Placentas were obtained within 20 min after both cesarean sections and vaginal deliveries. Patient characteristics are shown in Table 1. The study was approved by the institutional ethics committee of the Medical University of Graz (27-265 ex 14/15) and all mothers gave written informed consent. Placentas from healthy singleton pregnancies were used as controls. GDM macrophages were isolated from singleton pregnancies when GDM was diagnosed by an oral glucose tolerance test within the second trimester of pregnancy according to ADA criteria (35). Although the study groups were matched for maternal BMI, a predisposing factor for GDM and often considered a confounder in GDM studies (36), groups could not be matched for gestational age (GA, see Table 1). It is common obstetric practice to deliver GDM children a bit premature to avoid complications, such as macrosomia and shoulder dystortia (37–39). However, as placental weight and fetal ponderal index did not differ significantly, one might consider that placenta and children were equally well developed in both groups. The number of HBCs in placenta steadily declines from first trimester to full term, but polarization does not change intensely during this time (40); we, therefore, considered the apparent difference in GA negligible.

For isolation, the upper layer of maternal membranes was removed from the placenta to avoid contamination with decidual macrophages. The tissue was dissected, finely minced, and stored overnight in PBS buffer. Between 60 and 100 g tissue was used for isolation. The next day, tissue was digested in two steps, first employing trypsin (10×, Gibco) and DNase I (Roche), and second using Collagenase A (Roche) and DNase I. After digestion, cells were applied onto a Percoll gradient (Gibco) and centrifuged at 2,030 × g for 30 min, without brake. Macrophages appeared as band between the 30 and 35% Percoll layers. Cells were aspirated from the gradient and negative immune selection with magnetic beads (Dynabeads anti-goat IgG, Invitrogen) and antibodies against CD10 (abcam) and EGFR (NeoMarkers) was used to further purify the cells. After immune selection, cells were counted and seeded in macrophage medium [macrophage medium (MaM), ScienCell] supplemented with 5% FCS and macrophage growth supplements (ScienCell) at a density of 1 × 10⁶ cells/ml. Cells were cultivated at 21% oxygen, 37°C; quality control was carried out by loading HBCs with Ac-Dil-LDL after 2 days and monitoring fluorescence in the live cells; and immune cytochemistry (ICC) after 7 days on fixed cells.

Hofbauer cells isolated from control placentas were plated at a density of 1 × 10⁶ cells/ml in 6-well culture dishes (3 ml total volume). On day 3 post-isolation, cells were serum-starved for 12 h and thereafter switched to complete MaM containing either 25 mM d-glucose (Sigma) to mimic maternal and fetal hyperglycemia, 10 nM Insulin (Calbiochem) to mimic fetal hyperinsulinemia in response to maternal GDM, or a combination of both. Equimolar l-glucose (Sigma) was used as osmatic control, an untreated control grown in MaM only was included. Cells were cultivated for 72 h, receiving treatment every 24 h. Cells were harvested and lysed using RIPA buffer. Protein content was measured using bichinonic acid method (BCA assay, Pierce). 7.5 µg of protein was subjected to electrophoresis (4–20% Mini-Protean TGX gels, Biorad) and blotted onto nitrocellulose membranes (Trans-Blot Turbo System, Biorad). Membranes were incubated with antibodies against CD163 (Thermo Scientific), CD86 and CD209 (both NovusBio) and β-Actin as loading control (abcam); secondary antibodies against mouse and rabbit IgG were from Biorad. Detection was carried out using West Femto ECL substrate (Pierce) on a ChemiDoc XRS system (Biorad).

Secretion of pro- and anti-inflammatory cytokines, chemokines, and growth factors from placental macrophages was assessed using a multiplex ELISA-on-beads approach. Control (N = 5) and diabetic macrophages (N = 6) were cultivated in MaM up to 4 days, a time point taken every 24 h. Due to the low levels of cytokines in the supernatant, samples were concentrated fourfold using Amicon Ultra centrifugal units (3kD MwCo, Millipore), to be in the multiplex assay range. A panel of 23 factors of interest (see Table 2) was obtained as customized panel (Aimplex, YSL Bioprocess Development Co.) and the multiplex experiment was carried out according to the manufacturer’s instructions. Bead signals were quantified using a FACS Calibur instrument (Becton Dickinson) and FlowCytomixPro software (eBioscience) was used for calculation of standard curves and sample concentrations. For data normalization, total protein of the supernatant was assessed using BCA Kit and cytokine levels where normalized to total protein levels accounting for eventually divergent concentration of volume. ELISAs used for validation were carried out according to assay manuals. ELISAs used were IL-1β and IL-13 (both eBioscience), IL-1RA, IL-6, IL-8, IL-12, macrophage colony stimulating factor (M-CSF), TNFα, intracellular adhesion molecule 1 (ICAM-1), and vascular endothelial growth factor (VEGF) (all Peprotech) and TGFβ (eBioscience).

Macrophages cultured for 7 days were carefully harvested using Dispase enzyme solution and gentle scraping. Cells were counted and a minimum of 3 × 10⁵ cells were used for each FACS staining. Fc receptors were blocked in 3% BSA in 1× HBSS (Gibco) for 10 min. As this study focused on membrane proteins, permeabilization and fixation of cells were omitted. Instead, cells were re-suspended in PBS and incubated with the respective antibodies for 30 min. Antibodies used for FACS were αCD163-APC and αCD90-PE (both Biolegend), CD11b-V450, CD11c-PE, CD40-FITC, CD45-PE, CD209-Cy5.5, CD80-V450, CD86-V450, CD146-PE, and CD206-FITC (all BD Pharmingen). Unstained cells were used as control, as well as an IgG control. Cells were washed twice after antibody incubation and re-suspended in PBS for counting. Cell sorting was performed on a LSR-II instrument (BD Bioscience) using FACSDiva8 acquisition and analysis software. Cells were gated in three steps; first, cells were discriminated by size employing forward and side scatter (FSC and SSC, respectively). Second, doublets were removed by pulse-geometry gating the area and height of FSC. Lastly, the fluorescence signal of cells positive for respective markers was gated directly, plotting it against the SSC area.

Immune cytochemistry was carried out on 4-well glass chamber slides (Lab-Tek) using isotype controls and reagents from Dako. Briefly, cells were washed twice with HBSS and fixed with ice cold acetone. Antibodies were diluted to working concentration using Dako antibody diluent and cells were incubated with primary antibodies for 30 min. Cells were subsequently incubated with Dako antibody enhancer for 10 min. After a washing step, cells were incubated with large HRP polymer solution and AEC chromogen solution, for 20 and 10 min, respectively. Cells were counterstained with haemalaun solution and mounted with glycerin. Antibodies used were αCD163 (Thermo Scientific), αCD90 (Dianova), αCD68 (Dako), αCD11b (abcam), αCD11c (abcam), α CD14 (BD Pharmingen), αCD209 (Novus), and α mouse IgG (Dako) as isotype control. For quantification, pictures from four different HBC isolations per group were counted and positive cells were expressed as percentage relative to the total number of cells on each picture. To avoid investigator bias, all countings were done by one researcher alone.

Tissue sections of 5 µm thickness were cut from paraffin-embedded placental tissue (taken from the fetal side and a central region of the placental disk) and mounted onto glass slides. After eliminating the paraffin using xylene and rehydrating the tissue in an ethanol dilution series, antigen retrieval was carried out using 1 mM EDTA. For immune histochemical staining, the UltraVision LP detection system by Dako was employed: tissue was incubated for 15 min in Hydrogen Peroxide Block and washed four times in TBE buffer, followed by 5 min incubation with UltraV Block. Samples were subsequently incubated with the primary antibody, followed by incubation with primary antibody enhancer, for 30 and 10 min, respectively. Antibodies used were αCD163 (Pierce), αCD68 (Dako), αCD40 (Abcam), αCD11b (Abcam), and αCD11c (Abcam). Samples were washed again in TBE and HRP polymer was applied for 30 min. After another washing step, samples were incubated with AEC chromogen solution for 10 min. Samples were counterstained with haemalaun and mounted using glycerin. Spleen and lymph node tissue, embedded in paraffin, were prepared and stained in the same manner and used as positive control, as these tissues are supposed to express all investigated markers according to the Human Protein Atlas tissue bank. Photographs were taken using Zeiss AxioVision software v8.0 on an Olympus BX53 light microscope with an AxioCam MRc5 (Zeiss). For quantification, positive cells on the pictures were counted by one single histologist with the help of the Visiopharm Stereotopix™ software platform to avoid intra-individual bias. Pictures from four placentas per group were taken, and seven to ten different sites from each placenta were used for quantification. Positive cells were expressed relative to CD163, which proved to be the most consistent marker of HBCs.

Human placental arterial endothelial cells (pAECs) were isolated as described before (41). HBCs isolated from control and GDM placentas were cultivated for 6 days in MaM (1 × 10⁶ cells/ml medium), then supernatant was collected. To obtain conditioned medium (CM), equal parts of three different HBC isolations were mixed1:1 with endothelial basal medium (EBM, Lonza) without growth supplements and only 2% FCS. pAECs were exposed to macrophage CM from control and GDM HBCs to unconditioned MaM as negative control l. In addition, a positive control was made from EBM containing a cytokine mix of TNFα (1.5 ng/ml), IL-1β (0.2 ng/ml), and IL-6 (1 ng/ml). After 48 h treatment (37°C, 5% CO₂, 12% O₂), pAEC were harvested and lysed in cell lysis buffer (Raybiotech). A custom Quantibody ELISA array (Raybiotech) was used to detect ICAM-1, vascular adhesion molecule 1 (VCAM-1), P-selectin, and E-selectin, as well as vascular endothelial (VE)-cadherin and M-CSF in parallel in the lysates. The array way prepared according to the manufacturer’s instructions and fluorescence scanning was performed on an Agilent Microarray scanner G2565CA. Data extraction was done using GenePix Pro v6.0 and analysis was carried out using a software tool for MS Excel provided by the manufacturer.

GraphPad Prism 7 was used for all statistical calculations and plotting of graphs. Shapiro–Wilk test was used to test for normal distribution. For comparison of two groups, t-test was used. One-way ANOVA was performed, if more than two groups/time points were compared. In one-way ANOVA, Sidaks post hoc test was used to account for multiple comparisons. Kruskal–Wallis test with Dunns test for multiple comparisons was used instead of ANOVA if normality test failed. p-Values below 0.05 were considered statistically significant. For post hoc power analysis, G*Power 3.1 software (available from the University of Duesseldorf, Germany) was used.

Immunohistochemistry was used to characterize macrophages in placental tissue. Markers for M1 and M2 polarization were investigated in serial tissue sections. Although suggested as a pan-macrophage marker, CD68 staining was limited in placenta (Figure 1A, top panel), and CD163 (Figure 1A, second from top) appears to be more abundant and a more reliable HBC marker. On serial sections of placental tissue, we observed regions were CD68 staining was absent, but clearly a staining for CD163 was present; examples of such regions are indicated by the black arrow heads in Figure 1A. Quantification of CD68 did not show any difference in number of positive cells between control and GDM placenta in absolute quantification and also in quantification relative to CD163 (data not shown). Specifically, CD163 positive cells were found in the villous stroma of control and GDM placentas (Figure 1A, second from top). With respect to distribution, independent of the presence of GDM, CD163 positive cells were present in stem villi, intermediate and terminal villi, and positioned with in the stroma, but distant from vessel walls and vessel-surrounding connective tissue. Counting the numbers of CD163 and expressing them relative to control (i.e., control = 100%), we found similar numbers of CD163 positive cells in control and GDM tissue (Figure 1B). CD163 was used in quantification of IHC to account for the overall number of HBCs, and quantification of M2 markers CD209 and CD206 (Figure 1A, center panels) was performed relative to CD163. We observed a significant increase of CD209 positive cells in GDM placenta (p < 0.001; Figure 1C), and a slight but non-significant increase in CD206. However, calculating the ratio of CD206 to CD209 positive cells, we observed a decreased number of CD206 positive cells in GDM placenta (Figure 1D).

As M1 markers, we used CD80 and CD86; spleen and lymph node tissue served as positive control. In control placenta (Figure 2A, left), no specific staining against CD80 was observed. Also almost no CD86 positive cells were present (Figure 2A, right). In GDM placenta, a very limited number of cells showed CD80 (Figure 2B, left) and CD86 (Figure 2B, right) staining. In spleen (Figure 2C) and lymph node (Figure 2D), which were used as positive control tissues, a much higher number of positive cells were observed.

In addition, we also tested the M1 marker CD40 (Figure S1A in Supplementary Material) and CD11b (Figure S1B in Supplementary Material) in conjunction with CD11c (Figure S1C in Supplementary Material). No specific staining against these M1 markers was observed in control placenta (left panel) as well as GDM placenta (second from left). In the positive controls, spleen and lymph node, staining against these markers was observed.

In line with the slight differences in macrophage marker expression between normal and GDM placentas, also total macrophage cell number was not altered by GDM. Previous data suggest that inflammatory disease of the placenta leads to an infiltration of tissue with HBCs. This has been stated not only for villitis of unknown etiology but also for GDM (42, 43). Therefore, we calculated the yield of macrophages relative to the wet weight of tissue used for isolation and found that total macrophage cell number between control and diabetic isolations did not differ (Table 2). One might also take total placental weight before isolation into account; however, placental weight was not significantly different between control and GDM subjects (Table 1).

The purity of isolated HBCs was confirmed by positive staining for CD163 and CD68 and negative staining for CD90 (fibroblast marker) in immunohistochemistry. CD14 and CD209 were used as markers for M2 polarization, and CD11b together with CD11c as markers for M1 macrophages. We found that HBCs, independent of gestational diabetes, expressed M2 surface markers CD163, CD14, and CD209 (Figure 3A/upper panel: control HBC, Figure 3B/lower panel: diabetic HBC). In addition, HBCs expressed CD11b and also CD11c, which are associated with a pro-inflammatory M1 phenotype and the pan-macrophage marker CD68. ICC was also quantified, counting positive cells relative to total cells (Figure 3C); no significant changes for either M1 or M2 marker expression between control and GDM HBCs were observed, although CD209 appeared slightly increased in GDM. No relevant contamination with fibroblasts (CD90 staining) was observed.

For a more accurate quantification, the presence of surface markers was also assessed on isolated primary HBCs by flow cytometry. FACS revealed that CD163—the most prominent resident macrophage marker—was evenly distributed between control and diabetic HBC (Figures 4A,E; Table 3). Other scavenger receptors, such as M2 markers CD206 (Figures 4B,F) and CD209 (Figures 4C,G), were indeed significantly increased on HBCs isolated from diabetic placentae (Table 3; p = 0.03 and p < 0.001, respectively). CD11b and CD11c were found both on control and diabetic HBC and did not differ in the percentage of their population (Table 3). Also the ratio of CD11b positive to CD11c positive cells, which was proposed as a measure of a shift toward M1 macrophage populations (44), was unchanged. Moreover, in GDM-HBC a higher percentage of cells was positive for M1 markers CD80 and CD40, but these increases did not reach significance (Table 3); however, the M1 marker CD86 (Figures 4D,H) showed a trend toward increased levels in GDM (p = 0.08; Table 3).

Furthermore, we investigated whether hyperglycemia or hyperinsulinemia induce or alter HBCs polarization in vitro. HBCs isolated after normal pregnancies were exposed to 25 mM d-glucose, 10 nM insulin, or a combination of both for 72 h. l-Glucose was used for control conditions. Western Blot analysis revealed that none of the treatments altered protein levels of CD163, CD209, and CD86 (Figures 5A–C).

Using Multiplex ELISA-on-bead technology, we assessed the secretion of 23 cytokines, chemokines and growth factors by HBC into the medium after 24, 48, 72, and 96 h. IL-4, IFNγ, IL-1α, and epithelial growth factor were below the detection limit of the assay. The remaining 19 compounds could be quantified and were normalized to total protein of the respective supernatants. The results revealed only slight differences between normal and diabetic HBCs with respect to cytokine secretion (all data summarized in Table S1 in Supplementary Material). Results were further validated by ELISA (Figure 6).

The left panel of Figure 6 shows release of cytokines and chemokines associated with the M1 phenotype from HBC. GM-CSF release was only borderline difference between normal and GDM HBCs as measured by multiplex (Table S1 in Supplementary Material) and ELISA confirmed that there is no different in GM-CSF secretion (Figure 6A). Noteworthy, although GM-CSF is generally associated with the M1 phenotype, some studies have indicated that macrophages GM-CSF aids wound healing and tissue remodeling; therefore GM-CSF has also been linked with the M2b phenotype (45). IL-1β release from GDM-HBC was consistently elevated over all time points (Figure 6B; plus ~25%) compared to control HBC. However, this increase did not reach statistical significance, neither in multiplex (Table S1 in Supplementary Material) nor in ELISA. The same observation was made for IL-6 (Figure 6C; plus ~50%). Also, the release of the classic M1 cytokine TNFα was unchanged between normal and diabetic HBCs, although release from GDM HBC was higher at 48 and 72 h time points (Figure 6D).

Release of cytokines associated with the M2 phenotype is shown in the right panel of Figure 6. Release of TGFβ—a hallmark cytokine of the M2 phenotype—was similar from HBC independent of diabetes. Only at the 24 h time point, TGFβ release tended to be higher (+95%, p = 0.06) in GDM-HBC (Figure 6E). Notably, TGFβ was only assessed by ELISA as binding to the TGFβ epitope requires acidification of samples which precluded the use for the multiplex approach. Soluble ICAM-1 (sICAM-1) antagonizes the effects of ICAM-1/LFA-1 signaling and promotes angiogenesis (46, 47), so it can be considered an M2 marker, although it has to be noted that it is not a classical phenotypic marker of either M1 or M2 macrophages. sICAM-1 was significantly increased in the multiplex at 48 h (Table S1 in Supplementary Material); ELISA showed that although the increase in GDM persisted at all time points, it was non-significant (Figure 6F). Moreover, release of IL-10 from control and GDM HBC was not altered (Figure 6G). Levels of IL-12 were not changed between control and GDM HBC (Table S1 in Supplementary Material) and much lower than levels of IL-10 in multiplex (Table S1 in Supplementary Material), yielding IL-10^hi/IL-12^low macrophages, which are considered anti-inflammatory (44). Levels of IL-12 could not even be validated by ELISA because of limit of detection in conventional ELISA. Finally, VEGF release between control and GDM-HBC was not significantly different (Figure 6H). VEGF is considered a pro-angiogenic factor associated with the tissue remodeling phenotype of M2 macrophages.

In addition to the parameters in Figure 5, also other proteins have been validated by conventional ELISA. Like IL-12, IL-13 validation failed due to the detection limit of the ELISA. IL-8 was validated and unchanged between control and GDM HBC; as IL-8 is an ambiguous cytokine involved in neutrophil attraction (an M1 like feature) but also pro-angiogenic (an M2 like feature) it was not included in Figure 5. Also, secretion of macrophage chemoattractant protein (MCP-1), an M1 marker, and IL-1RA, an M2 marker, was not significantly different between control and GDM HBCs.

Cross talk between tissue macrophages and endothelial cells is a well-established phenomenon, especially in vascular dysfunction and atherosclerosis (48). GDM is known to cause endothelial dysfunction in mother, placenta, and fetus (49, 50). Therefore, we investigated placental endothelial cell activation by HBCs. Macrophage CM from control and GDM HBCs was used to treat primary pAECs. In addition, as a positive control, cells were treated with a cytokine cocktail containing TNFα, IL-1β, and IL-6, which were increased not only in GDM as found in our multiplex approach but also in other studies (51). Unconditioned medium served as negative control. After incubation with CM for 48 h, cells were lysed and lysates were used in a multiplexed, fluorimetric ELISA detecting ICAM-1, VCAM-1, E-selectin, P-selectin, VE-cadherin, and M-CSF.

Intracellular adhesion molecule 1 protein levels in endothelial cells were increased upon incubation with GDM CM when compared to control CM (+12%, p < 0.05). Also, the cytokine mix positive control induced EC activation (Figure 7A). VCAM-1 protein was only induced by the cytokine cocktail (threefold, p < 0.001), but no other treatment (Figure 7B). E-selectin protein levels were not changed by any treatment (Figure 7C). P-selectin, similar to VCAM-1, was only induced by the cytokine mix (+25%, p < 0.05; Figure 7D). VE-cadherin protein expression was unchanged by either treatment (Figure 7E). M-CSF levels were increased about twofold by Ctr-CM (p < 0.01), GDM–CM (p < 0.001), and cytokine cocktail (p < 0.0001) compared to unconditioned medium (Figure 7F).