Impaired mitochondrial medium-chain fatty acid oxidation drives periportal macrovesicular steatosis in sirtuin-5 knockout mice

Medium-chain triglycerides (MCT), containing C8–C12 fatty acids, are used to treat several pediatric disorders and are widely consumed as a nutritional supplement. Here, we investigated the role of the sirtuin deacylase Sirt5 in MCT metabolism by feeding Sirt5 knockout mice (Sirt5KO) high-fat diets containing either C8/C10 fatty acids or coconut oil, which is rich in C12, for five weeks. Coconut oil, but not C8/C10 feeding, induced periportal macrovesicular steatosis in Sirt5KO mice. 14C–C12 degradation was significantly reduced in Sirt5KO liver. This decrease was localized to the mitochondrial β-oxidation pathway, as Sirt5KO mice exhibited no change in peroxisomal C12 β-oxidation. Endoplasmic reticulum ω-oxidation, a minor fatty acid degradation pathway known to be stimulated by C12 accumulation, was increased in Sirt5KO liver. Mice lacking another mitochondrial C12 oxidation enzyme, long-chain acyl-CoA dehydrogenase (LCAD), also developed periportal macrovesicular steatosis when fed coconut oil, confirming that defective mitochondrial C12 oxidation is sufficient to induce the steatosis phenotype. Sirt5KO liver exhibited normal LCAD activity but reduced mitochondrial acyl-CoA synthetase activity with C12. These studies reveal a role for Sirt5 in regulating the hepatic response to MCT and may shed light into the pathogenesis of periportal steatosis, a hallmark of human pediatric non-alcoholic fatty liver disease.

www.nature.com/scientificreports/ the liver for disposal. Less than 1% of the ingested dose of MCT will reach peripheral circulation 11 . Consumption of large doses of MCT exposes hepatocytes to high concentrations of free medium-chain fatty acids, which must be either immediately catabolized or else elongated to a chain length that can be stored as triglyceride. This contrasts with long-chain fatty acids released from long-chain triglycerides (LCT), which are re-esterified into triglycerides, packed into chylomicrons, and trafficked from the gut through the lymphatic system and into peripheral circulation. Not all MCT-based nutritional products contain the same composition of fatty acids. Coconut oil is primarily C 12 (50%), but also contains minor amounts of C 14 , C 10 , and C 8 12 . Other MCT products contain either pure C 8 or a mixture of C 8 and C 10 obtained from fractionating coconut oil. C 8 /C 10 may undergo a different route of metabolism than C 12 due to the substrate specificity and intracellular localization of the acyl-CoA synthetase enzymes responsible for conjugating free fatty acids to acyl-CoAs, which are the biologically active form of fatty acids in the cell. C 8 , and to lesser extent C 10 , can diffuse into the mitochondrial matrix where they are activated into acyl-CoAs by the medium-chain acyl-CoA synthetases (ACSMs) 13 . C 12 cannot cross the mitochondrial membrane and is a poor substrate for ACSMs 13 . C 12 is activated to C 12 -CoA by the long-chain acyl-CoA synthetases (ACSLs), which reside on the outer mitochondrial membrane, plasma membrane, ER, and the peroxisomal membrane 14 . In keeping with this, C 12 can be catabolized by both mitochondrial and peroxisomal β-oxidation pathways 15 . C 12 is also metabolized by a minor pathway known as ω-oxidation in which C 12 is hydroxylated by cytochrome P450 enzymes in the endoplasmic reticulum, converted to a dicarboxylic acid in the cytosol, and finally chain-shortened in the peroxisome 16 . While peroxisomes can chain-shorten long-chain fatty acids down to C 6 17 and DCAs as far as C 4 18 , the capacity of the peroxisome to catabolize exogenous C 8 /C 10 is limited by the fact that all known ACSMs reside in the mitochondrial matrix 19 .
In rodent models, both C 8 and C 12 feeding have been linked to hepatic steatosis 15,20 . However, other studies have indicated either no fat accumulation or even a reversal of pre-existing fatty liver, leading to the suggestion that MCT may have utility for treating the metabolic syndrome 21,22 . Understanding the molecular mechanisms involved in hepatic adaptation to MCT is critical for determining the impact of this supplement on human health. Because of Sirt5's dual localization to mitochondria and peroxisomes and its known opposite regulation of FAO enzymes in these compartments, we used Sirt5 knockout mice to interrogate the role of this lysine deacylase on the hepatic response to dietary C 8 /C 10 versus coconut oil. Our results implicate Sirt5 as an important regulator of C 12 disposal in the liver.

Results
Sirt5KO mice develop hepatic macrovesicular steatosis on coconut oil diet. To study the role of Sirt5 in regulating the hepatic response to acute high-fat feeding, we placed Sirt5KO and wild-type control mice on three high-fat diets containing different chain-lengths of fatty acids: (1) a diet containing 60% C 8 and 40% C 10 ; (2) a coconut-oil diet containing primarily C 12 ; and (3) a long-chain triglyceride (LCT) diet containing primarily C 16 and C 18 fatty acids. The relative composition of these diets appears in Supplemental Fig. 1. A low-fat standard diet (SD) served as control. After five weeks on the high-fat diets, liver tissue was collected and analyzed for signs of steatosis. In H&E-stained liver sections, macrovesicular steatosis was noted only in Sirt5KO livers from mice maintained on the coconut oil diet (Fig. 1a,b). This was observed in both genders of Sirt5KO mice. While the C 8 /C 10 diet did not induce overt macrovesicular steatosis, increased microvesicular lipid deposition was visible in liver of both genotypes upon Oil-Red-O staining (Fig. 2). The intensity of the Oil-Red-O staining was even greater with coconut oil feeding. Little or no Oil-Red-O staining was visible in livers from mice fed LCT for five weeks. Together, the H&E and Oil-Red-O stains indicated that fat accumulated in the order of coconut > > C 8 / C 10 > > LCT for both genotypes of mice, with the lipid accumulation progressing to the point of macrovesicular steatosis only in Sirt5KO mice on coconut oil diet. This was supported by a biochemical analysis of liver triglyceride content. Both C 8 /C 10 and coconut oil feeding significantly increased the triglyceride content of liver tissue over that seen in liver from mice maintained on the SD, while LCT had no effect (Fig. 1c). Sirt5 deficiency was associated with significantly increased liver triglycerides for both the C 8 /C 10 and coconut oil diets (Fig. 1c). Again, both genders were evaluated, and no gender effect was seen for triglyceride accumulation.
The macrovesicular steatosis in coconut oil-fed Sirt5KO mice is periportal. In adult humans with non-alcoholic fatty liver disease (NAFLD) and NAFLD rodent models induced by chronic LCT feeding, macrovesicular steatosis first appears around the pericentral regions of the liver. In contrast, pediatric NAFLD is characterized by periportal steatosis 23 . Interestingly, while five weeks of coconut oil feeding induced pan-zonal microvesicular steatosis in both wild-type and Sirt5KO livers as visualized by Oil-Red-O staining (Fig. 2), the macrovesicular steatosis in Sirt5KO livers was strictly periportal (Fig. 3a,b). To confirm the periportal localization of the lipid droplets we immunostained the liver tissue for glutamine synthase (GS), a marker of the pericentral zone. No lipid droplets were observed near the GS-positive pericentral zones (Fig. 3c-f). Besides marking the pericentral zones, the GS immunostaining highlighted an altered morphology of the pericentral hepatocytes in Sirt5KO liver that was consistent with ballooning degeneration. Sirt5 has previously been reported to exhibit hepatic zonation, with higher expression in primary mouse hepatocytes isolated from the periportal zone 24 . We confirmed this using beta-galactosidase staining as a proxy for expression of the Sirt5 mutant allele in Sirt5KO mice, which contains a lacZ insert. Pericentral hepatocytes exhibited less staining for the Sirt5 promoter-driven beta-galactosidase enzyme (Fig. 3g).
Sirt5KO liver has decreased capacity for C 12 fatty acid β-oxidation (FAO). We next sought to investigate the mechanism(s) behind the observed acceleration in lipid storage in Sirt5KO livers during coconut oil feeding. Sirt5KO mice have previously been shown to be deficient in long-chain FAO but have not been Liver triglyceride content of mice maintained on the four diets for five weeks. N = 6 for C 8 /C 10 and LCT diets, and N = 10 for coconut and SD diets. There was no gender difference in triglyceride accumulation; all groups contained half males and half females, age 6-8 weeks at time of diet onset. *P < 0.01 Sirt5KO versus wild-type controls. www.nature.com/scientificreports/ www.nature.com/scientificreports/ evaluated for changes in medium-chain FAO 3 . To interrogate medium-chain FAO we used 14 C-labeled C 8 and C 12 and quantified flux in liver homogenates as well as mouse embryonic fibroblasts (MEFs). Sirt5KO MEFs, cultured in standard medium, showed a significant deficit in 14 C-C 12 catabolism (Fig. 4a). In contrast, C 8 oxidation rates were higher in Sirt5KO MEFs (Fig. 4c). The same pattern was observed in liver homogenates prepared from mice fed coconut oil diet for five weeks (Fig. 4b,d). The higher C 8 oxidation may be due to higher expression of the C 8 -specific enzyme medium-chain acyl-CoA dehydrogenase (MCAD) in Sirt5KO livers after coconut oil feeding (Fig. 4e,f).

Scientific Reports
Peroxisomal FAO is not altered in coconut-fed Sirt5KO mice. C 12 , the principal fatty acid in coconut oil, is the optimal substrate for the rate-limiting peroxisomal FAO enzyme acyl-CoA oxidase-1 (ACOX1) 25 .
In wild-type mouse liver homogenates from animals on standard diet, nearly 40% of the capacity for 14 C-C 12 oxidation was resistant to potassium cyanide (KCN) and therefore peroxisomal (Fig. 5a). In contrast, for the long-chain fatty acid (C 16 ), peroxisomes contributed only about 10% (Fig. 5b). Because of the strict mitochondrial localization of the acyl-CoA synthetases responsible for activation of C 8 to C 8 -CoA, C 8 metabolism is believed to be mitochondria-specific 19 . We therefore reasoned that the low C 12 oxidation in Sirt5KO liver, but normal/high C 8 oxidation (Fig. 4), might be due to a specific loss of peroxisomal FAO capacity. Western blotting indicated a significant increase in ACOX1 protein expression in both genotypes of mice when fed coconut oil diet for five weeks (Fig. 5c,d). Immunostaining for the peroxisomal marker protein PMP70 showed that peroxisomes are more enriched in the pericentral zones when mice are maintained on a standard diet but demonstrate a pan-zonal proliferation after five weeks of coconut oil diet (Fig. 5e). This occurred in both wild-type and Sirt5KO mice, but overall, the PMP70 staining was darker in Sirt5KO liver. However, the rate of KCN-resistant (peroxisomal) 14 C-C 12 FAO in liver homogenates of coconut oil-fed mice was not significantly different between genotypes (Fig. 5f). This indicates that the reduced capacity for C 12 oxidation seen in Sirt5KO liver is not attributable to suppressed peroxisomal FAO.
Fatty acid ω-oxidation is upregulated in Sirt5KO livers. In addition to mitochondria and peroxisomes, C 12 is also the optimal chain length for a minor FAO pathway known as ω-oxidation 16 . ω-oxidation initiates with C 12 hydroxylation by Cyp4a family members in the ER. Through a poorly characterized pathway, C 12 OH is converted to a dicarboxylic C 12 fatty acid (DC 12 ) and finally catabolized by peroxisomes, with only minimal contribution by mitochondria 26 . Immunoblotting revealed significantly higher protein expression of enoyl-CoA hydratase/3-hydroxy-CoA dehydrogenase (EHHADH), a peroxisomal enzyme known to be required for DC 12 catabolism 27 , in Sirt5KO liver following five weeks on coconut oil diet (Fig. 6a,b). DC 12 is chain-shortened to adipic acid which is excreted into the urine. Sirt5KO mice on coconut-oil diet excreted www.nature.com/scientificreports/ four-fold more adipic acid than wild-type (Fig. 6c). Finally, higher 14 C-DC 12 oxidation was observed in Sirt5KO MEFs (Fig. 6d). An attempt was made to measure 14 C-DC 12 oxidation in liver, but the substrate was found to be unsuitable for assaying broken cells or tissue homogenates. Together these data suggest increased utilization of the ω-oxidation pathway in the absence of Sirt5.
Loss of mitochondrial C 12 FAO causes periportal macrovesicular steatosis on coconut oil-diet. The observations of normal peroxisomal C 12 oxidation and elevated ω-oxidation in Sirt5KO mice suggested that the defect in C 12 metabolism must be mitochondrial. The first step in mitochondrial FAO is catalyzed by the acyl-CoA dehydrogenase (ACAD) enzyme family. There are four ACAD enzymes with partially overlapping substrate specificities 28 . To determine which ACADs contribute to C 12 -CoA metabolism, we compared the specific activities of all four recombinant enzymes with C 12 -CoA. Long-chain acyl-CoA dehydrogenase (LCAD) exhibited the highest specific activity with C 12 -CoA, followed by MCAD (Fig. 7a). Very long-chain acyl-CoA dehydrogenase (VLCAD) had low activity with C 12 -CoA and ACAD9 activity was negligible.
To test whether loss of mitochondrial C 12 FAO capacity is sufficient to recapitulate the macrovesicular steatosis phenotype, we maintained MCADKO and LCADKO mice on the various diets for five weeks. On standard diet, neither knockout strain exhibited visible macrovesicular steatosis upon H& E staining of liver tissue (data not shown). Surprisingly, while MCAD is the predominant ACAD enzyme metabolizing C 8 -CoA and C 10 -CoA, maintaining MCADKO mice on the high-fat C 8 /C 10 diet did not result in increased triglyceride storage (Fig. 7b) or any signs of macrovesicular steatosis following histological staining (data not shown). With the coconut oil diet, both MCADKO and LCADKO mice showed significantly increased liver triglycerides (Fig. 7c), but only LCADKO livers exhibited the periportal macrovesicular steatosis phenotype (Fig. 7d). Immunostaining for  www.nature.com/scientificreports/ PMP70 in liver from LCADKO fed coconut oil also revealed a pan-zonal distribution of peroxisomes similar to that observed in Sirt5KO liver (Fig. 7e).
Sirt5KO liver shows reduced activation of C 12 to C 12 -CoA. The experiments described above point to a mitochondrial FAO defect in Sirt5KO livers that causes reduced C 12 oxidation but not C 8 oxidation. C 12 is the transition point between medium-chain and long-chain fatty acids; it is utilized primarily by the long-chain FAO machinery while C 8 is the optimal chain length for the medium-chain FAO machinery. We therefore hypothesized that reduced function of either LCAD or trifunctional protein (TFP), which together catalyze the four reactions in the long-chain FAO cycle, may explain the C 12 -specific phenotype. Both LCAD and TFP are heavily acylated enzymes with multiple Sirt5-targeted lysine residues 3 , but their activity has not heretofore been measured in Sirt5KO liver. First, the ETF fluorescence reduction assay was used to measure acyl-CoA dehydrogenase activity in liver lysates from coconut-oil fed wild-type and Sirt5KO mice. Three substrates were tested-C 8 -CoA, which is specific for MCAD; C 12 -CoA, which is primarily metabolized by LCAD with overlapping activity from MCAD and very long-chain acyl-CoA dehydrogenase (VLCAD); and 2,6-C 7 -CoA, a branched-chain substrate specific for LCAD 29 . There was no statistically significant change in acyl-CoA dehydrogenase activity in Sirt5KO livers with any of the three substrates (Fig. 8a). Next, we measured activity of TFP with both a medium-chain substrate (3-keto-C 10 -CoA) and a long-chain substrate (3-keto-C 16 -CoA). As with LCAD, no significant change in TFP activity was observed (Fig. 8b). Another difference in substrate handling between C 12 and C 8 is at the acyl-CoA synthetase step, which activates fatty acids to acyl-CoAs prior to FAO. C 8 more readily diffuses through membranes than C 12 and is activated to C 8 -CoA inside the mitochondrial matrix by the ACSMs. C 12 is a poor substrate for the ACSMs 13 . This suggests that C 12 is activated by members of the ACSL family. There are no ACSLs in the matrix; rather, they are localized to the outer mitochondrial membrane facing the cytosol as well as on the ER membrane 14 . To determine whether loss of Sirt5 alters activity of the acyl-CoA synthetases, we isolated liver mitochondria and measured conversion rates of 14 C-C 8 and 14 C-C 12 to their respective acyl-CoAs. The rate of mitochondrial 14 C-C 8 activation was not different between coconut-fed wild-type and Sirt5KO livers while the rate of mitochondrial 14 C-C 12 activation was significantly lower (Fig. 8c).

Discussion
In the nutritional supplement industry, the term MCT is used a catch-all for coconut oil, pure C 8, and mixed C 8 /C 10 products. Our data indicate that these products may be metabolized very differently. C 8 /C 10 feeding resulted in periportal microvesicular steatosis. In mice consuming the C 12 -rich coconut oil diet this effect was greatly exacerbated, with steatosis spread throughout the liver. Loss of either Sirt5 or the mitochondrial FAO enzyme LCAD pushed the phenotype into macrovesicular steatosis in the periportal zone. Loss of MCAD, a mitochondrial FAO enzyme with high specificity for C 8 -CoA/C 10 -CoA, did not increase triglyceride storage over wildtype levels or cause macrovesicular steatosis on the high C 8 /C 10 diet. Thus, stressing mice partially deficient in mitochondrial C 12 FAO (Sirt5KO, LCADKO) with a diet rich in C 12 is a combination sufficient to produce periportal macrovesicular steatosis, while stressing mice deficient in C 8 /C 10 FAO (MCADKO) with a diet rich in C 8 /C 10 does not produce this phenotype. We postulate that this is because C 12 is metabolized differently than fatty acids that are just 2-4 carbons shorter.
Medium-chain fatty acids are often stated to enter mitochondria by diffusion and rapidly undergo β-oxidation 7 . C 8 , the prototypical medium-chain fatty acid, is well known to enter mitochondria independently of the carnitine transport system 30,31 . Etomoxir, an irreversible inhibitor of the key transport enzyme carnitine www.nature.com/scientificreports/ palmitoyltransferase-1 (Cpt1), is completely without effect on liver C 8 oxidation 32 . C 12 has been poorly studied in comparison to C 8 . In our experiments, etomoxir blocked C 12 oxidation by 33% but C 16 oxidation by 86% (Fig. 5). This indicates that C 12 partly behaves like a long-chain fatty acid in regard to mitochondrial membrane transport. Perhaps of greater significance is the difference between C 8 and C 12 in terms of the mechanism of activation to CoA. We observed normal C 8 -CoA synthetase activity in Sirt5KO liver but significantly reduced C 12 -CoA synthetase activity (Fig. 8). This data suggests that the two chain lengths are not activated to CoA by the same enzyme. Liver abundantly expresses two isoforms of medium-chain acyl-CoA synthetase known as ACSM1 and ACSM2. ACSM1 and ACSM2 are localized to the mitochondrial matrix where they activate free medium-chain fatty acids into medium-chain acyl-CoAs for β-oxidation. Vessey et al 13 purified ACSM1 and ACSM2 from human liver mitochondria and characterized reactivity against a range of fatty acid substrates. Interestingly, ACSM1 is 40-fold less active with C 12 than with C 8 , while ACSM2 is 90-fold less active with C 12 . Thus, if these enzymes are tasked with activating C 12 that has diffused into the mitochondria, the reaction rate would be predicted to be slow. In contrast, the long-chain acyl-CoA synthetase (ACSL) enzyme family, of which there are four isoforms expressed in liver, exhibit high reactivity with C 12 and low reactivity with C 8 33,34 . ACSLs are not present in the matrix, but rather are inserted into the outer mitochondrial membrane, ER membrane, and plasma membrane, facing into the cytosol 14 . We previously found multiple Sirt5-targeted lysine residues on ACSL1 and ACSL5 35 , but no Sirt5 target sites have been identified on ACSL3 or ACSL4, which are also expressed in liver. Each ACSL activates fatty acids to acyl-CoAs for a different metabolic purpose. For example, ACSL1 is thought to preferentially channel acyl-CoAs into the mitochondria, while ACSL3 has been implicated in lipogenesis 14 . We postulate that decreased activity of ACSL1, but normal activity of ACSL3, could cause partitioning of fatty acids like C 12 into lipid droplets rather than into mitochondria in Sirt5KO liver.
Another key difference between exogenous C 8 and C 12 fatty acids is the ability of the latter to be metabolized partially by peroxisomal β-oxidation. Peroxisomes were previously shown to be more abundant in the pericentral zone, which our PMP70 immunostaining confirms 36 . Interestingly, coconut oil feeding appeared to recruit peroxisomes into the periportal zone, such that the anti-PMP70 staining became azonal. PMP70 staining appeared stronger in coconut-fed Sirt5KO liver compared to wild-type (Fig. 5e) yet immunoblotting for the peroxisomal enzyme ACOX1 revealed a trend for lower ACOX1 in Sirt5KO liver (Fig. 5c,d). This apparent discrepancy between peroxisomal abundance (PMP70 staining) and peroxisomal enzyme content (ACOX1) has been reported previously in animals treated with peroxisome-proliferator activated receptor-α (PPARα) agonist drugs and may reflect newly formed peroxisomes that have membrane marker proteins like PMP70 but are largely devoid of peroxisomal matrix proteins 37 . Future work will address whether such a phenomenon may be occurring in Sirt5KO liver with coconut oil feeding. It must also be tested whether C12 can alter the distribution of peroxisomes in human hepatocytes. Peroxisomal biogenesis and the peroxisomal FAO pathway are well known to be more inducible in rodents than in humans, possibly due to the much higher abundance of PPARα in rodent liver 38,39 .
While peroxisomes have the capacity to metabolize 14 C-C 12 in liver homogenates when mitochondria are inhibited with KCN, it is not currently possible to determine the relative disposition of C 12 through peroxisomes versus mitochondria in vivo. What is clear is that increased peroxisome abundance does not prevent lipid accumulation or macrovesicular steatosis in either Sirt5KO or LCADKO livers. Peroxisomes are known to interact directly with lipid droplets 40 . Elimination of some peroxisomal membrane "Pex" proteins results in reduced lipid droplet size, thought to be due to shared biogenesis mechanisms between peroxisomes and lipid droplets 41 . Another interesting possibility requiring future investigation is that periportal peroxisomes in Sirt5KO and LCADKO livers are partially chain-shortening excess C 12 to acetyl-CoA, which is being released to the cytosol and used for fatty acid synthesis. Intriguingly, the peroxisomal membrane protein PMP70 physically interacts with fatty acid synthase 42 . Further, experiments with 13 C-fatty acid tracers showed significant labeling of cytosolic malonyl-CoA, consistent with peroxisomally-produced acetyl-CoA being converted to malonyl-CoA in the cytosol 43,44 . This would not only provide building blocks for fatty acid synthesis, but also further inhibit mitochondrial FAO at the level of Cpt1, leading to a vicious cycle that promotes steatosis.
In summary, this work shows that MCT, and particularly C 12 -rich coconut oil, is associated with a stronglyzoned pattern of lipid accumulation exacerbated by impaired mitochondrial FAO. People consuming MCT as a dietary supplement are not likely to consume the quantity of MCT that the animals did in our experiments. However, preterm infants, patients with epilepsy and FAO disorders, and other patient groups for whom longchain fatty acids are contraindicated may consume 40% or more of their calories from MCT. It remains to be seen whether long-term consumption of MCT in these patients leads to NAFLD or other hepatic complications. Finally, it is of note that the periportal macrovesicular steatosis seen in Sirt5KO and LCADKO mice consuming coconut oil resembles pediatric NAFLD 23 . For reasons that are not understood, adult NAFLD begins in the pericentral zones and slowly spreads towards the periportal zones as the disease progresses, while pediatric NAFLD shows the opposite pattern. It is tempting to speculate that impaired FAO may contribute to the development of pediatric NAFLD. The biological roles of both Sirt5 and LCAD in humans are poorly understood. LCAD exhibits a much more restricted expression pattern in human than in rodents 45,46 . LCAD is, however, expressed and active in human liver 47 . Our results here suggest it could play a role in the hepatic response to dietary medium-chain fatty acids. Further, common polymorphisms exist in both the Sirt5 and LCAD genes that may affect expression/activity of these enzymes 48,49 . In any case, the consumption of coconut oil, which is increasingly being incorporated into processed foods, may be contraindicated for children with other known NAFLD risk factors. www.nature.com/scientificreports/ Experimental procedures Animals experimentation. All animal protocols were approved by the University of Pittsburgh Institutional Animal Care and Use Committee (IACUC), and all experiments were conducted in accordance with the guidelines and regulations set forth in the Animal Welfare Act (AWA) and PHS Policy on Humane Care and Use of Laboratory Animals. SIRT5−/− and wildtype control mice were purchased from Jackson Laboratories (Bar Harbor, ME) and bred in-house. LCAD−/− mice and medium-chain acyl-CoA dehydrogenase (MCAD) −/− mice were obtained from the Mutant Mouse Regional Resource Center. All strains are on a mixed background of C57Bl/6 and 129. High-fat diets based on coconut oil (D12331), C 8 /C 10 (D17011004; 60% C 8 and 40% C 10 ), or lard (D12451) were purchased from Research Diets, Inc. Mice were placed onto the high-fat diets for five weeks beginning at age 6-8 weeks. Both genders were used; the gender for a given experiment is specified in "Results". All tissues were collected in non-fasted animals between 9:00 and 11:00 a.m. Euthanasia was conducted using inhaled CO 2 gas according to IACUC recommendations.
Fatty acid oxidation assays. 14 C-labeled lauric acid (C 12 ) and palmitic acid (C 16 ) were from PerkinElmer, while 14 C-octanoic (C 8 ) and dodecanedioc (dicarboxylic C 12 ) acids were from Moravek, Inc. All but C 8 were bound to fatty acid-free albumin; C 8 was dissolved in 10 mg/ml α-cyclodextrin. For experiments with Sirt5 knockout mouse embryonic fibroblasts (MEFs), which were a kind gift of Dr. Eric Verdin (Buck Institute), the cells were grown to near confluence in T75 flasks. The cells were harvested and resuspended in DMEM containing 5 mM glucose and 125 µM of labeled fatty acid. Cells were rotated in a 37 °C water bath for 1 h. Then, perchloric acid was added to a final concentration of 0.5 M and 14 CO 2 was evolved, captured, and counted as described 50 . FAO in liver lysates were conducted in similar fashion. Freshly isolated liver was weighed and homogenized in 10 volumes of Mir05 media 51  Immunoblotting. Western blotting was performed after electrophoresis on Criterion SDS polyacrylamide gels (BioRad, Hercules, CA) and transfer to nitrocellulose membranes. Antibodies used were: rabbit anti-succinyllysine (PTM Biolabs), anti-malonyllysine (PTM Biolabs), anti-acyl-CoA oxidase-1 (ACOX1; Abcam), antienoyl-CoA hydratase/3-hydroxyacyl-CoA dehydrogenase (EHHADH; Abcam), and anti-β-actin (Proteintech). After incubation with HRP-conjugated secondary antibodies (1:5000) blots were visualized with chemiluminescence. In some experiments the blots were scanned and subjected to densitometric analysis using ImageJ software.
Histology and immunohistochemistry. Fresh portions of liver were fixed in 4% paraformaldehyde and embedded in paraffin, sectioned at 4 μm, and stained with hematoxylin and eosin (H&E) using standard methodology. Another portion of liver was embedded in OCT. frozen, and sectioned at 5-10 μm thick for Oil-Red-O staining. Immunostaining was performed as previously described 50,52 using anti-peroxisomal membrane protein-70 (PMP70; Abcam ab85550 at 1:200) and anti-glutamine synthase (GS; Sigma G2781 at 1:1500). Assessment of histology was performed by the Biospecimen Repository and Processing Core of the Pittsburgh Liver Research Center.
Urine adipic acid. Urine was collected from experimental and control animals over a 24 h period using metabolic caging. Creatinine concentration was determined using a kit (Cayman Chemicals, Inc). Adipate was determined in the context of standard clinical urine organic acid assessment at the Children's Hospital of Pittsburgh Clinical Biochemical Genetics Laboratory 53,54 . Briefly, a volume of urine was utilized equal to 1.0 mM creatinine. A 2-phenylbutyrate internal standard is included. Organic acids were extracted by sequential ethyl ether and acetoacetate extractions. Trimethylsilane derivatization was employed. Analysis utilized Agilent 7890A gas chromatography and 5975C mass spectrometry. All peak areas were normalized to that of 2-phenylbutyrate. Fragmentation patterns of urine analytes were compared to an internally compiled fragmentation library and the fragmentation library of the National Institute for Standards and Technology.
Liver triglyceride content. Approximately 100 mg of liver was weighed, chopped finely and digested in 350 µl of ethanolic KOH at 55 °C overnight. After centrifugation to pellet any undigestible material, the sample volumes were brought to 1.2 ml with 50% EtOH. 200 µl was removed to a new tube and mixed with 215 µl of 1 M MgCl 2 . Samples were clarified once more by centrifugation at 8000 × g and the supernatant used for assaying glycerol content. Glycerol was assayed spectrophotometrically at 540 nm using a kit (Sigma). Triglyceride content was normalized to tissue weight.
Enzyme activity assays. The anaerobic electron transfer flavoprotein (ETF) fluorescence reduction assay was used to measure acyl-CoA dehydrogenase activities as described 55 . His-tagged recombinant enzymes were purified as described 47 . Assays contained either 150 ng of purified recombinant protein or 200 µg of total liver protein, 2 µM recombinant porcine ETF, and 25 µM acyl-CoA substrate. C 8  www.nature.com/scientificreports/ tion of ETF fluorescence was followed for 1 min and used to calculate specific activity normalized to protein concentration. Mitochondrial trifunctional protein (TFP) activity was measured in the reverse direction using 3-ketodecanoyl-CoA (3-keto C 10 -CoA) and 3-ketopalmitoyl-CoA (3-keto-C 16 -CoA). Reactions contained 30 µg of liver homogenate and 50 µM substrate in a reaction buffer consisting of 100 mM KPO 4 , 50 mM MOPS, 0.1 mM DTT, 0.1% Triton-X100, and 0.15 mM NADH. The conversion of NADH to NAD + was followed in a plate reader at 340 nm. C 8 and C 12 -CoA synthetase activities were measured using 14 C-C 8 and 14 C-C 12 . Liver mitochondria were isolated by differential centrifugation and resuspended in cold SET buffer (10 mM Tris-HCl, 250 mM sucrose, 1 mM EDTA). Synthetase reactions (200 µl) contained 5 µl of homogenate with 10 µM 14 C-fatty acid in a buffer of 40 mM Tris-HCl, 5 mM ATP, 5 mM MgCl 2 , 4 mM CoA, 0.8 mg/ml Triton WR1339, and 1 unit/ml inorganic pyrophosphatase. After 2 min incubation at 37 °C, reactions were stopped with sulfuric acid and extracted either four times with ether (C 8 -CoA synthetase) or chloroform/methanol (C 12 -CoA synthetase) to separate formed acyl-CoAs from excess 14 C-fatty acid. Activities were normalized to protein concentration.