Using transcriptomics to identify and validate novel biomarkers of human skeletal muscle cancer cachexia

Abstract

Background: Cancer cachexia is a multi-organ tissue wasting syndrome that contributes to morbidity and mortality in many cancer patients. Skeletal muscle loss represents an established key feature yet there is no molecular understanding of the disease process. In fact, the postulated molecular regulators of cancer cachexia originate largely from pre-clinical models and it is unclear how these translate to the clinical environment.

Methods: Rectus abdominis muscle biopsies were obtained from 65 upper gastrointestinal (UGI) cancer patients during open surgery and RNA profiling was performed on a subset of this cohort (n = 21) using the Affymetrix U133+2 platform. Quantitative analysis revealed a gene signature, which underwent technical validation and independent confirmation in a separate clinical cohort.

Results: Quantitative significance analysis of microarrays produced an 83-gene signature that was able to identify patients with greater than 5% weight loss, while this molecular profile was unrelated to markers of systemic inflammation. Selected genes correlating with weight loss were validated using quantitative real-time PCR and independently studied as general cachexia biomarkers in diaphragm and vastus lateralis from a second cohort (n = 13; UGI cancer patients). CaMKIIbeta correlated positively with weight loss in all muscle groups and CaMKII protein levels were elevated in rectus abdominis. TIE1 was also positively associated with weight loss in both rectus abdominis and vastus lateralis muscle groups while other biomarkers demonstrated tissue-specific expression patterns. Candidates selected from the pre-clinical literature, including FOXO protein and ubiquitin E3 ligases, were not related to weight loss in this human clinical study. Furthermore, promoter analysis identified that the 83 weight loss-associated genes had fewer FOXO binding sites than expected by chance.

Conclusion: We were able to discover and validate new molecular biomarkers of human cancer cachexia. The exercise activated genes CaMKIIbeta and TIE1 related positively to weight-loss across muscle groups, indicating that this cachexia signature is not simply due to patient inactivity. Indeed, excessive CaMKIIbeta activation is a potential mechanism for reduced muscle protein synthesis. Our genomics analysis also supports the view that the available preclinical models do not accurately reflect the molecular characteristics of human muscle from cancer cachexia patients.

Figures

Figure 1

Cluster analysis identifies high and low weight loss groups. Using SAM and limma, 83 genes were identified as correlating with weight loss. Expression data from these genes were used to drive cluster analysis. This revealed two clusters of subjects; high weight loss (≥ 5%) and low weight loss (<5%).

Figure 2

qRT-PCR validates array-identified genes covarying with weight loss. For each of the genes validated by qRT-PCR Pearson correlation coefficients were generated for expression and percentage weight loss for both the Affymetrix data and the qRT-PCR data. All genes except SGK1 validated the array data. P-values for the correlations ranged from 0.03 to below 0.01. Yellow indicates positive correlation; blue indicates negative correlation.

Figure 3

CAMkIIβ and TIE1 correlate with weight loss in cancer cachexia. In order to validate the findings from the rectus abdominis, qRT-PCR was used to examine mRNA expression of (a) CAMkIIβ and (b) TIE1 in diaphragm (open circles) and vastus lateralis (closed circles) in a separate clinical cohort. Correlation plots for mRNA level against rate of weight loss are shown. Correlation coefficients were significant with P < 0.05. CAMkII protein and phospho-protein levels are increased in subjects with weight loss. (c) Protein levels of CAMkII and (d) phosphoCAMkII were assessed in the rectus abdominis muscle from center 1 subjects by western blot. Intensity levels were normalized against alpha-skeletal actin and the mean ratio of CAMkII/actin or phosphoCAMkII (pCAMkII)/actin are shown for subjects with less than (black) or more than (white) 5% weight loss. *P-value < 0.05, one-sided Mann Whitney test; n = 59. Error bars represent SEM.

Figure 4

Gene expression signatures demonstrate lack of relationship between weight loss and muscle damage, muscle sepsis and exercise training status. The top 20 most regulated genes by (a) eccentric muscle damage, (b) muscle obtained from intensive care unit patients and (c) in response to exercise training were obtained from three published articles (see Methods). The mean values for these selected genes were then plotted for patients in the present study that had either less than or more than 5% weight loss. As can be observed, no single gene, for each of these 'comparative' conditions, was differentially expressed; thus, the gene expression profile of cancer cachexia does not resemble muscle damage, sepsis-induced degeneration or exercise training status. Error bars represent SEM.