ENPP2 Contributes to Adipose Tissue Expansion and Insulin Resistance in Diet-Induced Obesity

Until recently, adipose tissue was viewed as a passive energy storage organ, but with the discovery of leptin and the adipose-derived humoral factors now known as “adipokines,” it has become apparent that adipose tissue is an active endocrine organ that is essential for energy homeostasis (1). Moreover, obese adipose tissue secretes various inflammatory cytokines, including interleukin-6 (IL-6) and tumor necrosis factor-α (TNF-α), whose activities are known to contribute to the development of metabolic and cardiovascular diseases (2).

Enpp2, also designated autotaxin, phosphodiesterase I α/autotaxin, and nucleotide pyrophosphatase/phosphodiesterase 2, was originally discovered as an autocrine motility-stimulating factor released from cancer cells (3). ENPP2 catalyzes the conversion of lysophosphatidylcholine to lysophosphatidic acid (LPA), which exerts a variety of biological effects, in part via G-protein–coupled receptors (3,4). In addition, the COOH-terminal noncatalytic domain of ENPP2 also has biological effects independent of LPA (3). Homozygous Enpp2-deficient mice die in utero due to profound vascular defects, but heterozygous Enpp2-deficient (Enpp2^+/−) mice are apparently healthy, with plasma LPA levels about half those in wild-type (WT) mice (5). Enpp2 is reportedly expressed in mouse adipose tissue and 3T3-F442A preadipocytes, and medium conditioned by Enpp2-expressing COS7 cells increased proliferation of 3T3-F442A cells. Recently, Dusaulcy et al. (6) reported that adipocyte-specific Enpp2 knockout (KO) mice fed a high-fat diet showed greater adiposity and less systemic insulin resistance than control mice, with no difference in food intake. By contrast, Federico et al. (7) reported that fat pad weights were higher in mice overexpressing Enpp2, although locomotor activities, thermogenic profiles, and systemic metabolism were unchanged. These apparently contradictory results prompted us to examine the role of Enpp2 in systemic metabolism. Our findings demonstrate that ENPP2 is a key regulator of brown adipose tissue (BAT) function, energy expenditure (EE), and adipose tissue expansion associated with obesity and metabolic conditions.

Enpp2^+/− mice were provided by Shinichi Okudaira and Junken Aoki (Department of Molecular and Cellular Biochemistry, Graduate School of Pharmaceutical Sciences, Tohoku University, Miyagi, Japan) (8). Adipocyte-specific (Fat) Enpp2 KO mice were generated as previously reported (6). Enpp2^F/F mice carrying a conditional Enpp2-deleted allele (encoding for the catalytic site of Enpp2) flanked by two loxP sites were crossed with Fabp4-Cre mice (obtained from The Jackson Laboratory). Enpp2^F/F Fabp4-Cre (Fat-Enpp2 KO) mice were compared with control Enpp2^F/F littermates of the same generation.

Mice selectively overexpressing ENPP2 in their adipocytes were generated as follows. A 5.5-kb cDNA fragment corresponding to positions −5.4k to +62 of the murine Fabp4(aP2) promoter was ligated to a 2.5-kb fragment corresponding to positions +1 to +2,475 of Enpp2 (NM_001136077), which included a bovine growth hormone polyadenylation signal. The linearized 9.6-kb MluI-KasI FABP4-ENPP2 construct was injected into pronuclei of fertilized zygotes from C57BL/6 mice and transferred to pseudopregnant females. Offspring were then screened for genomic integration by PCR, and mice were generated by breeding F1 heterozygous transgenic males to WT females.

The background of all the mice was C57BL/6J. Littermates for Enpp2^+/− (shown as Enpp2^+/+) were also used as controls. All mice were housed under a 12-h light-dark cycle and allowed free access to food.

To examine the changes in metabolic phenotype seen in diet-induced obesity, we divided C57BL/6 mice into two groups and then fed one group a standard chow diet (6% fat; Oriental Yeast Co., Ltd.) and the other a high-fat diet (D12492, 60 kcal % fat; Research Diets) for 10 weeks, beginning when the mice were 7 weeks old. We performed glucose tolerance tests (oral, 1 g/kg, after 16 h of fasting) at age 14 and 17 weeks and insulin tolerance tests (i.p., 1 unit/kg, after 3.5 h of fasting) at 16 weeks to assess glucose intolerance and insulin resistance. Exercise activity, oxygen consumption, and carbon dioxide production were measured using an Oxymax system (Columbus Instruments).

Interscapular BAT was also surgically removed under microscopic observation for analysis. We were able to differentiate BAT from white adipose tissue (WAT) based on its color and confirmed this identification by histological analysis.

Interstitial fluid was sampled from epididymal fat pad using a modification of an earlier method (9). After placing microdialysis catheters inside the epididymal fat pads of anesthetized mice, saline was injected from the catheters at a rate of 2 μL/min. The interstitial fluid was then sampled 60 min after implantation of the catheters.

The body temperatures of the mice were measured using an electronic thermistor equipped with a rectal probe while the mice were exposed to room temperature and 4°C. Body core temperature was measured at selected times up to 3 h.

All experiments were approved by the ethics committee for animal experiments of the University of Tokyo and Jichi Medical University, and strictly adhered to the guidelines for animal experiments.

We isolated stromal vascular (SV) cells using previously described methods (10,11) with some modification. Mice were killed under general anesthesia following systemic heparinization. We then removed epididymal, subcutaneous, mesenteric, and BAT and minced them into small pieces, which were incubated for 20 min in collagenase solution (2 mg/mL of collagenase type 2 [Worthington] in Tyrode buffer) with gentle stirring. The digested tissue was then centrifuged, and the resultant pellet containing the SV fraction was resuspended in PBS and filtered through 70-µm mesh. We then washed the collected cells twice with PBS, incubated them for 10 min in erythrocyte-lysing buffer as previously described (12), and finally resuspended them in PBS supplemented with 3% FBS. Isolated adipocytes or SV fractions were further used for imaging, culture, and flow cytometric analysis.

The isolated cells were labeled with either a monoclonal antibody or an isotype control antibody and analyzed by flow cytometry using a Canto II, Aria (Becton Dickinson), or SP6800 cytometer (Sony), and FlowJo 7.6.5. software (Tomy Digital Biology). We used propidium iodide to exclude dead cells, and adipocyte numbers were determined after collagenase digestion by counting the floating round adipocytes stained with boron-dipyrromethene (BODIPY). Some cells were analyzed after cell sorting using a MoFlo cell sorter (Dako) or Aria cell sorter (Becton Dickinson).

The SV fraction from epididymal adipose tissue was harvested as described above and cultured until confluent. The cells were cultured in a standard adipogenic mixture containing dexamethasone, isobutylmethylxanthine (IBMX), and insulin (13). To examine their differentiation status, cells were harvested and real-time PCR analysis or histochemical study was carried out.

We isolated Pref1⁺ CD34⁺ preadipocytes from epididymal adipose tissue collected from lean 20-week-old C57BL/6 WT, Enpp2^+/−, Fat-Enpp2 KO, and adipocyte-specific Enpp2-overexpressing mice using a MoFlo or Aria cell sorter. Then using flow cytometry, we confirmed that >99.5% of the sorted cells were preadiopocyte cell fractions. Propidium iodide staining was used to exclude dead cells. Some cells were stained with 5 μmol/L CFSE (CellTrace CFSE Cell Proliferation Kit; Invitrogen) for 15 min, after which 2.5 × 10⁴ cells/mL were cultured in DMEM supplemented with 10% FBS plus recombinant murine ENPP2 (10 ng/mL; R&D Systems) and LPA (1 μmol/L; Sigma-Aldrich) for 48 h. Some cells were treated with si-Enpp2, si-CTRL, and si-EDG2 for 24 h prior to the incubation period. After incubation, the cultured cells were harvested and analyzed.

To visualize ectopic fat accumulation, we used in vivo multiphoton microscopy, which is a modification of conventional single photon methods (12). Mice were anesthetized by injection with urethane (1.5 g/kg), after which the skin was removed, and they were secured to the heated stage of an inverted microscope (Eclipse Ti; Nikon, Tokyo, Japan). BODIPY (Invitrogen), fluorescein-labeled isolectin (Vector Laboratories), and Hoechst 33342 (Invitrogen) were injected into the mice to visualize fat accumulations, vessels, and nuclei, respectively. TMRE (tetramethylrhodamine ethyl ester; Invitrogen) dye was used to monitor mitochondrial membrane potential. The tissue was excited at a wavelength of 860 nm using a Ti:Sapphire laser (Vision II; Coherent, Santa Clara, CA), and images were captured using a Nikon A1R MP system equipped with a 40× (N.A. 1.15) water immersion objective lens (Nikon). More than five animals were examined in each group. Images were quantified by observers blinded to the treatment group using NIS-Elements software (Nikon).

To image living adipose tissue ex vivo, we used a previously described method with some modification (14). Adipose tissues were stained with BODIPY (Invitrogen), Hoechst (Invitrogen), and isolectin (Vector Laboratories) (14), after which the cells were imaged using a two-photon microscope (A1R MP; Nikon). This approach enabled us to visualize adipose tissue structure in detail at depths up to >200 μm and to precisely quantify adipocyte cell numbers. Cell diameters and numbers were quantified using NIS-Elements software by observers blinded to the conditions.

For real-time PCR, we homogenized adipose tissue in Trizol (Invitrogen), after which total RNA was purified from the homogenates. A TaqMan fluorogenic reverse transcriptase PCR assay was used to determine relative mRNA levels according to the manufacturer’s instructions.

After obtaining informed consent using an institutional review board–approved protocol, we examined serum samples acquired at periodic health checks. This study was conducted according to the principles outlined in the Declaration of Helsinki, and all protocols were approved by the ethics review committee of Tokyo University School of Medicine. We excluded patients diagnosed with a malignancy, liver disease, chronic kidney disease, ischemic heart disease, or type 2 diabetes. In addition to the screening examination, we used a high-throughput ELISA system to assess levels of ENPP2 in serum collected in the morning, after the patients had fasted for at least 6 h (15). Human total and high-molecular-weight adiponectin levels were assessed using an ELISA kit (Sekisui Medical). We also measured the intimal-medial thickness in the carotid arteries as previously reported; the average of the values measured in the right and left carotid arteries was used as the intimal-medial thickness value (16).

We also acquired subcutaneous adipose tissue from healthy female donors undergoing liposuction of the abdomen or thighs (10). After digesting 1-g samples of each specimen using collagenase, the samples were centrifuged to isolate the SV fractions. Total RNA was then isolated using Trizol (Invitrogen), after which relative mRNA levels were determined using real-time PCR. This study was approved by the ethics committee of the University of Tokyo Hospital.

The results are expressed as means ± SEM. The statistical significance of differences between two groups was assessed using Student t tests. Differences among three groups were evaluated using ANOVA followed by post hoc Bonferroni tests. Differences among more than three groups were evaluated using Tukey-Kramer tests. Correlations were examined using the Pearson correlation coefficient test. Values of P < 0.05 were considered significant. Regression analysis was used to identify independent determinants of ENPP2 and the percentage of the variance that they explained.

To investigate the functions of ENPP2 in metabolism and obesity, we first assessed its expression in mice. We found that ENPP2 is expressed in various mouse tissues, but is particularly high in adipose tissue (Fig. 1A). Moreover, ENPP2 levels were higher in epididymal fat pads than in subcutaneous or brown fat. After separating the adipocyte and SV fractions, we found that levels of Enpp2 expression were higher in adipocytes than in the SV fraction, as was reported previously (Fig. 1B) (17). Among the cell types present in the SV fraction, Pref1⁺ CD34⁺ preadipocytes expressed higher levels of ENPP2 than CD11b⁺ F4/80⁺ macrophages or CD8⁺ T cells (Fig. 1C). However, because the preadipocyte fraction was relatively small (∼5–10%), ENPP2 levels in the whole SV fraction were lower than in the adipocyte fraction. Collectively, these results indicate that preadipocytes and adipocytes are the major cells producing ENPP2 in WAT.

To begin to elucidate the role played by ENPP2 in obesity, we examined the effects of feeding Enpp2^+/− and Enpp2^+/+ mice a high-fat diet. In Enpp2^+/− mice, serum ENPP2 levels were reduced by half (Fig. 1D), as were serum and interstitial LPA levels (Supplementary Fig. 1). Enpp2^+/− mice fed a high-fat diet showed smaller body weight gains and smaller fat pad weights than Enpp2^+/+ mice, although food intake did not differ between the two groups (Fig. 1E–G). Adipocyte numbers in obese epididymal fat pads were smaller in Enpp2^+/− diet-induced obese (DIO) mice than in WT DIO mice (Supplementary Fig. 1). In addition, the results of insulin and oral glucose tolerance tests showed that the systemic insulin resistance induced by a high-fat diet was diminished in Enpp2^+/− mice (Fig. 1H and I). ENPP2 deficiency similarly suppressed adipose mass expansion and ameliorated glucose intolerance and insulin resistance in db/db mice (Supplementary Fig. 2).

We compared the ectopic fat accumulation in heart, skeletal muscle, and liver using two-photon microscopy (Supplementary Fig. 3). This enabled us to confirm the remarkable fat accumulation seen in all three tissues in Enpp2^+/+ DIO mice, as compared with Enpp2^+/+ mice fed a normal diet (ND). Notably, Enpp2 haploinsufficiency did not affect ectopic fat accumulation in any of these tissues in DIO mice, which ruled out a contribution of ectopic fat to the phenotype of Enpp2-deficient mice. Thus, WAT from Enpp2^+/− DIO mice showed less expansion, improved metabolism, and less inflammation than WAT from Enpp2^+/+ DIO mice.

To assess the contribution of ENPP2 produced in adipocytes to the observed phenotype, we selectively deleted the ENPP2 gene from adipocytes by crossing Fabp4-Cre and Enpp2^fl/fl mice (Fig. 2 and Supplementary Fig. 4). This deleted ∼90, 85, and 80% of Enpp2 gene from adipocytes in the epididymal, inguinal, and brown fat pads, respectively (Fig. 2A). Ennp2 gene was also deleted by >90% from Pref1⁺ CD34⁺ preadipocytes in epididymal fat pads (Supplementary Fig. 4). Although it has been reported that Fabp4-Cre may also drive floxed gene deletion in macrophages and other cell types (18), the overall efficiencies of Enpp2 deletion from other cell types were <10% in epididymal fat pads in our models.

As in Enpp2^+/− mice, serum Enpp2 levels were reduced by half in adipocyte-specific Enpp2 KO (Fat-Enpp2 KO) mice, as compared with WT mice (Fig. 2B). In addition, they gained less body weight than control mice when fed a high-fat diet, with no change in food intake (Fig. 2C and D). The weights of the visceral epididymal, femoral subcutaneous, and interscapular brown fat pads were smaller in Fat-Enpp2 KO DIO mice than in control mice (Fig. 2E). The numbers of adipocytes in obese epididymal fat pads were smaller in Fat-Enpp2 KO DIO mice than in WT DIO mice (Supplementary Fig. 4), and Fat-Enpp2 KO mice showed lower serum triglyceride and cholesterol levels.

Induction of the inflammatory cytokines MCP-1 and TNF-α in epididymal fat pads was diminished in Enpp2^+/− DIO mice (Supplementary Fig. 4). Epididymal fat pads from Fat-Enpp2 KO DIO mice contained fewer CD8⁺ T cells, which reportedly amplify adipose inflammation (10), and fewer Pref1⁺ CD34⁺ preadipocytes and Lin⁻ CD29⁺ CD34⁺ Sca1⁺ progenitors than WT DIO mice (Fig. 2F) (19). Following injection of BrdU, the percentage of BrdU⁺ cells among the Pref1⁺ CD34⁺ preadipocytes was smaller in Fat-Enpp2 KO mice, and the size of the apoptotic preadipocyte cell fraction was increased (Fig. 2G and H). Expression in epididymal fat of Lpl and CD36(Fat), which are involved in fatty acid transport, was lower in Fat-Enpp2 KO than WT mice (Supplementary Fig. 4), and serum inflammatory cytokine (IL-6, MCP-1, and TNF-α) levels were also lower in Enpp2^+/− than Enpp2^+/+ DIO mice.

Adipocyte-specific deletion of Enpp2 also ameliorated glucose and insulin intolerance induced by a high-fat diet, with no changes in food intake (Fig. 2D and Fig. 3A and B). Our finding that Enpp2 deficiency reduces fat pad expansion in WAT and BAT and improves systemic metabolism prompted us to examine locomotion and EE. We found that the body weights of mice fed a high-fat diet were increased after 10 weeks but were unchanged after 4 weeks (Fig. 2C). On the other hand, an increase in EE was observed in Fat-Enpp2 KO DIO mice after as little as 4 weeks on the high-fat diet (Fig. 3C–F and Supplementary Fig. 4), and Fat-Enpp2 KO DIO mice exhibited greater spontaneous locomotive activity than WT DIO mice after 10 weeks on a high-fat diet (Fig. 3G and H). This suggests increased EE was at least partially responsible for the smaller body weight gain seen in Enpp2 KO DIO mice.

The expression of Enpp2 in BAT and the contribution of Enpp2 deficiency to improved systemic metabolism prompted us to analyze BAT functionality in these mice. Imaging BAT tissue in WT DIO mice revealed that nearly 30% of adipocytes showed multiple lipid droplets, a morphological feature of functional brown adipocytes, whereas the other adipocytes contained single lipid droplets (Fig. 3I). By contrast, the number of adipocytes containing multiple lipid droplets was increased to ∼70% in Fat-Enpp2 KO DIO mice. We therefore speculated that the number of functional brown adipocytes was increased in Fat-Enpp2 KO DIO mice. In addition, mitochondrial content is increased in BAT from both ND and DIO Fat-Enpp2 KO mice (Supplementary Fig. 5). The thermogenic function of BAT under cold conditions was improved in Fat-Enpp2 KO DIO mice, as compared with WT DIO mice (Fig. 3J). This may reflect the stronger expression of Ucp1, Ppargc1a, Cidea, and Mcpt1 in whole BAT from Fat-Enpp2 KO mice than WT mice (Fig. 3K), as well as the increased BAT functionality and lipid oxidation capacity in Fat-Enpp2 KO mice. The abovementioned difference in locomotion activity was not enough to fully explain the EE difference (20), and the unaltered food intake and thermogenic gene changes indicate that elevated energy dissipation reflecting more functional BAT also contributed to the improved metabolic phenotype.

To further test whether ENPP2 expressed by adipocytes plays a role in metabolic control, we also generated a transgenic mouse line in which Enpp2 was selectively overexpressed in adipocytes (Supplementary Fig. 6). Levels of ENPP2 expression were increased ∼25-, 30-, and 15-fold in epididymal, inguinal, and BAT, respectively (Supplementary Fig. 6). Previous studies showed that the Fabp4(aP2) promoter may also drive transgene expression in macrophages and other cell types (21). When we analyzed Enpp2 levels in macrophages within epididymal adipose tissue, we found that although Enpp2 expression was increased, the level was <5% of that in adipocytes. Serum ENPP2 and LPA levels were increased fivefold in Fabp4-Enpp2 transgenic mice (Supplementary Fig. 6). When fed a high-fat diet, the resultant Fabp4-Enpp2 transgenic DIO mice showed greater body and fat pad weights than WT DIO mice and greater numbers of adipocytes in WAT (Supplementary Fig. 6).

Collectively, the effects of adipocyte-specific deletion and overexpression of Enpp2 indicate that ENPP2 produced by adipocytes plays a key role in the development of adipose tissue obesity and the metabolic dysfunction induced by a high-fat diet.

We then addressed the mechanism by which ENPP2 affects adipose tissue expansion. We hypothesized that ENPP2 positively affects adipocyte hyperplasia, and because adipocyte progenitor cell fractions were reduced in Fat-Enpp2 KO DIO mice (Fig. 2F), we further hypothesized that ENPP2 is important for proliferation of adipocyte progenitors. Consistent with those ideas, recombinant ENPP2 promoted proliferation of 3T3-L1 cells, a line of Pref1⁺ CD34⁺ adipocyte progenitors (Fig. 4A–C). In addition, when subjected to hormonal stimulation, Pref1⁺ CD34⁺ adipocyte progenitors isolated from Enpp2^+/− WAT differentiated into lipid-bearing adipocytes less efficiently than WT preadipocytes (Fig. 4D–F).

We also examined the adipogenic gene expression profiles of isolated Pref1⁺ CD34⁺ preadipocytes in adipocyte-specific Enpp2-deficient mice and found that expression of Cebpa, Cebpb, Cebpd, and Pparg was reduced in preadipocytes from Enpp2-deficient mice (Fig. 4G). In addition, expression of Klf5 and Klf15, two adipogenic transcriptional factors, was significantly reduced in these mice, whereas expression of the antiadipogenic factor Klf2 was increased. Taken together, these results demonstrate that ENPP2 promotes preadipocyte proliferation and differentiation into adipocytes, thereby promoting adipocyte hyperplasia.

ENPP2 may affect preadipocyte function through production of LPA and may also act directly, in an LPA-independent manner, via its noncatalytic COOH-terminal domain (3). Consistent with an earlier report, we found that 5 μmol/L LPA promoted proliferation and differentiation of Pref1⁺ CD34⁺ adipocyte progenitors from epididymal fat pads (22) (Fig. 4H–J). However, knockdown of LPA receptor 1 (EDG2), the main LPA receptor expressed by Pref1⁺ CD34⁺ adipocyte progenitors (Supplementary Fig. 7), only partially inhibited the effects of ENPP2 on the proliferation and differentiation of Pref1⁺ CD34⁺ adipocyte progenitors (Fig. 4A–C). By contrast, EDG2 knockdown almost completely blocked cell proliferation promoted by LPA (Fig. 4H–J). These results suggest that, in addition to LPA production, ENPP2 may affect preadipocyte function through LPA-EDG2–independent pathways.

To elucidate the contribution of Enpp2 deficiency to functional LPA levels, we measured the interstitial concentration of LPA in epididymal fat using microdialysis catheters, as well as the serum levels (Supplementary Figs. 1 and 4). The interstitial levels measured in epididymal fat pad were lower in Fat-Enpp2 KO mice than WT mice. We therefore speculated that ENPP2 contributes to local LPA levels within fat.

One of the striking phenotypes of global Enpp2 haploinsufficiency and adipocyte-specific Enpp2 deletion is the suppression of adipose tissue inflammation. Although this may be primarily due to the reduction in obesity, it is also possible that ENPP2 produced by adipocytes and progenitors modulates inflammatory processes independently of its effects on adipocyte hyperplasia. To begin to address the possible effects of adipocyte-derived ENPP2 on immune cells, we cocultured 3T3-L1 adipocytes and bone marrow–derived macrophages. The coculture increased expression of TNF-α in macrophages (Supplementary Fig. 8), but this effect was diminished by siRNA-mediated knockdown of Enpp2 in the 3T3-L1 cells. In addition, recombinant ENPP2 increased TNF-α expression in macrophages. Recombinant ENPP2 also increased expression of CD44 and interferon-γ in CD8⁺ T cells in epididymal fat pads. These results indicate that ENPP2 from adipocytes contributes to immune cell function within epididymal fat pads, which is associated with adipose tissue inflammation in obesity.

To gain further insight into the role of ENPP2 in the development of obesity, we analyzed the regulation of Enpp2 expression in obese adipose tissue. Despite the fact that ENPP2 promotes obesity, the level of Enpp2 expression was lower in both the SV and adipocyte fractions of WAT from 20-week-old DIO mice than from lean mice (Fig. 1B). This seeming discrepancy between the obesity-promoting function of ENPP2 and its downregulation in obese adipose tissue prompted us to further analyze Enpp2 expression during adipocyte differentiation and hypertrophy. We first analyzed Enpp2 expression in Pref1⁺ CD34⁺ adipocyte progenitor cells. Although Enpp2 expression was increased during differentiation of Pref1⁺ CD34⁺ adipocyte progenitor cells, the levels peaked on day 6 and declined thereafter (Fig. 4K–M), suggesting Enpp2 is upregulated in differentiating adipocytes but downregulated in hypertrophied adipocytes.

To further elucidate the role of ENPP2 in humans, we also measured ENPP2 expression in subcutaneous fat from human subjects. It was previously reported (23) that EMR1 expression correlated positively with BMI (Supplementary Fig. 9), and we observed that ENPP2 expression was reduced in obese subjects (BMI >25.0). In addition, we used an ELISA-based, high-throughput system to assay serum ENPP2 levels in large numbers of human samples (15). The results showed that serum ENPP2 levels are reduced in obese subjects, and multivariate analysis indicated ENPP2 to be an independent parameter associated with BMI (Supplementary Tables 1 and 2). Thus, the ENPP2 levels are also reduced in obese human subjects.

In the current study, we showed that ENPP2 contributes to the metabolic phenotype associated with obesity. We speculated that ENPP2 directly modulates the function of both BAT and WAT. EE and expression of Ucp1 and Ppargc1a were higher in Fat-Enpp2 KO mice than WT mice (Fig. 3). In addition, microscopic visualization showed that the numbers of multiple lipid droplet–containing adipocytes and mitochondrial membrane potential were increased in both mice. These results suggest that while BAT mass was smaller in Fat-Enpp2 KO than WT DIO mice, the tissue remained functionally active. We therefore speculate that energy dissipation contributes to the improved metabolic profile and reduced weight gains seen in Fat-Enpp2 KO mice fed a high-fat diet.

Enpp2 gene manipulation also affected the immune cell populations in adipose tissue and the tissue inflammation (Fig. 2F), and our finding that ENPP2 mediates proinflammatory interaction between 3T3-L1 adipocytes and macrophages suggests ENPP2 modulates immune cell function (Supplementary Fig. 8). Future studies will be needed to address the precise mechanisms by which ENPP2 contributes to the regulation of adipose tissue inflammation. Although we found no significant differences between the fibroblast fractions in Enpp2^+/− and Enpp2^+/+ mice, or between their adipose Col1a1 or Col6a1 expression, a contribution of fibrosis within fat pads to the phenotype of Enpp2-deficient mice could not fully be excluded (Supplementary Fig. 1).

It was revealed that ENPP2 has a functionally active COOH-terminal domain (i.e., the MORFO2 domain) that functions independently of ENPP2 catalytic activity and LPA production (3,24). In fact, LPA activity in preadipocyte proliferation and differentiation of Pref1⁺ CD34⁺ adipocyte progenitors was only partially inhibited by EDG2 knockdown (Fig. 4). However, the role of the MORFO2 domain in preadipocytes remains to be clarified.

Despite our finding that ENPP2 is important to adipose tissue obesity, its expression was diminished within obese adipose tissue and hypertrophied adipocytes (Fig. 1C and Fig. 4). On the other hand, it has also been reported that ENPP2 expression is elevated in adipose tissue from diabetic db/db mice and that DIO mice show no significant changes in ENPP2 expression (17). What’s more, earlier studies produced contradictory results regarding BAT mass, locomotor activity, thermogenic profiles, and systemic metabolism (6,7). The precise mechanisms responsible for these discrepancies between the results of the current study and those earlier ones are not immediately clear, but there were several experimental differences. First, all of our mouse models had a C57BL/6 genetic background. In study by Dusaulcy et al. (6), it appears the mice had a mixed genetic background (FVB and C57BL/6). This difference in genetic background could reportedly contribute to the difference in adipose phenotype between obese mice (25). In addition, we found that the distributions of immune cells in epididymal fat pads and BAT significantly differed between these two mouse strains (data not shown), and we speculate that the genetic differences also influenced the effects of Enpp2 deficiency. Second, in Dusaulcy et al. (6), the mice were fed a high-fat diet for 13 weeks, starting when they were 10 weeks old. We started the high-fat diet when the mice were 7 weeks old because we found that early initiation of the diet was necessary to activate adipogenesis. In addition, levels of Enpp2 expression in the SV and preadipocyte fractions peaked at 8 weeks of age and declined thereafter (data not shown). In our experiments, therefore, the mice were started on a high-fat diet while Enpp2 levels were high, whereas in Dusaulcy’s study, the high-fat diet was started after Enpp2 levels had already declined to baseline. This difference in the timing of the high-fat diet may have led to the apparent difference in phenotypes. It is also possible that environmental factors affected the phenotypes. In fact, our examination showed that the effects of Enpp2 deficiency on metabolic profiles depend on the timing of experiments during DIO (Supplementary Fig. 4H).

In humans, we found that serum ENPP2 levels correlated negatively with BMI and that levels of ENPP2 expression were reduced in subcutaneous fat from obese subjects (Supplementary Fig. 9). Earlier studies found that increases in adipocyte size correlated with serum insulin concentrations, insulin resistance, and increased risk of developing type 2 diabetes (26,27). Although in the current study we did not directly measure adipocyte size in human subjects, it is likely that levels of ENPP2 expression in obese subjects were negatively related to adipocyte size. Another possibility is that the progression of diabetes affects ENPP2 expression in adipose tissue. In fact, our preliminary analysis showed that there was a tendency toward higher serum ENPP2 levels in diabetic human subjects (data not shown).

In conclusion, our findings suggest that ENPP2 is a key mediator of adipose tissue obesity and is involved in the systemic regulation of BAT metabolism, and could be a useful therapeutic target for the treatment of metabolic disorders.

Acknowledgments. The authors thank Dr. Kotaro Yoshimura (Department of Plastic Surgery, Graduate School of Medicine, University of Tokyo) for human sample preparations and Dr. Hironori Waki (Department of Metabolic Diseases, University of Tokyo) for insightful comments on the manuscript. The authors also thank M. Tajima, C. Yoshinaga, X. Yingda, T. Hirabayashi (Department of Cardiovascular Medicine, University of Tokyo), C. Nakamikawa, and M. Ito (Center for Molecular Medicine, Jichi Medical University) for their excellent technical help.

Funding. This study was supported by research fellowships from the Japan Society for the Promotion of Science Research Fellowships for Young Scientists (S.N.); the Funding Program for World-Leading Innovative R&D on Science and Technology (FIRST Program) (S.N. and R.N.); Grants-in-Aid for Scientific Research (R.N.); grants from the Translational Systems Biology and Medicine Initiative (T.L. and R.N.) and the Global Centers of Excellence Program (T.K. and R.N.) from the Ministry of Education, Culture, Sports, Science and Technology of Japan; and a research grant from the National Institute of Biomedical Innovation (R.N.).

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

Author Contributions. S.N. designed and performed experiments, analyzed data, and wrote the manuscript. M.N. performed experiments and reviewed the manuscript. S.O., T.O., R.O., K.N., and K.I. performed experiments. J.A. and Y.Y. performed experiments and contributed to discussion. H.Y., K.E., K.U., N.H., T.K., and I.K. contributed to discussion. R.N. directed this study. S.N. and R.N. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.