Visceral adipose tissue activated macrophage content and inflammatory adipokine secretion is higher in pre-eclampsia than in healthy pregnancys

Obesity increases pre-eclampsia (PE) risk. Adipose tissue inflammation may contribute to the clinical syndrome of PE. We compared adipose tissue macrophage infiltration and release of pro-inflammatory adipokines in PE and healthy pregnancy. Subcutaneous and visceral adipose tissue biopsies were collected from healthy (n=13) and PE (n=13) mothers. Basal and lipopolysaccharide (LPS) stimulated adipocyte TNFα, IL-6, CCL-2, and CRP release was measured. Adipose tissue cell densities of activated (cfms+) and total (CD68+) macrophages were determined. In PE only, visceral adipose tissue TNFα release was increased after LPS stimulation (57 [76] versus 81 [97] pg/ml/µg DNA, P=0.030). Basal TNFα release was negatively correlated insulin sensitivity of visceral adipocytes (r = −0.61, P=0.030) in PE. Visceral adipocyte IL-6 release was increased after LPS stimulation in PE only (566 [696] versus 852 [914] pg/ml/µg DNA, P=0.019). Visceral adipocyte CCL-2 basal (67 [61] versus 187 [219] pg/ml/µgDNA, P=0.049) and stimulated (46 [46] versus 224 [271] pg/ml/µg DNA, P=0.003) release was greater than in subcutaneous adipocytes in PE only. In PE, median TNF mRNA expression in visceral adipose tissue was higher than controls (1.94 [1.13–4.14] versus 0.8 [0.00–1.27] TNF/PPIA ratio, P=0.006). In visceral adipose tissue, CSF1R (a marker of activated macrophages) mRNA expression (24.8[11.0] versus 51.0[29.9] CSF1R/PPIA ratio, P=0.011) and activated (cfms+) macrophage count (6.7[2.6] versus 15.2[8.8] % cfms+/adipocyte, P=0.031) were higher in PE than in controls. In conclusion, our study demonstrates dysregulation of inflammatory pathways predominantly in visceral adipose tissue in PE. Inflammation of visceral adipose tissue may mediate many of the adverse metabolic effects associated with PE.


Introduction
Pre-eclampsia (PE) occurs in 2-4% of all pregnancies and is a leading cause of maternal and neonatal morbidity and mortality. There is considerable evidence that maternal obesity, increased insulin resistance, inflammation, and aberrant fatty acid metabolism are involved in the pathogenesis of PE [1,2]. The link between adiposity, inflammation, and insulin resistance has been increasingly acknowledged since Hotamisligil et al. [3] first demonstrated this relationship in 1993, and there has been extensive research into elucidating the mechanisms which link these conditions. White adipose tissue secretes a number of pro-inflammatory mediators which contribute significantly to the chronic inflammatory state and metabolic complications of obesity [4]. It is plausible that similar disturbances in adipocyte function could contribute to the development of the clinical syndrome of PE, a state of inflammation and insulin resistance. Indeed we have previously shown that third trimester subcutaneous adipocyte lipolysis is more resistant to insulin suppression in PE compared with healthy pregnancy [5]. Visceral instructions. HOMA was calculated as follows: [fasting insulin (mU/l) × fasting glucose (mmol/l)]/22.5. Plasma leptin, adiponectin, IL-6, TNFα, and VEGF R1/sFlt1 were carried out using commercial ELISA kits (all R&D Systems, Abingdon, U.K.).

Adipose tissue lipolysis assay
Adipocyte preparation was by a modification of the method described by Rodbell [20], with temperature maintained as near to 37 • C throughout. The method is described in detail in Huda et al. [5]. Fat cell number was measured indirectly by quantitating the DNA content in a known volume of adipocyte suspension. The number of adipocytes are directly proportional to the amount of adipocyte DNA content and lipolysis rates are expressed per μg of DNA. The DNA was isolated from a known quantity of adipocyte suspension that was previously frozen at −70 • C and thawed at room temperature using the Blood Prep DNA Purification protocol on the ABI Prism TM 6100 Nucleic Acid Prep-Station (Applied Biosystems). The concentration of DNA was quantitated using Nanodrop R ND 100. The diameters of 100 adipocytes were manually measured using a stage micrometer and a mean calculated for each preparation (Supplementary Figure S1). For information on the adipocyte profile from control and PE biopsies please refer to the Supplementary Results (Supplementary Figure S1) of Huda et al. [5]. The fat cell insulin sensitivity index (FCISI) was calculated as the percentage inhibition of isoproterenol-stimulated lipolysis by insulin.

Adipokine analysis
Paired samples of both basal (no reagent added to adipocyte cell suspension and buffer) and LPS (LPS concentration of 1 μM to adipocyte cell suspension) were incubated. An aliquot of the buffer layer below the adipocyte suspension at time 120 min was frozen at −80 • C for later analysis of cytokines. Adipokine quantitation, apart from sFlt-1, was carried out with Bio-Plex (BIO-RAD R ) system and suspension array technology. The assay was customized to detect and measure multiple adipokines (PAI-1 LOB1786, CCL-2 LUH279, CRP LOB1707, IL-6 LUH206, Leptin LUB398, TNFα LUH210, Adiponectin LOB1065, IL-110 LUH217 R&D Systems) using the R&D Systems Obesity Base Kit (LOB000). VEGF R1/sFlt1 was quantitated using a commercial ELISA kit (R&D Systems, Abingdon, U.K.).

Tissue mRNA expression quantitation
Total RNA was isolated from adipose tissue using the ABI PRISM 6100 Nucleic Acid Prepstation following manufacturer's instructions (Applied Biosystems, Warrington, U.K.). cDNA was reverse transcribed from RNA using a High Capacity cDNA Reverse Transcriptase Kit (Applied Biosystems, Warrington, U.K.) according to manufacturer's instructions. Target gene expression was quantitated relative to a control gene [5,21] (PPIA Hs99999904 m1) using commercial primer probe sets (CD68 Hs00154355-m1, CSF1R Hs99999197-m1, IL6 HS00174131-ms, TNF Hs00174128-m1, CCL2 Hs00234140 m1, FLT1 Hs01052961 m1) Applied Biosystems, Warrington, U.K.) in a final volume of 25 μl on an 7900HT Sequence Detection System (Applied Biosystems, Warrington, U.K.) according to manufacturer's instructions. Quantitation analysis was carried out using SDS Version 2.3 software (Applied Biosystems), which calculated the threshold cycle (C T ) values. The expression of target assays were normalized by subtracting the C T value of the endogenous control from the C T value of the target assay. The fold increase relative to the control was calculated using the 2 − C T . The expression of the target assay was then expressed as a percentage relative to the endogenous control assay.

Macrophage cell densities
Immunocytochemistry was performed on 7 μM sections of paraffin-embedded adipose tissue biopsies using the avidin: biotinylated enzyme complex method (see Supplementary Methods). The sample size was smaller (n=9) as there was insufficient stored adipose tissue in four of the PE cases. The antibodies used were polyclonal anti-human cfms (CBL776 Chemicon) and monoclonal anti-human macrophage CD68 (Dako-CD68, PG-M1 M876) which identify activated and total macrophages respectively. Macrophages were identified using histological analysis in ten randomly selected high powered fields (×400 objective magnification) and were counted by two independent observers who were blinded to the specimen details. The area for each high-powered field was 0.23 mm 2 . Macrophages within the blood vessels were not included in the counts. Tissue macrophage densities were expressed as cell count per adipocyte.

Statistical analysis
Data were assessed for normal distribution using a Ryan-Joiner test and transformed to achieve a normal distribution where necessary. Means with SD are presented for normalized data. Median and interquartile range are presented for non-parametric data. Comparison within individuals was by paired t-test (subcutaneous adipose tissue versus visceral adipose tissue and basal versus LPS stimulated) and between PE and control groups by two sample t-test or Mann-Whitney test as appropriate. Mann-Whitney and χ2 tests were used to test between differences in semi-categorical and categorical variables respectively. For adipocyte sFlt-1 secretion, which included many zero values, the Wilcoxon rank sum test was used to test differences between groups. Pearson's correlation coefficients and regression analysis were calculated to assess associations between variables, and results were expressed as r value, R 2 and P-value. A P-value of <0.05 was considered significant. The data were adjusted for potential cofounders using the General Linear Model. All statistical analysis was carried out in Minitab (version 17).

Maternal antenatal booking characteristics and third trimester plasma metabolic and inflammatory profile
Maternal antenatal booking characteristics (11-14 weeks) for women with PE and controls matched for age, BMI, and smoking are shown in Table 1. Women with PE had a tendency for a higher booking systolic and diastolic blood pres- sure, were delivered at an earlier gestation and offspring were of lower birth weight centile than controls. Pre-delivery diagnostic blood pressure was higher in PE (Table 1). Twelve out of thirteen had late onset PE (onset after 35 weeks of gestation) and eight out of thirteen had severe PE as defined in NICE guideline 107. Plasma sFlt-1 levels were significantly higher in PE than control women (P<0.001). In addition, mothers with PE had higher plasma triglyceride and NEFA than controls. Maternal serum TNFα was also 64% higher in PE than controls (P=0.024).

Adipocyte TNFα release
There were no differences in either basal or stimulated TNFα release between subcutaneous and visceral adipocytes.  Figure S2). This was independent of subcutaneous adipocyte size (P=0.002 and P=0.006 respectively). This association with BMI was not seen in subcutaneous adipocytes in PE (Supplementary Figure S1). In PE but not controls, both basal and stimulated TNFα release from visceral adipocytes negatively correlated with visceral adipocyte size (r = −0.65, R 2 = 41.8%, P=0.017, and r = −0.65, R 2 = 42.2%, P=0.016). Basal release of TNFα negatively correlated with a direct measure of visceral adipocyte insulin sensitivity (FCISI) in this depot (r = −0.60, R 2 = 30.3, P=0.030). This relationship was independent of BMI (P=0.023), but not of cell size, (P=0.060).  Figure S3). This correlation was independent of cell size.

Adipocyte CCL-2 release
Basal and stimulated CCL-2 from visceral adipocytes was higher than that from subcutaneous adipocytes in PE ( (Figure 3). There were no differences between basal and LPS stimulated release of CCL-2 from either subcutaneous adipocytes or visceral adipocytes in either the controls or the PE group. Basal and stimulated release of CCL-2 from subcutaneous adipocytes was correlated with maternal BMI (r=0.79, R 2 = 59.1%, P=0.01 basal, and r=0.82, R 2 64.9%, P=0.001 stimulated) in controls only (Supplementary Figure S4). This relationship was independent of cell size.

Adipocyte CRP release
There was no difference between subcutaneous adipocyte and visceral adipocyte CRP release in either pregnancy group. Nor was there any difference between basal and stimulated release of CRP from either subcutaneous adipocytes or visceral adipocytes in control and PE group (Supplementary Figure S5). Basal and stimulated CRP release from subcutaneous adipocytes were correlated with circulating maternal plasma levels of CRP in controls (basal r=0.69, P=0.010, and stimulated r=0.70, P=0.008). In PE this association appeared to be stronger (basal P=0.91, P<0.001, and stimulated r=0.91, P<0.001). CRP release from subcutaneous adipocytes in PE was correlated with maternal plasma IL-6 concentrations (basal r=0.72, P=0.006, and stimulated r=0.70, P=0.008). Basal and stimulated CRP

Adipokine gene expression
There was no difference in subcutaneous adipose tissue TNF mRNA expression between PE and controls. Median

Adipose tissue macrophage densities and mRNA expression in pre-eclampsia and controls
The mean CSF1R mRNA expression in visceral adipose tissue, but not subcutaneous adipose tissue, was higher in PE relative to controls matched for BMI (24.8[11.0] versus 51.0[29.9] CSF1R/PPIA ratio, P=0.011) (Figure 4). CD68 expression was not significantly different in subcutaneous adipose tissue or visceral adipose tissue between PE and controls. Similarly the mean percentage of cfms+ macrophages in visceral adipose tissue, but not subcutaneous adipose tissue, was higher in PE relative to controls (P=0.032) ( Figure 5). There were no differences in tissue densities of CD68+ macrophages between PE and controls in either subcutaneous adipose tissue or visceral adipose tissue.

Discussion
The key observations of our study were that, in PE, visceral adipose tissue had a higher activated macrophage content and higher mRNA expression of TNFα than in controls and that adipocytes from visceral adipose tissue in PE were more responsive to LPS stimulation in terms of releasing higher levels of TNFα and IL-6. This suggests that inflammation in visceral adipose tissue may be a hallmark of PE. Human fat compartments can be classified into lower body subcutaneous adipose tissue, upper body subcutaneous adipose tissue and visceral adipose tissue and each depot has different characteristics. Upper body subcutaneous adipose tissue (from which our biopsy of subcutaneous adipose tissue was taken) is relatively resistant to insulin suppression of lipolysis and is estimated to be the source of 60% of circulating NEFA. Visceral adipose tissue can be regarded as an 'ectopic' site of fat accumulation and high levels of visceral adipose tissue are related to metabolic obesity even at a low BMI [22]. We have previously shown that in pregnancy visceral adipose tissue adipocytes are more insulin resistant than subcutaneous adipose tissue adipocytes and that in PE subcutaneous adipose tissue adipocytes become as insulin resistant as their visceral adipose tissue counterparts [5]. Together with the data presented here, this focuses attention on adipocyte dysfunction as a key pathogenic pathway in PE.
TNFα appears to play a central role in the adipocyte inflammatory response in pregnancies complicated by PE. TNFα transcription was higher in PE adipose biopsies compared with controls and adipocytes exposed to LPS were Means and standard deviations are presented. *P<0.05 between the mean percentage of VAT cfms + /adipocyte in PE compared to controls by two sample t-test on log transformed data. stimulated to secrete significantly higher levels of TNFα. TNFα is known to be an important determinant of insulin sensitivity in pregnancy and can lead directly to insulin resistance by inhibiting insulin signaling through several mechanisms including inducing serine phosphorylation of the insulin receptor substrate 1 (IRS1) [23]. TNFα is also a potent stimulator of lipolysis through down-regulation of perilipin and suppression of the anti-lipolytic GTP-binding membrane proteins Gα i [24]. In our study, TNFα release from visceral adipose tissue was negatively correlated with adipocyte insulin sensitivity (as assessed by the ability of insulin to suppress lipolysis) of visceral adipose tissue adipocytes in PE thus linking inflammation, insulin resistance and lipolysis in PE. Furthermore, TNFα release was negatively correlated with fat cell size in visceral adipose tissue in PE in this cohort. We have previously shown that in PE visceral adipose tissue has a significantly larger proportion of small adipocytes than controls and our current data are consistent with observations in the non-pregnant that insulin resistance is associated with a greater proportion of small adipose cells. Higher visceral adipose tissue TNFα release in smaller adipocytes may contribute to their insulin resistance in PE.
We showed for the first time that FLT1 expression was higher in both subcutaneous and visceral adipose tissue in PE compared with control women although there was little evidence for secretion of sFlt-1 from adipocytes in culture. This would suggest the adipocyte does not act as a source additional to the placenta for circulating sFlt-1 in PE. Other stromal cells located in adipose tissue may express FLT1 and a potential paracrine role for adipose tissue VEGF receptor in angiogenesis related to gestational adipose tissue expansion is possible.
In addition to the higher production of TNFα in visceral adipose tissue, the mRNA expression of other adipokines in visceral adipose tissue biopsies, namely IL-6 and CCL-2, was also higher in PE than controls. Furthermore, secretion of both adipokines from visceral adipocytes was higher than in subcutaneous adipocytes regardless of whether from PE or control women. In PE, IL-6 release from visceral adipose tissue adipocytes was more responsive to stimulation with LPS. IL-6 is a stimulator of whole body lipolysis with anti-insulin effects [25,26] and has also been related to pregnancy-associated insulin resistance. Unlike with TNFα, we did not observe a relationship between IL-6 and direct measures of insulin sensitivity in adipose tissue in our cohort. In vivo, the CCL2 signal is important for macrophage infiltration of white adipose tissue. Indeed knockout of the receptor for CCL2 in mice protects against obesity induced inflammation and insulin resistance [27]. The higher expression of CCL2 in visceral adipose tissue may have important implications for macrophage recruitment.
We observed higher mRNA expression of CSF1R and a higher density of cfms positive macrophages in the visceral adipose tissue of women with PE, while the total number of macrophages both in terms of CD68 gene expression, and of tissue CD68 positive macrophage content, did not differ. These data suggest that there is no additional recruitment of macrophages to visceral fat in PE, despite the increased CCL2 expression, but that macrophages become activated to a more pro-inflammatory macrophage phenotype. Women with PE have more pro-inflammatory/activated macrophages in visceral adipose tissue than controls and these cells are another source of cytokines, including TNFα and IL-6, in PE. We observed higher maternal plasma concentrations of TNFα in this small cohort and previously we and others have found plasma levels of both TNFα and IL-6 to be elevated in PE [28,29]. In addition to the potential paracrine and endocrine effects on lipid and glucose metabolism discussed above, TNFα and IL-6 are also implicated in endothelial dysfunction, leukocyte activation, and alterations in coagulation-all features of PE [30][31][32]. The chronic infusion of TNFα or IL-6 into normal pregnant rodents significantly increases arterial pressure and impairs renal haemodynamics [33]. Previously, the placenta has been suggested as a source of maternal plasma TNFα and IL-6 with its overproduction being secondary to placental hypoxia. However, expression has not consistently been seen to be higher in the placentae of women with PE thereby implicating another tissue source for the elevated concentrations found in peripheral blood [34,35]. Visceral adipose tissue could be an alternative source for these cytokines in PE although it is yet to be determined whether adipocytes or infiltrating macrophages, or both, produce the cytokines.
While the adipose tissue inflammatory changes associated with PE appear to be located predominantly at visceral adipose tissue, there were also some important observations regarding subcutaneous adipose tissue. Basal levels of adipokine release were generally lower in subcutaneous adipose tissue compared to visceral adipose tissue and were refractory to stimulation by LPS, apart from TNFα. Interestingly subcutaneous adipose tissue adipocyte TNFα, IL-6, and CCL-2 release was strongly correlated with maternal BMI possibly reflective of the link between obesity and chronic low-grade inflammation. This association with maternal BMI was lost in PE suggesting that visceral adipose tissue located inflammatory mechanisms override the background chronic inflammation of obesity. This is consistent with the concept of 'benign' and 'metabolic' obesity where the amount of visceral adipose tissue is associated with the metabolic consequences of obesity rather than obesity per se [22]. The higher basal release of TNFα from subcutaneous adipose tissue in PE (equivalent to the amounts released from visceral adipose tissue), due to the large size of the subcutaneous adipose tissue depot, could potentially impact significantly on circulating NEFA and may be a mechanism through which early exaggerated rise of NEFA in PE occurs. We observed that in PE, TNFα release from visceral adipose tissue does relate to the insulin sensitivity of lipolysis in this adipose depot. The association between subcutaneous adipose tissue adipocyte CRP release and maternal plasma CRP concentrations suggests that subcutaneous adipose tissue is an important determinant of circulating CRP in the third trimester of pregnancy in both healthy and PE pregnancies. It is interesting to note that in PE this association also extends to maternal plasma IL-6 levels perhaps reflecting the wider metabolic disturbances associated with PE.
The strengths of the present study were the direct ex vivo assessment of adipocyte inflammatory function in subcutaneous adipose tissue and visceral adipose tissue derived from the same women. This could be related to the lipolytic function of the same adipose tissue and to plasma concentrations of inflammatory markers. Both total and activated macrophage content was assessed by two independent methods (immunohistochemistry and gene expression) and the data were consistent. We did not establish the M1/M2 phenotype of the infiltrating macrophages and this would be a useful future analysis to carry out. Although the sample size was small, BMI, smoking and age-matching with controls reduced confounding and there was sufficient power to detect consistent statistically significant differences. The gestational age was different between the two groups and this is a limitation of our study that is difficult to avoid due to the characteristics of the condition under study. If we had used preterm matched controls, this also raises issues as to whether these control can be considered as 'healthy' . However, the majority of women were late onset PE and near to term. The present study is a cross-sectional comparison therefore it is difficult to draw conclusions regarding which pathways are causal in PE and which are consequences of the condition and our data should be interpreted accordingly.

Conclusion
Our study demonstrates that in PE, dysregulation of inflammatory pathways is located predominately in visceral adipose tissue and not subcutaneous adipose tissue. In PE, visceral adipose tissue has a higher activated macrophage content, higher mRNA expression of TNFα and is more responsive to LPS stimulation in terms of releasing higher levels of TNFα and IL-6 than in control pregnancy. Inflammation at visceral adipose tissue may be a hallmark of