Polycystic ovary syndrome (PCOS) is one of the most common endocrine disorders, affecting 5–10% of women of child-bearing age (^{Knochenhauer et al., 1998}; ^{Diamanti-Kandarakis et al., 1999}; ^{Asuncion et al., 2000}; ^{Azziz et al., 2004}; ^{Lindholm et al., 2008}). Women with PCOS carry a number of risk markers for cardiovascular disease (CVD), such as obesity, type 2 diabetes, hypertension and dyslipidemia, and these risk factors remain, even after adjustment for age and BMI (^{Lo et al., 2006}).

Obesity is a common feature of PCOS, with prevalence rates varying between 38 and 88% (^{Barber et al., 2006}). The elevated metabolic risk profile associated with obesity is, to a large extent, dependent on body fat distribution. Central obesity is of importance in the definition of the metabolic syndrome, and especially visceral fat accumulation is associated with insulin resistance and CVD (^{Despres and Lemieux, 2006}).

Obesity is characterized by a systemic low-grade inflammatory state that is thought to originate from the adipose tissue (^{Kershaw and Flier, 2004}). The adipose tissue is an active endocrine organ that produces and releases hormones, pro- and anti-inflammatory cytokines and chemoattractant cytokines (chemokines). Pro-inflammatory molecules produced by the adipose tissue are proposed to be active participants in the development of insulin resistance and the increased risk of CVD associated with obesity (^{Mahadik et al., 2008}).

Recently, it was reported that adipose tissue in obesity is characterized by macrophage infiltration, and that macrophages are an important contributor to inflammation (^{Cancello and Clement, 2006}). An increased macrophage infiltration could be caused by different chemokines thought to attract inflammatory cells to the adipose tissue. One such example is chemokine (C-C motif) ligand 2 (CCL2) (also known as monocyte chemoattractant protein-1, MCP-1) and its cognate receptor chemokine receptor 2 (CCR2), both of which contribute to systemic insulin resistance, macrophage infiltration and maintenance of macrophages in the adipose tissue (^{Kershaw and Flier, 2004}; ^{Cancello and Clement, 2006}).

In addition to increased macrophage infiltration, the expression and release of pro-inflammatory cytokines, such as macrophage migration inhibitory factor (MIF), tumour necrosis factor-α (TNF-α) and interleukin-18 (IL-18), is increased in adipose tissue in the obese (^{Hotamisligil et al., 1995}; ^{Bruun et al., 2007}; ^{Skopkova et al., 2007}), and there is strong evidence that the expanded adipose tissue mass of the obese contributes, directly or indirectly, to the increased circulating levels of inflammatory markers found in obese patients (^{Hotamisligil, 2006}).

Circulating markers of inflammation, including MCP-1, MIF and high-sensitive C-reactive protein (hs-CRP), are elevated in overweight women with PCOS in comparison with weight-matched controls (^{Diamanti-Kandarakis et al., 2006}; ^{Glintborg et al., 2009}; ^{Gonzalez et al., 2009}). However, thus far no studies concerning inflammatory markers in adipose tissue have been conducted in patients with PCOS.

Our hypothesis was that overweight/obesity would be associated with increased adipose tissue expression of inflammatory markers, irrespective of a PCOS diagnosis.

This study demonstrates that the gene expression pattern of inflammation markers in abdominal superficial subcutaneous tissue is similar in overweight patients with PCOS and overweight control subjects. Furthermore, this finding remains when analyses are restricted to those patients with PCOS who displayed hyperandrogenism. In contrast, lean patients with PCOS display lower expression of adipose tissue inflammation markers compared with the two overweight groups. In concert with this finding, most of the associations between markers of adipose tissue inflammation and variables of the metabolic syndrome in patients with PCOS were lost when adjusted for age and BMI. Thus, overweight or obesity, rather than the PCOS diagnosis per se appears to be the main explanatory variable for increased subcutaneous adipose tissue inflammation in patients with PCOS.

Most phenotypes of PCOS—irrespective of degree of obesity—are hallmarked by a combination of androgen excess and insulin resistance. While the association with type 2 diabetes is well established, the precise CVD risk in PCOS women is still unclear, in part because there are no, or few, longitudinal studies examining cardiovascular events and/or mortality. However, biochemical hyperandrogenism is associated with CVD risk factors in young as well as middle-aged and post-menopausal healthy women (^{Sowers et al., 2005}; ^{Sutton-Tyrrell et al., 2005}; ^{Coviello et al., 2006}; ^{Shaw et al., 2008}) and in post-menopausal women, hyperandrogenism has also been associated with CVD events (although not independently of BMI) (^{Rexrode et al., 2003}). Previous studies have indicated that serum concentrations of inflammatory markers, such as MCP-1 (i.e. CCL2) and CRP, are increased in overweight patients with PCOS compared with weight-matched controls (^{Gonzalez et al., 2009}), and that the inflammatory status is correlated with levels of hyperandrogenism (^{Glintborg et al., 2009}; ^{Gonzalez et al., 2009}). Other studies have failed to display any differences between overweight controls and overweight patients with PCOS in terms of circulating TNF-α and IL-6 (^{Olszanecka-Glinianowicz et al., 2007}). In contrast with previous studies of inflammatory markers in PCOS patients, we were not able to confirm prior findings of increased circulating levels of hs-CRP, CCL2 and IL-18 in overweight PCOS patients, presumably because our study was powered primarily for the adipose tissue mRNA levels and not for the circulating levels of inflammatory markers. Similarly, we were unable to demonstrate a correlation between adipose tissue mRNA levels for inflammatory markers and FAI or testosterone. However, in line with previous studies SHBG, as an indirect measure of androgenicity and insulin resistance, was negatively correlated with most of the inflammatory markers (^{Glintborg et al., 2009}; ^{Gonzalez et al., 2009}).

Although most studies on PCOS report that obese women with PCOS are more insulin resistant than control obese women, there was no difference in our present study. The reason for this is unclear but may involve the fact that the overweight women with PCOS did not differ in waist circumference compared with controls. When women with PCOS are compared with control women matched for abdominal adiposity, the difference in insulin resistance between the groups is much less marked than if the two groups are matched for BMI (^{Holte et al., 1994}). It is also possible that our relatively small comparative study was unable to reveal true differences in insulin sensitivity. The fact that, based upon waist circumference measurements and waist-hip ratio, obese women with PCOS and obese controls did not differ in this study goes against the notion that anovulatory women build up more truncal-abdominal fat. However, data from recent studies utilizing advanced imaging techniques are ambiguous. Using magnetic resonance imaging, Barber and colleges found that overweight PCOS cases and BMI/fat mass-matched controls were indistinguishable with respect to distribution of fat within visceral, abdominal subcutaneous and gluteofemoral subcutaneous fat depots (^{Barber et al., 2008}). Furthermore, by use of computer tomography, it was recently demonstrated that obese women with PCOS only have a modest, but significant, increase in visceral fat compared with BMI/fat mass-matched non-PCOS women (^{Hutchison et al., 2011}). Notably, these studies (^{Barber et al., 2008}; ^{Hutchison et al., 2011}) included overweight and obese subjects, which may explain the discrepancy from previous studies where mainly lean subjects were included (^{Yildirim et al., 2003}). Concerning visceral adiposity and the close link to metabolic risk, it is unfortunate that we had only biopsies from the abdominal superficial subcutaneous adipose tissue depot. It has recently been reported that the different adipose tissue depots have different profiles of inflammatory markers (^{Harman-Boehm et al., 2007}; ^{Huber et al., 2008}; ^{Poulain-Godefroy et al., 2008}), and genomic and proteomic profiling studies of morbidly obese PCOS omental adipose tissue have revealed changes in components of several biological pathways (^{Corton et al., 2007}, ²⁰⁰⁸). Consequently, there might be differences in the inflammatory status of visceral adipose tissue in overweight PCOS versus weight-matched controls. However, it is not likely that the visceral tissue directly would account for the increased circulating concentrations of plasma cytokines and chemokines. Only a limited number of adipose-derived mediators are actually released into the circulation in amounts sufficient to result in the increased systemic levels which are associated with obesity (^{Dahlman et al., 2005}; ^{Murdolo et al., 2007}). Indirectly, however, the portal drainage of visceral fat provides direct hepatic access to free fatty acids and adipokines that may activate hepatic immune mechanisms with production of inflammatory mediators, such as CRP (^{Despres and Lemieux, 2006}). Further studies of adipose tissue depot-specific inflammation in relation to circulating levels of inflammatory markers in larger cohorts of women with PCOS are thus warranted.

Support for a possible causal relationship between obesity-associated low-grade inflammation and metabolic impairment is gained from epidemiological data indicating an association of markers of systemic inflammation (i.e. CRP, IL-6) with insulin resistance and future risk for development of diabetes type 2 (^{Pradhan et al., 2001}). In agreement with previous studies, we found a pronounced effect of increased body mass on adipose tissue expression of chemokines, pro-inflammatory cytokines and markers of inflammation (^{Hotamisligil, 2003}; ^{Bruun et al., 2007}; ^{Huber et al., 2008}; ^{Shakeri-Manesch et al., 2009}), whereas overweight patients with PCOS did not differ from overweight controls in this respect. We showed that MIF mRNA, however, was not influenced by fat mass in agreement with data from our laboratory where MIF expression in subcutaneous fat depots from white women was unaffected by fat mass (^{Evans et al., 2011}). ^{Skurk et al. (2005)} reported that secretion of MIF from isolated human subcutaneous adipocytes correlated with BMI, seemingly in conflict with our data. Importantly, the majority of MIF secreted from adipose tissue originates from non-fat cells in obese women (^{Fain et al., 2006}) and it is also likely that the adipocyte isolation and culturing procedure per se can influence the secretion pattern (^{Ruan et al., 2003}). Similar to MIF, we found no effect of fat mass on adipose tissue IL-6 mRNA. Although a minority of the body's circulating IL-6 is derived from adipose tissue (^{Mohamed-Ali et al., 1997}; ^{Weisberg et al., 2006}), secretion from adipose tissue and circulating IL-6 levels are suggested to be a function of the degree of fat mass (^{Kern et al., 2001}). However, since we did not measure IL-6 secretion, this issue should be investigated further.

Several pro-inflammatory cytokines are elevated in parallel with many metabolic risk markers, whereas the anti-inflammatory and adipocyte-specific hormone adiponectin is consistently lower (^{Spranger et al., 2003}). In agreement with this, our lean patients with PCOS had higher levels of mRNA for adiponectin versus the overweight groups. Adiponectin exerts anti-inflammatory effects on adipocytes primarily via inhibition of the nuclear factor кB pathway (^{Ajuwon and Spurlock, 2005}). As a result, release of a number of pro-inflammatory cytokines (e.g. MCP-1, IL-6) by adipocytes is suppressed (^{Ajuwon and Spurlock, 2005}; ^{Dietze-Schroeder et al., 2005}). Expression of the anti-inflammatory cytokine IL-10, on the other hand, was lower in the lean group which has been reported previously (^{Juge-Aubry et al., 2005}). The fact that anti-inflammatory cytokine IL-10 is upregulated in adipose tissue from obese humans suggests that adipose tissue may produce anti-inflammatory cytokines with the potential to limit the overall pro-inflammatory response commonly seen in obesity.

It is possible that adipokine secretion pattern might influence blood pressure parameters and in this respect, most focus has been on endocrine factors, such as leptin and adiponectin, which may activate and inhibit, respectively, the sympathetic nervous system (reviewed in ^{Mathieu et al., 2009}). In this study, adipose tissue TNF-α (mainly autocrine/paracrine action) correlated with systolic blood pressure. Since TNF-α may act on immune cells and induce local and systemic inflammation, endothelial as well as vascular function might be altered. Indeed, systemic TNF-α has been suggested as an independent risk factor for elevated blood pressure in apparently healthy subjects (^{Bautista et al., 2005}). Notably, the contribution of adipose tissue TNF-α to the circulation has not been fully resolved and there is only a modest elevation of serum TNF-α concentrations in obesity (^{Hauner, 2005}).

There are a number of limitations in this study. The study included a limited number of participants, which presumably is the reason why a difference in circulating levels of inflammatory markers was not detected. Furthermore, control subjects were significantly older than the patients with PCOS, but we adjusted for this age difference, and this appears not to have affected the primary outcome variables: as our study did not include a lean control group, we are unable to determine whether lean patients with PCOS differ in adipose tissue expression of inflammatory markers from weight-matched controls. Furthermore, the inclusion of a lean control group would have enhanced our understanding of how adipose tissue inflammation is influenced by hyperandrogenism and PCOS. Another limitation, discussed above, is that only adipose tissue biopsies from one anatomical location were available for study.

In conclusion, fat mass rather than a PCOS diagnosis per se seems to be the main determinant of increased inflammatory markers in subcutaneous adipose tissue. Further studies investigating the different clinical phenotypes of PCOS in terms of cardiovascular risk are warranted.

Å.L., M.B., I.S.P. and J.B. designed the study and collected the data, C.B., I.D., T.H. and J.B. performed the data analysis. All authors have participated in the drafting and revising of the manuscript and have given their final approval of the version to be published.

This work was supported in part by the Swedish Society for Medical Research, the Faculty of Medicine of Umeå University, the Selander Foundation and the Family Planning Foundation, Uppsala University and FoU Norrbotten. Funding to pay the Open Access publication charges for this article was provided by the Family Planning Foundation, Uppsala University.

This study could not have been done without the excellent help from our research nurses Sigrid Nyberg, Åsa Pettersson and Ingrid Ottenbring. We thank Maria Lindberg for creating a program for the Beckman Coulter Biomek FX workstation. We thank Prof. Tommy Olsson for initial ideas on this study and for critical reading of the manuscript.