Gestational diabetes impacts fetal precursor cell responses with potential consequences for offspring

Abstract Fetal programming has been proposed as a key mechanism underlying the association between intrauterine exposure to maternal diabetes and negative health outcomes in offspring. To determine whether gestational diabetes mellitus (GDM) might leave an imprint in fetal precursors of the amniotic membrane and whether it might be related to adverse outcomes in offspring, a prospective case‐control study was conducted, in which amniotic mesenchymal stem cells (AMSCs) and resident macrophages were isolated from pregnant patients, with either GDM or normal glucose tolerance, scheduled for cesarean section. After characterization, functional characteristics of AMSCs were analyzed and correlated with anthropometrical and clinical variables from both mother and offspring. GDM‐derived AMSCs displayed an impaired proliferation and osteogenic potential when compared with control cells, accompanied by superior invasive and chemotactic capacity. The expression of genes involved in the inflammatory response (TNFα, MCP‐1, CD40, and CTSS) was upregulated in GDM‐derived AMSCs, whereas anti‐inflammatory IL‐33 was downregulated. Macrophages isolated from the amniotic membrane of GDM mothers consistently showed higher expression of MCP‐1 as well. In vitro studies in which AMSCs from healthy control women were exposed to hyperglycemia, hyperinsulinemia, and palmitic acid confirmed these results. Finally, genes involved in the inflammatory response were associated with maternal insulin sensitivity and prepregnancy body mass index, as well as with fetal metabolic parameters. These results suggest that the GDM environment could program stem cells and subsequently favor metabolic dysfunction later in life. Fetal adaptive programming in the setting of GDM might have a direct negative impact on insulin resistance of offspring.


| INTRODUCTION
Gestational diabetes mellitus (GDM) is associated both with short-term adverse obstetric and perinatal complications and with long-term metabolic health consequences for offspring. 1 In this context, fetal programming has been proposed as a key mechanism underlying the link between the intrauterine exposure to maternal diabetes and an increased risk for metabolic dysfunction in adulthood, leading to type 2 diabetes, obesity, and cardiovascular disease. 2 However, the ultimate mechanisms involved are not well known.
As a natural interface between mother and fetus, the functional integrity of the placenta is crucial to support fetal growth and development as well as to adapt to the maternal nutritional and metabolic status. 3 Thus, beyond its recognized endocrine, immunological, and vascular functions, the placenta is a highly structured organ with multiple cell types dedicated to nutrient transport and energy metabolism. 4 Although the extent to which maternal glycemic control contributes to placenta dysfunction remains a matter of debate, 5 an enlarged size, altered vascular permeability, and the presence of low-grade inflammation seem to be typical features of GDM. 4 In this aspect, the potential of these disturbances to imprint fetal precursor cells found in the placenta and/or umbilical cord is emerging as a novel pathway playing a key role in the etiology of metabolic diseases in offspring of mothers with GDM. Indeed, recent studies have revealed the significant impact of GDM on human umbilical cord-derived stromal cells in terms of proliferation, 6 mitochondrial dysfunction, 7 and angiogenesis. 8 The amniotic membrane is the innermost layer of the placenta and possesses anti-inflammatory and antibacterial properties that contribute to materno-fetal tolerance. It also exerts many metabolic functions, such as the transport of water and soluble molecules and the production of bioactive factors. 9 Our study sought to investigate whether GDM might leave an imprint in fetal precursors found in the amniotic membrane and, if so, whether it might be related to adverse outcomes in offspring. More specifically, we focused on amniotic mesenchymal stem/stromal cells (AMSCs) which are mainly characterized by their low immunogenicity, immunomodulatory properties and mesodermal multilineage differentiation capacity in vitro. 10 We report evidence that maternal metabolic derangements during gestation disturb the biological properties of AMSCs. Of note, we found an association between the biological features of these fetal precursor cells and the maternal and fetal clinical and metabolic parameters, supporting the notion that fetal adaptive programming in the setting of GDM might have a direct impact on offspring.

| Study subjects
Eighteen pregnant women with a singleton pregnancy (9 with GDM and 9 with normal glucose tolerance acting as controls) scheduled for cesarean delivery were included in this study, which was performed at the Hospital Universitari de Tarragona Joan XXIII (HUJ23) according to the tenets of the Helsinki Declaration. The institutional review board (CEIm) approved the study protocol and all patients gave written informed consent before participating in the study. All mothers diagnosed with GDM were recruited at the Diabetes and Pregnancy Clinic of the HUJ23, while control subjects were recruited at the delivery suite. GDM was diagnosed according to the current criteria of the Spanish Diabetes in Pregnancy guidelines, which followed the National Data Group Criteria. 11,12 Four of the GDM women were treated only with diet whereas five were also treated with insulin.
Timing of delivery was based primarily on obstetric indications. Gestational age was confirmed in all pregnant women by a routine ultrasonographic examination performed before 20 weeks of gestation.
Exclusion criteria for all subjects were preexisting type 1 or type 2 diabetes, inflammatory or chronic diseases, current use of drugs known to affect carbohydrate metabolism, fetal anomalies identified at birth, smoking, or high blood pressure.

| Clinical and demographic data
Upon inclusion, demographic and historical data were collected paying special attention to third trimester HbA1c in GDM women, 1-hour 50-g glucose challenge test (GCT), pregravid weight,

Significance statement
Signatures of metabolic deregulation seem to remain in cells early in development. Given the location on the inner side of the placenta, amniotic membrane stem cells might be a good indicator of how the intrauterine environment impacts the fetus. To the best of authors' knowledge, this study showed for the first time how gestational diabetes disturbs both the phenotype and the functional characteristics of amniotic mesenchymal stem cells, and these alterations are related to maternal and fetal metabolic status, suggesting that fetal adaptive programming in the setting of gestational diabetes might have a direct impact on offspring.
gestational weight gain (GWG), and gestational age at delivery.
Maternal anthropometric measurements of height (measured to the nearest 0.5 cm) and weight (measured to the nearest 0.1 kg) were obtained using a medical scale. Body mass index (BMI) was calculated using the formula BMI = weight (in kilograms)/height (in meters) 2 . GWG was calculated as (final weight)−(pregravid weight). Neonatal length, weight, and waist circumference were measured at birth and the waist circumference/length ratio was calculated. Suprailiac skinfold thickness was measured at least three times using a Holtain skinfold caliper (Chasmors Ltd, London, UK) to obtain a consistent and stable reading.

| Sample collection and processing
Maternal blood samples were collected in the morning after an 8 hours fast and immediately before cesarean section, and umbilical vein cord blood was obtained at the time of delivery. Serum was immediately separated by centrifugation and stored at −80 C until analysis. Full-term placentas (37-39 weeks gestation) were collected after delivery and immediately processed under sterile conditions. The amniotic membrane was mechanically peeled free from underlying chorion and washed with phosphate-buffered saline (PBS) containing antibiotics.

| Laboratory measurements
Glucose, cholesterol, and triglycerides levels were determined using ADVIA 1800 and 2400 (Siemens AG, Munich, Germany) autoanalyzers by standard enzymatic methods. Fasting plasma insulin was determined by immunoassay on the Centaur XP platform (Siemens AG).

| Histological study of placenta
Tissue sections from the maternal (facing the decidua) and fetal (facing the amniotic cavity) sides of the placentas were collected, washed, fixed in a 4% paraformaldehyde, embebbed in paraffin, cut into slices using a rotator microtome and mounted onto glass microscope slides. Tissue sections were then dewaxed and stained with Masson's trichrome to highlight connective tissue and collagen fibers of the extracellular matrix. Slides were observed microscopically at ×2.5 magnification using a Leica DM LB2 bright-field microscope (Leica Microsystems GmbH, Wetzlar, Germany). Ten different nonoverlapping fields randomly selected from each slide were visualized and captured at ×40 magnification using a Leica DFC320

| Isolation of amniotic membrane-resident macrophages
Macrophages were purified by adherence to tissue culture plates following an established procedure that results in >90% purity. 15 Briefly, the cell pellet from the digested membrane was resuspended in complete medium and allowed to attach for 10 minutes. The supernatant was then collected and seeded again in different flasks to obtain AMSCs. Culture plates were then washed twice with PBS. Cells that did not express the macrophage marker CD68 were discarded.
Since amniotic macrophages and AMSCs are isolated from the same primary sample, in order to assess their purity, the expression levels of typical markers of monocyte-lineage and antigen presenting cells, such as CD80, CD86, CD163, CD206, and CD209 were compared in amniotic macrophages and AMSCs.

| Immunophenotyping
To assess mesenchymal lineage features and to determine whether they met the minimum criteria defined by the International Society of Cell Therapy, 16 AMSCs (1 × 10 5 ) were incubated with a panel of primary antibodies (BD Pharmingen, San Diego, California) and their surface markers expression was analyzed by flow cytometry (FacsAriaIII, BD Biosciences, San Jose, California). Data analysis was performed using the FACSDiva software (BD Biosciences).

| Multilineage differentiation capacity
To determine the multilineage differentiation capacity of AMSCs, specific differentiation conditions were used to trigger cell differentiation into adipocytes, chondrocytes, and osteocytes as described. 17

| Migration capacity
The migratory capacity of AMSCs was determined using a Transwell system (8-μm polycarbonate membrane; Corning, New York, New York), as described. 19 Briefly, lower chambers were filled with DMEM high glucose 0.2% BSA and cells (8 × 10 4 ) suspended in the same medium were added to the upper chambers. Transwells were then incubated for 24 hours at 37 C. Cells that migrated to the lower chamber were then collected and counted using the BioRad TC 10 Automated Cell Counter (BioRad, Hercules, California).

| Invasion capacity
The invasive capacity of AMSCs was determined as for the migration assay except that Transwell membranes were firstly coated with Matrigel (Corning) in DMEM high glucose 0.2% BSA for 2 hours at 37 C. Cells (8 × 10 4 ) were added to the upper chambers and incubated for 24 hours at 37 C and those ones that invaded into the lower chamber were collected and counted as above.

| Stimulation experiments
AMSCs from control pregnant women were cultured in 6-well plates (5 × 10 4 ) and allowed to attach for 24 hours. Cells were then stimulated for 24 hours with glucose, insulin and/or palmitic acid

| Cell proliferation
Proliferation rate of AMSCs was determined by standard colorimetric 3-(4,5-dimethylthiazol-2-y1)-2,5-diphenyltetra-zolium bromide (MTT) incorporation experiments. Cells (1.6 × 10 3 ) were cultured in 96-well plates and allowed to attach for 24 hours. A MTT assay at day 1 was performed to count the initial number of cells. After 5 days, a second MTT assay was performed (day 7) and the difference in absorbance between day 7 and day 1 was considered the proliferation rate. In the AMSCs stimulation studies, the proliferation rate was measured at 24 hours after the addition of the stimuli. Absorbance was measured by spectrophotometry at 540 nm in all cases.

| Prostaglandin E 2 determination
Prostaglandin E 2 (PGE2) concentrations were measured, at 24 hours, in the conditioned medium of AMSCs by ELISA (R&D Systems) following the manufacturer's instructions.  The macroscopic study of the placentas failed to find difference between GDM and control groups. No differences were observed in placental weight and area (Table 1), and the umbilical cord was marginally inserted in most of the placentas studied (7/9 of the control group and in 5/9 of the GDM group). However, the histological study of placental tissue showed differences in the distribution of collagen deposition in villous stroma between groups. Staining quantification indicated lower collagen deposition on the maternal side of placentas obtained from GDM women (P = .0173) whereas it was higher on the fetal side compared the controls (P = .0067) (Figure 1). These results indicate that a diabetic environment modifies the collagen deposition within the placenta, highlighting the presence of fibrosis on the placenta's fetal side, which is characterized by an excessive connective tissue accumulation in response to tissue injury and inflammation.     Figure 3B). At the same time, these findings are accompanied by an increase in GDM-AMSCs migratory and invasive capacities as compared to control cells ( Figure 3C).

MCP-1 is a potent chemoattractant and PGE2 has been involved
in the anti-inflammatory and immunosuppressive capacities of the AMSCs. According to these findings, the migration of monocytes and T-lymphocytes was significantly higher when exposed to the conditioned medium from GDM-AMSCs compared to those from control cells ( Figure 3D) and this increased chemotactic capacity was significantly reversed when we treated the conditioned medium of GDM-AMSCs with an anti-MCP-1 specific antibody ( Figure 3E). In  Figure 4A). Moreover, AMSCs exposed to the three insults showed a significantly higher migratory capacity and an increased chemotactic activity over monocytes compared with the unstimulated ones ( Figure 4B,C). Finally, in order to make sure that these results are specifically caused by the sum of the three stimuli tested and not by cell damage or death, we performed a MTT proliferation assay and found no significant differences in the proliferative capacity of the differentially stimulated AMSCs ( Figure 4D). These data confirm that GDM adverse nutritional and metabolic environment alter the inflammatory profile of the amniotic membrane-derived fetal precursors.

| Biological properties of fetal precursor cells are related to maternal and infant clinical and metabolic parameters
Finally, we explored the potential relationship between maternal clinical and analytical parameters and AMSCs phenotype, and also whether their functional characteristics were associated with offspring's anthropometric and metabolic parameters. MCP-1 expression levels were correlated with maternal biomarkers associated with adverse perinatal and long-term outcomes such as prepregnancy BMI and HOMA-IR, as well as with cord blood concentrations of triglycerides and insulin. A positive association was also observed between CD40 and CTSS with GCT and prepregnancy BMI (Table 3).
Next, we analyzed the groups separately and regarding the con- and prepregnancy BMI (r: 0.644; P = .061) followed a trend. The associations between CD40 with maternal prepregnancy BMI (r: 0.683; P = .042) and GCT (r: 0.800; P = .010) remained significant. Overall, our data indicate that maternal metabolic phenotype during pregnancy could determine the biological characteristics of fetal precursors, which may be linked to metabolic characteristics in the offspring.
These results become stronger in the context of GDM.

| DISCUSSION
It has been proposed that signatures of metabolic deregulation remain in cells early in development. 20  in AMSCs from GDM women, a key enzyme involved in the PGE2 synthesis pathway.
A pro-inflammatory state is known to be associated with the development of insulin resistance, obesity, type 2 diabetes, and atherosclerosis. 30 TNFα expression is increased in placental tissue obtained from pregnancies complicated by GDM 31 and it is also released by the placenta under hyperglycemic conditions. 32 MCP-1 is a key element involved in the modulation and recruitment of macrophages, participates in the induction of fat tissue inflammation in type 2 diabetes and obesity 33 but also seems to have a pivotal role in adipogenesis. 34 We have observed an increased expression of MCP-1 in AMSCs isolated from diabetic mothers along with an increased chemotactic capacity over monocytes and T-lymphocytes through their conditioned medium.
These phenomena may participate in the onset of placental inflammation that accompanies GDM. In addition to TNFα and MCP-1, AMSCs isolated from GDM women showed an upregulation of genes involved in the inflammatory and immune response such as CD40 and CTSS.
Cathepsin is a cysteine protease implicated in the regulation of inflammatory activity and it has been postulated as a potential biomarker of the development of insulin resistance and type 2 diabetes. 35 In agreement with epidemiological data, we observed an upregulation of CTSS, Our in vitro results confirm that AMSCs exposed to high concentrations of insulin, glucose and palmitic acid, simulating the hyperinsulinemia, hyperglycemia, and dyslipidemia observed in metabolic syndrome 40 but also in pregnancies complicated by diabetes, 41 reproduced the inflammatory expression profile, the migration capacity, and the chemotactic activity observed in GDM-derived AMSCs.
Although we did not observe changes in the gene expression profile of control AMSCSs when exposed to insulin or glucose separately, a trend was seen when they were treated with palmitic acid. These results became evident when we treated them with a combination of the three stimuli, suggesting that saturated fatty acids were the main drivers of the metabolic activation but this effect was only significant in the context of hyperglycemia and hyperinsulinemia.
According to the fetal programming hypothesis, 42 a disturbed metabolic environment could permanently affect the health of offspring exposed and predispose them to obesity and/or type 2 diabetes. GDM has been associated with an increased risk of obesity, type 2 diabetes, and cardiovascular disease in offspring, but the underlying pathogenic mechanisms are unclear. Despite GDM's association with macrosomia and excessive fat accretion, we found similar birth weight and adiposity in GDM offspring compared with control offspring. MCP-1 gene expression was positively correlated with both cord blood insulin and triglyceride concentrations, but not with birth weight or newborn adiposity. The lack of association between gene expression profiling and neonatal anthropometric parameters could be due, in part, to the sample size, the effect of nutritional therapy which normalizes fetal growth, 43 and/or other factors besides hyperglycemia. 44 On the other hand, we reported that maternal BMI and insulin resistance are related to gene expression of chemokines implicated in the modulation of migration and inflammation, but also involved in adipose tissue dysfunction and cardiovascular diseases. The upregulation of MCP-1 and CD40 by the adverse maternal environment seems to be specific to pregnancies complicated by GDM and suggests that maternal nutritional and metabolic status could induce changes in the placental gene expression profile even when they are adequately controlled. It suggests that normalizing birth weight and metabolic control is unable to revert the placental insult. Based on this evidence, we speculate that GDM environment could program stem cells and subsequently favor metabolic dysfunction later in life.

| CONCLUSION
For the first time, we demonstrate that maternal metabolic derange-

CONFLICT OF INTEREST
The authors indicated no potential conflicts of interest.

DATA AVAILABILITY STATEMENT
The data sets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.