Maternal Gestational Diabetes Mellitus increases placental and foetal lipoprotein-associated Phospholipase A2 which might exert protective functions against oxidative stress

Increased Lipoprotein associated phospholipase A2 (LpPLA2) has been associated with inflammatory pathologies, including Type 2 Diabetes. Studies on LpPLA2 and Gestational Diabetes Mellitus (GDM) are rare, and have focused mostly on maternal outcome. In the present study, we investigated whether LpPLA2 activity on foetal lipoproteins is altered by maternal GDM and/or obesity (a major risk factor for GDM), thereby contributing to changes in lipoprotein functionality. We identified HDL as the major carrier of LpPLA2 activity in the foetus, which is in contrast to adults. We observed marked expression of LpPLA2 in placental macrophages (Hofbauer cells; HBCs) and found that LpPLA2 activity in these cells was increased by insulin, leptin, and pro-inflammatory cytokines. These regulators were also increased in plasma of children born from GDM pregnancies. Our results suggest that insulin, leptin, and pro-inflammatory cytokines are positive regulators of LpPLA2 activity in the foeto-placental unit. Of particular interest, functional assays using a specific LpPLA2 inhibitor suggest that high-density lipoprotein (HDL)-associated LpPLA2 exerts anti-oxidative, athero-protective functions on placental endothelium and foetus. Our results therefore raise the possibility that foetal HDL-associated LpPLA2 might act as an anti-inflammatory enzyme improving vascular barrier function.


Results
LpPLA 2 is expressed in human placental tissue and secreted by Hofbauer cells. Using immune fluorescence staining of placental tissue sections, LpPLA 2 protein was localized predominantly in the villous stroma and to some extent also in the sub-endothelial space (Fig. 2a-c). Van Willebrand factor (vWF), a marker of endothelial cells, was expressed along placental vessel linings ( Fig. 2a) but vWF did not co-localize with LpPLA 2 in placental tissue (Fig. 2a). Cytokeratin 7 (CK7), which serves as marker of trophoblast, was localized all around the fused syncytial layer of the villus, but did not co-localize with LpPLA 2 either (Fig. 2b). However, co-localization of LpPLA 2 with CD163, a marker of HBCs, was observed in the villous stroma (Fig. 2c), suggesting that HBCs are the main cell type in the placenta producing LpPLA 2 .
To prove that GDM affects LpPLA 2 production and thereby activity from macrophages, HBCs from control and GDM placental tissue were isolated and cultivated under the same conditions for six days. LpPLA 2 activity was determined in the collected supernatants for each time point. After 48 h, GDM-HBCs secreted more active LpPLA 2 than control HBCs, and at 72 h the difference became statistically significant (Fig. 2d) and persisted until day 6 (2.3-fold increase, p = 0.002). To demonstrate that the enzyme activity corresponds specifically to LpPLA 2 , cells were exposed to Darapladib, a selective inhibitor of LpPLA 2 activity 38 . LpPLA 2 activity was completely absent (−93%) in the HBCs supernatants after inhibitor treatment (Fig. 2e).
Insulin and Leptin regulate LpPLA 2 activity in vitro. As maternal GDM appeared to affect LpPLA 2 activity in HBCs, we assessed whether glucose or insulin are contributing factors. HBCs isolated from control placentae were exposed to glucose or insulin for 72 h, each reagent was added three times, once every 24 h. Supernatants were used to assess LpPLA 2 activity and cells were harvested for Western Blot. Insulin caused an increase in LpPLA 2 activity (Fig. 3a) with a maximum effect at 20 nM (+22%, p = 0.004) and also LpPLA 2 protein increased (Fig. 3a, Western Blot insert). Interestingly, glucose levels did not affect LpPLA 2 activity (data not shown). In addition to glucose, we also stimulated cells with Leptin. Plasma leptin levels are increased in obese mothers and their children, and maternal obesity is a major predisposing factor for development of GDM. Indeed, leptin also led to a moderate yet significant increase in LpPLA 2 activity (+15%, p = 0.01 at 500 pg/ml; Fig. 3b) and an even more pronounced increase in LpPLA 2 protein measured by Western Blot (Fig. 3b, insert).

Pro-inflammatory cytokines and adhesion molecules stimulate LpPLA 2 activity in vitro.
Low-grade inflammation and placental endothelial dysfunction are characteristic for GDM pregnancies. Levels of TNFα (inflammation) or ICAM-1 and VCAM-1 (endothelial dysfunction) are classically elevated in these conditions and may therefore alter LpPLA 2 activity. To test this, HBCs were exposed to a range of concentrations of the respective cytokines for 72 h. The supernatants were used for LpPLA 2 activity assay and cell lysates for Western Blots. Although the effects were moderate, all three molecules had a positive effect on LpPLA 2 activity (Fig. 3c,d). TNFα had a maximum effect of 22% increase (p = 0.003, Fig. 3c), whereas both ICAM-1 and VCAM-1 (Fig. 3d) caused an increase around 9% each (p = 0.015 and p = 0.035, respectively). Concomitant increases in LpPLA 2 protein were observed for all three stimuli (see respective Western Blot inserts.)

IL-4 and IL-13 negatively regulate LpPLA 2 activity in vitro.
To also test possible effects of anti-inflammatory cytokines on LpPLA 2 activity, HBCs were exposed to either IL-4 or IL-13 for 72 h. Both cytokines significantly reduced LpPLA 2 activity (Fig. 3e) as well as LpPLA 2 protein (Fig. 3e, Western Blot insert). IL-4 treatment dose-dependently decreased LpPLA 2 activity up to 22% (p < 0.01). Even more pronounced reductions (up to −39%, p < 0.001) were observed with IL-13. Overview of the study design and experimental set-up. The study investigated LpPLA 2 in three sample matrices: (1) isolated placental Hofbauer cells, (2) total placental tissue, and (3) on lipoproteins isolated from foetal cord blood plasma drawn immediately after delivery. Tissue and plasma were sampled from both Control and GDM subjects and used as described in the Material and Method section. Abbreviations: HBCs = Hofbauer cells; GDM = gestational diabetes mellitus; LpPLA 2 = lipoprotein-associated phospholipase A2; LDL = low density lipoprotein; HDL = high density lipoprotein; ELISAs = enzyme-linked immunosorbent assays; ECIS = electrical cell-substrate impedance sensing; DHR = 123-dihydrorhodamine.
SCIEnTIfIC REPoRtS | 7: 12628 | DOI:10.1038/s41598-017-13051-6 HDL is the main carrier of neonatal plasma LpPLA 2 activity. In adults, LDL is the main carrier of LpPLA 2 activity, and its activity is related to the LDL-cholesterol content in adult plasma. In the neonate, however, HDL is the major cholesterol carrying lipoprotein species. For rodent species, were HDL is the main lipoprotein fraction, it has been demonstrated that the majority of LpPLA 2 activity is associated with HDL 31 . When we compared LpPLA 2 activity on LDL and HDL particles isolated from healthy adult subjects and foetuses of non-GDM pregnancies, HDL was indeed the main carrier of LpPLA 2 activity in the foetus ( Fig. 4a; 65% activity on HDL vs. 35% on LDL; p < 0.001).

LpPLA 2 activity on neonatal HDL is increased by GDM. To assess if GDM causes alterations in
LpPLA 2 activity and distribution in the foetus, LDL and HDL was isolated from cord blood plasma and LpPLA 2 activity was measured. No differences in the LDL-associated LpPLA 2 activity was observed between control and GDM foetuses (N = 21/group; Fig. 4b). In contrast, HDL-associated LpPLA 2 activity was significantly increased in the GDM foetus compared to healthy controls (+54%, p = 0.004, Fig. 4b).

Insulin and Leptin levels are increased in GDM foetuses in vivo. Given that insulin and leptin
increased LpPLA 2 activity in vitro, we assessed whether levels of these hormones are altered in foetal plasma in GDM by ELISA. Plasma insulin (Fig. 5a) and leptin (Fig. 5b) were both increased (+54%, p = 0.06 and +84%, p = 0.01, respectively) in GDM foetuses.
Pro-inflammatory cytokines and adhesion molecules are increased in foetal plasma. As these pro-inflammatory molecules stimulated LpPLA 2 activity in vitro, TNFα, ICAM-1, and VCAM-1 levels were measured in plasma of control and GDM neonates. TNFα levels were below the assay detection limit in the majority of samples and could not be reliably quantified. ICAM-1 (Fig. 5c) and VCAM-1 (Fig. 5d) levels were significantly increased in GDM foetuses (+14% p = 0.02 and +10% p = 0.006, respectively).
Anti-inflammatory cytokine plasma levels are unchanged between Control and GDM foetuses. In foetal plasma, IL-4 levels were not different between control and GDM group (Fig. 5e). IL-13 levels were below limit of detection of the ELISA and could not be quantified.
Foetal HDL-associated LpPLA 2 activity is correlated with maternal BMI. As our study cohort for isolation of foetal lipoproteins was not matched for maternal BMI and GDM mothers were significantly overweight compared to control mothers, we tried to investigate maternal BMI as confounding factor, probably affecting foetal LpPLA 2 activity. Pre-pregnancy BMI in control mothers was average 22.7 ± 3.4 kg/m 2 vs. 31.5 ± 7.7 kg/ m 2 in GDM women (p < 0.001). At term, BMI in control women was average 28.6 ± 4.0 kg/m 2 , and 34.9 ± 6.2 kg/ m 2 in GDM (p < 0.001). Whereas LDL-LpPLA 2 was neither associated with maternal pre-pregnancy BMI ( Fig. 6a) nor BMI at delivery (Fig. 6b), HDL-LpPLA 2 showed strong positive correlation with pre-pregnancy BMI (Fig. 6c, r = 0.5, p = 0.003) and moderate correlation with BMI at term (Fig. 6d, r = 0.4, p = 0.04). Overweight mothers are recommended to gain less weight during pregnancy, so gestational weight gain in GDM mothers was smaller compared to controls (average 9.3 ± 9.6 kg in GDM vs. 15.9 ± 9.7 kg in controls, p = 0.05) and as a consequence HDL-LpPLA 2 activity was inversely correlated with maternal gestational weight gain (Fig. 6e, r = −0.35, p = 0.05).
LpPLA 2 is inversely associated with surrogate markers of oxidative stress in placenta and foetal plasma. To test if LpPLA 2 action might be relevant in situations of placental and foetal oxidative stress, we measured surrogate markers of oxidative stress in placenta and cord blood plasma. In GDM placentae, LpPLA 2 protein was more abundant than in control placentae (Fig. 7a). This was paralleled by lower levels of oxPL-modified proteins (detected by the E06-oxPC antibody) in GDM placental tissue (Fig. 7b). LpPLA 2 is a highly N-glycosylated enzyme, producing multiple bands in Western Blot, ranging in size from 42-51 kDa, depending on the degree of glycosylated residues present 39 . Interestingly, in GDM tissue we detected more than one band, as opposed to control tissue, where only one defined band was detectable. The possible functional implications of the N-glycosylation is not fully clear, but may contribute to directing LpPLA 2 binding towards HDL instead of LDL 39 . We further investigated the glycosylation of LpPLA 2 using an enzymatic de-glycosylation kit, but could not find differences in the glycosylation of LpPLA 2 in GDM placenta and on GDM-HDL with respect to the shift in molecular weight or retention factor after de-glycosylation (data not shown).
Furthermore, we measured the concentrations of thiobarbituric acid reactive substances (TBARS) as index of lipid peroxidation in foetal plasma. They were significantly lower in GDM than in controls (Fig. 7d, p < 0.0001). TBARS levels were inversely correlated with HDL-associated LpPLA 2 activity (Fig. 7e, r = −0.5, p = 0.01) thus confirming our hypothesis that LpPLA 2 might act against oxidative stress in placenta and foetus.
HDL-associated LpPLA 2 contributes to anti-inflammatory, anti-oxidative functionalities of HDL. The influence of HDL-associated LpPLA 2 on foetal HDL function was assessed using cell-based and cell-free functional assays. First, real-time paracellular passage between the endothelial cells (the barrier function) was monitored by using electric cell-substrate impedance sensing (ECIS, Fig. 8a). Placental arterial endothelial cells were exposed to oxPL alone, oxPL that had been pre-incubated with native foetal HDL, or foetal HDL treated with Darapladib to inhibit LpPLA 2 activity. Additionally, to exclude possible off-target effects, cells were also treated with inhibitor alone. As negative control, untreated cells grown in endothelial basal medium (EBM) only were included. Figure 8a shows one representative ECIS experiment. After recording of a 5 h baseline, the respective compounds were added. Both set ups containing HDL protected the cellular barrier integrity compared to oxPL alone (p = 0.005 for oxPL-HDL-DMSO and p < 0.001 for oxPL-HDL-Darapladib at 25 h). However, only when LpPLA 2 was active on HDL, the impedance increased compared to EBM control (p = 0.04 at 25 h). Inhibition of LpPLA 2 caused a decrease in barrier function compared to EBM control (p < 0.001 at 25 h). No significant off-target effects of Darapladib were observed. Although a total of five experiments showed comparable results (Supplemental Fig. 1), considerable inter-individual variation between the five primary endothelial cell isolations (variance in baseline impedance, more immediate vs. more prolonged response to compound addition, etc.) precluded statistically significant results.
In another cell-based assay measuring lipid peroxidation, we compared native foetal HDL with inhibitor-treated HDL. Placental endothelial cells were grown on slides suited for fluorescence microscopy and incubated with either native HDL, inhibitor-treated HDL, oxPL alone (positive control) or BSA (negative control). Also, native GDM-HDL and inhibitor-treated GDM-HDL were used for incubation. ECs were subsequently exposed to linoleamide alkyne (LAA), which incorporates into cell membranes. Upon lipid peroxidation, it leads to the formation of aldehyde-protein-adducts, which can be detected and visualised by azide-modified fluorophores. This assay allows assessing lipid peroxides within cells. We found that oxPL alone leads to high fluorescence signals reflecting a high degree of lipid peroxidation in the cells, thus corroborating the ECIS results. In addition, oxPL exposure induced a change in morphology, likely due to an increase in apoptosis (Fig. 8b, upper right corner). Between BSA treated (Fig. 8b, lower right corner) and native HDL treated cells (Fig. 8b, upper left corner) no apparent morphological difference was observed, and fluorescence (=lipid peroxidation) in both treatments was much lower than in the oxPL positive control. Finally, in cells exposed to Darapladib-treated HDL (Fig. 8b, lower right corner), there was a higher degree of lipid peroxidation detectable compared to native HDL or BSA control, but clearly lower than in the positive control. For GDM-HDL (upper centre), lipid peroxidation seemed a bit more compared to Control HDL and BSA negative control; however, inhibition of LpPLA 2 by Darapladib (lower centre), further exacerbated cellular peroxidation, resulting in a similar effect as in Control HDL treated cells.

Discussion
Here we report that GDM-and obesity-associated metabolic and inflammatory derangements in pregnant mothers alter LpPLA 2 activity with functional consequences in the placenta and foetus. Our results suggest that the increased release of LpPLA 2 from placental HBCs is not caused by hyperglycaemia, but rather by hyperinsulinemia and inflammatory cytokines.
As placental HBCs are of foetal origin 41 we could corroborate in vitro findings with related in vivo data of foetal plasma parameters. This study design combining cellular in vitro with in vivo studies in human is a major strength, and to the best of our knowledge also unique in addressing why and how LpPLA 2 is altered in the foetus and foeto-placental tissues. Provided that our findings at the end of pregnancy reflect to some extent the later period of pregnancy, LpPLA 2 activity may have beneficial, protective effects for the developing placenta and foetus.
We found that LpPLA 2 activity was increased in foeto-placental macrophages from GDM pregnancies, as well as on foetal lipoproteins from an obese GDM cohort. Specifically, HDL-associated LpPLA 2 activity was increased, which may have functional implications for the HDL particles. Previous studies suggested that LpPLA 2 might exert pro-and anti-inflammatory activities dependent on its lipoprotein carrier. LDL-associated LpPLA 2 exerts pro-inflammatory actions, whereas HDL-associated LpPLA 2 exerts anti-inflammatory actions 6,42,43 . We identified HDL as the major carrier of foetal LpPLA 2 activity which is in line with other observations demonstrating HDL as the main cholesterol carrying lipoprotein sub-fraction in cord blood 44 , whereas it has been shown in adults that LpPLA 2 activity correlates strongly with LDL-C and ApoB.
It has to be noted, that a recent study failed to observe increased HDL-associated LpPLA 2 activity in GDM neonates, nor did the authors find HDL as the major carrier of LpPLA 2 activity 36 . Differences in the two study designs and methodology may explain this. First, clinical and metabolic parameters of investigated cohorts are substantially different. While Gao and colleagues included only lean GDM mothers, our cohort was biased by a pre-pregnant BMI > 25 kg/m 2 . Second, the investigated foetal cord blood was exclusively collected from the umbilical vein. The umbilical vein transports oxygenated, nutrient-rich blood to the foetal heart. In return, the two umbilical arteries contain deoxygenated and nutrient-poor blood which is transported back to the placenta. Our study, although smaller, used pooled cord blood from vein and arteries and therefore may reflect the systemic foetal environment more closely. Third, LpPLA 2 activity was assayed by distinct methods; whereas we used a commercially available standardized colorimetric kit, their study used an in-house method employing trichloroacetic acid precipitation in combination with a radioactive tracer. Finally, the LDL-LpPLA 2 activity was obtained by subtracting the HDL-LpPLA 2 activity from total plasma activity, whereas our LpPLA 2 activity was determined in each isolated lipoprotein fraction distinctly.
Overweight and even more so obesity are major pre-disposing factors for the development of diabetes throughout pregnancy 45,46 . Our cohort used for HDL isolation could not be matched for maternal BMI; mothers in the GDM group had significantly higher BMI, both before pregnancy and at time of delivery. Other studies have faced the same problem 36,47 .
We attempted to answer the question if maternal BMI affects LpPLA 2 , and performed correlation analysis of foetal LpPLA 2 activity with maternal BMI. For LDL-associated LpPLA 2 activity, making up only the minor  . Anti-oxidative potential of HDL-LpPLA 2 on neonatal placental endothelium. (a) Barrier function assay of placental endothelial cells exposed to (i) oxidized phospholipid mix (oxPL, 15 ug/ml, green), (ii) oxPL plus neonatal HDL (15 ug/ml + 200 ug/ml, dark blue) and (iii) oxPL plus neonatal HDL in the presence of Darapladip (15 ug/ml + 200 ug/ml + 150 nM, resp.; turquoise). Darapladib alone (light blue) did not show any off-target effects compared to endothelial basal medium (EBM, red). One out of five representative experiments is shown. *p < 0.05 compared to EBM, †p < 0.005 compared to oxPL. (b) Anti-oxidative effects of HDL-LpPLA 2 in a cell-based assay of lipid peroxidation (ClickIT ™ assay). Lipid peroxidation was visualized based on linoleamid alkyl and fluorophores on a laser scanning microscope using defined settings for all pictures taken to make them comparable. (c) Cell free assay of Control HDL (black) and GDM-HDL (grey) anti-oxidative potential based on the oxidation of 123-dihydrorhodamine (DHR). In addition to LpPLA 2 -inhibition by Darapladib, also Paraoxanase-1 (PON-1) was inhibited by 2-hydroxyquinolone (2-OHQ). One-way ANOVA was used to test for significance. portion of foetal total plasma LpPLA 2 , no correlations were found. We did, however, find a strong correlation of HDL-associated LpPLA 2 activity with maternal pre-pregnancy BMI. Correlation with BMI at time of delivery was more moderate, which can be explained by differential weight gain in the control and GDM group. Also, HDL-LpPLA 2 activity was inversely correlated with maternal weight gain. Additionally, in samples from a different obese, non-GDM cohort, we also found a correlation between LpPLA 2 protein expression in placenta and maternal BMI (Schliefsteiner et al., unpublished data). This corroborates the notion that our study -as well as others -might be flawed by a considerable (co-)effect of maternal obesity along with maternal diabetes.
Notably, several studies have shown that maternal obesity and GDM do not only impose a risk on maternal health but also affect long term health of their children. Neonates from GDM pregnancies are prone to be macrosomic, have increased fat depositions 48 and an increased risk for development of Type 2 diabetes later in life 49 . Several studies have investigated LpPLA 2 levels in obese children and adolescents and all found correlations between LpPLA 2 activity and BMI 50,51 as well as fat mass and waist circumference 52 . Other studies also showed that LpPLA 2 activity is a predictor of Type 2 diabetes development 19 . It is tempting to speculate if increased LpPLA 2 in GDM children might be causally related to development of obesity and diabetes later on, or if it could at least serve as a risk-predictor. However, more focused prospective studies regarding metabolic variations and follow-up of patients with and without risk factors are needed in order to clarify the role of LpPLA 2 in these settings.
We also sought to identify in vitro molecular regulators of LpPLA 2 activity on HBCs, which may contribute to the GDM associated changes. Elevated glucose and insulin levels in the cord blood are key features of a GDM pregnancy 15,47 . Foetal insulin is able to regulate placental gene expression 53 and HBCs are rich in insulin-receptors 54 . Glucose and insulin were therefore obvious candidates for regulating LpPLA 2 activity. Whereas glucose did not have any effect, insulin increased LpPLA 2 activity. Leptin as a major adipogenic hormone also increased LpPLA 2 activity in HBCs and foetal plasma leptin levels were significantly elevated in the obese GDM group. Our findings at the foeto-placental axis are consistent with the role of insulin and leptin as adiposity signals, which are both positively correlated with body weight in general and adipose tissue mass in particular 55 , further indicating that not only GDM but also maternal BMI influences perinatal outcome. Both hormones regulate LpPLA 2 and may thus account for chronic low-grade inflammation in the GDM placenta 22 as well as (placental) endothelial dysfunction 56 . Pro-inflammatory TNFα, and to a lesser degree also endothelial adhesion molecules ICAM-1 and VCAM-1, also induced LpPLA 2 activity in HBCs. TNFα could not be quantified in cord blood plasma, but both ICAM-1 and VCAM-1 levels were increased in the obese GDM group. ICAM-1 and VCAM-1 have been established as circulating markers of endothelial activation 57,58 and increased plasma levels may also contribute to increased HDL-LpPLA 2 activity in GDM neonates.
Importantly, anti-inflammatory cytokines such as IL-4 and IL-13 significantly decreased LpPLA 2 activity from HBCs, suggesting that LpPLA 2 expression is responsive to the macrophage micro-environment, which would also explain up-and down-regulation of LpPLA 2 during and after acute-phase response 59 . Similar to our observations, others found that peripheral blood monocytes upon stimulation with IL-4 secrete significantly less LpPLA 2 compared to cells stimulated with M-CSF (macrophage colony stimulating factor) 60 .
Unlike clinical studies in the past, which linked LpPLA 2 mass and/or activity with clinical parameters in an associative manner, we aimed to identify molecules causal for regulating LpPLA 2 activity in vitro. However, we did not investigate the signal transduction mechanisms by which these regulators orchestrate LpPLA 2 activity. One might consider this a limitation of our study. From the limited amount of studies on signalling pathways activating LpPLA 2 61,62 and current knowledge about signal transduction pathways activated by pro-inflammatory cytokines 63 and insulin and leptin 55 , we assume that the regulators of LpPLA 2 activity identified in our study act through mechanisms dependent on p38 and PI3K within the MAPK pathway.
The functional consequences of altered LpPLA 2 activity in placenta and foetus may be important for maintaining stress levels low at the foeto-placental interface. GDM and obesity are associated with oxidative stress in the placenta, which is paralleled by higher antioxidant levels 64 . The inverse relationship between LpPLA 2 protein and oxidized phospholipids in placental tissue suggests that lower oxPL levels could be a result of increased LpPLA 2 action, which is in line with an anti-oxidative role of LpPLA 2 . In cord blood, TBARS were measured as a surrogate marker of oxidative stress and their levels inversely correlated with HDL-LpPLA 2 activity. Collectively, these results suggest a local and specifically tight regulation of anti-oxidative defence mechanisms within the human placenta and foetus. Of note, different from our findings, increased TBARS or malondialdehyd levels in cord blood of GDM pregnancies were reported by others 65,66 . However, anti-oxidant enzymes (e.g. superoxide dismutase) were also increased in GDM 65 , so total anti-oxidative potential was unchanged in these studies.
LpPLA 2 circulating on LDL and HDL is in constant contact with macro-and microvascular endothelium. We therefore considered that LpPLA 2 might have an effect on endothelial function. Using electrical cell substrate impedance sensing (ECIS), we demonstrated the positive effect of foetal HDL-LpPLA 2 on placental endothelial barrier function and that this effect was abolished when HDL-LpPLA 2 activity was inhibited. One might speculate that elevated HDL-LpPLA 2 in GDM could be a protective counter-mechanism against the endothelial dysfunction commonly observed in GDM 58 . By pre-incubating HDL and LpPLA 2 -inhibited HDL with oxPL, we could also describe an anti-oxidative effect of LpPLA 2 . This points towards a role of LpPLA 2 as an anti-oxidant, and in the regulation of vascular permeability. Furthermore, a cell free assay demonstrated that HDL-LpPLA 2 , specifically on GDM-HDL, contributes to the total anti-oxidative potential of the HDL particle.
In clinical trials, the specific LpPLA 2 -inhibitor Darapladib offered no benefit for patients 67 . Nevertheless, physicians identified low LpPLA 2 activity as goal to improve patient health. Our in vitro study in human macrophages points towards other treatment options than inhibitors to achieve lower plasma LpPLA 2 activity, such as lifestyle interventions, i.e. diet and exercise 68 , to lower insulin and leptin as well as LDL levels; also lowering LDL levels by statin therapy will reduce LpPLA 2 69,70 . In addition, lowering pro-inflammatory cytokines or raising anti-inflammatory cytokines by pharmacological means could improve patient outcome, not only by regulating LpPLA 2 activity but by addressing inflammation more holistically. Supplementary Tables 1   and 2. All subjects gave written informed consent. The study design had been approved by the ethics board of the Medical University of Graz (24-529 ex 11/12). All women underwent an oral glucose tolerance test (OGTT) between 24 and 28 weeks of gestation. Gestational diabetes was diagnosed according to the guidelines of the American Diabetes Association 71 . All individuals in the GDM group were treated only by diet and lifestyle modifications; none of the patients administered insulin. All experimental methods were performed in accordance with the respective approved study protocols.

Study population. Clinical characteristics for the GDM study are summarised in
Sample collection and storage. For placenta tissue collection (N = 12/group), the placenta was divided into quadrants and a piece of 5-7 mm diameter was punched from each quadrant, reaching from the maternal to the foetal side. The punched piece was cut in half, so that one half contained the chorionic plate (=foetal side, used exclusively in this study) and the other the basal plate (=maternal side). Tissue was either snap frozen in liquid nitrogen and stored at −80 °C until RNA or protein isolation, or was formalin fixed and embedded into paraffin for immune histological examination.
For collection of neonatal cord blood plasma (N = 21/group), cord blood was obtained as mixed blood (from umbilical arteries and vein) directly after delivery of the placenta and cutting of the umbilical cord. Blood was collected in EDTA plasma tubes and centrifuged for 15 min at 2000× g at 4 °C. Plasma was carefully aliquoted and aliquots were stored at −80 °C until further use for ELISAs or lipoprotein isolation. Adult EDTA plasma for isolation of adult LDL and HDL was obtained from healthy female donors (N = 4) at child-bearing age from the blood bank at the General Hospital of Graz and was stored and processed like foetal plasma.
For Hofbauer cell isolation, the placenta was obtained and used as described below.
Hofbauer cell (HBC) isolation. Both placentae from caesarean section and vaginal delivery were used within 20 min after delivery (N = 13/Control group, N = 5/GDM group). Maternal membranes (decidua) were removed to avoid contamination with decidual HBCss. Tissue was dissected, finely minced and stored overnight in PBS. The next day, 60-100 g tissue was digested in two steps, employing trypsin, and thereafter collagenase A. After digestion, cells were applied onto a percoll gradient to separate cell populations. HBCs appear as band between the 30-35% percoll layers. They were aspirated from the gradient and purified by negative immune selection using antibodies against CD10 and EGFR. Cells were counted and plated in HBCs medium (MaM, ScienCell) supplemented with 5% lipoprotein-deficient serum (LPDS) at a density of 1 × 10 6 cells/ml. After five days, quality of the primary cells was controlled by immune cytochemistry, using CD163 as a marker for HBCs.

Time-course experiments.
HBCs isolated from control and diabetic placentae (N = 5/group) were cultured up to 6 days and the secreted LpPLA 2 activity in supernatant was monitored every 24 h. Additionally, control HBCs (N = 4) were also treated for 1, 3, or 5 days with a specific inhibitor of LpPLA 2 activity, Darapladib (Medchem Express), at a final concentration of 250 nM, and LpPLA 2 activity was measured. All experiments were carried out using three technical replicates per condition.
Exposure of HBCs to diabetic environment. For all treatments HBCs were seeded at a density of 1 × 10 6 cells/ml in 6-well plates; all treatments were performed in triplicates. An untreated control was included in every experiment. Glucose treatment: HBCs (N = 4) were exposed to 5, 15, and 25 mM of D-glucose for 72 hours, glucose was added daily. Equimolar controls with L-glucose were included. Insulin treatment: HBCs (N = 5) were exposed to 5, 10, 20, 30, and 50 nM of insulin daily for 72 h. Cytokine treatments: HBCs were exposed to TNFα (50, 100 and 250 pg/ml; N = 5), IL-4 and IL-13 (200, 600 and 1000 pg/ml, respectively; N = 4), and adhesion molecules ICAM-1 and VCAM-1 (500, 1000 and 3000 pg/ml, respectively; N = 5) each day for 72 hours. After all treatments, supernatants were collected for activity assay; cells were washed twice with 1x HBSS and lysed using RIPA buffer supplemented with proteinase inhibitor cocktail. Cell lysates were incubated for 30 min on ice, centrifuged at 16000 g for 20 min and stored at −20 °C. Protein concentration of lysates was determined using Bichononic acid (BCA) method (Pierce) following the instruction manual.

Placental tissue protein isolation.
For tissue lysates about one gram of Control or GDM placental tissue (N = 12/group) was homogenized in 2 ml of RIPA buffer with proteinase inhibitor cocktail using an ultra-turrax device. Lysates were centrifuged and supernatant was used to measure protein concentration using the BCA method.
Western Blot. 10 µg of total tissue protein were subjected to PAGE on 4-20% Bis acrylamide precast gels and protein was transferred onto nitrocellulose membranes. Membranes were probed against a polyclonal anti-PAF-AH antibody (Cayman Chemical) recognizing specifically the C-terminal region of LpPLA 2 and against oxPL using the E06-oxPC antibody (Avanti Polar Lipids). Anti-rabbit and anti-mouse IgG-HRP, respectively, was used to detect protein, ECL substrate for chemiluminescence and a Biorad LAS-400 camera. Protein signal was normalized against β-actin as loading control. To compare between blots, an internal control prepared from THP-1 macrophages was included in every experiment. Densitometric analysis was performed using DigiDoc1000 software (Alpha Innotech). The anti-PAF-AH antibody yields more than one band due to glycosylation of the enzyme, the E06 antibody yields more than one band because it recognizes PC-modified proteins via a phosphocholine headgroup of oxidized phospholipids. For densitometric analysis, the sum of these bands was considered.
Immune fluorescence staining. Placental tissue sections of 4-6 um thickness were prepared from paraffin-embedded blocks and mounted onto glass slides. Sections were de-paraffinised in xylene and rehydrated in an ethanol dilution series. Antigen retrieval was omitted to not destroy placental villus structure. Sections were blocked using 3% BSA in TBE buffer. Antibodies were mixed in Antibody Diluent and Background Reducing Component (both Dako). Sections were incubated with primary antibody over-night and secondary antibody in the dark for 2 h, respectively and washed several times in between steps. Coverslips were mounted onto glass slides using Prolong Gold Antifade Reagent with DAPI (LifeTechnologies), sealed and stored in the dark at 4 °C. Pictures of sections were taken with a Zeiss LSM510, AxioVert200M microscope.
Lipoprotein isolation from cord blood. Density of foetal cord blood plasma (8 ml volume) from control and GDM pregnancies (N = 21/group) was adjusted to ρ = 1.24 g/ml using potassium bromide. Plasma was transferred to ultracentrifuge tubes and potassium bromide solution of ρ = 1.006 g/ml was layered on top. Samples were centrifuged at 90.000 g for 4 h at 15 °C in a table top ultra-centrifuge. The LDL layer on the top was collected, the interphase was discarded, and HDL floating in the centre layer of the tube was collected. Lipoproteins were stored at 4 °C, light-protected and under a layer of argon gas to prevent oxidation. Each sample was concentrated to 1.5 ml volume using Vivaspin tubes (MwCo 5 kD, Satorius) and excess potassium bromide was removed using PD10 resin columns. Quality of HDL was assessed measuring total protein and cholesterol, calculating a Protein/ Cholesterol ratio (>2:1 were used).
LpPLA 2 activity assay. Enzymatic activity of LpPLA 2 in cell culture supernatant collected from HBCs, as well as on isolated foetal lipoproteins, was measured using a commercially available PAF-AH activity kit (Cayman Chemical). The assay was carried out according to the manufacturer's instructions and activity was calculated as suggested by the manufacturer and expressed as nmol/ml/min. Enzyme linked immunosorbent assays. All ELISAs, IL-4 and ICAM-1 (both Peprotech), IL-13 and VCAM-1 (both RnD Systems), Leptin (Millipore), for foetal cord blood plasma were carried out according to manufacturer's instructions. Insulin levels were measured by an automated ELISA (Advia Centaur, Siemens).
Thiobarbituretic acid reactive substances assay. Oxidative status in foetal plasma was measured by surrogate markers of lipid peroxidation. Thiobarbituretic acid reactive substances (TBARS) assay kit (Cayman Chemical), was carried out according to the manufacturer's instructions.
ClickIT Lipid Peroxidation assay. HDL anti-oxidative function by LpPLA 2 was assessed in a cell based assay (ClickIT ™ Lipid Peroxidation Detection Kit, Life Technologies). Placental endothelial cells were isolated as previously described 72 and grown on chamber slides to 80% confluence. Foetal HDL was used in its native form or pre-incubated with 250 nM Darapladib. Cells were exposed to HDL (200 ug/ml) with or without inhibitor for 2 h at 37 °C. Cells incubated with oxPL (15 ug/ml) or BSA (200 ug/ml) served as positive and negative control, respectively. Subsequently, cells were incubated with LAA for 2 h. For cell fixation, permabilisation and visualisation by Alexa488 fluorophores, the kit was carried out according to manufacturer's instructions. Laser scanning microscopy (LSM510 AxioVert200M, Zeiss) was used to detect peroxidation-induced fluorescence in the cells.
Electrical cell-substrate impedance sensing (ECIS). Human placental arterial endothelial cells were plated in Endothelial Basal Medium (EBM, Lonza) onto chamber slides suited for barrier function measurement (8W10E + PET, ibidi). On an ECIS Z instrument (Applied BioPhysics), baseline was recorded at 4000 Hz for 4 to 6 hours. A mixture of oxPL (15 ug/ml) plus foetal HDL (200 ug/ml) plus either Darapladib (250 nM) or DMSO (vehicle control) -which had been pre-incubated for 1 h at 37 °C -was added to cells. Additionally, cells were exposed to oxPL only, Darapladib only, and EBM only (untreated control). Impedance was monitored over 40 h at 4000 Hz. Analysis of experiments was done using ECIS software (Applied BioPhysics). Statistical analysis. All statistics were calculated and graphs prepared using GraphPad Prism v7.0. Where applicable, either two-tailed t-test, one-way or two-way ANOVA were performed, depending on whether two or more groups were compared, respectively. Normal-distribution was tested by Shapiro-Wilks test. If normal distribution failed, non-parametric tests were performed, either Mann-Whitney rank sum or ANOVA on ranks with Dunns test post hoc testing. All data are presented as mean ± SD in tables, bar charts and dot plots. Spearman Correlation was performed for correlation analysis. P-values < 0.05 were considered statistically significant.