ORIGINAL ARTICLE Human visceral adipose tissue and the ...

International Journal of Obesity (2007) 31, 1671–1679 & 2007 Nature Publishing Group All rights reserved 0307-0565/07 $30.00 www.nature.com/ijo

ORIGINAL ARTICLE Human visceral adipose tissue and the plasminogen activator inhibitor type 1 JHN Lindeman1, H Pijl2, K Toet3, PHC Eilers4, B van Ramshorst5, MM Buijs2, JH van Bockel1, T Kooistra3 1 Department of Vascular Surgery, Leiden University Medical Center, Leiden, The Netherlands; 2Department of Endocrinology, Leiden University Medical Center, Leiden, The Netherlands; 3TNO-Biomedical Research, Leiden, The Netherlands; 4Department of Medical Statistics, Leiden University Medical Center, Leiden, The Netherlands and 5Sint Antonius Ziekenhuis, Nieuwegein, The Netherlands

Objective: The objective of this study was to systematically evaluate the molecular basis of the association between visceral fat mass and plasma plasminogen activator inhibitor-1 (PAI-1) levels in man. Design: A comprehensive approach comprising observational, in vitro, and human intervention studies. Measurements and results: We confirmed an exclusive relationship between visceral fat and plasma PAI-1 levels (r ¼ 0.79, Po0.001) and corroborated preferential PAI-1 release from adipose tissue explants. Yet, messenger RNA analysis and in vivo measurement of PAI-1 release from visceral fat (AV-differences over the omentum) not only excluded visceral adipose tissue as a relevant source of circulating PAI-1, but also excluded visceral fat as a significant source of proinflammatory mediators such as tumor necrosis factor-a, IL-1 or transforming growth factor-b that could induce PAI-1 expression in tissues other than visceral fat. Short-term interventions with acipimox and growth hormone (GH) as well as statistical evaluation excluded free fatty acids and GH as metabolic links. Further analysis of the metabolic data in a stepwise regression model indicated that plasma PAI-1 levels and visceral fat rather are co-correlates that both relate to impaired lipid handling. Conclusion: Our PAI-1 studies show that visceral fat mass and plasma PAI-1 levels are co-correlated rather than causatively related, with lipid load as common denominator. International Journal of Obesity (2007) 31, 1671–1679; doi:10.1038/sj.ijo.0803650; published online 1 May 2007 Keywords: visceral fat; PAI-1; inflammation; TNFa

Introduction Plasma levels of the plasminogen activator inhibitor-1 (PAI-1) correlate with body mass; and obesity, in particular visceral obesity, is associated with elevated PAI-1 plasma levels.1,2 Increased PAI-1 levels are responsible for the impaired fibrinolysis that accompanies obesity and presumably contribute to increased cardiovascular risk in overweight and obese individuals.3 Moreover, the apparent association between PAI-1 plasma levels, and insulin resistance4 and incident type 2 diabetes mellitus5 characterizes PAI-1 as a biological marker of the metabolic risk that accompanies (visceral) obesity.

Correspondence: Dr JHN Lindeman, Department of Vascular Surgery, K6R, Leiden University Medical Center, PO Box 9600, 2300 RC Leiden, The Netherlands. E-mail: [email protected] Received 5 November 2006; revised 7 March 2007; accepted 29 March 2007; published online 1 May 2007

Despite a large body of epidemiological and experimental evidence in support of a direct link between (visceral) adipose tissue and plasma PAI-1 levels,1,2,6 clinical evidence is less obvious and the molecular basis for the association between (visceral) adipose tissue and plasma PAI-1 levels in man is still unclear.7 The most obvious explanation for the association between plasma PAI-1 levels and visceral adipose tissue is a direct link in which the elevated plasma PAI-1 levels in (visceral) obesity result from PAI-1 release from (visceral) fat mass.6 Support for such a direct association largely stems from animal and in vitro data,6 but evidence for such a mechanism in humans is still missing. A second, more indirect, explanation for the relationship is release of an inflammatory ‘mediator’, such as tumor necrosis factor (TNF)a or IL-1,8 from visceral fat, which induces PAI-1 production outside visceral fat. Then again, evidence is largely derived from animal studies and human data in support of such a mechanism is inconclusive.9 A third explanation for the association between PAI-1 levels and visceral obesity would be that the association reflects a

Human visceral adipose tissue and PAI-1 JHN Lindeman et al

1672 common cause outside the visceral fat compartment that underlies both accumulation of visceral adipose tissue as well as the elevated PAI-1 levels. Such a mechanism could involve metabolic factors such as increased circulating free fatty acids (FFAs) or triglycerides.10 Thus, in spite of the important role of PAI-1 in the metabolic risk associated with (visceral) obesity, the source and mechanisms of increased PAI-1 in obesity are incompletely understood. In this paper, we systematically evaluated the three possible explanations for the association between (visceral) adipose tissue and plasma PAI-1 levels in man.

Methods All clinical studies were approved by the Medical Ethical Committee of the Leiden University Medical Center.

Patients The following substudies were used to explore the putative association between visceral adipose tissue and plasma PAI-1 levels: Associations between plasma PAI-1 levels and anthropometric measures of adiposity and fat distribution as well as between plasma PAI-1 levels and the various components of the metabolic syndrome were explored in a well characterized group of premenopausal normal-weight (n ¼ 8, body mass index (BMI) 22.472.1 kg/m2; age 37.677.8 years (mean7s.d.)) and moderately obese (n ¼ 15, BMI 33.973.6 kg/m2, age 35.176.0 years (mean7s.d.)) women (this group is further referred to as the reference group). Subjects were recruited through advertisements in local media. The moderately obese women were carefully selected to cover a wide range of waist hip ratios (WHRs). Subcutaneous and visceral adipose tissue for studies on PAI-1 messenger RNA (mRNA) expression and in vitro studies used adipose tissue explants from male and female donors. Ad hoc analysis did not indicate a sex difference. For this reason only combined data is presented. Tissue was obtained during elective aneurysm repair (normal weight and overweight patients (BMI 20–32 kg/m2) or during gastric banding for morbid obesity (severely obese individuals (BMI440 kg/m2). Except for their obesity or their abdominal aortic aneurysm, all patients were healthy according to medical history, clinical examination and routine laboratory findings. Measurement of arteriovenous concentration differences over the omentum for in vivo assessment of PAI-1 and cytokine release from visceral adipose tissue was performed in 15 male and female patients (BMI 20–32 kg/m2) undergoing elective aortic repair for their abdominal aneurysm. Ad hoc analysis did not indicate a sex difference. For this reason only combined data is presented. Except for their abdominal aortic aneurysm, all patients were healthy according to medical history, clinical examination and International Journal of Obesity

routine laboratory findings. All sampling was performed between 0800 and 0900 hours.11 A possible association between increased plasma FFA availability and elevated plasma PAI-1 levels was studied in 11 healthy, obese (BMI 35.872.8 (mean7s.d.) kg/m2) premenopausal women (age 33.873.7 (mean7s.d.) years) who had been carefully selected to cover a wide range of WHRs. Women were randomly assigned to 250 mg acipimox or placebo in a double-blind crossover design by an independent investigator. The study was performed on two separate occasions in the early follicular state of the menstrual cycle with an interval of at least 8 weeks. Drug or placebo was taken four times daily (0100 and 0700 and 1300 and 1900 hours, respectively), starting 36 h before the first blood sampling. Eucaloric standard meals were prescribed during the study periods. To test for a direct relationship between the hyposomatropism in visceral obesity and elevated PAI-1 levels, we evaluated the effect of a single infusion human growth hormone (GH) (12 mU/kg) or placebo infusion on plasma PAI-1 levels in six healthy, premenopausal normal-weight (BMI 21.372.3 (mean7s.d.) kg/m2) and six obese (BMI 35.471.5 (mean7s.d.) kg/m2) premenopausal women (age 35.0711.4 (mean7s.d.) years). Women were randomly assigned to GH or placebo in a double-blind crossover design by an independent investigator. The study was performed on two separate occasions in the early follicular state of the menstrual cycle with an interval of at least 8 weeks. Infusion of GH was started at 0800 hours (t ¼ 0) and blood samples for PAI-1 analysis were obtained at t ¼ 0, 1 and 7 h (1500 hours). To limit confounding by nutritional factors, eucaloric standard meals were prescribed during the study periods.

Anthropometrics, fat distribution All anthropometric measurements were made with the subjects wearing underwear. Body weight was rounded to the nearest 0.1 kg and body height measured to the nearest 0.1 cm using a wall-mounted stadiometer. Fat distribution was assessed by magnetic resonance imaging. For the visceral fat mass, transverse abdominal scans were made at three levels with each scan 10 mm thick and a gap of 2 mm between levels. The highest scan was at the level of the cranial intervertebral facies of the forth lumbar vertebra. Similar scans were made at the level of the hip with the upper scan taken at the superior margin of the major trochanter. Images were analyzed on a SUN workstation and the visceral, abdominal subcutaneous, femoral subcutaneous and total (sum abdominal and subcutaneous fat area) subcutaneous adipose tissue areas were calculated. Blood and adipose tissue sampling For the adipose tissue sampling 2–5 g visceral and (abdominal) subcutaneous adipose tissue was obtained immediately after exposure of the visceral adipose tissue. Samples were snapfrozen in liquid N2 and stored at 701C until further analysis.


1673 All blood sampling was performed at 0800 hours after an overnight fast. Blood samples were immediately put on melting ice and centrifuged within 1 h of sampling. Plasma was stored at 701C until analysis. Blood sampling in the reference group was performed under highly standardized testing conditions (such as dietary lead in, standardized iso-caloric diet, follicular phase of the menstrual cycle).12 Assessment of arteriovenous concentration differences over the omentum, the largest visceral fat depot, was performed immediately on exposure of the omentum during elective transabdominal aneurysm repair. A 21-G winged needle infusion set (Venisystes, Abbott, Hoofddorp, The Netherlands) was retrogradely inserted in one of the principal veins draining the omentum. Retrograde blood flow during aspiration was prevented by manual compression of the distal part of the vein. Arterial blood samples were simultaneously drawn from the intra-arterial line, routinely inserted during aneurysm repair. Blood was gently drawn in a sterile 5 ml Sarstedt Monovette CTAD container and directly placed on ice. Blood samples were centrifuged at 41C (10 min, 1200 g), and the extracted plasma recentrifuged (10 min, 10 000 g) and stored at 701C until further analysis. Sequential blood sampling for integrated 24-h GH levels and FFA rhythm was performed as follows: a 20-gauge cannula was placed in an antecubital vein. Starting 1 h after insertion of the cannula (0700 hours), blood was drawn though a constant withdrawal pump (Conflo, Carmeda AB, Taeby, Sweden) into heparinized reservoir tubes. Tubes were changed every 10 min over a 24-h period. All samples were centrifuged within the hour of sampling and plasma stored at 701C. Plasma GH and FFA levels were measured every 10 and 20 min, respectively.

Adipose tissue explants Adipose tissue from subcutaneous area (from the abdominal wall) and visceral area (omentum) was obtained immediately after exposure of the omentum. Tissue was directly placed in ice-cold RPMI medium and processed within the hour of sampling. Adipose tissue was rinsed once with cold RPMI and chopped into small pieces (1 mm3) with a McIlwain tissue chopper (Mickle laboratory Engineering Co., Gromshell, UK), and subsequently incubated (100 mg tissue/ml) in RPMI–N-2-hydroxyethylpiperazine-N0 -2-ethanesulfonic, supplemented with 1% (w/v) human serum albumin (CLB), glutamine, penicillin and streptomycin. Incubations were carried out for 19 h at 371C in a 95% O2 and 5% CO2 atmosphere. At fixed time points, medium samples were drawn and replaced by an equal amount of fresh medium. Samples were immediately centrifuged and stored at 701C. At the end of the incubation, supernatants were collected, centrifuged to remove the debris and snap-frozen in liquid nitrogen. Adipose tissue explants were washed once with ice-cold phosphate-buffered saline, superficially dried and snap-frozen in liquid nitrogen.

Serum and plasma measurements Serum total cholesterol and triglycerides were measured with a fully automated Hitachi 704 system and highdensity lipoprotein (HDL) cholesterol was determined on a Hitachi 911 (Boehringer Mannheim, Almere, The Netherlands). Low-density lipoprotein (LDL) cholesterol was calculated with the Friedewald formula.13 GH concentrations were measured with a sensitive time-resolved fluoroimmunoassay (Wallac, Turku, Finland), specific for the 22-kDa GH protein, and plasma FFA levels were determined by the NEFA-C FFA kit (Wako Chemicals, Neuss, Germany). Plasma IL-1a, IL-6, TNFa, transforming growth factor-b (TGFb), soluble intercellular adhesion molecule-1 (sICAM-1) and tissue plasminogen activator (tPA) were assessed by commercially available ELISA kits: PeliKane compact kit (CLB, Amsterdam, the Netherlands) for IL-6, HS Quantikine (R&D Systems, Abingdon, UK) for IL-1a, TNFa, TGFb and ICAM-1 and Imulyse (Biopool, Umea˚, Sweden) for tPA. PAI-1 antigen was measured by a specific in-house sandwich ELISA.12

Statistics Values are given as mean7s.d. or median and range when indicated. All calculations were performed using the SPSS 11.5 statistical package. GH and HDL cholesterol values were normalized by using inverse (i.e., 1/GH or 1/HDL) concentrations. Differences in results were determined by a paired t-test or Wilcoxon rank-order test for paired observations when applicable. Correlation, multiple regression and partial correlation analysis were used to assess the degree of association between various anthropometric indices and PAI-1 plasma levels. P values less than 0.05 were considered significant.

Real time competitive lightcycler PCR for PAI-1, TNFa and IL-6 Real time PCR was performed according to the Taqman method of Applied Biosystems (Nieuwerkerk aan de IJssel, The Netherlands), using a combination of a forward and a backward primer and a specific (6-carboxy-fluorescein/ 6-carboxy-tetramethyl–rhodamine) double-labeled probe. Specific amplicons were chosen according to the manufacturer’s recommendations; mRNA was expressed relative to glyceraldehyde phosphate dehydrogenase (GAPDH) levels. Probes and primers were obtained from Isogen (Maarsen, The Netherlands) and the VIC-labeled GAPDH primer-probe combination was purchased from Applied Biosystems.

Results PAI-1 plasma levels exclusively relate to visceral adipose tissue but not any other anthropometric measure of adiposity of body-fat distribution The relationship between PAI-1 plasma levels and the anthropometric measures of adiposity and fat distribution International Journal of Obesity


1674 was evaluated in a group of 23 normal-weight to moderately obese women (‘reference group’), who were carefully selected to cover a broad spectrum of body-fat distribution (Table 1). This evaluation confirmed reported associations between plasma PAI-1 and measures of visceral fat mass (visceral fat area, WHR (Table 2)) and indicated a borderline significant relationship between plasma PAI-1 and measures of overall adiposity (BMI, % fat mass, Table 2). No relationship was found between plasma PAI-1 levels and the different measures of subcutaneous fat mass (Table 2). Further analysis of relationship between PAI-1 plasma levels and the different measures of fat distribution and adiposity in a multiple regression model incorporating BMI, and visceral and subcutaneous fat mass, showed that plasma PAI-1 levels exclusively relate to the visceral fat area.

Visceral adipose tissue is not a relevant source of plasma PAI-1 Prominent PAI-1 release from adipose tissue explants has frequently been used to demonstrate PAI-1 production by adipose tissue. Our incubation experiments confirmed PAI-1 production from adipose tissue explants and the superior production by visceral adipose tissue explants. Yet, we found indications that PAI-1 release from adipose tissue explants relates to an in vitro artifact. First, over the wide range of BMIs tested, PAI-1 mRNA expression was found to be similar and low in native, snap-frozen adipose tissue (number of normalized amplification cycles (Ct values): 28.1 (0.9) and 27.9 (2.0) (mean (s.d.)) in subcutaneous and visceral adipose tissue, respectively), and contrary to expectations we observed an inverse relationship between adipose tissue PAI-1 expression and BMI (r ¼ –0.85 (Po0.01) and 0.49 Table 1 Characteristics of the reference group (normal weight and moderately obese premenopausal women) (n ¼ 23)

Age (years) Weight (kg) BMI (kg/m2) Waist-hip ratio Total fat mass (kg) Percentage fat mass Total subcutaneous fat area (cm2) Total visceral fat area (cm2)

Mean

s.e.m.

Range

37.0 84.6 29.9 0.83 33.6 38.2 2244 348

1.49 3.64 1.3 0.02 2.7 1.7 188 49

23–50 52.0–120.8 19.6–40.4 0.72–1.01 14.4–60.5 24.2–51.4 722–3874 97–863

(P ¼ 0.08) for visceral and subcutaneous adipose tissue, respectively). Second, significant PAI-1 release from subcutaneous adipose tissue explants apparently contrasts with in vivo data of Yudkin et al.14 showing that human subcutaneous adipose tissue does not release PAI-1. Third, PAI-1 production by adipose tissue explants is preceded by several hour lag-phase with minimal PAI-1 protein release that is followed by an exponential increase in PAI-1 protein production and a concomitant several 100-fold increase in PAI-1 mRNA expression (data not shown). Owing to these controversial findings, it was decided to validate the in vitro findings through evaluation of in vivo PAI-1 release from visceral adipose tissue (arteriovenous concentration differences over the omentum, the principal visceral fat compartment). Although FFA, leptin and IL-6 release from visceral adipose tissue (Table 3) confirmed the validity of the method, we found no evidence for net PAI-1 release from visceral adipose tissue (Table 3), indicated by similar and highly correlated (r ¼ 0.95, Po0.001) PAI-1 levels in arterial and postomental venous blood. Hence, our findings exclude visceral adipose tissue as a relevant direct source of circulating PAI-1.

Visceral adipose tissue is not a relevant source of proinflammatory cytokines The role of (visceral) adipose tissue as an ample source of proinflammatory cytokines, in particular TNFa, is widely acknowledged. PAI-1 gene expression is strongly stimulated by a number of proinflammatory cytokines including TNFa. Hence, production and release of such a proinflammatory factor from visceral adipose tissue and the subsequent induction of PAI-1 outside the fat compartment could underlie the association between visceral adipose tissue and plasma PAI-1. We evaluated TNFa mRNA expression in native visceral and subcutaneous adipose tissue and found low and similar TNFa mRNA expression in both fat compartments over the broad range of BMIs tested (data not shown). These findings seemingly conflict with the concept of (visceral) adipose tissue as a prominent source of circulating TNFa. We therefore analyzed TNFa release from visceral adipose tissue Table 3 Arteriovenous (AV) concentration differences over the omentum (n ¼ 15, mean7(s.d.) or median (range))

BMI, body mass index.

Table 2 Correlations between plasma PAI-1 levels and the different measures of adiposity and fat distribution in the reference group (n ¼ 23)

2

Visceral fat area (cm ) Waist hip ratio BMI (kg/m2) Percentage fat mass Total subcutaneous fat area (cm2) Abdominal subcutaneous fat area (cm2)

r

P

0.68 0.65 0.40 0.47 0.32 0.38

0.0001 0.001 0.06 0.02 0.14 0.07

BMI, body mass index; PAI-1, plasminogen activator inhibitor-1.

International Journal of Obesity

PAI-1(mg/l) TNFa (ng/l) IL-1 (ng/l) IL-6 (ng/l) TGFb (ng/l) Leptin (ng/l) FFA (mmol/l) ICAM (mg/l)

Arterial

Venous

20.2 (4.3–108.8) 2.02 (0.77–4.80) 0.12 (0.00–2.30) 4.18 (1.8) 1.1 (0.41–21.4) 3.8 (0.8–23.7) 0.58 (0.14) 159 (39)

20.4 (8.3–90.3) 1.94 (0.55–5.11) 0.26 (0.00–2.92) 4.72 (1.70)* 0.92 (0.22–25.10) 5.2 (0.8–23.7)* 1.08 (0.35)* 160 (40)

AV difference –2.70 0.023 0.14 0.31 –0.25 0.69 0.45 –2

(2.03) (0.069) (0.21) (0.23) (0.40) (0.34) (0.06) (4)

FFA, free fatty acid; ICAM, intercellular adhesion molecule; PAI-1, plasminogen activator inhibitor-1; TGF, transforming growth factor; TNF, tumor necrosis factor. *Concentrations in venous blood significantly higher than in arterial blood (Po0.005).


1675 in vivo and measured arteriovenous concentration differences over the omentum. As shown for PAI-1, similar and highly correlated (r ¼ 0.97, Po0.001) TNFa levels were found in paired arterial and postomental venous plasma (Table 3), and our findings thus exclude visceral adipose tissue as a relevant source of circulating TNFa. Analogous results were obtained for IL-1 and TGFb (Table 3), two other established inflammatory inducers of PAI-1, excluding visceral adipose tissue as a relevant source of circulating IL-1 and TGFb. These data, along with the absence of PAI-1 release and release of sICAM-1 (Table 3), a sensitive marker of proinflammatory activity, from visceral adipose tissue do not identify human visceral adipose tissue as a proinflammatory environment and as a relevant source of proinflammatory mediator(s) that could result in the induction of PAI-1 production outside the fat compartment.

PAI-1 plasma levels do not relate to the increased FFA availability or to the secondary GH deficiency that accompanies visceral obesity The previous studies exclude visceral adipose tissue as a relevant source of circulating PAI-1 or as a source of inflammatory mediators, able to induce PAI-1 outside the fat compartment. We therefore sought evidence for a metabolic link between visceral adipose tissue and PAI-1. Accumulation of visceral adipose tissue has been associated with several metabolic disturbances including increased FFA availability, GH deficiency and dyslipidemia (elevated LDL cholesterol and triglycerides along with decreased HDL cholesterol), which have all been associated with elevated plasma PAI-1 levels. We found net release of FFA from the visceral adipose tissue (Table 3) and confirmed increased FFA availability in women with upper body (visceral) obesity (Figure 1). To evaluate a possible role of FFA as metabolic link between visceral adipose tissue and plasma PAI-1, we lowered plasma FFA levels in obese volunteers through short-term treatment with the antilipolytic drug acipimox. Although 36 h acipimox treatment reduced mean 24 h plasma FFA levels (0.5270.12 versus 0.4070.09 mmol/l (mean7s.d.), before and after acipimox, respectively), it did not influence plasma PAI-1 plasma levels: 14.9 (9–87) versus 14.6 (6–61) mg/l (median (range), P ¼ NS). Abdominal obesity is characterized by secondary GH deficiency. Elevated PAI-1 levels are found in primary GH deficiency. We explored a possible association between integrated 24 h GH and plasma PAI-1 levels and found robust inverse relationships between 24 h GH concentration and visceral fat mass (0.73, Po0.001) and 24 h GH concentration and plasma PAI-1 levels (r ¼ –0.58, Po0.01). Further analyzes of the relationships by a multiple regression model retained integrated 24 h GH levels but excluded visceral fat as variant. Evidence was sought for a direct link between GH deficiency and elevated PAI-1 levels, and we tested whether normalization of plasma GH levels through a single, adequate, bolus GH directly influenced PAI-1 levels.

Figure 1 Representative 24 h free fatty acid (FFA) profiles (sampled every 20 min) of two normal-weight individuals (normal), two moderate obese individuals with a small visceral fat mass (lower body obesity, LBO) and two moderate obese individuals with a large visceral fat mass (upper body obesity, UBO). Meals are indicated by the vertical dashed lines.

Figure 2 shows that this intervention did not reduce plasma PAI-1 levels but, on the contrary, resulted in a moderate increase in plasma PAI-1 levels in the group of obese women. Hence, these experimental findings along with findings for the acipimox intervention15 exclude GH deficiency as the direct cause of elevated PAI-1 levels in individuals with visceral obesity.

PAI-1 plasma levels and visceral fat mass are covariants and relate to impaired lipid handling Dysregulation of lipid metabolism resulting in elevated triglycerides and LDL cholesterol as well as decreased HDL International Journal of Obesity


1676 is another prominent characteristic of visceral obesity and close relationships are found between visceral fat mass and these variables (Table 4). We analyzed possible correlations between plasma PAI-1 and these variables and found a strong positive relationship between PAI-1 levels and plasma triglycerides on the one hand and an inverse relationship between PAI-1 levels and plasma HDL cholesterol (Table 4) on the other. No relationship was found between plasma PAI-1 and total or LDL cholesterol. Incorporation of triglycerides, HDL cholesterol, along with integrated GH levels and visceral fat area as covariants for PAI-1 in a multiple regression model retained plasma triglycerides and HDL cholesterol as dependent variables, yet significance was lost for visceral fat area or integrated GH secretion. Further evaluation of the relationships between PAI-1 plasma levels and triglycerides and HDL through assessment of partial correlations for the relationships between plasma PAI-1 and triglycerides and HDL cholesterol (after correction for HDL or triglycerides, visa versa) showed that this did not influence the individual relationships between plasma PAI-1, triglycerides and HDL cholesterol. Indicating that the observed relationships between these factors reflect a common pathway that results in

elevated PAI-1 and triglyceride and reduced HDL cholesterol plasma levels.

Discussion In this study, we sought insight in the molecular basis for the association between visceral adipose tissue and circulating PAI-1. We corroborated the strong relationship between visceral adipose tissue and plasma PAI-1 levels and confirmed in vitro data showing increased PAI-1 production by visceral adipose tissue explants. However, low and similar PAI-1 mRNA expression in subcutaneous and visceral adipose tissue, along with absent arteriovenous PAI-1 concentration differences across visceral adipose fat in vivo, exclude visceral fat as a relevant direct source of circulating PAI-1. Consequently, the association between visceral obesity and elevated PAI-1 levels must be caused by other factors. Elevated plasma levels of the cardiovascular risk factor PAI-1 are part of the metabolic disturbances that accompany obesity.1,2 PAI-1 is the primary endogenous inhibitor of tissue- and urokinase-type plasminogen activators and the main determinant of fibrinolytic activity.3 Increased PAI-1 plasma levels in obesity, and, in particular, an abdominal type of obesity, are responsible for the impaired fibrinolysis that accompanies visceral obesity and presumably contribute to the increased cardiovascular risk in (abdominal) obesity. Over and above, epidemiological studies identified PAI-1 plasma levels as early and independent predictors of insulin resistance and incident type 2 diabetes,4,5 also characterizing PAI-1 as a biological marker for the metabolic risk profile that accompanies visceral obesity. Animal studies and in vitro data from human adipose tissue explants suggest a direct and relevant contribution of visceral adipose tissue to circulating PAI-1.6,7,16 Yet, the observation of similar and minor PAI-1 mRNA expression in native human visceral and subcutaneous adipose tissue7,17,18 is not consistent with such a mechanism. Moreover, we found strong indications that the observed PAI-1 release from adipose tissue explants on in vitro incubation relates to an incubation artifact and may not necessarily reflect the in vivo situation. Yudkin et al14 previously pointed out that assessment of arteriovenous concentration differences has sufficient sensitivity to detect a relevant contribution of adipose tissue to plasma PAI-1 levels and, subsequently, Table 4 Correlations (r) between plasma lipid parameters with plasma PAI-1 and the visceral fat area respectively (n ¼ 23) PAI-1

Visceral fat area

Integrated 24 h growth hormone levels

0.30 0.29 –0.58w 0.67w

0.44* 0.44* –0.56w 0.59w

0.12 0.17 –0.52 0.52

Figure 2 The effect of a single bolus human growth hormone (hGH)

Total cholesterol (mmol/l) LDL cholesterol (mmol/l) HDL cholesterol (mmol/l) Triglycerides (mmol/l)

(12 mU/kg) on plasma PAI-1 plasma levels in normal weight (below) and obese (above) individuals. Placebo: (white boxes), 12 mU hGH/kg: (gray boxes) at t ¼ 0. *Po0.01.

HDL, high-density lipoprotein; LDL, low-density lipoprotein; PAI-1, plasminogen activator inhibitor-1. *Po0.05, wPo0.005.



1677 showed that subcutaneous adipose tissue is not a relevant source of plasma PAI-1 in vivo.14 We evaluated putative PAI-1 release from visceral adipose in vivo and measured arteriovenous concentration differences across the omentum, the principal visceral fat depot, and found similar arterial and postomental venous blood PAI-1 concentrations. Hence, our findings, along with that of Yudkin et al.,14 for subcutaneous adipose tissue exclude adipose tissue as a relevant, direct source of circulating PAI-1. On exclusion of a direct link between visceral adipose tissue and plasma PAI-1 levels, evidence was sought for an indirect link between adipose tissue and plasma PAI-1. Adipose tissue is considered to be an ample source of circulating inflammatory mediators,8,19 and release of proinflammatory mediators, in particular TNFa, from adipose tissue has been proposed as an indirect link between visceral obesity and plasma PAI-1 levels.20 However, such a concept is challenged by available human data, indicating negligible TNFa mRNA expression in human visceral and subcutaneous adipose tissue.19 We addressed this apparent controversy through direct evaluation TNFa release from visceral adipose tissue in vivo (arteriovenous concentration differences). Although observed net release of IL-6, FFA and leptin validated the approach, no evidence was found for release of TNFa, nor for IL-1 or TGFb, two other potent inducers of PAI-1 expression.21 These findings extend data from subcutaneous adipose tissue22 and exclude human adipose tissue as a relevant source of circulating TNFa, IL-1 or TGFb. We can, however, not exclude that other proinflammatory mediators (such as angiotensin)23 are released from visceral adipose tissue and result in the induction of PAI-1 release outside the fat compartment. Considering the vast and rapid induction of PAI-1 release from adipose tissue explants after exposure to proinflammatory stimuli (including angiotensin, Lindeman et al., unpublished data) and lack of PAI-1 release from adipose tissue in vivo, we consider such a mechanism unlikely. We extended these findings and evaluated the release of sICAM-1 from visceral adipose tissue. ICAM-1 shedding is a sensitive marker of proinflammatory activity24 and lack of sICAM-1 release from visceral adipose tissue further argues against characterization of visceral fat as a proinflammatory environment. Secondary GH deficiency,25 dysregulated lipid metabolism, in particular increased FFA availability,26 elevated triglycerides and low plasma HDL cholesterol27 as well as insulin resistance accompanied by high insulin levels are all central features of the metabolic disturbances that accompany (abdominal) obesity. These disturbances have all been associated with elevated PAI-1 plasma levels27–29 and may therefore well explain the association between (visceral) adiposity and plasma PAI-1.4,26 Hyperinsulinemia is among the most prominent metabolic disturbances accompanying accumulation of visceral adipose tissue. Although insulin has been shown to induce PAI-1 expression in vitro, experimental in vivo data is less

conclusive30 and clinical studies by Scelles et al.31 clearly show that elevated PAI-1 levels are a unique feature of insulin resistance seen in obesity and are not observed in other forms of chronic insulin resistance. Moreover, epidemiological data from the prospective insulin resistance atherosclerosis study4 as well as the diabetes offspring study30 clearly show that increases in PAI-1 plasma levels precede development of insulin resistance in man and thus further argue against a causative relationship between elevated insulin levels and plasma PAI-1 levels. We used an experimental approach to test whether elevated PAI-1 levels in visceral obesity relate to increased FFA availability or GH deficiency. Short-term treatment with antilipolytic drug acipimox15 has been shown to reduce FFA plasma levels and increase GH levels.15 Yet, despite reduced plasma FFA levels, acipimox treatment did not influence plasma PAI-1 levels. Hence, these findings, along with the data from the short-term GH supplementation study, exclude increased FFA availability or the hyposomatropism of obesity as the direct cause of elevated PAI-1 levels in visceral obesity. Most of the individual components of the dyslipidemia accompanying visceral obesity have been associated with elevated PAI-1 levels.32 Statistical analysis of the relationship between the individual components of the dyslipidemia and PAI-1 plasma showed strong positive relation between circulating PAI-1 levels and triglycerides on one hand and a negative relationship between HDL and plasma PAI-1 on the other. Incorporation of HDL cholesterol, triglycerides and visceral fat in a multiple regression analysis with PAI-1 as covariant identified triglycerides and HDL cholesterol as influential variables but excluded visceral fat mass as significant variable. This latter finding corroborates the experimental findings of this study and characterizes the association between visceral fat mass and plasma PAI-1 as indirect rather than causative. A critical question is whether the observed associations between triglycerides and HDL cholesterol, and plasma PAI-1 levels are causative (i.e., elevated triglycerides (very lowdensity lipoprotein)33 and/or decreased HDL cholesterol are directly responsible for the elevated PAI-1 plasma levels) or, alternatively, whether the relationships reflect a common, more upstream, factor that is responsible for increased PAI-1 and triglycerides as well as decreased HDL cholesterol. Studies from fibrate-treated individuals show that fibrate treatment decreases triglyceride levels and increases HDL levels, yet these changes are not paralleled by reduced PAI-1 levels.34,35 Hence, these studies are thus not consistent with a causative relationship between triglycerides/HDL and PAI-1. This notion is further supported by partial regression analysis performed on the triglyceride/HDL PAI-1 data in this study that characterizes the respective associations between triglycerides and HDL cholesterol and plasma PAI-1 as independent, indicating that the apparent relationship between plasma PAI-1 and these variables reflect a common upstream variable. International Journal of Obesity


1678 In conclusion, our data exclude a direct association between visceral adipose tissue and plasma PAI-1 levels and confirm epidemiological data from Hanley et al.,4 showing that the relationship between visceral adipose tissue and plasma PAI-1 relates to a metabolic factor rather than an inflammatory factor. We speculated that this association might reflect PAI-1 production within the visceral fat compartment. Our data is not consistent with such a scenario, but rather suggests that the apparent relationship between plasma PAI-1 and visceral fat reflects covariation.36 Alessi et al.37 showed that elevated PAI-1 levels in obesity are more closely related to the extent of hepatic steatosis (nonalcoholic steatohepatitis) than to adipose tissue. Intrahepatic fat accumulation is strongly associated with visceral fat mass38 and is associated with the features of the metabolic syndrome (including insulin resistance, elevated plasma PAI-1 and triglyceride levels and reduced HDL cholesterol), suggesting that intrahepatic fat accumulation itself, or the mechanism underlying hepatic fat accumulation in obesity, is the unknown upstream variable that is responsible for the association between PAI-1 and visceral adipose tissue.

Acknowledgements We thank Professor Hans Romijn for his helpful comments during the preparation of the paper. This work was supported by The Netherlands Heart Foundation (NHS 1997-100).

References 1 Alessi MC, Juhan-Vague I. PAI-1 and the metabolic syndrome. Links, causes, and consequences. Arterioscler Thromb Vasc Biol 2006; 26: 2200–2207. 2 Skurk T, Hauner H. Obesity and impaired fibrinolysis: role of adipose production of plasminogen activator inhibitor-1. Int J Obes Relat Metab Disord 2004; 28: 1357–1364. 3 Kohler HP, Grant PJ. Plasminogen-activator inhibitor type 1 and coronary artery disease. N Engl J Med 2000; 342: 1792–1801. 4 Hanley AJ, Festa A, D’Agostino RB, Wagenknecht LE, Savage PJ, Tracy RP et al. Metabolic and inflammation variable clusters and prediction of type 2 diabetes: factor analysis using directly measured insulin sensitivity. Diabetes 2004; 53: 1773–1781. 5 Festa A, D’Agostino Jr R, Tracy RP, Haffner SM. Insulin resistance atherosclerosis study. Elevated levels of acute-phase proteins and plasminogen activator inhibitor-1 predict the development of type 2 diabetes: the insulin resistance atherosclerosis study. Diabetes 2002; 51: 1131–1137. 6 Shimomura I, Funahashi T, Takahashi M, Maeda K, Kotani K, Nakamura T et al. Enhanced expression of PAI-1 in visceral fat: possible contributor to vascular disease in obesity. Nat Med 1996; 2: 800–803. 7 Bastard JP, Pieroni L. Plasma plasminogen activator inhibitor 1, insulin resistance and android obesity. Biomed Pharmacother 1999; 53: 455–461. 8 Zoccali C, Mallamaci F, Tripepi G. Adipose tissue as a source of inflammatory cytokines in health and disease: focus on end-stage renal disease. Kidney Int Suppl 2003; 84: S65–S68.


9 Moller DE. Potential role of TNF-alpha in the pathogenesis of insulin resistance and type 2 diabetes. Trends Endocrinol Metab 2000; 11: 212–217. 10 Eckel RH, Grundy SM, Zimmet PZ. The metabolic syndrome. Lancet 2005; 365: 1415–1428. 11 Kluft C, Jie AF, Rijken DC, Verheijen JH. Daytime fluctuations in blood of tissue-type plasminogen activator (t-PA) and its fastacting inhibitor (PAI-1). Thromb Haemost 1988; 59: 329–332. 12 Langendonk JG, Pijl H, Toornvliet AC, Burggraaf J, Frolich M, Schoemaker RC et al. Circadian rhythm of plasma leptin levels in upper and lower body obese women: influence of body fat distribution and weight loss. J Clin Endocrinol Metab 1998; 83: 1706–1712. 13 Hukshorn CJ, Lindeman JH, Toet KH, Saris WH, Eilers PH, Westerterp-Plantenga MS et al. Leptin and the proinflammatory state associated with human obesity. J Clin Endocrinol Metab 2004; 89: 1773–1778. 14 Yudkin JS, Coppack SW, Bulmer K, Rawesh A, Mohamed-Ali V. Lack of evidence for secretion of plasminogen activator inhibitor1 by human subcutaneous adipose tissue in vivo. Thromb Res 1999; 96: 1–9. 15 Kok P, Buijs MM, Kok SW, Van Ierssel IH, Frolich M, Roelfsema F et al. Acipimox enhances spontaneous growth hormone secretion in obese women. Am J Physiol Regul Integr Comp Physiol 2004; 286: R693–R698. 16 Morange PE, Alessi MC, Verdier M, Casanova D, Magalon G, Juhan-Vague I. PAI-1 produced ex vivo by human adipose tissue is relevant to PAI-1 blood level. Arterioscler Thromb Vasc Biol 1999; 19: 1361–1365. 17 Dusserre E, Moulin P, Vidal H. Differences in mRNA expression of the proteins secreted by the adipocytes in human subcutaneous and visceral adipose tissues. Biochim Biophys Acta 2000; 1500: 88–96. 18 Alessi MC, Peiretti F, Morange P, Henry M, Nalbone G, JuhanVague I. Production of plasminogen activator inhibitor 1 by human adipose tissue: possible link between visceral fat accumulation and vascular disease. Diabetes 1997; 46: 860–867. 19 Lau DC, Dhillon B, Yan H, Szmitko PE, Verma S. Adipokines: molecular links between obesity and atheroslcerosis. Am J Physiol Heart Circ Physiol 2005; 288: H2031–H2041. 20 Moller DE. Potential role of TNF-alpha in the pathogenesis of insulin resistance and type 2 diabetes. Trends Endocrinol Metab 2000; 11: 212–217. 21 Birgel M, Gottschling-Zeller H, Rohrig K, Hauner H. Role of cytokines in the regulation of plasminogen activator inhibitor-1 expression and secretion in newly differentiated subcutaneous human adipocytes. Arterioscler Thromb Vasc Biol 2000; 20: 1682–1687. 22 Mohamed-Ali V, Goodrick S, Rawesh A, Katz DR, Miles JM, Yudkin JS et al. Subcutaneous adipose tissue releases interleukin6, but not tumor necrosis factor-alpha, in vivo. J Clin Endocrinol Metab 1997; 82: 4196–4200. 23 Giacchetti G, Faloia E, Mariniello B, Sardu C, Gatti C, Camilloni MA et al. Overexpression of the renin–angiotensin system in human visceral adipose tissue in normal and overweight subjects. Am J Hypertens 2002; 15: 381–388. 24 Roebuck KA, Finnegan A. Regulation of intercellular adhesion molecule-1 (CD54) gene expression. J Leukoc Biol 1999; 66: 876–888. 25 Buijs MM, Romijn JA, Burggraaf J, De Kam ML, Cohen AF, Frolich M et al. Growth hormone blunts protein oxidation and promotes protein turnover to a similar extent in abdominally obese and normal-weight women. J Clin Endocrinol Metab 2002; 87: 5668–5674. 26 Arner P. Insulin resistance in type 2 diabetes: role of fatty acids. Diabetes Metab Res Rev 2002; 18: S5–S9. 27 Chan DC, Barrett HP, Watts GF. Dyslipidemia in visceral obesity: mechanisms, implications, and therapy. Am J Cardiovasc Drugs 2004; 4: 227–246.


1679 28 Bastard JP, Bruckert E, Robert JJ, Ankri A, Grimaldi A, Jardel C et al. Are free fatty acids related to plasma plasminogen activator inhibitor 1 in android obesity? Int J Obes Relat Metab Disord 1995; 19: 836–838. 29 Silveira A. Postprandial triglycerides and blood coagulation. Exp Clin Endocrinol Diabetes 2001; 109: S527–S532. 30 Gurlek A, Bayraktar M, Kirazli S. Increased plasminogen activator inhibitor-1 activity in offspring of type 2 diabetic patients: lack of association with plasma insulin levels. Diabetes Care 2000; 23: 88–92. 31 Scelles V, Raccah D, Alessi MC, Vialle JM, Juhan-Vague I, Vague P. Plasminogen activator inhibitor 1 and insulin levels in various insulin resistance states. Diabete Metab 1992; 18: 38–42. 32 Kakafika AI, Liberopoulos EN, Karagiannis A, Athyros VG, Mikhailidis DP. Dyslipidaemia, hypercoagulability and the metabolic syndrome. Curr Vasc Pharmacol 2006; 4: 175–183. 33 Lemieux I, Pascot A, Tchernof A, Bergeron J, Prud’homme D, Bouchard C et al. Visceral adipose tissue and low-density lipoprotein particle size in middle-aged versus young men. Metabolism 1999; 48: 1322–1327.

34 Milosavljevic D, Griglio S, Le Naour G, Chapman MJ. Preferential reduction of very low-density lipoprotein-1 particle number by fenofibrate in type IIB hyperlipidemia: consequences for lipid accumulation in human monocyte-derived macrophages. Atherosclerosis 2001; 155: 251–260. 35 Durrington PN, Mackness MI, Bhatnagar D, Julier K, Prais H, Arrol S et al. Effects of two different fibric acid derivatives on lipoproteins, cholesteryl ester transfer, fibrinogen, plasminogen activator inhibitor and paraoxonase activity in type IIb hyperlipoproteinaemia. Atherosclerosis 1998; 138: 217–225. 36 Frayn KN. Visceral fat and insulin resistance – causative or correlative? Br J Nutr 2000; 83: S71–S77. 37 Alessi MC, Bastelica D, Mavri A, Morange P, Berthet B, Grino M et al. Plasma PAI-1 levels are more strongly related to liver steatosis than to adipose tissue accumulation. Arterioscler Thromb Vasc Biol 2003; 23: 1262–1268. 38 Agarwal N, Sharma BC. Insulin resistance and clinical aspects of non-alcoholic steatohepatitis (NASH). Hepatol Res 2005; 33: 92–96.
