Angiocrine polyamine production regulates adiposity

Reciprocal interactions between endothelial cells (ECs) and adipocytes are fundamental to maintain white adipose tissue (WAT) homeostasis, as illustrated by the activation of angiogenesis upon WAT expansion, a process that is impaired in obesity. However, the molecular mechanisms underlying the crosstalk between ECs and adipocytes remain poorly understood. Here, we show that local production of polyamines in ECs stimulates adipocyte lipolysis and regulates WAT homeostasis in mice. We promote enhanced cell-autonomous angiogenesis by deleting Pten in the murine endothelium. Endothelial Pten loss leads to a WAT-selective phenotype, characterized by reduced body weight and adiposity in pathophysiological conditions. This phenotype stems from enhanced fatty acid β-oxidation in ECs concomitant with a paracrine lipolytic action on adipocytes, accounting for reduced adiposity. Combined analysis of murine models, isolated ECs and human specimens reveals that WAT lipolysis is mediated by mTORC1-dependent production of polyamines by ECs. Our results indicate that angiocrine metabolic signals are important for WAT homeostasis and organismal metabolism. Endothelial cells in white adipose tissue are shown to produce polyamines, which regulate adipocyte lipolysis, thus demonstrating how local angiocrine signals contribute to healthy adipose tissue homeostasis.

3,4,5-trisphosphate (PtdIns(3,4,5)P 3 , also known as PIP 3 ). Through the generation of PIP 3 , the PI3K/PTEN signaling pathway controls a plethora of cellular functions, including growth, proliferation, migration, metabolism and vesicular trafficking 8 . ECs are exquisitely sensitive to PI3K fluctuations, with both too much and too little PI3K signaling resulting in defects in vascular development 11,12 . Analysis at single-cell resolution has shown that genetic manipulation of the PI3K/PTEN pathway primarily impacts cell proliferation 11,13,14 . and Pten iΔEC mice (n = 9) after postnatal induction of Cre. c, Representative macroscopic and microscopic images of eWAT from control and Pten iΔEC mice at 5 and 12 weeks of age. d, eWAT pad weight in control and Pten iΔEC mice at 5 (control n = 5; Pten iΔEC n = 4) and 12 (control n = 8; Pten iΔEC n = 7) weeks of age. e, Quantification of adipocyte area of eWAT from control and Pten iΔEC mice at 5 (control n = 5; Pten iΔEC n = 6) and 12 (n = 7 each group) weeks of age. f, Representative macroscopic and microscopic images of iWAT from control and Pten iΔEC mice at 5 and 12 weeks of age. g, iWAT pad weight in control and Pten iΔEC mice at 5 (control n = 5; Pten iΔEC n = 4) and 12 (control n = 8; Pten iΔEC n = 7) weeks of age. h, Quantification of adipocyte area in iWAT from control and Pten iΔEC mice at 5 (control n = 5;, Pten iΔEC n = 6) and 12 (control n = 10; Pten iΔEC n = 8) weeks of age. i, Representative hematoxylin and eosin staining of muscle and liver sections from control and Pten iΔEC mice at 12 weeks of age. j, Masson trichrome fibrosis staining in eWAT and iWAT sections from control and Pten iΔEC mice at 12 weeks of age. Data represent mean ± s.e.m. Samples represent biological replicates. Statistical analysis was performed by the two-sided Mann-Whitney U-test (d,e,g,h) and two-way analysis of variance (ANOVA) with Bonferroni correction (b).
Here we show that ECs, through the release of specific angiokines, regulate the function of WAT with systemic metabolic implications. Forced activation of the PI3K pathway in ECs results in increased proliferation of the vessel compartment in the adipose tissue and reduced adiposity. The molecular connection between the endothelial function and adipocyte biology is underscored by a metabolite-based angiocrine communication, a process mediated by polyamines. We provide evidence showing that this unprecedented metabolic axis is relevant in pathophysiological conditions, such as obesity.

Results
Endothelial Pten deletion elicits a lean phenotype. To elucidate endothelium-regulated systemic metabolic processes, we developed a model of enhanced EC-autonomous angiogenesis. Given that EC proliferation is primarily governed by PI3K signaling 11,13,14 , we studied the systemic consequences of sustained PI3K activity in the endothelium by Pten loss 14 . We bred Pten flox/flox mice 15 with PdgfbiCreER transgenic mice 16 (hereafter referred to as Pten iΔEC ), which enables the deletion of Pten in ECs in a tamoxifen-inducible manner. We administered 4-hydroxytamoxifen (4-OHT) postnatally to mice carrying conditional Pten knockout alleles alone or in combination with PdgfbiCreER. To confirm Pten deletion selectively in the endothelium, we took advantage of the Ribotag model 17 . This strategy allows for the analysis of actively translated messenger RNAs in a cell-specific manner in vivo, when Cre recombinase induces the expression of hemagglutinin A (HA)-tagged ribosomal protein Rpl22. We crossed these mice with PdgfbiCreER (hereafter referred to as Ribotag iHAEC ) and demonstrated that only the endothelium was targeted in all tissues analyzed (Extended Data Fig. 1a,b). We could confirm that the deletion of Pten in the endothelium occurred in all tissues studied in Pten iΔEC -Ribotag iHAEC mice (Extended Data Fig. 1c). Also, we validated the efficient depletion of Pten in isolated ECs from several tissues of Pten iΔEC mice (Extended Data Fig. 1d-g).
Next, we evaluated the phenotypic consequences of endothelialspecific Pten deletion in adult mice. Endothelial Pten loss was associated with a remarkable reduction in body weight and adiposity compared to control littermates, thus resulting in a lean phenotype (Fig. 1a,b). We found a selective reduction in WAT mass and adipocyte area, including both epididymal (eWAT) and inguinal (iWAT) depots ( Fig. 1b-h and Extended Data Fig. 2a), a phenotype that was robust by 10 weeks of age. Nuclear magnetic resonance analysis further confirmed the reduction in overall fat mass in adult Pten iΔEC mice (Extended Data Fig. 2b,c). Detailed histopathological and molecular characterization did not reveal ectopic lipid deposition in muscle and liver (Fig. 1i), WAT fibrosis ( Fig. 1j and Extended Data Fig. 2d,e), altered adipocyte differentiation or cell death (Extended Data Fig. 2f-j). Of note, in line with the reduction in adipose tissue mass, the circulating levels of leptin (an adipose-derived adipokine) were markedly reduced in mutant mice (Extended Data Fig. 2k).
To explore the impact of loss of endothelial Pten on the vasculature, we performed whole-mount or OCT-embedded section staining with isolectin B4 (IB4) or anti-CD31, two well-known markers of ECs, in several tissues. The vast majority of tissues analyzed in Pten iΔEC mice remained phenotypically unaffected at 12 weeks of age (Extended Data Fig. 3a,b) despite the deletion of this gene (Extended Data Fig. 1c). However, we observed vascular hyper plasia in the Pten iΔEC WAT, based on a progressive increase of vascular content and EC proliferation, as illustrated by enhanced IB4, ERG (EC-specific nuclear marker) and Ki67 staining ( Fig. 2a-h). The vascular phenotype was already apparent by 5 weeks of age ( Fig. 2a-h), a time point at which adiposity remained largely unaltered (Fig. 1d,g), indicating that vessel growth precedes the reduction in adiposity. The histological data were largely consistent with the molecular evaluation of endothelial markers using quantitative real-time PCR (Extended Data Fig. 3c). Of note, we could not observe vascular endothelial growth factor (VEGF) signaling alterations in WAT by means of VegfA mRNA abundance, thus suggesting no major role of this growth factor in the enhanced vascularization of Pten iΔEC mice (Extended Data Fig. 3d). The enhanced vascular content in Pten iΔEC WAT was not associated with aberrant leakiness (Extended Data Fig. 3e). Due to the regulatory and functional analogies with WAT, we performed a detailed analysis of brown adipose tissue (BAT). Pten iΔEC BAT showed a mild increase in the vascular content compared to control littermates (Extended Data Fig. 3a,b), although the vascular phenotype was less prominent than in WAT. In line with this, we could not observe alterations in the morphology of brown adipocytes (Extended Data Fig. 3f).
To validate the contribution of endothelial-specific Pten deletion on vessel density and adiposity, we carried out two complementary approaches. First, we ruled out developmental effects associated with early activation of Cre by studying the consequences of Pten deletion in adult mice. Administration of tamoxifen at 8 weeks of age in Pten iΔEC mice recapitulated the phenotype in vascularity and adiposity (Extended Data Fig. 4a-h). We also took advantage of a second inducible Cre delivery system targeted to the endothelium, under the control of the VE-Cadherin promoter 18 ; the so-called Cdh5-CreER T2 . Administration of 4-OHT to Cdh5-CreER T2 postnatal mice resulted in a similar phenotype to that of the PdgfbiCreER model, albeit milder (Extended Data Fig. 4i-p). Together, these results reveal an adipose tissue-specific contribution of endothelial Pten that enables the control of adiposity.
Endothelial Pten loss protects from obesity. Excessive adiposity is a hallmark of obesity and is associated with suboptimal WAT vascularization 2,3,19-21 . Thus, we interrogated whether obesity-associated reduced vascularization in experimental systems and human specimens could be correlated with alterations in PTEN levels. The elevated adiposity elicited by high-fat diet (HFD) feeding in mice was associated with reduced vascularization and higher expression of Pten in the adipose endothelium (Extended Data Fig. 5a-c). We confirmed that reduced vascularization and elevated expression of PTEN is also observed in surgical specimens of visceral adipose tissue (VAT) from individuals with obesity (Extended Data Fig. 5d-f and Supplementary Table 4).
Given that induction of angiogenesis in WAT reportedly ameliorates obesity-associated metabolic complications [22][23][24][25][26] , we predicted that our mutant mouse model would be protected from obesity. To address this, we first challenged control and Pten iΔEC mice with HFD. Loss of Pten in this context led to reduced body weight, WAT mass and adipocyte area (Fig. 3a-d), with no significant changes in other tissues (Extended Data Fig. 6a). Similar to mice fed with chow diet, Pten iΔEC mice showed higher vascularization in WAT upon exposure to HFD (Fig. 3e,f). Moreover, Pten deletion in ECs  (a) and iWAT (c) wholemount staining with IB4 (red) from control and Pten iΔEC mice at 5 (left) and 12 (right) weeks of age. b,d, Quantification of IB4-positive area in eWAT (b) and iWAT (d) from control and Pten iΔEC mice at 5 (n = 15 per genotype) and 12 (control n = 8; Pten iΔEC n = 6) weeks of age. e,g, Representative images of flat-mounted eWAT (e) and iWAT (g) from 5-week-old control and Pten iΔEC mice stained with ERG (green), Ki67 (magenta) and IB4 (white). Yellow arrows indicate double-positive proliferating cells. f,h, Quantification of proliferating ECs (ERG and Ki67 double-positive cells) in images from flat-mounted eWAT (f) and iWAT (h) from control and 5-week-old Pten iΔEC mice (control n = 5; Pten iΔEC n = 4). Data represent mean ± s.e.m. Samples represent biological replicates. Statistical analysis was performed by the two-sided Mann-Whitney U-test (b,d,f,h).    Pten iΔEC n = 9). k, Quantification of adipocyte area in eWAT and iWAT (control n = 10; Pten iΔEC n = 8). l, Representative images of IB4-stained blood vessels (red) in flat-mounted eWAT and iWAT. m, Quantification of IB4-positive area in eWAT and iWAT from control and Pten iΔEC mice exposed to HFD for 34 weeks (control n = 8; Pten iΔEC n = 7). n, GTT performed in control (n = 10) and Pten iΔEC (n = 9) mice fed with HFD for 28 weeks. Bars to the right show AUC quantification of blood glucose monitoring during GTT (control n = 10; Pten iΔEC n = 9). Data represent mean ± s.e.m. Samples represent biological replicates. Statistical analysis was performed by the two-sided Mann-Whitney U-test (c,d,f,j,k,m) and two-way ANOVA with Bonferroni correction (a,g,h,n).
resulted in improved glucose metabolism in mice fed with HFD for 10 weeks (Fig. 3g and Extended Data Fig. 6b-d). In line with previous observations, we did not detect overt alterations in WAT fibrosis in Pten iΔEC mice under HFD (Extended Data Fig. 6e,f). Second, we evaluated the consequences of endothelial-specific deletion of Pten in established obesity (Fig. 3h). Administration of tamoxifen in mice fed with HFD for 12 weeks resulted in a robust correction of body weight, reduced adiposity, increased WAT vascula rity and improved glucose tolerance in the absence of WAT fibrosis ( Fig. 3h-n and Extended Data Fig. 6g-k). Together, these data indicate that endothelial PTEN is instrumental for the control of adiposity in pathophysiological conditions.
Endothelial Pten loss leads to enhanced WAT metabolic rate. To address how loss of Pten in the endothelium resulted in reduced adiposity, we monitored factors involved in nutritional uptake and energy consumption in the mouse model. Notably, the phenotype of Pten iΔEC mice was not due to differences in food intake, intestinal malabsorption, locomotor activity, BAT thermogenesis or Ucp1-dependent or independent WAT browning ( Fig. 4a-h). However, we did detect a significant increase in energy expenditure in mutant mice, consistent with reduced adiposity (Fig. 4i).
In line with this, the plasma triglyceride concentration was reduced in Pten iΔEC mice compared to controls (Fig. 4j). To associate the energy expenditure phenotype with the reduction in adiposity, we sought to assess metabolic activity at tissue resolution. We evaluated mitochondrial function ex vivo by high-resolution respirometry in various freshly isolated tissues. Only WAT, the tissue exhibiting the most prominent phenotype in Pten iΔEC mice, presented an increase in mitochondrial respiration ( Fig. 4k-n). In summary, our results reveal a causal association between EC proliferation, increased local and systemic energy expenditure and reduced adiposity.
Pten-null EC proliferation relies on fatty acid metabolism. The increased local energy consumption in WAT suggested that this phenotype could be due, at least in part, to enhanced metabolic activity by the mutant endothelium. Thus, we interrogated whether the unique access to lipids in the adipose tissue could favor the activation of lipid catabolic transcriptional programs in Pten iΔEC WAT endothelium to promote enhanced vascularization selectively in this tissue. First, we evaluated endothelial expression of genes involved in lipid catabolism in vivo (using Pten iΔEC -Ribotag iHAEC mice; Fig. 5a) and in isolated primary adipose ECs (Fig. 5b). Upregulated genes in Pten iΔEC -Ribotag iHAEC WAT endothelium included fatty acid transporters, genes involved in fatty acid β oxidation (FAO) and central regulators of cellular energy metabolic pathways. In line with our previous results, the activation of this program was unique to WAT endothelium, with no major differences in the endothelium of BAT, muscle and liver (Extended Data Fig. 7a-c). The lipid catabolic reprogramming was associated with enhanced FAO (Fig. 5c). Consistently, we observed increased mitochondrial respiration when lipids were supplemented in the medium (Fig. 5d,e), whereas no changes in glucose uptake (Extended Data Fig. 7d) or in glucose-induced oxygen consumption were observed ( Fig. 5f,g). We speculated that the vascular phenotype in WAT was the result of increased local fatty acid availability. Indeed, cultured Pten iΔEC ECs derived from lung, a tissue that did not manifest a phenotype in vivo (Extended Data Fig. 3c), exhibited biological and metabolic features similar to adipose ECs when lipids were supplied in the medium (Extended Data Fig. 7e-i). To further investigate this, we evaluated the impact of etomoxir (a pharmacological inhibitor of FAO) on cell proliferation 27 . Etomoxir hampered the proliferative burst elicited by Pten loss in adipose ECs in vitro and in vivo ( Fig. 5h-j) without overt macroscopic or weight alterations in any tissue analyzed (Extended Data Fig. 7j). Next, we undertook a genetic approach to selectively interfere with lipid catabolism in ECs in vivo. Our gene expression data showed a prominent increase in genes related to lipid transport, fatty acid β-oxidation and oxidative phosphorylation, which are transcriptionally regulated by PGC1 family factors ( Fig. 5a,b). Also, we identified a mild upregulation of Ppargc1b, the transcript that encodes for Pgc1β ( Fig. 5a,b). Hence, we took advantage of Pgc1β conditional knockout mice 28 , which we bred into the endothelial-specific Pten-deficient mice. Ppargc1b deletion in the context of endothelial Pten loss partially rescued the elevation in adipose vascular density (Extended Data Fig. 7k,l). Together, these data indicate that Pten deletion makes ECs prone to proliferate in the presence of fatty acids, which represents a microenvironmental prerequisite for these cells.

Polyamines are prolipolytic angiocrine metabolic mediators.
FAO was necessary to sustain adipose EC proliferation, whereas adiposity remained reduced both upon pharmacological and genetic blockade of fatty acid catabolism ( Fig. 5k and Extended Data Fig. 7m). These data suggest that reduced adiposity is not solely the consequence of the proliferative burst of Pten iΔEC adipose endothelium. We hypothesized that angiocrine signals induced by Pten deletion were instructing a lipolytic response in adipocytes that would explain the lack of rescue in adiposity by inhibition of FAO. To explore this, we measured basal and stimulated fatty acid secretion into the media. Explants from Pten iΔEC WAT exhibited higher release of free fatty acids (FFAs) than control counterparts (Extended Data Fig. 8a and Fig. 6a). A similar trend was observed in primary adipocyte cultures upon exposure to conditioned medium derived from control or Pten iΔEC ECs (Fig. 6b). In line with this, fasted Pten iΔEC mice exhibited elevated circulating FFAs (Fig. 6c). Collectively, these data demonstrate that endothelial-specific Pten deletion leads to elevated lipid mobilization from the adipose tissue. Stimulation of the β-adrenergic receptors (β-ADRs) activates lipolysis in WAT [29][30][31] . Thus, we tested the involvement of these membrane proteins in the lipolytic phenotype observed in Pten iΔEC mice.
We treated mutant animals with the pan-β-ADR inhibitor propranolol for 4 d and then measured lipolysis in WAT explants. The results showed that propranolol reduced lipolysis in the explants, thus demonstrating that β-ADR receptor activity is relevant for endothelial Pten loss-induced lipolysis in WAT (Fig. 6d). The lipolytic phenotype observed with conditioned media from Pten iΔEC ECs suggests that an angiocrine signal may be responsible for the metabolic effects described above. Angiocrine signals control a variety of pathophysiological processes and are emerging as key mediators of EC function and tissue homeostasis [32][33][34] . These molecular cues could be polypeptides, metabolites or extracellular vesicles [32][33][34][35][36] . To discern the identity of angiocrine signals in our model, we filtered conditioned medium to separate the metabolite fraction from that containing proteins and vesicles (Extended Data Fig. 8b). Notably, only the Pten iΔEC metabolite-rich conditioned medium fraction retained lipolytic activity (Fig. 6e,f). These data indicate that the angiocrine signal that stimulates lipolysis in Pten iΔEC WAT is a metabolite.
Next, we investigated in further detail the metabolic nature of the lipolytic angiocrine signal. To this end, we searched for metabolic pathways fulfilling three criteria: (1) to be previously linked to adiposity, (2) to exert paracrine biological effects and (3) to respond to PI3K activity. Polyamines satisfy these criteria, as they can function in a paracrine manner [37][38][39] , reduce adiposity in the context of obesity [40][41][42][43][44] and it was previously reported that PI3K activation enhances polyamine synthesis to sustain oncogenicity 45,46 . Thus, we hypothesized that polyamines may act as WAT angiocrine lipolytic signals. Polyamines are polycationic metabolites synthesized from methionine and ornithine 38 . 13 C-methionine tracing revealed a remarkable increase in metabolites related to polyamine biosynthesis in isolated Pten-deficient ECs, including 13 C-labeled spermidine and spermine ( Fig. 6g and Extended Data Fig. 8c). In line with these results, label-free metabolomics confirmed the elevation in polyamine levels in Pten iΔEC isolated EC cultures and in Pten iΔEC WAT (Fig. 6h,i and Extended Data Fig. 9a,b). Moreover, elevated polyamine levels were observed in the supernatant of Pten iΔEC EC, in agreement with their secreted nature (Fig. 6j).
Polyamines activate β-ADRs to promote lipolysis. We next ascertained the role of polyamines in regulating WAT biology. Supplementation of 13 C-spermidine in adipocyte cultures showed that this polyamine was efficiently taken up (Fig. 7a). Increased lipolysis in Pten iΔEC WAT is dependent on β-ADR activation (Fig.  6d) and polyamines reportedly promote β-ADR activity 47 . Hence, we tested whether polyamines promoted the release of adipose FFA in a β-ADR-dependent fashion. Supplementation of polyamines to WAT explants increased lipolysis, an effect that was prevented by propranolol ( Fig. 7b and Extended Data Fig. 10a). Also, we showed that polyamines stimulated lipolysis in freshly isolated primary adipocyte cultures in a β-ADR receptor-dependent manner ( Fig. 7c and Extended Data Fig. 10b). β-ADR receptors signal through PKA activation and cyclic AMP production 29 and we could confirm that spermidine increased the production of intracellular cAMP in adipocytes 47 (Fig. 7d and Extended Data Fig. 10c). The capacity of propranolol to block the action of spermidine on β-ADR in WAT explants suggests that there is sufficient natural agonist in the assay (tissue-intrinsic sympathetic innervation) to sustain a basal β-ADR activity. Indeed, we detected the presence of these agonists by liquid chromatography-mass spectrometry (LC-MS) in the adipose tissue explants employed in our assays (Extended Data Fig. 10d). Overall, our data suggest that polyamines are angiocrine mediators of lipolysis through the regulation of β-ADR activity in adipocytes.
To demonstrate that angiocrine polyamines are key regulators of adiposity, we evaluated the effect of polyamine supplementation in vivo. The administration of spermidine to diet-induced obese mice for 6 weeks tempered the obesogenic phenotype induced by HFD, selectively reduced WAT mass and adipocyte size and improved glucose tolerance (Fig. 7e-j and Extended Data Fig. 10e-g). Also, polyamine supplementation significantly reduced food intake in mice (Extended Data Fig. 10h). To rule out that the effect of polyamines was not solely due to reduced food intake, we performed pair-feeding studies in HFD-fed mice. After 10 weeks on HFD, vehicle-treated animals were heavier than the ones supplemented with spermidine, but when the control group was pair-fed with spermidine-treated mice, the differences in body weight were partially maintained (Extended Data Fig. 10i). We also examined whether obesity states are associated with reduced WAT polyamine levels. Notably, WAT from both mice subjected to HFD and humans with obesity exhibited reduced polyamine content (Fig. 7k,l, Extended Data Fig. 10j and Supplementary Table 4).

AMD1 inhibition restores reduced adiposity in Pten iΔEC mice.
We have previously identified a molecular link between the PI3K pathway and polyamine biosynthesis in prostate cancer 46 . Activation of mTOR complex 1 (mTORC1) as a consequence of Pten deletion results in elevated Amd1 protein levels and the elevation of decarboxylated dcSAM, thus promoting the synthesis of polyamines 46 . As shown in Extended Data Fig. 9a,b, dcSAM was significantly increased in cultured ECs and WAT from Pten iΔEC mice. In agreement, we confirmed that Amd1 protein levels, but not mRNA gene expression, were higher in Pten-null ECs ( Fig. 8a and Extended Data Fig. 10k), an effect that was counteracted by the inhibition of mTORC1 with rapamycin (Extended Data Fig. 10l,m).
To address the relevance of Amd1 activity in the phenotype elicited upon endothelial Pten loss and in the absence of a conditional Amd1 knockout mouse model, we took advantage of a selective inhibitor of this enzyme, SAM486A. We confirmed that the conditioned medium from Pten-deficient ECs pre-treated with SAM486A was unable to increase lipolysis in WAT explants (Fig. 8c,d). In addition, explants from Pten iΔEC mice treated with SAM486A exhibited lower lipolytic rates (Fig. 8e) and Pten iΔEC mice treated with SAM486A demonstrated greater body weight (Fig. 8f) and adiposity than vehicle-treated counterparts (Fig. 8g-k). Together, these data uncover polyamines as angiocrine metabolic regulators of lipolysis and adiposity under the control of endothelial PTEN.

Discussion
ECs have been long studied as structural components of blood vessels with a bystander role in systemic metabolic homeostasis. Our data, together with others 2 , challenge this view and show that manipulation of EC biology influences systemic metabolism with pathobiological implications. We provide evidence that EC function modulates adiposity through angiocrine production of polyamines acting directly on adipocytes in a paracrine fashion (Fig. 8l). Remarkably, we demonstrate that ECs exploit the metabolic angio crine communication mode to stimulate the release of FFAs from adipocytes to sustain cell proliferation and promote vascular  . e, Increase in FFA release of eWAT explants from wild-type mice following incubation with the retained fraction (protein-and vesicle-enriched) of conditioned medium from adipose ECs from control and Pten iΔEC mice (control n = 9; Pten iΔEC n = 16). f, Increase in FFA release from eWAT explants of wild-type mice following incubation with the metabolite-enriched fraction (filtered fraction) of conditioned medium from control and Pten iΔEC ECs (control n = 6; Pten iΔEC n = 12). g, Incorporation of carbon-13 ( 13 C) from 13 C-U 5 -methionine (2-h and 4-h pulse) into metabolites related to polyamines biosynthesis in adipose ECs from control and Pten iΔEC mice (n = 6 per group). MET, methionine; SAM, S-adenosylmethionine; dcSAM, decarboxylated S-adenosylmethionine; MTA, 5ʹ methylthioadenosine. h,i, Spermidine levels in adipose ECs (h) (control n = 6; Pten iΔEC n = 6) and in whole tissue eWAT (i) (control n = 8; Pten iΔEC n = 7) measured by LCMS. j, Spermidine (control n = 20; Pten iΔEC n = 23) and spermine (control n = 17; Pten iΔEC n = 22) levels in conditioned medium from adipose ECs from control and Pten iΔEC mice measured by LCMS. Total number of samples control (17 and 20) and Pten iΔEC (22 and 23) correspond to independent EC cultures from eight WAT depots of each genotype. Data represent mean ± s.e.m. Samples represent biological replicates. Statistical analysis was performed by the two-sided Mann-Whitney U-test (a,c,f), one-sided Mann-Whitney U-test for hypothesis-driven analyses (b,h-j), t-test (g) and one-way ANOVA with Bonferroni correction (d).
growth. By doing so, ECs protect WAT from pathological expansion in the context of obesity.
Angiocrine is a term used to define endothelial-derived paracrine signals that mediate parenchymal function and regenerative functions in an organ-specific manner [32][33][34]48 . Yet, little is known about the intercellular endothelial-to-adipocyte communication. We uncover that ECs communicate with adipocytes through metabolites, the so-called polyamines. Other examples of angiocrine-related organ regulation include liver and lung regeneration, neuronal stem cell function, tumor angiogenesis and heart development 32,49-54 ; however, angiocrine signals in those contexts are of a protein nature. Recent evidence has recognized that metabolites produced and secreted by ECs also function as angiocrine signals 35,55 . This is the case for muscle and retinal ECs, which secrete lactate, a product of glycolysis 35,55 . ). e, Experimental design of spermidine treatment in wild-type mice exposed to HFD. f, Body weight curves of wild-type mice fed with HFD and treated with vehicle or spermidine (vehicle n = 9; spermidine n = 10). g, representative macroscopic images of eWAT. h, eWAT pad weight (n = 9). i, Representative histological sections from eWAT. j, Quantification of adipocyte area in eWAT (vehicle n = 7; spermidine n = 9) of wild-type mice fed with HFD and treated with vehicle or spermidine. k,l, Spermidine levels in whole-tissue eWAT (k) from chow and HFD wild-type mice (chow n = 11, HFD n = 11) and omental VAT (l) from non-obese and obese human individuals (non-obese n = 6, obese n = 12) measured by LC-MS. Data represent mean ± s.e.m. Samples represent biological replicates. Statistical analysis was performed by the two-sided Mann-Whitney U-test (h,j,k), one-sided Mann-Whitney U-test for hypothesis-driven analyses (l), t-test for paired groups (a,d), one-way ANOVA with Bonferroni correction (b,c) and two-way ANOVA with Bonferroni correction for correction for three or more unmatched groups (f). However, this is not surprising given that ECs are largely glycolytic cells 56 . A notable observation from these studies is that lactate does not directly interact with parenchymal cells, but it engages macrophages, which act as mediators of tissue regeneration and angiogenesis 35,55 . In contrast, we identify that endothelial-derived polyamines, upon secretion, directly stimulate lipolysis in adjacent adipocytes; thus, demonstrating that angiocrine metabolic signaling directly regulate parenchyma function. A remarkable result from our study is that loss of Pten in the endothelium result in a WAT-restricted phenotype. This is consistent with the observation that ECs specialize in each type of organ to fulfill tissue-specific tasks by cues that are essential for organ function 33 . Given that Pten-null cells primarily use lipids to proliferate and that WAT serves as a reservoir of lipids, our results suggest that the environmental milieu educates EC behavior. This would also explain why in BAT, a tissue with a mild accumulation of lipids, Pten-null ECs exhibit a moderate increase in proliferation. While at present it is not clear how organotypic differences between ECs emerge, our data call for considering the uniqueness of each environmental milieu as a key determinant of this EC specialization. We propose that understanding which signals from each organ microenvironment regulate the unique properties of ECs could offer tissue-specific vascular therapies not only to repair malfunctional or degenerative tissues, but also to improve their homeostatic function.
The exhaustive pathophysiological analyses of the Pten iΔEC mice included here have allowed us to identify a causal relationship between reduced body weight and adiposity with enhanced local metabolic rate in WAT and increased usage of FFA by Pten iΔEC ECs. While we cannot rule out that the adipocytes contribute to the enhanced basal metabolic rate in Pten iΔEC mice, our data support the concept that adipose Pten iΔEC ECs actively contribute to energy consumption by lipid oxidation. This is consistent with the observation that ECs can use lipids to sustain a proliferative phenotype 26,57,58 . We believe that the observation that loss of endothelial Pten in BAT also results in a mildly enhanced vascularization relates to the lipid availability in this tissue. This, in fact, further supports the conclusion that Pten deletion makes ECs prone to proliferate in the presence of fatty acids, which represents a microenvironmental prerequisite for these cells.
We have not been able to confirm a role for improved BAT function or WAT browning in the phenotype of Pten iΔEC mice. This is surprising given that VEGF-related enhanced angiogenesis in iWAT leads to improved systemic metabolic health by stimulating browning in this tissue [22][23][24] . Although the reason for this difference is not clear, collectively these studies reinforce the notion that enhanced vascularity improves WAT function. This fits with recent studies showing that age-dependent organ decline is associated with reduced vessel density 59,60 and that VEGF signaling prevents age-associated capillary loss, improves organ perfusion and function and extends life span 60 .
Polyamines are small polycations that are considered instrumental in proliferative cells 38 . Indeed, we show that polyamines are produced in the proliferative adipose endothelium and extrapolate evidence acquired in cancer studies to the endothelial field 46 . We demonstrate that the connection between the PI3K/mTORC1 pathway and AMD1 activity is operative in ECs. Notably, beyond the role of these metabolites in cell proliferation, we show that they function as paracrine regulators of adipocyte biology. A follow-up question would be to identify whether polyamines are angiocrine mediators in any ECs that undergoes proliferation or whether they selectively function as angiocrine mediators of adipose ECs. A critical finding from our study is that EC-derived secreted polyamines regulate the activity of β-ADRs, a concept for which there was solely isolated evidence to date 47,61,62 . However, a relevant question that remains is whether polyamines require the presence of β-ADR ligands to promote the activation of the receptor, an idea the has been proposed in specific experimental settings 47 . We observed that both supplemented polyamines and conditioned medium from Pten-deficient ECs activate the release of FFA in WAT explants, which contain detectable levels of β-ADR ligands. However, similar experiments performed in primary adipocyte cultures showed that while polyamine supplementation recapitulated the elevated release of FFA, the conditioned medium from Pten-deficient ECs had limited differential lipolytic activity compared to that of wild-type ECs. Taken together, our results suggest that the presence of β-ADR ligands enhances the capacity of polyamines to promote lipolysis. In the absence of these ligands (such as adipocyte cultures) the lipolytic activity of Pten-deficient EC-conditioned medium would lose potency, whereas supplemented polyamines in this assay would retain activity, probably owing to their elevated relative abundance in the assay. Assuming that the conditioned medium best recapitulates the secreted polyamine conditions, we propose that polyamines function as fine-tuners of canonical β-ADR signals. Of note, our data agree with previous observations showing that exogenous administration of spermidine regulates lipid metabolism and in turn ameliorates the WAT pathophysiological response to HFD 63 . This provides further support that polyamines and in particular spermidine, may open new therapeutic avenues to treat obesity.
Taken together, we uncover an unappreciated mode of communication between ECs and adipocytes with implications for systemic metabolism. Our data provide evidence that this mode of communication is disrupted in obesity, thus, opening further research avenues to comprehend and treat this disease. Also, we propose that understanding the extent of the contribution of polyamines to adrenergic signaling-driven cellular responses may be important in many other biological contexts.

Methods
All our research compiles with the relevant ethical regulation. Experiments pertaining to mouse work were conducted in accordance with the guidelines and laws of the Catalan Departament d' Agricultura, Ramaderia i Pesca (Catalunya, Spain) and were approved by the Ethics Committees of Institut d'Investigació Biomèdica de Bellvitge. The Hospital Clinical of Barcelona Ethical Committee approved the human studies and all study participants provided informed consent to donate tissue samples. Supplementary Tables 1, 2 and 3, including commercial references and experimental dosage.

Reagents. All reagents used in this work are listed in
Mice. Mice were maintained under specific-pathogen-free conditions and kept in individually ventilated cages. Mice were fed ad libitum either with chow diet or HFD (45% fat-enriched diet). The transgenic mouse colony was kept under mixed background of C57BL/6J and 129. Pten flox mice 15 were crossed into the transgenic mice expressing the tamoxifen-inducible recombinase CreER under the control of Pdgfb promoter (PdgfbiCreER) transgenic mice 16 or into the control of Cdh5 promoter (Cdh5-CreER T2 ) 18 . Only male mice were used for our studies. For isolation of EC-specific actively translating mRNA, Ribotag flox mice (Rpl22 tm1.1Psam ) 17 were crossed with Pdgfb-iCre-ER (Ribotag iHAEC ) and Pten iΔEC (Pten iΔEC -Ribotag iHAEC ). Cre activity and gene deletion were induced by intraperitoneal (i.p.) injection of 4-OHT at 25 mg (2.5 µl of a 10 mg ml −1 solution in absolute ethanol) in pups at postnatal day (P)2 and P3 and tissues were collected at 5 and 12 weeks of age. In adult mice, gene deletion was induced by i.p. injections of tamoxifen (75 mg kg −1 ; resuspended in peanut oil) at 8 or 18 weeks of age for three alternate days. Both Cre-positive and -negative mice were treated with 4-OHT or tamoxifen. For combined EC-specific loss-of-function of Pten and Ppargc1b, we crossed Pten flox mice 15 with Pgc1β flox mice 28 and PdgfbiCreER 16 mice. For pharmacological rescue studies, control and Pten iΔEC mice were treated with vehicle (saline), etomoxir (25 mg kg −1 ) or SAM486A (5 mg kg −1 ) from 5 to 10 weeks of age. For explants, mice were treated for 4 d with vehicle (saline), SAM486A (5 mg kg −1 ), spermidine (20 mg kg −1 ) or propranolol (20 mg kg −1 ). Etomoxir and SAM486A in vivo experiments were planned by choosing a dose, route of administration and drug regimen previously published by us 27,46 . For spermidine treatment, wild-type C57BL/6J mice (purchased from Charles River) were fed with HFD and were treated with vehicle or spermidine (20 mg kg −1 ) from 6 to 12 weeks of age daily, excluding weekends. Of note, the spermidine supplementation strategy 20 mg kg −1 is within physiological ranges of spermidine supplementation 43,64 . All in vivo experiments of our study, including chow and HFD conditions, were performed without altering the normal presence of dietary polyamines. Before starting in vivo experiments, mice were weighed and homogenously distributed in experimental groups. All other criteria were not considered and as such, were randomized.
Human cohort. We studied omental VAT from morbid patients with obesity (body mass index (BMI) > 35) undergoing bariatric surgery and non-obese individuals (BMI < 30) undergoing abdominal surgery at the Hospital Clinic of Barcelona. Participant information including sex, number, BMI and age (Supplementary Table 4). Participants of the study did not receive any compensation.
RiboTag, RNA extraction, cDNA synthesis and qPCR. Enrichment of the active translating RNA was achieved by immunoprecipitation (IP) of ribosomes via hemagglutinin A (HA) as described 65  Vascular density analysis. Tissues were cut into 3-mm 3 cubes, permeabilized for 1 h with PBS + 1% Triton X-100 and blocked with blocking buffer (1× PBS + 0.3% Triton X-100 + 5% goat serum) for 2 h at room temperature. Primary and secondary antibodies (Supplementary Table 3) diluted in blocking buffer were incubated overnight at room temperature, with over-day washings with PBS + 0.3% Triton X-100 after both incubations. Mouse WAT pads were stained with Alexa-conjugated isolectin B4 (IB4). Human WAT pads were stained with anti-CD31 antibody. To study the vasculature of BAT, muscle, liver, heart and brain, we used cryosection immunostaining. Freshly isolated tissues were fixed overnight in 4% paraformaldehyde, washed in PBS, dehydrated in 30% sucrose overnight and embedded in OCT. BAT (14 μm), muscle, heart, brain (10 μm) and liver (5 μm) cryosections were cut using the Cryostat (Leica Microsystems). BAT sections were blocked with 1× PBS 0.3% Triton X-100 + 5% goat serum for 1 h at room temperature, followed by incubation with Alexa-conjugated IB4 overnight at 4 °C. Muscle, liver, heart and brain sections were incubated in Tris-EDTA solution (0.1 M EDTA + 0.001 M Tris-base + 0.05% Tween 20) in a steamer for 30 min. After cooling down, sections were blocked with 1× PBS + 5% donkey serum + 0.4% Triton X-100 for 1 h at room temperature, followed by incubating overnight at 4 °C with anti-CD31 antibody and with secondary antibody 1 h at room temperature (Supplementary Table 3) in blocking solution. To analyze vessel area and proliferation, images were taken with a Leica SP5 laser-scanning confocal microscope (Leica Application Suite) or Nikon 80i microscope (NIS-Elements) using ×20 and ×40 objectives. Confocal images are maximal intensity z-stack projections and images were processed using Volocity, Fiji and Adobe Photoshop CS5. Vessel density was quantified by measuring the IB4-positive or CD31-positive area using ImageJ software. Endothelial branch points were quantified in four images per each sample (taken with ×40 oil immersion objective) and represented as an average. Histology analysis. The 5-µm thick paraffin sections were stained with hematoxylin/ eosin and mounted with DPX mounting medium. Adipocyte area was calculated with the Adiposoft tool of ImageJ software. Fibrosis was evaluated by Trichrome Stain kit following manufacturer's instructions. Cell death was evaluated by immunostaining of cleaved caspase-3 in paraffin sections, following the standard protocol. Briefly, antigen retrieval was performed with citrate buffer and permeabilization with PBS + 0.2% Tween 20, primary antibody was incubated overnight at 4 °C and secondary 2 h at room temperature (Supplementary Table 3). Samples were mounted with Shandon Immu-Mount. Images were taken with a Nikon 80i microscope using a ×20 objective.
Glucose tolerance test, insulin tolerance test and HOMA-IR index. GTT and ITT were performed in 6-h fasted mice. Both compounds were i.p. injected, glucose (1.5 g kg −1 ) and insulin (0.75 UI kg −1 ) and blood glucose was monitored for 90-120 min. HOMA-IR index was calculated applying the following formula: (fasting glucose (mg dl −1 ) × fasting insulin (mU l −1 )) / 405. Insulin was measured with Ultra-Sensitive Mouse Insulin ELISA kit following manufacturer's instructions.
Daily food intake. Mice were single housed, acclimatized for 1 week before study and a weighed amount of food was provided. Food intake was measured for five consecutive days at 8 and 12 weeks of age. Pair feeding was carried out to determine the extent to which the body weight-reducing effect of spermidine treatment was the consequence of changes in food intake. All mice were individualized during the pair-feeding protocol. The average food intake of the spermidine-treated group was daily measured between 9:00 and 10:00. Subsequently, the pair-fed group was offered the same amount of food eaten by spermidine-treated mice on the previous day.
Body composition, Indirect calorimetry, locomotor and thermogenic activity. Body weight and discrete adipose tissue pad mass were measured using a precision scale. Whole body composition was measured using nuclear magnetic resonance imaging (Whole Body Composition Analyzer; EchoMRI). Indirect calorimetry and locomotor activity was assessed using a TSE LabMaster modular research platform (TSE Systems) as previously described 66,67 . Briefly, mice were acclimatized for 24 h into test chambers and monitored for additional 48 h. O 2 consumption and CO 2 production were measured every 45 min during 48 h, to indirectly determine EE. Locomotor activity was determined using a multidimensional infrared light beam system with the parameters defined by the LabMaster software. These analyses were performed at 7 weeks of age, before body weight differences between control and Pten iΔEC mice were apparent. Heat production was visualized using a high-resolution infrared camera (FLIR PM280; FLIR Systems), as previously described 68 .
Western blot. Cell lysis, SDS-PAGE and immunoblot were performed as previously described 14 . Antibodies used are listed in Supplementary Table 3. Quantification of band intensities by densitometry was carried out using ImageJ software.
Mitochondrial respiration. Mitochondrial respiration was assessed in freshly isolated eWAT, iWAT, muscle and liver by high-resolution respirometry in an Oroboros Oxygraph-2k system (Oroboros Instruments) as previously reported 69 . LEAK respiration was measured by the addition of NADH-linked substrates (complex I linked) malate (2 mM) and pyruvate (5 mM) in the absence of ADP. OXPHOS state was measured by adding ADP + MgCl 2 (5 mM) and cytochrome C (10 µM), followed by the subsequent addition of glutamate (10 mM) (NADH-linked pathway) and succinate (10 mM) (convergent electron flow through both, NADH-and succinate-linked pathways). FCCP (0.5 µM) was titrated to evaluate the maximal capacity of the electron transfer system (ETS; CI + CII). Finally, rotenone (0.5 µM) was used to inhibit CI and measure ETS fueled by succinate-linked pathway. Oxygen flux values were expressed relative to tissue wet weight per second (pmol O 2 mg −1 s −1 ). Residual oxygen consumption (ROX) was determined by the inhibition of complex III adding antimycin A (2.5 µM) and this value was subtracted from O 2 flux as a baseline for all respiratory states.
Feces collection and analysis. Feces were collected from individually housed mice cages for 4 d, stored at −20 °C and desiccated at 60 °C before processing. Energy content in feces was measured using a calorimetric bomb in the Laboratorio de Nutrición Animal SERIDA (Villaviciosa, Spain). For triglyceride measurement, frozen feces samples were pulverized under liquid nitrogen and 100 mg portions were digested in 3 M KOH for 1 h at 70 °C, followed by overnight incubation at room temperature. Samples were diluted to a final concentration of 100 mg tissue in 500 µl Tris-HCl 50 mM before using triglycerides-LQ kit under manufacturer's instructions.
Primary ECs. Selection and culture conditions for adipose-derived and lung-derived ECs were as previously reported 14 . To assess cell proliferation 1 × 10 4 primary ECs were cultured in DMEM/F12 with 1% FBS, 2 mg ml −1 of AlbuMAX and half dose of EC growth factors in the presence or absence of etomoxir and measured by staining with crystal violet (0.1% in 20% methanol) at day 2 and 3 and normalized by day 0 values. For conditioned medium collection, 8 × 10 5 cells were cultured in 1 ml of DMEM 1 g l −1 glucose + 0.5% FFA-free BSA. After 4 h, medium was collected, centrifuged for 5 min at 300g, filtered with a 0.22 filter and kept at 4 °C (up to 2 d). When indicated, cells were pre-treated 24 h with vehicle (H 2 O) or SAM486A (0.5 µM) in growing medium. For medium fractioning, 2.5 ml of medium was loaded into Amicon Ultra-4 filters (membrane PLBC Ultracel-3, 3 kDa) and centrifuged at 3,800g, 4 °C for 1 h. The flow-through (metabolite fraction) was collected and stored at 4 °C. The retained protein and vesicle fraction (150 µl), was washed with 2 ml of DMEM 1 g l −1 glucose, centrifuged again, diluted in 2 ml of DMEM 1 g l −1 glucose and stored at 4 °C. For Amd1 analysis, ECs were cultured for 24 h with vehicle (dimethylsulfoxide) or mTOR inhibitor (rapamycin, 1 µm). FAO was measured as previously described 70  To assess respiration, ECs were seeded in customized Seahorse 24-well plates. At 1 h before the assay, cells were maintained in XF Assay Medium Modified DMEM (Seahorse Bioscience) supplemented with 5 mM glucose in a non-CO 2 incubator and just before the assay, 10 µl of AlbuMAX were added (to a final concentration of 2 mg ml −1 ). OCR was measured using the Seahorse XFe24 analyzer (Agilent) following the manufacturer's protocols. OCR was calculated by plotting the O 2 tension of medium as a function of time (pmol min −1 ) and data were normalized by the protein concentration measured in each individual well. Calculations were performed using the Agilent Seahorse Wave Desktop software. Primary adipocytes. The stromal vascular fraction (SVF) was isolated from iWAT depots of 6-week-old male mice. Differentiation was induced 48 h after cells reached confluence by adding the induction medium: DMEM complete with insulin (100 nM), dexamethasone (1 µM), 3-isobutyl-1-methylxanthine (IBMX) (0.5 mM) and rosiglitazone (1 µM). After 48 h, induction medium was replaced by DMEM complete + insulin (100 nM). Experiments were performed at day 8 of differentiation. Lipolysis was evaluated by measuring FFA release in differentiated adipocytes. Briefly, differentiated adipocytes were washed with 1× PBS and stimulated by (1) conditioned medium from control and Pten iΔEC ECs or (2) DMEM 1 g l −1 glucose + 0.5% FFA-free BSA ± 1 µM spermidine. DMEM 1 g l −1 glucose + 0.5% FFA-free BSA + 1 µM CL316243 (β-ADR agonist) was included in all assays as positive control. When stated, cells were pre-treated with 100 µM propranolol in DMEM complete + insulin (100 nM) during 2 h before the experiment. FFA release was measured at min 0 and min 180 and the difference between time points was calculated. FFA were measured using FFA assay HR-NEFA kit following manufacturer's instructions. cAMP was measured in primary adipocytes cell lysate after 15 min incubation with DMEM 1 g l −1 glucose + 0.5% FFA-free BSA ± 1 µM spermidine. Then, 10 µM forskolin was used as positive control. cAMP was measured using cAMP colorimetric assay, following manufacturer's instructions.
Plasma leptin, TG and FFA measurement. Blood was collected in EDTA microtubes from the tail vein of mice before and after 16 h of starvation. Plasma was obtained after 20 min of centrifugation at 1200 g at 4 °C. Leptin was measured with Mouse Leptin ELISA Kit following manufacturer's instructions after 6 h starvation. TGs were measured by a chemistry panel following standard protocols. FFA were measured using FFA assay HR-NEFA kit following manufacturer's instructions.
Ex vivo eWAT explants. Fresh eWAT depots were collected in ice-cold DMEM 1 g l −1 glucose + 0.5% FFA-free BSA, cut into small pieces and explants of ~25 mg of tissue and were placed in 24-well plates (one piece per well) with 500 µl of DMEM 1 g l −1 glucose + 0.5% FFA-free BSA. Baseline medium samples were obtained after 1 min of incubating the tissue at 37 °C with gentle shaking. Then, spermidine (50 µM) or equal volume of DMEM 1 g l −1 glucose + 0.5% FFA-free BSA was added. CL316243 was used as a positive control in all assays. Plates were incubated at 37 °C while shaking and aliquots of media (15 µl) were taken at 30, 60, 90 and 120 min. Alternatively, explants were incubated with 500 µl of conditioned medium and aliquots of medium were collected at baseline and 90 min. FFA concentration was quantified using an FFA assay HR-NEFA kit following manufacturer's instructions. When stated explants were preincubated 10 min with 100 μM propranolol at 37 °C before adding spermidine.
Targeted metabolomics. For metabolomic flux analysis ECs were incubated 2 and 4 h with DMEM high glucose, no glutamine, no methionine, no cystine plus [U-13C5]l-methionine 30 µg ml −1 , washed with PBS and snap-frozen in liquid nitrogen. For in vivo metabolomic flux analysis, [U-13C5]l-methionine was injected through the tail vein at 100 mg kg −1 ; mice were killed 20 h after injection and adipose tissue was collected and snap-frozen in liquid nitrogen. For targeted metabolomic of conditioned medium, ECs were incubated 4 h in DMEM 1 g l −1 glucose + 0.5% FFA-free BSA, medium was collected and kept at −80 °C; protein content was used to normalize data (the experiment was performed with technical triplicates of six biological replicates for each genotype). To measure spermidine uptake, primary adipocytes were incubated with 13 C 4 spermidine in DMEM 1 g l −1 glucose + 0.5% FFA-free BSA, washed with PBS and snap-frozen in liquid nitrogen; samples were taken at 0, 15, 30 and 60 min. For ex vivo metabolomic analysis adipose tissue was collected and immediately snap-frozen in liquid nitrogen. Levels of dcSAM and spermidine in WAT and adipose-derived ECs were analyzed by UPLC-MS, as previously described 46 . Briefly, extraction and homogenization were conducted in methanol/acetic acid (80/20% v/v). Speed-vacuum-dried metabolites were solubilized in 100 μl of a mixture of water/acetonitrile (40/60% v/v) and injected onto the UPLC-MS system (Acquity and SYNAPT G2, Waters). The extracted ion traces were obtained for dcSAM, spermine and spermidine. Corrected signals were normalized to protein content or mg of tissue.
Epinephrine and norepinephrine measure in WAT. WAT was homogenized in 500 µl of ice-cold extraction liquid and the aqueous phase was transferred to a fresh aliquot and placed at −80 °C for 20 min. The chilled supernatants were evaporated with a speedvac in approximately 2 h. The resulting pellets were resuspended in 150 µl water/acetonitrile (MeCN)/formic acid (40/60/0.1 v/v/%). Samples were measured with a UPLC system (Acquity, Waters) coupled to a time-of-flight mass spectrometer (ToF MS, SYNAPT G2, Waters). Extracted ion traces were obtained for the epinephrine fragment (m/z = 166.0859) and for norepinephrine fragment (m/z = 152.0731), in a 20-mDa window and subsequently smoothed and integrated with TargetLynx software (Waters). These calculated raw signals were adjusted for by median fold-change adjustment.
Reporting Summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
Image source data and Excel files of all data presented in graphs within the figures and extended data figures have been supplied in Source Data files. Separate files for each figure have been supplied.