Lipid Characteristics of Pregnant Women With Subclinical Hypothyroidism and Their Correlation to the Pregnancy Outcomes

Subclinical hypothyroidism (SCH) in pregnancy has become an important complication of pregnancy. We used nontargeted lipidomics to compare differential metabolites between women with SCH and healthy women. The metabolic pathways of the differential metabolites were analyzed using the Kyoto Encyclopedia of Genes and Genomes (KEGG) database. We found: 1. Multivariate analysis revealed 143 lipid molecules with different levels between the SCH group and the control group. Based on fold change, 30 differential lipid metabolites were selected as potential biomarkers. 2. KEGG pathway enrichment analysis showed that the differential metabolites participate in such pathways as pathogenic Escherichia coli infection response, metabolic pathways, glycerophospholipid metabolism. 3. Correlation analysis showed that sphingomyelin (SM) and phosphatidylcholine (PC) were positively correlated with tumor necrosis factor-α (TNF-α), C-reactive protein (CRP), and interleukin-6 (IL-6), while phosphatidylglycerol (PG), and phosphatidylinositol (PI) were negatively correlated with them. And PG was positively correlated with birth weight. The lipid prole of pregnant women with SCH is signicantly different from that of healthy pregnant women. Lipid molecules with differential metabolism, such as SM, PE, and PI, might be targets for further investigation of the pathogenesis of SCH in pregnancy and reduce the incidence of adverse pregnancy outcomes.


Introduction
Subclinical hypothyroidism (SCH) in pregnancy refers to the elevation of serum thyroid-stimulating hormone (TSH) in pregnant women with normal levels of free thyroxine (FT4) and total T4. Patients with SCH usually do not have obvious clinical symptoms or physical signs, so the diagnosis of SCH depends mainly on laboratory examinations. The prevalence of SCH during pregnancy has been increasing year by year, and it has become an important complication of pregnancy. Its adverse effects on pregnancy outcomes are becoming clear, including spontaneous abortion [1] , preeclampsia [2] , placental abruption, premature rupture of membranes, low birth weight [3] , fetal distress, preterm birth [4] , and abnormal neuropsychological and intellectual development [5] . The pathogenesis of SCH remains unclear and might be associated with nitric oxide (NO) [6] , Nesfatin-1 [7] , gene silencing mediated by microRNAs [8] , inactivation and mutation of the thyroid-stimulating hormone receptor gene [9] , and the gut-thyroid axis [10] . In summary, SCH during pregnancy has high incidence, lacks de nite symptoms and physical signs, and causes obvious adverse pregnancy outcomes. However, research on its pathogenesis is still lacking.
Although previous serum and urine metabolomic studies provided some clues about the pathogenesis of SCH at the metabolic level, the pathogenesis of SCH during pregnancy is still unclear, especially in terms of lipid metabolism. Zhang et al [11] fed a high-fat diet (HFD) or chow diet to male Sprague-Dawley rats for 18 weeks. Their examination of thyroid function showed that HFD rats had hypothyroidism, and the liquid chromatography (LC)-mass spectrometry (MS) analysis showed that HFD rats had a disturbed thyroid lipid pro le. Lee et al and Shao et al [12,13] also pointed out that metabolic disorders induced by excessive nutrition might be a pathogenic factor of hypothyroidism, which revealed that the thyroid gland is another victim of lipotoxicity. The association between SCH and lipids has become more clear in recent years.
What is unclear is how lipid metabolism disorders contribute to the pathogenesis of SCH during pregnancy. Therefore, the main purpose of this study was to use LC-MS to analyze the characteristics of the lipid pro le of pregnant women with SCH in order to investigate its pathogenesis and reduce the incidence of adverse pregnancy outcomes.

Results
General clinical data and pregnancy outcomes This study totally enrolled 30 patients in the SCH group and 30 patients at the same stage of pregnancy in the control group. The general clinical data of subjects in the two groups are shown in Table 1. The differences in age, BMI before pregnancy, BMI at enrollment, gravidity, parity, delivery mode, and Fasting blood glucose between the SCH group and the control group were not signi cant.
The pregnancy outcomes of subjects in the two groups are shown in Table 2. The differences in neonatal sex, Apgar scores at 1 and 5 minutes, between the SCH group and the control group were not signi cant, but birth weight (BW) was signi cant (p=0.036).  Difference analysis on the SCH group and the control group To validate the signi cant differences between the SCH group and the control group, the supervised partial least squares-discriminant analysis (PLS-DA) multivariate method was used for remodeling and analysis of the two groups of data (Fig. 1A). The results showed that most members of the two groups could be separated (R2Y-0.862 and Q2=0.286). In model validation (Fig. 1B), R2 was (0.0, 0.74) and Q2 was (0.0, -0.55). The validation results showed that this model was stable and reliable. Differential lipid molecules were subjected to cluster analyses (Fig. 1C). The results showed that serum metabolites in these two groups had a certain aggregation trend.
Screening of differential lipid molecules The differential lipid molecule screening conditions in this project were as follows: 1) a variable importance of projection score of the rst two principal components of the PLS-DA model ≥1, 2) FC≥1.2 or ≤0.83, and 3) t.test_p.value_BHcorrect <0.05. Compared with the healthy group, a total of 143 differential metabolites were identi ed in the SCH group, including 46 upregulated metabolites and 97 downregulated metabolites. The volcano plot used log 2 (FC) as the X-axis andlog 10 (t.test_p.value_BHcorrect) as the Y-axis ( Fig. 2A). Pathways analyses and correlation with clinical data and pregnancy outcomes of differential metabolites This study used the Kyoto Encyclopedia of Genes and Genomes (KEGG) database as the basis for linking differential metabolites to metabolic pathways, and enrichment analyses were performed on the pathways. The results indicated that differential metabolites in the SCH group and the control group participated in many pathways (Fig. 3A). Some of the main ones were the response to pathogenic Escherichia coli infection, regulation of lipolysis in adipocytes, metabolic pathways, glycerophospholipid metabolism, and fat digestion and absorption pathways.
Spearman correlation analyses were performed on the 29 differential metabolites that had the largest differences and clinical data (Fig. 3B). LPA and LPEt had a positive correlation with BMI at enrollment, while LPC had a negative correlation. SM, PC, and LPC showed a positive correlation with TNF-α, CRP, and IL-6, while PG, LPEt, LPA, PI, and different subclasses of PC showed a negative correlation with them. PG, LPA and TG had a positive correlation with BW, while LPC(20:0e) had a negative correlation. Moreover, there was no signi cant correlation between 29 differential metabolites and Apgar scores at 1 minute and 5 minute.

Discussion
Lipids are essential metabolites for living organisms and have many critical cellular functions. They can re ect the metabolic status of cells. Lipid metabolism in pregnant women changes signi cantly to meet the needs of growth and development of the fetus and ensure the health of the mother. If lipid metabolism is disordered, it will cause vascular endothelial damage, energy supply impairment [14] , and imbalance of immune functions and thyroid functions. This study performed a nontargeted plasma lipidomic analysis on pregnant women with SCH to investigate the pathogenesis of SCH during pregnancy.
The association between the thyroid gland and the lipidome is not well understood. Guidetti et al [15] extracted total lipids from the mitochondria of rats with hypothyroidism and showed that the proportion of PC increased and the proportion of PE decreased. The LC-MS study of Liu et al [16] showed that the levels of PC, SM, and PE in patients with hypothyroidism increased. Our PLS-DA plot showed that the lipidomes of pregnant women with SCH and healthy pregnant women were different. The volcano plot and the bar graph showed that the levels of PE, SM, and PC were higher in the SCH group than in the control group. The results are consistent with previous studies. Some studies [17,18] have indicated that TSH and TG are positively correlated. Screening of differential metabolites between the two groups in our study showed that the levels of TG (16:0/16:1/18:2) and DG (16:0/16:0) decreased in the SCH group. The reasons for this discrepancy might be the small sample size and differences in experimental techniques. In addition, the TG (16:0/16:1/18:2) and DG (16:0/16:0) obtained in this study were subclasses of TG and DG and do not represent the serum levels of total TG and DG. Therefore, our results indicate that the lipidomes of the SCH group and the control group were signi cantly different, and these two groups could be distinguished by their lipid metabolism.
PE affects many cellular processes, including cellular stability and the functions of many membrane proteins, and is a critical regulatory factor of cell membrane uidity. Our KEGG pathway analysis showed that differences in the pathogenic Escherichia coli infection response pathway were signi cant. In response to pathogenic Escherichia coli infection, during the contact process between bacterial agellin and Toll-like receptor (TLR) 4 and TLR5 in gut epithelial cells, PE is an important stimulus of the Toll-like receptor signaling pathway to further induce in ammation. Our previous studies have shown that gut ora had signi cant differences between pregnant women with hypothyroidism and healthy pregnant women. The relative abundances of gut Proteobacteria, Gammaproteobacteria, and Provotella in women with hypothyroidism during pregnancy were all higher than those in the control group at the same stage.
In recent years, changes in gut microbiota composition and function have been considered a hallmark of metabolic damage [19] . The extremely diverse metabolites that are produced, such as lipopolysaccharides and endotoxins, could reduce the integrity of cell connections [20] . They enter the systemic circulation through the leaky gut mechanism to induce T cell activation, resulting in a series of immune and in ammatory responses [21,22] . In addition, studies [23,24] have indicated that in ammation and the TLR pathway are associated with thyroid functions.
SM levels are mainly altered by endoplasmic reticulum (ER) stress, and SM plays important physiological roles in cell growth, differentiation, senescence, and signal transduction [25] . ER stress [26] might be a very important mechanism during the process of induction of SCH by a HFD, and the IRE1α/XBP-1 s pathway is involved. In this study, we observed that SM in the SCH group was higher than that in the control group and was positively correlated with IL-6. SM can be hydrolyzed by ve different sphingomyelinases to exert various biological functions. Sphingomyelinases are in uenced by TNF-α [27] in vivo. Furthermore, the PI level in pregnant women with SCH during pregnancy was lower than that in healthy pregnant women. In thyroid cells, PI participates in intracellular TSH signal transduction through phosphatidylinositol (3,4,5)-trisphosphate (PIP-3) [28] . Lipid kinases in the phosphoinositide 3-kinase (PI3K) family can promote TH1 lymphocytes to produce higher levels of interferon-γ and TNF-α secretion to stimulate release of chemokine ligand (CXCL) 10 from thyroid cells to initiate and maintain the autoimmune process in the body [29][30][31] . In our correlation analysis, PI and TNF-α showed a negative correlation. SCH during pregnancy is an immune-metabolic disease. Zhu et al [32] showed that IL-6 and TNF-α levels in SCH patients were signi cantly higher than those in healthy individuals. Therefore, metabolic disorders of SM and PI in pregnant women might cause SCH during pregnancy by affecting the autoimmune process.
In conclusion, we summarizes a possible pathogenic mechanism of SCH during pregnancy . During the contact process between bacterial agellin and TLR4 and TLR5 on gut epithelial cells, PE is an important stimulator of the TLR signaling pathway leading to in ammation and further in uencing the development of the disease. ER stress [26] might be a very important mechanism during the process of induction of SCH by HFD. The IRE1α/XBP-1s pathway is involved, and SM plays an important role. PI participates in intracellular TSH signal transduction through the PIP-3 signaling pathway [28] . Disruption of PIP-3 signal transduction further in uences immune processes to affect thyroid functions.
Some studies have shown that hypothyroidism during pregnancy could increase the risk of premature delivery, low birth weight infants, fetal distress [3,4] . Labarre et al. [33] found that maternal blood lipid level during pregnancy can affect the umbilical cord blood lipid group and its relationship with body weight. Our study found that PG, LPA and TG were positively correlated with birth weight, while LPC(20:0e) was negatively correlated, and there was a signi cant difference in birth weight between the SCH group and the control group. Therefore, we can speculate that PG and other differential lipid metabolism molecules can affected birth weight of the newborn by umbilical cord blood . The results of our study did not nd any difference between the two groups of neonatal Apgar scores at 1 minute and 5 minute, and there was no correlation between the potential biomarkers and Apgar scores at 1 minute and 5 minute. Therefore, this study did not nd a signi cant impact of gestational hypothyroidism on fetal distress.
The strengths of this study are as follows: The in uence of drugs on lipid disorders was controlled in all pregnant women in the SCH group before treatment. This study also included clinical examination results at a time close to the sample collection. Therefore, the association between the serum lipid pro les and clinical characteristics in pregnant women was tested at the smallest possible interval. Of course, this study also had limitations: The sample size was small. This study had certain regional speci city due to factors such as dietary habits and geographical differences. Because specimens were collected after the thyroid function disorders were discovered, no causal relationship between lipid metabolism disorder and thyroid function abnormality could be con rmed. This study only included samples in late pregnancy before the subjects took medication. Therefore, the dynamic changes to serum lipid metabolism in the whole pregnancy and after drug treatment could not be studied. These considerations also provide direction for our future studies.
In short, we analyzed the serum lipidomes of two groups of pregnant women and found signi cant differences in the lipid pro les between pregnant women with SCH and healthy pregnant women. Lipid metabolism disorder in pregnant women might result in the development of SCH during pregnancy by affecting in ammation and immune responses through an ER stress mechanism, PIP-3 signaling pathway disruption, and gut ora metabolism. Lipid molecules with differential metabolism, such as SM, PE, and PI, could be studied for their roles in the pathogenesis of SCH during pregnancy and reduce the incidence of adverse pregnancy outcomes.

Material And Methods
Study subjects Pregnant women who received regular perinatal health care in the outpatient department of the Third A liated Hospital of Zhengzhou University and gave birth in the hospital between July 2019 and January 2020 were randomly selected. They included 30 pregnant women with SCH (SCH group) in late pregnancy and 30 healthy pregnant women (control group) in late pregnancy who met the inclusion criteria.
Inclusion criteria (1) The thyroid function levels of the SCH group met the diagnostic criteria in the 2017 guidelines of the American Thyroid Association for the diagnosis and management of thyroid disease during pregnancy and the postpartum [34]  Note: Central hypothyroidism refers to hypothyroidism caused by reduced production and secretion of thyrotropin-releasing hormone or TSH resulting from a disorder of the hypothalamus or pituitary gland.
Statement: (i) All enrolled subjects volunteered and signed informed consent forms. This study was approved by the Ethics Committee of the Third A liated Hospital of Zhengzhou University. (ii) All methods were carried out in accordance with relevant guidelines and regulations.
Specimen collection All pregnant women fasted for 8-12 h before blood collection. A total of 5 mL blood from the median cubital vein was collected in ethylenediaminetetraacetic acid-coated tubes. After collection, the collection tube was gently inverted four times, wrapped in aluminum foil, and temporarily stored in a 4°C refrigerator. Blood samples were centrifuged within 2 h of sample collection in a lowtemperature centrifuge at 4°C and 1600×g for 10 min. The supernatant (plasma) after centrifugation was aliquoted into several centrifuge tubes and stored in a -80°C freezer. Later, samples were sent to the BGI Group for lipidomic examination.
Data collection The basic data of the two groups of pregnant women, including body mass index (BMI) before pregnancy, BMI at enrollment,, gravidity, parity, and delivery mode, neonatal sex, birth weight (BW), Apgar scores at 1 and 5 minutes, were collected. In addition, clinical data were collected during sample collection, including serum TSH, fasting blood glucose (FBG), CRP, IL-6, TNF-α. MS materials and methods: The Q Exactive mass spectrometer (Thermo Fisher Scienti c, USA) was used for primary and secondary MS data acquisition. The range of the mass-to-charge ratio in MS scanning was 200-2000, the primary resolution was 70,000, the automatic gain control was 3e6, and the maximum injection time was 100 ms. According to the precursor ion intensity, Top3 was selected for fragmentation. The secondary MS data were collected. The secondary resolution was 17,500, the automatic gain control was 1e5, the maximum injection time was 50 ms, and the stepped normalized collisional energy was set as 15, 30, and 45 eV. The settings of electrospray ionization were a sheath gas ow rate of 40, an aux gas ow rate of 10, a spray voltage (|kV|) of 3.80 in the positive ion mode and 3.20 in the negative ion mode, a capillary temperature of 320°C, and an aux gas heater temperature of 350°C.
To provide more reliable experimental results during instrument detection, samples were randomly sorted to minimize systemic errors. One QC sample was inserted for every 10 samples.
Statistical methods Raw data generated by the LC-MS/MS detection were input into LipidSearch v.4.1 (Thermo Fisher Scienti c, USA) for MS data analysis. The data matrix contained the lipid molecule identi cation results and quantitative results. Identi cation and peak extraction of individual samples were rst performed using LipidSearch v.4.1, and then peak alignment was performed on all samples.
The raw data obtained from LipidSearch were input into metaX for data pretreatment and subsequent analyses. SPSS 25.0 software was used to process basic clinical data of pregnant women in two groups. Normally distributed quantitative data are described as mean ± standard deviation (χ±s), and differences between groups were compared using the independent-samples t-test. Nonnormally distributed quantitative data are expressed as median (M) and quartile (Q), and differences between groups were compared using the Mann-Whitney U-test. Categorical variables are described as frequency, and the differences between groups were compared by the chi-squared test. Correlations were calculated by Spearman analysis.