PPA1 Regulates Systemic Insulin Sensitivity by Maintaining Adipocyte Mitochondria Function as a Novel PPARγ Target Gene

White adipose tissue (WAT) plays a key role in whole-body energy homeostasis. Besides its function as an organ to store excess energy, adipose tissue also secretes a number of adipokines, such as leptin, resistin, and adiponectin, to control body energy balance (1,2). Deficiency or dysfunction of adipose tissue often leads to an overflow of fatty acids into nonadipose tissues. Consequently, the ectopic lipid accumulation can interfere with insulin signaling and other tissue functions (termed lipotoxicity) and finally result in insulin resistance and diabetes (3). Although the underlying mechanisms that impair adipose tissue function are not fully elucidated, downregulation of mitochondrial function or biogenesis is considered as a central driver during the process.

The essential role of mitochondria in adipogenesis and mature adipocyte function has been well documented in previous studies. Pharmacologically inhibiting complex I or ATP synthase impaired adipocyte differentiation and proliferation in mice (4,5). In line with that, in vitro study showed that inhibiting of complex III led to reduced expression of adipogenic markers and impaired cell differentiation (6). As an important adipokine, adiponectin stimulates fatty acid oxidation and improves glucose metabolism and insulin sensitivity. Studies also demonstrated that inhibition of electron transport chain or deletion of mitochondria transcription factor A leads to reduced adiponectin secretion, decreased Glut4 expression, and attenuated insulin signaling (7,8). The antidiabetes drug class thiazolidinediones, ligands of PPARγ, was reported to influence mitochondrial functions. Rosiglitazone (RGZ) was reported to change the morphological features and protein profiles of mitochondria in 3T3-L1 adipocytes and promote mitochondrial remodeling and increased energy expenditure in white fat in vivo (9,10). Pioglitazone treatment significantly increased mitochondrial copy number and expression of factors involved in mitochondrial biogenesis, such as PGC-1α and mtTFA in human adipocytes (11). Besides its role as the most important regulator of adipocyte differentiation and function, PPARγ can also regulate a majority of genes involved in mitochondrial biogenesis and network remodeling (10). Therefore, improved mitochondrial mass and function are speculated to make a great contribution to the antidiabetic effects of PPARγ agonists, although the detailed mechanism is still not fully elucidated.

Inorganic pyrophosphatase 1 (PPA1) is an enzyme that catalyzes the hydrolysis of inorganic pyrophosphate (PPi) to inorganic phosphate (Pi), which is a highly exergonic reaction that can be coupled to some energy-demanding biochemistry reactions. PPA1 was reported to play an essential role for growth and development in the Caenorhabditis elegans and roundworms Ascaris (12). PPA1 defects may result in cell cycle arrest and autophagic cell death through inhibiting NAD⁺ generation in fermenting yeast (13). In mammals, studies revealed that PPA1 regulated neurite growth via a JNK dephosphorylation manner in mouse neuroblastoma cells (14), induced type I collagen synthesis, and stimulated calcification by osteoblasts (15). In addition, many studies have reported that expression of PPA1 was increased in many different types of highly proliferated cells (16,17). However, the role of PPA1 in metabolic syndromes has never been reported.

Previous studies in our laboratory revealed that PPA1 may participate in whole-body glucose metabolism, since overexpression of PPA1 can protect pancreatic β-cells against palmitate-induced cell apoptosis (18). Therefore, in this study, we have generated global PPA1 knockout mice (PPA1^+/–) to further explore the role of PPA1 in metabolic homeostasis. To our surprise, instead of insulin deficiency, increased insulin level was observed in PPA1^+/– mice accompanied by severe insulin resistance, suggesting that PPA1 played a more important role in maintaining insulin sensitivity in peripheral tissues. Further studies revealed that PPA1 deficiency in adipose tissue may be mainly responsible for the whole-body insulin resistance that we observed. Mechanistic studies revealed that PPA1 played a crucial role in maintaining mitochondrial function in adipocytes as a PPARγ target gene.

PPA1^+/– mice were generated at Cyagen Biosciences (Suzhou, China) on the background of C57BL/6J. Age-matched male PPA1^+/– mice and their wild-type (WT) littermates were provided either a normal chow diet (NCD) or high-fat diet (HFD) (#D12492; Research Diets) for 12 weeks. All mice were provided with free access to commercial rodent chow and tap water in a temperature- and humidity-controlled environment (23°C, 12-h light/dark cycle, 60–70% humidity) in a specific pathogen–free facility at Nanjing Medical University. All procedures concerning the animal care and use were in accordance with the guidelines of the institutional animal care and use committee of Nanjing Medical University (permit number IACUC-NJMU 14030178-1).

Glucose tolerance tests (GTTs) were performed by intraperitoneal (i.p.) injection of glucose (2 g/kg) after overnight fasting. Insulin tolerance tests (ITTs) were performed by i.p. injection of 1 unit/kg insulin after 4 h of fasting.

Blood glucose levels were measured using a glucometer (Abbott, Alameda, CA). Commercially available kits were used to measure plasma levels of insulin (EZAssay, Cat. # MS200), adiponectin (EZMADP-60K; Millipore, Temecula, CA), leptin (MOB00B; R&D Systems), triglycerides (TGs) (E1013; Applygen); free fatty acids (cat. no. 294-63601; Wako Chemicals), total cholesterol (TC) (A111-1-1; Jiancheng Bioengineering), and lipoproteins (H139; Jiancheng Bioengineering).

Whole-body oxygen consumption (VO₂) and carbon dioxide output (VCO₂) were assessed using the automated home-cage system (TSE; PhenoMaster). Data on VO₂, VCO₂, and food intake as well as on locomotor activity were monitored and recorded every 40 min for 5 consecutive days. Respiratory exchange ratio (RER) was calculated as the ratio of VCO₂/VO₂ and was normalized to lean body mass.

WT and PPA1^+/– mice were subjected to the hyperinsulinemic-euglycemic clamp assay as described previously (19).

Body composition was measured by MRI (Minispec NMR analyzer; Bruker). Total adipose tissue in WT and PPA1^+/– mice was determined by high-resolution micro-CT (SkyScan 1176 X-ray; Bruker). Mice were anesthetized with isoflurane and scanned from the tailbone to the lungs. Fat images and body fat ratio were quantified semiautomatically on dedicated software (NRecon).

Tissues isolated from mice were fixed overnight in 4% paraformaldehyde for hematoxylin-eosin (H-E) staining. Liver tissues were further stained with oil red O on frozen sections. Microphotographs were captured with an Olympus microscope.

Adeno-associated viral (AAV) vectors (1 × 10¹² viral genomes/mL) were injected directly into the inguinal WAT (iWAT) of 4-week-old PPA1^+/– male mice. Each subcutaneous fat pad was injected twice with 15 μL AAV vector. Purified AAV vector serotype 2/9 (AAV2/9) encoding PPA1 was generated by Hanheng Biotechnology (Shanghai, China).

The embryonic fibroblast mouse cell line 3T3-L1 (ATCC) was cultured and differentiated as described previously (20). The cultures were maintained in a humidified atmosphere of 5% CO₂ and 95% air at 37°C. The medium was replaced every 2–3 days. Passages 5–6 were used for all experiments.

The primary mouse preadipocytes of the iWAT were isolated from 4-week-old mice as previously described (21).

The lentivirus-mediated PPA1-specific shRNA (shPPA1) was constructed by Genomeditech Ltd. The shRNA sequence targeting mouse PPA1 was as follows: 5'-GCTACAAAGGACCCTTTAAAC-3'. The pGMLV-SC5 lentivectors containing the shRNA sequences were transfected into 3T3-L1 cells before differentiation.

Total RNA samples were extracted using TRIzol reagent. Real-time quantitative PCR was performed as described previously (22). The primer sequences are shown in Supplementary Tables 1 and 2.

The protein content determination and Western blotting were performed as described previously (23). Antibodies used are listed in Supplementary Table 3.

Chromatin immunoprecipitation (ChIP) assay was performed as described previously (22) [Chromatin Immunoprecipitation (ChIP) Assay Kit; Millipore]. Primers used were as follows: forward, 5'-ATCAAAGGTGAGCGACCCGCAAGC-3', and reverse, 5'-CGCCGTGCCTTTAATGCGTGGG.

Total DNA was isolated with a DNeasy DNA isolation kit (QIAGEN). The DNA levels of mitochondrial ND1 (NADH dehydrogenase 1) gene and nuclear 18S rRNA were determined by real-time PCR quantification. The relative mtDNA content was reflected by the ratio of DNA levels between mitochondrial ND1 and nuclear 18S rRNA as described previously (24).

VO₂ was measured using the XF24 Extracellular Flux Analyzer from Seahorse Bioscience as described previously (25).

Cells and tissues were fixed with 5% or 2.5% glutaraldehyde in 0.1 mol/L sodium cacodylate buffer for 2 h and postfixed in 1% OsO₄, 1.5% K₄Fe(CN)₆, and 0.1 mol/L sodium cacodylate for 1 h. Cells were en bloc stained, dehydrated, embedded, and cut into ultrathin sections (50–80 nm) followed by imaging and analysis using a transmission electron microscope.

Metabolites of cell lysate were extracted from shGFP and shPPA1 adipocytes as described previously (26). Bioinformatic analysis was performed using the metabolites with a Human Metabolome Database accession number in the web tool MetaboAnalyst (27). Details are described in the Supplementary Material.

The ppl-gal4 driver line was backcrossed to the white1118 line (ID 60000) from the Vienna Drosophila RNAi Center for >10 generations. Male UAS-RNAi transgenic flies (ID 103776 [Vienna Drosophila RNAi Center], ID31341 [Bloomington Drosophila Stock Center]) were crossed with ppl-gal4 virgin flies. For Nile red staining, a 10% stock solution of Nile red in DMSO was diluted 1:10,000 in a mixture of 1× PBS, 30% glycerol, and Hoechst. Stop-wandering larvae (8–12) were dissected, and lipid droplets (LDs) were counted and visualized with an Olympus FV1200 confocal microscope. For transmission electron microscopy, fat bodies were processed following the malachite green staining protocol. Micrographs were acquired using an FEI Tecnai G2 Spirit transmission electron microscope. For starvation sensitivity studies in adults, 5- to –7-day-old flies were used. Starvation sensitivity assays were conducted by transferring 15–20 flies of each genotype into vials containing 1% agar. Flies were transferred to new tubes every day, and dead flies were counted daily. At least 100 flies per genotype were scored.

Data are presented as mean ± SEM. Significant differences were assessed using a two-tailed Student t test or one-way ANOVA followed by the Student-Newman-Keuls test. All data analyses were carried out using GraphPad Prism 8 (GraphPad Software). P < 0.05 was considered statistically significant.

The data sets, adenoviruses, and plasmids generated during and/or analyzed during the current study are available from the corresponding authors upon reasonable request.

To evaluate the effect of PPA1 in metabolic disorders, PPA1^+/– mice were generated. However, since no homozygous mice (PPA1^–/–) were identified in the offspring, PPA1 heterozygous mice (PPA1^+/–) were used in the study. Knockdown efficiency of PPA1 in different tissues was detected as shown in Supplementary Fig. 1. When feeding with NCD, PPA1^+/– mice displayed comparable physiological status to WT controls. No significant differences were observed in body weight, blood glucose level, GTT, or ITT (Fig. 1A–D). The hyperinsulinemic-euglycemic clamp, which has been used as a “gold standard” method to accurately measure insulin action, was further used to evaluate insulin sensitivity. However, decreased glucose infusion rate (GIR), which indicates impaired insulin sensitivity, was observed in PPA1^+/– mice (Fig. 1E and F). Notably, we found that the insulin resistance develops with aging in PPA1^+/– mice. Aging has been considered a major risk factor for most chronic diseases since individuals gradually lose the ability to maintain homeostasis (28). Here, we also found higher fasting blood glucose and insulin levels in aging PPA1^+/– mice (Fig. 1G and H). Although no differences in glucose tolerance were observed (data not shown), ITT results indicated impaired insulin sensitivity in aging PPA1^+/– mice with NCD-feeding (Fig. 1I), suggesting that PPA1 plays an important role in maintaining whole-body insulin sensitivity.

To better understand the role of PPA1 in metabolism, PPA1^+/– and WT mice were fed an HFD, which can induce the metabolic stress. As we expected, PPA1^+/– mice showed more body weight gain and higher fasting blood glucose level compared with WT (Fig. 2A–C). To determine the mechanisms by which PPA1^+/– mice were susceptible to HFD-induced obesity, food intake and energy expenditure were detected. In the absence of differences in food intake (data not shown), the VO₂, VCO₂, and RER were significantly decreased in PPA1^+/– mice (Fig. 2D–G), indicating that rather than increased energy intake, reduced energy expenditure contributed to the obesity of PPA1^+/– mice. Accordingly, impaired glucose tolerance and severe insulin resistance were observed in HFD-fed PPA1^+/– mice (Fig. 2H–K). Similarly, female PPA1^+/– mice also showed decreased insulin sensitivity as shown in Supplementary Fig. 2. Furthermore, insulin sensitivity in muscle, liver, and fat tissue was determined by detecting the insulin signaling transduction. As shown in Fig. 2L, the phosphorylation of Akt and IR was significantly repressed, indicating an inhibited insulin signaling pathway in PPA1^+/– mice. These results suggested that decreased insulin sensitivity occurred in mice with PPA1 deficiency and further led to whole-body metabolic disorders.

During HFD-induced obesity, excess lipids accumulated first in WAT and subsequently led to ectopic fat deposition in tissues such as liver and muscle. Given that the body weight was significantly increased in HFD-fed PPA1^+/– mice, the body composition was measured, and results indicated that the fat mass was increased while the lean mass was decreased in PPA1^+/– mice (Fig. 3A). However, the micro-CT scanning results revealed that despite the increased fat mass, the adipose tissue, including subcutaneous adipose tissue and visceral adipose tissue, were decreased in PPA1^+/– mice (Fig. 3B and C). These results prompted us to further determine the effect of PPA1 on fat distribution in vivo. Consistent with the micro-CT results, the liver weight was significantly increased while the epididymal fat weight was markedly decreased in PPA1^+/– mice (Fig. 3D and E). The histological analysis also showed more severe hepatic lipid accumulation in PPA1^+/– mice fed the HFD (Fig. 3F). In addition, increased nonesterified fatty acid (NEFA) and TG levels were enhanced in not only liver but also skeletal muscle of PPA1^+/– mice (Fig. 3G–J). Furthermore, HFD-fed PPA1^+/– mice showed increased serum TG and NEFA levels and higher TC content (Fig. 3K–M). These data suggested that more fat was accumulated in nonadipose tissue in PPA1^+/– mice, which may further lead to more severe insulin resistance and metabolic disorders.

Adipose tissue is the primary site to store excess energy and is often expanded during HFD feeding. However, in this study, reduction of fat tissue weight was observed in PPA1^+/– mice, which suggested that deficiency of PPA1 impaired adipose tissue development and function. Therefore, more detailed studies were carried out on the morphology and function of adipose tissues. Histological analyses showed that the average adipocyte size was decreased in iWAT and epididymal WAT (eWAT) from PPA1^+/– mice (Fig. 4A and B). More importantly, the adipocytes from PPA1^+/– mice were vastly variable in size, displaying increased frequency of large hypertrophic adipocytes or very small adipocytes, which were considered less metabolically favorable and were associated with pathophysiological conditions (29). In addition, multiple genes involved in adipocyte differentiation and function were downregulated dramatically in PPA1^+/– mice (Fig. 4C and D). Circulating adiponectin levels were suggested to have strong correlations with various disease states (30). Consistently, the serum adiponectin level was significantly decreased in PPA1^+/– mice, indicating the dysfunction of adipose tissue with PPA1 deficiency (Fig. 4E).

Given the above observation, we speculated that PPA1 deficiency could impair adipose function and further lead to severe nonadipose tissue lipid deposition and whole-body metabolic disorders. Therefore, to test this hypothesis, PPA1 was overexpressed with PPA1-AAV in iWAT of PPA1^+/– mice by in situ injection. As shown in Fig. 5A, PPA1 was highly expressed in iWAT from PPA1-AAV mice, and no significant nonspecific infection was observed (Supplementary Fig. 3A). In line with our hypothesis, PPA1-AAV mice showed an increased serum adiponectin level (Fig. 5F), better glucose tolerance, and improved insulin sensitivity compared with GFP-AAV mice (Fig. 5B–E), despite that no significant difference was observed in the body weight gains (Supplementary Fig. 4A). In addition, PPA1 restoration significantly increased the average adipocyte size and decreased the frequency of small or large hypertrophic adipocytes in iWAT (Fig. 5G and H). Adipocyte hypertrophy was suggested to result in cell death, macrophage infiltration, and chronic inflammation (31). In line with this, decreased macrophage infiltration and reduced inflammatory-associated gene expression was observed in iWAT from PPA1-AAV mice (Supplementary Fig. 4B and C). Although the average adipocyte area was not altered in eWAT, distribution of small and large adipocytes also showed decreased tendency. As a consequence, hepatic steatosis was significantly ameliorated in PPA1-AAV mice. H-E and oil red O staining results revealed much lower lipid content in liver from PPA1-AAV mice (Fig. 5I). Accordingly, decreased liver weight mass and reduced NEFA and TG levels were also observed (Fig. 5J). More importantly, insulin sensitivity was significantly restored in liver and muscle just like in iWAT, despite the fact that PPA1 expression in liver and muscle was not restored (Fig. 5K). These results support our view that deficiency of PPA1 in adipose tissue is the main reason responsible for the metabolic disorders in PPA1^+/– mice.

To better understand the role of PPA1 in adipocytes, we first evaluated PPA1 expression at various time points during differentiation and found that expression of PPA1, together with PPARγ, was time-dependently increased during cell differentiation. In addition, we observed that RGZ, the PPARγ agonist that enhances the cell differentiation, can further increase PPA1 expression in 3T3-L1 cells (Fig. 6A). Consistently, PPA1 protein levels showed a decreased tendency in GW9662 (PPARγ antagonist)-treated cells (Fig. 6B). These results prompted us to further explore whether upregulation of PPA1 during adipocyte differentiation may be partially caused by increased PPARγ activation. The search results from the GeneCards database suggested that there are possible binding sites of PPARγ in the PPA1 promoter region. ChIP-PCR was further performed, and as shown in Fig. 6C, PPARγ can directly bind to the promoter of PPA1. These results indicated that PPA1 could play an important role in adipocytes as one of the PPARγ target genes. Given that PPARγ is considered the key regulator of both adipogenesis and mitochondrial biogenesis, we further evaluated the effect of PPA1 interference on mitochondrial content during adipocyte differentiation, and the interference efficiency was evaluated as shown in Fig. 6D. Our results demonstrate that the mtDNA content in 3T3-L1 cells was repressed upon PPA1 interference, especially when the cells were subjected to adipogenic induction (Fig. 6E). Next, the ultrastructure of mitochondria in cells was examined using transmission electron microscopy, and we found that the mitochondria in shPPA1 3T3-L1 cells had a more rounded appearance, and cristae content was also diminished compared with the control group (Fig. 6F). In line with that, mitochondria in iWAT from PPA1^+/– mice also declined both in number and in size. Decreased mtDNA content as well as altered mitochondrial morphology was observed (Fig. 6J and K). As we expected, PPA1 restoration significantly increased mtDNA content and reversed morphological alterations of mitochondria in iWAT (Fig. 6J and K). To better describe the morphological alteration of mitochondria, we assessed the morphology with MitoTracker Red CMXRos in 3T3-L1 cells. We observed a significant reduction of mitochondrial area and perimeter, which suggested increased mitochondrial fragmentation in shPPA1 cells (Fig. 6H). Given that mitochondrial fragmentation can occur by disturbed mitochondrial fusion and fission (32), we further detected expression of mitochondrial dynamics–related proteins. Accordingly, expression of Drp1, a key mitochondrial fission protein, was increased, while expressions of mitochondrial fusion proteins Mfn1, Mfn2, and Opa1 were decreased, which suggested that mitochondrial dynamics appeared to be shifted toward fission in the absence of PPA1 (Fig. 6G). In further support, disturbed mitochondrial dynamics was also observed in iWAT from PPA1^+/– mice (Fig. 6M). Phosphorylation of Drp1, which can promote mitochondria fission, was significantly increased with PPA1 deletion, while expressions of Opa1 and Mfn2 were sharply decreased. And not surprisingly, disturbed mitochondrial dynamics was recovered upon PPA1 restoration in iWAT (Fig. 6N). Impaired mitochondrial function was also observed in the absence of PPA1. PPA1-silenced cells showed a reduced oxygen consumption rate (OCR) compared with the control group, indicating impaired respiratory capacity (Fig. 6I). In addition, we observed attenuation in the abundance of oxidative phosphorylation complexes in iWAT from PPA1^+/– mice, which was further recovered by PPA1 re-expression (Fig. 6L). Therefore, these data indicate that PPA1 deficiency may result in mitochondrial dysfunction caused by impaired mitochondrial dynamics in adipocytes both in vitro and in vivo.

To evaluate the specific effect of PPA1 on adipocytes in vivo, PPA1 was knock down in fat body, an organ that serves as adipose tissue, liver, and immune system of Drosophila melanogaster. Nurf38, a single Drosophila homolog of PPA1 (33), was depleted by expressing a UAS-RNAi construct under control of the fat body–specific driver Gal4. According to the quantitative PCR results, >70% of Nurf38 was depleted in siNurf38 mutants (Fig. 7C). Compared with WT, PPA1-null larvae (Nurf38) are lean and more transparent, suggesting that the mutant larvae possess less fat tissue (Fig. 7A). Then the fat body of the larvae was isolated and stained with Nile red and Hoechst to evaluate the effect of PPA1 on lipid storage by confocal microscopy. Results revealed that the larvae fat bodies of Nurf38 mutant had smaller and fewer LDs, suggesting that they have reduced lipid storage compared with WT (Fig. 7B). The number of LDs with different diameters was also quantified and showed significant reduction in Nurf38 mutant (Fig. 7D). As a consequence, Nurf38 mutant flies were more sensitive to the starvation challenge (Fig. 7E). Mutant flies died much faster than WT flies when they were placed on 1% agar without nutrients. Moreover, the ultrastructure of mitochondria in fat bodies was examined by transmission electron microscopy, and smaller mitochondria were observed in mutant flies compared with WT (Fig. 7F).

In this study, we explored the effect of PPA1 on obesity-related metabolic disturbances in mice for the first time. PPA1^+/– mice were more susceptible to HFD-induced obesity and insulin resistance, which might be attributed to impaired adipose tissue function. Mechanistically, we demonstrated that PPA1 works as a PPARγ target gene to maintain mitochondrial function in adipocytes. Taken together, our findings establish the important role of PPA1 in adipocyte function and whole-body energy regulation.

Previous studies in our laboratory first reported that PPA1 can protect β-cells against palmitate-induced cell apoptosis as a FoxO1 target gene, suggesting the protecting role of PPA1 in glucose homeostasis (18). Therefore, in this study, PPA1^+/– mice were generated to further explore the detailed role of PPA1 in metabolic syndromes. Heterozygous PPA1^+/– mice were generated and intercrossed in an attempt to obtain homozygous PPA1^–/–mice. However, no PPA1^–/– mice and 282 PPA1^+/– mice were identified among 521 weaned offspring. To investigate this finding, embryos were genotyped at embryonic day 13.5, and no PPA1^–/– embryos were identified. A previous study reported that a defect of soluble PPAs may lead to cell cycle arrest and cell death in fermenting yeast (13). Therefore, we speculated that biallelic deletion of PPA1 might lead to embryonic death, but this hypothesis still needs to be further explored in future studies. Although our previous study suggested an essential role of PPA1 in pancreatic β-cells, the primary phenotype of PPA1^+/– mice is decreased insulin sensitivity and reduced adiposity rather than impaired insulin synthesis and secretion, suggesting that even partial loss of PPA1 could lead to severe impairment of the adipose tissue function. Increased insulin and enlarged pancreatic islet area, which is a distinct feature of the prediabetes insulin-resistant state, was observed in PPA1^+/– mice (Supplementary Fig. 5A–C).

PPi can be generated by several important nucleotide triphosphate-dependent reactions necessary for synthesis of nucleic acid, protein, and lipid. PPi has to be further hydrolyzed to Pi to provide for the synthesis of ATP. Removal of PPi is carried out by PPA, and an excess of PPi was proven to inhibit a majority of the PPi-producing reactions (13). There are two members of PPAs in eukaryotes: cytoplasmic-soluble PPA1 and mitochondrial-located PPA2 (34,35). In mammals, alteration of PPA levels is associated with several illnesses, such as calcium phosphate homeostasis and cancer (15,36). PPA1 has now been found to be increased in several different origins of tumor cells (17,37). Recently, studies reported that PPA2 mutation caused a mitochondrial disease leading to sudden cardiac death in infants and suggested that PPA2 has a greater physiological importance in mitochondrial function than previously recognized (38,39). In the current study, expression of PPA2 was also detected in adipocytes, which excludes the possibility that deletion of PPA1 may also affect PPA2 expression and further lead to impaired mitochondrial dysfunction. As shown in Supplementary Fig. 6, interference of PPA1 has no effect on PPA2 gene expression both in adipocytes and in PPA1^+/– mice.

Given that PPA1 was identified as a negative regulator of cell apoptosis in pancreatic β-cells in our previous study (18), adipocyte apoptosis was also evaluated. However, we did not observe significant cell apoptosis with PPA1 interference in 3T3-L1 cells as well as in adipose tissue (Supplementary Fig. 7). Therefore, our results indicate that the mitochondrial dysfunction caused by PPA1 deficiency is independent of cell apoptosis in adipocytes.

To better understand how PPA1 deficiency affected mitochondrial function, the intracellular localization of PPA1 during cell differentiation was evaluated. However, we did not observe an overlap between endogenous PPA1 staining and the mitochondrial staining (Supplementary Fig. 8), which suggests that interference of PPA1 may indirectly affect mitochondrial biogenesis. Considering that PPi hydrolysis is involved in multiple reactions in cell metabolism, alteration in cell metabolites could be one possible explanation. Therefore, we performed a metabolomics study of cell lysate from shGFP and shPPA1 adipocyte to screen for metabolites involved in mitochondrial dysfunction. Among 1,184 metabolites identified, 226 metabolites were significantly increased and 132 metabolites were markedly reduced (Supplementary Fig. 9A and B) with PPA1 interference. Pathway analysis revealed a significant alteration of amino acid and lipid metabolism (Supplementary Fig. 9C). Some classes of metabolites were reported to closely relate with mitochondrial dysfunction (40); among these, two classes of lipids have raised our interests. One interesting class of lipids is acylcarnitines, which are free fatty acids coupled to carnitine. Interference of PPA1 has markedly increased levels of multiple acylcarnitines (Supplementary Fig. 9D). Elevated concentrations of acylcarnitines were reported to correlate with adipose tissue inflammatory state conditions associated with mitochondrial dysfunction and metabolic syndrome (41). The accumulation of acylcarnitines is suggested to contribute to lipid-induced mitochondrial stress and insulin resistance, while clearing of these lipids improves insulin sensitivity (42). However, a high concentration of acylcarnitines in mitochondria is also suggested as a biomarker of mitochondrial dysfunction since they could represent insufficient β-oxidation (41). Therefore, it is not yet clear whether acylcarnitines are “reflecting or inflicting” mitochondrial dysfunction and insulin resistance (43). Lysophosphatidylcholines (LPCs) comprise another interesting class of lipids that are also significantly increased in adipocytes with PPA1 knockdown (Supplementary Fig. 9D). LPCs are often considered to be cell death effectors and to play an important role in lipotoxicity (44). LPCs have been linked to mitochondrial dysfunction, as the insertion of LPCs in mitochondria membrane promotes its permeabilization (45) and inhibits the activity of mitochondrial COX1 and glycerol-phosphate dehydrogenase (46). Therefore, elevated levels of LPCs and acylcarnitines caused by PPA1 deletion may partially contribute to the mitochondrial dysfunction we observed; however, the detailed mechanism will be further investigated in our future studies.

Although there are dozens of transcription factors known to promote adipose tissue development and function, none is as crucial as PPARγ, which is considered the principal regulator in adipocytes (47). The central role of PPARγ in adipose tissue and glucose metabolism make it a desirable pharmacological target. PPARγ agonists thiazolidinediones have proven robust as insulin sensitizers and work as potent antidiabetes drugs. Therefore, in this study, we evaluated the effect of RGZ on insulin sensitivity in HFD-fed WT and PPA1^+/– mice. As shown in Supplementary Fig. 10A and B, the therapeutic effect of RGZ was attenuated in HFD-fed PPA1^+/– mice, supporting our hypothesis that PPA1 works as an important downstream target of PPARγ in adipocytes.

In addition to WAT, PPARγ also functions as a master regulator of brown adipocytes (48). In our study, decreased brown adipose tissue (BAT) weight was observed in PPA1^+/– mice (Supplementary Fig. 11A), suggesting impaired BAT development with PPA1 deficiency. In addition, the H-E results revealed that the BAT of PPA1^+/– mice displayed giant, white adipocyte–like droplets (Supplementary Fig. 11B). Given the role of BAT in thermogenesis, we exposed mice to a cold environment (4°C) for 6 h in order to active BAT. Rectal temperature was monitored every hour. As shown in Supplementary Fig. 11C and D, PPA1^+/– mice showed lower core body temperature following cold exposure compared with WT mice, which was at least partially caused by impaired BAT function. To blunt BAT activity, HFD-fed mice were housed in thermoneutrality for 8 weeks, and TG accumulation in BAT was further assessed by H-E staining. Increased lipid accumulation was observed in BAT from both WT and PPA1^+/– mice, and thermoneutrality eliminated the difference in TG accumulation between the two groups (Supplementary Fig. 11E). Therefore, the BAT morphology alteration and TG accumulation we observed are partly attributed to the impaired BAT thermogenesis caused by PPA1 deficiency.

In conclusion, results from this study demonstrate, for the first time, essential roles of PPA1 in adipocytes. Even partial loss of PPA1 could impair adipose tissue function and further lead to whole-body metabolic disorders. Results from in vitro studies suggested that deficiency of PPA1 may result in mitochondrial dysfunction through disruption of mitochondrial dynamics and biogenesis in adipocytes. Our data suggest that PPA1 may have a greater physiological importance in metabolic syndrome than previously recognized and may lead to novel therapeutic strategies against human obesity-associated disease.

Acknowledgments. The authors thank all the laboratory members for insightful discussion during the course of this study. X.H. is a fellow at the Collaborative Innovation Center for Cardiovascular Disease Translational Medicine.

Funding. This work was supported by the National Natural Science Foundation of China (81830024, 81870566, 81970709, 81970673) and the Joint Key Project Fund of Southeast University and Nanjing Medical University (2019DN0008).

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

Author Contributions. Y.Y., Y.W., and X.Z. designed and performed most experiments. Y.Y. and H.L. conceived the project and wrote the manuscript. Y.Z. and Y.S. performed in vitro experiments. J.Y. and Y.G. performed in vivo studies and analyzed the data. P.S. provided key expert advice that enabled parts of this study. X.H. contributed to the discussion and critically reviewed and edited the manuscript. All authors revised and approved the final version of the manuscript. X.H. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.