Article Mitochondrial TNAP controls thermogenesis by hydrolysis of phosphocreatine

https://doi.org/10.1038/s41586-021-03533-z Yizhi Sun1,2, Janane F. Rahbani3,4, Mark P. Jedrychowski1,2, Christopher L. Riley1,2, Sara Vidoni1,2, Dina Bogoslavski1, Bo Hu1,2, Phillip A. Dumesic1,2, Xing Zeng1,2, Alex B. Wang1,2, Received: 31 August 2020 Nelson H. Knudsen1,2, Caroline R. Kim1, Anthony Marasciullo1, José L. Millán5, Accepted: 11 April 2021 Edward T. Chouchani1,2, Lawrence Kazak3,4 & Bruce M. Spiegelman1,2 ✉

Published online: xx xx xxxx Check for updates Adaptive thermogenesis has attracted much attention because of its ability to increase systemic energy expenditure and to counter obesity and diabetes1–3. Recent data have indicated that thermogenic fat cells use creatine to stimulate futile substrate cycling, dissipating chemical energy as heat4,5. This model was based on the super-stoichiometric relationship between the amount of creatine added to mitochondria and the quantity of oxygen consumed. Here we provide direct evidence for the molecular basis of this futile creatine cycling activity in mice. Thermogenic fat cells have robust phosphocreatine phosphatase activity, which is attributed to tissue-nonspecifc alkaline phosphatase (TNAP). TNAP hydrolyses phosphocreatine to initiate a futile cycle of creatine dephosphorylation and phosphorylation. Unlike in other cells, TNAP in thermogenic fat cells is localized to the mitochondria, where futile creatine cycling occurs. TNAP expression is powerfully induced when mice are exposed to cold conditions, and its inhibition in isolated mitochondria leads to a loss of futile creatine cycling. In addition, genetic ablation of TNAP in adipocytes reduces whole-body energy expenditure and leads to rapid-onset obesity in mice, with no change in movement or feeding behaviour. These data illustrate the critical role of TNAP as a phosphocreatine phosphatase in the futile creatine cycle.

Adipose tissues have gained considerable interest because of the New pathways of thermogenesis in adipose tissues have been dis- tight linkage between obesity and metabolic disorders such as type covered in the past few years14,15. Previously, a pathway was described 2 diabetes, cardiovascular diseases and many cancers1. Two particu- in which creatine (Cr) and phosphocreatine (PCr) are in a futile cycle, lar types of fat cell, brown and beige adipocytes, are notable for their which dissipates the high energy charge of PCr without performing exceptional ability to oxidize biological fuels and to support adaptive any mechanical or chemical work5. This was named the futile creatine thermogenesis2,3. These two types of thermogenic fat cell utilize futile cycle (FCC). Biochemical, pharmacological and genetic evidence indi- cycles to dissipate chemical energy in the form of heat without perform- cates that this pathway is an important component of adipose tissue ing mechanical or chemical work4. Accordingly, they both serve as a thermogenesis5,16–18. Proteomic and transcriptomic data also suggest physiological defence against hypothermia, obesity and obesity-linked that this pathway is important for thermogenesis in human fat19,20, and metabolic disorders. Notably, the presence of functional thermogenic cultured human brown or beige fat cells show a marked suppression adipocytes in adult humans is now well established6–8. of thermogenic respiration when creatine metabolism is inhibited5. Despite the important metabolic functions of adaptive thermogen- Despite the apparent physiological importance of the FCC, ques- esis, its molecular mechanisms are not fully understood. Although tions concerning its mechanism remain to be answered. PCr is iso- uncoupling protein 1 (UCP1) is the most-recognized thermogenic energetic with ATP and serves both as a reservoir and a shuttle of the effector9,10, it is required for defence of body temperature only on a high-energy phosphate21. In theory, the FCC could involve a direct pure genetic background in mice11,12. Ucp1-knockout mice on a mixed enzymatic hydrolysis of PCr, but such an enzymatic activity has never genetic background can tolerate cold exposure with no loss of core been described in the biomedical literature from cellular sources. In body temperature. Moreover, recent work has demonstrated that even this study, we identify TNAP (encoded by Alpl) as a robust PCr phos- within inbred strains of mice, cold exposure results in severe damage phatase in thermogenic fat. We find that TNAP in thermogenic fat cells to the entire electron transport chain in the mitochondria of brown localizes within mitochondria—a highly atypical subcellular location. fat without UCP113. These findings suggest that at least some part of Chemical and genetic ablation of adipose TNAP activity demonstrates the loss of defence against hypothermia due to the genetic absence that this enzyme has a critical role in supporting the FCC and adaptive of UCP1 must now be reconsidered mechanistically. thermogenesis.

1Department of Cancer Biology, Dana-Farber Cancer Institute, Boston, MA, USA. 2Department of Cell Biology, Harvard Medical School, Boston, MA, USA. 3Goodman Cancer Research Centre, McGill University, Montreal, Quebec, Canada. 4Department of Biochemistry, McGill University, Montreal, Quebec, Canada. 5Sanford Children’s Health Research Center, Sanford Burnham Prebys Medical Discovery Institute, La Jolla, CA, USA. ✉e-mail: [email protected]

Nature | www.nature.com | 1 Article

a PCr Hydrolase Pi

  1+ ² 2 1+ 2 ² P ² H O 2 2– 2 + 2 + 1 1 + HPO4 ² 1 1  2 + 2 2 b c d PCr Sonication 400 Cold BAT Room-temperature BAT )

–1 Heart Buffer Ultracentrifugation 300 control 200 Pi Detergent 100 PCr phosphatase Mitochondrial SEC, IEX activity (nM min extract 0 4 2 0 –2 –4 –5 Mass spectrometry 078 910111213 31P (ppm) SEC elution volume (ml) e PCr phosphatase Protein ID Gene symbol Description Peptides Enrichment f 1,500 sp|Q9CZ13|QCR1_MOUSE Uqcrc1 Cytochrome b-c1 complex subunit 1, mitochondrial 17 17.82 Cold sp|Q62465|VAT1_MOUSE Vat1 Synaptic vesicle membrane protein VAT-1 homologue 6 11.67 Warm sp|Q3UWW6|GA2L3_MOUSE Gas2I3 GAS2-like protein 3 1 11.50 1,000 sp|O35114|SCRB2_MOUSE Scarb2 Lysosome membrane protein 2 24 8.72 sp|O08749|DLDH_MOUSE Dld Dihydrolipoyl dehydrogenase, mitochondrial 25 7.43 sp|P09242|PPBT_MOUSE Alpl Alkaline phosphatase, tissue-nonspeci c isozyme 23 7.19 500 sp|P51912|AAAT_MOUSE Slc1a5 Neutral amino acid transporter B 1 5.56 sp|P06909|CFAH_MOUSE Cfh Complement factor H 47 5.45 (normalized counts) sp|P11438|LAMP1_MOUSE Lamp1 Lysosome-associated membrane glycoprotein 1 29 5.25 Alpl mRNA expression 0 sp|P38647|GRP75_MOUSE Hspa9 Stress-70 protein, mitochondrial 81 4.72 Brown Beige

Fig. 1 | Isolation and identification of TNAP as a cold-inducible PCr errors of nonlinear regression that fits a straight-line model to the initial linear phosphatase from mitochondria of thermogenic fat. a, Chemical equation phase of PCr hydrolysis kinetics measured by 31P NMR over four time points for the hydrolysis of phosphocreatine, with the phosphorus-containing (shown in Source Data). e, The top 10 proteins enriched in the active fraction chemical species indicated at the top. b, Stacked one-dimensional 31P NMR compared with a nearby, less-active fraction from ion-exchange chromatography. spectra showing the hydrolysis of phosphocreatine 6 h after the addition of TNAP (encoded by Alpl) is highlighted in grey. f, UCP1 TRAP (translating mitochondrial protein extract (1.0 mg ml−1) from BAT of cold-acclimated mice. ribosome affinity purification) data showing the cold-sensitivity of Alpl c, Scheme showing the isolation and identification of PCr phosphatase(s) from expression in brown and beige adipocytes. Data were adapted from ref. 23 and the sucrose-gradient-purified mitochondrial preparations obtained from normalized with DESeq241, and are presented as mean ± s.e.m.; n = 5 for cold and mouse tissues. IEX, ion-exchange chromatography. d, The PCr phosphatase warm brown adipocytes, 4 for cold beige adipocytes, and 3 for warm beige activities of different fractions from SEC. Error bars represent the standard adipocytes (all are biologically independent samples).

spectrometry was performed on an active fraction (fraction a) and a PCr phosphatase activity in thermogenic fat nearby, much less active fraction (fraction b). The gene alkaline phos- To determine whether thermogenic fat displays PCr phosphatase activ- phatase, liver form (Alpl), which encodes TNAP, was sevenfold enriched ity, we incubated PCr with mitochondrial protein extracts prepared in the active fraction compared with the less active fraction (Fig. 1e), in from the interscapular brown adipose tissue (BAT) of mice that were agreement with the detected hydrolytic activities. The enrichment of chronically exposed to cold (4 °C for 2 weeks). 31P nuclear magnetic TNAP in the active fractions was validated by western blotting, using resonance (NMR) was used to simultaneously monitor the consump- an antibody that detects endogenous TNAP (Extended Data Fig. 1c). tion of PCr and the production of inorganic phosphate (Pi) (Fig. 1a, b). In addition, the PCr phosphatase activity detected in the total mito- Using this approach, we found that mitochondrial protein extracts chondrial protein extract derived from BAT was diminished by treat- of interscapular BAT and inguinal white adipose tissue (iWAT) (con- ment with SBI-425, a highly specific inhibitor of TNAP22 (Extended Data taining beige adipocytes) exhibited a robust PCr phosphatase activity Fig. 1d). Finally, the cold sensitivity of Alpl expression in adipocytes was (Extended Data Fig. 1a). To further characterize this activity, mitochon- examined using a published adipocyte-specific ribosomal profiling drial protein extracts from BAT were separated into soluble and insoluble dataset23. Figure 1f demonstrates that the expression of Alpl mRNA is fractions by ultracentrifugation, and the insoluble fraction was gently increased in cold compared with warm conditions, both in brown fat solubilized with the non-ionic detergent NP-40. A cold-inducible PCr (31.4-fold) and in beige fat (18.8-fold). phosphatase activity was found in this previously insoluble fraction, Alpl mRNA is broadly expressed in many tissues, especially in the which was further purified by fast protein liquid chromatography liver, kidney and bone24. In bone, TNAP is best known for its function in

(FPLC)-based size-exclusion chromatography (SEC) (Fig. 1c, d). This puri- facilitating the harvest of Pi from many substrates, including inorganic 25,26 fication procedure was performed on mitochondrial protein extracts pyrophosphate (PPi), for the synthesis of mineralized bone matrix . from brown fat of mice that were kept either at room temperature or We next investigated whether purified TNAP protein could hydrolyse at 4 °C. Equivalent fractions were also included from mitochondria iso- PCr in our assays, and found that recombinant human TNAP hydrolyses lated from the hearts of the same mice that were kept at 4 °C. As shown PCr robustly (Extended Data Fig. 1e–g). in Fig. 1d, a distinct peak corresponding to PCr phosphatase activity at an elution volume of 10 ml was observed only for material derived from brown fat mitochondria isolated from cold-acclimated mice. TNAP is localized to mitochondria TNAP is annotated as a cell-surface and secreted enzyme that does not contain an apparent mitochondrial targeting sequence (UniProt: TNAP is the cold-inducible PCr phosphatase P09242). However, Extended Data Figs. 1c and 3e reveal that it is present The active fraction from SEC was then further separated by ion-exchange within our mitochondrial preparations from BAT. We next investigated chromatography. The fraction that showed enzymatic activity eluted the intracellular distribution of TNAP using immunocytochemistry. as a single narrow peak (Extended Data Fig. 1b). Quantitative mass We first ectopically expressed full-length TNAP (Uniprot: P09242-1) in

2 | Nature | www.nature.com a Untransduced AdALPL TNAP TNAP Mitochondria Merge Brown adipocyte Hepatocyte

b Untransduced AdALPL–APEX2 Brown adipocyte

Fig. 2 | TNAP targets mitochondria in brown adipocytes. a, Confocal (red) and anti-HSP60 (green) were used to visualize TNAP and mitochondria, fluorescence images showing the subcellular localization of ectopically respectively. Scale bar, 10 μm. b, Transmission electron microscopy images expressed TNAP in immortalized brown adipocytes and primary hepatocytes. showing the detailed localization of TNAP–APEX2 in primary mature brown The ectopic expression was mediated by adenovirus (AdALPL). The insets show adipocytes. The ectopic expression was mediated by adenovirus a magnified view of the region outlined by the dotted box. Arrows denote (AdALPL-APEX2). Arrows denote the structures of mitochondrial inner selected TNAP signals that co-localize with mitochondrial signals. Anti-TNAP membrane. Scale bars, 500 nm (left, middle), 100 nm (right).

brown adipocytes and observed a substantial amount of this expressed studies27 (Extended Data Fig. 4b). We ectopically expressed TNAP– TNAP in mitochondria, co-localizing with the well-known mitochon- APEX2 in brown adipocytes that stably express a mitochondrial flu- drial protein HSP60 (Fig. 2a). Notably, this mitochondrial localization orescent reporter (Extended Data Fig. 4a), and first confirmed the was essentially absent when we expressed the same TNAP cDNA in mitochondrial localization and peroxidase activity of TNAP–APEX2 non-thermogenic fat cell types (Fig. 2a, Extended Data Fig. 2). by confocal microscopic imaging (Extended Data Fig. 4c). Notably, We next studied the subcellular location of endogenous TNAP in both the mitochondrial localization of TNAP–APEX2 was not observed in wild-type cells and Alpl-knockout cells. Extended Data Fig. 3a shows that hepatocytes (Extended Data Fig. 4c). a large fraction of the fluorescence signals, which correspond to the To enable examination by electron microscopy, brown adipocytes anti-TNAP antibody, co-localize with mitochondria in these brown adi- that expressed TNAP–APEX2 were stained using 3,3′-diaminobenzidine pocytes. In the control cells, genetic depletion of Alpl eliminated nearly (DAB)/H2O2 and fixed with OsO4. As shown in Fig. 2b, we found that all of the signals that were co-localized with mitochondria (Extended TNAP–APEX2 strongly labelled the inner mitochondrial membrane Data Fig. 3a–c). Moreover, we did not detect endogenous TNAP in the and the intermembrane space (IMS), with many dense areas coincident mitochondria of non-thermogenic fat cell types (Extended Data Fig. 3a, with the cristae structures. These images strongly resemble published d). This is in agreement both with our observations of ectopic TNAP staining patterns27,28 shown by an IMS-localized protein fused to APEX and with the extensive literature on this protein. Together, these data or APEX2. Additionally, protease protection assays of BAT mitochon- together provide evidence that TNAP is localized to mitochondria, and dria derived from cold-acclimated mice demonstrated that TNAP that this localization seems to be specific to thermogenic adipocytes, displayed a similar profile as known IMS-localized proteins such as at least within the cell types examined. glycerol-3-phosphate dehydrogenase and cytochrome c (Extended Data Canonical TNAP is tethered to biological membranes by a glyco- Fig. 4d, e), which also suggest that TNAP is likely to reside in the IMS. sylphosphatidylinositol (GPI) anchor that is installed within the lumen of the endoplasmic reticulum (ER). We therefore examined whether this is also the case for mitochondrial TNAP. The insoluble fraction of the Role of TNAP in the FCC and thermogenesis mitochondrial protein extract from BAT was treated with a phospho- Our findings suggest that TNAP might participate in adaptive thermo- lipase that is specific to phosphatidylinositol. As shown in Extended genesis by supporting the FCC. To test this, we silenced Alpl expression Data Fig. 3f, TNAP was readily dissociated from the mitochondrial in primary brown adipocytes by introducing a short hairpin (sh)RNA membranes by this treatment, whereas VDAC1—a mitochondrial trans- and found that this specifically decreased noradrenaline-stimulated membrane protein—remained associated with membranes. These data respiration (Extended Data Fig. 5a–c). To assess the metabolic function suggest that mitochondrial TNAP is indeed tethered to mitochondrial of TNAP in more detail, we used SBI-425 to ablate its enzymatic activity membranes via a GPI-anchor. (Fig. 3a). Mitochondria were prepared from the iWAT fat depot from cold-acclimated mice and creatine-dependent respiration was meas- ured. As shown in Fig. 3b and Extended Data Fig. 6a, SBI-425 treatment Ultrastructural location of TNAP eliminated creatine-stimulated respiration only under limiting ADP con- Because mitochondrial localization of TNAP is highly unusual, we ditions, which indicates that blocking TNAP activity abolishes the FCC5. next prepared a TNAP–APEX2 fusion protein to enable ultrastructural Consistent with this, creatine-stimulated mitochondrial respiration

Nature | www.nature.com | 3 Article

a SBI-425 b c 2 Cr H O P = 0.032 2 + 400 2 20 1 TNAP P = 0.018 Creatine &O 6 1+ PCr ) CKB –1 15 300 P = 1.000 2 Pi 1 10 ATP ADP 200 2 + H IMS 5 Cyt C 100 0 I III IV V ACC OCR (pmol mi n O 0 –5 NADH 2 Matrix ADP:creatine stoichiometry NAD+ H O Vehicle SBI-425 Vehicle SBI-425 H+ H+ H+ 2

ADP+Pi ATP ADP d ef Immortalized brown adipocytes Vehicle 3XHA-EGFP-OMP25 3 SBI-425 3 P = 0.036 P = 0.057 P = 0.968 2 P = 0.351 P = 0.133 1 h Vehicle/SBI-425 2 Phosphocreatine ( P value)

3 min Homogenization 10 P = 0.05 1

1 Fold change 8 min HA immunoprecipitation –log

Metabolite extraction 0 0 for LC–MS –2 –1 012 PCr Cr ATP ADP AMP

log2(fold change SBI-425 vs vehicle)

Fig. 3 | Ablation of TNAP activity abolishes the FCC by hydrolysis of independent samples per group. d, Scheme for the rapid extraction of phosphocreatine. a, Inhibition of FCC by the TNAP-selective inhibitor SBI-425. mitochondrial metabolites from immortalized brown adipocytes for b, Effect of SBI-425 treatment (10 μM) on the creatine-dependent stimulation metabolomic profiling. LC–MS, liquid chromatography coupled with mass of the oxygen consumption rate (OCR) of beige fat-derived mitochondria in the spectrometry. e, Volcano plot of mitochondrial metabolites of cells treated presence of 0.1 mM ADP, as measured by a Seahorse XF24 Extracellular Flux with SBI-425 (10 μM) and with vehicle. f, Changes in the mitochondrial levels of Analyzer; n = 10 independent measurements for the vehicle group, major metabolites involved in the FCC after SBI-425 treatment; n = 6

6 independent measurements for the SBI-425 group supplemented with H2O, biologically independent samples. Data are presented as mean ± s.e.m. and 8 independent measurements for the SBI-425 group supplemented with Statistical significance was calculated by either two-way ANOVA with creatine. c, Calculated ADP:creatine stoichiometry based on mitochondrial Bonferroni’s multiple comparisons test (b) or unpaired Student’s two-sided oxygen consumption during the FCC in the presence of 0.01 mM creatine and t-test (c, e, f). 0.1 mM ADP, as measured by a Clark-type oxygen electrode; n = 3 biologically was not observed in mitochondria derived from mice in which Alpl was Together, these data provide strong evidence that TNAP activity in knocked out in adipocytes (adipo-Alpl knockout mice) (Extended Data adipose tissues contributes to adaptive thermogenesis. Fig. 6a). Notably, TNAP inhibition had no effect on ADP-dependent mitochondrial respiration in the absence of creatine supplementation (Extended Data Fig. 6b). These data effectively exclude the possibility Ablation of TNAP stimulates obesity that TNAP works as an ATPase to affect respiration, despite its high The FCC has been shown to have an important role in diet-induced ther- ATPase activity in vitro29. Creatine also elicited the release of a molar mogenesis16–18. We therefore studied the role of TNAP in whole-body excess of ADP from the mitochondria of brown adipocytes18, an effect energy balance by feeding a high-fat diet (HFD) to adipo-Alpl knock- that was eliminated by SBI-425 treatment (Fig. 3c). These data suggest out mice and their littermate controls. The adipo-Alpl knockout mice that TNAP promotes the FCC in thermogenic adipocytes. showed a markedly faster increase in body weight compared to the con- To determine directly whether the acute inhibition of TNAP was trol mice (Fig. 4e), and also gained more fat mass than their littermate blocking PCr hydrolysis within mitochondria, we used a rapid mito- controls (Fig. 4g). The increase in weight and fat mass in the adipo-Alpl chondrial purification scheme (MITO-Tag) that has recently been devel- knockout mice occurred with no changes in lean mass, food intake oped to enable metabolite measurements30,31 (Fig. 3d). The successful or physical movement (Fig. 4f, h, Extended Data Fig. 9c). Moreover, enrichment of mitochondria by this approach was confirmed by the indirect calorimetry measurements showed that adipo-Alpl knockout increased levels of several mitochondrial metabolites and proteins mice that were fed a HFD had a lower whole-body energy expenditure (Extended Data Fig. 7a, b). Mass spectrometric profiling of the mito- compared with their littermate controls (Extended Data Fig. 9a, b). chondrial metabolites revealed a 65% increase in PCr levels as a result of These data indicate a role of TNAP in protecting mice from diet-induced TNAP inhibition, making it one of the most significantly altered metabo- obesity. We also observed no apparent deficiency of CL 316,243-induced lites (Fig. 3e, f, Supplementary Table 1). As expected, this blockade of thermogenesis in adipo-Alpl knockout mice (Extended Data Fig. 10a); PCr hydrolysis led to an increase (although non-significant) of ATP this might be due to an upregulation of UCP1—as well as other pro- levels. These data are consistent with a role for TNAP as a mitochondrial teins involved in oxidative metabolism—in a probably compensatory PCr phosphatase in thermogenic fat cells. pathway (Extended Data Fig. 10b–d, Supplementary Table 2). This has To examine the role of TNAP on energy expenditure in vivo, mice also been observed when the FCC is inhibited in mice by fat-specific were administered the TNAP inhibitor SBI-425 simultaneous with the knockout of Gatm16 or Ckb18. This upregulation of UCP1 occurs without stimulation of adaptive thermogenesis by injection of CL 316,243, a β3 an increase in mRNA abundance, which suggests that it is governed by adrenergic receptor agonist. Treatment with the TNAP inhibitor caused post-transcriptional regulation (Extended Data Fig. 10e, f). a large (30%) reduction in whole-body energy expenditure elicited by this β3 adrenergic stimulation (Fig. 4a, b); notably, this occurred without changes in basal respiration, physical movement or food intake (Fig. 4b, Discussion Extended Data Fig. 8a). By contrast, the drug-dependent suppression Stoichiometric evidence was previously provided that futile cycling of respiration was completely lost in mice bearing a fat-specific dele- between creatine and phosphocreatine was part of a newly described tion of TNAP (adipo-Alpl knockout) (Fig. 4c, d, Extended Data Fig. 8b). thermogenic pathway5. However, there was no direct demonstration

4 | Nature | www.nature.com a b Vehicle c d Vehicle 3.0 SBI-425 3.5 Vehicle SBI-425 Vehicle P = 0.018 3.0 P = 0.703 3.5 P = 0.018 3.0 SBI-425 P = 0.703

2.5 SBI-425 ) ) ) 3.0 ) 2.5 –1 –1 –1 –1 2.5 2.5 P = 0.966 2.0 2.0 P = 0.566 2.0 1.5 2.0 (ml mi n (ml mi n (ml mi n

(ml min 1.5

1.5 2 2 2 2 1.0 1.5 VO VO VO 1.0 VO 1.0 Wild type Adipo-Alpl KO 0.5 1.0 0.5 0.5 0 0.5 0 –4 –3 –2 –1 0123456 CL 316,243 PBS –4 –3 –2 –1 0123456 CL 316,243 PBS Time (h) Time (h) e f gh 25 WT 50 250 WT 30 P = 0.204 WT WT P = 0.008 KO KO 25 45 KO P = 0.043 200 KO P = 0.855 20 20 40 150 15 P = 0.016 15 35 100 10 10 Fat mass (g) Lean mass (g) Body weight (g) 30 50 P = 0.503 5 5

25 Cumulative food intake (g ) 0 024 6810 0246810 0 0 Week 0Week 4Week 8 Week 8 Time on HFD (weeks) Time on HFD (weeks)

Fig. 4 | Ablation of TNAP in fat represses adaptive thermogenesis and administration (25 mg kg−1 day−1) on CL 316,243 (1.0 mg kg−1 day−1)-stimulated stimulates obesity. a, Indirect calorimetric measurements (VO2, oxygen respiration. The arrow indicates the time of drug administration; n = 4 mice per consumption rate) on wild-type mice kept in metabolic cages at 22 °C, showing group; all mice were pre-treated with CL 316,243 (1.0 mg kg−1 day−1) for 5 days effect of SBI-425 administration (25 mg kg−1 day−1) on CL 316,243 (1.0 mg kg−1 before SBI-425 administration. d, Averaged respiration rates measured in c day−1)-stimulated respiration. The arrow indicates the time of drug over the time span from 1 h to 6 h after drug administration along with the administration; n = 10 mice per group; all mice were pre-treated with CL 316,243 administration of either CL 316,243 or PBS; n = 4 mice per group. e–h, Body (1.0 mg kg−1 day−1) for 5 days before SBI-425 administration. b, Averaged mass (e), cumulative food intake (f), fat mass (g) and lean mass (h) of adipo-Alpl respiration rates measured in a over the time span from 1 h to 6 h after drug knockout mice and their littermate controls measured over several weeks while administration along with the administration of either CL 316,243 or PBS; n = 10 being fed a high-fat diet and maintained at 22 °C; n = 7 mice per group. Data are mice per group. c, Indirect calorimetric measurement on adipo-Alpl knockout presented as mean ± s.e.m. Statistical significance was calculated by either (KO) mice kept in metabolic cages at 22 °C, showing the effect of SBI-425 two-way ANOVA (a, c, e, f) or unpaired Student’s two-sided t-test (b, d, g, h). of phosphocreatine dephosphorylation. Here we determine, for the substantially hydrolysed, as evidenced by the fact that state 3 respira- first time to our knowledge, the existence of an enzyme that hydrolyses tion is not affected by the inhibition of TNAP in the absence of creatine. phosphocreatine in a biological context. This enzyme is the well-studied Why ATP is not a substantial substrate of TNAP is not clear, but it seems phosphatase TNAP, and here we describe its previously unrecognized most likely to be because the effective transfer of ATP might be spatially role in thermogenesis. TNAP in thermogenic fat cells is localized in compartmentalized within the mitochondrial creatine kinase and ADP/ mitochondria, where it has not previously been observed. The cellular ATP translocase35. Finally, the FCC must operate without compromising mechanisms by which TNAP becomes localized in mitochondria in a the energetic requirements of normal cellular functions in thermogenic cell-selective manner are not clear, but probably involve one or more adipocytes. This is in agreement with our mitochondrial metabolomics factors that enable it to diverge from its usual ER–Golgi pathway32. The data, which show that when the FCC is operating, a substantial fraction role of TNAP in adipose thermogenesis is demonstrated by the following of the mitochondrial PCr remains unhydrolysed at steady state. This observations: first, the reduction of oxygen consumption by isolated remaining PCr is likely to be exported from mitochondria and used to thermogenic fat mitochondria and cells when TNAP is eliminated by enable normal cellular processes to proceed. chemical inhibition and by shRNA-mediated knockdown; second, the Previous genetic studies demonstrated the beneficial metabolic reduced energy expenditure in vivo after the chemical inhibition of functions of the FCC in fat cells and on whole-body energy expenditure TNAP—that this drug acts ‘on target’ was demonstrated using mice by altering the supplies of creatine or phosphocreatine5,16,17. These that lacked TNAP in fat cells; third, that mice that have undergone findings suggested that some people, in particular those with obesity, fat-selective ablation of TNAP develop rapid-onset obesity without might benefit from dietary creatine supplementation. Our current any change in food intake or physical movement. It is important to note work elucidates a direct molecular target that participates in the major that the levels of obesity that were observed in mice fed a high-fat diet heat-generating step of the FCC. Whether the mitochondrial TNAP far exceed those that have been reported after loss of UCP111,33,34. It undergoes any cell-selective modifications that render it suscepti- is worth noting that the portion of adrenergic thermogenesis that is ble to drugging in a more specific way remains to be determined. In attributable to FCC has not been fully determined owing to the second- addition to adaptive thermogenesis and anti-obesity activity, it is pos- ary effects of the genetic disruption of TNAP. Nevertheless, we believe sible that the FCC could be involved in pathological thermic events that the reduced energy expenditure in vivo caused by the chemical that are poorly understood—such as hyperthyroidism, burn injury, inhibition of TNAP represents a rough estimation of the contribution anaesthesia-induced hyperthermia and cancer cachexia36–40. These of the FCC; this is in agreement with observations in mice that bear a possibilities are worthy of further investigation. fat-specific deletion of CKB, which interferes with the FCC before the phosphorylation of creatine18. The results of the TNAP–APEX2 electron microscopy study and protease protection assay suggest that TNAP is Online content attached to the inner mitochondrial membrane and resides in the IMS. Any methods, additional references, Nature Research reporting sum- This is of interest because mitochondrial CKB is also likely to reside in the maries, source data, extended data, supplementary information, IMS18. Whether there is a physical association between TNAP and CKB in acknowledgements, peer review information; details of author con- thermogenic adipocytes remains to be determined. Although TNAP is tributions and competing interests; and statements of data and code active on both PCr and ATP, and both are present in the IMS, ATP is not availability are available at https://doi.org/10.1038/s41586-021-03533-z.

Nature | www.nature.com | 5 Article

1. Rosen, E. D. & Spiegelman, B. M. What we talk about when we talk about fat. Cell 156, 23. Roh, H. C. et al. Warming induces significant reprogramming of beige, but not brown, 20–44 (2014). adipocyte cellular identity. Cell Metab. 27, 1121–1137.e5 (2018). 2. Harms, M. & Seale, P. Brown and beige fat: development, function and therapeutic 24. Buchet, R., Millán, J. L. & Magne, D. Multisystemic functions of alkaline phosphatases. potential. Nat. Med. 19, 1252–1263 (2013). Methods Mol. Biol. 1053, 27–51 (2013). 3. Carobbio, S., Guénantin, A. C., Samuelson, I., Bahri, M. & Vidal-Puig, A. Brown and beige 25. Hessle, L. et al. Tissue-nonspecific alkaline phosphatase and plasma cell membrane fat: from molecules to physiology and pathophysiology. Biochim. Biophys. Acta Mol. Cell glycoprotein-1 are central antagonistic regulators of bone mineralization. Proc. Natl Acad. Biol. Lipids 1864, 37–50 (2019). Sci. USA 99, 9445–9449 (2002). 4. Chouchani, E. T., Kazak, L. & Spiegelman, B. M. New advances in adaptive thermogenesis: 26. Anderson, H. C. et al. Impaired calcification around matrix vesicles of growth plate and UCP1 and beyond. Cell Metab. 29, 27–37 (2019). bone in alkaline phosphatase-deficient mice. Am. J. Pathol. 164, 841–847 (2004). 5. Kazak, L. et al. A creatine-driven substrate cycle enhances energy expenditure and 27. Lam, S. S. et al. Directed evolution of APEX2 for electron microscopy and proximity thermogenesis in beige fat. Cell 163, 643–655 (2015). labeling. Nat. Methods 12, 51–54 (2015). 6. Cypess, A. M. et al. Anatomical localization, gene expression profiling and functional 28. Rhee, H. W. et al. Proteomic mapping of mitochondria in living cells via spatially characterization of adult human neck brown fat. Nat. Med. 19, 635–639 (2013). restricted enzymatic tagging. Science 339, 1328–1331 (2013). 7. Cypess, A. M. et al. Activation of human brown adipose tissue by a β3-adrenergic 29. Kiffer-Moreira, T. et al. Catalytic signature of a heat-stable, chimeric human alkaline receptor agonist. Cell Metab. 21, 33–38 (2015). phosphatase with therapeutic potential. PLoS ONE 9, e89374 (2014). 8. Leitner, B. P. et al. Mapping of human brown adipose tissue in lean and obese young men. 30. Chen, W. W., Freinkman, E., Wang, T., Birsoy, K. & Sabatini, D. M. Absolute quantification Proc. Natl Acad. Sci. USA 114, 8649–8654 (2017). of matrix metabolites reveals the dynamics of mitochondrial metabolism. Cell 166, 9. Bachman, E. S. et al. βAR signaling required for diet-induced thermogenesis and obesity 1324–1337.e11 (2016). resistance. Science 297, 843–845 (2002). 31. Chen, W. W., Freinkman, E. & Sabatini, D. M. Rapid immunopurification of mitochondria 10. Zeng, X. et al. Innervation of thermogenic adipose tissue via a calsyntenin 3β–S100b axis. for metabolite profiling and absolute quantification of matrix metabolites. Nat. Protoc. 12, Nature 569, 229–235 (2019). 2215–2231 (2017). 11. Enerbäck, S. et al. Mice lacking mitochondrial uncoupling protein are cold-sensitive but 32. Haarhaus, M., Brandenburg, V., Kalantar-Zadeh, K., Stenvinkel, P. & Magnusson, P. Alkaline not obese. Nature 387, 90–94 (1997). phosphatase: a novel treatment target for cardiovascular disease in CKD. Nat. Rev. 12. Hofmann, W. E., Liu, X., Bearden, C. M., Harper, M. E. & Kozak, L. P. Effects of genetic Nephrol. 13, 429–442 (2017). background on thermoregulation and fatty acid-induced uncoupling of mitochondria in 33. Liu, X. et al. Paradoxical resistance to diet-induced obesity in UCP1-deficient mice. J. Clin. UCP1-deficient mice. J. Biol. Chem. 276, 12460–12465 (2001). Invest. 111, 399–407 (2003). 13. Kazak, L. et al. UCP1 deficiency causes brown fat respiratory chain depletion and 34. Feldmann, H. M., Golozoubova, V., Cannon, B. & Nedergaard, J. UCP1 ablation induces sensitizes mitochondria to calcium overload-induced dysfunction. Proc. Natl Acad. Sci. obesity and abolishes diet-induced thermogenesis in mice exempt from thermal stress USA 114, 7981–7986 (2017). by living at thermoneutrality. Cell Metab. 9, 203–209 (2009). 14. Ikeda, K. et al. UCP1-independent signaling involving SERCA2b-mediated calcium cycling 35. Vendelin, M., Lemba, M. & Saks, V. A. Analysis of functional coupling: mitochondrial regulates beige fat thermogenesis and systemic glucose homeostasis. Nat. Med. 23, creatine kinase and adenine nucleotide translocase. Biophys. J. 87, 696–713 (2004). 1454–1465 (2017). 36. Shulman, G. I., Ladenson, P. W., Wolfe, M. H., Ridgway, E. C. & Wolfe, R. R. Substrate 15. Long, J. Z. et al. The secreted enzyme PM20D1 regulates lipidated amino acid uncouplers cycling between gluconeogenesis and glycolysis in euthyroid, hypothyroid, and of mitochondria. Cell 166, 424–435 (2016). hyperthyroid man. J. Clin. Invest. 76, 757–764 (1985). 16. Kazak, L. et al. Genetic depletion of adipocyte creatine metabolism inhibits diet-induced 37. Wolfe, R. R., Herndon, D. N., Jahoor, F., Miyoshi, H. & Wolfe, M. Effect of severe burn injury thermogenesis and drives obesity. Cell Metab. 26, 660–671.e3 (2017). on substrate cycling by glucose and fatty acids. N. Engl. J. Med. 317, 403–408 (1987). 17. Kazak, L. et al. Ablation of adipocyte creatine transport impairs thermogenesis and 38. Clark, M. G. et al. Accelerated substrate cycling of fructose-6-phosphate in the muscle of causes diet-induced obesity. Nat. Metab. 1, 360–370 (2019). malignant hyperthermic pigs. Nature 245, 99–101 (1973). 18. Rahbani, J. F. et al. Creatine kinase B controls futile creatine cycling in thermogenic fat. 39. Newsholme, E. A. & Crabtree, B. Substrate cycles in metabolic regulation and in heat Nature 590, 480–485 (2021). generation. Biochem. Soc. Symp. 61–109 (1976). 19. Müller, S. et al. Proteomic analysis of human brown adipose tissue reveals utilization of 40. Klein, S. & Wolfe, R. R. Whole-body lipolysis and triglyceride-fatty acid cycling in coupled and uncoupled energy expenditure pathways. Sci. Rep. 6, 30030 (2016). cachectic patients with esophageal cancer. J. Clin. Invest. 86, 1403–1408 (1990). 20. Svensson, P. A. et al. Gene expression in human brown adipose tissue. Int. J. Mol. Med. 27, 41. Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion 227–232 (2011). for RNA-seq data with DESeq2. Genome Biol. 15, 550 (2014). 21. Bessman, S. P. & Carpenter, C. L. The creatine–creatine phosphate energy shuttle. Annu. Rev. Biochem. 54, 831–862 (1985). Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in 22. Pinkerton, A. B. et al. Discovery of 5-((5-chloro-2-methoxyphenyl)sulfonamido) published maps and institutional affiliations. nicotinamide (SBI-425), a potent and orally bioavailable tissue-nonspecific alkaline phosphatase (TNAP) inhibitor. Bioorg. Med. Chem. Lett. 28, 31–34 (2018). © The Author(s), under exclusive licence to Springer Nature Limited 2021

6 | Nature | www.nature.com Methods Phosphatase activity assays Mitochondria isolation and purification 31P NMR spectra were acquired on Bruker 500 MHz Avance III Spec- Mitochondria isolation and fractionation were conducted at 4 °C. trometer equipped with a Smart probe for multi-nuclear detection Mouse tissues were excised, washed with ice-cold SHE buffer (250 mM and SampleJet for automated sample-handling. Each spectrum was sucrose, 10 mM HEPES, 0.1 mM EGTA, pH 7.2), minced to homogeneity, recorded on 0.5 ml of aqueous solution sample in a 5-mm NMR tube and further homogenized with a motorized Teflon pestle in complete at 298 K, with a spectral width of 50 kHz, relaxation delay of 2 s, and −1 SHE buffer (5 ml g , SHE buffer supplemented with 2% essentially fatty scan number of either 64 or 256. Ten per cent D2O was included in each acid-free BSA (Sigma Aldrich) and 1× EDTA-free Protease Inhibitor sample for the external field-frequency lock. TopSpin v.4.0.7 was used

Cocktail (Roche)). Homogenates were passed through two layers of for NMR data processing and analysis. The phosphorus in PCr and Pi cheesecloth and centrifuged at 8,500g for 10 min to remove lipids resonates at −3.2 ppm and 2.2 ppm, respectively. For continuous PCr in the supernatant and to pellet debris, nuclei and mitochondria. phosphatase activity measurement, each reaction was initiated by The pellet was resuspended in complete SHE buffer, transferred to mixing 10 mM PCr with a protein sample in an NMR tube right before a clean tube, and centrifuged at 700g for 10 min to spin down nuclei the recording of spectrum for the first time point. The progressing reac- and cell debris. The supernatant was transferred to a clean tube again tion was kept in the NMR tube held in SampleJet at room temperature and centrifuged at 8,500g for 10 min to pellet crude mitochondria. between measurements for the entire time course. Before the addition The mitochondrial pellet was then washed with complete SHE buffer of PCr for activity measurements, mitochondrial protein extract was twice and either pelleted for long-term storage at −80 °C or resus- passed through a PD-10 desalting column (GE Healthcare) for removal pended to yield a chosen concentration of mitochondrial protein of endogenous metabolites. with an appropriate buffer (10 mg ml−1 in complete SHE buffer unless The enzyme-coupled phosphatase assay was performed using otherwise indicated). EnzChek Phosphate Assay Kit (Molecular Probes) that couples the

For sucrose gradient purification, sucrose was dissolved to a con- generation of Pi to an enzymatic conversion of purine riboside to centration of 1 M and 1.5 M in sucrose gradient buffer (10 mM HEPES, purine that causes a spectrophotometric shift in maximal absorb- 5 mM EDTA, 2 mM DTT, pH 7.8) and gently layered in a 5-ml polyallomer ance from 330 nm to 360 nm. To assay the PCr phosphatase activ- centrifuge tube (first 2 ml of 1.5 M sucrose solution for bottom layer, ity of ion-exchange chromatography fractions, 90 μl of solution and then 1.5 ml of 1.0 M sucrose solution for top layer). A total of 5 mg of from each fraction was mixed with 10 μl of 10× PCr stock solution −1 crude mitochondria resuspension (10 mg ml in complete SHE buffer) (100 mM, dissolved in H2O), and incubated at room temperature for 48 h. was loaded on top of the sucrose solution layers and centrifuged at Endpoint measurements were performed in duplicate by mixing 20 μl 28,000 rpm for 30 min at 4 °C using the SW55Ti rotor. Purer mitochon- of the phosphatase reaction with 80 μl of coupling-reaction com- dria that banded at the interface of the sucrose cushion were carefully ponents to make a final solution containing 0.2 mM purine ribose extracted, washed twice in complete SHE buffer, and stored at −80 °C (MESG), 1 unit of purine nucleoside phosphorylase (PNPase), and in pellet form until further protein extraction for either activity assays 1× reaction buffer (50 mM Tris-HCl, 1 mM MgCl2, 0.01 mM sodium or biochemical fractionations. azide, and pH 7.5). After 10 min of equilibration, absorbance was measured at 360 nm using a FLUOstar Omega plate reader at room Fractionations of mitochondrial protein extract temperature. Sucrose gradient-purified mitochondria pellets were resuspended To derive the Michaelis–Menten curves of TNAP towards different in ice-cold resuspension buffer (100 mM Tris, 0.1 mM EGTA, pH 7.8), substrates, a continuous enzyme-coupled assay was performed in fragmented using a cold H2O-bath ultrasonicator at high power-level duplicate at 37 °C on a reaction mixture containing 0.05 μM of recombi- for 15 min, and centrifuged at 9,000g for 10 min to spin down nant human TNAP (R&D Systems, 2909-AP) and a serial concentration of unfragmented mitochondria. The supernatant was transferred substrate in assay buffer (50 mM Tris-HCl, 1 mM MgCl2, 20 μM ZnCl2, pH to a 5 ml polypropylene centrifuge tube, topped with resuspen- 7.8) supplemented with 0.2 mM MESG and 1 unit of PNPase. Nonlinear sion buffer, and centrifuged at 32,400 rpm for 1 h at 4 °C using a regression of the Michaelis–Menten model to measured activities was SW55Ti rotor. The resultant supernatant that contains the ‘soluble’ performed with GraphPad Prism v.8 for Mac (GraphPad Software). fraction was quantified using a bicinchoninic acid (BCA) assay and re-concentrated to 2.0 mg ml−1 of protein using an Amicon centrifu- Quantitative mass spectrometry gal filter unit (3k MWCO); the pellet that contained the ‘insoluble’ Proteins in selected ion-exchange chromatography fractions were pre- fraction was resuspended with a small volume of solubilization cipitated using 25% ice-cold HPLC-grade trichloroacetic acid overnight buffer (100 mM Tris, 0.1 mM EGTA, 0.5% NP-40, pH 7.8), ultrasoni- at 4 °C and were washed three times with 100% ice-cold HPLC-grade cated until the solution turned clear, quantified using a BCA assay, MeOH to remove detergent. Dried protein precipitates were trypsi- and diluted with solubilization buffer to yield a protein concentra- nized, Tandem mass tag (TMT)-labelled, and subjected to orthogonal tion of 2.0 mg ml−1. Both soluble and insoluble fractions were then bpHrp fractionation as previously described5. The subsequent liquid assayed for PCr phosphatase activity using 31P NMR methods (see chromatography–tandem mass spectrometry (LC–MS/MS) experi- ‘Phosphatase activity assays’). ments, data processing and quantitative analysis were all performed For further fractionation by SEC on FPLC, 1.0 mg of solubilized as previously described5. insoluble fraction was loaded on a Superdex 200 Increase 10/300 GL column (GE Healthcare) pre-equilibrated with solubilization buffer. Gene expression analysis by quantitative PCR with reverse The eluents were collected in 1.0-ml fractions and assayed for PCr transcription phosphatase activity using the 31P NMR assay (see ‘Phosphatase Total RNA was prepared from frozen tissue using TRIzol (Invitrogen) activity assays’). The active fractions were next fractionated by extraction and RNeasy Mini purification kit (QIAGEN), and reverse tran- ion-exchange chromatography on FPLC, using a Mono Q 5/50 GL scribed using HighCapacity cDNA Reverse Transcription kit (Applied column (GE Healthcare). Both the unbound and NaCl-eluted analytes Biosystems). The resultant cDNA was analysed by quantitative PCR. In were collected in 0.5-ml fractions and assayed for PCr phosphatase brief, 10 ng cDNA and 150 nmol of each primer were mixed with SYBR activity using enzyme-coupled phosphatase assay (see ‘Phosphatase GreenERTM qPCR SuperMix (Applied Biosystems). Reactions were per- activity assays’); the activity of the most-active fractions was verified formed in a 384-well format using an ABI PRISM 7900HT real time PCR by 31P NMR. system (Applied Biosystems). Relative mRNA levels were referenced Article to the mRNA level of Rplp0, unless otherwise indicated. A complete dexamethasone, 0.114 μg ml−1 insulin, 1 nM T3 and 125 μM indometha- list of primers and sequences can be found in the following section. cin for 2 days. Cells were re-fed every 2 days with 1 μM rosiglitazone, 0.5 μg ml−1 insulin and 1 mM T3, and were fully differentiated by day Primers 8 post-induction. For immunocytochemistry, preadipocytes were Alpl forward: TCA ACA CCA ATG TAG CCA AGA, reverse: GTA GCT GGC seeded on 12-mm PDL-coverslip (Corning, 354086) coated with 10 μg CCT TAA GGA TTC; Rplp0 forward: GGA GTG ACA TCG TCT TTA AAC ml−1 laminin (Sigma-Aldrich, L2020) and grown until 100% confluency CCC, reverse: TCT GCT CCC ACA ATG AAG CA; Pparg forward: TCG CTG before induced to differentiate. Cells were subjected to adenoviral ATG CAC TGC CTA TG, reverse: GAG AGG TCC ACA GAG CTG ATT; Pgc1a transduction on day 4 and imaging studies on day 8 post-induction. forward: CCC TGC CAT TGT TAA GAC C, reverse: TGC TGC TGT TCC TGT Immortalized brown preadipocytes were prepared using a 3T3 TTT C; Prdm16 forward: CAG CAC GGT GAA GCC ATT C, reverse: GCG immortalization protocol as previously described43. For differentiation, TGC ATC CGC TTG TG; Adipoq forward: GCA CTG GCA AGT TCT ACT immortalized brown preadipocytes were grown to 100% confluency and GCA A, reverse: GTA GGT GAA GAG AAC GGC CTT GT; Ucp1 forward: induced to differentiate with 1 μM rosiglitazone, 0.5 mM IBMX, 5 μM CAC CTT CCC GCT GGA CAC T, reverse: CCC TAG GAC ACC TTT ATA dexamethasone, 0.114 μg ml−1 insulin, 1 nM T3 and 125 μM indometha- CCT AAT GG; Lep forward: ATG TGC TGG AGA CCC CTG TG, reverse: cin for 4 days. Cells were re-fed every 2 days with 1 μM rosiglitazone, TCA GCA TTC AGG GCT AAC ATC C; Ckb forward: GCC TCA CTC AGA 0.5 μg ml−1 insulin and 1 mM T3, and were fully differentiated by day TCG AAA CTC, reverse: GGC ATG TGA GGA TGT AGC CC; Gatm forward: 10 post differentiation. For immunocytochemistry, preadipocytes GAC CTG GTC TTG TGC TCT CC, reverse: GGG ATG ACT GGT GTT GGA were seeded on 12-mm PDL-coverslip (Corning, 354086) coated with GG; Enpp1 forward: GGC CCG AAA TCA GAC AGG AA, reverse: CTG ACA 10 μg ml−1 laminin (Sigma-Aldrich, L2020) and grown until 100% conflu- AAA CCA GCG ACA GC; Phospho1 forward: AAG CAC ATC ATC CAC AGT ency before induced to differentiate. Cells were subjected to adenoviral CCC TC, reverse: TTG GTC TCC AGC TGT CAT CCA G. transduction on day 8 and imaging studies on day 10 post-induction. Stable integration of the mitochondrial reporter 3XHA-EGFP-OMP2530 Western blotting was mediated by retroviral transduction in preadipocytes right before Protein concentration was determined using the BCA assay (Pierce). A the initiation of differentiation. total of 7.5 μg of protein sample was loaded in each lane for western blot Primary mature brown adipocytes were collected and attached to analysis unless otherwise indicated. Protein samples were denatured coverslips as previously described9. Aclar 12-mm coverslips pre-coated in 1× NuPAGE LDS Sample Buffer (Invitrogen) without boiling, resolved with 50 μg ml−1 poly-d-lysine (Sigma-Aldrich, P4707) and 10 μg ml−1 by 4–12% NuPAGE BisTris SDS–PAGE (Invitrogen), and transferred to a laminin (Sigma-Aldrich, L2020) were used. The attached cells were polyvinylidene difluoride (PVDF) membrane. Primary antibody incuba- subjected to adenoviral transduction for 48 h before electron micro- tions were performed in TBS containing 0.05% Tween (TBST) and 5% scopic imaging studies. milk, for either 4 h at room temperature or overnight at 4 °C; second- ary antibody incubations were performed using either horseradish Hepatocyte, myotube, and PTEC cultures peroxidase (HRP)-conjugated or IRDye secondary antibodies in TBST Primary hepatocytes were prepared from perfused livers dissected with 5% milk for 2 h at room temperature, for visualization using Immo- from 8- to 12-week-old male C57BL/6J mice, as previously described44. bilon Crescendo Western HRP substrate (Millipore) or fluorescence Hepatocytes were seeded on PDL coverslips coated with 10 μg ml−1 signal, respectively. A complete list of antibodies can be found in the laminin at a density of 20,000 cm−2, in maintenance medium (DMEM following section. with 0.2% BSA, 2 mM sodium pyruvate, 1× penicillin/streptomycin, 0.1 μM dexamethasone and 1 nM insulin). The attached hepatocytes Antibodies were subjected to adenoviral transduction for 2 days before imaging The following commercially available antibodies were used for western studies. Primary myoblasts were prepared from limb muscle dissected blotting: TNAP (R&D Systems, AF2910), OXPHOS (Abcam, ab110413), from 8- to 12-week-old male C57BL/6J mice as previously described45. UCP1 (Abcam, ab10983), VDAC (Abcam, ab34726), COX1 (Abcam, To initiate fusion, myoblasts were seeded on PDL-coverslips coated ab14705), HSP60 (Cell Signaling, 12165) and VCL (Sigma-Aldrich, with 10 μg ml−1 laminin at a density of 50,000 cm−2 and were induced V9264). The following commercially available antibodies were used the next day by switching to fusion medium (5% horse serum in DMEM, for immunocytochemistry: TNAP (R&D Systems, MAB29092), HSP60 1× PenStrep). The resultant myotubes were subjected to adenoviral (Cell Signaling, 12165), OXPHOS (Abcam, ab110413), Gt-anti-Ms Alexa transduction on day 2 and imaging studies on day 4 post-induction. Fluor 568 (Invitrogen, A-11004), and Gt-anti-Rb Alexa Fluor 647 (Inv- AML12 and PTEC cells were obtained from ATCC and cultured based itrogen, A-21245). on the manufacturer’s instruction.

Treatment with phosphatidylinositol-specific phospholipase-C Confocal fluorescent microscopic imaging and analysis The insoluble fraction of mitochondria (1.0 mg) derived from Cells attached on PDL-coverslips (Corning, 354086) coated with 10 μg BAT of cold-acclimated mice was incubated with 1 unit of ml−1 laminin (Sigma-Aldrich, L2020) were fixed by 4% paraformaldehyde phosphatidylinositol-specific phospholipase-C (Sigma-Aldrich) for (1 volume of 16% paraformaldehyde (Electron Microscopy Sciences) 10 min at 30 °C. The reaction product was centrifuged at 17,000g for diluted with 3 volumes of PBS) for 10 min at room temperature, washed 10 min at 4 °C to remove protein precipitates and subjected to ultra- 3 times with PBS, permeabilized by permeabilization solution (PBS sup- centrifugation at 32,400 rpm for 1 h at 4 °C using a SW55Ti rotor. The plemented with 0.1% Triton X-100) for 10 min at room temperature, pellet was resuspended with solubilization buffer (100 mM Tris, 0.1 mM and blocked by blocking solution (PBS supplemented with 0.1% Triton EGTA, 0.5% NP-40, pH 7.8) and volume-matched to the supernatant. X-100 and 5% FBS) for 10 min at room temperature. For immunofluores- Both the solubilized pellet and the supernatant were then subjected cence staining, anti-TNAP mouse monoclonal antibody (R&D Systems, to western blot analysis. MAB29092) at 1:100 and anti-HSP60 rabbit monoclonal antibody (Cell Signaling, 12165) at 1:500 in blocking solution were used for primary label- Brown adipocyte cultures ling of TNAP and mitochondria, respectively, and anti-mouse Alexa Fluor Primary brown preadipocytes were prepared from stromal vascular 568 and anti-rabbit Alexa Fluor 647 at 1:1,000 (Invitrogen) were used for sec- fraction of interscapular brown adipose tissue as previously described42. ondary labelling. DAPI (3 μM) (BioLegend) was used to stain nuclei. Cover- In general, brown preadipocytes were grown to 100% confluency and slips were mounted in ProLong Diamond Antifade Mountant (Invitrogen). induced to differentiate with 1 μM rosiglitazone, 0.5 mM IBMX, 5 μM Fluorescent images were acquired using MetaMorph v.7.8.13.0 on a Nikon Ti motorized inverted microscope equipped with Yokogawa spinning 5–10 μg mitochondria isolated from the iWAT of cold-acclimated mice disk confocal, and stored and managed using OME Remote Objects were plated in each well of the XF24 V7 cell culture microplate, in a cold 46 (OMERO) . Pearson’s correlation coefficient was calculated using the respiration buffer (125 mM sucrose, 20 mM K-TES at pH 7.2, 2 mM MgCl2, 47 Coloc2 plugin in Fiji on individual cells that were outlined manually. 1 mM EDTA, 4 mM KH2PO4, 4% fatty-acid-free BSA, 10 mM pyruvate, 5 mM malate and 1 mM GDP); and adhered to the bottom by centrifugation APEX2-assisted fluorescence and electron microscopic imaging at 2,000g for 20 min at 4 °C. To inhibit TNAP, adhered mitochondria Immortalized brown adipocytes stably integrated with were treated with 10 μM final concentration of SBI-425 (100 mM stock in 3XHA-EGFP-OMP25 were attached to laminin-coated PDL-coverslips 100% DMSO) in respiration buffer for 8 min at 37 °C, before respiration before initiation of differentiation. On day 6 of differentiation, cells measurement. Creatine was added to a final concentration of 0.01 mM, were subjected to adenoviral transduction to express TNAP–APEX2, and coupled respiration was initiated by addition of 0.1 mM of ADP to and, on day 10 of differentiation, subjected to fluorescent staining using detect the creatine-dependent stimulation. Alexa Fluor 647 Tyramide SuperBoost Kit according to the manufac- To determine the ADP:creatine stoichiometry during the FCC, turer’s instruction (Invitrogen, B40936). Confocal fluorescent imaging mitochondrial respiration was monitored by a Clark-type oxygen was performed as described in the previous section. Immortalized electrode (Rank Brothers) as previously described18. Mitochondria hepatocytes (AML12) integrated with 3XHA-EGFP-OMP25 were attached were isolated from differentiated immortalized brown preadipocytes to PDL-coverslips and subjected to adenoviral transduction the next described above by differential centrifugation using an MSH buffer day to express TNAP–APEX2. Fluorescent staining using Alexa Fluor (210 mM mannitol, 70 mM sucrose, 20 mM HEPES, pH 7.8). For each 647 Tyramide SuperBoost Kit was performed 4 days after transduction. respiration measurement, 0.25 mg mitochondria in respiration buffer

Primary mature brown adipocytes described in ‘Brown adipocyte (125 mM sucrose, 20 mM K-TES at pH 7.2, 2 mM MgCl2, 1 mM EDTA, cultures’ were subjected to adenoviral transduction to express TNAP– 4 mM KH2PO4, 4% fatty acid-free BSA, 10 mM pyruvate, 5 mM malate APEX2. After 48 h, the localization of TNAP–APEX2 was detected by and 1 mM GDP) was used, and final concentrations of 0.01 mM creatine electron microscopy as previously described27. In detail, after 48 h of and 0.1 mM ADP were added to stimulate the FCC. Pilot experiments ectopic expression, cells were fixed by freshly prepared 2% glutaral- were conducted to determine the length of time required to reach state dehyde for 30 min, quenched with 20 mM glycine, washed with PBS, 4 respiration. State 4 respiration was defined as the respiratory rate, and incubated in solution containing 0.7 mg ml−1 diaminobenzidine after ADP addition, that closely matched the respiratory rate before −1 (DAB) and 0.7 mg ml H2O2 in Tris buffer (Sigma, D4293) for 6 h. After ADP addition. Once these parameters were established, the total num- extensive washing, samples were post-fixed with 1% osmium tetroxide, ber of nanomoles of oxygen consumed was measured over this time, dehydrated and embedded in EPON resin. Then, 48 h later, coverslips based on the oxygen consumption rate (measured as nanomoles per were removed and areas containing cells were randomly selected and min). The excess of oxygen consumed by creatine supplementation mounted. Ultrathin sections (70 nm) were cut and examined on a JEOL was used to quantify ADP liberation based on a P/O ratio of 2.7, which 1200EX - 80kV electron microscope. is then used to calculate the ADP:creatine stoichiometry.

Protease protection assay Adenoviral shRNA cloning, construction and use Differential centrifugation-purified mitochondria derived from Complementary DNA oligonucleotides for shRNA generation were cold-acclimated (7 days) mice were incubated with trypsin in pres- annealed and cloned into the pENTRTM/U6 entry vector according ence of varying concentrations of digitonin (see figure legends for to the manufacturer’s instructions (Invitrogen, K4944). Sequences details) for 30 min at room temperature. Mitochondria were treated for shRNA targeting Alpl were: sense, CAC CGC CTG GAT CTC ATC AGT rapidly with 1% SDS and boiled at 95 °C for 5 min to denature proteins ATT TCG AAA AAT ACT GAT GAG ATC CAG GC; antisense, AAA AGC and kept at −20 °C before western blot analysis. CTG GAT CTC ATC AGT ATT TTT CGA AAT ACT GAT GAG ATC CAG GC. Cloned shRNA was shuttled into the pAD/BLOCK-iTTM-DEST vector Adenovirus preparation and use and transfected into 293A cells. Adenovirus was produced according pAd-DEST expression clones were prepared using Gateway recombina- to the manufacturer’s instructions (Invitrogen, K4941), and purified tion cloning technology (Invitrogen). For AdALPL, the open-reading using the Fast Trap Adenovirus Purification and Concentration Kit frame (ORF) of mouse Alpl was amplified from a cDNA clone (Origene, (EMD Millipore) according to the manufacturer’s instructions. Primary MR208392), inserted into pDONR221 vector via a BP recombination adipocytes were transduced with purified adenovirus in the evening reaction, and subcloned into pAd/CMV/V5-DEST vector via an LR of day 1 and then again in the evening at day 3 of differentiation, with recombination reaction. For AdALPL–APEX2, site-directed mutagen- medium replacement the following morning each time. esis was used to insert the APEX2 DNA sequence (Addgene) into the pDONR221–ALPL construct, between nucleotide c1494 and t1495 of the Brown adipocyte respiration ORF of Alpl (the APEX2 peptide will therefore be expressed between Cellular respiration of primary brown adipocytes was measured using amino acid residues A498 and W499 of wild-type TNAP (UniProt: an XF24 Extracellular Flux Analyzer (Seahorse Bioscience). Preadi- P09242)). The resultant pDONR221–ALPL–APEX2 was subcloned into pocytes were plated at 6,000 cells per well of an XF24 V7 cell culture the pAd/CMV/V5-DEST vector via an LR recombination reaction. microplates, and differentiation was induced 12 h later. On day 6 of dif- Adenovirus was produced using the ViraPower Adenoviral Expres- ferentiation, the adipocyte culture medium was changed to respiration −1 sion System (Invitrogen). In brief, pAd-DEST expression clones were medium consisting of DMEM lacking NaHCO3 (Sigma), 1.85 g l NaCl, digested by PacI restriction enzyme (New England Biolabs), gel purified, 3 mg l−1 phenol red, 2% fatty-acid-free BSA, and 1 mM sodium pyruvate at and packed by Lipofectamine 2000 (Invitrogen) to transfect 293A cells. pH 7.4. Uncoupled respiration was determined by the addition of 2.5 μM The cells were then grown for 10–13 days to enable the appearance of oligomycin. Maximal respiration was determined by addition of 0.5 mM an 80% cytopathic effect before collection. The titre numbers were 2,4-dinitrophenol. Rotenone and antimycin (5 μM each) were used to determined by Adeno-X Rapid Titer Kit (Takara Bio), and an multiplicity inhibit complex-1-dependent and complex-3-dependent respiration. of infection of 500 was used to transduce adipocytes. Rapid mitochondria extraction Mitochondria respiration This was performed as per the procedures described in a previ- An XF24 Extracellular Flux Analyzer (Seahorse Bioscience) was used to ous study30. Immortalized brown preadipocytes stably expressing measure mitochondrial respiration as previously described5. In brief, 3XHA-EGFP-OMP25 described in ‘Brown adipocyte cultures’ were Article plated on 6-well plates and grown until 100% confluency before being induced to differentiate. On day 12 of differentiation, cells were treated Body temperature with 10 μM SBI-425 (as described in ‘Mitochondria respiration’) vs Mice were either housed at room temperature or pre-acclimatized at vehicle for 1 h, before rapid immunoprecipitations of 3XHA-tagged thermoneutrality (28–30 °C) for 1 week before being shifted to 4 °C. mitochondria as previously described30 with minor changes. In detail, Body temperature was measured with a microprobe thermometer cells from each well were washed once with 0.5 ml cold PBS, once with (Physitemp, BAT-12) equipped with a rectal probe (Physitemp, RET3).

0.5 ml cold KPBS (136 mM KCl, 10 mM KH2PO4, pH 7.25), collected in 1 ml cold KPBS, transferred to a pre-chilled 2-ml homogenizer glass High-fat feeding vessel (VWR, 89026-386), and disrupted by 20 stokes of a plain plunger Adipo-Alpl knockout mice and age-matched littermate controls were (VWR, 89026-398) on ice. The resultant homogenate was poured into housed at 22 °C and fed a chow diet until 10 weeks of age. Mice were then one 1.5-ml tube and centrifuged at 1,000g for 2 min at 4 °C, and the singly caged and switched to a high-fat diet (Research Diets, D12492) supernatant was collected and loaded onto Pierce anti-HA magnetic containing 60 kcal% fat, 20 kcal% carbohydrate and 20 kcal% protein. beads (Thermo Fisher Scientific) prewashed with cold KPBS. For cell Mice had free access to food and water. Food was changed every week. lysate from one well of a 6-well dish, a 20 μl slurry of anti-HA beads was used. The lysate/anti-HA beads mixture was incubated on a rota- Body composition analysis tor for 3.5 min in cold and pulse-spun to collect any residual liquid The fat mass and lean mass of mice were analysed by Echo MRI (Echo on the lid. The anti-HA beads were then collected on a magnet strip Medical Systems) using the 3-in-1 Echo MRI Composition Analyzer. and the liquid was aspirated. The beads were quickly resuspended with 1 ml cold KPBS, split into 0.8 ml for metabolite extraction and Statistics and reproducibility 0.2 ml for western-blot analysis, and collected again on a magnet strip Data are expressed as mean ± s.e.m. P values were calculated using before the liquid was aspirated. Metabolites were eluted with 55 μl two-tailed Student’s t-test for pairwise comparison of variables, pre-chilled 80% MeOH, and proteins were eluted with 50 μl 1× NuPAGE one-way ANOVA for multiple comparison of variables, and two-way LDS Sample Buffer (Invitrogen) supplemented with 1% BME. The time ANOVA for multiple comparisons involving two independent variables. that each step took was strictly controlled across different purifica- Sample sizes were determined on the basis of previous experiments tions to minimize the variations in IMS metabolite levels introduced using similar methodologies. For all cellular and animal experiments, by the leakage over time across the permeable outer mitochondrial all stated replicates are biological replicates. For SBI-425 treatment membrane during the isolation. The metabolite extraction mixture studies, mice were randomly assigned to treatment groups. For mass was centrifuged at 17,000g for 10 min at 4 °C, and 50 μl of superna- spectrometry analyses, samples were processed in random order and tant was collected, flash-frozen and stored in liquid nitrogen until experimenters were blinded to experimental conditions. All experi- further processing. ments were successfully repeated with similar results at least two or three times. Polar metabolomics profiling The mitochondrial metabolite extraction in 50 μl 80% MeOH was lyo- Data reporting philized using a SpeedVac system (Labconco) for 15 min at 4 °C. The No statistical methods were used to predetermine sample size. For resultant lyophilized samples were resuspended in 12 μl LC–MS grade in vivo experiments in which grouping was based on genotype, mice in water and kept on dry ice until they were subjected to polar metabo- each genotype were randomly picked for the experiments; for in vivo lomics profiling using the AB/SCIEX 5500 QTRAP triple quadrupole experiments in which grouping was based on pharmacological treat- instrument as previously described48. Peak integration was performed ments, mice of the same genotype were randomly allocated to treat- using MultiQuant v.2.0. The whole-cell metabolites were extracted as ment groups. The investigators were not blinded to allocation during previously described48 from fully differentiated immortalized brown experiments and outcome assessment. preadipocytes described in ‘Brown adipocyte cultures’. Reporting summary Mice Further information on research design is available in the Nature All mice that were used for mitochondrial isolation and physiologi- Research Reporting Summary linked to this paper. cal studies were male, C57BL/6J mice of 6–8 weeks of age and were obtained from The Jackson Laboratory unless otherwise indicated. To breed adipo-Alpl knockout mice, Alplfl/fl mice were obtained from Data availability Sanford Burnham Prebys Medical Discovery Institute49 and crossed with The mass spectrometry proteomics data have been deposited to adiponectin-cretg/0 line maintained by Beth Israel Deaconess Medical the ProteomeXchange Consortium via the PRIDE partner repository Center to generate Alplfl/−+adiponectin-cretg/0 mice; these were then (https://www.ebi.ac.uk/pride/) with the dataset identifier PXD025032. backcrossed with Alplfl/fl mice to generate Alplfl/fl+adiponectin-cretg/0 The published adipocyte-specific ribosomal profiling dataset can be (homo adipo-Alpl KO), Alplfl/-+adiponectin-cretg/0 (hetero adipo-Alpl downloaded at https://ars.els-cdn.com/content/image/1-s2.0-S155041 KO), and their littermate control mice. All mice were housed at 22 °C 3118301839-mmc2.xlsx. Full scans for all western blots are provided in under a 12 h light/dark cycle with free access to food and water. Animal the Supplementary Information. All other data are available from the experiments were performed according to procedures approved by the corresponding author on reasonable request. Source data are provided Institutional Animal Care and Use Committee (IACUC) of Beth Israel with this paper. Deaconess Medical Center.

Indirect calorimetry 42. Klein, J. et al. β3-adrenergic stimulation differentially inhibits insulin signaling and The whole-body metabolic rate was measured using the Promethion decreases insulin-induced glucose uptake in brown adipocytes. J. Biol. Chem. 274, 34795–34802 (1999). Metabolic Cage System (Sable Systems). Mice were housed individually 43. Wu, J. et al. Beige adipocytes are a distinct type of thermogenic fat cell in mouse and in metabolic chambers maintained at 22 °C under a 12 h light/dark cycle human. Cell 150, 366–376 (2012). with free access to food and water. Cold exposure or intraperitoneal 44. Sharabi, K. et al. Selective chemical inhibition of PGC-1α gluconeogenic activity −1 ameliorates type 2 diabetes. Cell 169, 148–160.e15 (2017). injection of CL316,243 (Sigma-Aldrich; 1 mg kg ) into mice started at 45. Springer, M. L., Rando, T. A. & Blau, H. M. Gene delivery to muscle. Curr. Protoc. Hum. the indicated times. Genet. 31, 13.4.1–13.4.19 (2002). 46. Allan, C. et al. OMERO: flexible, model-driven data management for experimental Leonard Kahn Foundation Fellow. P.A.D. is supported by a Damon Runyon Cancer Research biology. Nat. Methods 9, 245–253 (2012). Foundation Fellowship. This study was supported by NIDCR grant DE12889 to J.L.M., NIH grant 47. Schindelin, J. et al. Fiji: an open-source platform for biological-image analysis. Nat. DK 123095 to E.T.C., Canadian Institutes of Health Research (CIHR) grant PJT-159529 to L.K. and Methods 9, 676–682 (2012). JPB Foundation 6293803 and NIH grant DK123228 to B.M.S. 48. Yuan, M., Breitkopf, S. B., Yang, X. & Asara, J. M. A positive/negative ion-switching, targeted mass spectrometry-based metabolomics platform for bodily fluids, cells, and Author contributions Y.S. and B.M.S. conceived the study and designed the experiments. Y.S. fresh and fixed tissue. Nat. Protoc. 7, 872–881 (2012). performed experiments and analysed the data. Y.S. and J.F.R. performed cellular and 49. Foster, B. L. et al. Conditional Alpl ablation phenocopies dental defects of mitochondrial respiration experiments and analysed the data. M.P.J. performed mass hypophosphatasia. J. Dent. Res. 96, 81–91 (2017). spectrometry analysis. C.L.R. and S.V. helped with fluorescence imaging studies. S.V. assisted 50. Gettins, P., Metzler, M. & Coleman, J. E. Alkaline phosphatase. 31P NMR probes of the with protease protection assays. D.B. assisted with all experiments. B.H. assisted with animal mechanism. J. Biol. Chem. 260, 2875–2883 (1985). experiments. C.L.R., S.V., P.A.D., X.Z., A.B.W. and N.H.K. helped with animal and/or cellular 51. Subramanian, A. et al. Gene set enrichment analysis: a knowledge-based approach for experiments and data analyses. C.R.K. assisted with CLAMS studies. A.M. assisted with activity interpreting genome-wide expression profiles. Proc. Natl Acad. Sci. USA 102, 15545– measurements and animal experiments. J.L.M. contributed essential genetic and 15550 (2005). pharmacological reagents and discussed data. L.K. and E.T.C. contributed to experimental 52. Mootha, V. K. et al. PGC-1α-responsive genes involved in oxidative phosphorylation are discussions. Y.S. and B.M.S. wrote the manuscript with comments from all authors. All authors coordinately downregulated in human diabetes. Nat. Genet. 34, 267–273 (2003). provided input and reviewed the manuscript.

Competing interests The authors declare no competing interests. Acknowledgements We thank C. J. Rosen for sharing the Alplfl/fl mouse strain with permission of J.L.M.; R. Garrity for help with CLAMS studies; the NMR Core jointly operated by Harvard Additional information Medical School and Dana-Farber Cancer Institute for help with NMR data acquisition; Nikon Supplementary information The online version contains supplementary material available at Imaging Center at Harvard Medical School for help with fluorescence imaging studies; the EM https://doi.org/10.1038/s41586-021-03533-z. Core at Harvard Medical School for APEX2/EM imaging studies; and the Mass Spectrometry Correspondence and requests for materials should be addressed to B.M.S. Facility at Beth Israel Deaconess Medical Center for targeted metabolomics studies. Y.S. and Peer review information Nature thanks the anonymous reviewers for their contribution to the C.L.R. are supported by the American Heart Association postdoctoral fellowship. J.F.R. is peer review of this work. supported by the Charlotte and Leo Karassik Fellowship. B.H. was a Cancer Research Institute/ Reprints and permissions information is available at http://www.nature.com/reprints. Article

Extended Data Fig. 1 | PCr phosphatase activities of thermogenic fat and equivalent fraction prepared from room-temperature housed mice (RT). d, PCr TNAP. a, PCr phosphatase activities of total mitochondrial protein extracts phosphatase activities of total mitochondrial protein extracts from BAT of from different tissues of cold-acclimated mice. Mitochondrial protein extract cold-acclimated mice treated with vehicle or SBI-425 (10 μM), measured by 31P was prepared from tissues excised from 10 mice for BAT or 20 mice for iWAT. NMR. n = 2 technical replicates per group. Data are presented as mean ± s.e.m. Each reaction contains 10 mM of PCr and 0.4 mg ml−1 of mitochondrial protein e, Stacked traces of 31P NMR spectra recorded at indicated time points, extract, except the buffer control. Data are presented as the estimated demonstrating the kinetics of PCr hydrolysis catalysed by recombinant TNAP. parameters ± uncertainties. Uncertainties are represented by the standard The minor peak marked with an asterisk on top is from glycerol-3-phosphate, a errors of nonlinear regression that fits a straight-line model to the initial linear side-product of the phospho-transferase activity of TNAP50, that transfers the phase of PCr hydrolysis kinetics measured by 31P NMR over 11 time points for phosphoryl-group from PCr to glycerol present in the reaction buffer.

BAT and iWAT and 6 time points for the buffer control (shown in Source Data). f, Michaelis constant (Km) curves of hydrolysis of PCr (left) and PPi (right) b, Ion-exchange chromatography of the active fraction of SEC. The PCr catalysed by recombinant TNAP. Activities were measured by the phosphatase activity of each fraction was measured by enzyme-coupled assay. enzyme-coupled assay; n = 2 technical replicates. Data are presented as The activity of the most active fraction was also verified by 31P NMR. Red and mean ± s.e.m. g, Comparison of the Michaelis–Menten parameters blue bars denote the fractions used for isobaric labelling (TMT) and extrapolated from f. Data are presented as the estimated parameters ± quantitative mass spectrometric analysis. c, Western-blot analysis of the active uncertainties. Uncertainties are represented by standard errors derived from SEC fraction prepared from cold-acclimated mice (cold), compared with the the nonlinear regression fit of Michaelis–Menten model to the data in f. Extended Data Fig. 2 | Mitochondrial localization of ectopically expressed show a magnified view of the area outlined by the dotted box. Anti-TNAP and TNAP in non-thermogenic fat cell types. Confocal fluorescence microscopic anti-HSP60 were used to visualize TNAP and mitochondria, respectively. Scale images showing subcellular localization of ectopically expressed TNAP in bars, 10 μm. different cell types. PTEC, kidney proximal tubule epithelial cells. The insets Article

Extended Data Fig. 3 | Mitochondrial localization of endogenous TNAP in (VCL) blot was used as a sample preparation control. d, Confocal fluorescence BAT and non-thermogenic fat cells. a, Confocal fluorescence images showing microscopic images showing subcellular localization of endogenous TNAP in subcellular localization of endogenous TNAP in brown adipocytes (top different cell types. Anti-TNAP and anti-HSP60 were used to visualize TNAP and and middle panels) and hepatocytes (bottom panels). Primary brown mitochondria, respectively. Scale bars, 5 μm. e, Western-blot analysis on TNAP preadipocytes were prepared from Alplfl/fl mice, transduced with either AdGFP and mitochondrial markers in mitochondrial preparations from BAT of (WT) or AdCre (Alpl KO) on day 4 of differentiation, and fixed for imaging on cold-acclimated, wild-type vs adipo-Alpl knockout mice. Blots were processed day 8. Arrows denote selected peri-nuclear areas of TNAP signal that in parallel with samples derived from the same experiment. f, Western-blot co-localize with mitochondria signal. Antibodies for TNAP (red) and HSP60 analysis of the insoluble fraction of mitochondria extract treated with (green) were used to visualize TNAP and mitochondria. Scale bar, 5 μm. phosphatidylinositol-specific phospholipase-C (PI-PLC), followed by b, Pearson’s correlation coefficient (PCC) analysis showing the extent of ultracentrifugation, showing that PLC treatment releases TNAP from co-localization of TNAP with mitochondria in indicated cell types; n = 10 cells membranes. P, pellet; S, supernatant. Mitochondrial preparation was per group; data are presented as mean ± s.e.m.; statistical significance was fragmented by sonication before treatment. Blots were processed in parallel calculated by one-way ANOVA with Bonferroni’s multiple comparisons test. with samples derived from the same experiment. c, Western-blot analysis on TNAP in wild-type versus knockout cells. Vinculin Extended Data Fig. 4 | Proximity-based fluorescent labelling by TNAP– were fixed and treated with Alexa Fluor 647–tyramide/H2O2 for APEX2 and trypsin protection assay on mitochondria from BAT. a, Confocal proximity-based fluorescent labelling facilitated by the peroxidase activity of fluorescence microscopic images of immortalized brown adipocytes showing APEX2. Stably expressed 3XHA-EGFP-OMP25 was used as mitochondria co-localization of the GFP signal from 3XHA-EGFP-OMP25 construct (mGFP, reporter. Scale bars, 10 μm. d, Western blot analysis of the protease protection channel: 488 nm) with different mitochondria markers. Endogenous assay on mitochondria derived from BAT of cold-acclimated mice. TOMM20, antibodies, OxPhos (upper red, channel: 561 nm) and HSP60 (lower red, GPD2 (glycerol-3-phosphate dehydrogenase), Cyt C (cytochrome c) and CS channel: 640 nm), were used to visualize mitochondria. The insets show a (citrate synthase) are shown as markers of outer mitochondrial membrane magnified region of the image outlined by the dotted box. Scale bars, 5 μm. (OMM), intermembrane space (IMS) and mitochondrial matrix. Blots were b, Illustration of how APEX2 reports subcellular localization of TNAP by its processed in parallel with samples derived from the same experiment. peroxidase activity. X indicates either Alexa Fluor 647-conjugated tyramide e, Relative protein abundances in trypsin-digested mitochondria derived from (for confocal microscopy) or 3,3′-diaminobenzidine (for TEM studies). band intensities of intact protein quantified from d; n = 2 technical replicates. c, Confocal fluorescence analysis of immortalized brown adipocytes (top) and Data are presented as mean ± s.e.m. hepatocytes (bottom) ectopically expressing a TNAP–APEX2 construct. Cells Article

Extended Data Fig. 5 | Effect of Alpl silencing on cellular respiration. initiate different respiration states are as follows: stimulated, noradrenaline; a, Quantitative PCR with reverse transcription (qRT–PCR) of differentiated uncoupled, oligomycin; maximum, carbonyl cyanide m-chlorophenyl primary brown preadipocytes treated with shLacZ or shAlpl; n = 3 biologically hydrazone; n = 11 biologically independent samples for the shLacZ group and 18 independent samples per group. b, Western-blot analysis on TNAP in cells biologically independent samples for the shAlpl group. Data are presented as treated with shLacZ or shAlpl. Vinculin (VCL) blot was used as a sample mean ± s.e.m. Statistical significance was calculated by unpaired Student’s processing control. c, Effect of Alpl knockdown (adenoviral shAlpl) on the two-sided t-test. oxygen consumption rate (OCR) of primary brown adipocytes. Treatments to Extended Data Fig. 6 | Effect of TNAP inhibition on the futile creatine cycle. a, Effect of SBI-425 treatment (10 μM) on the OCR of beige fat-derived mitochondria from wild-type vs adipo-Alpl knockout mice in the presence of 0.01 mM creatine and 0.1 mM ADP (limiting ADP) or 1 mM ADP (saturating ADP), as measured by a Seahorse XF24 Extracellular Flux Analyzer; n = 7 independent measurements per group. b, Effect of SBI-425 treatment (10 μM) on OCR of beige fat-derived mitochondria in the presence of 0.1 mM ADP (limiting ADP) or 1 mM ADP (saturating ADP), but in the absence of creatine, as measured by a Seahorse XF24 Extracellular Flux Analyzer; n = 6 independent measurements per group. Data are presented as mean ± s.e.m. Statistical significance was calculated by either two-way ANOVA with Bonferroni’s multiple comparisons test (a) or unpaired Student’s two-sided t-test (b). Article

Extended Data Fig. 7 | Rapid mitochondria purification enriched b, Western-blot analysis on mitochondrial markers in immunoprecipitated mitochondrial metabolites and proteins. a, Relative abundances of citric mitochondria for metabolomics study. Blots were processed in parallel with acid cycle intermediates in mitochondria vs whole-cell metabolomics; n = 6 samples derived from the same experiment. biologically independent samples. Data are presented as mean ± s.e.m. Extended Data Fig. 8 | Movement and food intake of mice upon SBI-425 vehicle; n = 10 mice for wild-type and 4 mice for adipo-Alpl knockout. Data are treatment. a, b, Cumulative movement and food intake of wild-type mice (a) presented as mean ± s.e.m. Statistical significance was calculated using an and adipo-Alpl knockout mice (b) for 24 h after treatment of SBI-425 versus unpaired Student’s two-sided t-test. Article

Extended Data Fig. 9 | Energy expenditure and movement of wild-type and measured in a; n = 7 mice for wild-type and 6 mice for adipo-Alpl knockout. c, adipo-Alpl knockout mice on a HFD. a, Indirect calorimetric measurements Cumulative movement of mice over 24 h; n = 7 mice for wild-type and 6 mice for taken for wild-type and adipo-Alpl knockout mice that had been on a HFD for 4 adipo-Alpl knockout. Data are presented as mean ± s.e.m. Statistical weeks at 22 °C; n = 7 mice for wild-type and 6 mice for adipo-Alpl knockout; the significance was calculated by either two-way ANOVA (a) or unpaired Student’s grey area indicates the dark period. b. Averaged respiration rates over 24 h as two-sided t-test (b, c). Extended Data Fig. 10 | Compensatory thermogenesis in adipo-Alpl 6 mice for adipo-Alpl knockout. Enrichment analysis was performed with GSEA knockout mice. a, Indirect calorimetric measurement taken for wild-type and 4.1.051,52. The hallmark gene sets were surveyed, and oxidative phosphorylation adipo-Alpl knockout mice kept in metabolic cages at 22 °C, showing is the top hit for both BAT and iWAT. Family-wise error rate P value was stimulation of respiration by CL 316,243 administration (1.0 mg kg−1). Arrow presented for statistical significance (number of permutations, 1,000); NES, denotes the time point of drug administration; n = 12 mice for wild-type and 9 normalized enrichment score (enrichment statistics, ‘classic’). e, f, qRT–PCR of mice for adipo-Alpl knockout; all mice were pre-treated with CL 316,243 (1.0 mg BAT (e) and iWAT (f) from wild-type and adipo-Alpl knockout mice treated with kg−1 day−1) for 5 days. b, Western-blot analysis of BAT and iWAT from wild-type or CL 316,243 (1.0 mg kg−1 day−1) for 5 days; n = 4 mice for wild-type and 6 mice for adipo-Alpl knockout mice pre-treated with CL 316,243 (1.0 mg kg−1 day−1) for 5 adipo-Alpl knockout. Data are presented as mean ± s.e.m. Statistical days. c, d, Gene set enrichment plot of quantitative mass spectrometric significance was calculated by either two-way ANOVA (a) or unpaired Student’s analyses of BAT (c) and iWAT (d) from wild-type or adipo-Alpl knockout mice two-sided t-test (e, f). treated with CL 316,243 (1.0 mg kg−1 day−1) for 5 days; n = 4 mice for wild-type and