Obesity and gestational diabetes independently and collectively induce specific effects on placental structure, inflammation and endocrine function in a cohort of South African women

Maternal obesity and gestational diabetes mellitus (GDM) are associated with insulin resistance and health risks for mother and offspring. Obesity is also characterized by low‐grade inflammation, which in turn, impacts insulin sensitivity. The placenta secretes inflammatory cytokines and hormones that influence maternal glucose and insulin handling. However, little is known about the effect of maternal obesity, GDM and their interaction, on placental morphology, hormones and inflammatory cytokines. In a South African cohort of non‐obese and obese pregnant women with and without GDM, this study examined placental morphology using stereology, placental hormone and cytokine expression using real‐time PCR, western blotting and immunohistochemistry, and circulating TNFα and IL‐6 concentrations using ELISA. Placental expression of endocrine and growth factor genes was not altered by obesity or GDM. However, LEPTIN gene expression was diminished, syncytiotrophoblast TNFα immunostaining elevated and stromal and fetal vessel IL‐6 staining reduced in the placenta of obese women in a manner that was partly influenced by GDM status. Placental TNFα protein abundance and maternal circulating TNFα concentrations were reduced in GDM. Both maternal obesity and, to a lesser extent, GDM were accompanied by specific changes in placental morphometry. Maternal blood pressure and weight gain and infant ponderal index were also modified by obesity and/or GDM. Thus, obesity and GDM have specific impacts on placental morphology and endocrine and inflammatory states that may relate to pregnancy outcomes. These findings may contribute to developing placenta‐targeted treatments that improve mother and offspring outcomes, which is particularly relevant given increasing rates of obesity and GDM worldwide.


Introduction
Gestational diabetes mellitus (GDM) is associated with adverse outcomes for both the mother and child (McIntyre et al., 2019;Metzger et al., 2008). In the short term it is associated with multiple maternal and fetal complications, including preeclampsia, Caesarean delivery, fetal overgrowth, and neonatal hypoglycaemia. While in the long term, women who developed GDM have an elevated risk of developing type-2 diabetes mellitus (T2D) and cardiovascular disease (CVD). For the child, GDM also increases the risk of obesity and T2D in later life. The likely factor driving the rising prevalence of GDM is the obesity epidemic, which is fuelled by westernized diets and sedentary lifestyles (Ben-Haroush et al., 2004;Durnwald, 2015;Lao et al., 2006;Levy et al., 2010;Mwanri et al., 2015;Zhang et al., 2014). Obesity is linked to insulin resistance (Kahn & Flier, 2000) and elevates the risk of GDM and other adverse pregnancy outcomes, such as early pregnancy loss, preeclampsia, increased Caesarean section rate, macrosomia, congenital malformations, intrauterine growth restriction and stillbirth (Heslehurst et al., 2010;Jacobsen & Aars, 2015;Zhou et al., 2016). Maternal obesity is also associated with an increased long term risk of CVD, obesity and T2D in the child, as well as future cardiometabolic disease in the mother (Aune et al., 2014;Heslehurst et al., 2008;Meehan et al., 2014;Onubi et al., 2016). Understanding how GDM and obesity may affect pregnancy outcomes is thus imperative for improving human health, particularly so for low-middle income countries undergoing urbanisation/nutritional transitions.
A recent systematic review and meta-analysis identified the pooled prevalence of GDM in Africa as 13.6% when applying the 2013 World Health Organization (WHO) and the International Association of the Diabetes and Pregnancy Study Groups (IADPSG) diagnostic criteria (Muche et al., 2019). In South Africa, the prevalence of GDM ranged between 9.1% and 25.8%, depending on the diagnostic criteria used (Adam & Rheeder, 2017;Macaulay et al., 2018). Reports of maternal obesity in Africa showed an overall prevalence of up to 50.7% (Onubi et al., 2016), with a rate of 44% in South Africa; one of the highest on the continent (Anon (n.d.); Basu et al., 2010). The reasons for such high rates of obesity and insulin resistance in South African women are unclear; however, there has been demonstrable increased compensatory insulin secretion to maintain normoglycaemia among insulin resistant black South African women . Despite this, mechanistic studies of pregnancy outcome in the context of GDM and obesity have chiefly not been conducted in sub-Saharan Africa (Goedecke et al., 2017).
Prior work has suggested that obesity and GDM are accompanied by inflammatory and endocrine changes during pregnancy (Fasshauer et al., 2014;Hotamisligil et al., 1995). In pregnancy, obese women tend to show elevated circulating levels of pro-inflammatory cytokines, including tumour necrosis factor alpha (TNFα) and interleukin 6 (IL-6) (Lappas, 2014a;Pantham et al., 2015; 0 Ezekiel Musa was awarded his PhD from the University of Cape Town, South Africa in 2022. He is currently a Clinician Scientist, Senior Lecturer and Consultant Endocrinologist at Kaduna State University and Barau Dikko Teaching Hospital in Nigeria. His research is focused on improving maternal and fetal outcomes in the context of gestational diabetes and obesity, particularly in Africa. He is also keen to engage in mentoring, training and further collaboration opportunities. Esteban Salazar-Petres was awarded his PhD in 2018 from Universidad Austral de Chile, Chile. Then he undertook a Postdoctoral Research Fellowship (funded by Beca-Chile) at the University of Cambridge until 2021. Since 2022, he has been working as an early career researcher and teaching officer at Universidad Santo Tómas, Valdivia, Chile. His research is focused on the placenta and maternal adaptations under normal and adverse pregnancy conditions such as obesity, diet and stress. Segovia et al., 2014). Increased maternal serum levels of pro-inflammatory cytokines are also reported in GDM pregnancies with or without obesity (Morisset et al., 2011). The placenta is the functional interface between the mother and fetus and is both a target and source of such cytokines in the maternal circulation (Hauguel-de Mouzon & Guerre-Millo, 2006). TNFα and IL-6 have been shown to perturb nutrient transporter capacity, as well as insulin and insulin-like growth factor (IGF) signalling and cell survival pathways, in trophoblast cell lines (BeWO) and primary cultures of syncytiotrophoblast and cytotrophoblasts isolated from the term human placenta Hashimoto et al., 2010;Jones et al., 2009;Tanaka et al., 2018). In addition, placental production of TNFα has been described as an important mediator of insulin insensitivity during pregnancy, and mis-production of TNFα by the placenta may represent an important factor for the development of GDM in women in the presence or absence of obesity (Kirwan et al., 2002). This is because TNFα negatively regulates the insulin signalling transduction pathway that is responsible for glucose uptake in skeletal muscle and white adipose tissue (Borst, 2004;Nieto-Vazquez et al., 2008). The human placenta also secretes key hormones, including leptin, oestradiol, progesterone, growth hormone variant and placental lactogens (collectively termed placental lactogen family members), that regulate maternal β-cell adaptations, insulin secretion and insulin resistance and may be influenced by obesity and GDM (Napso et al., 2018;. These adaptations occur in healthy pregnancy but can be exacerbated by obesity. While placental endocrine dysfunction has been reported in obese women and is closely linked to GDM development (Lassance et al., 2015;Ngala et al., 2017;Sferruzzi-Perri et al., 2020), the extent to which maternal obesity and GDM interact at the level of the placenta to influence pregnancy metabolic adaptation and therefore outcomes for the mother and child requires further study. Therefore, we examined the expression of placental inflammatory cytokines and hormones in relation to placental structure and circulating TNFα and IL-6 concentrations in a South African cohort of non-obese and obese pregnant women with and without GDM.

Study participants
The study consisted of 71 participants over the age of 18 years who attended Groote Schuur and Mowbray Maternity Hospitals in Cape Town, South Africa for antenatal care and did not have known pre-gestational diabetes, multiple gestations, preeclampsia, HIV with unsuppressed viral load, or pregnancy secondary to assisted reproduction with gonadotropins or in vitro fertilisation. Each participant had undergone a standard oral glucose tolerance test (OGTT; 75 g) at 24−28 weeks gestation following a risk-based screening for GDM and GDM diagnosis was based on the IADPSG (FBG: ≥5.1 mmol/l or 1 h: ≥10 mmol/l or 2 h: ≥8.5 mmol/l) and the National Institute for Health and Care Excellence (NICE) (FBG: ≥5.6 mmol/l or 2 h: ≥7.8 mmol/l) diagnostic criteria adopted by the study hospitals where the participants were recruited. They were further subdivided by the presence or absence of obesity as defined by BMI >/< 30 kg/m 2 at booking, respectively (non-GDM non-obese (NGNO) = 14, and non-GDM obese (NGO) = 19, GDM non-obese (GNO) = 15, GDM obese (GO) = 23). Gestational ages at booking ranged from 15 to 37 weeks of gestation and were calculated using the last menstrual period and ultrasound scan. Socio-demographic, maternal clinical and metabolic data were obtained using a questionnaire and review of medical records. Neonatal outcomes, including APGAR scores, anthropometric measures and placental weight, were obtained at delivery. Sample size was calculated based on similar, previous work undertaken by others (Radaelli et al., 2003).

Collection and analysis of blood samples
At delivery, maternal and umbilical cord blood samples were collected, immediately centrifuged at 3500 rpm for 10 min, and the serum fraction collected into sterile tubes and stored at −80°C. TNFα and IL-6 concentrations in maternal and fetal serum were measured in duplicate using TNFα ELISA kit (ab181421, Abcam, UK) and IL-6 ELISA kit (ab46042, Abcam, UK), respectively, according to the manufacturer's instructions. The detection range of the TNFα ELISA was 15.63 pg/ml-1000 pg/ml (sensitivity: 4.32 pg/ml) with intra-assay variation and inter-assay coefficient variation of 2.5% and 3.1%, respectively. The detection range of the IL-6 ELISA was 1.6 pg/ml-50 pg/ml (sensitivity of 0.8 pg/ml) with intra-assay variation and inter-assay variation of 4.4% and <14%, respectively.

Collection of placental tissue samples
The placenta was collected and weighed immediately after delivery, and four random, centrally located representative placental tissue samples were collected for gene, protein and histological studies (NGNO =14, NGO = 19, GNO = 15, GO = 23). For gene expression analysis, placental samples were placed into RNA later solution (RNAlater, Cat. no. R0901, Sigma-Aldrich, USA), snap-frozen in liquid nitrogen and stored at −80°C. Placental samples for protein expression studies were snap-frozen and stored immediately at −80°C. Placental samples were immersion fixed in 4% paraformaldehyde or 10% neutral buffered formalin and subsequently processed into paraffin wax-embedded tissue blocks for sectioning and morphological and protein immunolocalisation studies.

Gene expression by real-time qPCR assessment
Total RNA was extracted from 30 mg of placental tissue using the GeneJet RNA purification kit (Thermo-Fisher Scientific, USA) according to the manufacturer's protocol. About 2 μg of RNA was reverse transcribed to cDNA using a High-Capacity RNA-to-cDNA kit (Applied Biosystems, USA). The synthesis of cDNA was confirmed by a conventional polymerase chain reaction using GAPDH primers. Three dilutions of each sample (1:10, 1:20 and 1:100) were run as triplicates with negative controls for real-time qPCR gene quantification. The real-time qPCR was performed using SYBR Green chemistry (Applied Biosystems, UK) and gene-specific primers (Table 1) using a standard thermal cycling protocol conducted as follows: 50°C for 2 min, 95°C for 10 min and 40 cycles of 95°C for 95 s and 60°C for 1 min, on a 7500 Fast Real-Time PCR thermocycler System (Applied Biosystems, UK). Ten-fold serial dilutions of cDNA for each primer set were used to generate a standard curve, and only those with amplification efficiencies above 0.90 and R 2 above 0.985 were included. The relative expression of each gene of interest was calculated by the 2 − Ct method, and GAPDH, β2M and ACTIN, which remained stably expressed across the groups, were used as housekeeping genes for data normalisation.

Protein abundance by western blot analysis
Protein extraction was performed on frozen placental tissues homogenized by sonication (Soniprep 150, UK) for 20 s in RIPA lysis buffer containing 50 mM Tris-HCl (pH 8), 1% Triton X-100, 150 mM NaCl, 0.1% sodium dodecyl sulphate (SDS), 0.5% sodium deoxycholate with a protease and phosphatase cocktail inhibitor (Halt protease and phosphatase inhibitor, Thermo Scientific, USA). The placental lysates were then centrifuged at 15,000 rcf for 20 min at 4°C (Labnet International, NJ07095, USA). The supernatant was collected, and protein concentration was measured using a bicinchoninic acid (BCA) protein assay kit (Thermo Scientific, Rockford, USA). Placental lysates were diluted in 4× Laemmli buffer (BioRad, USA) and water to a final protein concentration of 20 μg/well. The prepared samples were denatured at 95°C for 10 min and then separated by sodium dodecyl sulphate polyacrylamide gel electrophoresis. Then, proteins in the gel were electroblotted onto polyvinylidene difluoride membrane (Hybond-P, Amersham Biosciences) using a semi-dry technique (Semi-dry Blotter, Invitrogen) with protein transfer confirmed by Ponceau Red staining. The membrane was blocked with 5% milk in Tris Buffered Saline (TBS) with Tween 20 (TBST) for 60 min and subsequently incubated with rabbit polyclonal antibody against IL-6 (1:800; ab6672, Abcam, UK) or mouse monoclonal antibody against TNFα (1:1000; Ab1793, Abcam, UK). Both antibodies were diluted in 5% milk TBS and incubated overnight at 4°C. Two sets of gels were run simultaneously; one of the final membranes was incubated with primary antibody, while the other membrane was incubated with the blocking solution (5% milk TBST), serving as the negative control (to confirm antibody specificity). Following washing, membranes were incubated in goat anti-rabbit antibody (1:4000; SC-2004, Santa Cruz Biotechnology, USA) or goat anti-mouse antibody (1:4000; Cat. no. 1706516, BioRad Laboratory, USA) diluted in 5% milk TBST for 60 min, for detecting IL-6 and TNFα, respectively. Protein bands were visualised by chemiluminescence using Clarity TM Western ECL substrate (Cat. no. 1705061, BioRad, USA) and exposed on C400 Azure Biosystems Imaging System. The captured images were downloaded, and the signal intensity of protein bands quantified using GelQuantNet software. The abundance of each protein of interest was normalised to the abundance of GAPDH (detected by rabbit polyclonal antibody SC-25778, Santa Cruz Biotechnology, USA; diluted 1:1000 and incubated overnight at 4°C).
J Physiol 601.7

Morphological analysis
Wax-embedded placental tissue sections (5 μm) were de-waxed and stained in Haematoxylin and Eosin (H&E) according to standard protocols. Slides with stained sections were scanned using Olympus VS120 Digital Slide Scanner to create electronic images and analysed to determine the structure of the placenta as described previously (Yong et al., 2022). In brief, grids of equally spaced test points were superimposed on each placenta image at random positions and used to determine component volume and surface densities. Images of 5× magnification were used to calculate volume densities of stem villi, intermediate villi, terminal villi and intervillous space in the placenta (grid 7 × 5 = 35 points). Similarly, a grid (composed of 13 × 9 = 117 points) was used to estimate volume densities of placental terminal villi constituents (syncytiotrophoblast (STB), stroma, fetal capillaries (FC)) at 20× magnification. Surface densities of maternal blood spaces (MBS) and FC were estimated using 16 cycloids per image to count chance intersections between test lines and the surface traces of peripheral (intermediate + terminal) villi and their fetal capillaries. Component densities were converted into absolute volumes and component surfaces using the volume of the placenta as the reference space. Placental volumes were derived from weights taking tissue density to be 1.05 g/cm 3 . Analyses were performed blinded to the group.

Protein immunolocalisation analysis
Immunohistochemistry was performed in wax-embedded placental tissue sections (5 μm) using standard methods. In brief, following de-waxing, sections were incubated for 15 min in 3% hydrogen peroxide in distilled water and then subjected to antigen retrieval using 0.01 M sodium citrate buffer (pH 6) in a pressure cooker. Sections were then blocked for non-specific binding using 3% bovine serum albumin (BSA) for 30 minutes and 5% BSA for 15 minutes for TNFα and IL-6 immunostaining, respectively. Sections were then immuno-stained with either mouse monoclonal primary antibody against TNFα (1:100 dilution in 5% BSA, ab1793, Abcam, UK) or rabbit polyclonal primary antibody against IL-6 (1:100 dilution in 5% BSA, ab6672, Abcam, UK) at 4°C overnight. Following incubation, the sections were washed in TBST and incubated with anti-mouse and anti-rabbit EnVision+System-HRP (K4001 and K4003, Dako, USA) for TNFα and IL-6, respectively, for 30 min at room temperature as per the manufacturer's instructions. After washing, sections were incubated with liquid diaminobenzidine (DAB) + substrate chromogen system (K3468, Dako, USA) for 10 min, then quenched in tap water and counter-stained in Mayer's Haematoxylin. Sections were dehydrated, mounted with Entellan mounting media (Sigma-Aldrich), and visualisation was done with a light microscope. To determine differences in the amount of staining between groups, two methods were performed. Using a Macro designed with ImageJ, the percentage of DAB-stained area related to the Haematoxylin-stained area representing the terminal villi area was quantified in 10 images at 20× magnification and data were expressed as DAB area/terminal villi area. In addition, the IRS score semi-quantitative method (Fedchenko & Reifenrath, 2014) was used to evaluate the DAB staining area and intensity of the STB, stroma and FC of terminal villi within the section. To perform this analysis, score values were assigned to evaluate the percentage of positive staining area, with the score range from 0 (no staining), 1 (<10% of positive area), 2 (10-50% of positive area), 3 (51-80% of positive area) to 4 (>80% of positive area). In addition, the intensity of staining was scored as; 0 = no DAB positivity, 1 = mild DAB positivity, 2 = moderate DAB positivity, and 3 = intense DAB positivity. The final IRS score was calculated by multiplying both scores, with a final IRS score from 0−1 = negative, 2−3 = mild, 4−8 = moderate, and 9−12 = strongly positive. Analyses were performed blinded to the group.

Statistical analysis
Data were analysed using GraphPad Prism software version 8 for Windows, San Diego, CA, USA. Data were reported as means ± standard deviation (SD). The Shapiro-Wilk test assessed the distribution of continuous variables. A two-way analysis of variance (ANOVA) was used to determine the effect of GDM and obesity and their interaction on gene expression, protein abundance, immunostaining and circulating serum concentrations, followed by Bonferroni post hoc tests to identify significant differences between the four groups (NGNO, NGO, GNO, GO). Pearson´s correlation coefficient (r) analyses were used to assess whether relationships existed between data and the strength of their associations. P values of <0.05 were considered statistically significant.

Maternal and fetal demographic and clinical data
In general, our cohort of pregnant women was composed of mixed ancestry/coloured and black ethnic groups, most of them were in full employment or were students, and the majority possessed high school or tertiary education levels ( Table 2). A family history of diabetes mellitus was similar in the NGO and GNO groups but was at least twice as high in the GO patients. As expected, both non-GDM and GDM women with obesity were heavier at booking and delivery compared to non-obese (Table 3). They also had a greater BMI at booking (mean ± SD kg/m 2 ; NGNO: 25.24 ± 2.96, NGO: 38.13 ± 7.24, GNO: 26.52 ± 2.14, and GO: 37.72 ± 5.04) and showed an overall decrease in weight gain relative to the non-obese groups (when expressed as kg, percentage weight gain, and rate of weight gain) during gestation. Blood pressure was overall also higher in women who were obese, an effect significant by pairwise comparison for GO compared to GNO (P = 0.033). Also, as expected, fasting blood glucose and glucose concentration 2 h post OGTT (indicative of glucose intolerance) were higher in the GDM groups compared to non-GDM (Table 3). GDM was associated with decreased gestational age at delivery and reduced infant length (Table 4). Infant ponderal index was overall increased by both obesity and GDM, with significance detected by pairwise comparisons for the GDM versus non-GDM groups. However, infant and placental weight, and the ratio between them, did not differ among the groups.

Placental structure
Unbiased stereological analysis of placental histology identified an interaction between GDM and maternal obesity in determining the percentage of stem villi in the placenta, with obesity decreasing it in women without GDM, but obesity increasing it in women with GDM (Table 5). There was also a reduction in the percentage of terminal villi in obese women irrespective of GDM status (i.e. lower percentage terminal villi NGO and GO groups). There was also an effect of GDM, maternal obesity and interaction between them, with a greater percentage of intervillous space in the placenta for the NGO compared to NGNO, but a lower percentage for the GO versus NGO groups. Whilst immature stem villi and intermediate villi volumes were not affected by either GDM or maternal obesity, mature terminal villi volume was reduced by maternal obesity (NGO and GO groups), with pairwise comparisons revealing the most significant effect when comparing GO to GNO women (Table 5). Meanwhile, syncytiotrophoblast (STB), fetal capillaries (FC), and stroma volumes were all significantly decreased by obesity, irrespective of GDM status. Additionally, maternal blood space (MBS) and FC surface area were decreased by obesity, with the former most dramatically reduced in GO compared to GNO women. Furthermore, the theoretical diffusion capacity of the placenta was diminished in obese women, irrespective of the GDM diagnosis. Finally, placental intervillous space volume was reduced by GDM irrespective of BMI status. Hence, maternal obesity, and to a much lesser extent, GDM, had an effect on several placental morphological parameters.  Data are means ± SD. Differences between groups were determined using 2-way ANOVA and Bonferroni's multiple comparisons test highlighted in bold. # Significantly different to non-GDM (P < 0.05, pairwise comparisons). APGAR: appearance, pulse, grimace, activity and respiration.

Placental expression of hormone and cytokine genes
We sought to determine the relative mRNA expression of various inflammatory, growth factor and steroidogenic genes across obese and GDM study cohorts using quantitative real-time PCR, and found no significant differences in the expression of steroidogenic (HSD3B1, CYP11A1 and CYP19A1), placental lactogen (GHV, CSH, CSH1 and CSH2), pro-inflammatory cytokine (TNFα, IL-1β and IL-6), and growth factor (IGF2 and VEGF) genes by the placenta in response to GDM or obesity (Fig. 1). However, placental LEPTIN mRNA expression was overall downregulated by maternal obesity, an effect that was significant by pairwise comparisons between NGO and NGNO women (P = 0.011; Fig. 1C). These data suggest that obesity, particularly in women without GDM, has a specific effect of reducing the expression of LEPTIN by the placenta.

Placental TNFα and IL-6 protein abundance and localisation
Placental expression of TNFα and IL-6 may be subjected to post-transcriptional regulation. Moreover, the placenta is heterogeneous in nature, and cellular localisation of TNFα and IL-6 may be altered in response to different gestational environments. Hence, western blotting and immunohistochemistry were performed. This revealed that placental TNFα protein abundance was overall diminished by GDM (P GDM = 0.003), with pairwise comparisons showing a significant effect when comparing non-obese women (P = 0.042, Fig. 2A). Immunohistochemistry showed that TNFα was highly expressed by the placenta and particularly abundant in the STB ( Fig. 2B and C). Further analysis revealed that maternal obesity, rather than GDM affected placental terminal villi TNFα abundance, with pairwise analyses showing

Figure 1. Relative mRNA expression of steroidogenic enzymes (A), placental lactogen (B), growth factor (C), and pro-inflammatory cytokine (D) genes in the placenta of non-obese and obese pregnant women with and without GDM
Data are shown as individual values with mean ± SD and were normalized to the mean expression levels of GADPH, B2MA and β-ACTIN. Blue and orange are non-obese groups, purple and red are obese groups. Differences between groups were determined using 2-way ANOVA and Bonferroni post hoc test. * Significantly different to non-obese (P < 0.05, pairwise comparisons). n = 9-10 per group. [Colour figure can be viewed at wileyonlinelibrary.com] increased TNFα immunostaining in the GO compared to the GNO women (P = 0.045, Fig. 2B). Additionally, STB TNFα immunoreactivity was significantly increased by maternal obesity overall, with no differences found in the stroma and fetal capillaries (Fig. 2C). Neither GDM nor obesity affected the abundance of IL-6 protein in the placenta (Fig. 2D). IL-6 was immunolocalised to the STB, stroma, and fetal capillaries of the placenta (Fig. 2E and F), and further analysis identified that placental terminal villi immunostaining was significantly affected by an interaction between GDM and obesity. Pairwise comparisons further revealed that IL-6 immunostaining was lower in only the NGO compared to the NGNO group (P = 0.038, Fig. 2E). Subsequent analysis revealed that while no differences in STB immunolocalisation were found, stroma and fetal capillary IL-6 were overall lower in obese women, an effect significant by pairwise comparisons in the GO compared to the GNO group (stroma: P = 0.045 and fetal capillaries: P = 0.049, Fig. 2F). Thus, GDM and obesity Data are represented as means ± SD. Differences between groups were determined using 2-way ANOVA and Bonferroni post hoc test highlighted in bold. * significantly different to non-obese (P < 0.05, pairwise comparisons). # Significantly different to non-GDM (P < 0.05, pairwise comparisons). n = 9-10 per group for stereological analyses. Abbreviations: FC, fetal capillaries; MBS, maternal blood space; STB, syncytiotrophoblast. affect placental TNFα and IL-6 protein abundance and localisation.

Maternal and infant cord TNFα and IL-6 circulating concentrations
The placenta is bathed in maternal and fetal blood and may respond, as well as contribute, to changes in the abundance of circulating factors, namely TNFα and IL-6. Compared to other cytokines (like IL-1β), TNFα and IL-6 are particularly inflammatory and implicated in the development of insulin resistance during pregnancy (Kirwan et al., 2002;Morisset et al., 2011). Hence, we quantified the concentrations of TNFα and IL-6 in maternal and cord blood at delivery. Circulating concentrations of TNFα were overall decreased in the circulation of women with GDM (P GDM < 0.0001), whilst there was no difference between non-obese and obese women (Fig. 3A). There was no effect of GDM or maternal obesity on cord TNFα concentration (Fig. 3B). There were also no differences in IL-6 concentrations in maternal or cord blood with GDM or maternal obesity ( Fig. 3C and D). Thus, circulating levels of TNFα in the mother appear to be altered by GDM only.

Figure 2. Abundance and immunolocalisation of TNFα and IL-6 protein in placenta of non-obese and obese pregnant women with and without GDM
TNFα and IL-6 protein abundance quantification in placenta tissue by western blotting. The molecular weight of each protein was confirmed at 26 kDa for TNFα, 55−70 kDa for IL-6 and 37 kDa for GADPH (A and D). TNFα and IL-6 immunolocalisation measured by using an ImageJ macro (B and E) and IR score semi-quantitative determination (C and F). Images were obtained with 20× magnification. Data are shown as individual values with mean ± SD. Blue and orange are non-obese groups, purple and red are obese groups. Differences between groups were determined using 2-way ANOVA and Bonferroni post hoc test. * Significantly different to non-obese (P < 0.05, pairwise comparisons). n = 7−8 per group for western blot analysis and immunolocalisation analysis. Abbreviations: STB, syncytiotrophoblast; FC, fetal capillaries. [Colour figure can be viewed at wileyonlinelibrary.com]

Discussion
In this study, we comprehensively investigated the effect of GDM, maternal obesity and their possible interaction on placental endocrine, inflammatory and structural characteristics in a well-characterised cohort of South African women. We show that placental expression of endocrine and growth factor genes was not altered by obesity or GDM. However, LEPTIN gene expression was diminished, and TNFα and IL-6 protein immunolocalisation were affected in the placenta of obese women in a manner influenced by GDM status. In addition, placental TNFα protein abundance was reduced in line with decreased maternal circulating TNFα concentrations in GDM women. Moreover, both maternal obesity and, to a lesser extent, GDM were accompanied by specific changes in placental morphometry. Other pregnancy characteristics, including maternal blood pressure, weight gain and infant ponderal index were also modified in obese and/or GDM women. Together, these data suggest that placental phenotype may be associated with specific maternal and fetal clinical outcomes in the context of obesity and GDM. Structural alterations associated with maternal obesity included reduced villi maturation, vascularity, surface areas and diffusion gradients for maternal-fetal exchange of nutrients and oxygen. Whether this may be related to the elevated systolic blood pressure that is considered within the normal range (as women with pregnancy-induced hypertension and pre-eclampsia were excluded) and was observed in our obese women during pregnancy, is unclear. In the placenta of obese women, GDM specifically increased the density of intervillous spaces and select morphological features, namely the density of stem villi, were significantly affected by an interaction between obesity and GDM. Prior studies have only investigated the effect of obesity or GDM separately on placental morphology, and some of their findings are consistent with ours (Calderon et al., 2007). For example, previous work has shown that the placenta of obese women shows reduced maturity index in keeping with our findings, but maternal and fetal neutrophilic infiltration and villous lesions have also been reported in relation to obesity (Bar et al., 2012;Huang et al., 2014;Rosado-Yépez et al., 2020). Furthermore, maternal obesity was additionally linked to an increase in the muscularity of villous vessels in the placenta (Roberts et al., 2011). Thus, the placental structure is differentially altered by obesity and GDM. An increase in stem villi with less placental tissue available for hormone secretion (particularly evident in obese women with GDM) might also be a cause rather than a consequence of GDM. However, as morphological analyses of the placenta were performed at term, and GDM manifests much earlier in gestation, further work is required to assess the temporal relationships between changes in the placenta and maternal glucose handling.
The effect of maternal obesity on the placental structure may have clinical implications for fetal outcomes. In particular, it could contribute to the reported increased risk of perinatal death, fetal death, stillbirth and infant death for obese women (Aune et al., 2014;Meehan et al., 2014). It may also have implications for infant birth weight. Obese women delivering large babies exhibit increased placental abundance and activity of glucose (particularly GLUT1) and amino acid (including SNAT1) transporters (Illsley, 2000;Takahashi et al., 2017). Hence, the observations of reduced villi formation and surface area in our cohort of obese women might reflect a compensatory strategy for the placenta to prevent excessive fetal growth in obese women and may explain the observed lack of difference in birth weight and length between non-obese and obese women in our study. J Physiol 601.7 Moreover, obese women with GDM had increased intervillous space volume (compared to non-obese counterparts with GDM). Thus, it is expected that adaptative changes from the placenta could explain the elevated infant ponderal index in our study (which was greatest in the obese women with GDM), presumably due to increased nutrient supply to the fetus.
Our study showed no effect of obesity or GDM on the expression of placental lactogen genes (GHV,CSH1 and CSH2). These data are consistent with previous studies of the human placenta in GDM and non-GDM women (Männik et al., 2012;Mills et al., 1998;Ngala et al., 2017;Retnakaran et al., 2016;Tzingounis et al., 1978;Ursell et al., 1973). However, other work has also shown that placental CSH1 and GHV mRNA expression is aberrant in obese women, depending on their GDM status and clinical management (Jin et al., 2018). In the study by Jin et al. (2018), women were categorised into three groups, namely lean (BMI: 20−25 kg/m 2 ), obese (BMI: 35 kg/m 2 ), and obese GDM (BMI: 35 kg/m 2 and GDM diagnosis), which is different to our cohort of women (non-GDM and GDM classified as non-obese (BMI ≤ 30 kg/m 2 ) and obese (BMI ≥ 30 kg/m 2 )). Data on placental LEPTIN gene expression in the presence of obesity are also conflicting. In our study, we found a significant reduction in LEPTIN gene expression by the placenta in obese women without GDM, which is consistent with other studies (Lepercq et al., 2001). The role of placental leptin has yet to be fully elucidated; however, it is proposed to play a role in promoting placental amino acid transportation (Fowden et al., 2015), encouraging proliferation and survival of trophoblast cells via stimulation of JAK/STAT, MAPK and PI3K pathways in placental cells, as well as acquisition of maternal insulin and leptin resistance during pregnancy (Maymó et al., 2011;Napso et al., 2018;Sferruzzi-Perri et al., 2020). However, other studies have reported no significant differences in placental LEPTIN mRNA levels in obese compared to non-obese women regardless of their GDM status (Allbrand et al., 2015;Tsiotra et al., 2018). Other studies have also found elevated placental LEPTIN expression in GDM compared to controls, although the BMI status of the study population was not reported (Lepercq et al., 2001;Mrizak et al., 2014). The placental syncytiotrophoblast is the main source of circulating leptin in pregnancy. As leptin is typically increased in the circulation of obese individuals, this may be the explanation for the down-regulation of placental LEPTIN expression in our study, as leptin can act in negative feedback (Wang et al., 1999). In future work, it will be valuable to measure leptin receptor expression and signalling in the placenta, as well as circulating levels in our cohort of women. It would also be helpful to assess concentrations in cord blood, as data from rodents suggest that a late gestational/neonatal surge in leptin is required for early growth and neurotrophic effects that establish appetite control with important implications for later metabolic disease risk (Picó et al., 2007;Yau-Qiu et al., 2020).
Neither placental IGF2 nor VEGF gene expression were affected by obesity or GDM diagnosis in our cohort of women. This is consistent with findings of a lack of change in placental VEGF-A gene expression in GDM and non-GDM women with or without obesity (Lappas, 2014b). However, others have reported decreased VEGF-A mRNA and protein expression, as well as elevated syncytiotrophoblast and stromal VEGF immunoreactivity in women with GDM (Akarsu et al., 2017;Meng et al., 2016), which may be related to the specific population and clinical characteristics of the women studied (namely, the definition of obesity, BMI of women in each group, diagnostic criteria, GDM management and mode of delivery). There are also reports that the soluble form of the VEGF receptor (sFLT1) that is released by the placenta may be altered in the circulation of women with GDM in early pregnancy (Napso et al., 2021). Similarly, significantly higher placental IGF2 gene expression was found in GDM compared to normoglycaemic women (Su et al., 2016). The expression of select steroidogenic enzyme genes (HSD3B1, CYP11A1 and CYP19A1) in the placenta was also not affected, which would imply normal production of hormones, namely oestradiol and progesterone in our cohort of women, regardless of their obesity and GDM status. These results are congruent with findings by Maliqueo et al. (2017); however, others have demonstrated an association between circulating progesterone and oestradiol concentrations and the risk of GDM (Couch et al., 1998;Montelongo et al., 1992). In addition, plasma oestradiol and progesterone concentrations are lower in obese compared to lean pregnant women (Lassance et al., 2015). Finally, fetal sex has been shown to influence changes in placental steroidogenic capacity in the context of maternal obesity (Maliqueo et al., 2017). Whilst mechanisms underlying these differences are unclear, the role of fetal sex may be pivotal. Thus, increasing the sample size and splitting data by infant sex would be essential in future work.
Obesity is described as a low-grade inflammatory state. Moreover, TNFα and IL-6 are predominantly expressed in the placenta and adipose tissue, and have been demonstrated to be potent stimulators of inflammatory activities leading to insulin resistance (Kern et al., 2001;Riley & Nelson, 2010). In keeping with previous studies, we did not find differences in the expression of genes encoding TNFα, IL-6 and IL-1β by the placenta with obesity or GDM (Allbrand et al., 2015;Kleiblova et al., 2010). On the other hand, others have reported increased TNFα (Allbrand et al., 2015;Heinig et al., 2000;Kleiblova et al., 2010;Stirm et al., 2018) and IL-6 gene expression in the placenta from GDM women compared to healthy controls (Heinig et al., 2000;Mrizak et al., 2014;Roberts et al., 2011;Stirm et al., 2018). It is probable that variations between other studies and our findings are due to differences in GDM screening methods, mode and timing of delivery, and fetal sex (Barke et al., 2019;Heinig et al., 2000). Indeed, it must be noted that there was a higher proportion of female babies and longer gestational length in the non-GDM compared to the GDM groups in the present study.
Even though maternal obesity and GDM did not impact TNFα mRNA levels in the placenta, placental TNFα protein abundance and localisation were significantly influenced by GDM and obesity, respectively. Of note, GDM reduced placental lysate TNFα levels, and obesity increased placental STB TNFα localisation. Previous studies have also observed increased TNFα protein expression by the placenta in obese women (Basu et al., 2011;Challier et al., 2008;Oliva et al., 2012). Similar to placental lysate abundance, maternal circulating TNFα levels were significantly reduced by GDM, but unaffected by obesity. However, other work did not find a significant difference in the circulating concentration of TNFα between GDM and non-GDM women (Gomes et al., 2013). In line with no change in IL-6 mRNA expression, obesity and GDM had no effect on placental IL-6 protein abundance. This contrasts with the work of others who found higher levels of IL-6 (and TNFα) in placenta lysates of GDM compared to non-GDM women (Zhang et al., 2017). In addition, we identified a reduction in placental villi IL-6 immunostaining in obese women without GDM and decreased villous stroma and fetal capillary IL-6 immunostaining in obese women that was most pronounced if they had GDM. Maternal and cord IL-6 levels were not influenced by GDM, obesity, or their interaction. Maternal circulating IL-6 levels are reported to be higher in GDM pregnancies (Hassiakos et al., 2016;Kuzmicki et al., 2009;Morisset et al., 2011;Zhang et al., 2017), yet like our study, no differences were detected in IL-6 concentrations between obese and non-obese women (Friis et al., 2013;Lain et al., 2008). In our study, all GDM women received metformin alone or in combination with insulin to control their hyperglycaemia. Metformin is known to reduce inflammatory cytokine production (Tizazu et al., 2019), and a randomised control trial of type-2 diabetic patients found decreased circulatory levels of TNFα and IL-6, among other cytokines, after a year of treatment with metformin compared with baseline (Mo et al., 2019). Thus, a potential explanation for the significantly lower placental and maternal circulating TNFα levels is the downregulating effect of metformin on inflammation in our cohort of GDM women. However, the interaction between obesity, GDM and metformin treatment on villous IL-6 immunolocalisation requires further study. Furthermore, changes in placental genes and proteins that are known to contribute to the development of GDM, like placental lactogen genes, cytokines, IGF2 and sex steroids (Napso et al., 2018;Sferruzzi-Perri et al., 2020) may have been attenuated by a treatment like metformin and thus, explain minimal changes in the placenta at term. Together, our findings suggest that GDM impacts on placental inflammation in a different way compared to maternal obesity, although much further work is required to assess the contribution of metformin to placental alterations in women who have GDM by comparison to those who are not treated with metformin.
A key strength of the study is the combined use of clinical phenotyping, stereology and molecular assays of the placenta and biochemical assessments of maternal and cord blood to comprehensively investigate the effect of maternal obesity and GDM in a cohort of South African women. Thus, this study provides a unique intersection between clinical and basic science with the potential to advance our understanding of the mechanisms underpinning pregnancy outcomes and fetal programming. However, since sample sizes within the study sub-groups were relatively small, we were not able to examine whether there may be a sex-specific effect on the placental gene, protein and morphological changes. Increasing the sample size would also be worthwhile for performing more complex modelling that incorporates other potential confounders, such as maternal age and socioeconomic status in future work. We are also aware that our participants were categorised as obese and non-obese, with the latter including overweight women rather than only healthy/lean-weight women, which probably limited our ability to fully identify how obesity and GDM interact to mediate pregnancy outcomes. Indeed, the rates of overweight/obesity are incredibly high among South African women of childbearing age (Nglazi & Ataguba, 2022), and the proportion of women presenting to our hospitals who fall into this category is ∼80%, making it very challenging to have a GDM group that is lean. Undertaking a study with three distinct BMI categories, obese, overweight and lean, would also allow us to identify if and how GDM may cause worse outcomes than obesity alone. It may also reduce the variability in our findings. The variability in our study may have also been influenced by the different modes of delivery and efficacy of metformin treatment given to all the GDM women in our study. Furthermore, although the placental sampling was standardised across the women in our study, some variability may have been introduced, especially for the gene expression data, which is very sensitive, as the placenta is heterogeneous in nature (Cindrova-Davies & Sferruzzi-Perri, 2022) and morphological differences between groups were found.
In summary, we demonstrate that obesity and GDM have specific impacts on placental morphology, endocrine and inflammatory state that are in line with pregnancy and infant outcomes. Importantly, that obesity, more than GDM, has various effects on placental morphological J Physiol 601.7 parameters, probably with pathophysiological implications. Whilst further work is required to precisely ascertain how obesity and GDM interact to mediate pregnancy outcomes, our findings may be helpful for developing placental-targeted treatments that improve mother and offspring health, which is particularly relevant given that rates of obesity and GDM during pregnancy are increasing worldwide and have the most significant economic impacts in low-middle income countries.