Introduction
Fat accumulation outside adipose tissue, i.e. ectopic fat in liver and muscle, is linked to decreased insulin sensitivity and type 2 diabetes [
1,
2]. However, intervention studies are needed to test a putative causal relationship between changes in ectopic lipid deposition and insulin sensitivity. Diet-induced weight loss in obese individuals is associated with reduction of fat in liver and skeletal muscle [
3‐
5]. This has been linked to improved insulin sensitivity but has not been a universal finding. Notably, there are conflicting data regarding the effect of macronutrient composition on ectopic fat [
3‐
7]. Two recent studies on obese postmenopausal women found that a Paleolithic diet consumed ad libitum with a moderately decreased carbohydrate intake and a high content of mono- and polyunsaturated fatty acids effectively reduced liver fat [
8,
9]. Furthermore, a Paleolithic diet efficiently improved glucose tolerance in overweight individuals and in people with type 2 diabetes [
10‐
12].
In contrast to the well-established relationship between diet-induced weight loss and decreased liver fat, it remains unclear whether exercise training decreases liver fat independently of weight reduction. An earlier intervention study in participants with type 2 diabetes reported that 4 months of either aerobic or resistance training was associated with a slight decrease in liver fat [
13]. However, aerobic exercise combined with diet intervention does not appear to cause further liver fat reduction compared with diet intervention alone [
14,
15]. To our knowledge, no prior study has investigated liver fat changes associated with diet in combination with both aerobic and resistance training. Interestingly, fetuin-A, a multifunctional protein secreted from both liver and adipose tissue, has been suggested as a putative link between insulin resistance and liver and adipose tissue function [
16,
17].
Several studies show that individuals with obesity, insulin resistance and type 2 diabetes have higher intramyocellular lipid (IMCL) content compared with lean and healthy individuals [
18,
19]. In obese individuals, weight reduction decreases IMCL content and simultaneously improves insulin sensitivity [
20,
21]. However, lean endurance-trained athletes exhibit IMCL content measured at rest that is nearly as high as that of people with type 2 diabetes, but with concomitant normal insulin sensitivity, referred to as the athlete’s paradox [
18]. Moreover, IMCL content is reduced immediately following an acute bout of aerobic exercise in young lean individuals [
22]. This suggests that IMCL is an important intracellular source of energy during exercise in people with high insulin sensitivity. This dynamic response does not occur in obese individuals, who show unchanged IMCL content after 1 h of cycling [
23]. Notably, a 12 week exercise intervention in individuals with type 2 diabetes found an increased IMCL content with concomitantly increased insulin sensitivity [
24].
Owing to these inconclusive data, it is of major interest to study how a combination of diet intervention and aerobic and resistance training influences ectopic fat deposition and tissue-specific insulin sensitivity in individuals with type 2 diabetes. We therefore tested the hypothesis that overweight individuals with type 2 diabetes on a 12 week Paleolithic diet would exhibit a decrease in liver fat and IMCL content, associated with an improvement in hepatic and peripheral insulin sensitivity. Moreover, we hypothesised that combined aerobic and resistance exercise training would lead to a further improvement in liver fat and peripheral insulin sensitivity.
Discussion
This intervention with a Paleolithic diet in overweight individuals with type 2 diabetes showed decreased ectopic lipid accumulation in liver and soleus muscle, as well as improved peripheral insulin sensitivity. On a group level, the addition of combined resistance and aerobic exercise training to the diet intervention reduced the effect on muscle and liver fat content. This was due to the considerable heterogeneity in response to exercise. Decreased liver fat during the intervention was strongly associated with reduction in plasma fetuin-A levels in both intervention groups. This was linked to an improvement in adipose tissue insulin sensitivity.
Current guidelines for non-alcoholic fatty liver disease (NAFLD) recommend lifestyle interventions involving diet and exercise to decrease liver fat [
33]. However, earlier studies suggest a complex relationship between liver fat and the effect of lifestyle interventions. Two short-term studies showed that weight reduction by a low-carbohydrate diet effectively reduced liver fat [
4,
5]. We found a reduction in liver fat after a Paleolithic diet with a moderately reduced carbohydrate content. We anticipated that our combined intervention in individuals with type 2 diabetes would show an additional fat-decreasing effect on the liver, as a recent meta-analysis concluded that exercise reduces hepatic fat content [
34]. Unexpectedly, three individuals in the PD-EX group showed a clear increase in liver fat. After exclusion of these three individuals, we found that liver fat decreased significantly in both study groups, with no difference between groups, while all other comparisons were unaltered. There may be several explanations for the unexpected response regarding liver fat in some individuals, including a prolonged exercise-induced increase in liver fat in some individuals or increased metabolic flexibility. The three individuals whose liver fat increased during 12 weeks of exercise thus showed decreased or unchanged inflammation (plasma CRP levels), triacylglycerol levels and liver enzymes, indicating improved metabolic health in this subgroup. A third possibility is that the participants did not refrain from exercise for 48 h before the examination as they were supposed to, leading to a decrease in muscle fat and an increase in liver fat.
In healthy people with normal weight and in overweight individuals, an acute bout of aerobic exercise immediately increases liver fat [
35,
36]. This exercise-induced increase in liver fat seems to be mainly due to the rise in plasma NEFA during and after exercise [
35,
36]. All participants in our study were told to refrain from exercise for 48 h prior to liver fat examination. A possible explanation for the conundrum could be that an exercise-induced increase in liver fat may last longer than 2 days in some individuals.
Lipolysis and inflammation are closely linked [
37]. During our intervention, decreasing
TNFα expression in adipose tissue was associated with decreasing plasma NEFA levels. Moreover, we found an association between liver fat and suppressibility of NEFA production, highlighting the importance of plasma NEFA concentration for hepatic lipid content. In NAFLD, most triacylglycerols in the fatty liver originate from plasma NEFA, and most plasma NEFA originate from adipose tissue [
38]. Plasma NEFA uptake in liver cells and esterification into hepatic triacylglycerols are insulin-independent, depending only on the plasma NEFA concentration [
39]. Notably, the percentage of plasma NEFA taken up by the liver remains constant both during and after exercise [
40].
Plasma NEFA concentrations are also important for IMCL content, and we found that fasting plasma NEFA levels were closely associated with IMCL content of the tibialis anterior muscle. In healthy individuals, IMCL is used as an energy substrate during exercise and is replenished during recovery. Endurance athletes who performed 3 h of cycling exercise had a 20% decrease in IMCL content in the legs and a simultaneous 38% increase in IMCL content in the non-exercising arms [
22]. Importantly, obesity and type 2 diabetes are associated with a lack of this dynamic response to exercise, which may be related to continuously increased plasma NEFA levels [
23]. However, if adipose tissue lipolysis is reduced with a nicotinic acid analogue, plasma NEFA levels are reduced and the decrease in IMCL content during one bout of exercise is more pronounced [
41].
Another contributing factor to the heterogeneous response in the combined diet and exercise intervention is the intake of carbohydrates. Suppression of adipose tissue lipolysis through insulin administration or carbohydrate ingestion leads to a reduction in plasma NEFA levels. Indeed, glucose ingestion during exercise causes plasma NEFA levels to decrease below fasting levels during and after exercise [
36,
42]. Accordingly, glucose supplementation during and after cycling prevents an increase in liver fat during the recovery phase [
36]. Furthermore, studies with isoenergetic diets show a decrease in liver fat only with high-carbohydrate diets, not with high-fat diets [
6,
7]. Although energy intake was ad libitum in our study, participants reported decreases in carbohydrate and total energy intake. More detailed studies of macronutrient intake in relation to exercise are therefore of interest regarding the effects on hepatic lipid content.
Fetuin-A is secreted mainly from liver and adipose tissue and is elevated in type 2 diabetes and NAFLD [
17,
43,
44]. Circulating fetuin-A levels have been associated with severity of liver steatosis, independently of insulin resistance, and with non-alcoholic steatohepatitis [
45,
46]. Our results showed a strong association between fetuin-A levels and changes in liver fat content. Changes in circulating fetuin-A levels were also associated with the ability to suppress NEFA production. Since fetuin-A can be secreted by both hepatocytes and adipocytes, it remains unclear whether fetuin-A secreted by the liver influences adipose tissue or the other way round.
The PD group increased insulin clearance, which might have been due to the improvement in liver fat content. Hepatocytes thus show impaired insulin clearance when loaded with triacylglycerols in vitro [
47], and liver fat is inversely related to insulin clearance in vivo [
48].
Improvement of hepatic insulin sensitivity was less pronounced in our intervention: only if normalised by insulin, the PD group increased hepatic insulin sensitivity. This may relate to the relatively well-preserved hepatic insulin sensitivity in our study cohort. In most studies, diet-induced weight loss in people with type 2 diabetes causes an increase in hepatic insulin sensitivity, but some authors report it unchanged [
49,
50].
A limitation of our study is that the insulin dose might have been too high to detect changes in hepatic insulin sensitivity. Moreover, target plasma glucose during the euglycaemic clamp was 8 mmol/l, which may not represent euglycaemia, especially after the intervention when fasting glucose was normalised. A lower insulin dose and a lower glucose target during the clamp studies might have detected more subtle changes in hepatic insulin sensitivity. Indeed, we have previously demonstrated an improvement in HOMA-IR after 5 weeks and 6 months following a Paleolithic diet in healthy overweight participants [
8,
9].
Another limitation is that we had to exclude soleus muscle measurements from three participants because we could not separate the intramyocellular and extramyocellular lipid signals. This is a known technical difficulty related to the fact that soleus muscle fibres are not aligned in parallel to the main magnetic field. This shortage of data limits our ability to draw conclusions regarding the effects of the intervention on different skeletal muscle types. Finally, analyses of gene variants that may influence liver fat accumulation, e.g. PNPLA3, are of interest in future intervention studies.
In conclusion, our results indicate that an exercise intervention is associated with a heterogeneous response in liver fat content in obese individuals with type 2 diabetes, despite improved metabolic health. Further studies are needed to understand how exercise changes liver fat and hepatic insulin sensitivity in relation to energy balance and macronutrient intake among individuals with obesity and type 2 diabetes.
Acknowledgements
The authors thank the study participants; the research nurses I. Arnesjö, K. Iselid, L. Uddståhl, C. Ring and L. H. Bergman at the Clinical Research Centre, (Umeå University Hospital, Umeå, Sweden) for performing the clamp examinations; A. Tellström (Department of Public Health and Clinical Medicine, Umeå University, Umeå, Sweden) for guiding participants during the diet intervention; and K. Lundgren (Swedish Metabolomics Centre, Swedish University of Agricultural Sciences, Umeå, Sweden) for the GC-MS analysis of [6,6-2H2]glucose. Parts of this paper were presented at the 52nd Annual Meeting of the European Association for the Study of Diabetes, Munich, Germany, 12–16 September 2016.