Altered Molecular Regulation of TUG Is a Central Feature of Insulin-Resistant Human Adipose Tissue

Abstract

White adipose tissue (WAT) insulin resistance (IR) is a central feature of metabolic syndrome; however, data regarding defects in WAT insulin signaling in humans with IR is limited. To determine which defects in WAT insulin signaling are associated with human IR, WAT was obtained from three cohorts of patients with obesity. In a bariatric surgery cohort (RESOLVE), subcutaneous WAT (n = 24) was collected before and after weight loss, and RNA sequencing was performed. In another bariatric surgery cohort (SODA), glucose- or fructose-sweetened beverages were consumed before subcutaneous and omental WAT collection, and proteomic data were collected (n = 16). In an adolescent cohort, subcutaneous WAT (n = 14) was collected before and during hyperinsulinemic clamps, and both quantitative PCR and immunoblotting were performed. The TC10–tether containing a UBX domain for GLUT4 (TUG) pathway regulates GLUT4 translocation and glucose uptake in insulin-responsive tissues. Expectedly, in the adipose tissue from all three cohorts, GLUT4 content decreased in those with IR. TUG, which traps insulin-responsive GLUT4 vesicles in intracellular pools, was increased in the setting of IR in all three cohorts. Furthermore, expression of multiple components of the TC10–TUG pathway was altered with IR. Therefore, human WAT IR is characterized by altered molecular regulation of the TC10–TUG pathway, underscoring the importance of this pathway to WAT metabolic health.

Article Highlights

• There is a paucity of data regarding defects in insulin signaling in insulin-resistant human white adipose tissue (WAT).

• The tether containing a UBX domain for GLUT4 (TUG) protein, which retains GLUT4 vesicles, was increased in WAT of participants with greater insulin resistance in three different cohorts with obesity.

• Components of the TUG regulatory signaling pathway were differentially expressed between participants with greater insulin resistance and those with greater insulin sensitivity.

• TUG may provide an important pharmacologic target in the treatment or prevention of metabolic dysfunction in patients with obesity.

Graphical Abstract

Introduction

Adipose tissue insulin resistance is a central feature of metabolic syndrome (1). It is characterized by a diminished suppression of lipolysis, an impairment in lipid storage, and a reduction in insulin-stimulated glucose uptake (2–5). Together, these features of white adipose tissue (WAT) insulin resistance increase circulating nonesterified fatty acids (i.e., free fatty acids [FFAs]), a well-described component of human insulin resistance (6). Dysregulated WAT lipolysis leads to increased fatty acid flux to tissues subject to lipotoxicity, such as the liver, skeletal muscle, cardiac muscle, and kidneys. Increased fatty acid flux to the liver plays a role in metabolic dysfunction–associated steatotic liver disease (MASLD) and is involved in the increased hepatic gluconeogenesis seen in individuals with insulin resistance (7–9).

The mechanisms responsible for the development of adipose tissue insulin resistance and WAT dysfunction are not thoroughly described in humans. Low-grade inflammation, hypoxia, and plasma membrane sn-1,2-diacylglycerol accumulation are all promotors of insulin resistance in rodent WAT and are thought to cause dysfunction mainly by attenuating canonical insulin signaling through AKT (10–13); however, whether these mechanisms apply to human adipose tissue insulin resistance is yet to be determined.

Adipocyte insulin action suppresses lipolysis, induces adipocyte lipid uptake from circulating lipoproteins, and increases glucose uptake. In insulin resistance, WAT presents with an impaired suppression of lipolysis and a decrease in WAT insulin-stimulated glucose uptake (3,4). Dysregulation of adipocyte lipid handling is directly associated with metabolic disease; lipid accumulation in ectopic tissues instead of WAT leads to whole-body insulin resistance, a critical component of metabolic syndrome and type 2 diabetes (14). Although the connection between dysregulated adipocyte glucose uptake and metabolic dysfunction may be more indirect, it is associated with both dysfunctional adipose tissue lipid storage and whole-body insulin resistance (15).

Insulin stimulates glucose uptake through GLUT4 glucose transporter translocation from intracellular vesicle membranes to the plasma membrane, and this action is reduced in WAT in the setting of insulin resistance (4,16). Insulin acts through multiple signaling pathways to translocate GLUT4 (17,18). The canonical signaling pathway involves AKT-dependent phosphorylation of AS160 (TBC1D4), which modulates the activity of specific Rab GTPases to direct vesicle trafficking. Insulin also signals through an AKT-independent pathway to activate the ρ family GTPase TC10 (RhoQ), which is coupled through its effectors that act at various steps of vesicle trafficking (17,18).

The main pool of GLUT4 that is mobilized to the cell surface by insulin stimulation is sequestered in intracellular GLUT4 storage vesicles (GSVs; also termed insulin-responsive vesicles) within unstimulated cells. GSVs are retained intracellularly by tether containing a UBX domain for GLUT4 (TUG) proteins, which anchor them at the early secretory pathway in unstimulated cells (19–21). Insulin acts through the TC10 effector PIST to stimulate site-specific TUG cleavage, mediated by the protease Usp25m, releasing the intracellular anchor and mobilizing the GSVs (22–24).

In the current study, we sought to uncover defects in adipose tissue insulin signaling that are associated with human insulin resistance. To investigate these differences, we obtained human adipose tissue from three separate studies. In each of these studies, participants were carefully evaluated for whole-body insulin action. We assessed these tissues for alterations in the abundance of either canonical or noncanonical cellular mediators of insulin action at the level of RNA and protein. These results will provide insight into the specific molecular changes that underlie adipose tissue insulin resistance.

Research Design and Methods

Participants

RESOLVE Bariatric Surgery Study

The RESOLVE (A systems biology approach to RESOLVE the molecular pathology of two hallmarks of patients with metabolic syndrome and its comorbidities; hypertriglyceridemia and low HDL-cholesterol) study is a European multicenter, oberservational intervention study on the metabolic syndrome. The methods for this study and population details have been previously described (25). Participants with severe obesity underwent Roux-en-Y gastric bypass surgery. The mean preclinical age of the participants was 45.0 ± 10.8 years. Subcutaneous biopsies were collected at baseline and 1 year postsurgery from 24 participants. FFA suppression differences pre- and postsurgery are shown in Supplementary Fig. 6C. Clinical details are listed in Table 1 of the report by Yildrim et al. (25). Insulin sensitivity was assessed using a hyperinsulinemic–euglycemic clamps (HEC) pre- and postsurgery.

Adolescent Obesity Study

A total of 14 adolescents (eight male and six female participants) with obesity underwent an oral glucose tolerance test (OGTT) and HEC. Insulin sensitivity was assessed via the whole-body insulin sensitivity index (WBISI), M values, and the HOMA for insulin resistance (HOMA-IR), obtained from the OGTT, HEC, and fasting values, respectively (26). Participant WBISI score was significantly correlated with the clamp-obtained M values (Supplementary Fig. 8). An M value <5.0 and WBISI score <1.4 defined peripheral insulin resistance (both measures sorted participants identically). HOMA-IR >4.0 indicated hepatic insulin resistance. Based on these measures, participants were grouped by insulin sensitivity as sensitive, resistant, or intermediate (i.e., peripherally sensitive and hepatically resistant); sensitive and intermediate cohorts were often combined. FFA suppression data are provided in Supplementary Fig. 6A. Groups had similar ages and body composition, but participants with insulin resistance had higher fasting insulin, C-peptide, and hepatic fat levels. Furthermore, OGTT areas under the curve for glucose, insulin, and C-peptide were higher in those with insulin resistance. The average age was 16.3 ± 1.9 years. Additional details are reported in a previous article (27). Biopsies were collected at baseline and 30 min after insulin infusion. Protein and RNA levels were assessed through immunoblotting and quantitative PCR (qPCR). Immunoblot images are shown in Supplementary Fig. 1.

SODA Study

The SODA study was a randomized, controlled, single blinded, multicenter study registered in the Netherlands Trial Registry (NTR5351). Methods and demographics for this study have been previously published (28). Subcutaneous and omental adipose tissue was collected from 24 patients undergoing bariatric surgery (four men and 20 women). The average age of participants was 44.8 ± 11.2 years. Participants were grouped by MASLD status (>5.56% liver fat), consistent with HEC data (28,29). Twelve participants had MASLD, and 12 did not. MASLD status was correlated with whole-body glucose uptake (Rd) and FFA suppression differences (Supplementary Fig. 7). Details are listed in Table 1 of the report by ter Horst et al. (28). Two hours before biopsy, participants consumed either glucose (75 g) or fructose (75 g) in 300 mL water. Protein content was assessed with liquid chromatography–mass spectrometry proteomics. Samples were selected for sufficient tissue and balanced conditions. Sixteen patients were analyzed using two TMTpro-18 plexes. Of those 16 (three men and three women), eight had MASLD, and eight did not. Data are shown for the glucose group when insulin-stimulated changes occurred; otherwise, fructose and glucose groups were combined.

RNA Sequencing

RNA sequencing was used to assess transcriptomic differences from the RESOLVE study. Collection and processing of transcriptomic RNA sequencing data were previously described for these adipose tissue biopsies (30). Sequence libraries were generated with the Ovation Ultralow System V2 (Tecan Group, Ltd., Männendorf, Switzerland). Sequencing was outsourced to GenomeScan (Leiden, the Netherlands). Paired-end reads were produced using the Illumina HiSeq 4000 system. The quality of the sequenced reads was assessed using the FastQC (version 0.11.5), dupRadar (version 1.0.0), and Picard tools. The adaptor sequences and low-quality bases were trimmed from the reads using Trimmomatic (version 0.32). Quality trimmed raw reads were aligned to the human reference genome (hg38) using HISAT2 (version 2.0.4). Counts were obtained at the gene level using HTSeq (version 0.6.1). Reading depths for this analysis were between 4.3 and 20 million reads per sample.

Proteomics

Alterations in protein abundance between groups in the SODA study were assessed via a liquid chromatography–mass spectrometry proteomic approach with ion trap mass analysis, with mass spectra processing using a COMET-based software pipeline (Supplementary Material). Data have been deposited in the ProteomeXchange Consortium via the PRIDE partner repository with the data set identifier PXD061142 (31).

Immunoblotting

Protein abundance in the adolescent obesity study was assessed via immunoblotting. Proteins from tissue lysate were resolved by SDS-PAGE using a 4–12% gradient gel and electroblotted onto polyvinylidene difluoride membranes (EMD Millipore). Membranes were blocked with 3% BSA in Tris-buffered saline with Tween-20 (TBST) and incubated in primary antibodies overnight. After washing, horseradish peroxidase–conjugated secondary antibodies (Cell Signaling Technology, Danvers, MA; 1:5,000) were applied for 1 h, and detection was performed with enhanced chemiluminescence.

Primary antibodies included HSP90 (cat. no. 610419; BD Biosciences; 1:1,000), AKT (cat. no. 2920; Cell Signaling Technology; 1:1,000), phosphorylated AKT (cat. no. 9271; Cell Signaling Technology; 1:1,000), GLUT4 (Bogan Laboratory; 1:2,000), TUG (Bogan Laboratory; 1:1,000), and TC10 (cat. no. SAB5701368; Sigma-Aldrich; 1:1,000).

Real-Time qPCR

Differences in gene expression between groups in the adolescent obesity study were assed via real-time qPCR. After tissue homogenization, RNA was extracted using the RNeasy Mini Kit (cat. no. 74106; Qiagen). The QuantiTect Reverse Transcription Kit was used to produce cDNA (cat. no. 205311; Qiagen). Gene expression for genes of interest was measured with real-time qPCR. GAPDH was used as an internal reference. Primer sequences are listed in Supplementary Table 2.

Statistical Analysis

Statistical analysis was performed using GraphPad Prism 10 (GraphPad Software, La Jolla, CA). When two distinct groups were compared, a Student t test was performed. Most t test comparisons were unpaired; however, when assessing the effect of insulin on AKT phosphorylation (Fig. 1A), a paired t test was conducted. When three groups were analyzed, an ordinary one-way ANOVA was conducted, followed by a Tukey multiple comparisons test. All data are expressed as mean ± SEM. Proteomic data from the SODA study were assessed via simple linear regression analysis. Throughout experiments, a P value of <0.05 was used to determine significance. Analysis of the raw proteomic data is described above. RNA sequencing statistical analysis was performed as previously described by Oussaada et al. (30). Briefly, differential expression was assessed using an empirical Bayes moderated t test within the limma linear model framework, including the precision weights estimated by voom and correction for possible plate effects. The consensus intraindividual correlation (function duplicateCorrelation; package limma) was included in the linear model fit.

Figure 1.

AKT, GLUT4, and TUG in the setting of insulin resistance. AKT phosphorylation (pAKT) was assessed via immunoblot in adipose tissue obtained from both the adolescent obesity cohort (A, C, E, and G) and the SODA cohort (B, D, F, and H). Insulin-stimulated tissue was compared with basal conditions (A and B), and the insulin-stimulated phosphorylation ratio was compared between participants with insulin resistance and those with greater insulin sensitivity (C and D). GLUT4 protein abundance was assessed in the adolescent obesity cohort via immunoblot (E) and in the SODA cohort via proteomic analysis (F). By identical methods, TUG protein abundance was also measured in both the adolescent obesity cohort (G) and in the SODA cohort (H). Statistical comparisons between the insulin-stimulated and basal conditions in the adolescent obesity cohort were assessed by a paired Student two-tailed t test (n = 12) (A). Other statistical comparisons from bar graphs were assessed via an unpaired Student two-tailed t test (n = 5–7 per group) (B–E and G). When grouping in the adolescent obesity study based on insulin sensitivity, the sensitive and intermediate groups were pooled in the sensitive-labeled bar. Blue dots represent the insulin-sensitive group, purple dots represent the intermediate group, and red dots represent the insulin-resistant group. For all bar graphs, data are represented as mean ± SEM. Statistical comparisons of the SODA study were assessed based on whether the slope of a simple linear regression between protein abundance and whole-body glucose uptake (Rd) was significantly nonzero (n = 15) (F and H). In each graph, every dot represents a biologic replicate. ns, not significant. **P < 0.01.

Data and Resource Availability

Proteomic data have been deposited in the ProteomeXchange Consortium via the PRIDE partner repository with the data set identifier PXD061142 (31). Additional clinical data from the RESOLVE, SODA, and adolescent obesity cohorts presented in this article can be found in the original publications (25,30,32). Additional information will be provided by the corresponding author on request. All other data generated or analyzed during this study are included in the published article (and its online Supplementary Material).

Results

No Significant Attenuation of Insulin-Stimulated AKT Phosphorylation in Insulin-Resistant Adipose Tissue

Proximal insulin signaling in human adipose tissue was assessed via immunoblotting in both the adolescent obesity cohort and in the SODA study. Expectedly, in the adolescent participants, AKT phosphorylation was increased by insulin infusion (Fig. 1A). In the SODA study, the glucose-consuming group (higher insulin levels) displayed a trend toward increased AKT phosphorylation compared with the fructose-consuming group (lower insulin levels) (Fig. 1B). In the adolescent study, insulin-stimulated AKT phosphorylation was not significantly attenuated in participants with insulin resistance, as compared with those with greater insulin sensitivity, in either cohort (Fig. 1). Similarly, in the SODA study, no significant difference in AKT phosphorylation was seen between patients with insulin resistance and MASLD and control participants (Fig. 1D).

GLUT4 Is Decreased in Insulin-Resistant Adipose Tissue

The GLUT4 glucose transporter, which facilitates insulin-stimulated glucose uptake in adipocytes, is decreased with insulin resistance (33,34). Consistent with these reports, decreased adipose tissue GLUT4 was associated with systemic insulin resistance in all three cohorts. In the RESOLVE bariatric surgery study, GLUT4 expression decreased by 27% (P = 7.8E^-10) in adipose tissue taken before weight reduction compared with tissue collected 1 year after surgery (Table 1). In the adolescent study, WAT GLUT4 protein was 60.3% lower (P < 0.05) in the insulin-resistant group (Fig. 1E). Similarly, in the SODA study, GLUT4 protein in subcutaneous adipose tissue was lower (R² = 0.37; P < 0.05) in all participants with insulin resistance (Fig. 1F).

Table 1.

RESOLVE study gene expression assessed by RNA sequencing

Gene	P	Log2 fold change (pre- vs. postsurgery)
CBL	0.045	0.11
CRK	0.12	0.09
RHOQ/TC10	0.0003	0.31
GOPC/PIST	0.0002	−0.34
GOLGA3/Golgin160	0.04	0.26
ASPSCR1/TUG	0.04	0.39
SLC2A4/GLUT4	1.47E−10	−1.54
ADRB1	0.0152	−0.345
ADRB2	0.099	−0.162

TC10 pathway and related genes from RESOLVE study biopsies were assessed before vs. after weight loss and expressed as log₂ fold change, so positive fold change values correspond to increased expression in more insulin-resistant adipose tissue before surgery. Statistical comparisons were assessed using an empirical Bayes moderated t test within the limma linear model framework.

TUG Is Increased in Insulin-Resistant Adipose Tissue

With no major changes in proximal insulin signaling, we examined an alternative mediator of insulin action, TUG. TUG, which traps insulin-responsive GSVs intracellularly, was significantly increased in participants with insulin resistance. In the RESOLVE study, insulin-resistant WAT taken before bariatric surgery expressed 39% more TUG (P < 0.05) than the insulin-sensitive tissue biopsied 1 year after surgery (Table 1). In the adolescent cohort, TUG protein from insulin-resistant adipose tissue was increased by 49% (P < 0.01) compared with the more insulin-sensitive tissue (Fig. 1G). In the SODA study, TUG protein abundance in subcutaneous adipose tissue was significantly correlated (R² = 0.28; P < 0.05) with insulin resistance (Fig. 1H).

Many Proteins and Genes in the TC10 Pathway Have Altered Abundance and Expression in Insulin-Resistant Adipose Tissue

In addition to increased TUG expression and abundance in participants with insulin resistance, many proteins and genes throughout the TC10 pathway are also altered (Fig. 2A). In the RESOLVE cohort, the insulin-resistant adipose tissue samples (before bariatric surgery) had increased gene expression in Cbl (P < 0.05), TC10 (P < 0.001), and PIST (P < 0.001) (Table 1). In the adolescent cohort, insulin-stimulated TC10 protein abundance was greater (P < 0.001) in participants with insulin sensitivity (Supplementary Fig. 2A). In the SODA study, the subcutaneous adipose tissue of those with greater insulin resistance had a significant increase in PIST protein abundance (P < 0.05) after participants consumemd a glucose-sweetened beverage (Supplementary Fig. 4E). TRARG1, a regulator of GLUT4 trafficking, was also significantly altered in both the adolescent (Supplementary Figs. 2B and 3G) and SODA cohorts (Supplementary Fig. 4H); however, the change had no clear directionality.

Figure 2.

Heat map of significant alterations in the TC10–TUG pathway. The heat map allows visualization of significant changes in either protein or RNA assessed in all studies discussed. White boxes signify that the step was not measured in that study, whereas gray boxes signify no significant change. The table at the left provides the color code for different levels and directions of significance. When a box is outlined in green, that protein was assessed only in an insulin-stimulated condition. Statistical comparisons used for each study are outlined in the Research Design and Methods section. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

TUG and the TC10 Pathway Are Altered in Subcutaneous Adipose Tissue With Insulin Resistance, but Not in Omental Adipose Tissue

In the SODA study, both subcutaneous and omental adipose tissue were collected. Although proteomic analysis determined that both adipose tissue beds had a downregulation of GLUT4 protein in participants with greater insulin resistance (subcutaneous P < 0.05; omental P < 0.01), only the subcutaneous adipose tissue displayed any differential regulation of TUG or TC10 pathway proteins in relation to insulin resistance (Fig. 2A and Supplementary Fig. 4).

β-Adrenergic Receptors (ADRB1 and ADRB2) Are Altered in Insulin-Resistant Human Adipose Tissue

Adipose tissue lipolysis is suppressed with insulin stimulation and increased with catecholamine signaling. In the adolescent obesity study, ADRB1 protein was significantly upregulated (P < 0.01), and ADRB1 gene expression was significantly downregulated (P < 0.01), in insulin-resistant adipose tissue (Fig. 3A and B). In the same participants, ADRB2 gene or protein expression was not significantly altered by insulin sensitivity. Furthermore, in the RESOLVE bariatric surgery study, ADRB1 was significantly decreased (P < 0.05) before weight loss (Table 1).

Figure 3.

ADRB1, but not ADRB2, is altered with insulin resistance. Both ADRB1 and ADRB2 protein abundance (A and C) and gene expression (B and D) were assessed in the adolescent obesity cohort between the insulin-sensitive (n = 3–4), intermediate (n = 2–3), and insulin-resistant groups (n = 7). Data are represented as mean ± SEM. Statistical comparisons were completed by an ordinary one-way ANOVA, followed by a Tukey multiple comparisons test. Each dot represents a biologic replicate.

Discussion

Insulin resistance is a central feature of WAT dysfunction; however, the molecular defects in insulin-resistant human adipose tissue are yet to be fully elucidated (3,15). To better understand how insulin signaling is altered in insulin-resistant human adipose tissue, we assessed WAT obtained from three human cohorts comprising participants with a wide range of insulin sensitivity: 1) the RESOLVE study, 2) an adolescent obesity study, and 3) the SODA study. In the RESOLVE study, RNA sequencing was used to assess adipose tissue transcriptomic differences before and 1 year after bariatric surgery. In the adolescent obesity study, adipose tissue biopsies were collected from adolescents with obesity with a wide range of insulin sensitivity both at baseline and 30 min into an insulin infusion. In the SODA study, adipose tissue biopsies were collected from adults with obesity who had a wide range of insulin sensitivity after either a fructose or a glucose challenge.

Consistent with previous reports (33), GLUT4 gene expression and protein abundance were substantially decreased in insulin-resistant human adipose tissue. Unique to this study, we found TUG, the protein tether that retains GSVs intracellularly, to be significantly upregulated in both gene expression and protein abundance in more insulin-resistant adipose tissue across all three human cohorts. Furthermore, the molecular regulation of the insulin-responsive TC10 pathway, which regulates TUG action and GLUT4 translocation, was also significantly altered in insulin-resistant adipose tissue.

Defects in proximal insulin signaling are well established in insulin-resistant human liver and skeletal muscle, and the dampening of insulin-stimulated proximal phosphorylation events in these tissues is dramatic (28,35,36). In contrast, in the current study, the difference in proximal insulin signaling between human participants with varying degrees of insulin resistance was subtle. This suggests that the current study was underpowered to detect differences in proximal insulin signaling, an unexpected result considering the prior findings in the liver and muscle. A different study design may allow future investigations to tease out the role of proximal insulin resistance in human adipose tissue. For example, a lower dose of insulin or a shorter period of insulin stimulation may be useful, analogous to the design of rodent studies (13,37). In the current study, the relatively insulin-sensitive control groups were largely composed of individuals with obesity; however, a study including lean control participants with insulin sensitivity might demonstrate larger differences in insulin-stimulated phosphorylation events. Finally, although it seems likely that defects in classical proximal insulin signaling are conserved between insulin-resistant rodents and humans, perhaps proximal defects do not have the same primacy in human adipose tissue as they do in human muscle or liver. Future investigations will be required to clarify the relative importance of proximal insulin-signaling defects in insulin-resistant human adipose tissue.

In contrast with the small inconclusive differences in proximal insulin signaling between participants with insulin sensitivity and those with insulin resistance, large differences were seen in GLUT4 abundance, consistent with previous reports. Adipocyte glucose uptake occurs through GLUT4, an insulin-regulated glucose transporter, which has been established to be downregulated in the adipose tissue of humans with insulin resistance (33,38). Data from this study are consistent with this finding; GLUT4 was decreased in insulin-resistant adipose tissue across all three human cohorts.

In the absence of insulin, the main pool of GLUT4 remains in intracellular GSVs retained by TUG proteins (19–21). The TC10–TUG pathway is a well-described insulin-signaling pathway in rodents, and this pathway directly influences tissue glucose uptake; however, the clinical relevance of this pathway has not been previously established, and defects in this pathway in humans with insulin resistance have yet to be described. In this study, we observed that TUG was increased in insulin-resistant human adipose tissue. This is consistent with increased intracellular GSV retention, decreased GLUT4 translocation, and a subsequent decrease in glucose uptake, as previously observed in insulin-resistant adipose tissue (4). In addition to the increase in TUG in insulin-resistant adipose tissue, significant changes were observed in the insulin-responsive TC10 pathway in all three human cohorts. Although there were distinct clusters of insulin sensitivity to compare in all three cohorts, these studies were otherwise unique in terms of patient populations and methods of analysis. Despite the differences in study design, robust and consistent changes to the molecular regulation of the TC10 pathway were observed across all three human cohorts, consistent with an important role for this pathway in the pathogenesis of adipose tissue insulin resistance.

The importance of TUG in patients with obesity is consistent with what is known from other preclinical studies. In muscle, TUG cleavage accounts for ∼85–90% of insulin-stimulated GLUT4 translocation to the cell surface and therefore the majority of insulin-stimulated glucose uptake (39). Of note, in insulin-resistant rodents, TUG cleavage is impaired, and Usp25m abundance is reduced, both in adipose tissue and muscle (24,39). Conversely, sustained caloric restriction in humans ameliorates insulin resistance and is accompanied by selectively decreased TUG and increased Usp25m and AS160 protein abundance in adipose tissue, consistent with improved insulin-stimulated TUG cleavage (40). This is not accompanied by altered abundances of PI3K pathway proteins, such as AKT2 or PDK1, suggesting that the improvement in insulin-stimulated glucose uptake may be mediated, at least in part, by increased GLUT4 targeting to GSVs during the fasted state. This notion is consistent with older literature, which described altered GLUT4 targeting during fasting in those with insulin resistance, both in adipose tissue and in muscle (34,41).

The adipose tissue biopsies collected from the SODA study allowed for a direct comparison between subcutaneous and omental adipose tissue depots. Previous studies have shown differences in insulin action between subcutaneous and omental adipose tissue. Overall, the antilipolytic effect of insulin is greater in subcutaneous adipose compared with omental tissue, but conversely, insulin-stimulated glucose uptake seems lower (42,43). Notably, differences in depot-specific insulin action resulting from insulin resistance have been inconsistent (44,45). In the current study, neither insulin receptor nor AKT content in either subcutaneous or omental adipose tissue was correlated with insulin resistance. Nevertheless, the TC10 pathway proteins PIST and TUG were both significantly increased in the setting of insulin resistance in subcutaneous adipose tissue. No differences in any insulin-signaling proteins were seen in omental adipose tissue. The presence of molecular changes in insulin resistance in subcutaneous but not visceral adipose tissue was consistent with a model in which adipose tissue insulin action at a systemic level was largely attributable to the subcutaneous adipose tissue compartment. This is consistent with studies in mice where insulin-stimulated glucose uptake was blunted in subcutaneous adipose tissue compared with mouse epididymal adipose tissue, another visceral depot located near the epididymis and often studied in rodents (4). Although the current study primarily pertains to the regulation of GLUT4 trafficking, that molecular hallmarks of insulin sensitivity and resistance were primarily found in subcutaneous adipose tissue may one day be extrapolated to the relative importance of the two depots in terms of insulin-mediated suppression of lipolysis as well.

With the data presented, a connection between insulin resistance in human participants and alterations in the molecular regulation of the TC10–TUG pathway has been established; however, the key connection between adipose insulin resistance and whole-body insulin resistance lies in the failure of adipose tissue to properly suppress lipolysis and offload triglyceride-rich lipoproteins, functional deficits that drive lipid-induced insulin resistance and lipotoxicity (46). The TC10–TUG pathway directly regulates glucose uptake, not lipolysis, but a potential connection between the two fluxes is through TC10 action and beta1 adrenergic receptor recycling. PIST stabilizes the β1 adrenergic receptor in the trans-Golgi, reducing catecholamine-stimulated adenylate cyclase activation (47). Therefore, a reduction in TC10 pathway action would be expected not only to decrease GLUT4 translocation but also to increase β1 adrenergic receptor translocation. The increase in plasma membrane adrenergic receptors would increase intracellular cAMP for any given catecholamine stimulus, thereby reducing adipocyte insulin sensitivity (and reducing the inhibition of adipose tissue lipolysis) in a postprandial setting, where circulating catecholamines are increased (48,49). Our findings of increased β1 adrenergic receptor protein, with a potentially feedback-related decrease in gene expression, may be consistent with this global model of adipocyte insulin resistance. Further investigation is required to validate this model.

Our data establish a connection between the regulation of the TC10–TUG pathway and human adipose tissue insulin resistance; however, as with any study, there are limitations that should be recognized and addressed in future studies. Although the examination of adipose tissue from three different human cohorts is a strength of this study, human tissue collection is often suboptimal for the purposes of molecular biology and cellular signaling analyses. In both the RESOLVE and SODA studies, where adipose tissue was collected during bariatric surgery, the focus was clinical, so the timing of tissue collection may not have been optimal for examining proximal insulin signaling. Furthermore, the quantity of tissue that was collected was limited, so we were not able to apply all techniques (i.e., mass spectrometry, immunoblotting, quantitative PCR, and RNA sequencing) to all samples. The use of different technologies to address the same set of questions is a strength of this study; however, the inability to use every technology in the analysis of all samples precluded a direct comparison of each of our cohorts. When measuring TUG protein abundance, only intact TUG protein is reported in this article. Through the action of TC10 on the TUG pathway, TUG protein is cleaved to translocate GLUT4; however, the experimental conditions used in this study did not allow reliable detection of the cleaved form in human tissue samples. Furthermore, because of the limited number of tissue samples, the analyses could not be segregated by sex. Because there was no difference in TUG expression between the male and female participants (Supplementary Fig. 5), this weakness is unlikely to have led to erroneous conclusions. Rather, the presence of a correlation of TUG expression with insulin sensitivity in diverse cohorts emphasizes the robust nature of this finding.

In conclusion, the molecular regulation of TUG and its upstream TC10 regulatory pathway is altered in humans with insulin resistance, a finding that is particularly remarkable considering that in the same tissues classic proximal insulin signaling through AKT phosphorylation was only modestly altered (and not statistically significantly so). These changes were observed only in subcutaneous adipose tissue, implying that changes in whole-body adipose tissue insulin action are most likely driven by changes in subcutaneous rather than visceral adipose tissue. Because few prior studies of human adipose tissue have reported differences in insulin-signaling genes and proteins, this study provides valuable data that may inform future therapies targeting insulin-resistant adipose tissue.

This article contains supplementary material online at https://doi.org/10.2337/figshare.30610430.

Funding Statement

This study was supported by National Institutes of Health grants R01 DK124272 (D.F.V.), T32 DK007058 (J.W.S.), and R01 DK129466 (J.S.B.) and European Union grant FP7-EU 305707 (M.J.S.).