Multi-omics reveals a relationship between endometrial amino acid metabolism and autophagy in women with recurrent miscarriage^†

Abstract

Deterioration of the endometrial environment is an essential cause of recurrent miscarriage (RM). However, current studies in terms of endometrial amino acid metabolic characterization and autophagy are still inadequate. We tried to (1) identify the alternation in metabolite profiles in the RM endometrium; (2) investigate the expression of autophagy-related proteins in RM; and (3) elucidate the association between amino acid metabolism and autophagy in RM. Our results showed that glutamine metabolites were up-regulated in the endometrium of RM women. The levels of autophagy-associated proteins, LC3B, ATG12, and Beclin-1, were significantly higher in RM. Hemostasis, autophagy and IFNα signaling were the top three differentially activated signaling pathways between women with RM and normal pregnancy. Interestingly the expression of AMPK and GCN2 was significantly up-regulated in the endometrium of women with RM, and the same expression trend was also observed in the human endometrial stromal cells cultured in glutamine deprivation medium. Furthermore, inhibition of AMPK decreased the level of GCN2, indicating a positive correlation between GCN2 and AMPK. The expression of GCN2 was consistent with the expression of ATG12 and beclin-1; however, it was opposite to that of p62. Exposure to glutamine deprivation increased the level of LC3B, GCN2, ATG12, and beclin-1. Altogether, these findings suggested significant crosstalk between amino acid metabolism and autophagy. In summary, our data suggested that aberrant crosstalk between amino acid metabolism and autophagy may contribute to the impaired endometrial microenvironment of RM. Our study may provide new insight into the diagnosis of RM due to endometrial factors.

Introduction

Recurrent miscarriage (RM), the loss of three or more consecutive pregnancies, affects ~1–3% of reproductive couples [1–3]. Diverse factors, including uterine anomalies, chromosomal abnormalities, endocrine and immune defects, thrombophilia, and infections, could be associated with an increased risk of RM. With an increase in the number of miscarriages, maternal factors especially endometrial factors may become more and more involved in pregnancy failures. The endometrium is one of the essential components of the uterus and plays a pivotal role in reproduction and continuation. The physiological functions of the endometrium are preparation for implantation and maintenance of pregnancy if implantation occurs [4]. A tightly regulated interaction between the semi-allogeneic conceptus and the maternal receptive endometrium is responsible for the establishment and maintenance of pregnancy. Factors affecting the immune responses and metabolic environment of the endometrium may eventually affect the pregnancy.

The endometrium has the sophisticated metabolic capability to sustain nutrient supply to an embryo during development in the first trimester of pregnancy. Endometrial cells rewire metabolic pathways to adapt to their increased nutritional demands for energy, reducing equivalents, and cellular biosynthesis [5]. Endometrial regulation of amino acid metabolism is one of the most essential metabolic processes for satisfying those demands in preparation for implantation and early pregnancy [6]. Amino acid metabolism can not only provide components of proteins but also intermediate metabolites fueling multiple biosynthetic pathways [7], which contributes to establishing the metabolic microenvironment of the endometrium prepared for implantation and early pregnancy. Once amino acid metabolism is abnormal, especially the lack of amino acid, the production of AMP increases indirectly and activates AMPK [6]. Autophagy, a highly conserved catalytic and adaptive process, is induced to decompose the aging organelles and misfolded protein into amino acids, fatty acids, and other components under conditions of nutrient deprivations (especially amino acid starvation), ischemia, hypoxia or stress, and maintain cell homeostasis [8]. Eukaryotes are equipped with nutrient sensors and have developed a vital mechanism to maintain amino acid balance [9]. Among them, the general regulation of general control nonderepressible 2 (GCN2) can effectively sense the depletion of amino acid, which in turn activates cellular machinery to promote catabolic processes, including autophagy, to suppress the harmful effects of intracellular amino acid deficiency. However, the association between GCN2 regulation and autophagy in the endometrium of RM patients remains elusive. Some studies showed an association between placental defect and disruption of autophagy in miscarriage patients [10–12]. Moreover, studies have revealed that autophagy can increase the expression of matrix metalloproteinase (MMP), which may cause miscarriage [13, 14]. V Deretic reported that autophagy might cause miscarriage by affecting the maternal and fetal interface immune tolerance [15]. Accumulating studies have indicated abnormal autophagy in villi of miscarriage patients [16]; however, autophagic events in endometrium tissue from patients with RM remain to be elucidated. Considering these findings, we extended to illustrate the crosstalk between autophagy and amino acid metabolism to reveal the specific etiology associated with early miscarriage.

High throughput omics provide massive experimental data and multi-omics makes it easier to describe complex diseases by a single theoretical model. Metabolomics enables the differential assessment of the levels of a broad range of endogenous and exogenous metabolites and the identification of perturbed pathways due to disease. Next-generation RNA sequencing (RNA-Seq) technology enables the simultaneous analysis of all differentially expressed genes involved in the implantation and early embryonic developmental process [17]. Although several cases of miscarriage have been analyzed using RNA-Seq technology, up to now, the genome-wide study of the endometrium in unexplained RM remains to be investigated. The integration of metabolomics and transcriptomics can realize the mutual verification of the changes in the metabolome and transcriptome. It not only helps to profoundly analyze the association between the metabolism and transcription profile but also reveals the metabolic mechanism of various biological systems.

Therefore, by integrating the metabolomic and transcriptomic approaches, the present study aims to gain a better understanding of the metabolic and transcriptomic profile of endometrium in RM to decipher the molecular mechanisms underlying RM. Our results revealed increased autophagy in the endometrium of women with RM compared with women with normal pregnancy. Furthermore, we employed metabolomics on endometrial samples during the window of implantation (WOI) between RM patients and normal pregnant women to determine differential metabolite profiles and identify the specific alterations in the endometrial metabolic microenvironment associated with increased autophagy. Next, we combined this finding with the results of RNA-Seq to identify the key genes and potential signaling pathways related to RM. The findings may be beneficial in facilitating the early diagnosis of unexplained RM. Besides, the study further provides insight into the potential molecular mechanism underlying RM.

Materials and methods

Study population and sample collection

This cohort study retrospectively examined 97 infertile women who underwent their first IVF cycle in the Fertility Center of Shenzhen Zhongshan Urology Hospital from January 2017 to December 2019. The infertility factors of all participants include ovulation disorders, fallopian tube factors, pelvic inflammatory disease, male factor, and endometrial dysfunction. The control participants attended the clinic for the IVF/ICSI and the sole cause of marital infertility was male azoospermia based on clinical evaluation. All controls had regular menstrual cycles and normal sex hormone level, none had a history of miscarriage. The patients with endometriosis, endometritis, autoimmune- or thyroid-related disease, abnormal karyotypes, positive infectious disease tests, uterine malformation, and ultrasonographic evidence of hydrosalpinx were excluded from the control group. Patients were eligible for inclusion if the following criteria were met: (1) age under 40 years; (2) a regular menstrual cycle from 26 to 35 days; and (3) the endometrial thickness ≥ 7 mm on the day of biopsy and the endometrium was in the mid-luteal phase confirmed by hematoxylin and eosin (H&E) staining. The samples were collected within half an hour of Intrauterine Curettement, and biopsies of the endometrium were subjected to routine paraffin embedding and sectioning or snap-frozen in liquid nitrogen. Additional details on the clinical characteristics of the enrolled patients are summarized in Table 1. Of 97 infertile women, 51 RM patients and 46 normal pregnancy controls were recruited. This study was approved by the ethics committee of Shenzhen Zhongshan Urology Hospital. Written informed consent was obtained from individual patients for the use of the endometrial specimens.

Table 1

Baseline and cycle characteristics of women with normal pregnancy and RM^a

Characteristics	Normal pregnancy (n = 46)	RM (n = 51)	P-value
Maternal age (year)	30.69 ± 4.13	33.69 ± 4.35	0.605
FSH (mIu/mL)	6.09 ± 1.09	6.99 ± 4.19	0.217
E2 (pg/mL)	33.46 ± 12.49	32.37 ± 14.17	0.282
P (ng/mL)	0.70 ± 0.52	0.69 ± 0.48	0.354
PRL (ng/mL)	18.28 ± 6.80	19.05 ± 6.17	0.152
LH (mIu/mL)	3.94 ± 1.79	4.19 ± 1.76	0.545

Abbreviations: FSH, follicle-stimulating; E2, estrodiol; P, progesterone; PRL, prolactin; LH, luteinizing hormone.

^aStudent t-test, data are shown as mean ± SD.

Total DNA/RNA extraction and purification

Total RNA was isolated from the endometrium tissues using the Qiagen RNeasy Kit and QIAamp DNA/Blood Mini Kit following the manufacturer’s protocol. High-quality RNA was obtained by the RNase-Free DNase digestion. The purity and concentration and quantification of RNA samples were determined with a nanodrop spectrophotometer (Shimadzu, Japan). Total RNA was then stored at −20 °C for preparation.

Microarray hybridization and GO/KEGG analysis

For microarray hybridization, the Agilent Human whole-genome 4 × 44 K Oligo microarray (Agilent Technologies, Santa Clara, CA, USA) following the manufacturer’s instructions. Briefly, 200 ng of total RNA was labeled using the Low Input Quick Amp Labeling kit and then hybridized to Human whole-genome 4 × 44 K array containing ~35 000 human transcripts with cRNA probes at the core facility of GenoCheck. After fragmentation, the labeled cRNAs were hybridized to the microarrays for 17 h at 65 °C and scanned as described in the manufacturer’s protocol. Overexpressed mRNA was identified when the ratio of the mRNA expression level in unexplained miscarriage to the control group was 2 or more and under-expression was identified when the ratio was 0.5 or less. Gene Ontology (GO; geneontology.org) and Kyoto Encyclopedia of Genes and Genomes (KEGG; www.genome.jp/kegg/pathway.html) pathway analyses were performed. The P-value was calculated using a right-side hypergeometric test and adjusted for multiple testing using the Benjamini Hochberg. An adjusted P-value of <0.05 was considered to indicate a statistically significant deviation from the expected distribution, and the corresponding GO terms and pathways were explored by enrichment analysis. We analyzed all of the differentially expressed genes.

Sample preparation and GC–MS analysis

Frozen endometrium sample (20 ± 2 mg) was homogenized in 400 μL of extraction buffer (chloroform: methanol: water solvent, 2:5:2) containing 10 μg/mL of L-norleucine (NOR, Sigma Aldrich) as an internal standard on ice bath by using a TissueLyser (JX-24, Jingxin, Shanghai) with zirconia beads for 3 min at 30 Hz. After centrifugation at 14 000 ×g for 15 min at 4 °C, a total of 320 μL supernatant was collected. A second extraction step was performed using 320-μL methanol. And samples were centrifuged at 14 000 ×g for 5 min at 4 °C. Supernatants from the two extractions were combined. A 150 μL of combined supernatant was dried under a gentle nitrogen stream. The dry residue was reconstituted in 30 μL of methoxyamine hydrochloride (20 mg/mL) in pyridine, and the resulting mixture was incubated at 37 °C for 90 min. Subsequently, 30 μL of BSTFA (with 1% TMCS) was added into the mixture and incubated at 70 °C for 60 min before GC–MS analysis. Quality control samples pooled from all samples were prepared and analyzed by GC–MS as described above. Then, derivatized samples were analyzed, an Agilent 7890A/5975C GC–MS system (Agilent Technologies Inc., CA, USA). An OPTIMA® 5 MS Accent fused-silica capillary column (30 m × 0.25 mm × 0.25 μm; MACHEREY-NAGEL, Düren, GERMAN) was utilized to separate the derivatives. Helium (>99.999%) was used as a carrier gas at a constant flow rate of 1 mL/min through the column. The injection volume was 1 μL, respectively. The solvent delay time was 5.4 min. The initial oven temperature was held at 60 °C for 1 min, ramped to 240 °C at a rate of 12°C/min, to 320 °C at 40 °C/min, and finally held at 320 °C for 4 min. The temperatures of the injector, transfer line, and electron impact ion source were set to 250, 260, and 230 °C, respectively. The electron ionization energy was 70 eV, and data were collected in a full scan mode (m/z 50–600). Additional details on the clinical characteristics of the enrolled patients are summarized in Table 2.

Table 2

Baseline and cycle characteristics of women with normal pregnancy and RM in the metabolomic experiments^a

Characteristics	Normal pregnancy (n = 10)	RM (n = 10)	P-value
Maternal age (year)	31.1 ± 3.38	31.80 ± 2.86	0.623
FSH (mIu/mL)	5.60 ± 2.40	5.57 ± 2.08	0.324
E2 (pg/mL)	35.65 ± 12.57	36.78 ± 14.12	0.892
P (ng/mL)	1.63 ± 2.10	0.66 ± 0.39	0.092
PRL (ng/mL)	14.67 ± 4.82	18.71 ± 5.35	0.093
LH (mIu/mL)	4.49 ± 2.02	4.22 ± 1.88	0.761

Abbreviations: FSH, follicle-stimulating; E2, estrodiol; P, progesterone; PRL, prolactin; LH, luteinizing hormone.

^aStudent t-test, data are shown as mean ± SD.

Western blot and real-time quantitative PCR analysis

To validate the results of the expression profile, western blot analysis and quantitative PCR (qRT-PCR) analysis were performed. Whole lysates from HESC cells were extracted with RIPA buffer containing protease inhibitor cocktail (Beyotime, China). Protein concentrations were determined using the BCA Protein Assay Kit (Beyotime, China). The proteins were resolved to perform a western blot analysis on the Protein Simple (USA) machine with LC3B antibody (cell signaling technology, USA) using a standard procedure; 500 ng of endometrium total RNA was used for cDNA synthesis, which was performed using the M-MLV reverse transcriptase kit (Toyobo, Japan) according to the manufacturer’s instructions. Synthesized cDNA was utilized for PCR with primers at optimized cycles (Supplementary Table S1). qRT-PCR was performed using ABI7500 cycler (USA) with the QuantiTect SYBR Green PCR kit (Toyobo, Hilden, Japan). All experiments were performed at least in triplicate for each gene.

Immunohistochemistry

Endometrial tissues were collected during the mid-luteal phase using an endometrial curette. The endometrial samples were fixed with 4% paraformaldehyde for 6–12 h at room temperature after removing blood and washed with phosphate-buffered saline (PBS). Then, all samples were processed into paraffin within 48 h; 4-μm-thicked formalin-fixed, paraffin-embedded endometrial tissue sections were deparaffinized with three successive passages through xylene, and rehydrated through decreasing concentrations (100, 95, 80, 70, and 50%) of ethanol. The endogenous peroxidase activity was blocked by 3% hydrogen peroxide and the nonspecific binding was blocked by 5% bovine serum albumin for 20 min. The sections were incubated with primary antibodies specific for LC3B (Nanotools, 0231–100) at 37 °C for 1 h, dissolved in 1% bovine serum albumin (w/v in TBS) at the final concentration of 25 μg/mL. After three washes in 0.1% Tween 20 (v/v in PBS), sections were incubated for 30 min with secondary antibodies (Dako Cytomation), washed again as before. Antibodies binding were detected with a brown precipitate after stained with peroxidase substrate 3, 3-diaminobenzidine (DAB; Dako Cytomation) and counterstained with hematoxylin to allow visualization of the nuclei and dehydrated. Finally, the immunohistochemistry (IHC) section was scanned by the Vectra automated quantitative pathology imaging system (Perkin Elmer, Waltham, Massachusetts) and then quantified by the software Qupath 0.2.0-m9. Only the images with over 80% of tissues were included in the analysis. The numbers of cytoplasmic LC3+ puncta/cell indicated the conversion of LC3-I into LC3-II.

In vitro experiments

Human endometrial stromal cell line HESC cells were purchased from American Type Culture Collection and cultured in a medium of regular high-glucose DMEM (Gibco, 11995040). For glutamine deprivation experiments, the cells were maintained in a medium of high-glucose DMEM depleted of glutamine (Gibco, 10313021), supplemented with 10% FBS and sodium pyruvate. For inhibition of AMPK experiments, the HESC cells were cultured in a medium of regular high-glucose DMEM with Dorsomorphin (APExBIO, 866405-64-3).

Statistical analysis

For analysis of metabolites, orthogonal projections to latent structures-discriminant analysis (OPLS-DA) were performed for the evaluation of data. The t-test was used for univariate statistical analysis, and PCA and OPLS-DA were used for multivariate statistical analysis to identify significantly different metabolites. The variables with VIP values OPLS-DA model larger than 1 and P-values of univariate statistical analysis lower than 0.05 were considered statistically significant differential metabolites. Fold change was calculated as binary logarithm of average normalized peak intensity ratio between RM and control groups, where the positive value indicated that the average mass response of Group 1 was higher than Group 2. Statistical analyses were performed using the Statistical Program for Social Science (SPSS Inc., Chicago, IL, USA) software. Student t-test was performed to examine statistical significance (^*P < 0.05; ^**P < 0.01).

Figure 1

An abnormal expression of autophagy in endometrium of women with RM. (A) the mRNA expression level of MAPLC3B between endometrium of women with normal pregnancies (n = 6) and RM (n = 6). (B) Immunohistochemical staining of LC3B in RM endometrial tissue. Quantitative analysis of LC3+ dots/cells (RM: n = 18; normal pregnancies: n = 17) was performed using QuPath-0.2.0 m9. Stained LC3+ dots/cells were significant higher in the early RM group. Data are represented as means ± SD; ^*P < 0.05, Student t-test. (C) Statistics of KEGG PATHWAY Enrichment. The abscissa Rich Factor represents the ratio of Input frequency to Background frequency, the bubble size represents the number of genes annotated by the differential gene, and the color corresponds to the P value. (D) Bioinformatics analysis of differentially expressed mRNAs which include GO, pathway and disease enrichment (top30). (E) The protein–protein interaction network of all differently expressed genes. b: binding/association; a: activation; i: inhibition; p or + p: phosphorylation; —: indirect effect; e: expression; r: repression; u: ubiquitination; c: compound; a + p —: activation and phosphorylation and indirect effect.

Figure 2

Expression of mRNA was validated using real-time PCR analysis of 10 up-regulated genes and 6 down-regulated genes. All values were normalized against the mRNA expression level of β-actin. Ten mRNAs are up-regulated and six mRNAs are down-regulated in RM compared with the normal pregnancy group (n = 20). Results are consistent with the microarray.

Results

Overexpression of autophagy in the endometrium of women with early RM

The autophagy marker microtubule-associated protein1 light chain 3B (MAPLC3B) was markedly overexpressed in the RM endometrium (Figure 1A), indicating that autophagy might be activated in the RM group. Furthermore, the expression of autophagy in the endometrium during WOI was determined by the IHC experiment (RM: n = 18; normal pregnancies: n = 17). Sampling is taken on LH + 7 days, and the secretory period is determined by H&E staining. LC3B was used as the autophagy marker. The results showed that compared with the control group, the expression of LC3B was significantly increased in the RM group (Figure 1B), indicating an abnormal autophagy phenomenon in the endometrium of the RM group. To identify differential mRNA expression in the endometrium tissue of early miscarriage, we performed a heatmap filtering on the microarray data between the two groups (RM: n = 3; normal pregnancies: n = 3). A total of 595 differentially genes, including 452 genes up-regulated and 143 down-regulated genes, were identified to exhibit 5-fold (P < 0.01) changes, as compared with the control group. GO and KEGG pathway enrichment analyses were performed to determine the functions of the identified differentially expressed mRNAs. GO analyses included biological processes, cellular components, and molecular function. The differentially expressed genes were mainly related to the following cellular components: component of membrane, vesicle, extracellular exosome, and organelle (Figure 1D). Additionally, molecular pathways were annotated from the pathway database. Hemostasis, regulation of autophagy, and regulation of IFNα signaling were the top three signaling pathway; among these biological processes, autophagy was the most frequently activated as the related genes were significantly up-regulated (Figure 2B); the diseases enrich system revealed the top 30 diseases for analysis (Figure 1C), which indicated that heart failure, protein quantitative trait loci, and neurofibrillary tangles were the top three related disease terms according to the differentially expressed genes. Protein interactions of genes with significantly altered expression (P < 0.05) were delineated using the online Search Tool for the Retrieval of Interacting Proteins (the STRING 9.1 software) to predict functional associations and generate networks of the differentially expressed mRNAs. We identified a network of 452 up-regulated and 143 down-regulated genes which were functionally related to each other. The results demonstrated those interacting proteins are mainly involved in metabolic, developmental, cell killing, rhythmic processes, and response to the stimulus (Figure 1E).

To validate the microarray data, we further performed a qRT-PCR assay on cDNA obtained from endometrium tissues and analyzed 10 up-regulated genes and 6 down-regulated genes. All 10 up-regulated mRNAs (HIF1α, STAT3, LIF, LIFR, ATG12, eIF2α, MAPLC3B, HK2, beclin-1, MMP2) were significantly increased and 6 down-regulated mRNAs (ATG7, p62, ADCY10, ULK1, MMP9, TNFα) were decreased in unexplained miscarriage group compared with the control group (Figure 2). All the primer sequences were attached in Table 3.

Aberrant metabolite profile in the endometrium of women with early RM

To investigate different alterations in metabolic microenvironment between normal pregnancy and RM, we employed non-targeted metabolomics to analyze the different metabolites (RM: n = 10; normal pregnancies: n = 10). SIMCA (version 14.1) software was used to perform principal component analysis (PCA) on the samples, and a PCA model (R²X = 0.635) with six effective principal components was established to characterize metabolic differences between samples. However, the samples named HT2–5, HT2–6, and HT2–10 in this model deviate from other samples in the RM group. Therefore, the HT2–5, HT2–6, and HT2–10 samples were removed, and the PCA model of five effective principal components was re-established to reflect the differences between RM and control groups. The PCA model (R²X = 0.605) can be reliably used to characterize the metabolic differences (Figure 3A). Then, a cross-validation PLS-DA model with satisfactory discriminating ability was established to determine the metabolic differences between the two groups. The parameters describing the PLS-DA model were significantly enhanced (R²Y = 0.938, Q² = 0.805), revealing that the quality of the PLS-DA model is reliable (Figure 3B). The OPLS-DA model (R²X = 0.438, R²Y = 0.98, Q² = 0.833) indicated that the model could effectively distinguish between the two groups (Figure 3C). The permutation test showed that the quality of the current model is reliable (Figure 3D). The differentially expressed metabolites were identified based on the VIP value (threshold > 1) of the first principal component of the OPLS-DA model and the P-value (threshold < 0.05) of the single-dimensional test. Based on the non-targeted metabolome standard library, a total of 3500 credible metabolite standards were identified, including amino acids and their derivatives, glycerophospholipids, glycerides, coenzymes and vitamins, benzene and its derivatives, sphingolipids, carbohydrates and its metabolites, hormones and related substances, alcohols, amines, organic acids and their derivatives, and other metabolites. By searching the metabolome standard library with chromatographic retention time and mass spectrum, we screened and identified 41 metabolites exhibiting significant changes, including 19 decreased and 22 increased metabolites. To further explore the relationship between the differentially expressed metabolites of the two groups, we analyzed the quantitative data, and the result was displayed as a heatmap (R platform, version 3.3.0). The heatmap showed an increased concentration of glutamine while the decreased concentration of glutamic acid (Figure 3E). Next, we conducted a Pearson correlation analysis (R platform, version 3.3.0) based on the quantitative data of these substances to further characterize the correlation between different metabolites (Figure 3F). According to the correlation matrix, we found that there was a significant negative correlation between glutamine and glutamic acid, indicating that the pathway of glutamine to glutamic acid may be inhibited, leading to the suppression of the amino acid synthesis pathway. Metabo-analyst (version 4.0) was used to analyze the correlation between different metabolic pathways (Figure 3G). The results illustrated that alanine, aspartate, and glutamate metabolism pathways were significantly altered and might be involved in the occurrence of RM. Moreover, the amino acyl-tRNA biosynthesis pathway was also significantly associated with RM.

Abnormal glutamine metabolism induced autophagy via AMPK

Starvation and hypoxia induce autophagy via a particular pathway. Under the condition of starvation, cells sustain intracellular ADP/ATP ratio to activate the adenosine 5′-monophosphate-activated protein kinase (AMPK) pathway resulting in autophagy. In general, AMPK, a potent activator of autophagy in the catabolic process of hypoxia and energy starvation, plays a crucial role in the regulation of cellular energy homeostasis and responds to the low level of ATP. Besides, activation of AMPK senses the nutritional status of the body. However, previous studies about the relationship between nutrient sense and autophagy mainly focused on the response to glucose starvation, and there is a paucity of studies on amino acid starvation. Therefore, it remains elusive whether amino acid starvation induces autophagy through the activation of AMPK. In this study, we reasoned that glutamine deprivation could signal nutrient deprivation and induced endometrium autophagy through endometrium tissue and cell experiments in vitro. The transcriptome and qRT-PCR assays revealed that autophagy was activated in the endometrium of the RM group (Figure 4A and B). Combined with metabolomics data, we further performed experiments in vitro to analyze the relationship between glutamine deprivation and autophagy. We cultured human endometrium stromal cell HESC cells in a well-established generic glutamine deprivation medium depleted of glutamine to mimic a restricted nutrient stress to the endometrium. The results showed that the expression of LC3B protein was up-regulated in HESC cells cultured in a medium depleted of glutamine (Figure 4C). Furthermore, exposure to glutamine deprivation increased the baseline level of MAPLC3B, ATG12, and beclin-1 and enhanced expression level of GCN2 and AMPK while reduced the expression level of mTORC1 (Figure 4D). These results suggested that glutamine deprivation may induce autophagy in the endometrium.

Table 3

Related primer sequences

Genes	Primer sequences (5′–3′)	Reverse primer (5′–3′)
ADCY10	TAGGTACATGGAGGGGCAAG	GACGTAAGCCATCAGGTGGT
AMPK	GGGTCATTCTCTATGCTTTGC	GTCCTGGTGGTTTCTGTTGTA
ATG7	GCCAGGTACTCCTGAGCTGT	GGTCTTACCCTGCTCCATCA
beclin-1	ATCTAAGGAGCTGCCGTTATAC	CTCCTCAGAGTTAAACTGGGTT
eIF2α	GCACAGTTGGTGAAGTATGGC	AGGTACAGCCCTTTGCCTTC
HK2	CATCTGCCTGTCCATGTCACT	TACTCCAGTATTGCAGGTTCCA
LIFR	GCTATGTGCGCCTAACATGA	AGTGGGGTTCAGGACCTTCT
MAPLC3B	TTATTCGAGAGCAGCATCCAACC	CCGTTCACCAACAGGAAGAAGG
MMP2	AACGCCATCCCTGATAACCT	GCTTCCGAACTTCACGCTC
p62	CTCACCGTGAAGGCCTACCT	TAGCGGGTTCCTACCACAGG
TNFα	CATTGTTCTCGGCTATGACAG	GGAACAGCTCGGATTTCAG
MMP9	GGGACGCAGACATCGTCATC	TCGTCATCGTCGAAATGGGC
ULK1	TTACCAGCGCATCGAGCA	TGGGGAGAAGGTGTGTAGGG
HIF1α	TCAGCTATTTGCGTGTGAGG	AAAACCATCCAAGGCTTTCA
STAT3	CCCAGGTAGCGCTGCCCCATAC	GGAGGTAGCGCACTCCGAGGT
LIF	GTACCGCATAGTCGTGTACCT	CACAGCACGTTGCTAAGGAG
ATG12	CCACAGCCCATTTCTTTGTT	GTCCTCGGCTGCAGTTTC
GCN2	CTCCTGGTTGTAAGTGTTGGTC	AGTTTCTGGGTTAGGTTGATGG
β-actin	GTATCGTGGAAGGACTCATGAC	ACCACCTTCTTGATGTCATCAT

Figure 3

The metabolites profile of endometrium of women with normal pregnancies and RM. (A) PCA scores plot. The abscissa represents the first principal component PC1, which is represented by T [1]; the ordinate represents the second principal component PC2, which is represented by T [2]. (B) PLS-DA scores plot. (C) OPLS-DA scores plot. (D) Permutation test. It is mainly used to indicate whether the model is over fitted. (E) Heatmap of differentially expressed metabolites (n = 10). Each row represents the differential metabolites, and each column represents the sample number, and the tree structure on the left represents the similarity clustering relationship between the differential metabolites. Red indicates the increased concentration of differential metabolites, while green indicates the decreased concentration of differential metabolites. (F) Pearson correlation of different metabolites. Each row and column in the graph represent a differential metabolite. The correlation coefficient measures are shown on the right. The correlation between the circle size and color depth is related to the difference of metabolites. Red indicates a positive correlation between the differential metabolites, while green indicates a negative correlation between the differential metabolites. The darker the color, the larger the square and the greater the correlation. (G) Different metabolites metabolome view between RM and normal pregnancy women. The vertical coordinate [−log(P)] indicates the correlation between the metabolic pathway and the group, and the redder the color, the greater the correlation; the horizontal coordinate value indicates the importance of the pathway to the occurrence of the difference; the greater the circle, the greater the importance.

Figure 4

Amino acid deprivation induces autophagy by activating AMPK. (A) The mRNA profile of autophagy-related genes by RNA-Seq (n = 3). The red refers to the up-regulated genes, while the green refers to the down-regulated genes. (B) The mRNA expression level of autophagy related genes ATG12, beclin-1 and p62 (n = 20). (C) The increased expression level of LC3B protein in HESC cells cultured in a medium depleted of glutamine. (D) The increased mRNA expression level of ATG12, beclin-1and MAPLC3B, GCN2, and AMPK and the decreased mRNA expression level of mTORC1 in HESC cells exposure to glutamine deprivation. (E) The increased mRNA expression level of GCN2 and AMPK (n = 20). (F) Positive correlation of GCN2 mRNA and AMPK mRNA expression by qRT-PCR (n = 20). (G) The increased mRNA expression level of GCN2 after inhibition of AMPK in HESC cells. (H) Positive correlation of GCN2 mRNA and ATG12 mRNA expression by qRT-PCR (n = 20). (I) Positive correlation of GCN2 mRNA and beclin-1 mRNA expression by qRT-PCR (n = 20). (J) Negative correlation of GCN2 mRNA and p62 mRNA expression by qRT-PCR (n = 20).

Besides, the mRNA expression of GCN2 and AMPK was identified to be consistently increased in the RM group (Figure 4E). To clarify whether there is a correlation between the increased expression levels of AMPK and GCN2, we firstly examined the expression of GCN2 and AMPK in early RM endometrium tissue using qRT-PCR and identified a positive correlation in RM group (Figure 4F). Moreover, the inhibition of AMPK activity decreased the expression of GCN2 in vitro HESC cells experiments (Figure 4G). These results suggested that AMPK played an important role in inducing GCN2 in the endometrium. Interestingly, the results demonstrated that GCN2 exhibited a positive correlation with autophagy-related genes (ATG12 and beclin-1) (Figure 4H and I) and a negative correlation with p62 in the endometrium of women with RM (Figure 4J). However, further research is needed to elucidate the biological mechanisms underlying the associations between GCN2 and autophagy. These data indicate that autophagy is induced by abnormal glutamine metabolism via activation of AMPK.

Discussion

Early RM is the loss of three times or more consecutive spontaneous pregnancies within the first 12 weeks of pregnancy and with the same sexual partner [18]. Multiple etiologies, including parental chromosomal anomalies, endocrine disturbances, uterine pathologies, maternal thrombophilic disorders, autoimmune diseases, infectious agents, and non-coding RNA, have been associated with early RM. However, specific etiology underlying RM remains to be elucidated in many cases. In this study, we attempted to investigate the relationship between amino acid metabolism and autophagy in the RM endometrium. Interestingly, our data revealed that autophagy is induced by aberrant cellular metabolism, particularly glutamine metabolism, through activation of AMPK.

Autophagy is a highly conserved catabolic process that degrades cytoplasmic constituents, dysfunctional, and damaged organelles through the autophagy-lysosomal pathway [19]. Autophagy transpires at low basal levels in all cells to homeostatic functions, including protein and organelle turnover and bioenergetic management [11]. However, it can be considerably upregulated by many factors, such as conditions of hypoxia, starvation, growth factor deprivation, and when there is an urgent need for nutrients and energy for cell proliferation [20]. It has become increasingly recognized that autophagy plays a crucial role in aging, infectious diseases, and failure of the autophagic process has also been associated with worsen aging-related diseases, such as neurodegeneration cancer. However, how autophagy affects the endometrium during WOI remains obscure. Importantly, inhibition of autophagy with small molecular drugs or short interfering RNAs can suppress the expression of vascular endothelial growth factor (VEGF). Besides, pharmacological or genetic inhibition of core components of the autophagic pathways induces cell death that distresses homeostasis [21]. Liu et al. [22] reported a new mechanism for diabetic angiogenesis impairment in endothelial cells and diabetic mice models. The study also suggested that the increased level of autophagy-related genes can inhibit the expression of VEGFR2, while the decreased level of autophagy-related genes can restore the expression of VEGFR2 and angiogenesis. Furthermore, autophagy-related proteins, LC3 and beclin-1, have been demonstrated in the villous trophoblast from spontaneous abortion, and increased levels of LC3-II were detected in placentas from pregnancies complicated by severe preeclampsia (PE) compared with those from normal pregnancies [23]. The previous study showed a differential change in the autophagy of trophoblast between normoxia and hypoxia-reoxygenation [24]. In the present study, we extended these findings by demonstrating increased autophagy in the RM endometrium samples using qRT-PCR and RNA-Seq analyses. The results indicated that autophagy might play an important role in ER.

Depletion of nonessential amino acids may trigger autophagy [8]. The metabolic profile found that 19 metabolites were decreased and 22 metabolites were increased in the endometrium of women with RM, and the altered metabolites included pyruvic acid, glutamine, glutamic acid, and succinic acid. Amino acid starvation results in the accumulation of uncharged tRNA species, which can bind eukaryotic translation initiation factor 2α kinase 4 (eIF2αK4, best known as GCN2) and facilitate the activation of its kinase function, leading to suppression of protein synthesis and induction of autophagy [25]. Glutamine is an important energy supply substance for rapid cell growth and differentiation, as well as an important regulator of protein metabolism. Studies have reported that the lack of glutamine can lead to a certain degree of autophagy, but there are also researches revealed that glutamine metabolism may induce autophagy by releasing NH3 [23]. Consistently, in this study, the results of the mRNA expression profile indicated abnormal glutamine and glutamic acid metabolism and increased autophagy in the endometrium of women with RM. Moreover, glutamine deprivation increased autophagy in vitro experiments. These findings suggested an aberrant regulation of amino acid metabolism and autophagy in RM and a correlation between glutamine deprivation and autophagy; however, further studies are warranted to decipher the underlying mechanism.

AMPK, a cellular energy sensor of changes in intracellular levels of AMP, ADP, and ATP, is a heterotrimeric complex comprised of a catalytic α subunit, a scaffolding β subunit, and a regulatory γ subunit. AMPK functions as a master regulator of metabolism and triggers the initiation of the autophagic cascade through multiple mechanisms. Presumably, in response to a decreased cellular energy status, AMPK activates the kinase function upon binding of the γ subunit with two molecules of AMP to inhibit the protection from Thr172 dephosphorylation of the α subunit [26]. As the basic structure of protein synthesis and the substrate of energy production, amino acids change the levels of AMP and ATP in cells by producing α-ketoglutarate (AKG) participated in the TCA cycle and regulate energy metabolism through the interaction of AMPK and mTORC1. Our study showed that women with RM exhibited higher endometrial levels of AMPK compared with normal pregnant women. Overexpression of AMPK was associated with increased levels of amino acid sensor GCN2 and autophagy-related proteins, including ATG12 and beclin-1. Inhibition of AMPK decreased the expression of GCN2. Taken together, these findings revealed important crosstalk between amino acid metabolism and autophagy in RM. Although further studies are warranted to elucidate the interaction between amino acid metabolism and AMPK in the human endometrium, these findings of this study suggested that amino acid deprivation may induce autophagy via the activation of AMPK.

Authors’ contributions

LH: conceptualization, validation, and draft the manuscript. YCZ: designed the study and performed some experiments. SL: review and editing. THW and LJY: data collection. YYL: formal analysis. YZ and LD: supervised and supported the study. All authors read and approved the final manuscript.

Acknowledgment

We are very grateful to all participants in this study, and thank all the clinical staff. We thank all the members of Dr Zeng’s group for helpful suggestions, interesting discussions, and funding and technical support. We also wish to acknowledge Dr YOZ, Professor of medicine, University of Tsinghua, for her help in interpreting the significance of the results of this study.

Conflict of interest

The authors declare that there is no conflict of interest that could be perceived as prejudicing the impartiality of the research reported.

Footnotes

^†Grant support: This work was supported by the National Natural Science Foundation of China (31972910, 31571185, 31401224), the Natural Science Foundation of Guangdong Province (2017A030310178), the Pearl River S&T Nova Program of Guangzhou City (201710010040), the Health & Family Planning System Research Project of Shenzhen City (SZXJ2017002), and the Sanming Project of Medicine in Shenzhen (ZDSYS201504301534057).

Author notes

Ling Hong and Yuan Chang Zhu authors contributed equally to this work.