The Glycosyltransferase EOGT Regulates Adropin Expression in Decidualizing Human Endometrium

Abstract

In pregnancy, resistance of endometrial decidual cells to stress signals is critical for the integrity of the fetomaternal interface and, by extension, survival of the conceptus. O-GlcNAcylation is an essential posttranslational modification that links glucose sensing to cellular stress resistance. Unexpectedly, decidualization of primary endometrial stromal cells (EnSCs) was associated with a 60% reduction in O-linked β-N-acetylglucosamine (O-GlcNAc)‒modified proteins, reflecting downregulation of the enzyme that adds O-GlcNAc to substrates (O-GlcNAc transferase; OGT) but not the enzyme that removes the modification (O-GlcNAcase). Notably, epidermal growth factor domain–specific O-linked GlcNAc transferase (EOGT), an endoplasmic reticulum-specific OGT that modifies a limited number of secreted and membrane proteins, was markedly induced in differentiating EnSCs. Knockdown of EOGT perturbed a network of decidual genes involved in multiple cellular functions. The most downregulated gene upon EOGT knockdown in decidualizing cells was the energy homeostasis–associated gene (ENHO), which encodes adropin, a metabolic hormone involved in energy homeostasis and glucose and fatty acid metabolism. Analysis of midluteal endometrial biopsies revealed an inverse correlation between endometrial EOGT and ENHO expression and body mass index. Taken together, our findings revealed that obesity impairs the EOGT-adropin axis in decidual cells, which in turn points toward a mechanistic link between metabolic disorders and adverse pregnancy outcome.

During the midluteal phase of the menstrual cycle, the endometrium becomes transiently poised to transit from a cycling into a semipermanent tissue that is maintained throughout pregnancy (1). During this window, the luminal endometrial epithelial cells acquire a receptive phenotype, and the underlying stromal cells start to differentiate into secretory decidual cells. After breaching of the luminal epithelium, migratory decidual cells rapidly encapsulate the implanting embryo (2) and form a nutritive and immune-privileged matrix that enables trophoblast invasion and placenta formation (3). Once the process of interstitial and endovascular trophoblast invasion begins, the placental-maternal interface is intensely remodeled and exposed to profound fluctuations in oxygen tension associated with changes to the vascular tree (1, 4). Decidual cells are programmed to resist a range of stressors, thus ensuring integrity of the interface and survival of the conceptus. Several molecular mechanisms underpin this quasi-autonomous state of decidual cells, including silencing of circadian gene expression (5), inhibition of stress pathways such as c-Jun N-terminal kinase (4, 6), attenuated inositol trisphosphate signaling (7), global cellular hypoSUMOylation (8), resistance to microRNA-mediated gene silencing through loss of argonaute proteins (9), and marked upregulation of free radical scavengers (10).

Posttranslational modification of proteins with O-linked β-N-acetylglucosamine (O-GlcNAc) is an integral component of the cellular stress response (11, 12). O-GlcNAcylation refers to the covalent addition of a GlcNAc sugar moiety to hydroxyl groups of serine and/or threonine residues of cytosolic, nuclear, and mitochondrial proteins. The O-GlcNAc transferase (OGT) transfers the O-GlcNAc moiety from uridine diphosphate (UDP)–GlcNAc to target proteins, whereas O-GlcNAcase (OGA) removes O-GlcNAc from proteins. UDP-GlcNAc is an end product of the nutrient-dependent hexosamine biosynthetic pathway (HBP), a branch pathway in glycolysis. Increased glucose flux through the HBP elevates UDP-GlcNAc and drives increased cellular O-GlcNAcylation (13, 14). OGT targets in excess of 3000 proteins (15), enabling it to regulate multiple processes, including signal transduction and transcription, in a manner akin to—and cooperative with—protein phosphorylation (11, 16). Of note, increased O-GlcNAcylation is important for cell survival in response to a variety of stressors, including osmotic (12, 17), oxidative (18), genotoxic (12, 19, 20), endoplasmic reticulum (ER) (21), and hypoxia/reoxygenation stress (21, 22).

In addition to OGT, a second enzyme has been identified that catalyzes the transfer of GlcNAc from UDP-GlcNAc to epidermal growth factor (EGF) repeats of extracellular proteins (23, 24). By contrast to OGT, this glycosyltransferase, termed EGF domain–specific O-linked GlcNAc transferase (EOGT), resides in the ER and targets seemingly only a very limited number of secreted and membrane receptors, including Notch receptors (25–27).

In this study, we examined the expression of OGT, OGA, and EOGT upon decidual transformation of primary endometrial stromal cell (EnSCs). Although increased O-GlcNAcylation has been implicated in stress resistance, decidualization was associated with a marked reduction in O-GlcNAc‒modified proteins, reflecting downregulation of OGT but not OGA. However, EOGT expression in differentiating EnSCs was increased. Although the EOGT target proteins in decidual cells remain to be determined, we demonstrated that EOGT knockdown perturbs the expression of numerous genes, most prominently energy homeostasis–associated gene (ENHO), which encodes the newly discovered metabolic hormone adropin, which regulates lipid metabolism, confers insulin sensitivity, and protects against vascular disease (28, 29). Finally, we demonstrated that obesity, a major risk factor for reproductive failure, is associated with lower midluteal endometrial EOGT and adropin expression.

Methods

Patient selection and endometrial sampling

The study was approved by the National Health Service National Research Ethics – Hammersmith, Queen Charlotte’s & Chelsea Research Ethics Committee (1997/5065). Endometrial samples were obtained during the luteal phase of an ovulatory, nonhormonally stimulated menstrual cycle using a Wallach Endocell^TM sampler, starting from the uterine fundus and moving downward to the internal cervical ostium. Written informed consent was obtained from all participants in accordance with the guidelines in The Declaration of Helsinki 2000. A total of 193 biopsies were used in this study, including 24 fresh endometrial biopsies processed for primary culture. In addition, 112 biopsies stored in RNAlater (Sigma-Aldrich) were used to measure messenger RNA (mRNA) expression, and a further 57 snap-frozen and formalin-fixed biopsies were used for Western blot analysis and immunohistochemistry, respectively. All endometrial biopsies were timed between 6 and 10 days after the preovulatory luteinizing hormone (LH) surge. Demographic details are summarized in Supplemental Table 1. None of the subjects had received hormonal treatment for at least 3 months before the procedure.

Primary cell culture

EnSCs were isolated and established from endometrial tissues as described previously (30). Confluent EnSC monolayers were decidualized in Dulbecco’s modified Eagle medium /F-12 containing 2% dextran-coated charcoal‒fetal bovine serum with 0.5 mM 8-bromo-cyclic adenosine monophosphate (cAMP; Sigma-Aldrich) and 10⁻⁶ M medroxyprogesterone acetate (MPA; Sigma-Aldrich) to induce a differentiated phenotype. Culture medium was refreshed every 48 hours. All experiments were carried out before the third cell passage.

Transient transfections

Primary EnSCs were transfected with small interfering RNA (siRNA) using the jetPRIME Polyplus transfection kit (VWR International). Undifferentiated EnSCs were transiently transfected with 50 nM EOGT-siGENOME SMARTpool or siGENOME Non-Targeting siRNA Pool 1 (GE Healthcare). Transfection studies were performed in triplicate and repeated on primary cultures from four subjects.

Real-time quantitative polymerase chain reaction

Total RNA was extracted from EnSC cultures using RNA STAT-60 (AMS Biotechnology). Equal amounts of total RNA were treated with DNase and reverse transcribed using the QuantiTect Reverse Transcription Kit (QIAGEN), with the resulting complementary DNA used as a template in real-time quantitative polymerase chain reaction (qRT-PCR) analysis. Detection of gene expression was performed with Power SYBR^® Green Master Mix (Life Technologies) and the 7500 Real Time PCR System (Applied Biosystems). The expression levels of the samples were calculated using the Δ cycle threshold method, incorporating the efficiencies of each primer pair. The variances of input complementary DNA were normalized against the levels of the L19 housekeeping gene. All measurements were performed in triplicate. Melting curve analysis confirmed product specificity.

Western blot analysis

Protein extracts were prepared by lysing cells in RIPA buffer containing protease inhibitors (cOmplete, Mini, EDTA-free; Roche). Protein yield was quantified using the Bio-Rad Protein Assay Dye Reagent Concentrate. Equal amounts of protein were separated by sodium dodecyl sulfate–polyacrylamide gel electrophoresis before wet-transfer onto nitrocellulose membrane. Global O-GlcNAcylation was determined by spotting 10 μg and 2 μg of total protein lysate directly onto nitrocellulose membranes. Nonspecific binding sites were blocked by overnight incubation with 5% nonfat dry milk in Tris-buffered saline with 1% Tween (TBS-T; 130 mmol/L NaCl, 20 mmol/L Tris, pH 7.6, and 1% Tween). The antibodies used in this study are listed in Table 1. Protein complexes were visualized with ECL Plus chemiluminescence. Densitometry was performed using Gene Tools software.

Table 1.

Primary Antibodies Used Throughout the Study

Protein Target	Manufacturer, Catalog No.	Dilution	RRID
OGT	Cell Signaling Technology, 24083	WB 1:1000	AB_2716710
OGA	Sigma-Aldrich, SAB4200267	WB 1:500	AB_10797267
EOGT	Sigma-Aldrich, HPA019460	WB 1:500; IHC 1:500	AB_1844628
O-GlcNAc (CTD110.6)	Sigma-Aldrich, O7764	WB 1:500	AB_1079524
NOTCH1	Cell Signaling Technology, 3608	WB 1:1000	AB_10691684
NOTCH3	Cell Signaling Technology, 5276	WB 1:1000	AB_10560515
Adropin	Abcam, ab122800	WB 1:500; IHC 1:300	AB_11132112
β-Actin	Abcam, ab8227	WB 1:50000	AB_2305186

Abbreviations: IHC, immunohistochemistry; RRID, Research Resource Identifier; WB, Western blot.

Immunohistochemistry

Paraffin-embedded, formalin-fixed endometrial specimens were immunostained for EOGT using the Novolink polymer detection systems (Leica) per manufacturer’s instructions. Universal LSAB Plus kits (DAKO) were used as previously described (31) with primary antibodies against EOGT (1:500 dilution) and ENHO (1:200 dilution). Bright-field images were obtained on a Mirax Midi slide scanner and visualized using Pannoramic Viewer software for analysis.

RNA sequencing and data analysis

Total RNA was extracted using RNA-STAT-60 from primary EnSC cultures first transfected with either EOGT or nontargeting (NT) siRNA and then decidualized with 8-br-cAMP and MPA for 4 days. Three biological repeats were performed to allow for interpatient variability. RNA quality was analyzed on an Agilent 2100 Bioanalyzer. RNA integrity number score for all samples was ≥8.0. Transcriptomic maps of paired-end reads were generated using Bowtie-2.2.3, SAMtools 0.1.19, and TopHat 2.0.12 against the University of California, Santa Cruz, hg19 reference transcriptome (2014) from the Illumina iGenomes resource using the fr-firststrand setting. Transcript counts were assessed by HTSeq-0.6.1. Transcripts per million were calculated as recently described (32). Differential gene expression analysis was performed using DEseq2-1.14.1. Significance was defined as an adjusted P value (q value) of <0.05 after Benjamini-Hochberg false discovery rate correction. Expression data have been submitted to the Gene Expression Omnibus (GEO) repository (accession number: GSE104720). Gene Ontology (GO) analyses were carried out using DAVID Bioinformatics Resources 6.8 (33, 34) and visualized using REVIGO online software (35). GO Term Gene Set Enrichment Analysis (GSEA) was performed using piano R package (36). Briefly, GO ID was extracted for each Ensembl gene ID using biomaRt package in R (37). Gene ID and GO ID were loaded into correct format using the load GSC function, and GSEA was performed using the runGSA function. Genes were ranked according to the adjusted P value, and log2-fold change was used to determine upregulated or downregulated transcripts.

Statistical analysis

In vitro experiments were analyzed with the statistical package Graphpad Prism 6. Unpaired Student t test and one-way analysis of variance with Tukey post hoc test were used when appropriate. The association between EOGT and ENHO mRNA in endometrial biopsies and body mass index (BMI) was analyzed using Pearson rank correlation. Statistical significance was assumed when P < 0.05.

Results

Loss of OGT-dependent O-GlcNAcylation in decidualizing EnSCs

O-GlcNAcylation of target proteins is enhanced in response to diverse stress signals and tissue injury (15). To test whether O-GlcNAcylation plays a role in decidualization, primary EnSCs were decidualized with 8-br-cAMP and MPA for 2, 4, or 8 days, and the expression of O-GlcNAc‒processing enzymes, OGT, EOGT, and OGA, were examined at both mRNA and protein levels. Analysis of four independent primary cultures demonstrated that decidualization resulted in downregulation of the canonical O-GlcNAc transferase OGT at both mRNA and protein levels (Fig. 1A and 1B), whereas expression of OGA (encoded by MGEA5) was unchanged. In contrast to OGT, expression of EOGT increased significantly upon decidualization. In fact, induction of EOGT was more marked at the protein than the mRNA level, with levels increasing approximately fivefold after 8 days of decidualization (Fig. 1A and 1B). Densitometric analyses of Western blots are shown in Supplemental Fig. 1.

Figure 1.

Regulation of O-GlcNAcylation in decidualizing EnSCs. (A) OGT, OGA, and EOGT transcript levels were measured in undifferentiated EnSCs, and cells were decidualized with 8-br-cAMP and MPA for 2, 4, or 8 days. Expression was normalized to L19 mRNA levels, and the data show mean ± standard error of the mean relative to that in undifferentiated cells in four independent primary cultures. Different letters above the error bars indicate that groups are significantly different from each other at P < 0.05. Group comparison by analysis of variance (ANOVA) and Tukey post hoc test. (B) Total protein lysates from parallel cultures were subjected to Western blotting. β-Actin served as a loading control. (C) Representative dot blot of total cellular O-GlcNAcylation from total protein lysates from cells decidualized for 2, 4, or 8 days. Lower panel denotes total cellular O-GlcNAcylation relative to β-actin by densitometry and expressed as arbitrary units (A.U.s). Group comparison by ANOVA and Tukey post hoc test. EnSC, endometrial stromal cell.

Although thousands of OGT substrates have been identified, only a handful of EOGT targets have been described to date (38, 39). To determine the effect of decidualization on total cellular O-GlcNAcylation, protein lysates from undifferentiated EnSCs and cells decidualized for 8 days were subjected to dotblot analysis using a primary antibody directed against serine and threonine residues with attached β-O-linked GlcNAc. This analysis revealed ∼60% reduction in global O-GlcNAcylation in decidual cells (Fig. 1C), reflecting the relative shift to OGA over OGT. Thus, decidualization was associated with decreased OGT-mediated cellular O-GlcNAcylation but increased expression of EOGT, a highly selective transferase that targets secreted and membrane-bound proteins (39).

EOGT expression in midluteal endometrium

Mining of the Genotype-Tissue Expression and FANTOM (Functional Annotation of Mammalian Genomes) projects revealed that EOGT was highly expressed in the endometrium compared with other tissues (40, 41). Furthermore, analysis of GEO (profile ID: 24476716) demonstrated that EOGT mRNA levels in cycling endometrium increased sharply upon transition from the early to the midsecretory endometrium (Fig. 2A). Laser microdissection of glandular endometrial epithelium coupled to RNA sequencing revealed a transient threefold increase in EOGT mRNA levels during the midluteal phase, coinciding with the putative window of implantation (Fig. 2B) (42). Immunohistochemistry was performed to assess the spatiotemporal expression of EOGT in the endometrial stromal compartment. In timed early‒secretory phase (LH + 5) biopsies, EOGT immunoreactivity was largely confined to endometrial glands (Fig. 2C, upper panel). During the midluteal implantation window (LH + 9), stromal cells were strongly EOGT positive (Fig. 2C, lower panel). Interestingly, EOGT was also expressed in endothelial cells lining the emerging terminal spiral arteries, although the surrounding perivascular cells often appeared devoid of this glycosyltransferase. Thus, EOGT was expressed in the endometrial epithelial compartment, decidualizing stromal cells and vascular endothelial cells at the time of embryo implantation.

Figure 2.

EOGT expression in midluteal endometrium. (A) EOGT transcripts, expressed as arbitrary units (A.U.s) in proliferative (P), early secretory (ES), midsecretory (MS), and late secretory (LS) endometrium. The data were derived from in silico analysis of GDS2052 microarray data. (B) Expression of EOGT, in transcripts per million (TPM), in ES, MS, and LS endometrial glandular epithelium. The data were derived from in silico analysis of published RNA-Seq data and show mean ± standard error of the mean (42). Different letters above the error bars indicate groups are significantly different from each other at P < 0.05. Group comparison by analysis of variance and Tukey post hoc test. (C) EOGT immunohistochemistry of midluteal endometrial biopsies obtained 5 and 9 days after the LH surge, representing the early and midluteal phases, respectively. Inset in the upper panel shows no staining upon omission of primary antibody (negative control). Arrowheads indicate positive staining of endothelial cells. Scale bar = 50 μm.

Impact of EOGT knockdown on decidual marker genes and Notch signaling

Induction of decidual marker genes, such as PRL and IGFBP1, in response to cAMP and progestin signaling is mediated, at least in part, by the auto/paracrine actions of a host of cytokines and morphogens (1). We speculated that the strong induction of EOGT was essential for the expression of decidual marker genes in differentiating EnSCs. To test this conjecture, four primary cultures were first transfected with NT or EOGT siRNA and then decidualized with 8-br-cAMP and MPA for 2, 4, or 8 days. Although EOGT knockdown was highly efficient (Fig. 3A, leftmost panel), there was no significant effect on the induction of either PRL or IGFBP1 in decidualizing cultures (Fig. 3A, middle and rightmost panels, respectively).

Figure 3.

EOGT knockdown in decidualizing EnSCs. (A) Four independent primary cultures were transfected with either nontargeting siRNA (siNT) or siRNA targeting EOGT (siEOGT). After 24 hours, the cultures remained either undifferentiated or were decidualized for 2, 4, and 8 days. Total mRNA was subjected to qRT-PCR analysis to determine transcript levels for EOGT (left panel), PRL (middle panel), and IGFBP1 (right panel). Expression was normalized to L19 mRNA levels, and data show mean ± standard error of the mean (SEM) fold change relative to undifferentiated EnSCs. **P < 0.01. Group comparison by unpaired t test. (B) Western blot analysis of EOGT, NOTCH1, and NOTCH3 intracellular domain levels in total protein lysates from EnSCs first transfected with siNT or siEOGT and then decidualized for the indicated time points. β-Actin served as a loading control. Densitometric analyses of NOTCH1 and NOTCH3 expression are shown in the middle and right panels, respectively. (C) Transcript expression of HES1 and HEY1, normalized to L19 mRNA, from three independent primary EnSC cultures transfected with siNT or siEOGT and decidualized for 4 days. Data are mean ± standard error of the mean of three biological replicates. P > 0.05 (unpaired t test). A.U., arbitrary unit.

Notch receptors are perhaps the best-characterized targets of EOGT (27). O-GlcNAc modification of EGF-like repeats of NOTCH1 enhances signaling by potentiating interaction with Delta-like (DLL) 1 (DLL1) and DLL4 ligands in a cell-specific context (27). Ligand binding to the extracellular domain of Notch receptors induces proteolytic cleavage and releases Notch intracellular domain, which enters the cell nucleus to regulate gene expression. To test whether induction of EOGT in decidualizing EnSCs modulates Notch signaling, total protein lysates of undifferentiated cells and cells treated with 8-br-cAMP and MPA for 2, 4, or 8 days were subjected to Western blot analysis for NOTCH1 and NOTCH3 intracellular domains. As shown in Fig. 3B, decidualization was associated with gradual silencing of Notch signaling, and EOGT knockdown had no discernable effect on this response. Further, expression of HEY1 and HES1, target genes of the canonical Notch signaling pathway (43), was not significantly altered upon EOGT knockdown in EnSCs decidualized for 4 days (P > 0.05) (Fig. 3C). Taken together, these observations indicate that Notch activity was not likely regulated by EOGT-mediated O-GlcNAcylation in decidualizing cells.

EOGT knockdown perturbed decidual gene expression

To gain insight into the role of EOGT in decidual cells, total RNA harvested from three independent cultures, first transfected with either EOGT or NT siRNA and then treated with 8-br-cAMP and MPA for 4 days, was subjected to RNA sequencing. Approximately 26 to 36 million paired-end reads were sequenced per sample. After accounting for variations between primary cultures, the effect of EOGT knockdown on decidual gene expression was highly consistent with principal components 1 and 2, accounting for 52% and 36% of variance in gene expression, respectively (Fig. 4A). Based on q ≤ 0.05, we identified 340 genes that were significantly altered upon EOGT knockdown (Fig. 4B), of which 178 (52%) were upregulated and 162 (48%) were downregulated. Several highly induced decidual genes were downregulated significantly upon EOGT knockdown, including LEFTY2 (q = 3.13 × 10⁻⁴), CDKN1C (q = 5.10 × 10⁻⁸), GADD45G (q = 9.58 × 10⁻⁹), and GPX3 (q = 7.17 × 10⁻¹²) (Supplemental Fig. 2). EOGT knockdown also downregulated ESR1, coding the estrogen receptor α, in decidualizing cells (q = 1.07 × 10⁻³). However, the most repressed gene upon EOGT knockdown was ENHO (−2.03 log2-fold change; q = 5.67 × 10⁻¹¹), coding adropin, a recently discovered peptide hormone implicated in the regulation of energy homeostasis, insulin resistance, and lipid metabolism (28). Interestingly, IL1RL1, which encodes the interleukin (IL)-33 receptor, is strongly upregulated upon decidualization (44), yet EOGT knockdown amplified induction of this gene in differentiating EnSCs (1.8 log2-fold change; q = 2.31 × 10⁻¹⁷).

Figure 4.

EOGT knockdown perturbed distinct gene networks in EnSCs. (A) Principal component analysis of RNA-Seq data from three independent primary EnSC cultures first transfected with siNT and siEOGT and then decidualized for 4 days. (B) Clustered heatmap of RNA-seq data for differentially expressed transcripts between siEOGT and siNT transfected cultures. (C) Semantic clustering of significantly overrepresented GO terms (P < 0.05) of differentially upregulated and downregulated genes (left and right panels, respectively) upon EOGT knockdown. Circle size is proportional to the percentage of genes in the GO term, whereas the color indicates the log10 P value. Color key is on the right. (D) Word cloud presentation of overrepresented disease associations based on GO terms upon EOGT knockdown in decidualizing EnSCs. Size of the word is proportional to log10 q value. A.U., arbitrary unit; siEOGT, siRNA targeting EOGT; siNT, nontargeting siRNA.

GO term enrichment analysis, using both DAVID (Fig. 4C) and GSEA (Supplemental Fig. 3), revealed that EOGT knockdown resulted in upregulation of genes involved—among other categories—in cell adhesion, extracellular matrix organization, and signal transduction (Fig. 4C, left panel, and Supplemental Fig. 3; Supplemental Table 2). Notable GO terms enriched in downregulated genes include oxidative-reductive process, response to estrogen/estradiol, and inflammatory responses (Fig, 4C, right panel, and Supplemental Fig. 3; Supplemental Table 2). We also annotated genes perturbed upon EOGT knockdown by their disease association. GO analysis yielded a conspicuous association between EOGT-responsive decidual genes and vascular and metabolic disorders, most prominently type 2 diabetes (Fig. 4D).

Obesity perturbed the endometrial EOGT-adropin axis

To explore the putative link with metabolic disorders, we measured EOGT transcript level by qRT-PCR in 112 midluteal (LH + 7 to 10) endometrial biopsies. Demographic details are summarized in Supplemental Table 1. Interestingly, endometrial EOGT mRNA levels correlated inversely with BMI (Pearson r = −0.194; P = 0.043) (Fig. 5A). By contrast, no association was found between either OGT or OGA mRNA levels and BMI (Supplemental Fig. 4). Western blot analysis of total protein lysates of 48 biopsies (LH + 7 to 9) substantiated the inverse correlation between endometrial EOGT levels and BMI (r = −0.335; P = 0.02) (Fig. 5B), with levels significantly lower in patients who were clinically obese compared with control subjects (P < 0.03) (Fig. 5C).

Figure 5.

Obesity perturbed endometrial EOGT expression. (A) Pearson correlation (r) between normalized EOGT transcript levels in midluteal endometrial biopsies and BMI (kg/m²) in 112 subjects. (B) Pearson correlation between normalized EOGT protein expression and BMI in midluteal endometrial biopsies from 48 subjects. (C) Normalized EOGT protein levels in timed endometrial biopsies obtained from women with BMIs <25, 25 to 30 (overweight), and >30 (obese). *P < 0.05 (unpaired t test). A.U., arbitrary unit.

Expression of adropin in the endometrium has not yet been reported. As shown in Fig. 6A, ENHO mRNA levels increased significantly in primary EnSCs decidualized with 8-br-cAMP and MPA for 4 days, although the level of induction varied markedly between primary cultures (Fig. 6A and Supplemental Fig. 5). Furthermore, immunohistochemistry of serial endometrial sections (LH + 8) revealed that the tissue distribution of adropin was indistinguishable from that of EOGT, characterized by strong expression in glands, differentiating stromal cells, and endothelial but not perivascular cells (Fig. 6B). Furthermore, a strong positive correlation was observed between EOGT and ENHO transcript levels in 112 timed endometrial biopsies (r = 0.327; P = 0.0003) (Fig. 6C), as well as a negative correlation between ENHO mRNA expression and BMI (r = −0.178; P = 0.044) (Fig. 6D).

Figure 6.

Endometrial ENHO expression related to BMI. (A) ENHO mRNA levels were measured in undifferentiated EnSCs and cells decidualized with 8-br-cAMP and MPA for 4 days (upper panel). Expression was normalized to L19 mRNA levels, and the data show fold change relative to expression in undifferentiated cells in six independent primary cultures Total protein lysates from parallel cultures were subjected to Western blotting (lower panel). β-Actin served as a loading control. **P < 0.01 (unpaired t test). (B) Immunohistochemistry of EOGT (left panel) and adropin (right panel) on serial tissue sections, demonstrating colocalization of both proteins in midluteal endometrium. Inset in the left panel shows no staining upon omission of primary antibody (negative control). Arrowheads indicate EOGT and adropin immunoreactivity in endothelial cells of terminal spiral arteries. Scale bars = 50 μm. (C) Pearson correlation between EOGT and ENHO transcripts in endometrial biopsies from 112 subjects. (D) Pearson correlation between ENHO mRNA levels in midluteal endometrial biopsies and BMI in 112 subjects. *P < 0.05; ***P < 0.001.

Discussion

Dynamic changes in protein O-GlcNAcylation enable cells to homeostatically balance energy supply and demand by modulating the stability, localization, and function of a myriad of proteins (45). Here, we report that decidualizing EnSCs downregulated OGT expression and intracellular O-GlcNAcylation but upregulated the highly selective glycosyltransferase EOGT. Increased canonical O-GlcNAcylation is a well-characterized prosurvival response (15), rendering the downregulation of OGT in differentiating EnSCs counterintuitive, especially as initiation of decidual differentiation coincides with a burst of endogenous reactive oxygen species production and release of various inflammatory mediators (46, 47). However, small ubiquitinlike modifier modification of proteins is also dramatically reduced in decidualizing cells and uncoupled from c-Jun N-terminal kinase‒mediated stress signaling though the induction of mitogen-activated protein kinase phosphatase 1 (4, 6, 8). Hence, by silencing selective pathways that converge on the posttranslational modification code of numerous proteins, decidual cells appear to prioritize cellular homeostasis over an adaptive response to stress signals. Further, recent studies have shown that decidualization is critically dependent on glucose utilization via the pentose phosphate pathway (48), suggesting that loss of OGT may be integral to the metabolic reprogramming of the endometrium in preparation for pregnancy.

Induction of EOGT in the endometrium during the midluteal phase of the cycle coincides with the window of implantation. At this time, EOGT is expressed in the glandular epithelium, vascular endothelial cells, and stromal cells that are poised to decidualize. Decidualization is characterized by an unfolded protein response that underpins ER expansion and acquisition of a secretory phenotype (49). In fact, multiple secreted factors, including IL-11, leukemia inhibitory factor, and bone morphogenetic protein 2, have been implicated in the auto/paracrine propagation of the decidual response (1). However, the identity and role of EOGT target proteins, whether secreted or expressed on the cell surface, in differentiating EnSCs requires further investigation. We showed that Notch signaling was attenuated upon decidualization (50), irrespective of EOGT knockdown. Furthermore, other known EOGT target proteins (26), including thrombospondin (THBS1), peptidase domain containing associated with muscle regeneration 1 (PAMR1), and laminin alpha 5 (LAMA5), are also downregulated upon decidualization, at least at the mRNA level (GEO accession number: GSE104720).

Although EOGT knockdown did not significantly affect PRL or IGFBP1 expression in differentiating cells, RNA sequencing uncovered a robust set of 341 EOGT-dependent genes. EOGT knockdown upregulated several genes encoding inflammatory mediators, including IL-1β (IL1B) and complement component 3 (C3), but downregulated key genes involved in decidual stress defenses, such as GPX3 (coding extracellular glutathione peroxidase), GLXR (glutaredoxin), and GADD45G (growth arrest and DNA damage inducible gamma) (1). EOGT knockdown also blunted the induction of other cardinal decidual genes, including F3 (tissue factor) (1), LEFTY2 (left-right determination factor 2, also known as endometrial bleeding-associated factor, or EBAF) (51), and CDKN1C (cyclin-dependent kinase inhibitor 1C, p57^kip2) (52).

Most striking, however, was the repression of ENHO upon loss of EOGT. ENHO encodes adropin, a recently discovered peptide hormone implicated in energy homeostasis, glucose and fatty acid metabolism, and vascular protection (53). Although ENHO is expressed primarily in the liver, pancreas, and brain (28), we showed that this gene is also induced upon decidualization of human EnSCs, although the magnitude of induction varied markedly between primary cultures. We further showed a strong positive correlation between EOGT and ENHO transcript levels in whole endometrial biopsies; and immunohistochemistry on serial tissue sections revealed that the cellular distribution of adropin in midluteal endometrium was indistinguishable from EOGT.

GO analysis revealed a putative association between decidual genes perturbed upon EOGT knockdown and metabolic and cardiovascular disorders. To explore this possible link further, we measured the transcript levels of the three O-GlcNAc‒processing enzymes in 112 randomly selected midluteal endometrial biopsies from women ranging in BMI (kg/m₂) from 18 to 42. A weak but significant negative correlation was observed between BMI and EOGT mRNA levels but not OGT or OGA expression. Western blot analysis confirmed that obesity is associated with impaired endometrial EOGT expression. Furthermore, endometrial ENHO transcript levels correlated negatively with BMI.

Obesity increases the risk of a spectrum of pregnancy disorders, including obstetrical syndromes such as preeclampsia, fetal growth restriction, and preterm labor (54, 55), which are caused by impaired endovascular trophoblast invasion and spiral artery remodeling (56). In the absence of physiological remodeling, these uterine vessels are prone to develop acute atherosis, characterized by changes in lipid metabolism, intravascular inflammation, macrophage infiltration, and endothelial cell dysfunction (57). Adropin promotes various indices of vascular health, including increased endothelial cell proliferation, migration, and angiogenesis, and diminishes permeability and apoptosis (58). Although as yet untested, these observations suggest that adequate decidual adropin production may be essential for successful spiral artery remodeling in pregnancy. Notably, low circulating adropin levels have been associated not only with high BMI, insulin resistance, endothelial dysfunction, and coronary atherosclerosis but also with severe preeclampsia (59, 60).

In summary, the shift from OGT to EOGT dominance in decidualizing EnSCs resulted in intracellular hypo‒O-GlcNAcylation, whereas glucose utilization through the HBP for modification of selective secreted and/or membrane proteins was likely enhanced. We demonstrated that EOGT upregulation was critical for normal decidual function and identified ENHO as a major EOGT-responsive gene. Further, our observation that obesity impaired the EOGT-adropin axis in the endometrium intimates a mechanistic pathway that links metabolic disorders to vascular placental pathology and adverse pregnancy outcome.

Abbreviations:

 
• BMI

  body mass index

 
• cAMP

  cyclic adenosine monophosphate

 
• EGF

  epidermal growth factor

 
• ENHO

  energy homeostasis–associated gene

 
• EnSC

  endometrial stromal cell

 
• EOGT

  epidermal growth factor domain–specific O-linked GlcNAc transferase

 
• ER

  endoplasmic reticulum

 
• GEO

  Gene Expression Omnibus

 
• GO

  Gene Ontology

 
• GSEA

  Gene Set Enrichment Analysis

 
• HBP

  hexosamine biosynthetic pathway

 
• IL

  interleukin

 
• LH

  luteinizing hormone

 
• MPA

  medroxyprogesterone acetate

 
• mRNA

  messenger RNA

 
• NT

  nontargeting

 
• OGA

  O-GlcNAcase

 
• O-GlcNAc

  O-linked β-N-acetylglucosamine

 
• OGT

  O-GlcNAc transferase

 
• qRT-PCR

  real-time quantitative polymerase chain reaction

 
• siRNA

  small interfering RNA

 
• UDP

  uridine diphosphate.

Acknowledgments

We are grateful to all participating women. We also thank Drs. Siobhan Quenby and David Snead for valuable advice.

Financial Support: This study was funded by Diabetes UK (15/0005207) (to M.W.) and Tommy’s National Miscarriage Research Centre.

Disclosure Summary: The authors have nothing to disclose.