Nitrate-Sialin2 axis couples ER-mitochondrial calcium signaling with fatty acid metabolism to drive white adipose browning

Abstract

White adipose browning is a promising route to restore energy balance; however, how inorganic anion signals engage intracellular organelle networks to drive this process remains unclear. Here, we identify Sialin2 as a nitrate sensor that converts dietary nitrate into a spatially confined thermogenic program by coupling ER-mitochondria Ca²⁺ transfer with lipid routing into mitochondrial oxidation. Sialin2 localizes to mitochondria and the endoplasmic reticulum (ER), where it strengthens ER-mitochondria contacts and engages the inositol 1,4,5-trisphosphate receptor type 1 (IP3R1)-voltage-dependent anion channel 1 (VDAC1)-mitochondrial calcium uniporter 1 (MCU1) conduit to enhance inducible mitochondrial Ca²⁺ uptake. In parallel, Sialin2 associates with lysosomal acid lipase (LIPA), acyl-CoA synthetase long-chain family member 3 (ACSL3), and carnitine palmitoyltransferase 1 A (CPT1A) to channel lipid-droplet-derived fatty acids into β-oxidation, thereby fueling the tricarboxylic acid cycle and uncoupling protein 1 (UCP1)-dependent respiration. Loss of Slc17a5 abolishes nitrate-evoked browning and metabolic benefits, whereas nitrate supplementation improves adipose thermogenesis and systemic metabolic indices in male mice with diet-induced obesity without adrenergic stimulation. Together, these findings identify an organelle-specific nitrate-sensing mechanism that couples inorganic anion signalling to substrate routing in adipocytes and establish a non-hormonal pathway for restoring metabolic homeostasis.

Dietary nitrate activates Sialin2 in adipocytes, enhancing ER-mitochondria calcium signalling and lipid metabolism to drive beige fat formation, increase energy expenditure, and promote metabolic homeostasis under high-fat diet conditions.

Introduction

The global epidemic of obesity and metabolic syndrome represents a major public health challenge worldwide^1–3. A central feature of this crisis is adipose tissue dysfunction, because adipose tissue functions as both a lipid depot and an endocrine organ whose remodeling capacity strongly influences whole-body energy homeostasis^4–7. Beige adipocytes, which emerge within white adipose tissue (WAT) in response to stimuli, such as cold or β-adrenergic activation, possess thermogenic features similar to those of classical brown adipocytes and dissipate energy through uncoupling protein 1 (UCP1)-mediated mitochondrial uncoupling^8–10. Because beige fat is inducible and present in adult humans, therapeutically harnessing its activation represents a physiologically grounded strategy with substantial potential to mitigate obesity and its comorbidities. However, clinical efforts to promote beige adipocyte formation have been largely restricted to β-adrenergic agonists, which often show limited efficacy and adverse cardiovascular effects¹¹. This limitation has prompted growing interest in endogenous nutrient-derived cues as alternative drivers of adipose remodeling.

Mitochondria serve as thermogenic engines and dynamic signaling hubs^12,13, and their activity is shaped by physical and functional coupling with multiple organelles, including lipid droplets (LDs) and the endoplasmic reticulum (ER)¹⁴. LD-mitochondria contacts facilitate vectorial lipid transfer that supports β-oxidation, and this coupling becomes more prominent during states of energetic demand, such as fasting and exercise, thereby enhancing lipolysis and oxidative output¹⁵. In parallel, specialized ER-mitochondria contact sites, known as mitochondria-associated membranes (MAMs), generate Ca²⁺ microdomains that regulate mitochondrial respiration, ATP production, and lipid oxidation^16,17. Disruption of ER-mitochondria Ca²⁺ exchange impairs adipocyte thermogenic capacity and energy expenditure^18–20. The IP3R1-VDAC1-MCU1 complex constitutes a core route for Ca²⁺ flux across the MAM interface²¹, yet the physiological cues that engage this axis remain unclear. More broadly, the nutrient-derived signals and intracellular sensors that coordinately regulate LD-mitochondria lipid flux and ER-mitochondria Ca²⁺ transfer during white-fat browning remain poorly defined^22,23. Clarifying this interface is crucial for understanding how adipocytes translate environmental and metabolic inputs into mitochondrial adaptations that sustain thermogenesis.

Inorganic nitrate, once considered an inert byproduct of nitrogen metabolism, has emerged as a bioactive regulator of systemic metabolic homeostasis²⁴. Accumulating evidence suggests that dietary nitrate improves mitochondrial efficiency, mitigates high-fat diet (HFD)-induced obesity, and supports vascular, hepatic, and musculoskeletal homeostasis under metabolic stress^25–27. These findings collectively reposition nitrate as a bioactive inorganic anion signal with coordinated multi-organ effects on metabolic flexibility, extending beyond its classical role as a precursor of nitric oxide (NO). In our recent work, nitrate was shown to function as an inorganic anion signal sensed by Sialin2, a proteolytically generated Sialin (SLC17A5) isoform with multi-organelle localization, thereby initiating compartmentalized signaling outside the canonical NO framework^28,29. Importantly, Sialin2 is not confined to lysosomes. It is dynamically localized to multiple organelles, including mitochondria, the ER, and lysosome-associated membranes, suggesting that it may function as a distributed inorganic anion sensor that integrates extracellular nitrate availability into organelle-specific responses²⁸. Our previous work identified mitochondrial Sialin2 as a key modulator of AMP-activated protein kinase (AMPK) signaling and metabolic stress adaptation, but whether Sialin2 coordinates organelle crosstalk, particularly ER-mitochondria Ca²⁺ signaling, to regulate adipocyte thermogenic programming remains unknown.

Here, we identify an organelle-centered signaling mechanism in which dietary nitrate engages Sialin2 to drive white adipocyte browning. Using adipocyte-specific genetics, metabolic tracing, and proximity-labeling proteomics, we demonstrate that Sialin2 is essential for nitrate-induced mitochondrial activation, fatty acid oxidation, thermogenic gene induction, and beige programming. Mechanistically, nitrate enhances Sialin2-dependent ER-mitochondria tethering and engages the IP3R1-VDAC1-MCU1 Ca²⁺ transfer axis, thereby increasing mitochondrial Ca²⁺ uptake, membrane potential, and TCA cycle flux while promoting LD-mitochondria coupling to direct fatty acids into β-oxidation. These findings establish Sialin2 as a nitrate sensor that links extracellular nitrate availability to compartmentalized Ca²⁺ dynamics and substrate routing, thereby defining a mechanistic framework for activating beige thermogenesis and improving energy homeostasis in obesity.

Results

Nitrate promotes adipose browning and alleviates HFD-induced metabolic dysfunction

To determine whether increasing nitrate availability can counteract diet-induced obesity and stimulate adipose browning, we supplemented mice with dietary nitrate during an 8-week HFD challenge (Supplementary Fig. 1a). Under HFD conditions, nitrate supplementation increased nitrate concentrations in saliva and blood (Supplementary Fig. 1b, c), confirming systemic exposure and bioavailability during diet-induced metabolic stress. Mice receiving nitrate together with high-fat feeding gained less weight and accumulated less fat than those on HFD alone, as indicated by a lower cumulative body-weight area under the curve and reduced adiposity (Fig. 1b and Supplementary Fig. 1d). Systemic metabolic indices were also improved, including reductions in plasma triglycerides, fasting insulin and glucose, a lower homeostasis model assessment for insulin resistance (HOMA-IR), and enhanced glucose tolerance during the oral glucose tolerance test (OGTT), consistent with broad mitigation of HFD-evoked metabolic dysfunction (Fig. 1c–g and Supplementary Fig. 1e). To determine whether these improvements were associated with altered whole-body energy balance, we performed indirect calorimetry. Nitrate-treated mice exhibited increased energy expenditure during high-fat feeding, accompanied by a shift in respiratory exchange ratio, whereas food intake and locomotor activity were not increased (Fig. 1h–k). When the same analysis was performed under thermoneutral conditions, the effect of nitrate on energy expenditure and respiratory exchange ratio was attenuated, whereas food intake and locomotor activity remained unchanged (Supplementary Fig. 1f). These findings support the interpretation that nitrate promotes metabolic improvement in part through enhanced adaptive thermogenesis under HFD conditions.

Fig. 1. Nitrate supplementation induces adipose tissue browning and improves metabolic phenotypes in HFD mice.

Representative body morphology (a) and AUC for body weight (b) of mice in the Con, Nit, HFD, and HFD+Nit groups before euthanasia. The corresponding weight curves are shown in Supplementary Fig. 1d. c Plasma TG levels. Plasma insulin (d), fasting blood glucose (e), and HOMA-IR (f). g AUC of the OGTT. The corresponding OGTT curves are shown in Supplementary Fig. 1e. Quantitative analysis of energy expenditure (EE) (h), respiratory exchange ratio (RER) (i), total locomotor activity (XT) (j), and cumulative food intake (FEED) (k) over the 24 h monitoring period at room temperature. Representative images of eWAT (l) and eWAT weight normalized to body weight (m). Representative H&E-stained sections of eWAT (n), quantification of average adipocyte area (o), and quantification of crown-like structure (CLS) counts (p). Representative Sirius Red-stained sections of eWAT (q) and quantification of fibrotic area (r). Representative UCP1 immunohistochemical staining of iWAT (s) and quantification of UCP1-positive area (t). Immunoblot quantification of full-length Sialin (u), Sialin2 (v), and UCP1 (w) in adipose tissue. Representative blots are shown in Supplementary Fig. 1h. x Quantitative analysis of CS activity in subcutaneous white adipose tissue from mice. CS activity was normalized to total protein content and is presented as nmol min⁻¹ mg⁻¹ protein. y Schematic illustration showing that concomitant nitrate administration during 8 weeks of HFD feeding attenuates obesity in mice by stimulating adipose tissue browning. Created in BioRender. Jiang, O. (2026) https://BioRender.com/z98kmo9. For all panels, data are represented as the mean ± SD, and p-values denote a two-sided t-test. Scale bar: 100 μm. n = 9 biologically independent mice per group, except h–k, r, u–w, where n = 3 independent experiments.

At the tissue level, nitrate reduced the relative mass of WAT and attenuated adipocyte hypertrophy (Fig. 1l, m). H&E-based quantification further showed a reduction in crown-like structure density, indicating alleviation of inflammatory remodeling in adipose tissue (Fig. 1n–p). In parallel, Sirius Red staining revealed decreased collagen deposition and a smaller fibrotic area in nitrate-treated WAT, consistent with reduced stromal remodeling under HFD conditions (Fig. 1q, r). Consistent with activation of a thermogenic program under HFD conditions, the UCP1-positive area in beige adipose depots was expanded in the nitrate-supplemented group (Fig. 1s, t). Molecular analyses further supported these histological changes. Slc17a5 mRNA was increased in adipose tissue from nitrate-supplemented HFD mice, indicating expansion of the transcript pool that gives rise to Sialin and its processed derivatives (Supplementary Fig. 1g). Consistent with this, immunoblot analysis showed that both full-length Sialin and Sialin2, the cleaved nitrate-responsive form, were elevated under these conditions (Supplementary Fig. 1h). In parallel, UCP1 was increased at both protein and transcript levels, and thermogenic and mitochondrial gene programs, including Pgc-1α, Cidea, Cycs, and Cox8b, were upregulated (Fig. 1u–w and Supplementary Fig. 1g, h). Consistent with these molecular changes, citrate synthase (CS) activity in WAT was increased by nitrate under HFD conditions, providing additional in vivo support for enhanced mitochondrial oxidative capacity in adipose tissue (Fig. 1x).

Collectively, these findings demonstrate that dietary nitrate is bioavailable and metabolically effective in vivo under HFD conditions and promotes adipose browning while alleviating obesity-associated metabolic dysfunction (Fig. 1y). The coordinated induction of Sialin2 together with thermogenic programs under diet-induced metabolic stress supports Sialin2 as a nitrate-responsive effector in adipose tissue.

Adipocyte-specific Slc17a5 deletion exacerbates obesity-associated metabolic dysfunction by impairing adipose browning

To test whether nitrate acts through Sialin2 in adipocytes, we generated adipocyte-specific Slc17a5 knockout mice (Slc17a5^ΔFabp4, cKO) and confirmed efficient deletion by genotyping and protein analysis (Fig. 2a and Supplementary Fig. 2a). Compared with wild-type (WT) littermates, cKO mice subjected to the same 8-week HFD challenge gained more body weight and developed a worsened metabolic profile under HFD conditions, with higher plasma triglycerides, elevated fasting insulin and glucose, an increased HOMA-IR, and impaired glucose tolerance (Fig. 2b–h and Supplementary Fig. 2b, c). These abnormalities were not meaningfully improved by nitrate treatment in the cKO background (Fig. 2b–h and Supplementary Fig. 2b, c). These findings indicate that adipocyte Slc17a5 is required for the metabolic benefits of nitrate in vivo and that its loss exacerbates systemic metabolic dysfunction under dietary stress.

Fig. 2. Adipocyte-specific Slc17a5 deficiency abolishes nitrate-responsive white adipose browning and impairs metabolic homeostasis in mice.

a Schematic illustration of the generation of adipocyte-specific Slc17a5 knockout mice. Created in BioRender. Jiang, O. (2026) https://BioRender.com/k10nsy8. Representative body morphology (b) and AUC for body weight (c) of mice in the WT, cKO, and cKO+Nit groups before euthanasia. The corresponding body weight curves are shown in Supplementary Fig. 2b. d Plasma TG levels. Plasma insulin (e), fasting blood glucose (f), and HOMA-IR (g). h AUC of the OGTT. The corresponding OGTT curves are shown in Supplementary Fig. 2c. Quantitative analysis of EE (i), RER (j), XT (k), and FEED (l) over the 24 h monitoring period at room temperature. m Heatmap of differential lipid metabolites in beige adipose tissue from WT and cKO mice. n Cardiolipin levels in beige adipose tissue from WT and cKO mice. Representative images of white adipose tissue (o) and white adipose tissue weight normalized to body weight (p) in the WT, cKO, and cKO+Nit groups. Representative H&E-stained sections of white adipose tissue (q) and quantification of CLS counts (r). Representative Sirius Red-stained sections (s) and quantification of fibrotic area (t) in adipose tissue. Representative UCP1 immunohistochemical staining (u) and quantification of UCP1-positive area (v) in beige adipose tissue. w Relative mRNA expression of Ucp1, Cidea, Pgc-1α, Cycs, and Cox8b in beige adipose tissue. mRNA levels were normalized to β-actin mRNA. x Schematic illustration showing suppression of adipose tissue browning by adipocyte-specific Slc17a5 deficiency during 8 weeks of HFD challenge. Created in BioRender. Jiang, O. (2026) https://BioRender.com/25yshl4. For all panels, data are represented as the mean ± SD, and p-values denote a two-sided t-test. Scale bar: 100 μm. n = 6 biologically independent mice per group, except n, r, t, w, where n = 3 independent experiments.

To determine whether these phenotypic differences were associated with altered whole-body energy balance, we performed indirect calorimetry. cKO mice exhibited reduced energy expenditure during high-fat feeding, accompanied by altered respiratory exchange ratio, whereas food intake and locomotor activity were not increased (Fig. 2i–l). Nitrate did not restore these defects in cKO mice (Fig. 2i–l), indicating that the aggravated metabolic phenotype was associated with impaired systemic energy dissipation rather than with altered caloric intake or physical activity. To further evaluate the contribution of thermogenesis, we repeated the metabolic assessment under thermoneutral conditions (30 °C). Under thermoneutrality, the genotype-dependent differences were markedly attenuated compared with those observed at standard housing temperature, while food intake and locomotor activity remained comparable between groups (Supplementary Fig. 2d–g). These findings support a substantial contribution of impaired adaptive thermogenesis to the metabolic phenotype caused by adipocyte-specific Slc17a5 deficiency.

Lipidomic profiling of beige adipose tissue revealed a pronounced perturbation of lipid homeostasis in cKO mice, with triglyceride accumulation accompanied by reductions in diglycerides, free fatty acids (FFAs), and phospholipids (Fig. 2m). Notably, cardiolipin (CL), a mitochondria-restricted phospholipid essential for respiratory chain organization, ATP synthase assembly, and ER and mitochondria contact integrity³⁰, was markedly decreased (Fig. 2n). The reduction in cardiolipin is consistent with impaired mitochondrial structure and function coupling in the absence of Sialin2. At the tissue level, adipose expansion and remodeling were evident in cKO mice under the same HFD conditions. White adipose depots displayed increased relative mass, and adipocytes were markedly hypertrophic despite nitrate treatment (Fig. 2o, p). H&E-based quantification showed a significant increase in crown-like structure density, indicating aggravated inflammatory remodeling that persisted in nitrate-treated cKO mice (Fig. 2q, r). In parallel, Sirius Red staining further revealed increased collagen deposition and a larger fibrotic area in subcutaneous WAT from cKO mice, and these changes were not corrected by nitrate administration (Fig. 2s, t). Consistent with these pathological changes, thermogenic capacity was impaired, as reflected by a reduction in the UCP1-positive area within beige depots (Fig. 2u, v). Concordantly, UCP1 protein and mRNA were diminished, with parallel decreases in thermogenic and browning programs, including Pgc-1α and Cidea, and in mitochondrial biogenesis genes (Fig. 2w and Supplementary Fig. 2h–k). In parallel, nitrate failed to induce either full-length Sialin or the cleaved Sialin2 form in cKO adipose tissue. These nitrate-responsive browning programs were largely lost in cKO adipose tissue (Supplementary Fig. 2h–j). In line with these defects, CS activity in WAT was reduced in cKO mice, and nitrate failed to effectively rescue this reduction, consistent with impaired mitochondrial oxidative capacity upon adipocyte-specific Slc17a5 deletion (Supplementary Fig. 2l). These coordinated deficits link Slc17a5 loss to impaired adipose browning, inflammatory remodeling, and mitochondrial dysfunction.

Taken together, adipocyte-intrinsic Slc17a5 is required to maintain thermogenic programming and metabolic balance under dietary stress and to confer nitrate responsiveness in vivo. The failure of nitrate to improve adipose and metabolic phenotypes in cKO mice establishes Sialin2 as a necessary mediator of nitrate-dependent adipose remodeling (Fig. 2x).

Sialin2 orchestrates mitochondrial fatty acid utilization and adipocyte thermogenesis through LIPA and CPT1A-associated lipid routing

To delineate the intracellular mechanisms by which Sialin2 mediates nitrate-induced browning, we used 3T3-L1 adipocytes with stable Slc17a5 knockdown together with Sialin2 reconstitution. Because cold exposure promotes adipocyte thermogenesis by enhancing mitochondrial fatty acid oxidation through β₃-adrenergic activation³¹, we asked whether Sialin2 drives a comparable metabolic rewiring program in response to nitrate. Untargeted metabolomics and stable isotope tracing revealed that Slc17a5 knockdown created a distinct metabolic bottleneck at the alpha-ketoglutarate to succinate step of the TCA cycle and shifted adipocytes toward glycolytic dependence (Fig. 3a, b, and Supplementary Fig. 3a, b). Seahorse analysis further confirmed that basal respiration, maximal respiration, and ATP-linked oxygen consumption were significantly reduced in Slc17a5 knockdown 3T3-L1 adipocytes, and that these defects were restored by Sialin2 reconstitution (Fig. 3c and Supplementary Fig. 3c).

Fig. 3. Nitrate-Sialin2 promotes lipid mobilization, mitochondrial respiration, and thermogenic remodeling in 3T3-L1 adipocytes.

a Heatmap of ¹³C₆-glucose-derived metabolites in 3T3-L1 adipocytes from the NC (negative control) and sh (Slc17a5 knockdown) groups. b Schematic illustration of ¹³C₆-glucose tracing through glycolysis and the TCA cycle. Created in BioRender. Jiang, O. (2026) https://BioRender.com/r05sup7. c Basal (top), maximal (middle), and ATP-linked (bottom) OCR in 3T3-L1 adipocytes from the NC, sh, sh+Vec (Slc17a5 knockdown+vector), and sh+OE (Slc17a5 knockdown+OE-Sialin2) groups. The corresponding OCR traces are shown in Supplementary Fig. 3c. Representative PLA images (d) and quantification (e) of PLIN1-TOMM20 proximity in 3T3-L1 adipocytes from the NC, sh, sh+Vec, and sh+OE groups. Representative PLA images (f) and quantification (g) of Sialin2-LIPA proximity in 3T3-L1 adipocytes from the Con (control) and Nit (nitrate) groups. h Quantification of Sialin2-CPT1A PLA puncta in 3T3-L1 adipocytes from the Con and Nit groups. The corresponding representative PLA images are shown in Supplementary Fig. 3j. Oil Red O staining (i), quantification of Oil Red O-positive area (j), BODIPY staining (k), and quantification of BODIPY-positive area per cell (l) in 3T3-L1 adipocytes from the NC, sh, sh+Vec, and sh+OE groups. Quantification of intracellular TG (m) and FFA release into the culture medium (n) in the NC, sh, sh+Vec, and sh+OE groups. Basal (o), maximal (p), and ATP-linked (q) OCR together with FAO-associated OCR (r) in 3T3-L1 adipocytes from the NC, sh, sh+Vec, sh+Mut (Slc17a5 knockdown+OE-Sialin (KRI/AAA)), and sh+OE groups. Representative TEM images of mitochondrial morphology (s) and quantification of mitochondrial aspect ratio distribution (t) in 3T3-L1 adipocytes from the NC, sh, sh+Vec, and sh+OE groups. Quantification of PGC-1α (u) and UCP1 (v) protein abundance in HPA-v adipocytes from the NC, sh, sh+Vec, and sh+OE groups. Representative immunoblot data are shown in Supplementary Fig. 4u. w Schematic illustration showing that nitrate-Sialin2 signaling promotes mitochondrial respiration, lipid mobilization, and thermogenic remodeling in adipocytes. Created in BioRender. Jiang, O. (2026) https://BioRender.com/9wm6ww3. For all panels, data are represented as the mean ± SD, and p-values denote a two-sided t-test. Confocal main image scale bar: 20 μm, confocal magnified view scale bar: 5 μm, TEM view scale bar: 0.5 μm. For a–c, m–r, u, v, n = 3 independent experiments per group. For d–h, n = 12 fields from 3 independent experiments. For i–l, n = 18 image fields from 3 independent experiments. For s, t, n = 9 fields from 3 independent experiments.

Because LDs supply fatty acids to mitochondria during thermogenic activation³¹, we next examined whether Sialin2 regulates the lipid droplet and mitochondria interface. Confocal imaging and PLA revealed that nitrate enhanced LD and mitochondria tethering in wild-type adipocytes, that this response was lost after Slc17a5 knockdown, and that it was restored by Sialin2 reconstitution (Fig. 3d, e, and Supplementary Fig. 3d–g). These structural findings suggested that Sialin2 facilitates the mobilization and transfer of fatty acids from LDs to mitochondria. To identify molecular mediators linking lipid mobilization to mitochondrial entry, we performed TurboID proximity labeling and identified lysosomal acid lipase (LIPA) and carnitine palmitoyltransferase 1A (CPT1A)³² as candidate Sialin2-associated proteins that could connect lipid breakdown to mitochondrial utilization (Supplementary Data 1). Their association with Sialin2 was further supported by co-immunoprecipitation and PLA and was enhanced by nitrate stimulation (Fig. 3f–h and Supplementary Fig. 3h–j). These findings support a sequential mechanism in which Sialin2 coordinates triglyceride hydrolysis with mitochondrial fatty acid entry.

Consistent with this model, Slc17a5 knockdown impaired nitrate-induced LD breakdown and promoted intracellular lipid accumulation, whereas Sialin2 reconstitution reversed these defects (Fig. 3i–l and Supplementary Fig. 3k–r). Biochemical quantification further showed that intracellular triglyceride content increased after Slc17a5 knockdown and decreased after Sialin2 reconstitution, consistent with the imaging based lipid accumulation phenotypes (Fig. 3m). Measurement of FFA release into the culture medium further showed that Slc17a5 knockdown impaired lipid mobilization, whereas Sialin2 reconstitution restored this response (Fig. 3n). To assess whether these changes in lipid remodeling translated into enhanced fatty acid oxidation, we performed seahorse analysis under palmitate-supported conditions. Nitrate increased palmitate-dependent oxygen consumption in control adipocytes, whereas Slc17a5 knockdown markedly blunted this response. Sialin2 reconstitution restored palmitate-supported respiration, and this signal was abolished by etomoxir, confirming that the measured OCR reflected mitochondrial fatty acid import and oxidation (Fig. 3o–r). These changes were accompanied by improved mitochondrial network features and mitochondrial functional readouts, including elongation, membrane potential, and oxidative performance, in a Sialin2-dependent manner (Fig. 3s, t, and Supplementary Fig. 4a–l). Functionally, Sialin2 reconstitution restored the expression of thermogenic genes and proteins, including UCP1, PGC-1α, and mitochondrial biogenesis-associated regulators, and promoted overall browning capacity (Fig. 3u, v, and Supplementary Fig. 4m–v).

Importantly, this response was maintained under lipotoxic stress. In palmitate-treated differentiated 3T3-L1 adipocytes, nitrate still reduced lipid accumulation and improved mitochondrial readouts, indicating that the nitrate-Sialin2 axis remains functionally effective in a metabolically stressed context (Supplementary Fig. 5a–d). To directly test whether cleavage-generated Sialin2 is required for this program, we introduced a cleavage-resistant mutant, Sialin (KRI/AAA), into Slc17a5-deficient adipocytes. Unlike Sialin2 reconstitution, Sialin (KRI/AAA) failed to restore nitrate-responsive lipid clearance and did not recover total OCR or fatty acid oxidation-associated OCR (Fig. 3o–r and Supplementary Fig. 5e, f). These findings demonstrate that cleavage-dependent Sialin2 generation is required to couple lipid remodeling with mitochondrial activation and thermogenic output in adipocytes. To determine whether these effects reflected altered canonical β3-adrenergic signaling, we measured intracellular cAMP accumulation following stimulation with the β₃-adrenergic agonist CL316,243. CL316,243 induced a robust increase in cAMP in both control and Slc17a5 knockdown adipocytes, with no significant difference between groups (Supplementary Fig. 5g). These data indicate that proximal β₃-adrenergic signaling remains intact at the level of cAMP generation despite Slc17a5 deficiency, supporting the interpretation that Sialin2 regulates thermogenic remodeling through a pathway that is mechanistically distinct from canonical adrenergic activation.

Together, these findings identify Sialin2 as a key organizer of lipid metabolic flux and mitochondrial functional remodeling during nitrate-induced adipocyte browning. By associating with LIPA and CPT1A, Sialin2 facilitates the mobilization and channeling of LD-derived fatty acids toward mitochondrial oxidation, thereby linking substrate supply to TCA cycle engagement, respiratory activation, and thermogenic reprogramming (Fig. 3w).

Sialin2 promotes ER and mitochondrial coupling and calcium transfer to support mitochondrial function in adipocytes

Building on the observation that Sialin2 enhances fatty acid oxidation and mitochondrial function, we next investigated whether Sialin2 coordinates ER and mitochondria communication to support these metabolic effects. TurboID-based proximity labeling identified a set of Sialin2-associated proteins enriched in pathways related to the ER and mitochondria (Fig. 4a, b). Consistently, high-resolution structured illumination microscopy (HIS-SIM) revealed that nitrate increased the spatial association of Sialin2 with both the ER and mitochondria (Fig. 4c, d). Moreover, HIS-SIM and PLA demonstrated that nitrate significantly increased ER and mitochondria apposition, as reflected by elevated CRT and TOMM20 PLA signals (Fig. 4e, f, and Supplementary Fig. 6a, b). To assess whether Sialin2 is required for this organelle interface, we examined Slc17a5 knockdown 3T3-L1 adipocytes. Transmission electron microscopy revealed a reduced frequency of ER and mitochondria contact sites, which was consistent with the decreased interorganelle proximity detected by HIS-SIM and PLA (Fig. 4g–l and Supplementary Fig. 6c). Reintroduction of Sialin2 restored ER and mitochondria connectivity in Slc17a5 knockdown adipocytes (Fig. 4g–l), indicating that Sialin2 is required to maintain this interorganelle coordination. Consistent with these cell-based observations, WAT from cKO mice exhibited reduced ER and mitochondria apposition, as reflected by decreased IP3R and VDAC PLA signal intensity, and this defect was not rescued by nitrate treatment in the cKO background (Supplementary Fig. 6d). These data extend the cellular findings to adipose tissue in vivo and support the conclusion that adipocyte Slc17a5 deficiency disrupts MAM organization and abolishes the corresponding nitrate response.

Fig. 4. Sialin2 enhances endoplasmic reticulum and mitochondrial interactions and promotes mitochondrial Ca²⁺ uptake in 3T3-L1 adipocytes.

a Schematic illustration of the Sialin2-TurboID proximity labeling workflow. Created in BioRender. Jiang, O. (2026) https://BioRender.com/slwg2ui. b Top 15 enriched GO pathways for Sialin2-associated proteins in 3T3-L1 cells, ranked by p-value based on DAVID analysis. c Confocal images of mCherry-Sialin2 together with ERTracker-labeled endoplasmic reticulum or MitoTracker-labeled mitochondria in 3T3-L1 adipocytes from the Con and Nit groups. d Colocalization analysis of Sialin2 with the endoplasmic reticulum or mitochondria, quantified by Pearson’s correlation coefficient (left) and line-scan plot profile (right). e Colocalization analysis of the endoplasmic reticulum and mitochondria, quantified by Pearson’s correlation coefficient (left) and line-scan plot profile (right). Representative images are shown in Supplementary Fig. 6a. f Quantification of PLA puncta for CRT-TOMM20 in 3T3-L1 adipocytes from the Con and Nit groups. Representative images are shown in Supplementary Fig. 6b. Representative TEM images of endoplasmic reticulum and mitochondria contact sites (g), quantification of contact length normalized to mitochondrial perimeter (h), HIS-SIM images of ERTracker-labeled endoplasmic reticulum together with MitoTracker-labeled mitochondria (i), and colocalization analysis (j) in 3T3-L1 adipocytes. Representative PLA images of CRT-TOMM20 proximity (k) and quantification of PLA puncta per cell (l) in HPA-v adipocytes. Representative traces (m) and quantitative analysis (n) of ATP-stimulated mitochondrial Ca²⁺ accumulation in 3T3-L1 adipocytes. Quantification of basal mitochondrial Ca²⁺ levels and corresponding time-lapse images are shown in Supplementary Fig. 7a, b. o Schematic illustration showing that nitrate-Sialin2 signaling enhances endoplasmic reticulum and mitochondria interactions and promotes calcium flux from the endoplasmic reticulum to mitochondria in adipocytes. Created in BioRender. Jiang, O. (2026) https://BioRender.com/pezcfs8. For all panels, data are represented as the mean ± SD, and p-values denote a two-sided t-test. Confocal image scale bar: 20 μm, confocal magnified view scale bar: 5  μm, HIS-SIM main view scale bar: 10 μm, HIS-SIM magnified view scale bar: 2 μm, TEM view scale bar: 0.5 μm. For b, m and n, n = 3 independent experiments per group. For d–f, j: n = 18 fields from 3 independent experiments. For h, n = 9 fields from 3 independent experiments. For l, n = 12 from 3 independent experiments.

Because ER and mitochondria coupling facilitates calcium transfer³³, we next monitored mitochondrial matrix calcium uptake using a mitochondria-targeted 4mtD3cpv probe (Supplementary Fig. 6e). In wild-type adipocytes, nitrate enhanced ATP-triggered mitochondrial Ca²⁺ uptake (Supplementary Fig. 6f–h). Across the indicated genetic conditions, Slc17a5 knockdown reduced ATP-evoked mitochondrial Ca²⁺ uptake capacity, whereas Sialin2 reconstitution restored it (Fig. 4m, n, and Supplementary Fig. 7a). These findings indicate that Sialin2 promotes efficient Ca²⁺ transfer across the ER and mitochondria interface. Baseline Ca²⁺ levels remained unchanged (Supplementary Fig. 7b), excluding nonspecific calcium overload. Together, these data indicate that nitrate-responsive Sialin2 signaling strengthens ER and mitochondria coupling, enhances mitochondrial Ca²⁺ transfer, and supports mitochondrial activity (Fig. 4o).

Sialin2 couples ER and mitochondria calcium transfer with fatty acid metabolism through ACSL3 to drive adipocyte browning

Beyond regulating calcium dynamics, we further investigated whether Sialin2 links ER and mitochondria communication to fatty acid metabolism. TurboID and PLA assays identified a nitrate-enhanced association between Sialin2 and acyl-CoA synthetase long chain family member 3 (ACSL3) (Fig. 5a, b, and Supplementary Data 1), an ER-resident enzyme that catalyzes the conversion of FFAs into acyl-CoA, a key step in lipid utilization³⁴. These findings suggested that Sialin2 may coordinate calcium and lipid flux across organelles.

Fig. 5. Artificial ER-mitochondria tethering rescues calcium transfer and metabolic defects in Slc17a5-deficient 3T3-L1 adipocytes.

Representative PLA images of Sialin2-ACSL3 proximity (a) and quantification of PLA puncta per cell (b) in 3T3-L1 adipocytes from the Con and Nit groups. Representative PLA images of CRT-TOMM20 proximity (c) and quantification of PLA puncta per cell (d) in 3T3-L1 adipocytes from the NC, NC+Linker, sh, and sh+Linker groups. Representative Oil Red O staining images (e), quantification of Oil Red O-positive area (f), representative BODIPY staining images (g), and quantification of BODIPY-positive area per cell (h). Representative MitoTracker staining images (i), quantification of MitoTracker fluorescence intensity (j), representative JC-1 staining images (k), and quantification of the JC-1 aggregate to monomer fluorescence ratio (l). Quantitative analysis of basal (m), maximal (n), and ATP-linked (o) OCR. p Schematic illustration showing that Sialin2 supports adipocyte thermogenic remodeling by maintaining ER-mitochondria communication, calcium transfer, and metabolic function. Created in BioRender. Jiang, O. (2026) https://BioRender.com/ykuu8z9. For all panels, data are represented as the mean ± SD, and p-values denote a two-sided t-test. PLA and BODIPY image scale bars, 20 μm. Oil Red O, MitoTracker, and JC-1 main image scale bars, 50 μm. MitoTracker and JC-1 inset scale bars, 5 μm. For b, d, n = 18 fields from 3 independent experiments. For f, h, j,l, n = 12 fields from 3 independent experiments. For m–o, n = 3 independent experiments per group.

To determine whether forced ER and mitochondria tethering could compensate for the defect caused by Sialin2 deficiency, we expressed an artificial ER and mitochondria linker, mAKAP1(34–63)-mRFP-yUBC6³⁵, in control and Slc17a5 knockdown adipocytes. HIS-SIM and PLA analyses showed that linker expression produced only limited changes in ER and mitochondria proximity in control adipocytes, but restored the reduced interorganelle apposition induced by Slc17a5 knockdown (Fig. 5c, d, and Supplementary Fig. 8a–c). Consistent with this, mitochondrial Ca²⁺ uptake after ATP stimulation was substantially rescued by linker expression in Slc17a5 knockdown adipocytes, whereas the effect in control cells remained comparatively modest (Supplementary Fig. 8d–g).

We next examined the downstream metabolic consequences of restoring ER and mitochondria coupling. In Slc17a5 knockdown adipocytes, linker expression reduced lipid accumulation, as shown by Oil Red O and BODIPY staining, whereas its effect in control adipocytes remained limited (Fig. 5e–h). Linker expression also restored mitochondrial morphology and membrane potential in Slc17a5 knockdown cells, as reflected by MitoTracker and JC-1 analyses (Fig. 5i–l). In parallel, seahorse analysis showed recovery of basal respiration, maximal respiration, and ATP-linked oxygen consumption (Fig. 5m–o), and immunoblotting further demonstrated restoration of PGC-1α and UCP1 expression (Supplementary Fig. 8h). Together, these results indicate that enforced ER and mitochondria tethering preferentially compensates for the defects caused by Sialin2 loss rather than broadly enhancing mitochondrial function in otherwise intact adipocytes.

Together, these findings define Sialin2 as a critical nitrate-responsive sensor that links ER and mitochondrial calcium signaling with fatty acid activation and mitochondrial metabolism. By sustaining ER-mitochondria coupling and integrating ACSL3-associated lipid utilization with mitochondrial oxidative function, Sialin2 promotes thermogenesis and adipocyte browning (Fig. 5p).

Sialin2 promotes browning through the IP3R1-VDAC1-MCU1 calcium shuttle

To elucidate how Sialin2 engages mitochondrial calcium transfer during nitrate-induced adipocyte browning, we focused on the canonical IP3R1-VDAC1-MCU1 axis, which mediates ER and mitochondrial calcium transfer across contact sites. TurboID-based interactome profiling identified IP3R1 as a candidate Sialin2-associated protein, and nitrate stimulation markedly enhanced the association between Sialin2 and IP3R1 (Fig. 6a–c). Furthermore, nitrate strengthened IP3R1-VDAC1 coupling (Fig. 6d, e, and Supplementary Fig. 9a). Consistent with these findings, immunofluorescence analyses showed increased spatial association of Sialin2 with IP3R1 and of IP3R1 with VDAC1 after nitrate stimulation, supporting formation of a nitrate-responsive ER and mitochondria calcium transfer microdomain (Supplementary Fig. 9b, c). In contrast, Slc17a5 knockdown disrupted IP3R1-VDAC1 coupling, indicating that Sialin2 is required for proper organization of this ER and mitochondria calcium transfer interface (Fig. 6f and Supplementary Fig. 9d).

Fig. 6. Inhibition of IP3R1 or MCU1 disrupts Sialin2-mediated thermogenic activation in 3T3-L1 adipocytes.

Representative co-IP images (a) and quantitative analysis (b) of the association of Sialin2 with IP3R1 or VDAC1 in 3T3-L1 adipocytes treated with or without nitrate (4 mM). Representative data from N = 3 or 4 independent experiments are shown. Representative PLA images and quantification of Sialin2-IP3R1 proximity (c) and Sialin2-VDAC1 proximity (d). Representative PLA images and quantification of IP3R1-VDAC1 proximity in HPA-v adipocytes from the Con and Nit groups (e) or the NC and sh groups (f). Representative traces (g) and quantitative analysis (h) of ATP-stimulated mitochondrial Ca²⁺ accumulation in 3T3-L1 adipocytes from the Vec, OE, OE + 2-APB (OE-Sialin2 + 2-APB), and OE + RR (OE-Sialin2+ruthenium red) groups (N = 3 independent experiments). Quantification of basal mitochondrial Ca²⁺ levels and corresponding time-lapse images are shown in Supplementary Fig. 10a, b. Representative BODIPY staining images (i) and quantification of BODIPY-positive area per cell (j). k Quantification of lipid droplet area. Representative Oil Red O staining images are shown in Supplementary Fig. 10c. Representative MitoTracker staining images (l) and quantification of fluorescence intensity (m). n Quantification of JC-1 aggregate to monomer fluorescence ratio (n). Representative images are shown in Supplementary Fig. 10d. o Mitochondrial respiration quantitative analysis of basal, maximal, and ATP-linked OCR. Representative mitochondrial respiration traces are shown in Supplementary Fig. 10e. p Quantitative analysis of PGC-1α and UCP1 protein abundance. Representative immunoblot images and quantitative analysis of PGC-1α are shown in Supplementary Fig. 10f, g. q Schematic illustration showing that Sialin2 drives adipocyte thermogenic remodeling by orchestrating ER and mitochondria communication through the IP3R1-VDAC1-MCU1 calcium axis. Created in BioRender. Jiang, O. (2026) https://BioRender.com/nof2d92. For all panels, data are represented as the mean ± SD, and p-values denote a two-sided t-test. PLA and BODIPY image scale bar: 20 μm, MitoTracker main view scale bar: 50 μm, MitoTracker magnified view scale bar: 5 μm. For a, b, n = 3 or 4 independent experiments. For c, d, i–n, n = 18 fields from 3 independent experiments. For e, f, n = 12 fields from 3 independent experiments. For g, h, o, p, n = 3 independent experiments.

We next tested whether this calcium signaling cascade extends to the mitochondrial matrix through the mitochondrial calcium uniporter MCU1, which is located in the inner mitochondrial membrane. Functional imaging revealed that ATP-induced mitochondrial calcium uptake was significantly elevated by Sialin2 overexpression, and this enhancement was abolished by inhibition of IP3R1 with 2-APB or of MCU1 with ruthenium red (Fig. 6g, h, and Supplementary Fig. 10a, b). These results indicate that Sialin2 is sufficient to enhance inducible mitochondrial Ca²⁺ uptake, and that this gain-of-function requires an intact IP3R1-VDAC1-MCU1 conduit. Disruption of this axis had broad consequences for adipocyte function. Both 2-APB and ruthenium red abolished the Sialin2-mediated reduction in intracellular lipid accumulation (Fig. 6i–k and Supplementary Fig. 10c), suppressed the Sialin2-induced improvements in mitochondrial readouts (Fig. 6l–o and Supplementary Fig. 10d, e), and attenuated the expression of UCP1 and other thermogenic markers (Fig. 6p and Supplementary Fig. 10f, g). Thus, the overexpression phenotype does not reflect a nonspecific effect of Sialin2 abundance. Instead, it reveals a pathway-dependent program that requires ER and mitochondria Ca²⁺ transfer. These findings establish the IP3R1-VDAC1-MCU1 calcium shuttle as a critical conduit through which Sialin2 promotes mitochondrial activation, fatty acid oxidation, and thermogenic reprogramming.

Together, these results reveal that nitrate strengthens the upstream engagement of Sialin2 with ER and mitochondria calcium-transfer machinery, whereas Sialin2 gain-of-function is sufficient to drive downstream thermogenic remodeling through the same IP3R1-VDAC1-MCU1 axis (Fig. 6q).

Discussion

Adipose tissue remodeling is central to metabolic adaptation in obesity^36–39, yet the intracellular mechanisms that couple nutrient sensing to beige thermogenic programming remain incompletely understood. In this study, we identify a nitrate-responsive Sialin2 pathway that links extracellular nitrate availability to compartmentalized organelle signaling in adipocytes. We show that dietary nitrate promotes adipose browning and improves systemic metabolic parameters under high-fat diet challenge, whereas adipocyte-specific Slc17a5 deletion abolishes these effects. Mechanistically, Sialin2 promotes ER and mitochondria coupling, enhances inducible mitochondrial Ca²⁺ uptake through the IP3R1-VDAC1-MCU1 axis, and coordinates lipid droplet-derived fatty acid routing into mitochondrial oxidation. These coordinated actions support mitochondrial activation, thermogenic gene expression, and acquisition of a beige adipocyte phenotype. Together, these findings define a mechanistically integrated nitrate-sensing framework for adipocyte browning and expand current views of metabolic regulation toward organelle-centered homeostatic control.

Mechanistically, the main contribution of this study is the identification of Sialin2 as a nodal organizer that couples two processes that are both required for thermogenesis but are often discussed separately. One process is mitochondrial activation through the ER and mitochondria Ca²⁺ transfer. The other is sustained substrate supply through lipid droplet-derived fatty acid mobilization and mitochondrial oxidation. Previous studies established that MAM-associated Ca²⁺ exchange regulates mitochondrial respiration, ATP production, and lipid oxidation, and that disruption of ER and mitochondria Ca²⁺ transfer impairs adipocyte thermogenic competence and energy expenditure^21,22,40,41. Our findings extend this framework by showing that a nitrate-responsive Sialin2 module lies upstream of this machinery and enhances inducible rather than basal mitochondrial Ca²⁺ uptake. At the same time, Sialin2 associates with LIPA, ACSL3, and CPT1A in a manner that functionally links local lipolysis, acyl-CoA generation, and mitochondrial entry of oxidation substrates^42–45. These observations support a model in which Sialin2 does not simply activate a calcium pathway or a lipid pathway in isolation. Instead, it synchronizes mitochondrial responsiveness with fuel availability so that mitochondrial activation and oxidative throughput rise together. This feature is particularly important in thermogenic adipocytes, where enhanced Ca²⁺ signaling without adequate substrate delivery would be inefficient, whereas fuel mobilization without sufficient mitochondrial activation would fail to generate a robust beige program^46,47.

This interpretation also helps place the present study in the broader context of thermogenic regulation. Canonical β-adrenergic signaling drives browning through cAMP and PKA-dependent mechanisms and promotes lipolysis and thermogenic gene expression, but its therapeutic exploitation remains constrained by systemic actions and cardiovascular liabilities⁴⁴. A distinct model has emphasized UCP1-independent heat generation through SERCA-dependent Ca²⁺ cycling in the ER⁴⁸. Our data support a different principle. The Sialin2 pathway is centered on organelle-confined signaling that enhances IP3R1-VDAC1-MCU1-mediated Ca²⁺ transfer and couples this event to substrate oxidation within mitochondria⁴⁹. In this sense, it is spatially restricted, metabolically efficient, and mechanistically different from diffuse hormonal activation or ATP-consuming futile cycling. The preserved cAMP response in Slc17a5-deficient adipocytes further argues that Sialin2 does not act by changing proximal adrenergic responsiveness. Instead, it defines a nitrate-responsive layer of control that converges on mitochondrial metabolism at the level of organelle communication. Our findings do not exclude contributions from the classical nitrate-nitrite-NO axis in other physiological settings, including vascular or microbiome-related nitrate metabolism^25–27. Rather, they define an adipocyte intrinsic, Sialin2-dependent mechanism that is sufficient to explain the major phenotypes observed under the experimental conditions used here. This distinction is conceptually important because it suggests that beige programming can be enhanced by tuning organelle interfaces and substrate routing rather than by globally increasing adrenergic drive.

Viewed more broadly, this work also positions Sialin2 as an organelle-embedded nitrate sensor rather than a conventional cytosolic metabolic sensor. Cytosolic systems, such as AMPK and mTOR, monitor cellular energetic state and nutrient sufficiency at the whole cell level^50,51, whereas Sialin2 appears to operate in spatially restricted membrane environments. This feature places it closer to emerging classes of organelle-associated sensing modules, such as PERK at the ER and SLC38A9 at the lysosome^52,53. Our data further suggest that Sialin2 is distributed across multiple organelle-associated pools and engages distinct molecular partners at different interfaces, including ER and mitochondria contact sites and lipid handling compartments. This organization provides a plausible explanation for how one nitrate-responsive factor can coordinate Ca²⁺ transfer and substrate routing without requiring a single static complex containing all partners simultaneously. It also raises the possibility that nitrate-responsive Sialin2 signaling may have consequences beyond acute bioenergetics. Because α-ketoglutarate, succinate, and related TCA intermediates influence chromatin-modifying enzymes, the Sialin2-dependent rewiring of oxidative metabolism may intersect with transcriptional and epigenetic control of adipocyte identity^54,55. That possibility remains speculative in the present study, but it provides a compelling framework for future work on how metabolic flux is translated into stable beige remodeling⁵⁶.

Several limitations should be acknowledged. First, the spatial organization of Sialin2 relative to the IP3R1-VDAC1-MCU1 conduit and to the lipid routing proteins LIPA, ACSL3, and CPT1A remains incompletely defined. Higher resolution live cell imaging, FRET-based distance measurements, and cross-linking mass spectrometry will be needed to resolve these interfaces more precisely⁵⁷. Second, our study focuses primarily on subcutaneous WAT, leaving open whether similar nitrate-responsive Sialin2 programs operate in brown fat, skeletal muscle, liver, or other metabolically active tissues. This issue is particularly relevant because nitrate supplementation under lean control diet conditions did not clearly elevate systemic nitrate or induce Sialin2 in vivo, making the contribution of Sialin2 to classical brown adipocyte function unresolved in the current dataset. Although this does not exclude a role in brown fat, it suggests that Sialin2-dependent effects may become more evident under conditions of greater nitrate exposure or thermogenic demand. Third, although our data support a coordinated role for Sialin2 in calcium transfer and lipid oxidation, additional components, such as GRP75, MICU1, EMRE, ATGL, and HSL, may further shape this pathway and should be tested through staged genetic or pharmacologic perturbation combined with isotope-based flux analysis^44,58. Finally, because the present work was performed in male mice and did not examine cold exposure paradigms, future studies should evaluate sex dependence, thermogenic context, and dose-relevant nitrate exposure to strengthen physiological interpretation and translational relevance.

In conclusion, our findings identify Sialin2 as a nitrate-responsive organelle sensor that links extracellular inorganic anion availability to adaptive thermogenic remodeling in adipocytes (Fig. 7). Through coordinated control of organelle communication and metabolic flux, the nitrate Sialin2 axis enables WAT to maintain energetic flexibility under metabolic stress. These findings extend current models of adipose regulation by introducing an organelle-centered mechanism for homeostatic control. More broadly, they suggest that inorganic anion sensing through Sialin2 may represent a general strategy for preserving metabolic homeostasis and provide a mechanistic basis for developing homeostasis-oriented interventions in obesity.

Fig. 7. Dietary nitrate attenuates obesity through Sialin2-dependent coordination of mitochondrial calcium transfer and lipid metabolism.

Following dietary intake, nitrate enters adipocytes through Sialin-mediated transport and engages a Sialin2-dependent thermogenic program. In this model, Sialin2 coordinates two functionally coupled arms of adipocyte remodeling. One arm enhances ER-mitochondria communication by strengthening the IP3R1-VDAC1-MCU1 calcium-transfer axis, thereby increasing inducible mitochondrial Ca²⁺ uptake and supporting mitochondrial activation, tricarboxylic acid cycle activity, and UCP1-linked thermogenesis. The other arm promotes lipid mobilization and utilization by associating with LIPA, ACSL3, and CPT1A, thereby facilitating fatty acid release, acyl-CoA formation, and mitochondrial fatty acid oxidation. Through this coordinated control of calcium transfer and substrate routing, the nitrate-Sialin2 axis reduces lipid storage, promotes browning of white adipose tissue, enhances energy expenditure, and improves systemic metabolic homeostasis. Created in BioRender. Jiang, O. (2026) https://BioRender.com/6ow88g4.

Methods

Mice

C57BL/6J mice were used for all experiments (The Jackson Laboratory) and bred under specific-pathogen-free conditions at the Capital Medical University animal facility. Adipose-specific Slc17a5 knockout (cKO) mice were generated by crossing Fabp4-Cre mice (C57BL/6 background; Cyagen, #C001442) with mice harboring loxP-flanked Slc17a5 alleles. F1 offspring were genotyped and verified for germline transmission, and homozygous Cre-positive mice were used for experiments. All procedures were approved by the Animal Care and Use Committee of Capital Medical University (protocol AEEI-2024-053) and were performed in accordance with the NIH Guide for the Care and Use of Laboratory Animals.

All experiments were performed using 8- to 10-week-old male mice (n ≥ 6 per group for each independent experiment) to avoid confounding effects of female sex hormones on adipose metabolism and obesity-related phenotypes. Mice were maintained at 18–24 °C with 40-60% humidity under a 12 h light/dark cycle (06:00–18:00), with ad libitum access to food and water. Mice were fed either a control diet (Research Diet, #D10012M) or a HFD (Research Diet, #D12492) for 8 weeks, with or without 4 mM nitrate in the drinking water. Body weight was recorded weekly. An OGTT was performed after 7 weeks of treatment in freely moving mice. After 8 weeks, mice were anesthetized with isoflurane following a 6 h fast. Portal and cava vein blood samples were collected, mice were euthanized by cervical dislocation, and tissues were dissected, weighed, snap-frozen in liquid nitrogen, and stored at −80 °C for subsequent analyses.

Indirect calorimetry

Whole-body energy metabolism was assessed by indirect calorimetry using a comprehensive laboratory animal monitoring system (TSE, PhenoMaster). Mice were individually housed in metabolic chambers under a 12 h light/12 h dark cycle with free access to food and water. After an acclimation period of 24 h, oxygen consumption, carbon dioxide production, energy expenditure, respiratory exchange ratio, locomotor activity, and food intake were recorded continuously for 24 h. Energy expenditure and respiratory exchange ratio were calculated automatically by the instrument software based on oxygen consumption and carbon dioxide production. Locomotor activity was quantified by infrared beam breaks, and food intake was monitored throughout the recording period. Data were analyzed over the entire recording period and, where indicated, separately for the light and dark phases. For thermoneutral experiments, mice were transferred to 30 °C after completion of the HFD intervention, and indirect calorimetry was performed under otherwise identical conditions.

Sirius Red staining and fibrosis quantification

Subcutaneous WAT was collected and fixed in 4% paraformaldehyde, followed by routine tissue dehydration, clearing, paraffin infiltration, and embedding. Samples were cut into 4 μm thick paraffin sections. For deparaffinization, sections were immersed in xylene for 5–10 min, repeated once with fresh xylene, then sequentially rehydrated in absolute ethanol, 90% ethanol, 80% ethanol, and 70% ethanol, and rinsed with distilled water. Sections were circled with a PAP pen, stained with hematoxylin for 10 min, rinsed with distilled water, differentiated with hydrochloric acid-ethanol solution for 30 s, and rinsed with running tap water for 10 min for bluing. Subsequently, sections were stained with Sirius Red working solution for 10–15 min and briefly rinsed with distilled water within 10 s. Sections were rapidly dehydrated through a graded ethanol series, cleared three times in xylene for 1–2 min each, and mounted with neutral balsam. Bright-field images were acquired under identical exposure settings using the Pannoramic SCAN DX150 with Pannoramic Viewer 2.10. Fibrosis was quantified by measuring the Sirius Red-positive area and normalizing it to the total tissue area using ImageJ 1.52v.

Hematoxylin and eosin staining and crown-like structure quantification

Paraffin-embedded subcutaneous WAT sections (4 μm) were deparaffinized, rehydrated, and stained with hematoxylin and eosin (H&E) using standard protocols. Images were acquired using Pannoramic SCAN DX150 with Pannoramic Viewer 2.10. Crown-like structures (CLS), defined as macrophage-rich cellular aggregates surrounding individual adipocytes on H&E-stained sections, were counted manually in a blinded manner. CLS density was calculated as the number of CLS per unit adipose tissue area using ImageJ 1.52v.

Triglycerides and fasting insulin measurement

After a 12-h fasting period, mice were subjected to blood samples collection into 1.5 mL tubes, followed by incubation at room temperature for 2 h. The blood was then centrifuged at 2000 × g for 20 min, and the serum was collected and stored at −80 °C. Triglycerides (TG) were measured using commercial kits (Nanjing Jiancheng, #E1025-105) in accordance with the manufacturer’s instructions. Fasting insulin was measured using commercial kits (Crystal Chem, #90080).

Glucose metabolism testing

For oral glucose tolerance testing (OGTT), all mice were placed in clean cages and provided with water but without food. After 6 h of fasting, the mice were administered an oral glucose solution (2 g/kg; Sigma-Aldrich, #158968). Blood glucose levels were measured before (fasting glucose) and after oral glucose administration at 30, 60, 90, and 120 min and were determined using a glucose meter (Accu-Chek Performa, Roche) on blood samples collected from the tip of the tail vein during week 7. HOMA-IR = (glucose × insulin)/405.

Citrate synthase activity assay

CS activity in WAT was measured using a commercial assay kit (COIBO BIO, #CB13548-Mu) according to the manufacturer’s instructions. Adipose tissue was homogenized, and the resulting lysates were used for enzymatic activity measurement. Protein concentration was determined by BCA assay, and CS activity was normalized to total protein content. Absorbance was measured with a microplate reader (Molecular Device, SpectraMax i3x) with SoftMax Pro 7.3.1. CS activity was analyzed in the indicated dietary and genetic cohorts.

Immunoprecipitation and immunoblotting

Cells were removed from the medium after the indicated treatment, washed once with ice-cold PBS, and lysed in an ice-cold RIPA lysis buffer system (Santa Cruz Biotechnology, #sc-24948) with 1 mM Na₃VO₄, 5 mM NaF, and 1× protease inhibitor cocktail (Sigma-Aldrich, #P8340). The cell lysates were then sonicated at 15% amplitude for 15 s. After sonication, the cell lysates were incubated at 4 °C with continuous rotation for 1 h and subsequently centrifuged at 16,000 × g for 10 min to collect the supernatant. The Pierce BCA Protein Assay (Thermo Fisher Scientific, #A65453) was used to measure the protein concentration in the supernatant, following the manufacturer’s instructions. Equal amounts of protein were used for further analysis.

For immunoprecipitation, 0.5–1 mg of cell lysate protein was incubated with 20 µL of antibody-conjugated agarose (Santa Cruz Biotechnology) at 4 °C overnight. For endogenous FLAG-tagged Sialin2 immunoprecipitation, 0.5–1 mg of cell lysates were incubated with anti-FLAG (Santa Cruz Biotechnology, #sc-166355AC) mouse monoclonal IgG antibody-conjugated agarose at 4 °C for 24 h. Normal immunoglobulin (IgG)-conjugated agarose was used as a negative control (Santa Cruz Biotechnology, #sc-2343). After washing five times with lysis buffer, the protein complex was eluted with SDS sample buffer.

For immunoblotting of immunoprecipitated complexes, horseradish peroxidase (HRP)-conjugated antibodies were used to avoid nonspecific detection of immunoglobulin in the immunoprecipitated samples. HRP-conjugated FLAG (Proteintech, #HRP-66008) antibodies were purchased.

For immunoblotting, 5–20 µg of protein was loaded. Equal amounts of total protein was separated on a 4–20% precast polyacrylamide gel (Bio-Rad, #4561096) and transferred onto 0.2 µm nitrocellulose membranes using the Trans-Blot Turbo system. The membrane was blocked with StartingBlock Blocking Buffer (Thermo Fisher Scientific, #37542) for 30 min at room temperature, and then incubated with primary antibody at 4 °C overnight. The membrane was then washed and incubated with horseradish peroxidase-conjugated secondary antibodies at room temperature for 1 h. The membrane was washed and exposed to SuperSignal West Pico PLUS Chemiluminescent Substrate (Thermo Fisher Scientific, #34577) in a ChemiDoc Imaging System (Bio-Rad). The intensity of protein bands was quantified using ImageJ 1.52v. The unsaturated immunoblot images was used for quantification with the appropriate loading controls as standards. Statistical data analysis was performed with Microsoft Excel using data from at least three independent experiments.

Quantitative real-time PCR (qRT-PCR)

Total RNA was prepared from cells using TRIzol reagents (Invitrogen, #15596018CN) according to the manufacturer’s instructions. RNA samples were reverse-transcribed using the iScript cDNA Synthesis Kit (Bio-Rad) according to the provided protocol. Quantification of gene transcripts was performed by qPCR using the CFX96 Touch Real-time PCR detection system (Bio-Rad CFX Manager 3.1). β-actin served as an internal control. A list of the PCR primers used to amplify the target genes is provided in Supplementary Table 1.

Cell culture and differentiation

The 3T3-L1 embryonic fibroblasts (ATCC, #CL-173) were cultured in DMEM medium (Corning, #10-013-CV) containing 10% calf serum (Gibco, #16010159) at 37 °C with 5% CO₂. The cell line was quarterly authenticated to ensure it was free of mycoplasma contamination. Sex was not considered as an experimental variable in the cell-based experiments performed in this study. For the differentiation of 3T3-L1 preadipocytes, when cell confluence reached 100%, prepared induction medium (90% DMEM, 10% calf serum, 0.5 mM 3-isobutyl-1-methylxanthine (Merck, I5879), 1 μM dexamethasone (Merck, D4902), 1 μM rosiglitazone (Merck, R2408), and 1 μM insulin (MCE, #HY-P0035)) was added, cells were cultured for 3 days, and then switched to maintenance medium (90% DMEM containing 10% calf serum, 1 μM rosiglitazone, 100 nM triiodothyronine and 1 μM insulin) for 4 days. Subsequently, the culture medium was changed to DMEM medium containing 10% calf serum for continued maintenance, with fluid changes every 2 days, until 3T3-L1 preadipocytes were induced into mature adipocytes on day 8–10. Cells were pretreated with 2-APB (2-aminoethyl diphenylborinate, 2 μM) or ruthenium red (0.5 μM) for 24 h before the indicated stimulation or functional analysis.

The HPA-v cells (Procell, #CP-H114) were cultured in DMEM/F12 medium containing 10% fetal bovine serum (FBS, Hyclone, #SH30910.03), 1% penicillin/streptomycin (Gibco, #15140-122), supplemented with basic fibroblast growth factor (bFGF) (4 ng/mL) (MCE, #HY-P5321) at 37 °C with 5% CO₂. For differentiation, HPA-v cells were grown and then switched to growth media without bFGF until they became 100% confluent. At 48 h after achieving 100% confluence, we initiated differentiation. Before induction, cells were incubated in pre-induction medium (DMEM/F12 medium supplemented with 2% FBS, 1% antibiotics, 0.5 μM human insulin), 2 nM triiodothyronine (T3, MCE, #HY-A0070A), and 3.3 nM bone morphogenetic protein 7 (BMP7, MCE, #HY-P7008) for 6 days, then switched to Induction Medium 1 (DMEM/F12 medium supplemented with 2% FBS, 1% antibiotics, 0.5 μM human insulin, 0.1 μM dexamethasone, 0.5 mM 3-isobutyl-1-methylxanthine), 33 μM biotin (MCE, HY-B0511), 2 nM T3, 30 μM indomethacin (MCE, #HY-14397), 17 μM pantothenate (MCE, #HY-B0430), 2 μM rosiglitazone for 7 days followed by Induction Medium 2 (Induction Medium 1 without rosiglitazone) until full adipocyte maturation (typically day 25 to day 30).

Plasmids, lentivirus, transfection, and stable cell line construction

For stable expression of proteins in this study, all DNA sequences were of human origin unless otherwise specified. Site-directed mutagenesis (MCLAB) was employed to generate all mutants.

The cDNA for Sialin2 was inserted into the lentivirus vector GV341 (pGC-FU-3FLAG-SV40-puromycin-library) by restriction enzymes using AgeI/NheI and also into the lentivirus vector GV287 (pGC-FU-3FLAG-SV40-EGFP) by restriction enzymes using AgeI. The cDNA for mCherry-Sialin2 was inserted into vector GV348 (pGC-FU-3FLAG-SV40-puromycin-library) by enzymatic digestion using AgeI and EcoRI. The fusion of mCherry with Sialin2 was achieved using the following primers: forward, 5′-caaggaggtggaggatcaatggtgagcaagg-3′ and reverse, 5′-attgatcctccacctccttgtactccagcc-3′. To generate knockdown cell lines, shSLC17A5 (target seq: TGTGAATCTGAGTGTTGCGTTAGTGGATA) was inserted into the lentivirus vector GV493 (hU6-MCS-CBh-gcGFP-IRES-puromycin) by restriction enzymes using AgeI/EcoRI. The lentivirus was purchased from Shanghai Genechem Co., Ltd. Target cells were infected with lentiviruses for stable cell line generation and selected with puromycin (1–2 μg/ml) for 5–7 days. To generate the Sialin2-TurboID construct (pUC-Sialin2-FLAG-TurboID), the Sialin2-FLAG sequence was amplified from Addgene Plasmid 34831 and fused with the TurboID sequence into pCDH-CMV-MCS.

To construct the OMM-ER linker³⁵, mRFP was targeted to the ER by using the C-terminal ER localization sequence of the yeast UBC6 protein (X73234, residues 233–250: MVYIGIAIFLFVGLFMK), through the linker (SGLRSRAQ-ASNSRV). This construct was complemented with the N-terminal mitochondrial localization sequence of the mouse AKAP1 protein (V84389, residues 34–63: MAIQLRSLFPLALPGLLALLGWWWFFSRKK) with the linker (DLELKLRILQSTVPRARDPPVAT).

The pcDNA-4mtD3cpv ⁵⁹plasmid was kindly provided by Dr. Qiaochu Wang, who is a fellow at Beijing Children’s Hospital, Capital Medical University. All newly generated constructs are available upon request from the corresponding author.

BODIPY and Oil Red O staining

Cells were fixed with 4% paraformaldehyde for 15 min and then washed twice with PBS. After rinsing with 60% isopropanol (Sigma-Aldrich, #PX1834), cells were stained with a filtered Oil Red O working solution (Servicebio, #G1015) for 20 min at room temperature. Cells were washed twice with PBS and then photographed.

For BODIPY staining, cells were washed with PBS after fixation and stained with 1X BODIPY (Beyotime, #C2053S) for 15 min, nuclei were counterstained with 1 µg/mL 4′,6-diamidino-2-phenylindole (DAPI) (Invitrogen, #D3571) for 5 min at room temperature in the dark. The images were captured using a Nikon AX confocal microscope controlled by NIS-Elements AR 5.42.06 (Nikon). For Oil Red O staining, cells were washed extensively with ddH₂O after fixation and then stained with the Oil Red O working solution at room temperature for 30 min. A Zeiss microscope acquired images. The lipid droplet area was measured with ImageJ 1.52v.

Intracellular triglyceride assay

Total intracellular triglyceride (TG) content was measured in differentiated adipocytes using a commercial triglyceride assay kit (Elabscience, #E-BC-K261-M). Briefly, cells were washed with PBS and lysed in the supplied extraction buffer. Cell lysates were collected for TG quantification, and protein concentration was determined by BCA assay. TG content was measured colorimetrically using a microplate reader (Molecular Devices, SpectraMax i3x) with SoftMax Pro 7.3.1 and normalized to total cellular protein.

Free fatty acid release assay

FFA release into the culture medium was quantified using a commercial non-esterified fatty acid assay kit (Elabscience, #E-BC-K013-S). Briefly, after the indicated treatments, conditioned medium was collected and centrifuged at 1000 × g for 5 min at 4 °C to remove cell debris. The resulting supernatant was used for subsequent colorimetric reaction and absorbance detection. FFA concentration was measured colorimetrically using a microplate reader (Molecular Devices, SpectraMax i3x) with SoftMax Pro 7.3.1 and normalized to total cellular protein from the corresponding cell samples.

Live cell imaging

Mito Deep Red (Genvivo, #PKMDR-2) and JC-1 staining (Sigma-Aldrich, #CS0390) were used to detect mitochondrial content and mitochondrial membrane potential according to the vendor’s instructions. In brief, cells with different treatments were incubated with Mito Deep Red (1:1000) and JC-1 (1.0 µg/mL) at 37 °C for 30 min. After three washes with PBS, relative fluorescence intensity and changes in the ratio of JC-1 polymer (590 nm) to JC-1 monomer (520 nm) were used to indicate mitochondrial activity. The images were taken using a Nikon AX confocal microscope controlled by NIS-Elements AR 5.42.06.

3T3-L1 cells, successfully and stably expressing mCherry-Sialin2,and were stained PK Mito Deep Red (Genvivo, #PKMDR-2), and ER-Tracker Green dye (Thermo Fisher Scientific, E34251) according to the manufacturer’s instructions. The procedure for Hessian imaging was performed as described previously⁶⁰. A commercial structured illumination microscope (HIS-SIM) was used to acquire and reconstruct the cell images. To further improve the resolution and contrast of the reconstructed images, sparse deconvolution was applied as described previously⁶¹.

TurboID-mediated biotinylation and mass spectrometry analysis

HEK293T cells were seeded in 10 cm culture dishes and co-transfected with plasmids encoding Sialin2-TurboID to prepare samples for mass spectrometry analysis. All mass spectrometry experiments were performed with three independent biological replicates. Twenty-four hours after transfection, biotin (Sigma, #B4501) was immediately added to the culture medium at a final concentration of 50 μM for 15 min at 37 °C. Cells without biotin addition were used as a negative control. Biotin labeling was terminated by transferring the cells onto the ice. Cells were washed three times with cold 1× PBS, then lysed for 20 min on ice with RIPA lysis buffer (Santa Cruz Biotechnology, #sc-24948) containing protease inhibitor (Sigma-Aldrich, #P8340) and RiboLock RNase Inhibitor (Thermo Fisher Scientific, #EO0381). The lysates were centrifuged at 12,000 × g, 4 °C for 10 min. The supernatant was then collected and transferred to new tubes.

For affinity purification, 300 μL streptavidin magnetic beads (BEAVERE, #22305-10) were washed three times with RIPA buffer before use. The prepared beads were added to each tube of supernatant, and the mixture was incubated at 4 °C with gentle rotation overnight. The next day, the supernatant was discarded, and the beads were washed sequentially with 1 ml of RIPA lysis buffer (twice), 1 mL of 1 M KCl (once), 1 mL of 0.1 M Na₂CO₃ (once), 1 mL of 2 M urea in 10 mM Tris-HCl (pH 8.0) (once), and 1 mL RIPA lysis buffer (twice). Biotinylated proteins were then eluted from the beads through boiling in a 4 × SDS sample buffer (GenScript, #M00676). The protein band of interest was excised from the SDS-PAGE gel, cut into small pieces (~1 cm²), and destained by incubation with ultrapure water for 10 min with shaking, followed by removal of the solution. The gel pieces were then washed three times with 50% acetonitrile (ACN)/100 mM ammonium bicarbonate (NH₄HCO₃, pH 8.0) solution for 10 min each with shaking, removing the solution after each wash. Subsequently, the gel pieces were dehydrated by incubating them in 100% ACN for 10 min with shaking, then completely drying them in a vacuum concentrator. Reduction was performed by incubating gel pieces with 10 mM dithiothreitol (DTT)/50 mM NH₄HCO₃ (pH 8.0) at 56 °C for 1 h, after which the solution was removed. Alkylation was then carried out by incubating with 55 mM iodoacetamide (IAA)/50 mM NH₄HCO₃ (pH 8.0) in the dark at room temperature for 30 min, followed by removal of the solution. Gel pieces were again dehydrated with 100% ACN for 20 min with shaking, then dried in a vacuum concentrator. Digestion was performed overnight at 37 °C by adding an appropriate amount of trypsin solution supplemented with 50 mM NH₄HCO₃ buffer to cover the gel pieces fully. Peptides were then extracted twice by incubating with the extraction solution (60% ACN/5% formic acid) using ultrasonication for 10 min each time. After centrifugation, the supernatants from each extraction were combined and dried using a vacuum concentrator. Finally, the peptides were desalted using a C18 column and stored at −20 °C until subjected to LC-MS/MS analysis.

Mass spectrometry data were acquired using the latest-generation Orbitrap Astral high-resolution mass spectrometer coupled with a Vanquish NEO UHPLC system. Peptide samples were dissolved in loading buffer and injected by an autosampler onto an analytical column (75 μm × 25 cm, C18, 2 μm, 100 Å) for separation. A 24-min gradient was established using two mobile phases (mobile phase A: 0.1% formic acid in water; mobile phase B: 0.1% formic acid in 80% acetonitrile), with a flow rate of 300 nL/min. Mass spectrometric data were acquired in data-independent acquisition (DIA) mode. Mass spectrometry data were processed using DIA-NN v1.8.1 against the UniProt human proteome reference database (release date: 2022-03-29; containing 20,377 protein sequences)⁶². The main search parameters were set as follows: experiment type, DIA; variable modifications, oxidation (M) and acetylation (protein N-term); fixed modification, carbamidomethylation (C); protease, trypsin/P; precursor mass tolerance, initially 20 ppm for the first search and 4.5 ppm for the main search; and fragment mass tolerance, 20 ppm. The Sialin2-interacting proteome obtained by mass spectrometry was analyzed for GO and KEGG functional enrichment by DAVID.

Lipidomics analysis

Lipids were extracted using a modified methyl tert-butyl ether (MTBE)-based protocol. A total of 12 beige adipose tissue samples (WT, n = 6; KO, n = 6) were subjected to untargeted lipidomics analysis, with approximately 40 mg of tissue used for lipid extraction from each sample. Samples were stored at −80 °C before lipid extraction, and repeated freeze–thaw cycles were avoided. Tissue samples were homogenized in 200 μL ultrapure water using a tissue homogenizer under low-temperature conditions, followed by the addition of 800 μL MTBE and 240 μL pre-chilled methanol with continuous vortexing. The mixture was ultrasonicated in a low-temperature water bath for 20 min and incubated at room temperature for 30 min. After centrifugation (14,000 × g, 10 °C, 15 min), the upper organic phase was collected and dried under nitrogen. The dried residues were reconstituted in 200 μL of 90% isopropanol/acetonitrile solution. After vortexing, a 90 μL aliquot of the reconstituted solution was centrifuged under the same conditions, and the supernatant was transferred to an autosampler vial for UHPLC-MS/MS analysis. Relative lipid abundance was determined based on extracted peak intensities. Pooled quality control (QC) samples were prepared by mixing equal aliquots from all study samples and analyzed periodically to monitor system stability.

Lipid separation was performed on a Nexera LC-30A UHPLC system using a Waters CSH C18 column (1.7 μm, 2.1 mm × 100 mm) maintained at 45 °C. The injection volume was 3 μL, and the flow rate was set to 300 μL/min. The mobile phase consisted of solvent A, acetonitrile/water (6:4, v/v) containing 0.1% formic acid and 0.1 mM ammonium formate, and solvent B, acetonitrile/isopropanol (1:9, v/v) containing the same additives. The gradient elution was set as follows: 0–2 min, 30% B; 2–25 min, a linear increase to 100% B; and 25–35 min, 30% B for column re-equilibration.

Mass spectrometry analysis was performed on a Thermo Q Exactive mass spectrometer equipped with an electrospray ionization (ESI) source. Data were acquired in both positive and negative ionization modes using the same acquisition parameters. The ESI parameters were as follows: heater temperature, 300 °C; sheath gas, 45 arbitrary units; auxiliary gas, 15 arbitrary units; sweep gas, 1 arbitrary unit; spray voltage, 3.0 kV; capillary temperature, 350 °C; and S-Lens RF level, 50%. The MS1 scan range was m/z 200–1800. Ten HCD-MS/MS spectra were acquired after each full MS1 scan in data-dependent acquisition (DDA) mode. The resolution was set to 70,000 for MS1 and 17,500 for MS2 at m/z 200.

Lipid identification, peak extraction, peak alignment, and relative quantification were performed using LipidSearch v4.2 with the built-in LipidSearch database⁶³. The search parameters were set as follows: precursor ion tolerance, 5 ppm; product ion tolerance, 5 ppm; and product ion intensity threshold, 5%. Lipid annotations were assigned based on accurate precursor mass, MS/MS fragment ions, and neutral loss information.

For data preprocessing, lipid features with a null or zero value ratio ≥50% in QC samples were excluded. Missing values were imputed using the k-nearest neighbor (KNN) algorithm. The resulting peak intensity matrix was normalized before downstream statistical analysis. Orthogonal partial least squares discriminant analysis (OPLS-DA) was performed for multivariate analysis, and variable importance in projection (VIP) values were obtained from the OPLS-DA model. Differentially abundant lipids were screened using OPLS-DA-derived VIP values combined with univariate statistical analysis, with VIP > 1 and P < 0.05 considered significant. Lipidomics data were analyzed and visualized using R version 3.6.3 and GraphPad Prism (version 10.2.0).

Analysis of U-¹³C₆-Labeled glucose-derived metabolites by UHPLC‑HRMS

For isotopic tracing metabolic flux analysis, fully differentiated beige adipocytes (NC and shSlc17a5 3T3-L1 cells) were cultured in glucose-free DMEM supplemented with 10% dialyzed fetal bovine serum and 2 g/L U-¹³C₆-glucose (Sigma-Aldrich, 110187-42-3, isotopic purity ≥99%) for 24 h to achieve isotopic steady state. All groups were set with three independent biological replicates.

After incubation, cells were quickly washed with ice-cold PBS and metabolically quenched with 400 μL ice-cold methanol (Merck, 1.06035, suitable for LC/MS, ≥99.9%), followed by freezing at −80 °C for 30 min. Then 100 μL ice-cold water was added, and cells were scraped into the methanol-water solvent and transferred to a sealed tube, and metabolites were extracted with 200 μL 80% methanol (v/v) under 5 cycles of sonication (1 min per cycle) in an ice-water bath. Extracts were incubated at −40 °C for 30 min, then centrifuged at 15,000 × g for 15 min at 4 °C. The supernatant (400 μL) was evaporated to dryness under gentle nitrogen flow and reconstituted in 50 μL of 50% acetonitrile (Thermo Fisher, 75-05-8, LC-MS grade, ≥99.9%). After centrifugation, the supernatant was transferred into a sampling vial with an insert before UHPLC‑HRMS analysis. A QC sample was prepared by pooling equal volumes of all prepared samples.

Chromatographic separation was performed on a Thermo Fisher Ultimate 3000 UHPLC system with a Waters ACQUITY UPLC BEH Amide column (2.1 mm × 100 mm, 1.7 μm). The flow rate was 0.30 mL/min, and the column temperature was kept at 25 °C. Mobile phases consisted of (A) 90% acetonitrile and (B) ultrapure water, both with 15 mM ammonium acetate (Merck, 73594, suitable for LC/MS) and 0.2% ammonium hydroxide. The linear gradient was as follows: 0–0.5 min, 5% B; 8 min, 30% B; 9 min, 50% B; 10.5 min, 50% B; 10.6 min, 5% B; held to 13 min. The injection volume was 2 μL.

Mass spectrometry was performed on a Thermo Fisher Q Exactive Hybrid Quadrupole‑Orbitrap Mass Spectrometer (QE) in Heated Electrospray Ionization Negative mode. Spray voltage was 4000 V. Capillary and probe heater temperatures were 320 °C. Sheath gas flow rate was 35 Arb, aux gas flow rate was 10 Arb, and S‑Lens RF level was 50 Arb. Full scan was acquired at 70,000 FWHM (m/z = 200) over m/z 70–1050 with an AGC target of 3 × 10⁶. Data‑dependent acquisition (DDA) was used to collect MS/MS spectra of the top 8 precursors per cycle, with HCD collision energies at 15, 30, and 45 eV, MS/MS resolution at 17,500 FWHM, and AGC target of 1 × 10⁵.

Raw data were processed using Thermo Fisher Xcalibur software (version 4.0.27.19). Peak integration of target metabolites was verified manually. Natural isotope correction and mass isotopomer distribution analysis were performed using the IsoCor (version 2.1.1)⁶⁴. All data were normalized to cell number. Statistical significance was analyzed by two-tailed Student’s t test using GraphPad Prism (version 10.2.0), with P < 0.05 considered statistically significant.

As for acyl-CoA metabolic flux analysis, cells were collected and lysed using the same procedure described above. The supernatant (400 μL) extracted as above was purified by a solid phase extraction cartridge of 2-(2-pyridyl) ethyl-functionalized silica gel with sample loading, washing, and elution with 80% methanol containing 50 mM ammonium formate (Merck, 70221, suitable for LC/MS). The eluate was evaporated to dryness under mild nitrogen gas and reconstituted in 50 μL of 50% acetonitrile before performing UHPLC-MS/MS analysis.

The UHPLC-MS/MS analysis was performed on a 1290 Infinity II UHPLC system (Agilent Technologies) coupled to a 6470A Triple Quadrupole mass spectrometer (Agilent Technologies). Samples were injected onto an ACQUITY UPLC BEH Amide column (100 mm × 2.1 mm, 1.7 μm, Waters Corp) at a flow rate of 0.2 mL/min. The mobile phase consisted of (A) 90% acetonitrile and (B) 15 mM ammonium acetate and 0.3% ammonium hydroxide. The chromatographic separation was conducted by a gradient elution program as follows: started from 10% B and held 1 min, increased to 25% B at 6 min, to 45% B at 7 min and held to 9 min, and finally returned to 10% B at 9.1 min and equilibrated to 12 min for the next injection.

The eluted analytes were ionized in an electrospray ionization source in positive mode. The temperatures of the source drying gas and sheath gas were 300 and 350 °C. The flow rate of the source drying gas and sheath gas was 5 and 11 L/min, respectively. The pressure of the nebulizer was 40 psi, and the capillary voltage was 4000 V. The multiple reaction monitoring (MRM) was used to acquire data in MRM transitions using the [M+H]⁺ as precursor ion and [M-507+H]⁺ as quantifier product ion or m/z 428 as qualifier› product ion at optimized fragmentor and collision energies. The Agilent MassHunter software (version B.08.00) was used to control instruments and acquire data.

Analysis of mitochondrial stress test through Seahorse XFe96

Oxygen consumption rates of primary inguinal cells were determined through an XFe96 Extracellular Flux Analyzer (Seahorse Biosciences). Cells were seeded at 4000 cells per well on a 96-well plate (103680-100, Agilent) for the mitochondrial stress test. Mitochondrial stress testing was performed according to the manufacturer’s guidelines (103680-100, Agilent). In brief, on the day of analysis, cells were washed twice, equilibrated with basal respiration media (10 mmol/L glucose, 1 mmol/L sodium pyruvate, 2 mmol/L glutamine in DMEM basal media), and incubated for 45 min in a CO₂-free incubator before testing. Port injection solutions were prepared as follows (final concentration): 1.5 μM oligomycin (Port A), 2 μM FCCP (Port B), and 0.5 μM Rotenone/Antimycin A (Port C). Each cycle consisted of 4 min of mixing and 2 min of measurement. Data were analyzed through Seahorse Wave Desktop software (ver 2.6.3, Agilent) and presented as mean ± standard deviation (SD).

Proximity ligation assay (PLA)

PLA was used to detect in situ molecular interactions between proteins or between protein and lipid^10,65. After fixation and permeabilization, cells subjected to the indicated treatments were blocked and incubated with primary antibodies, as in routine IF staining. PLA was then performed using the Duolink kit (MilliporeSigma, #DUO92101) according to the manufacturer’s instructions. Slides were mounted with Duolink® In Situ Mounting Medium with DAPI (MilliporeSigma, #DUO82040). PLA signals were acquired using a Nikon AX confocal microscope with NIS-Elements AR 5.42.06 and appeared as discrete punctate foci, indicating intracellular proximity events. For quantification, images were collected under identical acquisition settings within each experiment, and PLA puncta were counted on a per-cell basis using ImageJ 1.52v.

For PLA in WAT, paraffin sections were deparaffinized, rehydrated, and subjected to antigen retrieval before blocking and incubation with primary antibodies. PLA was then performed using the same kit and protocol as described above, according to the manufacturer’s instructions. Signals were imaged under identical settings using a confocal microscope, and PLA signal intensity was quantified using ImageJ 1.52v in the indicated groups.

Calcium imaging

All Ca²⁺ measurements were performed in the absence of extracellular Ca²⁺ in the following buffer (NaCl 140 mM; KCl 50 mM; MgCl₂ 10 mM; HEPES 100 mM; glucose 100 mM; EGTA 1 mM, pH 7.4) after 36 h of infection with pcDNA-4mtD3cpv overexpressing the FRET-based mitochondrial ratiometric Ca²⁺ probe⁵⁹. After 90 s basal fluorescence measurement, Na-ATP (100 µM, added in puff) was added to stimulate ER-mitochondria Ca²⁺ exchange, and acquisitions were continued for 6 min on 1 field/dish. The fluorescence ratio YFP/CFP was analyzed with ImageJ 1.52v after removing background fluorescence. Results represent the average of all analyzed cells from 3 independent experiments.

Statistics and reproducibility

Statistical analyses were performed using GraphPad Prism (version 10.2.0). An unpaired two-sided Student’s t test was used to determine the significance between two groups of normally distributed data. Welch’s correction was used for groups with unequal variances. An ordinary one-way ANOVA was performed for multiple comparisons between groups, followed by Tukey’s or Dunnett’s test as specified in the legends. Brown–Forsythe and Welch’s correction were used for groups with unequal variances. P < 0.05 was considered a statistically significant difference. The sample size was determined based on the previous studies and literature in the field using similar experimental paradigms. Each experiment was repeated at least three times independently, and the number of repeats is specified in the figure legend for each experiment. We used at least three independent experiments or biologically independent samples for statistical analysis.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

S.W. and X.L.: conceived the project, acquired funding, provided direction, and supervised the research; O.J. and X.L.: coordinated the group, designed and conducted experiments, analyzed data, and interpreted results; X.L.: assisted with cellular experiments and figure preparation; O.J.: conducted animal studies; O.J. and X.L.: performed proteomics analysis; Z.C., S.K., Q.W., T.Z., X.C., S. Wu., B.Z., Y.F., and J.W.: helped the preparation of experiments; M.C.: guided experimental design and assisted manuscript and model preparation; O.J., X.L., and S.W.: wrote the manuscript, with input from all authors.

Peer review

Peer review information

Nature Communications thanks Naresh Babu Sepuri, Xavier Prieur, and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The mass spectrometry proteomics data generated in this study have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD076314 and accession code MwV6L40ahc53 (https://www.ebi.ac.uk/pride/). The metabolomics data are deposited at MetaboLights under accession code MTBLS14192 [https://www.metabolights.org/]. Source data are provided with this paper. All other data supporting the findings of this study are available from the corresponding author upon reasonable request. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains Supplementary material available at 10.1038/s41467-026-74256-w.