The alteration of metabolic profiling in the second trimester of GDM and high risk pregnancy, and a nomogram for predicting macrosomia

Objective Dyslipidemia in the second trimester and associated gestational diabetes are increasing worldwide. Carnitine plays a key role in lipid metabolism. We aim to describe metabolic profiling in the second trimester based on carnitine related metabolomics in GDM and high risk pregnancy, and to find the potential risk factors in GDM and candidate metabolites for diagnosing GDM induced macrosomia. We have randomly investigated 450 pregnant women and their neonates in this retrospective study and 56 (12.4%) GDM cases were diagnosed. We used LC-MS/MS performing metabolic profiling about 12 amino acids and 31 acylcarnitines (containing C0) to assess circulating metabolites concentration in different subgroup according maternal and newborn clinical characteristic. We also calculated the correlation coefficient between maternal and newborn. GDM potential metabolic risk factors were screened by PLS-DA. Multivariate regression analyses were used in identifying independent risk factors for GDM and macrosomia. Based on these carnitine-related factors, a nomogram for estimating macrosomia was developed. Method This study is an observational retrospective study. Between June 2017 and 2018,a total of 450 pregnant women in the second-trimester participated this study at Women’s Hospital, University in Hangzhou. These women were blind, random chosen. The study design has been approved by the Ethics Committee of the hospital. All participants were of Chinese Han ethnicity. We obtain maternal characteristics information as follow: gravidity, parity, age, height, pre-pregnancy BMI, weight gain during pregnancy, using assisted reproductive technique (ART), method of pregnancy termination. The follow information of the offsprings were collected: termination of pregnancy weeks, sex, birth length, birth weight, head circumference, abdomen circumference and abdomen minus head circumference (data from fetal growth measurement ultrasound before labor). GDM was defined according to the Chinese Current Care Guidelines for GDM as one or more pathological glucose values in a standard oral glucose tolerance test (OGTT). [11] The diagnostic thresholds were: fasting plasma glucose: 5.1mmol/L, 1 H 10.0 mmol/L, and 2 H 8.5 mmol/L.[12] According to the guideline, 56 GMD cases were diagnosed. We randomly sampled 112 women without GDM from residual pregnant women participated in this study, frequency-matched to cases on birth day, age, delivery mode, and number of fetuses with 2:1 ratio. The common definition of macrosomia is birth weight over 4000g.


Introduction
Pregnancy is a complex process accompanied with substantial changes in lipid metabolism. In the second trimester, maternal metabolism is an anabolic state combined with increasing lipogenesis, accelerating maternal fat stores. [1] In the third trimester, there is a catabolic state breakdown in fat depot for fetal growth.
Dyslipidemia during the second trimester is associated with metabolic disorder (GDM, hyperlipemia, hypertension) and adverse neonate outcomes (macrosomia). [2] Carnitine and related amino acid (AA) play key roles in fatty acid metabolism. [3] So, it is essential to investigate carnitine related metabolomics in the second trimester and find metabolomic risk factors for metabolic disorder diseases such as GDM.
GDM is a common metabolic disorder during pregnancy, occurs in approximately 8-10% of pregnancies in global.
[4] Pregnant women with GDM are more prone to hypertension and metabolic syndrome. The risks to fetus include macrosomia, respiratory distress syndrome, childhood obesity and T2DM in adults. [5] The main causes of GDM are still undefined, a number of studies have reported that the serum levels of branched-chain and aromatic amino acids (BCAAs and AAAs), including leucine, isoleucine, valine, phenylalanine, and tyrosine, are significantly different among lean, obesity, and diabetes, and closely correlated to insulin resistance, which highlighted their potential impact on diabetes diagnosis and risk assessment.
[6] Carnitine worked as a critical role in energy metabolism of transporting long-chain fatty acid from the cytosol into the mitochondria.
[7] Some studies propose free carnitine and acetyl carnitine (AC) are decreased in pregnancy at first trimester compared with non-pregnancy. [8] However, studies of metabolic profiling about carnitine in second trimester among diverse pregnant women are rare. Herein, we provide a study of metabolism of carnitine and related AA about normal, high risk pregnancy and GDM, and investigate the potential risk factors in GDM.
Fetal macrosomia is clearly related with GDM. Macrosomia can induce multiple perinatal complication, such as shoulder dystocia, brachial plexus injury.
In the present study, LC-MS-based metabolomics approach was used to quantified the metabolic alteration in second trimester. We found the characteristic of metabolomics combined with clinical information and detected the potential risk factors of GDM. Moreover, we established a prognostic nomogram incorporating AA and AC for predicting macrosomia. Our finding provides a new insight that metabolic risk factors play a crucial role in GDM and associated macrosomia.

Subject
This study is an observational retrospective study. Between June 2017 and April 2018,a total of 450 pregnant women in the second-trimester participated this study at Women's Hospital, school of medicine, Zhejiang University in Hangzhou. These women were blind, random chosen. The study design has been approved by the Ethics Committee of the hospital. All participants were of Chinese Han ethnicity. We obtain maternal characteristics information as follow: gravidity, parity, age, height, pre-pregnancy BMI, weight gain during pregnancy, using assisted reproductive technique (ART), method of pregnancy termination. The follow information of the offsprings were collected: termination of pregnancy weeks, sex, birth length, birth weight, head circumference, abdomen circumference and abdomen minus head circumference (data from fetal growth measurement ultrasound before labor).
GDM was defined according to the Chinese Current Care Guidelines for GDM as one or more pathological glucose values in a standard oral glucose tolerance test (OGTT).
[11] The diagnostic thresholds were: fasting plasma glucose: 5.1mmol/L, 1 H 10.0 mmol/L, and 2 H 8.5 mmol/L. [12] According to the guideline, 56 GMD cases were diagnosed. We randomly sampled 112 women without GDM from residual pregnant women participated in this study, frequency-matched to cases on birth day, age, delivery mode, and number of fetuses with 2:1 ratio. The common definition of macrosomia is birth weight over 4000g.

Metabolic profiling detection by LC-MS/MS
We aim to investigate the plasma AA and AC level of pregnant women at the second-trimester. We also obtain their neonate blood plasma sample. The blood samples were stored -20℃.
[13] Furthermore, sample preparation used in tandem mass spectrometry (4000 QTrapTM; AB Sciex, Darmstadt, Germany) testing the concentration. The method used in the present study was essentially a modification of the procedure described elsewhere. [14] AA and AC were quantified using appropriate isotope-labelled standards. LC separation was performed on an Acquity UPLC HSS T3 column (2.1*100 mm, 100A°, 1.8 lm particle size; Waters Corporation, MA) using water with 0.1% formic acid, 5mM ammonium acetate and 0.015% heptafluorobutyric acid as solvent A and methanol with 0.1% formic acid and 5Mm detected with a Xevo-G2-QTOF MS (Waters Corp.) operating in positive mode. Raw data were processed using Targetlynxas described previously. Accuracy of quantification was below 6% for all quantified metabolites except glutamic acid (13.9%). Quantitative data were obtained using MetIDQTM Software. [15]

Statistical analysis
Data in figures and data were shown as mean±SD. Statistical methods Evidence for a difference between the two groups was tested by Student's t-test. One-way ANOVA was used in more than two groups. For correlation analysis, the Spearman correlation coefficient was calculated in the level of AA and AC between the second trimester pregnancy adults and their neonates. In PLS-DA, metabolomics data were log transformed to ensure a normally distributed data set, R2 (goodness of fitness) and Q2 (goodness of prediction) were assessed in the PLS-DA models. In the univariate analyses, A Mann-Whitney U test was used to compare the variables GDM and control group. The values of serum AA and AC were log-transformed and standardized before entered into a logistic regression model. Odds ratios (ORs) and 95% confidence intervals (CIs) are reported per standard deviation. All statistical analysis was performed using IBM SPSS 23.0 edition, SIMCA 14.0, R vision 3.6.0. A significance level of 0.05 was used for all statistical tests.

Comparison of stages of pregnancy
Baseline characteristics of 450 study participants divided into different categories according to natural conditions of pregnant women (gravidity, parity, age, height, pre-pregnancy BMI, weight gain, ART, method of pregnancy termination). The statistically significant results are shown in Table 1. In age subgroup, there was a trend that the level of Ala, Met, BCAA leucine, isoleucine, valine), AA (phenylalanine, tyrosine) ,several acylcarnitines (C2,C3, C4DC+C5OH, C16, C18, C18:1) were higher in age >35 group, whereas the level of C0(free carnitine), Gly in age>35 group was lower. In pre-pregnancy BMI ≥ 25.0 and 30 subgroup, ALA, BCAA leucine, isoleucine, valine), AAA (phenylalanine, tyrosine), several acylcarnitines (C2 C3 C5 C16 C18:1) were higher. These metabolites were increased with pre-pregnancy BMI. In weight gain group, BCAA leucine, isoleucine, valine) and AAA(phenylalanine, tyrosine), C3, C5, C16, C18:1 were higher in weight gain ≥ 20 kg group than other group, while C0 was lower. In ART group, Ala, Arg C2, C3, C5 were higher while C0, Gly were lower. Our clinical characteristics stratified that the plasma level of AA and AC (contain free carnitine) showed no statistical difference in gravity, parity, height, method of pregnancy termination subgroup.

Comparison of newborns
Additionally, we investigated the maternal level of AA and AC in second-trimester pregnancy under the different neonate subgroup termination of pregnancy weeks, sex, birth length, birthweight, head circumference, abdomen circumference and abdomen minus head circumference . The result was shown in table 2. In birth weight >4000g group, we found Ala, Arg, BCAA and AAA and several acylcarnitines (C2, C3, C5, C16, C18:1) were higher, these metabolites increased with birth weight, while C0 was lower. In abdomen circumference> 35cm group, Ala, BCAA and AAA and several acylcarnitines (C2, C3, C5, C16) were higher, C0 and Gly were lower. In abdomen minus circumference subgroup, the metabolites characters have the similar trend with abdomen circumference >35cm subgroup. AA and AC have no significant difference in other subgroups.

Correlation between maternal and newborn in amino acids and carnitine.
We calculated the correlation between the corresponding values of the carnitine profile of the mothers at second trimester and their own newborns. Significant positive correlations were seen for leu, Val, Phe, C0, C2, C3, C4DC+C5OH, C16, C18:1 (Table 3). Other AA and AC have no statistic significantly, these were not shown in table. We found the level of carnitine in neonate is less than adult.

Different metabolite distribution in incorporated group
We examined the serum metabolite in GDM patients and control. Our study used PLS-DA model (R2=0.527, Q2=0.464) to analyze difference between two groups. The PLS-DA scatterplot showed a clear class separation with GDM at the left ang the control group at the right. Furthermore, we used variable importance for the projection (VIP) to estimate the contribution of every metabolite to class separation (GDM vs. control). A VIP value >1 were considered with a high contribution to class separation. The VIP analysis showed that C0 plays a main role in class separation, followed by LEU+ILE+PRO-OH, C3, Phe, C18, TYR, C16, C2, GLY, ALA, VAL, C4DC+C5OH, C6DC, C8, C18:1.

Cluster correlation heat map
The heatmap (Fig. 1)   (OR=·1.27 95%CI= 1.00-3.27) were statistically associated with GDM. These factors can work as independent risk factors involve in the process of GDM.

A nomogram for predicting macrosomia
Macrosomia is relatively associated with GDM. Here, we investigated the clinical characteristics, AA and AC metabolite between GDM with macrosomia and GDM without macrosomia (Table 5 and Figure 2). We found pre-pregnancy weight, BMI, weight gain, LEU+ILE+PRO-OH, TYR, Val, C0, C2, C3, C16, C18 were higher (P< 0.05). In multivariate analysis, we found pre-pregnancy BMI, weight gain, C0, C3, C16, LEU, TYR were evaluated (P< 0.05). These factors are independent risk factors involve in the process of GDM-induced macrosomia. The nomogram predicting GDM induced macrosomia incorporated these significant variables pre-pregnancy BMI, weight gain, C0, C3, C16, LEU, TYR. Among these metabolites, C0 deficiency showed highest OR (OR=0.759, 95%CI= 0.50-0.87). Vertical lines should be drawn from the correct location from each prognostic factor. "Total points" which could be obtained by add all points axis to the bottom axes to make the conversion into a macrosomia probability.

Discussion
In summary, we have identified metabolic alteration in the second trimester. We found 15 metabolites significantly related to GDM by using PLS-DA approach.
Combined these metabolites with clinical information, multivariable logistic regression demonstrated that pre-pregnancy BMI, weight gain, LEU + ILE + PRO-OH, TYR, C0/acylcarnitine, C0, C3, C16, C18 are independent metabolic risk factors associated with GDM. At last, we developed a nomogram predicting probability of macrosomia based on carnitine related metabolites. Metabolomics is increasingly used in GDM capturing disease-relevant metabolic changes. To our knowledge, this is the first study report metabolic alteration during the second trimester pregnancy and metabolic risk factors in GDM during the second trimester based on carnitine related metabolomics and use nomogram predicting macrosomia.
Several experimental works have indicated characters of carnitine related metabolites during pregnancy. [8,16,17] Total carnitine consisted of free carnitine and acylcarnitine. Free carnitine is required for fatty acid transferring into mitochondrial membrane, the ester carnitines are formed in the process. Ester carnitine presented as acylcarnitine was released into plasma. Fatty acid βoxidation is associated uniquely with energy maintenance for maternal and neonate.
[18] If fatty acid β-oxidation dysfunction, fetal growth and development will occur serious damage, such as macrosomia, preterm, neonatal seizure. Previous studies reported, the level of free carnitine is low not due to more carnitine in ester form.
(16-18) An interesting finding appears there is a pronounced fall of the plasma content of free carnitine, acylcarnitine, total carnitine during pregnancy.
[20] Therefore, metabolomic profiling studies need to found China discipline.
GDM is defined as glucose intolerance dysfunction during pregnancy. The prevalence of GDM is increased from 5% up to 14% in USA.
[21] GDM has adverse effects on both maternal and neonatal health. Identification metabolomics risk factor of GDM is critical to medical diagnosis and to minimize the negative consequence of GDM mother and offspring. In Chen's cohort study the evaluated serum level of five AAs were closely associated with people who were diagnosed with diabetes 10 years later.
[6] Wang's study reported BCAA and AAA level progressively upraised in obese, hyperlipemia, high fructose diet and high fat diet group.
[25] Valine acted as a critical role stand out with 251% increase in the GDM onset process.
[6] Our study is consistent with these previous studies. In our study, the class separation of GDM and normal control was determined by C0, LEU + ILE + PRO-OH, C3, Phe, C18, etc. However, conflict observation were reported metabolic markers in T2DM may not have the same predictive power in GDM, the BCAA did not differ significantly between GDM and control in the first trimester.
[26] The discrepancies may arise from different stage involved in the studies. We think it more appropriate to directly metabolomics in the second trimester.
One model clarified there is a phenomenon obese GDM patients compensate βoxidation increased. This process accelerates catabolism of fatty acids, resulting the acylcarnitine accumulation in mitochondria. Acylcarnitine was transported from mitochondria into plasma. The excess of acylcarnitine in mitochondria is harmful for metabolism dynamic balance due to aerobic glycolysis disturbed. Acylcarnitine C16:1, called hexadecenoy-L-carnitine, for determine step of transporting long-chain fatty acid into mitochondria.
[27] Moreira's study investigated C16:1 and shortchain acylcarnitine increased with two weeks orange juice consumption. These results are in accordance with our observation. We hypothesize that the evaluated AC level may be an early maker and driver in aerobic glycolysis dysfunction.
[17] In our study, we did not find the medium-chain acylcarnitine difference with GDM and normal control. Our Medium-Chain Acylcarnitine data is relatively low in absolute value, and it is difficult to draw a conclusion that the difference is statistically significant. Medium-Chain Acylcarnitine have higher metabolic efficiency and accumulates into short-chain acylcarnitine.
[29] A number of studies have reported the association between AC and insulin resistance. Accumulated AC impaired β cell function and reduced insulin synthesis.
[17] Another AC maybe impaired signaling through mammalian target of rapamycin uncoupling downstream signal transduction of insulin as one possible mechanism.
[30] L-carnitine is beneficial for lowering lipid level by activating fatty acid fatty oxidation reducing the accumulation of AC and also effective in insulin sensitivity. [10] Our study found C0 is a protective factor in macrosomia. C3, C16, LEU, TYR are risk factors in macrosomia. Therefore, the characterization of metabolism may enable us to discover reliable biomarkers and enable us to set up a nomogram model to predict macrosomia.
Carnitine metabolite was investigated in assisted reproduction, obstetrics, pediatrics and other fields. Some papers reported administration L-carnitine induced the production of PGE1 and PGE2 by macrophages impairs fertility in female mice, [34] but oral L-carnitine may be beneficial for total sperm concentration, sperm total motility, sperm morphology.
[35] Serum acylcarnitine metabolomics in first trimester were used to improve the accuracy of preeclampsia diagnosis.
[36] In fetal, carnitine was involved in the production of pulmonary surfactant. [37] Our study first found the AA and AC metabolic characteristics in second trimester pregnancy, the significant difference in the GDM metabolites and independent risk factor affecting GDM. Furthermore, we induced a model to predict macrosomia. We speculated carnitine related metabolomics may be helpful to investigate GDM etiology and prophylactic treatment in GDM induced macrosomia. Understanding metabolic-related signaling pathways can identify novel therapeutic targets. Early pregnancy is a critical period of organ development, which is easily affected by external influences and causes malformation. [38] In the third trimester, it is too late to intervene.
[39] The second trimester is considered a safe phase and some intervention given, such as oral L-carnitine, to reduce the incidence of GDM. Oral Lcarnitine supplementation induce some side effects, more intimated researches are needed.
[40] Our study has several limitations. We should expand the sample size and detect more metabolites to understand the correlation between various metabolites. Our study was a retrospective case control study and just finding GDM risk factor, the real metabolomics value need cohort to confirm.
Our observations have important significance in public health and clinical implications. We confirmed the difference metabolites in GDM. These may participate in the process of pathological formation. We can develop targeted therapies of these metabolic patterns in future. The macrosomia model can estimate birthweight more accurate. This will guide the termination of pregnancy, avoid the risk of shoulder dystocia, and reduce the rate of cesarean section. We will give interventions in the second trimester to reduce the incidence of pregnancy complications.

Conclusion
In conclusion, by analyzing the metabolic alteration in the second trimester, we discovered the metabolic profiling in different subgroups according to clinical information. We obtained the metabolic trend is similar with maternal result in fetal clinical information. The major novel finding of this study was that pre-pregnancy BMI, weight gain, LEU + ILE + PRO-OH, TYR, C0/acylcarnitine, C0, C3, C16, C18 are independent risk factors associated with GDM. Pre-pregnancy BMI, weight gain, C0, C3, C16, LEU, TYR are independent risk factors involve in the process of GDMinduced macrosomia. Based on these metabolic factors and clinical information, a nomogram was conducted to predict the macrosomia. Our findings from this research serve as a reminder in the future to study and help understanding the underlying biochemical pathology of GDM and reduce the incidence of GDM.

Acknowledgements
We thank all patients for their participation.

Qiong Luo and Man Sun conceived and designed the experiments. Binqiao Wang and
Yunping Ding collected the data. Ruopeng Weng and Sainan He analyzed and interpreted the data. Man Sun and Baihui zhao wrote the paper. All authors reviewed and approved the paper.

Funding
This research was supported by Natural Science Foundation of Zhejiang province, China (Grant No. LQ20H040008 LY20H040009).

Availability of data and materials
The datasets during and/or analyzed during the current study available from the corresponding author on reasonable request. Ethics approval and consent to participate This study was approved by the Ethics Committee of Women's Hospital, School of Medicine, Zhejiang University. All study participants were included in the study after giving written informed consent.

Consent for publication
Not applicable.

Competing interests
The authors declare no conflicts of interest.  tables        Correlation heatmap of all targeted metabolites. In red are two clusters of strongly intercorre Figure 3 Carnitine-based nomogram for predicting macrosomia, predictor points (Points^ scale; top) c