sets (see below) identified by microarray analysis were used in further investigation of the regulatory mechanisms using promoter analysis. For all genes the region up to 1,500 bp upstream of the annotated gene start was used as the proximal promoter region. Both strands were then scanned with the JASPAR  matrices representing various mammalian transcription factor binding sites (89 in total). A matrix specific threshold corresponding to 0.8 of the scoring range of the matrix was used on the log-ratio matrix. All log-ratio transformations were done using a zero order uniform background model and a pseudo-count of one to avoid zero-entries in the original JASPAR matrix. The number of hits per base-pair and the number of sequences with one or more hits were registered and used for over-representation statistical analysis. We used a background set of promoter sequences extracted in a similar manner from the 'all genes expressed' present/absent call in skeletal muscle from this array technology [24, 27]. A sequence-specific over-representation was calculated using Fisher's exact test and a base-pair-specific over/under-representation was calculated using a Z-score. Finally, using the base-pair-specific over- and under-representation values, a heatmap was generated for visualization purposes. For all analyses the ASAP  framework was used in conjunction with R.