Introduction

Mitochondria are highly dynamic organelles that frequently fuse and divide in response to cellular energy demands, differentiation or pathological conditions [1–3]. The endoplasmic reticulum (ER) is an extensive, morphologically continuous network of membrane tubules in the eukaryotic cell. It performs a variety of functions in cells including lipid synthesis, intracellular Ca2+ regulation and protein secretion [4]. Recently, it was reported that the ER plays an active role in defining the sites of mitochondrial division [5]. In fact, the two organelles physically interact, forming specialised contacts referred to as mitochondria-associated ER membrane (MAM). MAM has been linked with lipid transfer, apoptotic calcium signalling and mitochondria–ER calcium homeostasis [6]. Furthermore, MAM has recently been shown to regulate mitochondrial shape and motility, energy metabolism and redox status and to be central to the modulation of various key processes such as ER stress, autophagy and inflammasome signalling [7].

In vertebrates, mitofusin-1 and -2 (MFN1 and MFN2) are involved in mitochondrial fusion and dynamin-related protein 1 (DRP1) and mitochondrial fission factor (MFF) control mitochondrial fission [8, 9]. During the fission cycle, DRP1 first binds to MFF on the surface of mitochondria, followed by entry into a complex that includes ER proteins at the ER–mitochondria interface [10]. Reciprocally, ER-localised inverted formin 2-mediated actin polymerisation is required for efficient mitochondrial fission [11], suggesting that ER contact is required for initiation of mitochondrial fission. Recently, MFN2 has been shown to be involved in MAM architecture and to regulate insulin signalling and insulin sensitivity in muscle and liver [12]. Furthermore, MAM integrity is required for insulin signalling and is implicated in hepatic insulin resistance [13]. These results led us to hypothesise that DRP1 plays a role in glucose homeostasis through organelle communication between the ER and mitochondria. To investigate such a role of mitochondrial fission, we generated mice with liver-specific deletion of Drp1 (Drp1LiKO). Drp1LiKO mice showed decreased fat mass and were protected from high-fat diet (HFD)-induced obesity. Further analysis revealed that defects of mitochondrial fission in the liver cause ER stress, leading to induction of fibroblast growth factor 21 (FGF21), a metabolic regulator of glucose and lipid homeostasis. In this study, we show that mitochondrial fission in the liver plays an important role in energy homeostasis through induction of FGF21.

Methods

Animal experiments

Drp1LiKO mice were generated by crossing Drp1 flox/+ mice [14] with Alb-Cre mice. Mice were fed ad libitum with a normal chow diet (NCD) (5.4% fat, CRF-1; Orient Yeast, Tokyo, Japan) and kept under a 12 h light–dark cycle. For the HFD study, 4-week-old mice were put on an HFD (24% fat, lard fat, 188.28 kJ% fat, D12451; Research Diets, New Brunswick, NJ, USA) for 16–24 weeks. See electronic supplementary material (ESM) Methods for further details.

Western blot

The in vivo insulin sensitivity was assessed by quantification of Akt phosphorylation in response to exogenous insulin. See ESM Methods for details.

Microscopic analyses

For histochemical and immunohistochemical analyses, adipose tissue and livers were fixed, processed, stained and quantified. The morphologies of mitochondria and ER were analysed by electron microscopy. See ESM Methods for details.

mRNA analyses

Real-time quantitative RT-PCR was used to determine the relative expression levels of mRNA. Results were normalised to Gapdh expression. Liver mRNA from control and Drp1LiKO mice was subjected to microarray analysis. See ESM Methods for details.

Biochemical assays

Total liver lipid content was determined by the Folch method [15]. For VLDL secretion assays, mice were fasted for 4 h before being injected with 0.5 g/kg tyloxapol (Triton-1339; Sigma, St Louis, MO, USA). See ESM Methods for details.

Metabolic studies

Body weight, food intake and locomotor activity were measured. Glucose and insulin tolerance tests were performed as described [16]. See ESM Methods for details.

Oxygen consumption

The rate of oxygen consumption of mitochondria was measured using an oxygen electrode as described [14]. See ESM Methods for details.

Statistical analysis

All data are expressed as means ± SE and evaluated by Student’s t test using Microsoft Excel software (Mac 2011 version 14.3.5; Microsoft Japan, Tokyo, Japan), two-way ANOVA performed with Tukey’s post hoc test or repeated measures two-way ANOVA performed with the Bonferroni post hoc test using GraphPad Prism 6.0 software (GraphPad; San Diego, CA, USA). To detect an outlier, we performed the Grubbs’ test using the free statistical calculators: QuickCals, http://www.graphpad.com/quickcalcs/. Significance levels were set at p < 0.05, p < 0.01 and p < 0.001.

Results

Ablation of Drp1 protects mice from HFD-induced obesity

As shown in Fig. 1a, DRP1 protein was ablated in the liver of Drp1LiKO mice. RT-PCR revealed approximately 80% reduction of Drp1 mRNA levels in the liver of Drp1LiKO (Fig. 1b). On the NCD, no apparent difference was found in body weight of the Drp1LiKO mice compared with controls (ESM Fig. 1a). In striking contrast, when fed an HFD, although the food intake and locomotor activity were nearly identical (Fig. 1c, d), the body weight of Drp1LiKO mice was significantly lower than that of control mice (Fig. 1e). Liver and epididymal white adipose tissue (WAT) weights were decreased in Drp1LiKO mice (Fig. 1f, g), while there were no significant differences in the weight of other tissues such as kidney and testes (Fig. 1h). Collectively, Drp1LiKO mice showed decreased fat mass and were protected from HFD-induced obesity. When fed an HFD, hematoxylin–eosin staining of liver sections revealed that disorganised lobular parenchyma displayed inflammatory cell infiltration in Drp1LiKO mice (Fig. 1i). The inflammatory cells were identified as small cells with little cytoplasm and formed a cluster.

Fig. 1
figure 1

Ablation of Drp1 protects mice from HFD-induced obesity. (a) Immunoblot analyses of DRP1 protein from the livers of two control and two Drp1LiKO mice. The bar graph shows the intensities of protein bands quantified by densitometric scanning. (b) Drp1 expression levels determined by real-time quantitative RT-PCR analysis. (c) Food intake in 24-week-old HFD-fed Drp1LiKO and control mice (n = 5). (d) Locomotor activity in 18-week-old HFD-fed Drp1LiKO and control mice (n = 5). (e) Body weight of Drp1LiKO (grey squares, n = 8) and control mice (black squares, n = 8) fed an HFD from 6 weeks of age. (f) Representative image of 24-week-old HFD-fed Drp1LiKO and control mice. (g, h) Tissue weights of 24-week-old HFD-fed Drp1LiKO and control mice: liver and epididymal WAT (eWAT) (g, n = 5); kidney and testes (h, n = 5). (i) Haematoxylin–eosin staining of livers from 20-week-old control and Drp1LiKO mice on either NCD or HFD. Arrows indicate inflammatory cell infiltration. Scale bar, 50 μm. (j, k) Serum AST (j) and ALT (k) values in control and Drp1LiKO mice (n = 5). Black bars, control mice; grey bars, Drp1LiKO mice. Values are expressed as means ± SE. *p < 0.05, **p < 0.01 and ***p < 0.001, as calculated by Student’s t test (a–c, e, g, h, j, k) or two-way ANOVA (d)

Full size image

When fed an NCD, reduced inflammatory cell infiltration was found in Drp1LiKO mice (Fig. 1i). Reflecting hepatic inflammation, serum aspartate aminotransferase (AST) and alanine aminotransferase (ALT) levels were 3.17-fold and 2.93-fold higher in the HFD Drp1LiKO mice, respectively, when compared with control mice (Fig. 1j, k).

Hepatic lipid content in Drp1LiKO mice is comparable with that in control mice

We analysed the hepatic lipid content and the expression of the genes involved in lipid metabolism in the livers of HFD-fed mice. There were no significant differences in triacylglycerol, total cholesterol, total NEFA or total lipid contents in the livers of control and Drp1LiKO mice (Fig. 2a–d). The mRNA levels of the genes involved in lipogenesis and β-oxidation were analysed. The gene encoding sterol regulatory element binding transcription factor 1 (Srebp1c, also known as Srebf1) was expressed at lower levels in Drp1LiKO mice than in controls in both the fasting and refed condition (Fig. 2e), whereas expression levels of the genes encoding fatty acid synthetase (Fasn), acetyl-CoA carboxylase (Acc) and carnitine palmitoyltransferase Ia (Cpt1a) were no different (Fig. 2f–h). Collectively, lipogenesis was slightly impaired, while lipolysis was not affected in the liver of Drp1LiKO mice on an HFD.

Fig. 2
figure 2

Lipid profile. (a–d) Triacylglycerol (a), total cholesterol (b), total NEFA (c) and total lipid (d) levels in the livers of Drp1LiKO and control mice under either fasted or refed (for 4 h) conditions (n = 6). (e–h) Hepatic gene expression of Srebp1c (e), Fasn (f), Acc (g) and Cpt1a (h) was determined by real-time quantitative RT-PCR analysis (n = 5 per group). Values were normalised to Gapdh and are presented relative to fasting control set at 1. (i–k) Serum concentrations of triacylglycerol (i), total cholesterol (j) and total NEFA (k) in fasted and refed Drp1LiKO and control mice (n = 6). (l) Serum triacylglycerol levels in control (black squares) and Drp1LiKO mice (grey squares) after refeeding for 4 h, measured after tyloxapol injection. The bar graph shows the rate of hepatic triacylglycerol secretion (n = 5–7 per group). Data are expressed as means ± SE. Black bars, control mice; grey bars, Drp1LiKO mice. *p < 0.05, **p < 0.01 and ***p < 0.001, calculated by Student’s t test (l, triacylglycerol level), two-way ANOVA (a–k) or repeated measures two-way ANOVA (l, VLDL secretion)

Full size image

Decreased hepatic VLDL secretion in Drp1LiKO mice

The levels of serum triacylglycerol and total cholesterol were significantly lower in Drp1LiKO mice than in control mice after 17 h fasting, while the total serum NEFA level was no different (Fig. 2i–k). Since food intake increases hepatic fatty acid uptake and VLDL secretion, these lipid profiles were also analysed after refeeding for 4 h. The levels of serum triacylglycerol and total cholesterol were significantly lower in Drp1LiKO mice than in control mice. It was noted that the increase in serum triacylglycerol after feeding was completely negated in Drp1LiKO mice (Fig. 2i), suggesting that the defect may also reside in VLDL secretion in Drp1LiKO mice. Serum triacylglycerol values, reflecting hepatic triacylglycerol secretion, were measured after intraperitoneal tyloxapol injection. Drp1LiKO mice showed significantly reduced rates of hepatic triacylglycerol secretion (Fig. 2l). Reduced levels of serum triacylglycerol might be attributed to, in part, defective VLDL secretion.

Drp1LiKO mice display improved glucose metabolism

Drp1LiKO mice showed significantly improved glucose tolerance (assessed by an intraperitoneal glucose tolerance test [IPGTT]) compared with control mice (Fig. 3a), while no significant difference was found in plasma insulin levels (Fig. 3b). As a result, the HOMA-IR was markedly lower in Drp1LiKO mice (Fig. 3c). Consistent with this, an insulin tolerance test (ITT) revealed high insulin sensitivity in Drp1LiKO mice (Fig. 3d). This improved glucose profile was not observed in the NCD-fed mice (ESM Fig. 1 b, c). Analysis of the mRNA expression of genes encoding enzymes involved in glycolysis (glucokinase [Gck], pyruvate kinase [Pk1]) and gluconeogenesis (glucose 6-phosphatase [G-6-pase, also known as G6pc], phosphoenolpyruvate carboxykinase 1 [Pck1], peroxisome proliferator-activated receptor γ co-activator 1 α [Ppargc1a]) showed that Gck, G-6-pase and Ppargc1a were significantly downregulated in the liver of fasted Drp1LiKO mice (Fig. 3e–i).

Fig. 3
figure 3

Amelioration of glucose metabolism in Drp1LiKO mice. (a, b) Blood glucose (a) and plasma insulin (b) values during IPGTT of Drp1LiKO and control mice under HFD conditions (n = 11–16). (c) HOMA-IR scores in HFD-fed Drp1LiKO and control mice (n = 6). (d) Blood glucose profile during ITT of Drp1LiKO and control mice (n = 6). (e–i) Hepatic expression levels of the genes encoding glucose kinase (Gck) (e), pyruvate kinase (Pk1) (f), glucose-6-phosphatase (G-6-pase) (g), phosphoenolpyruvate carboxykinase (Pck1) (h) and peroxisome proliferator-activated receptor γ co-activator 1α (Ppargc1a) (i) determined by real-time quantitative RT-PCR analysis (n = 3–8 per group). Values were normalised to Gapdh and are presented relative to fasting control set at 1. Data are expressed as means ± SE. Black bars and squares, control mice; grey bars and squares, Drp1LiKO mice. *p < 0.05, **p < 0.01 and ***p < 0.001, calculated by Student’s t test (c), two-way ANOVA (e–i) or repeated measures two-way ANOVA (a, b, d)

Full size image

Loss of Drp1 altered the mitochondria–ER positional relationship

Electron microscopy analysis of mitochondria showed a dispersed pattern in the cytosol of hepatocytes from both control and Drp1LiKO mice, while the morphology of Drp1LiKO mitochondria was swollen (Fig. 4a–d). As shown in Fig. 4a, c, ER morphology in control hepatocytes showed a long tubular network located along mitochondria, and there were relatively equal intervals between mitochondria and ER. In striking contrast, the ER in hepatocytes from Drp1LiKO mice was swollen and irregular-shaped, disrupting the relative positional relationship between ER and mitochondria (Fig. 4b, d). A change in ER morphology was also evident, although less severe, in the liver of Drp1LiKO mice fed the NCD (ESM Fig. 2). Morphometric analysis revealed that the total number of mitochondria in hepatocytes from Drp1LiKO mice was significantly reduced but the number of large mitochondria (>10,000 μm2) was significantly increased (Fig. 4e, f). Despite the morphological change in mitochondria, the expression levels of the mitochondrial import machinery components, TOM20 and TOM40, and the mitochondrial respiratory chain complexes were not changed by loss of Drp1 (Fig. 4g, h). Consistent with these observations, the ATP content and mitochondrial respiratory activity were not altered in Drp1LiKO mice (Fig. 4i–k).

Fig. 4
figure 4

Altered positional relationship between mitochondria and ER in hepatocytes of Drp1LiKO mice on HFD. (a–d) Electron micrographs of livers from control (Ctrl) and Drp1LiKO mice on HFD at magnification ×3,000 (a, b) and ×7,000 (c, d). Traces of ER and mitochondrial profiles are also shown. M, mitochondria; dotted lines, ER outline. Scale bar, 1 μm. (e) Total mitochondria (Mt) number in 24 graphs in Drp1LiKO mice is presented as the percentage of control mice. (f) Distribution of mitochondrial areas <5,000 μm2 (black bars), 5,000–10,000 μm2 (grey bars) or >10,000 μm2 (white bars) is shown as a percentage of mitochondria measured. Control, n = 439; Drp1LiKO, n = 289. (g) Western blots representative of two mice showing the expression of mitochondrial membrane proteins, TOM20 and TOM40. The bar graph shows intensities of protein bands, quantified by densitometric scanning. (h) Western blots representative of two mice showing the expression of each subunit of mitochondrial respiratory complex. The bar graph shows intensities of protein bands, quantified by densitometric scanning. Values were normalised to β-actin and are presented relative to control set at 1. (i) Hepatic ATP content in Drp1LiKO and control mice under either fasted or refed conditions (n = 6). (j, k) Mitochondrial respiratory rates of complex II (j) and complex IV (k) are shown by oxygen consumption in control and Drp1LiKO mice liver mitochondria. Data are expressed as means ± SE. Black bars, control mice; grey bars, Drp1LiKO mice. *p < 0.05 and **p < 0.01, calculated by Student’s t test (e–h, j, k) or two-way ANOVA (i)

Full size image

Loss of Drp1 causes ER stress and activates the eukaryotic translation factor 2α–activating transcription factor 4–FGF21 pathway

Liver mRNA from control and Drp1LiKO mice was subjected to microarray analysis. Genes were selected using the criterion of a z score of ≥2, which identified 526 upregulated and 640 downregulated genes in Drp1LiKO mice. The top five upregulated genes are shown in Table 1. Among these, P8, Atf3 and Trb3 (also known as Trib3) are known to be ER stress-response genes [17–19], indicating that Drp1 defects induce ER stress. Meg3, a maternally expressed imprinted gene, has a function of tumour suppression by activation of P53 [20]. Insulin-like growth factor binding protein 1 is potently induced during ER stress by activating transcription factor 4 (ATF4) [21]. Further confirmation by real-time RT-PCR analysis revealed that ER stress-response genes were significantly increased even in mice on the NCD, and were further increased by the HFD (Fig. 5a–d). Phosphorylation of eukaryotic initiation factor 2 (eIF2α) is a critical early step in the regulation of protein synthesis during ER stress. Western blotting analy-sis revealed an increase in phosphorylation of eIF2α in the liver of Drp1LiKO mice along with a concomitant increase in its downstream targets ATF4 and C/EBP homologous protein (CHOP) (Fig. 5e). Recently, MAM dysfunction was found to be associated with hepatic insulin resistance [13]. In addition, liver-specific Mfn2-knockout mice showed glucose intolerance with hepatic insulin resistance [12]. Therefore, we evaluated the level of phosphorylated Akt (pAkt) in Drp1LiKO mouse liver. As shown in Fig. 5f, addition of exogenous insulin increased pAkt levels in both control and Drp1LiKO mouse livers. However, a significant reduction was found in the amount of pAkt after insulin stimulation in the liver of Drp1LiKO mice compared with control mice, indicating hepatic insulin resistance in Drp1LiKO mice.

Table 1 Top upregulated genes in Drp1LiKO mouse liver
Full size table
Fig. 5
figure 5

Increased ER stress and FGF21 expression in Drp1LiKO mouse liver. (a–d) Expression of the genes involved in ER stress. The expression levels of P8 (a), Atf3 (b), Trb3 (c) and Chop (also known as Ddit3) (d) were determined by real-time quantitative RT-PCR analysis of total RNA from liver samples taken from NCD- and HFD-fed mice (n = 4 per group). (e) Western blots representative of two mice showing the expression of phosphorylated eIF2α, total eIF2α, ATF4, CHOP and GAPDH. Bar graphs show intensities of protein bands, quantified by densitometric scanning. (f) The phosphorylation levels of Akt (pAkt) in the livers of control and Drp1LiKO mice fed either NCD or HFD. Bar graph shows intensities of protein bands, quantified by densitometric scanning. (g) The hepatic gene expression of Fgf21 in control and Drp1LiKO mice fed either NCD or HFD (n = 5 or 6). (h) Serum concentration of FGF21 in control and Drp1LiKO mice under the fasted condition (n = 3–8). (i, j) The expression of Fgfr1 (i) and Klb (j) in WAT of control and Drp1LiKO mice (n = 8). (k) Immunohistochemical analyses of FGF21 protein in liver sections from control (Ctrl) and Drp1LiKO mice. Arrowheads show concentrated cytoplasmic FGF21. Scale bar, 50 μm. Black bars, control mice; grey bars, Drp1LiKO mice. Data are expressed as means ± SE. *p < 0.05, **p < 0.01 and ***p < 0.001, as calculated by Student’s t test (e, i, j) or two-way ANOVA (a–d, f–h)

Full size image

FGF21 has recently emerged as a hepatic stress hormone and an effective metabolic regulator of glucose and lipid homeostasis [22–24]. Furthermore, Fgf21 expression has shown to be induced by the protein kinase R-like ER kinase (PERK)–eIF2α–ATF4 axis [25–27]. Activation of the elF2α–ATF4 axis was demonstrated in the liver of Drp1LiKO mice (Fig. 5e) so we examined Fgf21 gene expression in Drp1LiKO mice. When mice were fed with NCD, no statistical difference in Fgf21 mRNA expression was found between control and Drp1LiKO mice (Fig. 5g). Feeding with HFD resulted in a significant increase in Fgf21 mRNA expression in Drp1LiKO mice but not in control mice (Fig. 5g). Serum FGF21 concentrations in Drp1LiKO mice were significantly higher than those in control mice on NCD (Fig. 5h); these levels were increased further by an HFD burden. As shown in Fig. 5i, j, mRNA expression of the receptor complex for Fgf21, Fgfr1 and Klb, was increased in WAT of Drp1LiKO mice, suggesting that FGF21 has a possible feedforward effect on its own signalling pathway [28]. The expression of FGF21 was further assessed by immunohistochemistry. There was scattered and faint staining in control liver (Fig. 5k). In contrast, a remarkable increase in FGF21 protein was observed in the cytoplasm of hepatocytes from Drp1LiKO mouse livers (Fig. 5k).

Increased energy expenditure in Drp1LiKO mice

The adipocytes in WAT were smaller in Drp1LiKO mice than in control mice (Fig. 6a). The expression of genes involved in lipolysis, such as those encoding hormone sensitive lipase (Lipe) and adipose triacylglycerol lipase (Atgl, also known as Pnpla2), was significantly increased in adipose tissue of Drp1LiKO mice (Fig. 6b, c). The expression of the gene encoding adiponectin (Adipoq) was significantly increased in Drp1LiKO mice (Fig. 6d). However, no statistical difference was found in serum adiponectin levels when comparing control mice (40.39 ± 7.89 μg/ml, n = 9) with Drp1LiKO mice (40.01 ± 7.87 μg/ml, n = 10) (p = 0.94). Given that we observed no appreciable attenuation of food intake (Fig. 1c), we hypothesised that FGF21 might increase energy expenditure. As measured by indirect calorimetry, oxygen consumption in Drp1LiKO mice was significantly increased compared with that in control mice (Fig. 6e). The respiratory exchange ratio (RER) of Drp1LiKO mice was significantly lower during both light and dark cycles (Fig. 6f), indicating an increased use of lipids in Drp1LiKO mice. The effect of FGF21 on energy expenditure is attributed to its regulation of the thermogenic genes in WAT and brown adipose tissue (BAT) [29]. We therefore investigated the expression of the genes such as Cidea, Pgc1a (also known as Ppargc1a), Ucp1 and Cpt1b in adipose tissue. The expression of Pgc1a and Ucp1 was significantly induced in Drp1LiKO BAT (Fig. 6g). The expression of Cidea and Ucp1 was significantly increased in Drp1LiKO WAT, suggesting browning of WAT (Fig. 6h). Furthermore, expression of the target genes for FGF21, such as those encoding dual specificity phosphatase 4 (Dusp4), leptin receptor (Lepr), FBJ osteosarcoma oncogene (cFos) and early growth response 1 (Egr1), have been analysed [30–32]. We found that Dusp4 and Lepr mRNA expression levels were significantly increased in the WAT of Drp1LiKO mice, while cFos and Egr1 expression levels were not increased (Fig. 6i), consistent with the previous observation that exogenous FGF21 failed to increase the expression of immediate early genes such as cFos and Egr1 in obese mice [32].

Fig. 6
figure 6

Increased energy expenditure in Drp1LiKO mice. (a) HE staining of WAT from 20-week-old control (Ctrl) and Drp1LiKO mice on HFD. Scale bar, 50 μm. Adipocyte size was measured and presented relative to control set at 1 (bar graph). (b, c) The expression of genes involved in lipolysis, Lipe (b) and Atgl (c), in WAT. (d) The expression of adiponectin in WAT. (e, f) Oxygen consumption (e) and RER (f) in 16-week-old mice on HFD. (g, h) The thermogenic gene expression in BAT (g) and WAT (h) (n = 5). (i) The expression of the target genes for FGF21 in WAT (n = 5). For mRNA expression, values were normalised to Gapdh and are presented relative to control set at 1. Black bars, control mice; grey bars, Drp1LiKO mice. Data are expressed as means ± SE. *p < 0.05, calculated by Student’s t test

Full size image

Discussion

In this study, we showed that lack of DRP1 in mouse hepatocytes caused ER stress, which was dramatically exacerbated by an HFD. In addition, we found that the PERK–eIF2α–ATF4 pathway was dramatically activated in the liver of Drp1LiKO mice (Fig. 5a–e). Loss of DRP1 in the liver had no effect on the inositol-requiring 1α (IRE1α) and ATF6 pathways (data not shown). Although ATF6, PERK and IRE1α share functionally similar luminal sensing domains and are activated in cells treated with ER stress inducers in vitro, they are selectively activated and regulate diverse downstream target genes in vivo [33]. Our results suggest that mitochondrial fission may act upstream to the cascade of the ER stress response and that communication between mitochondria and the ER is indispensable to the maintenance of cellular homeostasis. In this study, we provided direct evidence that loss of DRP1 alters ER architecture. Mitochondrial dynamics, which are changed in response to metabolic burden, may play an important role in maintaining ER morphology and function.

We observed that a defect in mitochondrial fission and subsequent ER stress promoted the expression of FGF21 in the liver and that the increased FGF21 in turn functions as a metabolic regulator, exhibiting anti-obesity and anti-diabetes effects [22–24]. A recent report by Kim et al demonstrated that autophagy deficiency in muscle leads to protection against obesity and insulin resistance by inducing FGF21 [34]. Interestingly, autophagy deficiency also results in morphologically abnormal, swollen mitochondria, reminiscent of those found in the liver of Drp1LiKO mice on an HFD. These results led us to hypothesise that defects in the inter-organelle communication between mitochondria and ER might depend on FGF21 signalling as a self-defence system to maintain metabolic homeostasis. Recently, Sebastian et al reported that liver-specific Mfn2-knockout mice showed glucose intolerance with hepatic insulin resistance [12]. Loss of MFN2 causes impaired mitochondrial respiration, indicating that MFN2 has multiple roles beyond regulating mitochondrial morphogenesis. There may be different roles for mitochondrial fusion and fission in organelle communication inducing Fgf21. FGF21 increases expression of thermogenic genes in adipose tissue and utilisation of lipids. In addition, FGF21 improves insulin sensitivity by increasing glucose uptake in adipocytes [22] and FGF21-deficient mice are insulin resistant and hyperinsulinaemic [35]. It is thus conceivable that the favourable whole-body insulin sensitivity in Drp1LiKO mice can be explained, at least in part, by increased FGF21, although they have hepatic insulin resistance.

Energy homeostasis is achieved by multiple mechanisms, including inter-organ communication. As well as humoral factors, recent studies have demonstrated that neuronal pathways play an important role in energy homeostasis as a feedback mechanism, preventing excessive energy intake and enhancing energy expenditure [36]. The interesting possibility that the signal from the liver of Drp1LiKO mice uses neuronal pathways in addition to FGF21 cannot be excluded. Nevertheless, our results suggests that mitochondrial dynamics are relevant to understanding the pathogenesis of human metabolic diseases caused by perturbations resulting from overnutrition.

In conclusion, we provide new insight into the role of mitochondrial fission in energy homeostasis. DRP1 may be a potential target for intervention in metabolic diseases such as diabetes and obesity.