Hepatic CPT1A Facilitates Liver–Adipose Cross Talk via Induction of FGF21 in Mice

The global epidemic of obesity is accompanied by a rising burden of nonalcoholic fatty liver disease (NAFLD), affecting about one out of three people in the Western world (1,2). Obesity leads to excessive hepatocellular lipid accumulation, a benign condition called hepatic steatosis (fatty liver). When combined with “secondary hits,” including chronic liver inflammation or injury, hepatosteatosis develops into more aggressive forms, such as nonalcoholic steatohepatitis and hepatocellular carcinoma (3).

Mechanistically, abnormal lipid accumulation within the liver stimulates endoplasmic reticulum (ER) stress, which is mediated by three integral ER proteins (4): protein kinase RNA-like ER kinase, inositol-requiring enzyme-1, and activating transcription factor-6 (ATF6). ER stress activation underlies hepatic inflammation and fibrosis, which are characterized features of advanced fatty liver diseases, such as nonalcoholic steatohepatitis and hepatocellular carcinoma. Furthermore, liver dysfunction promotes systemic inflammation and insulin resistance via interorgan communication (5). However, in contrast to the unambiguous consensus on fatty liver diseases at their late stages, it is less clear how the early stage of hepatic steatosis impacts on liver biology and systemic metabolism. In particular, it is still unclear whether lipid accumulation in hepatocytes per se is an etiological driver of ER stress.

In addition to regulating glucose and lipid metabolism, the liver plays a central role in the maintenance of systemic metabolic homeostasis, through interorgan cross talk (6). This communication is mediated mainly through various signaling molecules secreted from liver, including hormones (also called hepatokines), which integrate liver metabolic status with systemic needs. Fibroblast growth factor 21 (FGF21) is a key hepatokine that exerts pleiotropic metabolic actions on various organs, such as adipose tissue and skeletal muscles, to antagonize obesity and diabetes (7). The hepatic gene expression of FGF21 is transcriptionally regulated by peroxisome proliferator–activated receptor α (PPARα), a transcription factor activated by free fatty acid (8). PPARα response elements have been identified within both human and mouse FGF21 promoters. Pharmacological administration of FGF21 elicits favorable effects on metabolism, including decreasing body weight, increasing lipid utilization, promoting adipose browning and energy expenditure, improving glucose disposal and insulin sensitivity, and decreasing ER stress in mice (7,9–11).

Carnitine palmitoyltransferase 1 (CPT1) is a mitochondrial outer membrane enzyme that catalyzes the transfer of the acyl group of a long-chain fatty acyl-CoA from CoA to l-carnitine. This reaction permits movement of the acyl carnitine from the intermembrane space to the mitochondrial matrix, where the acyl carnitine is converted back to acyl-CoA for subsequent β oxidation. CPT1A is the predominant isoform in liver that catalyzes the rate-limiting step for long-chain fatty acid β oxidation (12). However, the exact physiological function of CPT1A within the hepatocytes, in terms of its role in progression liver pathology and maintenance of whole-body metabolic homeostasis, has not been fully characterized. In this study, we generated liver-specific Cpt1a knockout (LKO) mice to investigate the role of hepatic fatty acid oxidation on the pathogenesis of NAFLD and systemic energy metabolism. Unexpectedly, despite more pronounced hepatic steatosis, LKO mice were resistant to high-fat diet (HFD)–induced obesity and to developing insulin resistance and liver injury. We show that these beneficial metabolic effects are mediated by activation of hepatic PPARα-FGF21 signaling axis that promotes interorgan communication between liver and adipose tissue. Our findings demonstrate that in the early stages, the liver compensates against fatty acid deposition per se through increased secretion of FGF21, which confers protection against obesity-evoked ER stress and cellular injury in liver, as well as overt adiposity and systemic insulin resistance.

Cpt1a flox mice in the C57BL/6J background were generated by Shanghai Model Organisms Center, Inc. and mated with Albumin-Cre transgenic mice (purchased from The Jackson Laboratory) to obtain hepatocyte-specific Cpt1a knockout (KO) mice. The wild-type (WT) sibling littermates were used as the control. Eight-week-old male mice were housed in a controlled environment (22 ± 1°C and 60–70% humidity with 12-h light/dark cycle) and fed ad libitum with standard chow (LabDiet) or 45 kcal% HFD (D12492; Research Diets) for 12 weeks. Food intake and body weight of the mice were measured weekly. For insulin-stimulated Akt phosphorylation, the mice were fasted overnight before peritoneal injection of insulin (0.75 IU/kg body weight). The tissues were harvested 20 min later for subsequent analysis. All animals received humane care according to the criteria outlined in the “Guide for the Care and Use of Laboratory Animals” conducted in accordance with the Guide for the Care and Use of Laboratory Animals Committee of Guangzhou Institutes of Biomedicine and Health, Chinese Academy of Sciences.

Rabbit polyclonal antibody against mouse FGF21 was obtained from Antibody and Immunoassay Services, The University of Hong Kong. For in vivo neutralization, after 4 weeks of HFD feeding, the mice were injected with FGF21 antibody (250 µg/kg body weight) every 2 weeks via tail vein.

Cells were lysed in lysis buffer containing 1% Nonidet P-40, 150 mmol/L NaCl, 10 mmol/L Tris-Cl (pH 7.5), and 1 mmol/L EDTA, resolved by 10% SDS-PAGE, and transferred to a polyvinylidene difluoride membrane. The membrane was then blotted with antibodies against CPT1A (NBP1–59608; Novus Biologicals); UCP1 (ab10983; Abcam); and GAPDH (cat. no. 5174T), AKT (cat. no. 4685), p-AKT (cat. no. 4060), eIF2α (cat. no. 5324T0), p-eIF2α (cat. no. 5324T), ATF6 (cat. no. 65880S) and CHOP (cat. no. 5554T) (all from Cell Signaling Technology).

Total RNA was isolated with TRIzol (Invitrogen). The first-strand cDNA was synthesized with Superscript III Reverse Transcriptase (Invitrogen) with 0.5 μg of RNA as the template for each reaction. mRNA levels were quantified under optimized conditions with SYBR Premix Ex Taq (Takara Bio) following the manufacturer’s instructions. 18s ribosomal RNA was used as the reference gene.

Liver tissue and brown, inguinal, and epididymal adipose tissues were fixed in 4% formaldehyde overnight at room temperature, embedded in paraffin, and cut into 5-μm sections with a microtome. Slides were deparaffinized, rehydrated, and stained with hematoxylin-eosin (H-E) (Sigma-Aldrich) using a standard protocol. Alternatively, sections were stained with anti-UCP1 antibody (1:200; ab23841; Abcam) and developed with SIGMAFAST DAB with Metal Enhancer (Sigma-Aldrich). Sections were examined by light microscopy (Motic BA600) and photographed with a Moticam Pro 285A. Photomicrographs were scanned with an Abaton Scan 300/Color scanner.

Whole-body oxygen consumption was measured with an open-circuit indirect calorimetry system with automatic temperature and light controls (Comprehensive Lab Animal Monitoring System; Columbus Instruments) as described (13). Mice had ad libitum access to chow and water in respiration chambers, and data were recorded for 48 h, including 24 h of acclimatization. Energy expenditure was calculated in CLAX 2.2 as recommended by the manufacturer. VO₂ is normalized to the body weight of the mice.

For glucose tolerance tests (GTTs), mice were fasted overnight and injected i.p. with 20% glucose at a dose of 2 g/kg body weight. For insulin tolerance tests (ITTs), mice were starved for 6 h before injecting recombinant human insulin (0.5 units/kg body weight; Eli Lilly and Company). Blood glucose was monitored from tail vein blood using a glucometer (Accu-Chek Advantage; Roche Diagnostics) at various time points.

Inguinal subcutaneous adipose tissues were dissected from mice, rinsed in PBS, minced, and digested for 40 min at 37°C in 0.1% (w/v) type I collagenase solution (Sigma-Aldrich) with D-Hanks’ buffer as described previously. Digested tissue was filtered through a 250-μm nylon mesh and centrifuged at 800g for 3 min. The sediment was resuspended in DMEM (Gibco) with 10% FBS (HyClone) and used for cell differentiation.

Total RNA was extracted from tissue using TRIzol Reagent according the manufacturer’s instructions (Invitrogen) and genomic DNA removed using DNase I (Takara Bio). An RNA-sequencing (RNA-seq) transcriptome library from 5 μg of total RNA was prepared using a TruSeq RNA sample preparation Kit (Illumina, San Diego, CA). Following quantification by TBS380, a paired-end RNA-seq library was sequenced with an Illumina HiSeq X Ten (2 × 150-bp read length). To identify differentially expressed genes (DEGs) between two different samples, the expression level of each transcript was calculated according to the fragments per kilobase of exon per million mapped reads method. RSEM (https://deweylab.biostat.wisc.edu/rsem/) was used to quantify gene abundances. R statistical package software EdgeR (Empirical analysis of Digital Gene Expression in R, https://www.bioconductor.org/packages/2.12/bioc/html/edgeR.html) was used for differential expression analysis.

WT and LKO mice were anesthetized and perfused with cold PBS. The left lateral lobe of liver was then isolated and placed in Hanks’ balanced salt solution in a noncoated Petri dish in a sterile hood and cut into fine pieces (2 mm³). The liver tissues were cultured in serum-free DMEM with 5.5 mmol/L glucose at 25 mg liver tissue/500 mL medium. Conditioned medium was then harvested 24 h later and filtered through a 0.22-μm filter (Millex). The conditioned medium (1:3 dilution) was used to culture in vitro–differentiated adipocytes for 24 h.

The primary hepatocytes were isolated from the mice as described (14). The primary hepatocytes were seeded in XFe24 Seahorse Bioanalyzer and subjected to the FAO/XF Cell Mito Stress Test assay as described by the instructions from the provider. Briefly, the primary hepatocytes were pretreated with etomoxir (40 μmol/L) for 15 min prior to the assay. The measure protocol was set as: three measurement cycles in each step, and each cycle included 3 min of mixing, 2 min of waiting, and 3 min of measuring. A total of 5 μmol/L oligomycin (Sigma-Aldrich), 0.5 μmol/L carbonyl cyanide-4-(trifluoromethoxy) phenylhydrazone (Sigma-Aldrich), and 2 μmol/L rotenone/antimycin A (Sigma-Aldrich) was sequentially injected to assess ATP production–, proton leak–, maximal respiration–, and nonmitochondrial respiration–related oxygen consumption rate.

Concentrations of FGF21 in serum and conditioned medium were determined by ELISA (R&D Systems). Levels of adiponectin and leptin were measured by ELISA kits purchased from BioVendor.

The acyl carnitines in the mouse liver tissues were measured by liquid chromatography-mass spectrometry–based lipidomics in Metabo. Sample preparation was performed according to a modified protocol based on Chang et al. (14). The acyl carnitines were analyzed using an ACQUITY ultra-performance liquid chromatography system (Waters Corporation, Milford, MA) equipped with a binary solvent delivery system and an autosampler (Waters Corporation), coupled to a tandem quadrupole-time-of-flight mass spectrometer (Waters Corporation).

Data are expressed as means ± SEM. ANOVA and unpaired two-tailed t tests were used for most comparisons using Prism 5 software (GraphPad, San Diego, CA). Post hoc tests were run only when F achieved P < 0.05 and there was no significant variance in homogeneity. A P value <0.05 was considered significant.

The data sets used and/or analyzed during the current study are available from the corresponding author upon request.

To investigate the physiological function of CPT1A in liver, hepatocyte-specific Cpt1a knockout mice were generated (Supplementary Fig. 1A). Liver-specific deletion of Cpt1a was confirmed at mRNA and protein levels, respectively (Supplementary Fig. 1B and D). The mRNA levels of CPT1B were induced in the liver of LKO mice, compared with the WT littermates under both standard chow and HFD conditions (Supplementary Fig. 1B and C), but still no CPT1B protein was detected (data not shown). Despite the compensatory induction of CPT1B, the fatty acid β oxidation was significantly mitigated in the liver of the LKO mice, as evidenced by the reduced rate of hepatic fatty acid oxidation in primary hepatocytes (Supplementary Fig. 2A and B). The short-chain acyl-carnitines were higher and the long-chain acyl-carnitines, especially palmitoyl-carnitine, were lower in LKO mice on HFD, indicating the liver of the LKO mice shifted to use more short-chain fatty acid (Supplementary Fig. 2C). Quantitative PCR analysis revealed that some other fatty acid oxidation genes and the glycolytic genes were slightly induced (Supplementary Fig. 2D and E). No fatty acid synthesis and gluconeogenic genes or liver glycogen content were significantly changed (Supplementary Fig. 2F–H). Consistently, circulating concentration of ketone bodies and respiratory exchange ratio were elevated in LKO mice (Supplementary Fig. 2I–K). Collectively, these data suggest that LKO mice adapt to a compensatory energy utilization strategy and use relatively more carbohydrate than lipid as their fuel source, compared with the WT mice.

On standard chow diet, LKO mice developed normally with no apparent differences to WT mice with regard to body weight, food intake, liver weight, and lipid content (Supplementary Fig. 3). Interestingly, when the mice were challenged with HFD, mean body weights of LKO mice were significantly lower than WT mice despite a comparable level of food intake (Fig. 1A–C). X-ray computed tomography analysis further showed that the relative decrement of body weight in LKO mice was attributed to a reduced percentage of body fat mass, but not lean mass (Fig. 1D). In addition, HFD-evoked glucose intolerance was substantially mitigated upon liver-specific deletion of Cpt1a, as determined by GTT (Fig. 1E and F). Furthermore, obese LKO mice displayed a decrease in fasting insulin concentrations (Fig. 1G), indicating that hepatic deletion of Cpt1a rendered mice resistant to obesity-associated hyperinsulinemia. ITT showed a greater glucose-lowering effect of insulin in obese LKO mice as compared with WT mice, suggesting increased peripheral glucose disposal (Fig. 1H and I). Consistent with these results, Western blot analysis showed that insulin-stimulated phosphorylation of AKT was enhanced in skeletal muscle, liver, and white adipose tissue (WAT) (Fig. 1J–M). Collectively, these data demonstrate that upon HFD feeding, the absence of hepatic CPT1A improves systemic glucose tolerance and insulin resistance compared with WT mice.

When challenged with HFD, the liver was significantly increased in weight and size in LKO mice, compared with WT mice (Fig. 2A and B). The liver from LKO mice also appeared paler in color, implying a higher content of lipid deposition. Subsequent H-E and Oil Red O staining analyses of tissue section qualitatively confirmed increased lipid deposition in the liver of LKO mice (Fig. 2C and D). Increased lipid deposition in liver was also confirmed by quantitative biochemical assay of triglyceride (TG) content (Fig. 2E). In contrast, the lipid contents of diacylglycerol were not different between WT and LKO (Supplementary Fig. 2L). The circulating nonesterified fatty acid level was lower in HFD-fed LKO mice, compared with the WT mice (Supplementary Fig. 4A). Interestingly, despite the higher hepatic lipid deposition, LKO mice were resistant to HFD-evoked liver damage, as shown by reductions in serum ALT and AST concentrations (Fig. 2F and G). LKO liver also exhibited a lower level of inflammation, as evidenced by the reduced levels of inflammatory genes (Fig. 2H). In addition, ER stress markers, including Grp78, Grp94, Erdj4, Atf4, Chop, and the spliced form of Xbp1 (sXbp1), were significantly reduced in obese LKO livers, compared with age- and sex-matched control mice (Fig. 2I). Likewise, Western blotting showed protein levels of key ER stress markers, such as ATF6 and CHOP, and phosphorylation of eIF2α were comparatively lower in LKO liver (Fig. 2J and K). Collectively, these results demonstrate that CPT1A deficiency in hepatocytes potentiates HFD-induced hepatic steatosis but protects against liver injury, inflammation, and ER stress. The top 100 up- and downregulated genes in the liver of LKO mice are listed in Supplementary Table 2.

In contrast to the enlarged liver, Cpt1a LKO mice displayed reduced weight of white adipose depots, including both inguinal WAT (iWAT) and epididymal WAT, compared with WT mice (Fig. 3A and B). The adipocyte size in these two adipose depots was also smaller in LKO mice than in WT mice (Fig. 3C and D). Furthermore, oxygen consumption rate was elevated in LKO mice (Fig. 3E–H) and associated with increased expression of thermogenesis-related genes, including Ucp1, Prdm16, Dio2, Cidea, and Cox8b, in interscapular brown adipose tissue (BAT) and iWAT (Fig. 3I and K). This was further validated by Western blotting and immunochemistry, which showed that protein expression of UCP1 was increased in both BAT and iWAT (Fig. 3J and L). These results imply that deletion of Cpt1a in liver boosts the metabolic rate by enhancing browning within adipose tissues.

We proceeded to examine the mechanism underlying the metabolic effects of liver-specific CPT1A deficiency. RNA-seq analysis showed that the PPAR signaling pathway was identified as the most prominent one induced in liver upon deletion of Cpt1a (Fig. 4A). In support of this, multiple genes downstream of PPARα were strongly induced in Cpt1a-deficient liver (Fig. 4B and C). Intriguingly, one of the downstream targets of PPARα, fibroblast growth factor 21 (Fgf21), was found among one of the most significantly upregulated genes in Cpt1a KO liver (Fig. 4C). FGF21 is a hepatokine well-recognized as a key player to enhance adipose browning and metabolic health (15). The elevation of Fgf21 in liver of LKO mice was further verified by real-time PCR and Western blotting (Fig. 4D–F). Indeed, plasma FGF21 concentration in LKO mice was almost doubled, compared with WT mice (Fig. 4G). In contrast, serum concentration of adiponectin and leptin was not significantly changed in LKO mice (Supplementary Fig. 4B and C). Subsequent experiments confirmed that this increased circulated FGF21 was liver-derived. In this study, ELISA measurements of culture media from in vitro–cultured liver explants showed obese LKO mice secreted higher amounts of FGF21 as compared with WT mice (Fig. 4H). To gain additional insight into the mechanism by which CPT1a deficiency causes induction of FGF21, hepatocytes isolated from obese WT and LKO mice were treated with PPARα antagonist GW6471. We found upon inhibition of PPARα, the elevation in Fgf21 transcription was abolished in CPT1A-deficient hepatocyte (Fig. 4I), suggesting that activation of the PPARα signal is responsible for the elevated FGF21 in LKO liver.

To address whether the phenotypic changes in LKO mice were due to elevated circulating FGF21, a neutralizing antibody against FGF21 was administered to HFD-fed WT and LKO mice (16,17). Compared with the mice receiving species-matched control IgG, mice administered with FGF21 neutralizing antibody displayed increased total body and adipose tissue weights, especially in LKO mice (Fig. 5A and B). Most significantly, the difference in body weight and adiposity between WT and LKO mice was lost upon blockade of FGF21 (Fig. 5A and B), indicating that the elevation of circulating FGF21 in LKO mice was casually linked to the lean phenotype. Similarly, FGF21 neutralization exacerbated glucose intolerance and insulin resistance to comparable levels between WT and LKO mice (Fig. 5C and D). Furthermore, mRNA expression of brown markers was suppressed in BAT and iWAT of both WT and LKO mice after antibody-mediated neutralization of FGF21 (Fig. 5E and F), demonstrating that FGF21 serves as an intermediatory molecule responsible for potentiated adipose tissue browning in LKO mice. The Cpt1b gene was also induced in the BAT and iWAT of the LKO mice, and the induction was largely reversed upon neutralization of FGF21. This conclusion was further corroborated by in vitro experiments in which conditioned medium taken from LKO liver explants enhanced Ucp1 expression in primary adipocytes; this enhancement was abolished upon addition of the FGF21 neutralizing antibody to the medium (Fig. 5G).

Whether hepatic FGF21 acts autonomously was also examined. Neutralization of FGF21 exacerbates diet-induced hepatic steatosis in WT mice, as shown by higher liver weight and TG content (Fig. 6A and B). But these effects were absent in LKO mice, which suggests that CPT1A is required for FGF21-evoked hepatic lipid clearance. Nevertheless, in both WT and LKO mice, serum levels of AST and ALT were elevated upon FGF21 neutralization (Fig. 6C and D). Furthermore, ER stress markers were uniformly upregulated after treatment with the FGF21 antibody in both WT and LKO mice and were indistinguishable between two genotypes, in comparison with the littermates receiving control IgG (Fig. 6E). Taken together, our data demonstrate that Cpt1a deficiency–evoked steatosis stimulates overproduction of FGF21, which serves as a rescue signaling molecule to alleviate hepatic ER stress and cell damage, boosts adipose tissue browning, and antagonizes both obesity and insulin resistance (Fig. 7).

Obesity is commonly accompanied by NAFLD. In the current study, we found that hepatic-specific Cpt1a knockout mouse develop hepatic steatosis but are resistant to HFD-induced obesity and insulin resistance. These protective effects are largely contributed by the elevated liver production of FGF21, which enhances the energy expenditure by stimulating the adipose browning. In addition to its role in glucose and lipid metabolism, the liver is increasingly recognized as a highly active endocrine organ, secreting a number of bioactive molecules that regulate whole-body energy metabolism and cardiovascular tone (6). In particular, accumulating evidence supports multilevel cross talk between the liver and adipose tissue that coordinates whole-body energy metabolism. This communication acts as a key player to modulate the initiation and development of metabolic diseases via secretion of an array of hepatokines and adipokines (18). The current study highlights the role of the liver–adipose axis in the regulation of whole-body energy metabolism during the onset of NAFLD. Notably, FGF21 production from lipid-laden hepatocytes acts as a rescue signal sent by the liver to counteract the lipid overload by increasing adipose browning and energy expenditure.

In humans, there is dissociation between obesity and NAFLD. A recent meta-analysis of 84 studies (n = 10,530,308) revealed that 19.2% of people with NAFLD were lean and 40.8% were nonobese (19,20). On the basis of our findings, it would be intriguing to compare the liver secretomes between these nonobese subjects and obese patients to see whether circulating concentrations of FGF21 or other hepatokines are altered.

Other findings show that mice with liver-specific ablation of Cpt2, the mitochondrial inner membrane protein that converts acyl-carnitine back to acyl-coA in the mitochondria matrix for β-oxidation (12), inhibit hepatic gluconeogenesis, which is sufficient to alleviate obesity-associated glucose intolerance (21). In the case of Cpt1a KO mice, our results showed that the improved glucose disposal in LKO mice was at least partially attributed by the increment in liver-derived FGF21 in circulation. Interestingly, the gluconeogenic genes were not inhibited in LKO mice (Pck1 even exhibited a modest increase), indicating that deficiency of CPT1A and CPT2 poses different mechanisms of actions on liver metabolism, which shall be addressed in future studies.

We also found that the mRNA expressions of Cpt1b and Cpt2 were induced in the iWAT and BAT in LKO mice, while Cpt1a is not significantly altered. But additional caution shall be taken when translating this finding to other species considering the species differences on CPT1 isoforms, especially in white adipocytes. Brown et al. (22) reported that the mouse white adipocytes differ from those of the other species, including human and rat, in terms of its predominant enzyme in fatty acid metabolism, although all three rodents mainly express CPT1B. It is also worth mentioning that despite similar biochemical functions, the physiological consequences of CPT1 deficiency were depot-specific. Skeletal muscle-specific Cpt1b KO induces adaptive remodeling of muscle metabolism, accompanied by resistance to obesity and insulin resistance (23). Interestingly, whole-body heterogenous Cpt1b KO mice exhibited an exacerbated insulin resistance under the long-term HFD conditions and aggravated stress-induced cardiac hypertrophy (24,25). Collectively, these studies suggest that CPT1 isoforms work by different mechanisms in a tissue-specific manner.

Our antibody-neutralization experiments demonstrate that upregulated circulating FGF21 concentration in simple steatosis also protects against ER stress and liver injury, in both WT and LKO mice. Notably, the beneficial action of FGF21 is independent of the reduction in hepatic lipid deposition, since improvement was also evident in the liver of LKO mice, where hepatosteatosis remained unchanged after FGF21 antibody treatment due to lack of CPT1A. Consistent with these findings, FGF21 inhibits ER stress–induced cellular injury via the FGFR1-ERK1/2 signaling pathway in rat cardiomyocytes (26). Fgf21-null mice also exhibit increased expression of ER stress marker genes and augmented hepatic lipid accumulation after tunicamycin treatment; these changes were attenuated in inducible Fgf21-transgenic mice (10). Furthermore, FGF21 is the target gene for ATF4 and CHOP, such that FGF21 expression is regulated by ER stress via ATF- and CHOP-dependent mechanisms in hepatocytes (27). Therefore, there exists a feedback mechanism between FGF21 and ER stress responses to ensure metabolic homeostasis in liver. The glucose-mediated expression of Fgf21 is also controlled by CHREBP (28–30). Whether this transcription factor coordinates with PPARα to contribute to the elevation of FGF21 warrants future investigation.

In addition to FGF21, a number of other genes downstream of PPARα are also upregulated in the liver of LKO mice. For example, Cyp4 subgene family members in the cytochrome P450 family, including Cyp4a10, Cyp4a12b, Cyp4a14, Cyp4a31, and Cyp4a32, were upregulated upon deletion of CPT1A. These genes are responsible for ω-oxidation of fatty acid metabolism, which is normally a minor catabolic pathway for medium- to long-chain fatty acids. Ehhadh, the gene in peroxisomal lipid metabolism, is also induced, likely as another alternative way of lipid metabolism in LKO liver. Acyl-CoA thioesterases, such as Acot1 and Acot2, are also higher in CPT1A-deficient liver, possibly to potentiate the re-esterification of excess fatty acids into triacylglycerol (TAG). Considering that some lipid metabolites are also potential signaling molecules, it remains an intriguing question whether the alteration in lipid metabolism directly influences the liver biology by production of alternative lipid signaling molecules. Furthermore, the detailed mechanism by which PPARα is activated in LKO liver remains unknown. It is also possible that certain lipid metabolites that lead to the activation of PPARα. More comprehensive lipidomics study is warranted to tackle this question.

A growing body of evidence suggests that lipid compartmentalization in hepatocytes and in particular the type or quality, as opposed to the quantity, of lipids play a predominant role in determining the risk for progressive NAFLD. By using the mouse model that selectively overexpresses diacylglycerol acyltransferase 2 (DGAT2), the enzyme in the final step of TG formation, Monetti et al. (31) demonstrated that hepatic TG accumulation is insufficient to cause hepatic or systemic insulin resistance. Conversely, inhibition of TG synthesis by DGAT2 antisense oligonucleotide treatment improves hepatic steatosis but exacerbates liver damage and fibrosis in obese mice with nonalcoholic steatohepatitis (32). Moreover, emerging data now imply that TG accumulation per se even represents a protective mechanism against free fatty acid–induced lipotoxicity (33). Exposure of cultured cells to unsaturated free fatty acids results in a significant increase in TG content without a decrease in cell viability. In contrast, cells incubated with saturated fatty acids showed a significant increase in apoptotic death in conjunction with absence of TG accumulation (34). These studies are consistent with our observations in HFD-treated LKO mice, which exhibit higher liver TG content compared with the WT mice. Thus, the liver with deficient CPT1A function likely represents a “healthy” fatty liver that fights against the deleterious actions of HFD on hepatic injury. In the early stages of fatty liver in which the liver is otherwise healthy, the liver compensates by decreasing CPT1A activity, which protects by increasing FGF21 circulating concentrations.

Our findings provide compelling evidence to support the feasibility of targeting hepatic CPT1A for the management of type 2 diabetes and its related disorders. Inhibition of CPT1 conceivably leads to decreased hepatic glucose output as supported by the development of hypoglycemia in patients with mutated hepatic CPT1A protein (35–37). CPT1A inhibitors have also been reported to improve hyperglycemia and insulin resistance in mice (38–40). Our study provides the first genetic evidence in an animal model that abolishment of hepatic CPT1A function is efficacious against obesity-evoked impairment in both liver tissue and peripheral organs. Significantly, the metabolic benefits extend beyond modulation of liver lipid and glucose metabolism. Instead, decreased CPT1A function in liver engages the PPARα-FGF21 axis to orchestrate an interorgan communication between liver and adipose tissue to propagate improve systemic metabolism.

Funding. This study was supported by the Frontier Research Program of Bioland Laboratory (Guangzhou Regenerative Medicine and Health Guangdong Laboratory, 2018GZR110105019), Guangzhou International Collaborative Grant (2019A050510027), International Partnership Program of the Chinese Academy of Sciences (154144KYSB20180063), National Natural Science Foundation of China Excellent Young Scientists Fund (Hong Kong and Macau) (81922079), National Natural Science Foundation of China (81970729, 81670800, and 81774134), Science and Technology Planning Project of Guangdong Province, China (2020B1212060052), Guangdong Provincial Public Interest Research and Capacity Building Projects (2014A010107024), and Natural Science Foundation of Jiangsu Province of China (BK20171331).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. W.S., T.N., K.L., W.W., Q.Lo., T.F., L.M., Y.G., Q.Li., X.G., D.Y., K.Y., P.G., X.Z., and K.C. carried out the research and analyzed the results. S.L., D.W., and X.H. supervised experiments and analyzed data. T.N., Y.X., K.M.L., S.L., D.W., and X.H. wrote and revised the manuscript. S.L., D.W., and X.H. are the guarantors of this work and, as such, had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.