Very low-calorie diets (VLCD) induce significant weight loss and metabolic benefits, such as improvements in whole-body glucose disposal, and are commonly employed in the treatment of obesity and its related comorbidities (i.e diabetes). However, 25% of VLCD induced weight loss is attributable to declines in lean body mass, comprised largely of muscle (1). Given the essential role of muscle in maintaining whole-body metabolic health it is desirable to maintain muscle mass during caloric restriction. Furthermore, this may be achievable with chronic resistance exercise training (RET). However, it is unclear whether adjuvant RET has an additive metabolic health benefit when provided alongside VLCD. In the present study we implemented an unbiased approach, in the form of untargeted metabolomics, to determine the impact of exercise training in the form of RET and high intensity interval training (HIIT) adjuvant to VLCD, and to further understand the biology potentially underlying the metabolic benefits of VLCD.
Obese and overweight men (n=26) were recruited to a 6-week intervention study. Participants were randomly allocated into VCLD (800kcal/day) only (n=10; 46±8y; 104±12kg; BMI 32±4, mean±SD), VLCD+RET (n= 8; 40±8y; 99k±10kg; BMI 32±2) or VLCD+HIIT (n=8; 46±11y; 101±12kg; BMI 33±3). Body composition was determined by dual-energy X-ray absorptiometry (DXA) before and after the intervention period. Fasted plasma samples collected before and after intervention were analysed using ultra-high performance liquid chromatography operating dual phase HILIC and C18 RP separation coupled to a high resolution mass spectrometer, with data acquired in both positive and negative ionisation modes. The Random Forest (RF) algorithm was used to identify metabolite features most informative in classifying pre- and post-intervention samples which formed a reduced dataset for analysis of differences in metabolite abundance, metabolite identification and network mapping. Annotation of metabolite identity was assigned using network integration performed with the PIUMet algorithm.
Loss of lean body mass was larger in VLCD only (4.4±1.9kg) than VLCD+RET or VLCD+HIIT (3.9±1.8kg and 3.9±1.3kg, respectively), although not significantly. The RF algorithm identified the top 10 metabolites from each polarity and ionisation mode which classified pre and post intervention. Principal component analysis (PCA) using the reduced dataset resulted in clear separation of pre- and post-intervention samples (Fig. 1A). When further stratified into groups, excellent clustering was seen in VLCD+HIIT (Fig. 1D) with low separation in VLCD (Fig. 1B) and no separation in VLCD+RET (Fig. 1C). Metabolite abundance revealed similar patterns of expression for all groups from pre- to post-intervention, however differences were larger and significant (p<0.05) in VLCD only and VLCD+HIIT while expression changes in VCD+RET were not significant.  Putative metabolite identification found benzenoids, derivatives of carboxylic acids, and several species of lipids were primarily involved in separation of samples before and after intervention.
These results demonstrate that although the addition of exercise training to a VLCD intervention did offer some protection against the loss of lean body mass, albeit non-significantly, adjuvant RET did not alter the fasting state metabolomic signature from VLCD alone. Conversely adjuvant HIIT did elicit some differences which may warrant further investigation.