White adipocyte-specific Rala deletion protects mice from high fat diet-induced obesity
RNA-seq analysis from isolated mature adipocytes derived from control and HFD-fed mice 28 revealed that Rala expression is significantly upregulated in adipocytes from eWAT and iWAT during obesity development, while Ralgapa2 expression is downregulated (Fig. 1a,b). In addition, RalA protein content is increased in mature adipocytes from iWAT of obese mice (Fig. 1c, Extended Data Fig. 1a), accompanied by elevation of RalA-GTP binding (Fig. 1d, Extended Data Fig. 1b). We also observed a trend towards a positive correlation of the expression of the Ral GEF RGL2 in adipose tissue with BMI in a large dataset of obese patients (Extended Data Fig. 1c,d). Together, these observations support the notion that adipocyte RalA activity is constitutively elevated in obesity.
To explore further whether RalA plays a role in glucose homeostasis and energy metabolism, we generated adipocyte-specific Rala knockout (RalaAKO) mice by crossing Rala-floxed mice with adiponectin-Cre transgenic mice. Compared to Ralaf/f littermates, RalaAKO mice had a greater than 90% decrease of RalA protein in primary adipocytes from WAT and BAT, and an approximately 50% decrease in whole WAT, without changes in liver (Extended Data Fig. 1e). Insulin-stimulated GTP binding of RalA was diminished in WAT of RalaAKO mice compared to control mice, and the same result was observed in primary adipocytes (Extended Data Fig. 1f).
We generated primary white adipocytes by differentiation of iWAT stromal vascular cells from control and KO mice. As previously seen in 3T3-L1 adipocytes 22, knockout of RalA completely prevented the translocation of GLUT4 from intracellular sites to the plasma membrane in response to insulin (Extended Data Fig. 1g). Moreover, insulin-stimulated glucose uptake in KO cells was significantly reduced in knockout cells (Extended Data Fig. 1h).
Adipocyte specific deletion of Rala had no effect on body weight in chow diet (CD)-fed mice, although these mice displayed a reduction in fat mass and depot weight (Extended Data Fig. 2a-c). Generally, adipocytes from iWAT were considerably smaller than those found in epididymal WAT (eWAT) from mice fed CD. Moreover, RalaAKO mice had smaller adipocytes in iWAT compared to control mice fed with CD, while adipocytes were comparable in eWAT and BAT between the genotypes (Extended Data Fig. 2d). While RalaAKO mice on chow diet showed no difference in glucose tolerance, there was a slight reduction in insulin tolerance when compared to Ralaf/f mice (Extended Data Fig. 2e,f). Insulin levels and HOMA-IR in RalaAKO mice were not different from control mice fed with CD (Extended Data Fig. 2g,h). However, RalaAKO mice gained significantly less weight than control littermates when challenged with 60% HFD (Fig. 1e). RalaAKO mice showed a marked reduction of fat mass, with no change in lean body mass (Fig. 1f). Further analyses revealed that iWAT weight was significantly reduced in RalaAKO mice, with no difference in eWAT and BAT (Fig. 1g). HFD increased adipocyte size in all fat depots, but the effect was most pronounced in iWAT; HFD-fed RalaAKO mice displayed a trend towards smaller adipocytes in iWAT compared to control mice, but not in eWAT or BAT (Extended Data Fig. 2d). HFD-fed RalaAKO mice exhibited a marked improvement in glucose tolerance compared to control mice, with no change in insulin tolerance (Fig. 1h,i), but with reduced insulin levels and improved HOMA-IR (Fig. 1j,k). Fasting glucose levels were comparable between the genotypes on either HFD or CD (Extended Data Fig. 2i,j).
To investigate further which adipose tissue depot is responsible for the reduced weight gain in RalaAKO mice fed HFD, we generated BAT-specific Rala knockout (RalaBKO) mice by crossing RalA-floxed mice with UCP1-Cre transgenic mice (Extended Data Fig. 2k). Although CD-fed RalaBKO mice showed a reduction in BAT weight, presumably due to reduced glucose uptake, there were no differences in fat mass or depot weight compared with control mice (Extended Data Fig. 2l,m). Glucose and insulin tolerance tests (GTT and ITT) were identical between the genotypes on control diet (Extended Data Fig. 2n,o). Moreover, no differences in body weight, fat mass, tissue weight, GTT, or ITT were observed in HFD-fed RalaBKO mice (Extended Data Fig. 2p-t). These results suggest that specific Rala deletion in WAT, especially in iWAT, protects mice against obesity.
Loss of RalA in WAT ameliorates HFD-induced hepatic steatosis
Since HFD-fed RalaAKO mice showed an improved GTT without markedly altering insulin tolerance, we speculated that the improved glucose handling is due to reduced hepatic glucose production. To test this assumption, we performed a pyruvate tolerance test (PTT) in HFD-fed Ralaf/f and RalaAKO mice. RalaAKO mice exhibited substantially lower glucose excursions following pyruvate challenge compared to control mice (Fig. 2a). There was a significant downregulation of the hepatic gluconeogenic genes G6pc and Pepck (Fig. 2b). These data suggest that adipose tissue-specific Rala deletion improved glucose homeostasis partially through reduced hepatic glucose production.
Liver weights and triglyceride (TG) content were significantly reduced in HFD-fed RalaAKO mice when compared to control mice (Fig. 2c,d). Both H&E and Oil-Red-O staining indicated less lipid accumulation in the liver of RalaAKO mice (Fig. 2e). In line with histology results, lipogenic genes (Acc, Fasn, Scd1 and Acsl1) were expressed at significantly lower levels in the liver of RalaAKO mice (Fig. 2f). However, plasma leptin levels (Fig. 2g) and hepatic expression of genes related to fatty acid oxidation (FAO) (Fig. 2h) were unchanged in RalaAKO mice. In addition, inflammatory (Adgre1) and fibrosis-related (Col1a1 and Col3a1) genes were expressed at lower levels in livers of RalaAKO mice (Fig. 2i), as were aspartate aminotransferase (AST) and aminotransferase (ALT) activities (Fig. 2j,k). Of note, we did not observe a difference in liver weights in RalaBKO compared to controls fed with HFD (Extended Data Fig. 2r). Together, these observations suggest that WAT-specific deletion of Rala systemically regulates lipid metabolism to ameliorate liver steatosis and damage in obesity.
RalA deficiency in WAT increases energy expenditure and mitochondrial oxidative phosphorylation
To explore why adipose tissue Rala deletion protects mice from HFD-induced hepatic steatosis, weight gain, and glucose intolerance, we investigated energy metabolism in RalaAKO mice with metabolic cage studies. While Rala ablation in adipocytes did not affect energy metabolism and food intake in mice fed CD (Extended Data Fig. 3a-e), HFD-fed RalaAKO mice displayed a significant increase in energy expenditure (EE) during the dark phase as determined by ANCOVA using body weight as a covariate (Fig. 3a). Concordantly, oxygen consumption in RalaAKO mice was similarly increased compared to controls (Extended Data Fig. 3f), although there was no difference in respiratory exchange rate (RER), locomotor activity, or food intake between the genotypes (Extended Data Fig. 3g-i). In contrast, RalaBKO mice fed either control or HFD were identical to control littermates in EE, O2 consumption, RER, locomotor activity, and food intake (Extended Data Fig. 3j-n). These observations demonstrate that Rala deficiency specifically in WAT increases energy expenditure.
Increased energy expenditure is an indirect reflection of increased mitochondrial oxidative activity. Thus, we assessed the expression of mitochondrial proteins in fat depots. Oxidative phosphorylation (OXPHOS) proteins were markedly increased in iWAT of RalaAKO mice (Fig. 3b,c), but not in eWAT (Extended Data Fig. 3o,p). Complex I and Complex II levels were modestly increased in BAT of RalaAKO mice (Extended Data Fig. 3q,r). This may occur because of systemic metabolic improvement in RalaAKO mice rather than a cell-autonomous BAT function, since HFD-fed RalaBKO mice did not show an improved metabolic phenotype. In this regard, plasma FFA and TG levels in HFD-fed RalaAKO mice were significantly lower (Fig. 3d,e). To test the possible involvement of a generalized browning of iWAT, we also examined thermogenic markers. Ucp1, Cidea, and Prdm16 expression was identical between the genotypes in all three fat depots, indicating that the improvement in energy expenditure in RalaAKO mice did not reflect the development of beige adipose tissue (Extended Data Fig. 3s).
RalA knockout in white adipocytes increases mitochondrial activity and fatty acid oxidation
We sought to evaluate further the mechanisms underlying improved energy metabolism in RalaAKO mice, and directly assessed mitochondrial activity in adipocytes. Measurements of basal respiration revealed that oxygen consumption rate (OCR) was significantly increased in mitochondria isolated from KO iWAT compared to that from control mice, but was similar in eWAT mitochondria of Ralaf/f and RalaAKO mice (Fig. 3f). We also noted that both basal and maximal respiration were significantly higher in primary differentiated adipocytes from KO mice, and the difference in maximal respiration was blunted by the addition of the CPT1 inhibitor etomoxir that blocks fatty acid oxidation (FAO) (Fig. 4a, Extended Data Fig. 4a). To investigate directly whether RalA plays a role in controlling FAO, we incubated cells with (14C)-labeled palmitic acid (PA) and measured its oxidation to either acid-soluble metabolites (ASM) or CO2 in WT and KO white adipocytes. In agreement with the OCR results, fatty acid oxidation was significantly higher in KO compared to WT adipocytes (Fig. 4b). These data indicate that RalA knockout in WAT increases energy expenditure due to increased mitochondrial oxidation activity.
To ensure that these studies reflected the activity of RalA, we also generated an immortalized preadipocyte line from Ralaf/f mice and induced Rala deletion by transducing cells with Cre lentivirus. The Cre recombinase completely ablated RalA in preadipocytes and fully differentiated adipocytes (Extended Data Fig. 4b). BODIPY staining demonstrated that both primary and immortalized preadipocytes from WT and KO mice were fully differentiated. As an orthogonal approach, we performed live cell imaging using the cell permeant fluorescent dye, TMRM, to detect mitochondrial membrane potential (MtMP), which reflects electron transport and oxidative phosphorylation in active mitochondria. KO adipocytes exhibited a higher TMRM signal intensity than did their WT counterparts (Fig. 4c, Extended Data Fig. 4c). To specify the ability of TMRM to detect mitochondrial depolarization in active mitochondria, we applied the β3-adrenergic receptor agonist CL316,243 (CL) to induce mitochondrial membrane depolarization 29. The TMRM signal declined quickly after administration of the agonist, which confirms that TMRM stains only active mitochondria (Extended Data Fig. 4d).
We previously reported that lipolysis drives mitochondrial oxidative metabolism in adipocytes 30. To rule out a possible role for lipolysis as the primary driver of increased oxidative capacity of Rala-KO adipocytes, we performed in vitro and in vivo lipolysis assays. CL robustly stimulated FFA and glycerol release to the same extent in KO and WT immortalized adipocytes, and the molar ratio of FFA to glycerol was approximately 3:1 (Extended Data Fig. 4e,f). Additionally, there was no difference in CL-induced FFA and free glycerol production in Ralaf/f and RalaAKO mice (Extended Data Fig. 4g,h). We further tested whether RalaAKO mice are defective in the suppression of FFA release by insulin. Insulin suppressed CL-induced FFA release by approximately 50% in both WT and KO cells (Extended Data Fig. 4e). A single injection of insulin reduced FFA levels in control and RalaAKO mice to the same extent (Extended Data Fig. 4i). Interestingly, KO adipocytes displayed a mild increase in glycerol release in the presence of CL, while RalaAKO mice showed a mild decrease of plasma glycerol levels either in the presence of CL or after fasting (Extended Data Fig. 4f,h,j). Taken together, these results suggest that the absence of RalA in adipocytes enhances mitochondrial oxidative activity without affecting FFA supply.
Targeted Rala knockout protects against obesity-induced mitochondrial fission in iWAT
The increased mitochondrial oxidative activity observed in HFD-fed RalaAKO mice could result from increased mitochondrial biogenesis. Expression of genes related to mitochondrial biogenesis was comparable between the genotypes (Extended Data Fig. 5a,b) in WAT. The activity of AMPK, the master regulator of mitochondrial biogenesis 31,32, was also comparable between control and RalaAKO mice fed with HFD (Extended Data Fig. 5c-f). In addition to biogenesis, mitochondrial function can also be regulated by dynamic changes in morphology through tightly controlled fusion and fission events that shape the organelle to comply with energy demands 19,33. Electron microscopy (EM) revealed that HFD feeding of WT mice induced the appearance of smaller, spherical mitochondria in iWAT (Fig. 4d), consistent with previous reports that mitochondrial function and morphology is impaired in obese adipocytes 34,35. In agreement with the in vivo metabolic phenotypes, adipocyte Rala deletion did not grossly affect mitochondrial morphology in iWAT of CD-fed mice (Fig. 4d), but the HFD-induced change in mitochondrial morphology was completely prevented in Rala KO iWAT; mitochondria in iWAT from these mice displayed an elongated shape that was indistinguishable from CD-fed mice (Fig. 4d). Indeed, tissue weight (Fig. 1f), OXPHOS content (Extended Data Fig. 3o,p), and mitochondrial OCR (Fig. 3f) were not affected by RalA deletion in eWAT, corresponding to the observation that the appearance of fragmented mitochondria in this depot was not reversed by RalA KO in HFD mice (Extended Data Fig. 5g). In fact, mitochondria in eWAT do not undergo significant fragmentation in response to HFD, possibly because of their already fragmented shape, consistent with the overall anabolic function of visceral adipocytes (Fig. 4d, Extended Data Fig. 5g). We also examined mitochondrial morphology in immortalized adipocytes differentiated from iWAT. As shown in Fig. 4e, mitochondria in KO adipocytes appeared longer than those in WT cells. There was a higher frequency of elongated mitochondria (1.0-1.5 µm) in KO cells (Fig. 4f), and the mean maximal mitochondrial length was significantly higher than in WT cells (Fig. 4g).
Inhibition of RalA increases Drp1 S637 phosphorylation in white adipocytes
Opa1 and Drp1 have been identified as key regulators of mitochondrial fusion and fission, respectively 36. Opa1 undergoes proteolytic cleavage to generate long (L-Opa1) and short (S-Opa1) forms that together fuel mitochondrial fusion 37,38 39. Protein levels of both forms of Opa1 were downregulated in iWAT after HFD feeding (Extended Data Fig. 5h-j); only S-Opa1 was downregulated in eWAT from RalaAKO mice (Extended Data Fig. 5k-m), indicating the likelihood of reduced fusion in KO mice compared to WT littermates. However, the observation of elongated mitochondria in KO mice (Fig. 4d) suggests that this change in Opa1 processing is likely compensatory. We then focused on Drp1 as a key regulator of fission. Interestingly, Drp1 phosphorylation at the anti-fission S637 site was significantly increased in Rala-KO iWAT (Fig. 5a, Extended Data Fig. 6a), whereas Drp1 S637 phosphorylation was comparable between the genotypes in eWAT (Extended Data Fig. 6b,c). To establish whether this effect is cell-autonomous, we examined Drp1 phosphorylation in both immortalized and primary adipocytes. Drp1 S637 phosphorylation is catalyzed by PKA, activated by the b-adrenergic/cAMP pathway 40,41. Drp1 S637 phosphorylation was triggered by CL after 5 minutes and was maximal after 15 minutes in adipocytes (Fig. 5b, Extended Data Fig. 6d). Consistent with in vivo results, Rala-KO adipocytes showed a significantly higher Drp1 S637 after β-adrenergic stimulation compared to WT cells (Fig. 5c, Extended Data Fig. 6e). We also explored the effect of RalA on Drp1 S637 phosphorylation state using a specific Ral inhibitor that prevents activation and retains the GTPase in the GDP-bound, inactive state. Pretreatment with the pan-Ral inhibitor RBC8 26,42 significantly increased forskolin-stimulated Drp1 S637 phosphorylation in 3T3-L1 adipocytes (Extended Data Fig. 6f,g). Inhibition of RalA activity with RBC8 also increased forskolin-stimulated Drp1 S637 phosphorylation in the human primary adipocyte cell line (SGBS) (Fig. 5d,e). Thus, RalA specifically modulates Drp1 S637 phosphorylation downstream of PKA activation across multiple adipocyte cell lines of both murine and human origin. To determine whether RalA influences CL-induced PKA activation or cAMP breakdown, we measured cAMP production and phosphorylation of hormone sensitive lipase (HSL) in adipocytes. There was no difference in cAMP production between WT and KO primary adipocytes after 5 minutes of CL stimulation (Extended Data Fig. 6h). Similarly, HSL S660 phosphorylation was identical in WT and KO adipocytes (Extended Data Fig. 6i-l).
To examine the relevance of Drp1 as a regulator of metabolism in human obesity, we analyzed microarray data of abdominal subcutaneous WAT from obese and non-obese women. In human subcutaneous WAT, DNM1L (encoding human Drp1 protein) expression was positively correlated with BMI and HOMA (Fig. 5f,g), and its expression was significantly upregulated in obese subjects (Fig. 5h), indicating that increased expression of DNM1L may contribute to mitochondrial dysfunction in obesity. Moreover, bioinformatic analysis of published microarray data (GEO: GSE7053) from 770 human males further confirmed that DNML1 is associated with obesity (Extended Data Fig. 6m-o). Together, these in vivo and in vitro data suggest that upregulated Drp1 activity in adipose tissue may be an important contributor to mitochondrial dysfunction during obesity and further that RalA deficiency protects mitochondria from excessive fission by increasing Drp1 S637 phosphorylation.
RalA interacts with Drp1 and protein phosphatase 2A, promoting dephosphorylation of Drp1 at S637
To understand the molecular mechanism by which RalA regulates Drp1 S637 phosphorylation, we used proteomics to search for proteins interacting with wildtype (WT), constitutively active (G23V), or dominant negative (S28N) forms of RalA ectopically expressed in liver. Among the binding proteins was protein phosphatase 2A subunit A alpha (PP2Aa), the scaffolding subunit encoded by the Ppp2r1a gene, which preferentially bound to the RalAG23V constitutively active mutant. To confirm these mass spectrometry data, we purified RalAWT-Flag protein from HEK293T cells and pulled down PP2Aa from lysates (Fig. 6a). To determine whether this interaction is dependent on the activation state of the G protein, we co-expressed WT and mutant RalA constructs with PP2Aa in HEK293T cells. As a positive control, the effector Sec5 only bound to active RalAG23V. Similarly, this mutant form of RalA had the highest affinity for PP2Aa (Fig. 6b). We also loaded a RalA-Flag fusion protein in vitro with GTPγS or GDP to evaluate the specificity of effector binding. Both Sec5 and PP2Aa were pulled down by RalA loaded with GTPγS but not with GDP (Fig. 6c). In addition, because PP2Aa and Drp1 did not independently interact (data not shown), we investigated whether RalA directly modifies Drp1 phosphorylation via PP2Aa. When co-expressed, Drp1 and RalA interacted directly with each other, although there was no preference for the activation state of RalA (Extended Data Fig. 7a). Activation of the cAMP/PKA axis by addition of forskolin increased Drp1 S637 phosphorylation, while co-expression of PP2Aa promoted the dephosphorylation of S637 (Fig. 6d), although overexpression of PP2Ab had no effect (Extended Data Fig. 7b). These data suggest that Drp1 is constitutively associated with RalA independent of activation state, and upon activation, RalA recruits PP2Aa to promote the dephosphorylation of Drp1 S637.
To understand further the effects of RalA activation state on Drp1 phosphorylation and mitochondrial function, we transduced immortalized RalA KO cells with RalAWT and RalAG23V lentivirus prior to differentiation into adipocytes. RalAG23V expressing adipocytes showed a robust increase in RalA GTP binding (Fig. 6e), and these cells had significantly less Drp1 S637 phosphorylation (Fig. 6f, Extended Data Fig. 7c). Expression of either RalAWT or RalAG23V significantly reduced mitochondrial potential in KO adipocytes (Fig. 6g, Extended Data Fig. 7d). To confirm that this reduction in mitochondrial potential is associated with reduced oxidative function, we performed a seahorse assay. Consistent with results in primary adipocytes, RalAWT and RalAG23V expressing adipocytes displayed reduced basal and maximal oxygen consumption rate (OCR) in comparison to KO adipocytes (Fig. 6h, Extended Data Fig. 7e). In addition, EM revealed that overexpression of WT or constitutively active RalA in adipocytes resulted in fragmented mitochondria, indicating increased fission compared to RalA KO adipocytes (Fig. 6i).
RalA has previously been reported to promote fission in proliferating cells, and Rala knockdown led to a long, interconnected mitochondrial network and reduced proliferation 43. Partially in agreement with this study, we found that RalA deficiency resulted in elongated mitochondria in adipocytes, with increased oxidative phosphorylation that dramatically impacted whole body lipid metabolism. However, unlike the previous study, we did not observe an interaction between RalBP1 and Drp1. Interestingly, total PP2Aa protein levels were increased in Rala KO iWAT compared to control iWAT, without a difference in PP2Ab and PP2Ac content (Extended Data Fig. 7f,g), perhaps reflecting a compensatory pathway. Taken together, our data suggest that obesity drives RalA expression and GTP binding activity, leading to its association with PP2Aa, which in turn recruits the catalytic subunit PP2Ac to dephosphorylate Drp1 S637. We also note that catecholamine resistance, an inherent trait of the obese state 28, is also expected to lead to reduced PKA-catalyzed S637 phosphorylation. Together, these effects result in constitutive mitochondrial translocation of Drp1 and fragmented mitochondria in adipocytes from obese subjects (Extended Data Fig. 8).