LRG1 is an adipokine that promotes insulin sensitivity and suppresses inflammation

While dysregulation of adipocyte endocrine function plays a central role in obesity and its complications, the vast majority of adipokines remain uncharacterized. We employed bio-orthogonal non-canonical amino acid tagging (BONCAT) and mass spectrometry to comprehensively characterize the secretome of murine visceral and subcutaneous white and interscapular brown adip ocytes. Over 600 proteins were identified, the majority of which showed cell type-specific enrichment. We here describe a metabolic role for leucine-rich α–2 glycoprotein 1 (LRG1) as an obesity-regulated adipokine secreted by mature adipocytes. LRG1 overexpression significantly improved glucose homeostasis in diet-induced and genetically obese mice. This was associated with markedly reduced white adipose tissue macrophage accumulation and systemic inflammation. Mechanistically, we found LRG1 binds cytochrome c in circulation to dampen its pro-inflammatory effect. These data support a new role for LRG1 as an insulin sensitizer with therapeutic potential given its immunomodulatory function at the nexus of obesity, inflammation, and associated pathology.


Introduction
Obesity is a major threat to human health due to its association with serious comorbidities (Angelantonio et al., 2016;Poirier et al., 2006). Now considered an important endocrine organ, adipose tissue secretes a constellation of bioactive peptides, or adipokines, many of which regulate whole-body energy homeostasis and in ammation (Funcke and Scherer, 2019). Dysregulation of adipose tissue endocrine function is a key feature of obesity and a major contributor to its sequelae. However, the vast majority of adipokines remain unstudied. Addressing this knowledge gap calls for a detailed characterization of the adipose secretome to better understand how adipocytes communicate with other cells to coordinate systemic metabolism.
Mammals possess white (WAT) and brown adipose tissues (BAT), with divergent effects on whole-body metabolism. Visceral (Visc) WAT is particularly associated with obesity-related diseases, whereas subcutaneous (SubQ) WAT is comparatively benign (Fox et al., 2007). On the other hand, active BAT is associated with improved cardiometabolic health (Becher et al., 2021;Lee et al., 2014). Whereas Visc fat contains predominantly white adipocytes that e ciently store energy, SubQ depots contain a mixture of white and thermogenic beige adipocytes which, like brown adipocytes, dissipate energy as heat via uncoupled respiration (Chouchani et al., 2019). In addition to these bioenergetic properties, transplantation studies in mice have demonstrated that secretory mediators likely convey the metabolic effects of different adipose tissues (Tran et al., 2008).
Transcriptomic pro ling coupled with secretion prediction algorithms is often utilized to identify putative secreted proteins, but this approach could omit non-classically secreted factors. Alternatively, direct detection of proteins in conditioned medium (CM) can be performed using mass spectrometry (MS).
Such proteomic analyses are commonly performed in fetal bovine serum (FBS)-free conditions because FBS proteins interfere with detection of relatively low abundance secreted proteins in CM. Serum sample can estimate total moles of proteins detected (Shin et al., 2013). Σ iBAQ values closely correlated with in-gel uorescence, both of which showed highest abundance of secreted proteins in SubQ CM (Fig. 1d).
We next examined how many of these AHA-labeled proteins are reported or predicted secreted factors. Referencing gene ontology (GO) cellular component terms for 594 genes from 604 detected proteins, a large majority of proteins (413/594, 70%) were annotated to be extracellular (Fig. 1e). We performed retrospective secretion prediction analysis on 604 proteins, evaluating the proportion of proteins predicted to be secreted by the classical ER/Golgi pathway (SignalP5. 0 Pierleoni et al., 2008). Over two-thirds of the proteins (413/604, 68%) were predicted to be secreted via the classical pathway, while non-classical secretion constituted about 10% (62/604) (Fig. 1f). Overall, 79% of detected proteins (479/604) met at least one criterion of secretion prediction. Notably, more than 20% of identi ed proteins would not have been predicted to be secreted by these algorithms, highlighting the imperfect nature of in silico predictions and the value of directly measuring secreted protein levels.
In-gel uorescence analysis showed differential band patterns across the three types of adipocytes, suggesting cell type-speci c secretory pro les (Fig. 1c). Consistent with this observation, principal component analysis (PCA) showed that each cell type formed a distinct cluster (Extended Data Fig. 1e). Comparing all 9 AHA-pulsed samples against each other showed higher correlations within biological replicates (R > 0.97) than those across different cell types (0.74 < R < 0.91) (Extended Data Fig. 1f). For quantitative analysis, we focused on 424 proteins detected in at least 2/3 replicates (Extended Data Fig. 1d) and compared their Log2-transformed label-free quanti cation (LFQ) intensities, imputing any missing values with a left-shifted Gaussian distribution. One-way ANOVA yielded 348 proteins with signi cantly different (q < 0.01) secretory pro les across cell types, and unbiased clustering was performed based on protein levels in CM (Fig. 1g). Clusters 1 and 4 contained proteins enriched in CM of Visc and Brown, respectively. Cluster 2 proteins were most abundant in CM of SubQ followed by Visc, both white adipocytes. Cluster 3 was highly secreted by SubQ and Brown adipocytes with proteins such as SLIT2 (Svensson et al., 2016), suggesting enrichment of the beige/brown secretome. Cluster 4 comprised the largest number of proteins (125/348), consistent with Brown CM containing the greatest number of unique proteins (Fig. 1h). Ranking proteins by decreasing order of abundance, we found high levels of many well-described adipocyte-derived factors such as CFD, RETN, ADIPOQ, and RBP4 (Fig. 1i). Interestingly, many cluster 2 proteins ranked highly in abundance, while cluster 4 proteins skewed towards lower abundance (Fig. 1i). Hence, while SubQ adipocytes demonstrate high secretory capacity, the brown adipocyte secretome is characterized by a diverse array of proteins, many of which are secreted at lower levels.
We next performed pathway analysis on each of the clusters (Fig. 1j). Visc CM-enriched cluster 1 was overrepresented with proteins involved in immune response and complement activation, such as C1QA and CFH (Fig. 1k). Proteins with a role in collagen bril organization and extracellular matrix formation, such as COL1A1 and SFRP, were enriched in cluster 2 (Fig. 1l). Cluster 3 (beige/brown enriched) showed overrepresentation of angiogenesis regulators (Fig. 1m). Finally, cluster 4 Brown CM-speci c proteins were enriched for axon guidance factors such as NRP1 and NGF, consistent with the importance of innervation in thermogenesis (Fig. 1k) (Chi et al., 2018;Wang et al., 2020;Zeng et al., 2019).

Pro ling of the serum proteome in vivo
Previous studies have shown that AHA can be administered in vivo to label tissue proteins (Calve et al., 2016;McClatchy et al., 2015). However, it has not been tested whether this method can label the nascent serum proteome. We administered 0.1 g/kg/day of AHA IP to chow-fed B6 mice for two days (Fig. 2a) and did not notice any major adverse effects based on body weight (Fig. 2b). CuAAC conjugation of serum proteins with TAMRA-alkyne and in-gel uorescence analysis showed increased signal across most bands in the AHA group along with some AHA group-speci c uorescent bands (Fig. 2c). We used alkyne-agarose beads to enrich the azide-labeled nascent proteome and performed MS analysis. Σ iBAQ showed successful enrichment in animals injected with AHA (Fig. 2d). Even without dietary Met restriction or depletion of abundant serum proteins, we were able to identify and quantitate 180 proteins, including classical adipokines such as ADIPOQ, adipsin (CFD), and RBP4 (Fig. 2e).
Because AHA can be incorporated into the proteome of any tissue, we employed bioinformatic analyses to predict the source of serum proteins. We cross-referenced 177 genes from 180 detected proteins with publicly available transcriptomic datasets, such as ENCODE (RNA-Seq based) and BioGPS (microarray based) (Davis et al., 2018;Wu et al., 2016). A t-distributed stochastic neighbor embedding (t-SNE) plot was generated with each protein and its tissue mRNA levels from the ENCODE/LICR dataset (Fig. 2f). The majority of proteins were most highly expressed by the liver, which comprised the largest cluster (98/177, 55%). BAT, the only adipose tissue pro led in the dataset, formed a cluster of 14 genes (7.9%). As expected, classical adipokines known for highly adipose-speci c expression such as adiponectin (Adipoq) and adipsin (Cfd) belonged to this group, along with a recently described batokine, Kng2 (Peyrou et al., 2020).
BioGPS offers microarray-based transcriptomic data across a much wider variety of mouse tissue and cell types, including eWAT and BAT. We numerically scaled the degree of adipose tissue enrichment for each detected serum protein. Adipose enrichment of a gene was de ned as the number of tissues with expression lower than that of eWAT, BAT, or combined. We divided this value by the total number of pairwise comparisons (i.e., total number of tissues − 2 adipose tissues) to obtain the percentage adipose tissue enrichment score. Among the top-enriched genes were well-described adipokines such as Adipoq, where 76 of 86 tissues (88.4%) expressed Adipoq at signi cantly lower levels than eWAT and BAT combined (Fig. 2g). We also identi ed genes yet to be described as adipokines, such as Lrg1 and H2-Q10 (Fig. 2g). Many genes, such as Rbp4, Agt, and Lrg1, showed high expression in adipose tissues as well as liver. On the t-SNE plot, these genes were grouped with other liver-speci c genes, but located closer to the adipose tissue cluster (Fig. 2f). Still, their percentage enrichment scores were > 80%, as few other tissues express those genes. LRG1 is secreted by mature adipocytes and increased in obesity To identify uncharacterized adipokines with a putative role in whole-body metabolism, we prioritized factors 1) detected in adipocyte CM, 2) present in the nascent serum proteome, and 3) enriched in adipose tissues with scores > 80% (Fig. 3a). This analysis yielded 10 proteins (Fig. 3a). ADIPOQ, CFD, and RBP4 have already been identi ed as adipokines, validating our search strategy. Haptoglobin (HP), ceruloplasmin (CP), and angiotensinogen (AGT) have well-characterized biological functions. We therefore focused on LRG1, a protein with relatively unknown metabolic function. LRG1, or leucine-rich α-2 glycoprotein 1, has been shown to be expressed in endothelial cells, where it promotes angiogenesis by modulating TGF-β signaling (Wang et al. 2013). However, LRG1 has not been characterized as an adipokine, nor has its role in whole-body metabolism been studied. Consistent with our tissue enrichment analysis, qPCR of mouse tissues showed that Lrg1 mRNA was mostly expressed in adipose tissues and liver (Fig. 3b). To determine which cell type(s) within adipose tissue express Lrg1, we fractionated eWAT, iWAT, and BAT to separate oating mature adipocytes from the SVF. Adipoq, a mature adipocyte marker, was signi cantly co-enriched with Lrg1 in the adipocyte fraction in all three depots ( Fig.  3c and Extended Data Fig. 2a). Finally, we cultured primary Visc, SubQ, and Brown adipocytes and con rmed that Lrg1 mRNA is induced > 70-fold in all three cell types during in vitro adipogenesis (Fig. 3d). Consistent with mRNA data, we found robust levels of LRG1 protein in CM of mature adipocytes, whereas preadipocytes did not secrete detectable LRG1 (Fig. 3e). Of note, LRG1 protein levels detected by western blot were consistent with MS results (Extended Data Fig. 2b).
To assess whether adipose tissues are a signi cant contributor to circulating LRG1 levels, we collected serum from chow-fed lean mice or diet-induced obese (DIO) mice on high fat diet (HFD) for 4 or 9 weeks. Serum LRG1 protein levels increased with age and obesity (Fig. 3f). qPCR of major Lrg1-expressing tissues showed signi cant induction of Lrg1 mRNA in iWAT of DIO mice, but not in liver (Fig. 3g). With iWAT expansion in obesity, this induction likely contributes to elevated serum LRG1 levels in this state.
Obesity is characterized by chronic low-grade in ammation, with elevated circulating in ammatory cytokines (Lackey and Olefsky, 2016). We observed that treatment of primary SubQ adipocytes with recombinant TNFα induced expression of Lrg1 mRNA and protein in CM (Fig. 3h,i). Taken together, these results demonstrate that LRG1 is an obesity-induced adipokine. LRG1 overexpression improves glucose homeostasis in diet-induced obesity To explore whether LRG1 as an adipokine affects whole-body energy homeostasis, we used viral vectors to overexpress LRG1 in vivo. Adenovirus encoding eGFP (Ad-eGFP) or C-terminally FLAG-tagged LRG1 (Ad-LRG1-FL) was administered to obese B6 mice on HFD for 10 weeks (Fig. 4a). Plasma western blot con rmed LRG1 overexpression in the Ad-LRG1-FL group 5 days after infection (Fig. 4b). We observed no difference in body weights between groups (Fig. 4c), but an insulin tolerance test (ITT) showed that the Ad-LRG1-FL group had signi cantly enhanced insulin response (P = 0.045) (Fig. 4d).
To study the longitudinal effects of chronic LRG1 overexpression, we used AAV8, which has tropism for liver and adipose tissue. We administered AAV-eGFP or AAV-LRG1-FL to 6-week-old male B6 mice (Fig. 4e) and con rmed increased plasma LRG1 in the latter (Fig. 4f). During 3 months of HFD, both groups gained an equivalent amount of weight to around 50 g (Fig. 4g), with no difference in tissue weights (Fig. 4h). Fasting glucose measurements from 4 to 14 weeks on HFD showed that the AAV-LRG1-FL group had signi cantly lower fasting glucose levels (P = 0.0097) with a dampened peak (LRG1: 225.9 ± 6.5 mg/dL vs. eGFP: 263.7 ± 10.8 mg/dL, mean ± SEM) at week 18 ( Fig. 4i). A glucose tolerance test (GTT) demonstrated that AAV-LRG1-FL mice had markedly improved glucose (P = 0.0029) (Fig. 4j) and insulin tolerance (P = 0.023) (Fig. 4k). These observations suggest that LRG1 overexpression prevents obesityrelated dysregulation of glucose homeostasis by insulin sensitization. LRG1 loss of function elevates fasting blood glucose in diet-induced obesity To test whether LRG1 loss of function affects glucose homeostasis, we generated whole-body LRG1-KO animals using CRISPR-Cas9 targeting exon 2 of Lrg1 (Extended Data Fig. 3a). This led to a frameshift mutation in Lrg1 (Extended Data Fig. 3b) and absence of LRG1 protein in plasma (Extended Data  Fig. 3f). Interestingly, male LRG1-KO animals demonstrated signi cantly higher fasting glucose levels compared to WT littermates throughout the HFD challenge (P = 0.010) with a higher peak (KO: 261.6 ± 11.3 mg/dL vs. eGFP: 225.0 ± 10.9 mg/dL, mean ± SEM) at week 12 on HFD (Extended Data Fig. 3g). Therefore, fasting blood glucose levels are reciprocally regulated by LRG1 gain and loss of function. At the time points tested, however, GTT and ITT did not show signi cant differences between the genotypes (Extended Data Fig. 3h-k).
LRG1 overexpression delays diabetic phenotype and promotes WAT expansion in db/db mice While B6 mice develop severe obesity upon HFD feeding, they demonstrate only transient and mild hyperglycemia with moderate insulin resistance (Winzell and Ahrén, 2004;Kleinert et al., 2018). Because LRG1 overexpression in B6 animals mitigated hyperglycemia, we explored whether LRG1 can improve glucose homeostasis in C57BLKS/J-Lepr db/db (db/db) mice, a more extreme model of obesity-related type 2 diabetes. Due to a leptin receptor (Lepr) mutation and genetic background, db/db animals demonstrate hyperphagia and early-onset obesity, along with profound hyperglycemia and hyperinsulinemia (Kleinert et al., 2018). We con rmed that obesity and hyperglycemia over 400 mg/dL develop in db/db animals as early as 7 weeks of age, while littermate misty mice with a WT Lepr gene (m/m) maintain fasting glucose levels below 200 mg/dL (Extended Data Fig. 4a,b). Similar to B6 DIO mice, db/db mice showed higher circulating LRG1 levels compared to lean littermates (Extended Data Fig. 4c).
We administered AAV-eGFP or AAV-LRG1-FL to db/db mice at 4 weeks of age, before development of severe hyperglycemia (Fig. 5a,b). Starting two weeks post-injection, LRG1-overexpressing animals demonstrated accelerated weight gain, such that by 10 weeks of age the LRG1 group weighed 19.6% more (P = 0.004) than eGFP controls (Fig. 5c). During this period, cages housing LRG1-overexpressing animals tended to have greater food intake (Extended Data Fig. 4d). Concomitantly, the AAV-LRG1-FL group showed delayed onset of hyperglycemia. At week 6, we observed frank hyperglycemia in eGFP animals, whereas glucose levels in the LRG1 group were 33.6% lower (eGFP: 399.5 ± 31.7 mg/dL vs. LRG1: 265.4 ± 19.6 mg/dL, mean ± SEM; P = 0.0020) (Fig. 5d). Fasting plasma insulin levels were almost halved in the LRG1 group (eGFP: 12.5 ± 1.3 ng/mL vs. LRG1: 6.3 ± 0.6 ng/mL, mean ± SEM; P = 0.0008) (Fig. 5e). Consistent with these ndings, an ITT showed that LRG1-overexpressing animals demonstrated signi cantly improved insulin sensitivity (P = 0.019) ( Fig. 5f and Extended Data Fig. 4e). By 8 weeks of age, body weights continued to diverge between the groups, but we no longer observed signi cant differences in fasting blood glucose or plasma insulin concentrations ( Fig. 5c-e). Tissue weight measurements revealed that LRG1-overexpressing animals had accelerated gain of eWAT (P < 0.0001) and iWAT (P = 0.0001) mass, such that by week 10, their eWAT and iWAT were 53% and 40% heavier, respectively, than those of eGFP controls (Fig. 5g,h). We observed no difference in weights of BAT, liver, or gastrocnemius muscle (Fig. 5g). These results reveal that LRG1 overexpression in db/db animals promotes insulin sensitivity and WAT expansion-driven weight gain.

LRG1 suppresses obesity-associated systemic in ammation
Based on the accelerated eWAT and iWAT expansion in LRG1-overexpressing db/db animals, we hypothesized that WAT could be a target of LRG1 action. We analyzed para n-embedded, hematoxylin and eosin (H&E) stained tissue sections and found that eWAT from B6 animals sacri ced at 21 weeks of age (13 weeks on HFD) was characterized by accumulation of macrophages forming crown-like structures (CLS) (Fig. 6a). The distal portion of eWAT was especially susceptible to CLS formation in eGFP controls, while the same region in LRG1-overexpressing animals displayed an 82% reduction in CLS number (P < 0.0001) (Fig. 6a,b). B6 iWAT contained fewer CLS compared to eWAT and did not show major morphological differences between groups (Extended Data Fig. 5a,b). In db/db eWAT, we observed signi cantly reduced CLS in the LRG1 group by 81% at week 7 (P = 0.0001) and 85% at week 10 (P < 0.0001) (Fig. 6c,d). Similarly, CLS in db/db iWAT from LRG1-overexpressing animals were reduced by 83% at week 7 (P = 0.0043) and 88% at week 10 (P < 0.0001) (Fig. 6c,e).
Obesity is associated with non-alcoholic fatty liver disease (NAFLD), characterized by hepatic steatosis with or without in ammation (Farrell et al., 2019). While HFD-fed B6 mice rarely demonstrate liver injury or in ammation, mild necroin ammation can be observed in db/db liver as early as 1 month of age (Febbraio et al., 2019; Trak-Smayra et al., 2011). Liver sections from the B6 cohort showed a similar degree of steatosis in both groups, without any in ammatory lesions (Extended Data Fig. 5c). In the db/db cohort, while both eGFP and LRG1 developed a similar degree of hepatosteatosis, in ammatory foci were found only in eGFP-expressing animals (Fig. 6c).
In ammation is a key link between obesity and insulin resistance (Saltiel & Olefsky, 2017). We hypothesized that LRG1 may mediate its insulin sensitizing effect via attenuation of in ammation in susceptible organs. We performed RNA-Seq analysis on eWAT from AAV-treated db/db animals harvested at week 7 (midpoint) and 10 (endpoint). We performed differential gene expression analysis between the eGFP and LRG1 groups at midpoint and subjected the list of signi cant genes (P < 0.01) to GO gene-set enrichment analysis (GSEA). All of the top 20 differentially regulated pathways showed highly signi cant enrichment (P = 0.0067) and were immune-related, including leukocyte activation, innate immune response, and cytokine production (Fig. 6f). The enrichment score for each of these pathways was negative in the LRG1 group, indicating down-regulation of in ammatory processes in these animals. GSEA of signi cantly regulated genes at the 10-week endpoint yielded similar down-regulation of immune-related pathways in the LRG1 group (Extended Data Fig. 5d). To visualize which genes are most signi cantly differentially regulated, we plotted a heatmap of 68 genes that showed signi cant difference (adjusted P < 0.01) with Log2 fold-change of > 2 or <-2 between the two groups at both time points. Consistent with the pathway analysis, the LRG1 group showed signi cant down-regulation of cytokines and chemokines such as Ccl3, Ccl4, and Cxcl9; metalloproteinases such as Mmp12 and Mmp13 known to be highly expressed by macrophages; and various immunoglobulin subunit genes (Fig. 6g).
Our histological analysis demonstrated signi cantly reduced CLS in eWAT of LRG1-overexpressing animals. Deconvolution algorithms such as CIBERSORTx allow estimation of cell populations from bulk RNA-Seq datasets (Newman et al., 2019). We performed CIBERSORTx analysis on the expression dataset to gain further insight into differences in immune cell populations. CIBERSORTx estimated that the LRG1 group contains fewer immune cells (Fig. 6h). Macrophages were predicted to constitute the majority of immune cells in every eWAT sample analyzed, and the LRG1 group demonstrated a lower absolute quantity of macrophages ( Fig. 6h), without affecting their relative proportions (Extended Data Fig. 5e). Many of the differentially regulated genes between eGFP-and LRG1-overexpressing animals encode chemokines and cytokines. We performed a multiplex cytokine assay to assess whether these differences are re ected in serum levels, and at midpoint (week 7), the LRG1 group showed 28-86% reduction of circulating chemokines such as MCP-1, MIP-1α, MIG, and IP-10 and cytokines such as TNFα and IFNγ (Fig. 6i). Many of these differences subsided by week 10, mostly due to a reduction of cytokines in the eGFP group (Extended Data Fig. 5f). Taken together, these results suggest that LRG1-overexpression attenuated pro-in ammatory processes associated with obesity.
LRG1 binds extracellular cytochrome c and blocks its pro-in ammatory effect on macrophages LRG1 contains leucine-rich repeat (LRR) domains, which form a structural framework for protein-protein interactions. We hypothesized that LRG1's immunomodulatory function is mediated by protein-protein interactions. LRG1 has been reported to bind cytochrome c (Cyt c), a mitochondrial protein (Cummings et al., 2006), but the physiological signi cance of this interaction has not been studied. In addition to its role in the respiratory chain and intrinsic apoptosis pathway, Cyt c is released into the extracellular space following cell death (Jemmerson et al., 2002;Renz et al., 2001) and mediates pro-in ammatory signals as a damage-associated molecular pattern (DAMP) (Grazioli & Pugin, 2018). Adipocyte death in obesity is a key event promoting macrophage in ltration and WAT in ammation (Cinti et al., 2005), but the exact triggers of metabolic in ammation remain unidenti ed. We explored whether extracellular Cyt c released by dead/dying adipocytes could be a key mediator of macrophage recruitment/activation, and if LRG1 in turn modulates the pro-in ammatory action of Cyt c.
We rst examined whether LRG1 binds extracellular Cyt c in the circulation. We utilized the C-terminal FLAG-tag of overexpressed LRG1 and α-FLAG antibody-conjugated beads to co-immunoprecipitate (IP) LRG1-FL and interacting partners. In the serum of db/db animals transduced with AAV-LRG1-FL, we were able to successfully co-IP both LRG1-FL and Cyt c from the serum, indicating the two proteins indeed circulate as a complex (Fig. 7a) , it is not known whether obesity is associated with increased serum Cyt c levels. Western blot analysis of serum revealed that Cyt c is increased in db/db animals compared to lean littermates (Fig. 7b). Interestingly, extracellular Cyt c levels trended higher at 7 than at 10 weeks of age (Fig. 7b), correlating with the time point when LRG1 overexpression has a potent physiological effect. In B6 mice, HFD led to increased Cyt c levels compared to chow-fed lean animals (Fig. 7c). To test if dying adipocytes can contribute to circulating Cyt c, we subjected primary SubQ adipocytes to treatments that induce apoptosis or necrosis (Fig. 7d). In CM of SubQ adipocytes treated with staurosporine (STS) for 24 h, we detected a robust increase in Cyt c compared to vehicle controls (Fig. 7d). CM collected from adipocytes incubated in hypoxic conditions or treated with TNFα also showed increased Cyt c compared to normoxic or vehicle controls, respectively (Fig. 7d). Extracellular Cyt c has been shown to exert a pro-in ammatory effect by acting on the toll-like receptor 4 (TLR4)-mediated innate immune signaling pathway in astrocytes (Wenzel et al., 2019). Because circulating monocytes contribute to adipose tissue macrophages in obesity (Oh et al., 2012), we tested whether bone-marrow derived macrophages (BMDMs) respond to extracellular Cyt c by upregulating proin ammatory genes. Treating BMDMs with horse Cyt c led to > 30-fold induction of lipopolysaccharide (LPS)-responsive pro-in ammatory genes such as Il1b, Cxcl10, Il6, and Nos2 (P < 0.0001), while pretreatment of macrophages with a small molecule TLR4 inhibitor, TAK-242, prevented this induction (Fig.  7e). These data con rm that Cyt c in the extracellular space acts as a DAMP that activates innate immune signaling and polarizes macrophages into a more pro-in ammatory state. We then explored how co-treatment of Cyt c and recombinant human LRG1 (rhLRG1) affects Cyt c's pro-in ammatory effect on macrophages. Prior to treatment, media containing Cyt c, rhLRG1, or both were incubated for one hour to allow protein-protein interactions to occur. Compared to single treatment, co-treatment with rhLRG1 signi cantly dampened Cyt c-mediated induction of Il1b (P < 0.0001), Cxcl10 (P = 0.024), and Il6 (P = 0.0008), while Nos2 induction was not affected by co-treatment with rhLRG1. (Fig. 7f). Interestingly, pretreatment of BMDMs with rhLRG1 followed by addition of Cyt c did not attenuate the latter's proin ammatory effect, suggesting that Cyt c-LRG1 complex formation is necessary for LRG1's modulatory effect (Extended Data Fig. 6a). Hence, LRG1 can modulate the pro-in ammatory gene expression responses induced by Cyt c.
Cyt c is a small (12.5 kDa), globular protein with a short half-life in circulation (Radhakrishnan et al., 2007). Small proteins such as Cyt c are expected to be rapidly cleared via glomerular ltration in the kidney. With the tight binding interaction between LRG1 and Cyt c, we hypothesized that much of extracellular Cyt c circulates as a complex with LRG1. To assess whether this interaction affects Cyt c clearance, we performed retroorbital injection of horse Cyt c into WT and LRG1-KO mice and collected blood at time points from 5 minutes to 6 hours post-injection. Plasma western blot showed that LRG1-KO animals almost completely cleared Cyt c by 2 hours, whereas in WT animals, an initial decrease in Cyt c was followed by nearly constant levels between 1 and 4 hours post-injection (Fig. 7g). Fitting one-phase decay functions to relative Cyt c intensities showed that while both WT and LRG1-KO curves have similar half-lives (WT: 16.10 min vs. KO: 17.32 min), the WT trace plateaus at 0.2791, whereas the LRG1-KO curve approaches zero ( Fig. 7h and Extended Data Fig. 6b). Concomitantly, appearance of Cyt c in the urine was observed as soon as 30 minutes after the injection, demonstrating that clearance of excess Cyt c is indeed through the kidneys (Extended Data Fig. 6c). Urine Ponceau S staining showed that detected proteins were all below 25 kDa, the apparent size limit for urinary excretion (Extended Data Fig. 6c). We reason that in LRG1-KO animals, Cyt c circulates as an unbound, free form that can be rapidly excreted by glomerular ltration, whereas the presence of LRG1 in WT animals prevents a portion of Cyt c from excretion due to formation of a complex larger than the size cutoff for glomerular ltration. Taken together, these data directly demonstrate that Cyt c is a native ligand of LRG1, and that this interaction suppresses the pro-in ammatory effect of extracellular Cyt c (Fig. 7i).

Discussion
We applied a chemoproteomic technology to comprehensively pro le the secretome of three major types of murine primary adipocytes, revealing unique secretory pro les of each cell type with differential functional enrichment. This method was also applied in vivo to pro le the nascent serum proteome, and bioinformatic analysis demonstrated that adipose tissue is an important contributor to the serum proteome. The intersection of the adipocyte and serum proteome included classical adipokines such as adiponectin ( , all of which have important roles in whole-body energy homeostasis. LRG1 shares a similar expression signature with these adipokines, but its metabolic function has not been characterized. We demonstrate, using two types of viral vectors and two different mouse models of obesity, a novel metabolic role for LRG1 as a regulator of glucose homeostasis by promoting insulin sensitization. This insulin-sensitizing effect in LRG1 gain of function is associated with a dramatic reduction in systemic in ammation and with LRG1's ability to bind Cyt c and modulate its pro-in ammatory effect as a DAMP. The low rate of intracellular proteins detected in our dataset demonstrates the advantage BONCAT provides by enabling FBS supplementation. Of the remaining 20% of proteins not predicted or annotated to be secreted, many are non-classically secreted proteins such as FABP4/5 (Hotamisligil and Bernlohr, 2015) and other novel factors we have independently con rmed as bona de secreted factors. With careful validation, our dataset can greatly expand the scope of the adipose secretome that includes nonclassical modes of secretion. We have also demonstrated that BONCAT can label nascent serum proteins in vivo. This technique can be further utilized to study the nascent serum proteome in response to stimuli di cult to model in vitro, such as obesity and changes in ambient temperature. Novel technologies for in vivo pro ling of cell type-speci c proteomes using non-canonical amino acids (Alvarez-Castelao et al. Adipose tissue expansion in mice is characterized by periods of rapid adipocyte turnover, during which there is a sharp increase in adipocyte death and monocyte/macrophage accumulation (Rosen and Spiegelman, 2014;Weisberg et al., 2003). In B6 eWAT, adipocyte death and CLS formation peak around 12-16 weeks on HFD (Strissel et al., 2007), which corresponds to when a parallel surge in fasting blood glucose levels is observed. In db/db mice, the remodeling seems to be initiated earlier, as CLS in WAT emerge at 7 weeks of age, and chemokine/cytokine expression and circulating levels are more highly induced at week 7 than week 10. Periods when LRG1's insulin-sensitizing and anti-in ammatory effects are particularly effective coincide with these times of rapid adipocyte death and turnover. As in ammation is a key mechanistic link between obesity and insulin resistance (Saltiel & Olefsky, 2017), these ndings strongly suggest LRG1 promotes insulin sensitization via modulation of in ammation during adipose tissue remodeling.
The biological context and function of LRG1 and extracellular Cyt c binding have been unknown. We propose LRG1 functions as a buffer against deleterious effects of extracellular Cyt c released from dying/dead cells. During obesity-driven adipose tissue remodeling, a rapid increase in adipocyte death contributes to increased circulating Cyt c, which could be an important pro-in ammatory signal triggering monocyte/macrophage recruitment. The pro-in ammatory setting of obesity also induces LRG1 in other adipocytes, and this increase in circulating LRG1 modulates in ammation by directly binding Cyt c. Of note, both TLR4 and LRG1 contain LRR domains, so the Cyt c-neutralizing effect by LRG1 could involve steric interference as the two proteins compete for binding Cyt c. While the current study focused on LRG1's effect on adipose tissue, it is important to emphasize that LRG1-mediated anti-in ammatory effects were systemic, as observed by the decrease in in ammatory foci in LRG1-overexpressing db/db liver. As hepatocyte apoptosis contributes to progression of NASH (non-alcoholic steatohepatitis) in NAFLD (Feldstein et  Our study suggests that induction of LRG1 in these settings could be a compensatory mechanism to modulate in ammation.
Enhanced renal clearance of Cyt c in LRG1-KO mice provides further evidence that LRG1 complexes with Cyt c in circulation and highlights the need to measure LRG1 levels in future studies utilizing extracellular Cyt c as a biomarker. Similar serum Cyt c half-lives between WT and LRG1-KO suggest LRG1 loss of function does not affect kidney function. Divergence in Cyt c clearance pro les occurs towards the plateau, and we believe this difference represents the portion of Cyt c that stays bound to LRG1 in WT animals. In ammation in obesity, especially in B6 DIO mice, is chronic and relatively low grade, so enhanced clearance of Cyt c in LRG1-KO mice could serve to counteract Cyt c release and protect from its deleterious effects. It is possible therefore that the effect of LRG1 loss of function may be more pronounced in conditions involving acute or high-grade in ammation.
In summary, we identi ed LRG1 as a novel adipokine and discovered its metabolic role as an insulin sensitizer and suppressor of in ammation. LRG1's in vivo function is associated with Cyt c binding, and this interaction sheds light on previously unappreciated molecular players of adipocyte-macrophage interaction during adipose tissue expansion and remodeling. LRG1 or targeting extracellular Cyt c could be an attractive therapeutic approach for treatment of not only obesity but a variety of in ammatory conditions.

Metabolic characterization of mice
For studies involving diet-induced obesity, mouse body weights were monitored once a week, providing fresh 60% high fat diet (Research Diets) at least once a week. For fasting blood glucose measurements, mice were single housed in the morning in cages with fresh bedding and access to water but without food. Mice were kept in a procedure room free of noise or vibration throughout the experiment. After 6 h, blood was collected from the tail vein, and glucose levels were measured using a glucose meter. For plasma insulin ELISA, blood was also collected in EDTA-coated capillary tubes, which were centrifuged at 2000×g for 15 min at 4ºC to collect plasma. Insulin ELISA was performed using Ultra Sensitive Mouse Insulin ELISA Kit (Crystal Chem). Glucose (GTT) and insulin tolerance tests (ITT) were performed following a similar 6 h fasting procedure and started by intraperitoneally injecting at time 0 indicated doses of glucose or Novolin R human insulin (Novo Nordisk). Following injection, blood glucose measurements were taken from the tail vein at indicated timepoints. During an ITT procedure, mice with glucose measurements below 20 mg/dL or showing signs of hypoglycemia were rescued by 1 g/kg glucose IP injection and excluded from the study.
Generation of LRG1-KO mice CRISPR guide RNAs were designed using CRISPOR.org (Concordet and Haeussler, 2018) and were used as two-part synthetic crRNA and tracrRNA (Alt-RTM CRISPR guide RNA, Integrated DNA Technologies, Inc). Cas9 protein, crRNA, and tracrRNA were assembled to ctRNP using protocols described previously (Shola et al., 2021). Two crRNAs were assembled to ctRNPs and electroporated to one-cell-stage mouse embryos to assess their e ciency in generating indels on the Exon 2 of Lrg1 gene. To prepare for the microinjection mix, crRNA-B which binds to genomic target sequence "AATCTCGGTGGGACCATGGCAGG" was selected for its high on-target e ciency and low off-target potential. The nal injection mix was made of 0.6 µM of guide RNA (crRNA + tracrRNA) and 0.3 µM of Cas9 protein according to protocols

Cells
Mouse primary stromal vascular fraction (SVF) cells were obtained from adipose tissues of 6-to 8-weekold male mice by collagenase digestion and plated on collagen I-coated dishes. SVF cells from epididymal white adipose tissue (eWAT) were grown in ITS media containing 1.5:1 mixture of low-glucose DMEM:MCDB201 supplemented with 2% FBS (Gemini), 1% ITS premix (Corning), 0.1 mM L-ascorbic acid 2-phosphate (Sigma), 10 ng/mL bFGF (Thermo), 0.5% penicillin/streptomycin (P/S, Gibco), and 0.2% primocin (InvivoGen). SVF cells from inguinal white adipose tissue (iWAT) and interscapular brown adipose tissue (BAT) were grown in DMEM/F-12 GlutaMAX medium (Gibco) containing 10% FBS and 1% P/S. Once grown to con uence, differentiation and maintenance of primary adipocytes across cell types were done using DMEM/F-12 GlutaMAX medium containing 10% FBS and 1% P/S. eWAT and iWAT SVF cells were induced to differentiate with an adipogenic cocktail (0.5 mM IBMX, 1 µM dexamethasone, 1 µM rosiglitazone, and 850 nM insulin) for the rst 2 days, followed by 2 days of 1 µM rosiglitazone and 850 nM insulin, after which the cells were maintained in 850 nM insulin for additional 2-4 days. SVF cells from BAT were differentiated as above but with 17 nM insulin. Experiments with primary adipocytes were performed between days 6 and 8 of differentiation. All cultured primary adipocytes were checked for lipid accumulation under a phase-contrast microscope before studies. Primary SVF and adipocytes were maintained at 37ºC with 10% CO 2 .
Bone marrow-derived macrophages (BMDMs) were obtained from 8-to 10-week-old males. Femurs and tibias were dissected, cleaned, and sterilized with ethanol before ushed of bone marrow cells, which were plated onto petri dishes. Bone marrow cells were differentiated in RPMI-1640 medium (Gibco) supplemented with 20% heat-inactivated FBS (Sigma), 1% P/S, and 100 ng/mL M-CSF (Biolegend) for 6-7 days, changing media every 2-3 days. Differentiated BMDMs were washed, trypsinized, and plated onto TC-treated culture plates for overnight before studies were performed. Experiments with BMDMs were performed between days 6 and 7 of differentiation. BMDMs were differentiated and maintained at 37ºC with 5% CO 2 .
HEK293A cells were purchased from Invitrogen and grown using 4.5 g/L glucose DMEM (Gibco) supplemented with 10% FBS and 1% P/S at 37ºC with 5% CO 2 . Cells with passage number under 20 were used for adenovirus production. Cells were validated to be mycoplasma free.

Conditioned medium generation
On day 6 of differentiation, primary adipocytes on collagen-coated 6-well plates were washed twice with warm PBS and pulsed with 1 mL/well of Met-free DMEM containing 10% dialyzed FBS, 1% P/S, 17 nM or 850 nM insulin, and either 0.1 mM AHA or 0.1 mM Met. Following 24 h incubation at 37ºC with 10% CO 2 , conditioned media (CM) from 6 wells (1 plate) were collected and pooled, ltered through a 0.22 µm PES membrane syringe lter unit, and supplemented with ½ tabs of EDTA-free cOmplete mini protease inhibitor cocktail (Roche) and PhosSTOP (Roche). CM was concentrated using a 3 kDa centrifugal lter unit (Millipore).

AHA administration and serum collection
Mice were injected with 0.1 g/kg/day AHA or PBS IP for 2 consecutive days and sacri ced 24 hours following the second injection. Following decapitation, truncal blood was collected, allowed to clot for 15 min at room temperature, and centrifuged at 2000×g for 15 min at 4 ºC to collect serum.
In-gel uorescence analysis Concentrated CM or serum was dialyzed with phosphate-buffered RIPA (10 mM phosphate buffer pH 7.2, 1% Triton X-100, 0.1% Na deoxycholate, 0.1% SDS, 140 mM NaCl) supplemented with EDTA-free cOmplete mini protease inhibitor cocktail (Roche) and PhosSTOP (Roche) using a 3 kDa centrifugal lter unit (Millipore) and protein concentration was determined using Pierce BCA Protein Assay Kit (Thermo Scienti c) using a dilution series of bovine serum albumin as protein standards. Copper(I)-catalyzed azide-alkyne cycloaddition reaction with TAMRA-alkyne (Invitrogen) was performed by mixing 200 µg of CM proteins with 0.1 mM TAMRA-alkyne, 1 mM TCEP, 0.1 mM TBTA, and 1 mM of CuSO 4 in phosphatebuffered RIPA and rotated end-over-end for 1 h at room temperature under protection from light.
Following methanol/chloroform precipitation, the dried protein pellet was dissolved in Laemmli loading buffer. Following polyacrylamide gel electrophoresis, the gel was brie y washed with distilled H 2 O and imaged with a Typhoon 5400 imager (GE Healthcare) using 532 nm excitation and a 580 nm detection lter.

Enrichment of labeled proteins
Azide-labeled nascent protein in the concentrated CM was enriched using Click-iT™ Protein Enrichment Kit (Invitrogen). Serum was diluted 1:1 with the lysis buffer provided with the kit and subjected to enrichment. Enrichment and resin wash was performed following the protocol from Eichelbaum and Krijgsveld (2014).

On-bead digestion
Extensively washed beads were incubated with Lys-C endopeptidase (Wako) in 4 M urea and 0.14 M ). Oxidation of methionine and protein N-terminal acetylation were allowed as variable modi cations, cysteine carbamidomethyl was set as a xed modi cation, and two missed cleavages were allowed. The "match between runs" option was enabled, and false discovery rates for proteins and peptides were set to 1%. Protein abundances measured using label free quantitation (Tyanova et al., 2016).

Proteomic data analysis
Proteomic datasets were analyzed using Perseus v1.6.14.0 (Tyanova et al., 2016). Of the detected proteins, those agged as reverse, only identi ed by site, and potential contaminants were excluded from the analysis. For quantitative analysis, LFQ or iBAQ intensities were employed as indicated; LFQ intensities were used for comparisons across samples, while iBAQ intensities were used to compare abundances across different proteins. Imputation of undetected data points for Log 2 (LFQ) intensities

Secretion prediction analysis
UniProt accession IDs of the detected proteins were submitted to Retrieve/ID mapping tool on the UniProt website (https://uniprot.org) to obtain the FASTA sequences, which were used as inputs for various secretion prediction algorithms using the web-based query system. We de ned classically secreted proteins as having SignalP5.0 score > 0.5 and 0 or 1 predicted transmembrane domains by TMHMM2.0. Subcellular localization prediction analysis was performed using DeepLoc1.0 and searched for proteins whose predicted location is extracellular. PredGPI speci city score > 99% was used as the threshold to determine if a protein is expected to be GPI-anchored. Finally, proteins with SecretomeP2.0 score > 0.6 and SignalP5.0 score ≤ 0.5 were considered non-classically secreted.

Cluster analysis and functional annotation
Hierarchical clustering was performed on z-score-transformed Log 2 (LFQ) values using the complete- Genes with a Log2 fold change greater than 4 and a Benjamini-Hochberg-corrected FDR of 0.05 within pair-wise comparisons were considered signi cantly enriched. Genes were further scored by the total number of pair-wise comparisons where genes were found to be enriched in both adipose tissues, brown adipose tissue or white adipose tissue compared to other tissues in the tissue atlas.
RNA isolation, cDNA synthesis, and RT-qPCR Total RNA was extracted from cultured cells using RLT buffer (Qiagen) and from tissues using TRIzol (Invitrogen) and puri ed using RNeasy Mini Kit (Qiagen). cDNA was synthesized from 1 µg of RNA using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosciences). Power SYBR Green (Life Technologies) was used for RT-qPCR reactions performed with QuantStudio 6 Flex Real-Time PCR System (Thermo Scienti c) in a 384 well format. Relative fold changes of mRNA levels were calculated using the ΔΔCT method with 18S rRNA as loading control. qPCR primers are provided in Supplementary  Table 1.

Adipose tissue fractionation
Adipose tissues from 8-week-old C57BL/6J WT male mice were dissected and minced. eWAT and iWAT were digested in a buffer containing 10 mg/mL collagenase D (Roche), 2.4 mg/mL Dispase II (Roche), and 10 mM CaCl 2 in PBS. For BAT, 2x BAT digestion buffer containing 125 mM NaCl, 5 mM KCl, 1.3 mM CaCl 2 , 5 mM glucose, 1% P/S, and 4% BSA was prepared, which was diluted 1:1 with PBS and used to dissolve collagenase B (Roche) at a nal concentration of 1.5 mg/mL. Following collagenase digestion of the tissues in a 37ºC water bath, the mature adipocyte fraction was separated from the SVF pellet by centrifugation at 500×g for 10 min at 4ºC. The two fractions were transferred to two separate tubes, washed with DMEM/F-12 GlutaMAX containing 10% FBS and 1%P/S, and vortexed in TRIzol for RNA extraction.

RNA-Sequencing and immune cell deconvolution
Extracted RNA samples were analyzed for RNA integrity number (RIN) using the Bioanalyzer (Agilent) and sequenced using Illumina NovaSeq at the Rockefeller University Genomics Resource Center. Reads were trimmed with Cutadapt, aligned to mm10 reference genome using STAR, and quanti ed using featureCounts. Differential gene expression analysis was performed using DESeq2 (Love et al., 2014).

Tissue origin prediction
Mouse tissue mRNA sequencing data from ENCODE was downloaded from GSE36026. Reads were mapped and quanti ed as above and gene expression was normalized using DESeq2. Genes detected in mouse nascent serum were selected for t-SNE analysis. Brie y, average normalized expression in a tissue was divided by summed expression across tissues. Tissue with the highest relative expression was designated as the highest expressing tissue for a gene. t-SNE analysis was performed on relative expression values with R package Rtsne (https://github.com/jkrijthe/Rtsne) using a perplexity of 30 and maximum iteration of 1,000.

Immunoblot
Upon collection, conditioned medium (CM) was ltered using a 0.22 µm PES membrane syringe lter unit to remove cell debris. CM was concentrated using a 3 kDa centrifugal lter units (Millipore), and protein concentration was determined using Pierce BCA Protein Assay Kit (Thermo Scienti c) using a dilution series of bovine serum albumin as protein standards. Mouse serum samples were loaded at equal volume. Pre-cast polyacrylamide gels were used for electrophoresis, after which protein was transferred to PVDF membrane using standard techniques. Immunoblots were incubated with indicated primary antibodies and developed using Western Lightning Plus-ECL (PerkinElmer) and imaged on an autoradiographic lm or using a Bio-Rad Gel Doc system.
Crude adenovirus was produced by transfecting PacI (NEB)-linearized pAd vectors into HEK293A cells (Invitrogen), which were incubated at 37ºC with 5% CO 2 for 10-14 days with media supplementation every 3-5 days until most cells showed cytopathic effect/detachment. Both cells and the culture medium were collected, lysed by 3 cycles of freeze-thaw between dry ice-ethanol and room-temperature water baths, and centrifuged at 3500×g for 15 min at 4ºC to obtain the supernatant crude virus. Round 1 ampli cation product was obtained by transducing HEK293A cells with the crude virus and repeating the above collection, lysis, and centrifugation steps.
To obtain round 2 ampli cation product, twelve 15 cm plates of HEK293A cells were transduced with round 1 adenovirus and incubated at 37ºC with 5% CO 2 until most cells demonstrated cytopathic effect.
As with previous rounds, cells and media were collected, lysed by freeze-thaw cycles, and centrifuged to obtain the supernatant. The supernatant was treated with benzonase, and adenoviral particles were puri ed from the crude mixture using the Vivapure AdenoPACK 100 kit (Sartorius). Puri ed virus was dialyzed with buffer containing 20 mM Tris pH 8, 25 mM NaCl, and 2.5% (w/v) glycerol and concentrated using a 100 kDa centrifugal lter unit provided with the kit. Titer of the adenovirus was determined using Adeno-X Rapid Titer Kit (Takara).
In vivo adenovirus/AAV8 transduction In vivo adenoviral transduction studies were performed using puri ed adenovirus from second round of ampli cation. Adenovirus was injected at a dose of 10 10 pfu/mouse. AAV8 was injected at 10 11 GC/mouse. The mice were brie y anesthetized with iso urane for virus injection via the retroorbital route.
Following injection, the mice were quarantined in an ABSL-2 housing room for 72 h before transferred back to regular housing conditions.

H&E section preparation/CLS quanti cation
Dissected tissues were xed in 10% neutral buffered formalin for 3 days at room temperature and and kept in ice until further processing. During blood collection, each mouse was placed on a metal grating above a clean plastic wrap to allow collection of excreted urine, if any. Plasma was isolated via centrifugation at 2000×g for 15 min at 4ºC. Immunoblot against Cyt c was performed with WT and KO plasma samples run pairwise to enable relative quanti cation of Cyt c signal. Quanti cation was performed using ImageJ (Schneider et al., 2012).

Statistical analysis
Unless otherwise noted, data are presented as mean ± SEM, with n number speci ed in the gure legends. Statistical analyses were performed with GraphPad Prism 9. Binary comparisons were performed with Welch's t-test to account for possible difference in variance. Statistical analysis of data involving 3 or more conditions (levels) of a single variable was performed using one-way ANOVA followed by Dunnett post hoc tests to compare every mean with a control mean. Data measured across multiple time points as in GTT and ITT were analyzed with repeated measures two-way ANOVA, reporting group factor Pvalues. Analysis of data from a two-factor experimental setup was performed with two-way ANOVA or two-way mixed effects ANOVA in the case of an uneven n number, reporting group factor P-values. For post hoc tests, Tukey method was used when comparing every mean with every other mean and Šídák method was employed when a selected set of means were compared.

COMPETING INTERESTS
The authors declare no competing interests.

DATA AVAILABILITY
The data that support the ndings of this study are available from the corresponding author upon request. Reagents including unique biological materials are available from the corresponding author upon request. Proteomic and RNA-Seq data will be deposited to a public repository with accession codes available before publication.

CODE AVAILABILITY
Publicly available codes are available in the relevant references. Custom script is available upon request.    AAV injection was performed in 6-week-old B6 males. HFD was started at 8 weeks of age. f, Plasma western blot of LRG1 2 weeks after AAV injection. g, Body weights of AAV-transduced B6 male mice during HFD challenge. h, Weights of dissected tissues from AAV-transduced B6 male mice at 21 weeks of age (13 weeks on HFD). i, 6 h fasting blood glucose levels of AAV-transduced B6 male mice during HFD feeding. j, Intraperitoneal glucose tolerance test (1.5 g/kg) in AAV-transduced B6 males at 18 weeks of age (10 weeks on HFD). k, Insulin tolerance test (1.5 U/kg) in AAV-transduced B6 males at 20 weeks of age (12 weeks on HFD). AAV cohort consisted of n = 12 per group. Data are presented as mean ± SEM.
Scale bars indicate 1 cm. Data are presented as mean ± SEM.  LRG1 binds extracellular cytochrome c and blocks its pro-in ammatory effect on macrophages a, α-FLAG co-IP of C-terminally FLAG-tagged LRG1 and Cyt c from the serum of AAV-transduced db/db mice at 7 weeks of age. b, Western blot of Cyt c in the serum of db/db or littermate m/m mice at 7 or 10 weeks of age. c, Western blot of Cyt c in the serum of B6 mice on standard chow or HFD for indicated weeks. HFD was started at 6 weeks of age. Developed from the same membrane in Fig. 3f. d, Western blot of Cyt c in