Relationship Between a Sedentary Lifestyle and Adipose Insulin Resistance

Insulin resistance is a major driving force behind type 2 diabetes mellitus (T2DM) (1). Many factors may influence insulin action, such as physical activity (2,3). People with a sedentary lifestyle are more prone to insulin resistance, and they also have a higher risk for developing T2DM than physically active people (4,5).

The relationship between physical activity and insulin-induced glucose uptake in skeletal muscle has been extensively examined (2,3); however, little is known about the relationship in adipose tissue. Only a small portion of glucose uptake takes place in fat cells (6,7). The major metabolic role of insulin in these cells is to lower fatty acid release through inhibition of hydrolysis of triglycerides (lipolysis) (8,9). Inhibition of lipolysis occurs at a considerably lower hormone concentration than needed for stimulation of glucose uptake (9,10). Therefore, the relationship between physical activity and antilipolysis may be different from that between physical activity and glucose metabolism.

Insulin acts through so-called spare receptors, meaning that only a fraction of the total number of receptors must be activated to obtain a maximum insulin effect (11). Therefore, insulin resistance can be due to either low hormone sensitivity (i.e., a high half-maximal effective insulin concentration) or low maximum effect (responsiveness). In the former case, the resistance is localized at receptor near events. Decreased responsiveness is localized at distal intracellular signal steps. Whether physical activity influences either or both hormone actions is not known. By performing concentration response experiments, it is possible to measure insulin sensitivity and responsiveness in a single subject.

The impact of physical activity status on insulin action in fat cells has been directly examined in one small study of men (12). Responsiveness of insulin-induced glucose metabolism was higher in 8 athletes compared with 8 untrained men. In another small study of men, lipolysis was indirectly examined by measuring interstitial glycerol (end product of lipolysis) in subcutaneous adipose tissue using microdialysis (13). Hyperinsulinemia lowered the glycerol concentration more prominently in the 8 athletes compared with the 8 untrained men, but how this relates to sensitivity/responsiveness is not known. Another open question is how the two published findings (12,13) relate to a sedentary lifestyle, sex, the relative roles of the different insulin actions on fat cells, or the influence of other factors that are important for insulin action.

Herein, we performed a cross-sectional examination of a large number of subjects to elucidate these questions about fat cells. Between 1993 and 2020, we investigated insulin action on fat cell lipolysis and lipogenesis in 540 men and women. The subjects were asked about their daily physical activity patterns, including work time, transportation, and leisure time, and were divided into sedentary (no important physical activity at any time) or active (some or more intense overall activity). In a subgroup of 69 women, the ability of hyperinsulinemia to suppress circulating fatty acid levels and stimulate glucose disposal was also examined to evaluate the in vivo impact of the in vitro findings. Retrospective information on adipose gene expression was obtained from a subgroup of 126 women.

Starting in 1993 and ending in 2020, insulin action in white, abdominal, subcutaneous fat cells was determined in 540 subjects. They were asked for their physical activity pattern. Exclusion criteria were severe chronic disease and type 1 diabetes. Subjects came to the laboratory in the overnight fasting state at 8:00 a.m. Height and body weight were determined. After resting in the supine position for 15 min, waist and hip circumferences were determined, followed by blood pressure and venous blood sampling for routine clinical chemistry analysis as previously described (14). Glucose and insulin values were used to calculate overall insulin resistance of glucose metabolism with the homeostasis model assessment of insulin resistance (HOMA-IR) method (15). Fatty acid and insulin values were used to estimate overall adipose insulin resistance (Adipo-IR) (16). The latter reflects insulin suppression of circulating fatty acid levels and hormone action on lipolysis and lipogenesis in isolated subcutaneous fat cells, as well as re-esterification, fatty acid trapping in adipose tissue, and clearance from circulation (17–19). Neither HOMA-IR nor Adipo-IR can distinguish between sensitivity and responsiveness of insulin action. Finally, abdominal, subcutaneous, adipose tissue specimens were obtained by needle aspiration. The study was approved by the regional ethics committee in Stockholm, Sweden, and explained in detail to each subject. Informed written consent was obtained from all subjects.

Isolated fat cells were prepared and used for determination of fat cell size, lipolysis, and lipogenesis as described previously(20). For lipolysis, diluted fat cell suspensions (5,000–10,000 cells/mL) were incubated for 2 h at 37°C and pH 7.4 in buffers containing glucose, albumin, adenosine deaminase (to remove adenosine, which inhibits lipolysis), and the phosphodiesterase-sensitive cAMP analog 8-bromo-cAMP (as insulin inhibits lipolysis by stimulating phosphodiesterase). The medium also contained insulin (0–70 nmol/L). Glycerol release into the medium was analyzed as a lipolysis indicator (21). Lipogenesis was measured in fat cells incubated in a final concentration of 2% (v/v) using a buffer containing glucose (1 μmol/L), albumin, trace amounts of 3-³H glucose (˜5 × 10⁶ disintegrations per min/mL), and insulin at different concentrations (0–70 nmol/L). After 2 h of incubation at 37°C (pH 7.4), radioactivity in the total incubate was determined by liquid scintillation counting. The lipogenesis method has been evaluated and discussed previously in detail (22). At the presently used glucose, concentration and type of isotope data with radioactive uptake reflect glucose incorporation into fat cell lipids under conditions when glucose transport is the major rate-limiting step. There is no consensus about how to express absolute rates of lipolysis/lipogenesis, which depends on the use of a denominator (fat cell number, fat cell surface area, lipid content, per DNA, or per protein). Therefore, we expressed the maximum action as relative using the ratio of presence of insulin divided by no insulin, which is independent of other denominators. Responsiveness was the ratio at the maximum effective insulin concentration. Insulin sensitivity was the half maximal effective hormone concentration from the concentration-response curves and transformed to the negative log₁₀ molar value (pD₂). This measure does not depend on the use of a denominator. Responsiveness and pD₂ reflect receptor distal and near events, respectively, for hormones acting through spare receptors (23). When the yield of adipose tissue was low, we prioritized lipolysis and fat cell size because these analyses required less tissue compared with lipogenesis analysis. We also performed a retrospective analysis of abdominal subcutaneous adipose tissue microarrays from two of our published studies on women having physical activity scores of 0–3 (n = 128) (24,25). In both studies, subjects were investigated repeatedly. Herein, we used data from the baseline investigation. In brief, the Affymetrix Expression Console (version 1.4.1) was used, and mRNA expression for genes coding for the insulin receptor (IR) and IR substrate 1 (IRS1) and IRS2 were retrieved as an average log₂-transformed array signal. These genes were investigated because our pharmacological data (see Results) suggested involvement of early receptor-related steps in insulin signaling events.

We retrospectively reanalyzed published data from 69 women with or without obesity (BMI ≥30 kg/m²) (26). In brief, these women were subjected to a hyperinsulinemic-euglycemic clamp. The insulin infusion rate was 0.12 units/m² body surface area following a bolus dose (1.6 units/m² body surface area/min). At −5 and 115 min of the clamp, venous blood was removed for insulin and fatty acid analysis. Whole-body uptake of glucose was determined during the 2nd hour of the clamp. In women with obesity, we used data obtained before metabolic surgery.

We used self-report to determine total physical activity (e.g., at home, at work, at leisure time, during transportation) according to a four-point scale with a score of 1 defined as predominantly sedentary activities at any time and a score of 4 defined as competitive athletic performance and/or strong exercise for >30-min periods occurring at least five times per week. Intermediate scores (2,3) were defined according to the combined activities at home, work, or leisure plus mode of transportation. A score of 2 required at least one strong 30-min bout of physical activity per week, >30 min cycling to and from work at least once per week, or at least one >30-min period of hard work at the job per week. The self-reports were collected by the same three research nurses who did the scoring. Based on the scores, the subjects were divided into two groups: sedentary (score of 1) and active (scores of 2–4). For validation, the scores were also analyzed separately. There is no consensus about how to report physical activity, and most studies use self-report (27–29). At the start of our study, there were no validated publications available on physical activity scoring measures. However, 159 of the subjects also participated in another study (Swedish Cardiopulmonary Bioimage Study [SCAPIS]), where validated questionaries of physical activity were used (30). In brief, the data of the Swedish version of the Saltin-Grimby Physical Activity Level Scale (SGPALS) score (31) were extracted from this study and compared with the presently used scoring.

Values are median with interquartile range and shown using box plots or by individual plots in the figures. The primary comparison was between the sedentary and active group using Wilcoxon rank sum test. The secondary analysis was the influence of various factors on values for insulin action using a multiple regression model that included physical activity group, sex, age, cardiometabolic diagnosis, BMI, waist-to-hip ratio, and fat cell size. Finally, individual values of physical activity score and insulin action or SGPALS score were compared using Spearman rank correlation and receiver operating characteristic (ROC) curve. Prior to termination of inclusion, we made a power calculation using linear methods and published values for antilipolysis pD₂ (20). From 92 apparently healthy subjects, the mean (SD) for pD₂ of the antilipolytic effect of insulin was 14.9 (1.1). Assuming an equal distribution of sedentary and active subjects, we could detect a 0.4 difference in pD₂ in 120 subjects of each group with 80% power and P = 0.05 using a two-sided t test, suggesting that we had adequate statistical power in the current study. One extremely high value for responsiveness of lipogenesis with a z-score of 53 was removed from the analysis.

The data sets generated and/or analyzed during the current study are available from authors D.P.A. or P.A. upon reasonable request. No applicable resources were generated or analyzed during the current study.

Table 1 displays the clinical data for all subjects. The sedentary group was younger than the active group, with less favorable values for several cardiometabolic parameters, including BMI, waist-to-hip ratio, blood lipids, fat cell volume, and diastolic blood pressure. The frequency of hyperlipidemia (defined as treatment of high total cholesterol or triglycerides) and T2DM was lower in sedentary than active subjects. Details on medication are provided in Supplementary Table 1. The proportion of women was higher in the sedentary group compared with the active group.

Insulin action data are shown in Fig. 1. Values for HOMA-IR and Adipo-IR were higher in the sedentary versus the active group. The opposite was true for pD₂ values of antilipolysis and lipogenesis. When pD₂ was transformed to mol/L, there was a 10-fold difference between groups for those measures. Maximum percentage effect of insulin (responsiveness) on inhibition of lipolysis was 2% higher in the sedentary group, but there was no statistically significant difference between groups in responsiveness of lipogenesis (twofold increase). There was a strong positive correlation between HOMA-IR and Adipo-IR (ρ = 0.89; P < 0.0001).

To investigate whether insulin action measures were independent of other factors, we used a comprehensive multiple regression model that included BMI, waist-to-hip ratio, age, sex, fat cell volume, metabolic disease (hyperlipidemia, hypertension, or T2DM) and a sedentary/active phenotype (Table 2). Physical activity was independently associated with HOMA-IR and Adipo-IR. The same was true for pD₂ (measuring insulin sensitivity) for antilipolysis but not for lipogenesis. Physical activity did not contribute to variations in responsiveness of antilipolysis (P = 0.11). Other factors contributing to variation in insulin action were sex for pD₂ of antilipolysis, BMI, waist-to-hip ratio, and T2DM for Adipo-IR and BMI, waist-to-hip ratio, hyperlipidemia, and T2DM for HOMA-IR.

The influence of obesity (BMI ≥30 kg/m²) was investigated (Supplementary Table 2). For those with or without obesity, a sedentary lifestyle was associated with lower pD₂ values for antilipolysis and higher values for Adipo-IR or HOMA-IR. However, lipogenesis was not influenced by physical activity status in those with or without obesity.

We also investigated physical activity score as a continuous variable. Only pD₂ for antilipolysis, HOMA-IR, and Adipo-IR were associated with physical activity (ρ = 0.3; P < 0.0001). Using the multiple regression model described above, the three measures of insulin action remained independently associated with physical activity score (P = 0.0004 for pD₂, P = 0.0025 for HOMA-IR, and P = 0.0055 for Adipo-IR).

The number of men was too small for a valid separate analysis. However, the results in women were essentially the same as for the whole group (data not shown).

The data with pD₂ and responsiveness suggest that the impact of physical activity score on insulin action is located at the most immediate insulin signal steps (IR, IRS1, and IRS2). We therefore correlated physical activity score with mRNA expression for these three genes by combining microarray data from two cohorts into one group (n = 128) (Fig. 2). A strong positive correlation was observed for the expression of all the genes (ρ ≥0.35).

We examined the in vivo effect of insulin in 69 women. Clinical data are provided in Supplementary Table 3. Sedentary women had a more adverse profile for many cardiometabolic measures compared with women having an active lifestyle. The subjects were investigated before (−5 min) and 115 min after the start of a hyperinsulinemic-euglycemic clamp (Fig. 3). Glucose uptake during the last hour of the 2-h clamp was BMI dependent and higher in active than in sedentary subjects, although circulating insulin levels before and at the end of clamp were higher in the sedentary group. Circulating fatty acid levels were similar in both groups before the clamp and decreased markedly during the clamp to levels that were approximately two times higher in the sedentary compared with the active group. Circulating glycerol concentrations before and during clamp were similar in both groups (data not shown).

The physical activity score was compared with the validated SGPALS score (Fig. 4). The two scores correlated strongly (ρ = 0.6; P < 0.0001). The specificity to distinguish between active and sedentary subjects was also good according to ROC analysis (area under the curve 0.73; P < 0.0001).

This study provides new information about insulin action in fat cells of sedentary subjects. We found that subjects with a self-reported sedentary lifestyle had impaired insulin inhibition of lipolysis. Insulin-induced effects on glucose metabolism in fat cells (i.e., triglyceride formation from glucose through lipogenesis) is also decreased. However, the latter depends on other common factors that may influence insulin action. In other words, Adipo-IR among sedentary subjects only directly involved antilipolysis.

We confirm that sedentary people are resistant to the overall action of insulin on glucose utilization (2,3). This notion is supported by findings with HOMA-IR in all subjects and with glucose disposal during a hyperinsulinemic-euglycemic clamp in a subgroup. Of greater interest is the novel observation that the in vivo overall antilipolytic effect of insulin on fatty acid suppression is impaired in an independent way in sedentary subjects, as shown by Adipo-IR values.

Insulin action on the cellular level can be pharmacologically distinguished into early receptor near events, as reflected by pD₂ values, and distal signal events, as reflected by the maximum effect of insulin (responsiveness). We observed that only the receptor near events (i.e., insulin sensitivity) were associated with a sedentary lifestyle in an important and independent way. Although a weak significant association between a sedentary lifestyle and the maximum antilipolytic effect of the hormone was observed, it became statistically insignificant after correction for confounders. The average differences in pD₂ for antilipolysis between sedentary and active subjects was ˜1 log₁₀ unit, which corresponded to an ˜10 times lower hormone concentration in active compared with sedentary subjects.

Although pD₂ for insulin-induced lipogenesis differed between the sedentary and active groups in univariable analysis, this difference became statistically insignificant in a comprehensive multiple regression model. One reason for the discrepancy could be that fat cells contribute little to glucose uptake during hyperinsulinemia (6,7). Another explanation could be that antilipolysis is the most sensitive of the various metabolic actions of insulin (9,10); this is presently confirmed. The pD₂ values for lipogenesis in sedentary and active subjects were, on average, 1 log₁₀ unit higher than the pD₂ values for the antilipolytic action. We speculate that antilipolysis might be more sensitive to variations in physical activity than effects of insulin on glucose metabolism in fat cells, as it occurs at lower hormone levels. These conclusions are based on multivariable analysis, and we acknowledge that it is not possible to say anything definitive if independency implies causality. With this kind of analysis, factors such as stability and precision of a measurement may influence the association beyond the true biological relationship.

A sedentary lifestyle was independently associated with the overall insulin effect on glucose and fatty acid metabolism according to our Adipo-IR and HOMA-IR results, and the two in vivo measures correlated with each other. This finding confirms a previous in vivo study comparing adipose tissue lipolytic activity and skeletal muscle glucose disposal during hyperinsulinemic-euglycemic clamp (32). Our results also suggest that physical activity status has multiple effects on insulin action in several insulin target tissues.

Obesity has a profound influence on insulin action (1). However, our results are not influenced by body weight status in an important way. The findings held true after correction for BMI in multiple regression, and similar results were obtained when obese and nonobese subjects were analyzed separately.

Our in vivo data suggest that an impact of physical activity on insulin action has a physiological meaning. Thus, during hyperinsulinemia, circulating fatty acid deceased to ˜50% lower concentrations in active versus sedentary subjects despite higher circulating insulin levels among the sedentary subjects and comparable fatty acid levels before insulin infusion. However, glycerol levels decreased in a similar fashion during hyperinsulinemia in both groups. As discussed in detail above and previously (26), circulating glycerol is mainly influenced by lipolysis and liver metabolism, whereas additional factors affect circulating fatty acid concentrations. Taken together, the data herein suggest that lipolysis during the hyperinsulinemic-euglycemic clamp was maximally suppressed. Therefore, the impaired fatty acid suppression in the sedentary group is likely to involve additional factors, such as differences in fatty acid clearance.

Sedentary subjects showed many differences from physically active subjects with regard to cardiometabolic profile. BMI, waist-to-hip ratio, fat cell volume, and serum insulin levels were higher, and the lipid profile was less favorable despite sedentary subjects being somewhat younger on average than active subjects. Unexpectedly, the occurrence of hyperlipidemic disease or T2DM was less common among sedentary subjects. The reason for this finding is unclear but might be a random event because the number of subjects with these disorders was relatively small (˜10% of the total study group). Furthermore, the study was not designed to investigate the relationship between physical fitness and clinical features. The observation that, on average, sedentary subjects were younger and more prone to obesity than physically active subjects could directly affect the clinical profiles. We also found that a sedentary lifestyle was more common in women than in men, which is in line with several studies suggesting that men are more physically active than women (33). In other words, we cannot exclude that a selection bias may explain some of the clinical differences between sedentary and active subjects.

There are some caveats with the current study. It is not population based. Such an investigation had to be initiated when this study started ˜30 years ago. It would not be representative for more recently included subjects because of changes in physical activity and other factors influencing insulin sensitivity occurring in recent years. We investigated both men and women. In theory, sex may have an impact on the level of insulin resistance in sedentary or active people. On the other hand, results were not influenced in an important way by sex in multiple regression and were similar in the large female group as in men and women together. We used an in-house–developed questionnaire of self-reported physical activity and acknowledge that self-reported physical activity does not always reflect physical activity measured with, for example, an accelerometer. Although more thorough and validated methods have been developed recently, we could not, for scientific reasons, make continuous changes in how to measure physical activity during the study. However, the results were essentially the same if we used physical activity as a dichotomous or continuous variable. Furthermore, our physical activity score correlated strongly with validated scores in a subgroup and was particularly specific in identifying sedentary people. We only examined abdominal subcutaneous adipose tissue in vitro. In theory, other adipose depots (gluteal, femoral, visceral, renal) might be differently influenced by physical activity. Unfortunately, it is not practical or ethical to perform a multisite adipose investigation using the current study protocol. Many depots can only be investigated in connection with general surgery. Such procedures induce rapid insulin resistance in fat cells (34), which may influence findings. Finally, we cannot say which molecular events in insulin signaling are altered in fat cells of sedentary people. Unfortunately, the methods require far too much adipose tissue than can be removed by needle aspiration in an outpatient setting. On the other hand, some insight can be obtained from our retrospective analysis of the adipose expression of genes involved in early steps of insulin signaling. Thus, the expression of IR, IRS1, and IRS2 correlated strongly with physical activity score. These data support the findings with pD₂ for antilipolysis and lipogenesis, which share these early steps in hormone signaling.

With these caveats in mind, we propose the following model for how the metabolic actions of insulin in fat cells are altered in people with a sedentary lifestyle. The antilipolytic and lipogenic effects are decreased, but only antilipolysis is independent of other factors that may influence insulin action. At the cellular level, the insulin resistance of antilipolysis is primarily localized as a receptor near event in signal transduction (IR, IRS1, and IRS2) causing decreased antilipolytic sensitivity and, thereby, an elevated half maximal effective insulin concentration, but the maximum action is preserved. The impaired antilipolytic effect of insulin in sedentary people may lead to less efficient suppression of circulating fatty acids in sedentary people, and elevated fatty acid concentrations confer insulin resistance and an increased risk of T2DM (8,35,36). The positive effect of physical training on the action of prolipolytic hormones in humans is well established (37,38). Recent studies on Adipo-IR have suggested that antilipolysis may also be improved by changing from a sedentary lifestyle to a more active one (39).

In summary, subjects with a sedentary lifestyle had markedly decreased sensitivity, but no maximum effect, of the antilipolytic action of insulin in vitro, which probably involved early steps in hormone signaling. This is independent of other factors influencing insulin action and may influence the ability to lower circulating fatty acid levels following hyperinsulinemia. However, the effect of insulin on fat cell glucose metabolism (i.e., lipogenesis) is not independently influenced by a sedentary lifestyle.

Acknowledgments. The authors acknowledge the excellent assistance of Eva Sjölin, Kerstin Wåhlén, Elisabeth Dungner, and Ana Maria Suzuki (Department of Medicine, Karolinska Institutet at Karolinska Hospital–Huddinge) for the fat cell studies and Katarina Hertel, Britt-Marie Leijonhufvud, and Yvonne Widlund (Department of Medicine, Karolinska Institutet at Karolinska Hospital–Huddinge) for the examination of the subjects. The authors also express gratitude to Prof. Tomas Jernberg (Department of Clinical Sciences, Danderyd Hospital, Karolinska Institutet) for data on physical activity score from the SCAPIS study.

Funding. The study was supported by grants from the Swedish Research Council, Novo Nordisk Foundation, the Strategic Research Programme in Diabetes at Karolinska Institutet, the Center for Innovative Medicine at Karolinska Institutet, and Stockholm County.

Duality of Interest. No potential conflicts of interest relevant to this study were reported.

Author Contributions. D.P.A., A.G.K., and P.A. performed the statistical analyses. D.P.A., I.D., M.R., and P.A. participated in recruitment and examination of the subjects and/or collection of data. A.G.K., I.D., and P.A. performed array analyses. P.A. designed the study and wrote the first version of the manuscript. All authors were involved in the subsequent writing of the manuscript and approved the final version. P.A. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.