Short-term overfeeding of zebrafish with normal or high-fat diet as a model for the development of metabolically healthy versus unhealthy obesity

Obesity is characterized by an increase in adipose tissue (AT) accumulation in presence of positive energy balance and is a risk factor for metabolic and cardiovascular diseases, and premature mortality [1‐3]. Depending on body fat distribution and AT phenotype obese individuals differ in their susceptibility to obesity-associated diseases, such as hepatic steatosis and type 2 diabetes [4‐6]. Compared to the metabolically healthy obese (MHO) phenotype, which is characterized by a beneficial AT distribution and AT phenotype, the metabolically unhealthy obese (MUO) phenotype shows a deleterious AT distribution and phenotype with more visceral fat, bigger adipocyte size, and inflammatory processes [7]. However, the identification of mechanisms involved in the determination of body fat distribution and/or AT phenotype during obesity development are limited by the lack of suitable in vivo models, in which metabolically healthy versus metabolically unhealthy AT accumulation can be specifically induced.

Recently, the zebrafish (Danio rerio) has emerged as an alternative vertebrate model for the study of lipid metabolism and metabolic diseases, such as obesity, type 2 diabetes and hepatosteatosis [8, 9]. Organs and tissues of zebrafish are similar to those of humans in structure and function [10], and regulation of energy homeostasis at the neural and endocrine level is conserved [8, 11]. Moreover, diet-induced obesity in zebrafish shares common pathophysiological pathways with obesity in mammals [12‐15]. Because of this, the zebrafish has become increasingly important for the identification of genes or potential drugs regulating lipid metabolism, AT accumulation and associated processes [9, 16].

We aimed to establish the zebrafish as a model for the analyses of mechanisms involved in the development of MUO or MHO, respectively. To this end, we evaluated different models of short-term diet-induced obesity by overfeeding of adult zebrafish with normal fat or high fat diet, respectively, and assessed the effect on AT accumulation, AT phenotype and the occurrence of associated metabolic alterations.

We assessed diet-induced obesity in zebrafish by overfeeding adult male zebrafish with normal-fat-diet (NFD-OF) or high-fat-diet (HFD-OF) compared to physiological control diet for 8 weeks (Fig. 2a). Both, NFD-OF and HFD-OF fish had an enlarged belly compared to control fish (Fig. 2b). Control zebrafish maintained their weight during the 8-week feeding experiment. While fish in the NFD-OF group almost doubled their weight, the weight gain of HFD-OF fish was not as prominent (Fig. 2c). The weight gain observed under overfeeding conditions was accompanied by an increased growth in both groups (Fig. 2d). Both NFD-OF and HFD-OF fish showed a significant increase in body mass index (BMI, Fig. 2e) and Fulton’s condition index (Fig. 2f), which was considerably higher in NFD-OF compared to HFD-OF fish. Body fat percentage was increased in both NFD-OF and HFD-OF fish but did not significantly differ between the OF groups (Fig. 3a). HFD-OF but not NFD-OF fish showed significantly elevated blood glucose levels compared to control zebrafish (Fig. 3b) and presented significantly elevated plasma triglyceride (Fig. 3c) and cholesterol (Fig. 3d) levels as well as a prominent ectopic accumulation of lipids in liver (Fig. 3e) and muscle (Fig. 3f). In line with this, immunoblot analysis of phospho-Akt and total Akt protein levels in liver lysates demonstrated higher levels of phospho-Akt in HFD-OF zebrafish compared to NFD-OF or control zebrafish indicating an early state of insulin resistance (Fig. 3g). Analyses of fat distribution by MRI and haematoxylin-eosin staining of zebrafish cross sections revealed an increase in both subcutaneous and visceral AT in NFD-OF and HFD-OF compared to control zebrafish (Fig. 3f). Both, NFD-OF and HFD-OF groups showed a tendency towards an increase in adipocyte number in visceral AT. In subcutaneous AT adipocyte number was significantly increased in NFD-OF fish, while there was only a marginal increase in adipocyte number in HFD-OF fish (Fig. 3h). Visceral and subcutaneous adipocytes were significantly larger in NFD-OF and HFD-OF compared to control fish. Importantly, HFD-OF zebrafish had larger visceral but smaller subcutaneous adipocytes compared to NFD-OF zebrafish indicating the presence of a metabolically unhealthy AT phenotype (Fig. 3i).

To assess the molecular differences between NFD-OF and HFD-OF fish in AT and liver in more detail, we determined the expression of marker genes for lipid storage and inflammation, which had been previously associated with different obesity phenotypes in rodents and humans [24]. Due to the small amount of subcutaneous AT especially in control zebrafish, we had to restrict the analyses of marker gene expression to visceral AT. In visceral AT, we did not detect significant alterations in the expression of the adipocyte differentiation marker pparg after overfeeding with NFD (NFD-OF) or HFD (HFD-OF). However, in HFD-OF fish pparg was significantly lower compared to NFD-OF fish. Furthermore, we analysed expression of the lipid storage markers fabp11a, which is the zebrafish ortholog of human FABP4 [25‐27], as well as fasn and srebf1, which is an upstream regulator of fasn expression. While fabp11a expression was only slightly increased in response to overfeeding, fasn and srebf1 were upregulated in NFD-OF but not altered in HFD-OF fish. Similarly, expression of lpl, which is a lipoprotein lipase and a central regulator in lipid metabolism, was significantly higher in NFD-OF zebrafish but remained unchanged in HFD-OF zebrafish compared to the control group. In contrast, expression of the lipases atgl and hsl was not affected by NFD and HFD overfeeding. Analyses of the inflammatory markers il1b and tnfa revealed a tendential upregulation of il1b in HFD-OF fish indicating slightly increased AT inflammation, while tnfa remained unchanged (Fig. 4a). Analyses of the liver expression profile showed significantly increased fabp11a expression in HFD-OF but not NFD-OF zebrafish. Moreover, fasn expression in the liver was upregulated in NFD-OF but significantly downregulated in HFD-OF when compared to control zebrafish. In line with these data, srebf1 was increased in NFD-OF and decreased in HFD-OF fish when compared with control fish. Expression of the lipases atgl, hsl and lpl and the inflammatory marker il1b was not changed in fish included in the NFD-OF or HFD-OF group. In contrast, tnfa expression in the liver was increased under both overfeeding conditions. In addition, col1a1a expression was increased only in response to HFD but not NFD overfeeding suggesting the presence of fibrosis in the liver of HFD overfed zebrafish (Fig. 4b).

We show here that overfeeding of adult zebrafish with either NFD or HFD results in an increase in body weight, which is at least partially due to an increased accumulation of body fat. Interestingly and importantly, the biological and metabolic phenotypes differ depending on the type of the diet. In contrast to NFD-overfed zebrafish, HFD-overfed zebrafish present adipocyte hypertrophy, especially in the visceral AT depot, ectopic lipid accumulation in the liver, and hyperglycaemia and crucial differences in the expression of marker genes for lipid metabolism, inflammation and fibrosis,– hence a phenotype commonly referred to as MUO.

Different models of diet-induced obesity in zebrafish provided evidence that the pathophysiological pathways underlying diet-induced obesity in zebrafish are comparable to that of mammals [13, 14]. Most of those studies compared lean and obese animals and were based on excessive overfeeding of adult zebrafish with their “natural” diet in aquaculture conditions, i.e. flakes or artemia. We have substantially extended the published protocols by subjecting zebrafish to specific diets, i.e. NFD or HFD, over an 8-week-period. Our results show that dependent on the type of diet, one can not only induce obesity but also specifically trigger the development of a MHO versus a MUO AT phenotype as indicated by crucial differences in numbers and sizes of subcutaneous and visceral adipocytes, blood glucose, triglyceride and cholesterol levels, liver lipid accumulation and transcript levels of lipid metabolism and inflammation markers.

Compared with the MHO phenotype, the MUO phenotype is associated with characteristic alterations in the metabolic and immune function in human AT. These alterations include decreased expression of genes involved in lipid metabolism and increased expression of markers of AT inflammation [28]. Similarly to human studies, we observed a significant downregulation in the AT expression of several marker genes involved in lipid metabolism including pparg, fasn and lpl in HFD-OF compared to NFD-OF zebrafish. However, based on the striking increase in fasn expression we observed in both AT and liver upon overfeeding of zebrafish with a normal fat diet one would have expected a similar increase in srebf1 expression, which is an upstream transcription factor regulating fasn expression. However, fasn expression is not only regulated by Srebf1 but also by other upstream regulators, such as Usf [29]. Hence, we cannot exclude that these factors exert an additional impact on fasn expression. Nevertheless, the fact that we observed increased AT lpl expression in NFD-OF but not in HFD-OF fish further supports our model for the induction of metabolically distinct obesity phenotypes in zebrafish. Previous data suggested that lpl expression and activity in obese subjects are downregulated in an insulin-resistant compared to an insulin-sensitive state [30, 31], and that an increase of lipoprotein lipase in adipocytes improves glucose metabolism in HFD-induced obesity [32].

We observed an increase in plasma triglyceride and cholesterol levels and ectopic liver lipid accumulation in zebrafish overfed with HFD compared to zebrafish overfed with NFD. These results point to distinct metabolic phenotypes of the two groups. However, the difference in blood glucose levels between the groups was only modest and might reflect the short duration of the study. We assume that this modest increase in blood glucose might represent early obesity-related alterations in glucose metabolism but not a manifestation of the disease. In fact, this assumption is underlined by the results from immunoblot analyses showing enhanced Thr308 phosphorylation of Akt (Thr307 for the zebrafish protein) in the HFD-OF group. Unfortunately, we cannot provide data on Ser473 phosphorylation of Akt since we were lacking an antibody specifically detecting the zebrafish protein. Furthermore, it would be interesting to analyse how Akt phosphorylation in liver tissue of zebrafish subjected to NFD and HFD overfeeding is affected by an insulin challenge. Both, enhanced Thr308 and Ser473 phosphorylation of Akt have been associated with an early state of insulin resistance upon HFD overfeeding [33‐36]. In our opinion, the observation of increased Thr308 phosphorylation of Akt in liver tissue upon HFD overfeeding might represent a valuable strength of the study protocol allowing the analyses of obesity-related alterations in AT and liver which occur with AT accumulation before or during the development of obesity-related metabolic alterations. Hence, the model described in this study provides a unique tool to study physiological mechanisms involved in maintaining or disrupting metabolic health in obesity. The better understanding of these mechanisms might enhance the identification of novel therapeutic approaches specifically targeting the metabolic phenotype of obese patients. In this context, the zebrafish might be a particularly well-suited model organism. They share a considerable amount of genetic identity with humans and sophisticated mutagenesis, transgenesis, and screening strategies are available and can be used with an economy that is not possible in other vertebrates [10].

It should be noted that for the establishment of the short-term feeding protocol described in this study, we used male zebrafish only. Previous short-term and long-term analyses suggested that female zebrafish show a similar response to diet-induced obesity compared to male zebrafish with increased body length, body weight and BMI although to different extent [13, 14]. Reason for this might be that female zebrafish constantly produce eggs, which contain large amounts of lipids. In fact, it has been demonstrated that in adult female zebrafish ovaries can account for up to 29% of body weight compared to less than 2% for the testis in male zebrafish [13]. Because of this measurement of weight and body fat content in female adult zebrafish does not necessarily reflect AT accumulation but may be an indicator of oocyte growth as well. Since all zebrafish included in this study were either histologically analysed or subjected to organ preparation for expression analyses we can exclude that the occurrence of diet-induced sex reversal as a potential bias. It might be of interest for future studies to analyse whether the here provided feeding protocol is informative for both sexes. In this regard, the analysis of never-mated female zebrafish might provide important information on the transferability of the method to female zebrafish. Furthermore, obesity development and progression might be influenced by the age of fish, the genetic background and the body weight state at the start of the experiment. The here described protocol was established using 3 to 6 months old zebrafish of the AB strain. For the use of a different zebrafish strain, such as Tu, TL or WIK, specific parameters such as amount of food or duration of feeding experiment may need to be adjusted. Given the simplicity of the feeding protocol and the detailed phenotypic characterization described in our study, we believe that it can be easily applied to answer the questions whether sex, age or genetic background of zebrafish influence the response to NFD and HFD overfeeding and the associated development of metabolically distinct obesity phenotypes.

Our study is strengthened by the extensive phenotypic characterization including determination of total body fat content, adipocyte numbers and sizes in both subcutaneous and visceral AT, blood glucose, triglyceride and cholesterol levels, ectopic lipid accumulation in the liver and expression levels of marker genes for lipid metabolism and inflammation, which allows the identification of diet-dependent metabolically distinct obesity phenotypes.

In summary, we have established a protocol for the induction of MHO versus MUO in zebrafish providing an attractive model to study regulatory mechanisms underlying the determination of distinct obesity-related AT phenotypes and metabolic state.