This study, using both non-targeted as well as targeted proteomic approaches applied to biofluids and blood cells to elucidate changes in their proteomes induced by extended fasting, revealed substantial between-subject variability which may mask the true biological effect of fasting on the proteome. This suggests that any nutrigenomic experiment require a sufficient number of replicates and matched samples to identify the usually rather subtle changes that dietary maneuvers may impose. Levels of between-subject variability were higher in all biofluids samples as compared to the cellular proteomes. In this study, PBMC exhibited the largest number of significant changes in protein levels following the extended fast. RF analysis on the entire data set revealed that in PBMC the RhoGDI2 and plasma ApoA4 proteins were the most significant biomarkers for extended fasting. RF analysis of the multiplex immunoassay data set revealed leptin and MMP-3 as potential biomarkers for extended fasting.
It should be noted that the between-subject variation may in part be due to the heterogeneous composition of our study cohort with regard to various parameters, including gender and BMI. Consequently, this type of variability can be reduced by selecting a more homogeneous study population. We observed markedly less inter-individual difference in the PBMC proteome than we observed previously in the PBMC transcriptome of PBMC from human volunteers who were participating in a study with a similar baseline design . In our study 0.4% of the PBMC proteins detected differed significantly between individuals compared with 39% of PBMC transcripts .
Most interesting biomarkers that are involved in metabolic pathways, as well as those related to inflammation and oxidative stress, are present only in very low concentrations in biosamples. In these cases, their low abundance prevents detection by classical proteomic techniques such as 2D-electrophoresis . Multiplex immunoassays are therefore an interesting alternative because they are more sensitive compared to 2D-electrophoresis . Indeed, a wide range of plasma protein changes caused by extended fasting were detected using the targeted multiplex immunoassay method. However, a big advantage of 2D-electrophoresis is that it can visualize isoforms and possible protein modifications, whereas with an antibody-based approach this depends on the availability of isoform-specific antibodies.
CLIC1 in PBMC was the protein that exhibited the largest fold-change as a result of the extended fasting. In general, chloride channels play important roles in the regulation of cellular excitability, transepithelial transport, cell volume regulation, and acidification of intracellular organelles . CLIC1 is responsible for stabilization of the membrane potential, which could be influenced by the nutrient status in PBMC. Moreover CLIC1 seems to be associated with cellular stress response mechanisms and starvation may be interpreted as a metabolic stress condition .
Besides CLIC1, levels of RhoGDI2 and ApoA4 were the best indicators of the extended fasting state based on analysis of all the proteomic data. Insulin has the ability to relocate RhoGDI2 to the membrane . Since plasma glucose and insulin levels are low during fasting, the relative amount of RhoGDI2 in the cytosol of PBMC might be predicted to increase during extended fasting. Since by 2D preferentially cytosolic proteins are detected, RhoGDI2 is expected to rise in spot intensity. However, a decrease was noticed in our experiment indicating that cytosolic RhoGDI2 concentration is influenced by other factors like posttranslational modification or turn-over rate.
ApoA4 is a major component of chylomicrons and is synthesized by the small intestine. It is proposed to represent a circulating satiety signal  and after fasting, therefore, may decrease its levels .
IGF-1, leptin and CRP were the proteins with the biggest negative and positive fold change upon the fasting challenge in the RBM data. Circulating IGF-1 levels are associated with dietary protein intake [25, 26]. This could explain why after an extended fast without protein intake, IGF-1 is decreased . The decrease of leptin is as expected because it is known to be a positive marker for fasting or weight loss [28–30]. Chan et al.  have shown that leptin and IGF-1 levels decrease after a 72 h fasting. Our correlation analysis clustered IGF-1 and leptin suggesting that adipose tissue, as the main source of leptin, and liver, as the origin of circulating IGF-1, undergo metabolic adaptation via a closely linked mechanism. CRP is an acute phase protein and its elevation after this short term fasting might therefore be expected, although after long term fasting or caloric restriction CRP is usually down-regulated . RF analysis applied to the RBM data revealed leptin and MMP-3 as the most obvious biomarkers for a fasting challenge. No direct interaction between leptin and MMP-3 has been reported so far. A possible mechanism could be the inverse correlation between leptin and adiponectin levels [19, 33] since the leptin/adiponectin ration changes during extended fasting. Adiponectin by itself has the ability to increase TIMP1, which is able to inhibit the activity of MMPs [34, 35]. However, adiponectin plasma levels did not change significantly (p = 0.45) upon the extended fasting.
Most interestingly, the correlation analysis clustered CRP closely together with adiponectin, serum amyloid P and alpha-2-macroglobulin. Except for adiponectin, a networking interaction analysis using MiMI, places those proteins together with leptin and RhoGDI2. An interaction between leptin and CRP, which are positively correlated in normal weight, overweight and obese subjects , was described by Chen et al . Alpha-2-microglobulin is described as a leptin binding protein . This interaction may be involved in the clearance of leptin by the alpha-2-microglobulin receptor.
Another observation from the networking analysis was an interaction between PAI-1, MMP-3, MCP1 and TIMP1. As mentioned above, TIMP1 inhibits the activity of MMPs [34, 35], whereas both PAI-1 and MCP1 are inactivated through proteolysis mediated by MMP-3 [39, 40]. This interaction between MMP-3 and MCP-1 may be the reason that MMP-3 is ranked relatively high by the RF analysis. An inhibition or decrease of PAI-1 leads to a reduction of fat depots and adipocyte volume . Several studies have shown that PAI-1 is increased in adipose tissue and plasma in obese humans and that plasma PAI-1 concentrations decrease after weight loss or fasting [42, 43], which suggests a role for this proteinase inhibitor in the development and maintenance of obesity. Jensen et al.  reported that a low glycemic index diet in overweight adults could be beneficial in regulating fasting concentrations of the cardiovascular disease risk factor PAI-1 showing that PAI-1 concentrations are sensitive to the nutritional status. PAI-1, TIMPs and MMPs play a role in extracellular matrix remodeling [41, 45, 46]. Interestingly, a correlation has been described between CRP and PAI-1 [47, 48], linking both interaction networks.