Nuclear magnetic resonance spectroscopy based investigation of placenta and serum metabolic signatures in an L-NAME-induced preeclampsia rat model

doi:10.21203/rs.3.rs-4620605/v1

Download PDF

Research Article

Nuclear magnetic resonance spectroscopy based investigation of placenta and serum metabolic signatures in an L-NAME-induced preeclampsia rat model

https://doi.org/10.21203/rs.3.rs-4620605/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

Introduction

Despite numerous studies investigating potential metabolic biomarkers of preeclampsia (PE), there remains a dearth of comprehensive animal model validation to ascertain the utility of these metabolic markers as predictors for PE risk.

Objective

To characterize the serum and placental metabolome profiles in a N-nitro-L-arginine methyl ester (L-NAME) induced PE rat model, and identify potential differential metabolites to improve disease classification based on existing diagnostic markers.

Methods

The serological indicators and uterine artery blood flow indexes were measured in L-NAME-induced early-onset PE (EOPE) and late-onset PE (LOPE) rat models. Untargeted proton nuclear magnetic resonance metabolomics analysis was employed to compare serum and placenta tissue extracts, aiming to identify distinguishing metabolites. The random forest decision tree algorithm was utilized for validating biomarker metabolites and establishing classification models.

Results

Several characteristic metabolites associated with the development of preeclampsia in humans can be validated using L-name-induced PE as a model. EOPE and lope exhibit distinct and overlapping metabolic signatures, indicating unique and shared pathways. A combination of two serum metabolites (aspartate/ tryptophan/ 3-hydroxybutyrate) could effectively discriminate between different groups. Moreover, the classification accuracy of placental metabolites (creatine/acetate) combined with serum tryptophan surpasses that of existing serum markers and uterine artery blood flow indexes.

Conclusion

The identification of serum and placental metabolites holds potential to augment the classification accuracy of existing biomarkers. Nevertheless, comprehensive validation of metabolic markers necessitates further support from additional PE animal models.

Metabolomics

Proton nuclear magnetic resonance

Preeclampsia

Placenta

Serum

Preeclampsia (PE) is a severe multisystem disorder that occurs in 2–8% of pregnancies and represents a leading cause of maternal and fetal morbidity and mortality worldwide (Creswell et al. 2023). Based on the timing of onset and clinical presentation, PE can be further classified into two distinct phenotypes: early-onset PE (EOPE, < 34 weeks of gestation) and late-onset PE (LOPE, ≥ 34 weeks of gestation) (Mészáros, Kukor, and Valent 2023). Currently, the integration of soluble fms-like tyrosine kinase-1 (sFlt-1), placental growth factor (PlGF), pregnancy-associated plasma protein A (PAPP-A), uterine artery blood flow (UABF), and other variables into mathematical algorithms has the potential to enhance the prediction and early detection of PE (O’Gorman et al. 2016; Black et al. 2019). However, it is important to note that there is still no predictive method with sufficient performance characteristics to reliably identify individuals who will develop PE (Rybak-Krzyszkowska et al. 2023).

In recent years, metabolomics has emerged as a powerful tool for identifying specific metabolic signatures and pathways related to PE (Mayrink et al. 2022). This advancement may contribute to a more comprehensive understanding of the pathophysiology of PE and facilitate its risk assessment. Several metabolites such as creatinine, glycine, isoleucine, glucose decanoylcarnitine, 3-hydroxyisovaleric acid, and octenoylcarnitine have frequently been reported in biological samples from patients with PE including serum, plasma urine, and placenta (Yao et al. 2022). However, it should be noted that factors such as sample size, age, diet, BMI, and comorbidities can significantly influence the metabolome composition in human biospecimens. Therefore, independent validation studies, such as animal and cell experiments, must be conducted before translating these results into clinical practice (Yao et al. 2022).

The nitric oxide synthase inhibitor, N-nitro-L-arginine methyl ester (L-NAME), has been demonstrated to induce typical symptoms resembling PE in animal models, including hypertension, proteinuria, renal dysfunction, and an anti-angiogenic state (Ramesar et al., 2011; Bakrania, George, and Granger, 2022). In this study, our aim is to investigate the metabolic alterations in placental and serum samples of PE rats using an untargeted proton nuclear magnetic resonance (NMR) metabolomic approach. Additionally, we seek to identify reliable metabolic markers capable of accurately distinguishing between early-onset and late-onset PE.

Animal handling and tissue collection

The animal experiments adhered to the protocols outlined in the NIH Guide for the Care and Use of Laboratory Animals. A total of 36 female Sprague-Dawley rats, weighing 250 to 300 g, were individually housed under controlled conditions at a temperature of 25°C and subjected to a 12-hour light/dark cycle with ad libitum access to food and water. Following acclimatization, the female rats were co-housed with male rats in a 1:1 ratio, and the presence of a vaginal plug was designated as gestational Day 0 (GD 0). The induction of preeclampsia (PE) was achieved by administering L-NAME at a dose of 50 mg/kg/day. On gestational Day 7 (GD 7), pregnant rats were randomly assigned to three groups: the control group (n = 12), early-onset preeclampsia group (EOPE, n = 12), and late-onset preeclampsia group (LOPE, n = 12). The procedures for each group were as follows: i) The control group received daily subcutaneous injections of an equal volume of 0.9% normal saline from GD 7 to GD 19; ii) The EOPE group was subcutaneously injected with L-NAME from GD 7 to GD 19; iii) The LOPE group received an equal volume of 0.9% normal saline from GD 7 to GD 13 and then was subcutaneously injected with L-NAME starting from GD 14 to GD 19 (Kasture et al. 2019).

On GD 6, 11, and 18 of pregnancy, blood pressure, body weights, and 24-hour urine total protein were measured. On GD 19, blood flow within the left uterine artery was assessed using a MyLab™ 9VET ultrasound system (Esaote, Italy) equipped with a multifrequency linear transducer (10–22 MHz). Peak systolic velocity (PSV), end diastolic velocity (EDV), uterine artery resistance index (RI), and pulsatility index (PI) were measured over three cardiac cycles and then averaged.

Blood samples were obtained on GD 19 through abdominal aorta puncture after the rats were anesthetized with 3% intraperitoneal pentobarbital sodium (50 mg/kg). The collected blood was then subjected to centrifugation at 2000 rpm for 10 minutes to obtain serum. Serological indicators (sFLT, PLGF and PAPP-A) were measured using rat sFLT ELISA kit (Jianglai Biological, China, catalog number JL48077), rat PLGF ELISA kit (Jianglai Biological, China, catalog number JL11559), and rat PAPP-A ELISA kit (Jianglai Biological, China, catalog number JL10729) following the manufacturer's recommendations, respectively.

Placental tissues were meticulously dissected from each rat, rinsed with cold saline solution, blotted to remove excess fluids and blood, weighed, and rapidly frozen in liquid nitrogen. Both serum and placental tissues were stored at -80°C until further analysis.

NMR spectra acquisition

For serum sample preparation, 500 µL of rat serum were centrifuged at 13,000 rpm for 2 minutes. For placental samples, three frozen placentas from each rat were trimmed into smaller pieces to facilitate the freeze-drying process using a vacuum freeze dryer (Shunzhi, China). The freeze-dried placental samples were weighed and suspended in 1000 µL of purified water. Subsequently, the suspension was extracted using eight cycles of four seconds on/three seconds off cycling in an in-solution ultrasonic process (Sonics VX-130, USA), followed by a 15-minute centrifugation at 13,000 rpm. Both serum and placental supernatants were transferred to separate 0.5 mL ultrafiltration filters with a molecular weight cutoff of 3 kDa (Millipore, USA). After centrifugation at 13,000 rpm for 45 minutes, a total of 450 µL of filtrate from each sample was collected and transferred to clean 2 mL centrifuge tubes. Prior to transfer to an NMR tube (Norell, USA), samples were thoroughly mixed with a DSS standard solution (Anachro Canada) consisting of 50 µL.

The acquisition of NMR spectra for both serum and placenta was performed using a Bruker AV III 600 MHz spectrometer. Serum samples were analyzed using the Carr-Purcell-Meiboom-Gill (CPMG) with a free relaxation duration of 100 ms and an echo time of 250 µs. For placental samples, 1D Nuclear Overhauser Effect Spectroscopy (1D NOESY) was used with a water suppression of 100 ms mixing time along with a 990 ms pre-saturation (~ 80 Hz γB1). For data acquisition, a total of 128 scans were performed with 64K data points under a line-broadening function of 0.3 Hz for each free induction decay (FID). All acquired NMR spectra were manually phase- and baseline-corrected and then referenced to the doublet at 1.32 ppm.

Metabolomics data processing

The collected FID signal was automatically zero-filled and Fourier transformed using the processing module in Chenomx NMR Suite 8.1 software (Chenomx Inc., Canada). Subsequently, experienced technicians using the Chenomx Processor carefully performed phase and baseline correction. All spectra were referenced to the internal DSS standard at 0.0 ppm and analyzed by experienced analysts against the Chenomx Compound Library.

Next, the quantified NMR data were imported into the MetaboAnalyst 6.0 online platform (www.metaboanalyst.ca) for multivariate data analysis (Pang et al. 2024). Log transformation was applied to normalize all metabolites, followed by scaling using the Pareto scaling method. Principal component analysis (PCA) was employed to provide an overview of data distributions. Partial least squares-discrimination analysis (PLS-DA) models were built to classify all three groups, while orthogonal partial least squares-discrimination analysis (OPLS-DA) was used to discriminate between the EOPE or LOPE group and the control group (Mora-Ortiz et al. 2019). Model validation was performed with a 5-fold stratified cross-validation approach (Huang et al. 2021). Metabolites with variable importance in the projection (VIP) values greater than 1.0 and adjusted P-values less than 0.05 were selected for further investigation of their quantitative changes across all three studied groups (Van De Wetering et al. 2023). Methods based on fold-change (FC) were also utilized to evaluate log ratio concentrations among groups and identify significant metabolite variations above a cut-off value. Metabolic pathways were calculated based on Kyoto Encyclopedia of Genes and Genomes (KEGG) annotation.

Statistical analysis

Statistical comparisons were analyzed using GraphPad Prism version 9. Data normality was assessed using the Shapiro-Wilk method. For normally distributed data, one-way analysis of variance (ANOVA) was performed, while non-normally distributed data were analyzed using the Kruskal-Wallis test for multiple comparisons. P-values were adjusted for multiple comparisons using the Bonferroni correction for ANOVA or the Dunn correction for the Kruskal-Wallis test. A significance level of 0.05 was considered to indicate statistical significance. Additionally, a Random Forest classification model was implemented using Python's sklearn library version 0.16.0 to address complex interactions between serum and placental biomarkers. The aim was to determine the optimal number of biomarkers required for efficient discrimination among the study groups.

Pathophysiological features of L-NAME-induced PE rats

As presented in Table 1, prior to L-NAME injection (GD6), there were no significant differences (P > 0.05) in BP levels and proteinuria among the groups. Administration of L-NAME resulted in a substantial increase in maternal blood pressure and proteinuria. Compared to the control group, the EOPE group exhibited significantly higher systolic BP, diastolic BP, mean arterial pressure, 24h total urinary protein, and embryonic mortality rates. On GDs 18, Systolic BP, Diastolic BP, mean arterial pressure and 24h total urinary protein showed significant increases in the LOPE group.

Clinical screening algorithms have been developed and validated to predict PE by combining Doppler ultrasound measurements of maternal uterine artery resistance with levels of circulating sFLT, PLGF, and PAPP-A. To further validate our PE model, we also investigated the effects of L-NAME treatment on uterine artery blood flow UABF parameters and serological indicators (Table 1). Rats in the EOPE group that received L-NAME from GD7 to GD19 exhibited significantly increased RI, PI, and sFlt-1 levels but decreased PSV, EDV, PAPP-A, and PLGF levels compared to the control group. Additionally, significant changes were observed in indicators such as EDV, sFLT, and PAPP-A in the LOPE group, while other indicators showed similar trends as those seen in the EOPE group. These findings suggest that administration of L-NAME induces pathophysiological changes resembling a PE-like syndrome in rats, with varying severity between the EOPE and LOPE groups.

Table 1

Pathophysiological features of L-NAME-induced EOPE and LOPE rats
	control (n = 12)	EOPE (n = 12)	LOPE (n = 12)
Maternal body weight (g), mean ± SD
GD6	265.58 ± 4.04	267.67 ± 4.09	278.58 ± 4.74
GD11	294.33 ± 3.90	290.17 ± 3.91	307.33 ± 4.99^△
GD18	311.92 ± 10.03	317.50 ± 6.23	332.42 ± 4.56
24h urinary protein (mg/24h), median ± IQR
GD6	136.62 ± 42.60	149.95 ± 72.01	110.95 ± 65.86
GD11	116.40 ± 89.15	216.11 ± 160.17^*	133.65 ± 73.68^△
GD18	116.65 ± 25.26	195.04 ± 107.01^**	259.73 ± 105.56^##
Systolic BP (mmHg), mean ± SD
GD6	105.79 ± 1.83	101.67 ± 2.22	107.52 ± 1.97
GD11	108.21 ± 2.74	128.52 ± 3.17^**	108.89 ± 2.83^△△
GD18	103.47 ± 1.98	136.50 ± 4.60^**	131.41 ± 2.88^##
Diastolic BP (mmHg), mean ± SD
GD6	85.15 ± 2.03	80.95 ± 2.31	84.21 ± 2.27
GD11	78.38 ± 3.59	103.07 ± 2.89^**	82.41 ± 2.19^△△
GD18	79.92 ± 3.21	107.47 ± 4.19^**	103.30 ± 3.50^##
Mean arterial pressure (mmHg), mean ± SD
GD6	92.03 ± 1.86	87.85 ± 2.20	91.98 ± 2.09
GD11	88.32 ± 3.18	111.55 ± 2.93^**	91.23 ± 2.07^△△
GD18	87.77 ± 2.68	117.15 ± 4.13^**	112.67 ± 3.15^##
Embryonic mortality (%), mean ± SD	2.55 ± 1.51	16.82 ± 6.70^**	7.06 ± 2.80^△
Embryo weight (g), mean ± SD	3.57 ± 0.35	3.72 ± 0.38	3.14 ± 0.43
PSV (cm/s), mean ± SD	66.32 ± 3.61	49.79 ± 4.74^*	54.31 ± 4.12
EDV (cm/s), mean ± SD	25.47 ± 1.72	12.34 ± 1.41	17.58 ± 2.31^#
RI, mean ± SD	0.62 ± 0.02	0.76 ± 0.02^**	0.68 ± 0.02^△
PI, mean ± SD	1.10 ± 0.07	1.80 ± 0.13^**	1.45 ± 0.12
sFlt-1 (pg/mL), mean ± SD	47.57 ± 2.08	64.04 ± 1.90^**	56.20 ± 1.63^#△
PLGF (pg/mL), mean ± SD	32.53 ± 0.88	24.89 ± 1.05^**	30.44 ± 0.96^△△
PAPP-A (ng/ml), mean ± SD	198.73 ± 5.01	136.37 ± 4.70^**	172.45 ± 4.64^#△△

The data were analysed using One-way ANOVA followed with Bonferroni multiple comparison or Kruskal-Wallis test followed with Dunn correction depending on data distribution. ^*P < 0.05, ^**P < 0.001 for comparisons between EOPE and control; ^#P < 0.05, ^##P < 0.001 for comparisons between LOPE and control; ^△P < 0.05, ^△△P < 0.001 for comparisons between EOPE and LOPE.

L-NAME-induced metabolic changes

In our study, we quantified a total of 49 metabolites in serum samples and 38 metabolites in placental samples. After acquiring NMR spectra, we performed exploratory analyses using a PCA model. However, the established models failed to distinguish between the three groups for both serum and placental samples. Consequently, we employed a PLS-DA model to investigate group discrimination. Figure 1 exhibits the three-dimensional PLS-DA score plot, which clearly demonstrates separation between the control group and the LOPE group in both sample types, with some observed overlap between the EOPE group and others. VIP scores indicate the most crucial chemical shifts for discrimination. In serum samples, the PLS-DA metabonomics model yielded an R² value of 0.97, with 13 metabolites significantly contributing to group separation, including tryptophan, aspartate, betaine, acetone, creatine, isobutyrate, and mannose. For placental samples with an R² of 0.76, we identified 12 metabolites involved in group segregation: phenylalanine, glycine, tyrosine, isobutyrate, acetate, isoleucine, glutamine, 2-hydroxybutyrate, pyroglutamate, choline, propylene glycol, and lactate.

To further investigate the metabolic changes induced by L-NAME at different time points, we subsequently conducted OPLS-DA to compare pairwise groups. OPLS-DA analysis allowed us to separate the control and L-NAME treated groups into distinct clusters in both serum and placental samples. Similarly, the discrimination models of the EOPE and LOPE groups remained robust following a 5-fold cross-validation (Fig. 2).

Compared to the control group, serum samples from the EOPE group exhibited significant increases in tyrosine, betaine, 3-hydroxybutyrate, creatinine, and acetate; along with reductions in tryptophan, isobutyrate, phenylalanine, glutamine, succinate, valine, lactate, and lysine. The LOPE group showed significantly increased levels of betaine but decreased levels of glutamine, choline, acetate, isoleucine, glycine, isobutyrate, and tyrosine. Four significant metabolites were identified with the same trend in both EOPE and LOPE groups (Fig. 3a).

The placental metabolome was found to be quite different from serum metabolism. Early-onset L-NAME stimulation significantly increased creatine, lysine, and lactate levels while decreasing ethanol levels. On the other hand, later-onset L-NAME stimulation was characterized by abundant changes, including decreased concentrations of phenylalanine, glycine, isoleucine, tyrosine, acetate, and isobutyrate, accompanied by an increase in lactate. Notably, lactate-an important indicator of hypoxia-was elevated in both EOPE and LOPE placental samples but showed an opposite regulatory trend in serum samples (Fig. 3b).

We also observed significant differences in metabolic profiles between the EOPE and LOPE groups. Particularly in placental tissue, 11 metabolites showed significant concentration differences between the LOPE and EOPE groups (Fig. 3b). In addition, there were significant changes in 6 metabolites in serum samples between the LOPE group and the EOPE group.

Pathway enrichment analysis

A metabolic pathway analysis was conducted for serum and placenta, as depicted in Fig. 4. Enriched metabolic pathways were considered significant if defined by a -log10(P) ≥ 2. In serum samples, the EOPE group showed significant alterations in the biosynthesis and metabolism of phenylalanine, tyrosine, and tryptophan, as well as butanoate metabolism. In contrast, the LOPE group exhibited notable changes in alanine, aspartate, and glutamate metabolism; glyoxylate and dicarboxylate metabolism; arginine biosynthesis; butanoate metabolism; the tricarboxylic acid (TCA) cycle; glycine, serine, and threonine metabolism; nitrogen metabolism; and histidine metabolism. Furthermore, distinct differences in metabolic pathway characteristics were observed between the EOPE and LOPE groups. Six metabolic pathways were downregulated, including pyruvate metabolism, glycolysis/gluconeogenesis, alanine, aspartate, and glutamate metabolism, glyoxylate and dicarboxylate metabolism, cysteine and methionine metabolism, and glycine, serine, and threonine metabolism.

Regarding placental samples, both pyruvate metabolism and glycolysis/gluconeogenesis significantly influenced glycine levels in both the EOPE and LOPE groups. Additionally, differences were observed in the LOPE group for glyoxylate and dicarboxylate metabolism, phenylalanine, tyrosine, and tryptophan biosynthesis, as well as phenylalanine metabolism. When compared to the EOPE group, four metabolic pathways were found to be altered in the LOPE group: phenylalanine, tyrosine, and tryptophan biosynthesis; glycine, serine, and threonine metabolism; phenylalanine metabolism; and pyruvate metabolism.

Validation of key biomarkers for PE discrimination

We employed a series of machine learning models to assess classification performance using serological indicators, UABF parameters, serum metabolites, and placental metabolites. Performance evaluation was conducted using metrics such as accuracy, precision, recall, and the F1 score obtained from 5-fold cross-validations. Among these models, the random forest model demonstrated the best performance and was selected for further optimization and feature subset selection.

As presented in Table 2, satisfactory classification accuracy and precision were achieved using two- or three-feature models. These included models based on serum 3-hydroxybutyrate combined with either serum aspartate or serum tryptophan, or placental creatine along with both serum tryptophan and placental acetate. Models using serological indicators alone or in combination with the placental metabolome also showed good classification accuracy, exceeding 90%. However, the model constructed using UABF parameters alone exhibited the lowest performance compared to other models.

Furthermore, models employing only the placental metabolome or combining it with placental hemodynamic parameters for preeclampsia (PE) differentiation were found to be insufficient. This was due to a decrease in classification accuracy as the model complexity was reduced.

Table 2

Performance comparison for random forest models with different combinations
Model component	Group	Accuracy	Precision	Recall	F1-score	Selected markers
Serological indicators	control	0.92	0.91	0.83	0.87	sFLT, PLGF, PAPP-A
	EOPE		1.00	1.00	1.00
	LOPE		0.85	0.92	0.88
UABF parameters	control	0.42	0.64	0.58	0.61	PI, RI, PSV, EDV
	EOPE		0.43	0.5	0.46
	LOPE		0.18	0.17	0.17
Placenta metabolites	control	0.83	0.71	0.83	0.77	Top10 important features
	EOPE		0.92	0.92	0.92
	LOPE		0.9	0.75	0.82
Serological indicators + UABF parameters	control	0.81	0.82	0.75	0.78	sFLT, PAPP-A, EDV, PLGF
	EOPE		0.86	1.00	0.92
	LOPE		0.73	0.67	0.7
Serological indicators + Placenta metabolites	control	0.94	0.92	0.92	0.92	sFlt-1, glycine, lysine
	EOPE		0.92	1.00	0.96
	LOPE		1.00	0.92	0.96
UABF parameters + Placenta metabolites	control	0.86	0.82	0.75	0.78	Top10 important features
	EOPE		0.86	1.00	0.92
	LOPE		0.91	0.83	0.87
Serological indicators + UABF parameters + placenta metabolites	control	0.89	0.92	0.92	0.92	glycine, PAPP-A, uracil, choline, acetate
	EOPE		0.92	0.92	0.92
	LOPE		0.83	0.83	0.83
Serum metabolites	control	1.00	1.00	1.00	1.00	Model 1: 3-hydroxybutyrate, aspartate Model 2: aspartate, tryptophan Model 3: 3-hydroxybutyrate, tryptophan
	EOPE		1.00	1.00	1.00
	LOPE		1.00	1.00	1.00
Serum metabolites + placenta metabolites	control	1.00	1.00	1.00	1.00	Model 1: placenta creatine, serum tryptophan, placenta acetate
	EOPE		1.00	1.00	1.00
	LOPE		1.00	1.00	1.00

To validate the metabolomic and metabolic biomarker findings associated with preeclampsia (PE) in humans, we conducted a comprehensive metabolomic analysis using NMR spectroscopy on the L-NAME-induced rat model of PE. Our results revealed both shared and distinct metabolic alterations induced by L-NAME, observable in both serum and placental samples.

A total of 18 and 11 differential metabolites significantly impacting PE, including EOPE and LOPE, were identified in the serum and placenta, respectively. When compared to the 33 high-frequency metabolic biomarkers previously reviewed, our findings indicate that succinate, valine, creatinine, phenylalanine, lactate, glutamate, choline, and 3-hydroxybutyric acid show consistent regulation trends in rat serum samples as observed in human serum; However, this consistency does not extend to glutamine (Yao et al. 2022). In contrast, the regulation trends of phenylalanine, glycine, and isoleucine in rat placenta samples are contrary to those in human placenta (Yao et al. 2022; Zhou et al. 2017). These findings suggest that the L-NAME-induced PE model partially elucidates certain aspects of human metabolomic changes associated with PE development. Moreover, our results indicate a higher degree of concordance between the rat serum metabolome and human serum compared to rat placenta samples and human placenta. Among the identified serum metabolites, choline, phenylalanine and glutamine are implicated in vascular dysfunction and hypertension development (Golzarand et al. 2021; Yang et al. 2020; Heikal et al. 2018; Wang et al. 2021; Durante 2019); glutamine and succinate are associated with mitochondrial dysfunction during PE (Bornstein et al. 2023; Murphy and O’Neill 2018); valine and 3-hydroxybutyric acid levels can serve as biomarkers for insulin resistance and impaired glucose tolerance (Yoon 2016; Zhang et al. 2016); creatinine, creatine, and lactate contribute to the decline in renal function process (Yao et al. 2022). Additionally, we compared our results with the placenta metabolome of the reduced uterine perfusion pressure model of PE (McClements et al. 2022). In both models, the expression trends of two differential metabolites, glutamine and acetate, were consistent; however, the expression trends of three metabolites, lactate, isoleucine, and glycine, exhibited opposite patterns. Therefore, there are both consistencies and differences between the two models that highlight the need for furt

The utilization of L-NAME-induced rats as models for EOPE and LOPE is well-established in the literature (Ramesar et al. 2011; Bakrania, George, and Granger 2022). However, to our knowledge, no study has yet explored the metabolomic differences between these two subtypes. In our investigation of PE models, significant variations in metabolic profiles were observed among the control, EOPE, and LOPE groups. Both PLS-DA and OPLS-DA effectively discriminated between serum and placental samples from all three groups. In pairwise comparisons, we identified tryptophan, isobutyrate, lactate, and betaine as four differential metabolites with consistent expression trends in both LOPE and EOPE within normal serum samples. Notably, lactate was the only metabolite showing common upregulation in both LOPE and EOPE within placental tissue, yet it exhibited an opposite trend in serum. Lactate is pivotal in pyruvate metabolism and glycolysis/gluconeogenesis pathways during pregnancy, providing an essential energy source for fetal tissues and regulating immune cells, thus promoting immune tolerance towards the allogeneic fetus (Ma et al. 2020). In cases of PE, where maternal spiral artery remodeling is deficient, leading to persistent hypoxia, lactate production is increased (Peguero et al. 2019). Lactate also acts as an epigenetic regulator at the maternal-fetal interface and serves as an immune modulator and angiogenic promoter, highlighting its importance as a fetal energy source for maintaining pregnancy (Ma et al. 2020; Witkin 2018; Fan et al. 2023). Studies have shown that serum lactate dehydrogenase (LDH) levels are elevated in PE patients and can serve as a biomarker for disease severity (Deeksha et al. 2024; Jaiswar et al. 2012). We hypothesize that decreased serum lactate levels may result from increased LDH activity, catalyzing the oxidation of lactate to propionic acid. Tryptophan, an essential amino acid, is involved in the regulation of inflammatory responses, immune tolerance, and vascular tone during pregnancy (Broekhuizen et al. 2020). Decreased serum tryptophan levels have been correlated with the development of gestational hypertension and PE (Gupta et al. 2023). Betaine, a key choline metabolite and methyl donor, plays a role in fat metabolism, fetal gene expression, epigenetic regulation, maternal glucose metabolism, and redox balance, contributing to normal fetal growth and neurodevelopmental outcomes in children (Salahi et al. 2020). Elevated betaine concentrations and its derivatives have been observed in PE patients (Xu et al. 2024).

In addition to the common metabolic differences between early-onset preeclampsia (EOPE) and late-onset preeclampsia (LOPE) mentioned earlier, our findings revealed significant downregulation of six metabolites in LOPE compared to EOPE, primarily associated with gluconeogenesis and amino acid metabolism in serum samples. Specifically, pyruvate, a key intermediate in various cellular metabolic pathways, can be generated from glucose via glycolysis under hypoxic conditions or converted back into carbohydrates through gluconeogenesis, and further transformed into glucose and fatty acids by interacting with acetyl-CoA (Yao et al. 2022). The substantial decrease in pyruvate levels during the LOPE period suggests a less severe hypoxic status and less disrupted energy metabolism in LOPE compared to EOPE. Moreover, analysis of placental samples identified 11 differential metabolites between EOPE and LOPE, including upregulated levels of seven amino acids in EOPE. These amino acids include both non-essential ones, such as glycine, tyrosine, and creatine, and essential ones like phenylalanine, isoleucine, tryptophan, and lysine. These altered metabolites are primarily involved in the glycine/serine/threonine metabolism pathway and the phenylalanine/tyrosine/tryptophan biosynthesis pathway, indicating a disrupted amino acid metabolism previously observed in PE patients (Prameswari et al. 2022). The elevated nitrogen demand by the placenta during EOPE corresponds to increased oxidative stress levels and a greater need for maternal placental vascular remodeling, which may account for the higher levels of glycine and tyrosine detected in maternal blood (Liu et al. 2019). During the EOPE period, the damaged placenta may adapt by increasing tissue creatine concentrations, likely reflecting a heightened reliance on the creatine kinase pathway to stabilize placental bioenergetics under chronic hypoxic conditions (Ellery et al. 2019). The upregulation of isoleucine in EOPE may play a pivotal role in regulating cellular growth processes, oxidative stress, lipid dysfunction, and energy metabolism (Youssef et al. 2021; Wang et al. 2023). Elevated levels of lysine also contribute to the regulation of blood pressure homeostasis during the EOPE period (Dasgupta et al. 2023). From a metabolomic perspective, our findings suggest that L-NAME-induced PE occurring at different stages may have distinct pathogeneses, thereby confirming that EOPE and LOPE might possess unique pathological foundations.

Identifying potential predictive markers for preeclampsia (PE) is essential for the early detection and prevention strategies in high-risk pregnancies, attracting considerable research interest. Numerous screening and preventive models have been proposed to reduce the impact of PE and improve maternal and perinatal health outcomes. Notably, the Fetal Medicine Foundation (FMF) method combines maternal characteristics with biochemical indicators such as PlGF and PAPP-A, along with biophysical measures like the uterine artery pulse index and mean arterial pressure (O’Gorman et al. 2016). On the other hand, the International Society for the Study of Hypertension in Pregnancy (ISSHP) advocates for the joint assessment of PlGF and sFlt-1 (Black et al. 2019). Yet, many of these predictive models suffer from low sensitivity or high false-positive rates, which limits their utility for accurate prediction (Mészáros, Kukor, and Valent 2023). In our animal validation study, we applied machine learning to analyze the aforementioned methods and observed significant variations in serum markers and ultrasound indicators across different groups. However, when analyzed using a random forest model under anesthesia, ultrasound indicators only achieved 0.42 accuracy in distinguishing between rat groups, whereas serum markers showed a higher accuracy of 0.92. The combination of both methods led to an overfitted model with an accuracy of 0.81. Our results indicate that resting uterine artery blood flow may not reliably differentiate between disease groups in rats due to its instability, whereas serum markers offer higher classification accuracy, thus revealing potential weaknesses in current PE classification systems when viewed from the perspective of a single animal model. Concurrently, we recognize the outstanding capability of metabolomic analysis in classifying animal models of PE. Notably, satisfactory classification accuracy has been achieved either by examining the serum metabolome alone or by integrating it with the placental metabolome. Our proposed models, which rely on two or three features, require validation through analysis of human serum and placental samples.

Although the current findings offer valuable insights into the metabolic mechanisms and subtype differentiation of preeclampsia (PE), it is crucial to recognize the study's limitations. PE, being a multifactorial syndrome, cannot be fully represented by a single L-NAME induction model, which may only confirm select metabolic alterations observed in human PE. This underscores the need for a diverse range of animal models to provide robust external validation for metabolic markers associated with PE. While existing reports on human placental metabolomes have relied on isolated samples, our study also utilized in vitro detection methods for animal tissue samples to ensure consistency. However, for blood or placental specimens, a dynamic monitoring approach should be considered for metabolic analysis. The growing availability of clinical MRI and the advancement of non-invasive in vivo nuclear magnetic resonance (NMR) technology offer promising avenues for obtaining more comprehensive insights into the metabolic characteristics and diagnostic markers of PE.

We employed NMR methodology to explore the metabolic differences in serum and placental samples between L-NAME-induced EOPE and LOPE rats, with the goal of identifying and validating potential biomarkers that are consistent with findings from human studies. Our study represents the inaugural comprehensive metabolic profiling analysis conducted on both serum and placental samples in L-NAME-induced PE rats. We identified significant variations in serum and placental metabolic patterns across different stages of L-NAME-induced PE, as well as common metabolic compounds present throughout various induction stages. Notably, the majority of differentially expressed metabolites, particularly those identified in serum, are in alignment with commonly reported findings from human studies. This concordance provides robust support for further understanding the pathogenesis of PE and for validating differential metabolites observed in humans through the use of animal models. Moreover, we discovered metabolic biomarkers that exhibit superior performance in distinguishing between animal model groups compared to existing clinical diagnostic markers. This finding could potentially streamline the process of screening metabolites for human serum or in vivo NMR analysis.

Ethics approval and consent to participate

All experimental procedures involving animals were approved by Jiangxi Zhonghong Boyuan Biotechnology Co. Ltd Animal Ethics Committee (Approval ID: 2022062001).

Consent for publication

Not applicable.

Availability of data and material

The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.

Competing interests

The authors declare no conflict of interest.

Funding

This study was funded by Zhejiang Provincial Natural Science Foundation of China (Grant number LQ21H040007).

Authors' contributions

The concept for this study was originally conceived by XMX and YYZ. XMX, CCS, YJL, XQL, WHW, and LDG were responsible for conducting the experiments and analyzing the data. The initial draft of the manuscript was prepared by XMX. XKY provided supervision.All authors have reviewed the manuscript, provided feedback, and approved the final version that is submitted for publication.

Acknowledgements

Not applicable.

Bakrania, B. A., George, E. M., & Granger, J. P. (2022). Animal models of preeclampsia: investigating pathophysiology and therapeutic targets. American Journal of Obstetrics and Gynecology, 226(2S), S973-S987. https://doi.org/10.1016/j.ajog.2020.10.025
Black, C., Al-Amin, A., Rolnik, D. L., Kane, S. C., Stolarek, C., White, A., da Silva Costa, F., & Brennecke, S. (2019). Midpregnancy testing for soluble fms-like tyrosine kinase 1 (sFlt-1) and placental growth factor (PlGF): An inter-assay comparison of three automated immunoassay platforms. Placenta, 86, 11-14. https://doi.org/10.1016/j.placenta.2019.08.083
Bornstein, R., Mulholland, M. T., Sedensky, M., Morgan, P., & Johnson, S. C. (2023). Glutamine metabolism in diseases associated with mitochondrial dysfunction. Molecular and Cellular Neuroscience, 126, 103887. https://doi.org/10.1016/j.mcn.2023.103887
Broekhuizen, M., Klein, T., Hitzerd, E., de Rijke, Y. B., Schoenmakers, S., Sedlmayr, P., Danser, A. H. J., Merkus, D., & Reiss, I. K. M. (2020). l-Tryptophan-Induced Vasodilation Is Enhanced in Preeclampsia: Studies on Its Uptake and Metabolism in the Human Placenta. Hypertension, 76(1), 184-194. https://doi.org/10.1161/HYPERTENSIONAHA.120.14970
Creswell, L., O'Gorman, N., Palmer, K. R., da Silva Costa, F., & Rolnik, D. L. (2023). Perspectives on the Use of Placental Growth Factor (PlGF) in the Prediction and Diagnosis of Pre-Eclampsia: Recent Insights and Future Steps. International Journal of Women's Health, 15, 255-271. https://doi.org/10.2147/IJWH.S368454
Dasgupta, S., Subramani, E., Mitra, I., Bhattacharya, A., Sherpa, D. D., Joshi, M., Chakraborty, P., Ray, C. D., & Chaudhury, K. (2023). Discovery of novel metabolic signatures for early identification of women at risk of developing gestational hypertension. Metabolomics, 19(5), 50. https://doi.org/10.1007/s11306-023-02012-y
Deeksha, H. S., Pajai, S., Reddy Eleti, M., & Navalihiremath, V. U. (2024). A Comprehensive Review on Serum Lactate Dehydrogenase (LDH) and Uric Acid in Preeclampsia: Implications for Maternal Health and Disease Severity. Cureus, 16(3), e56928. https://doi.org/10.7759/cureus.56928
Durante, W. (2019). The Emerging Role of L-Glutamine in Cardiovascular Health and Disease. Nutrients, 11(9), 2092. https://doi.org/10.3390/nu11092092
Ellery, S. J., Murthi, P., Davies-Tuck, M. L., Della Gatta, P. A., May, A. K., Kowalski, G. M., Callahan, D. L., Bruce, C. R., Alers, N. O., Miller, S. L., Erwich, J. J. H. M., Wallace, E. M., Walker, D. W., Dickinson, H., & Snow, R. J. (2019). Placental creatine metabolism in cases of placental insufficiency and reduced fetal growth. Molecular Human Reproduction, 25(8), 495-505. https://doi.org/10.1093/molehr/gaz039
Fan, H., Yang, F., Xiao, Z., Luo, H., Chen, H., Chen, Z., Liu, Q., & Xiao, Y. (2023). Lactylation: Novel Epigenetic Regulatory and Therapeutic Opportunities. American Journal of Physiology-Endocrinology and Metabolism, 324(4), E330-E338. https://doi.org/10.1152/ajpendo.00159.2022
Golzarand, M., Bahadoran, Z., Mirmiran, P., & Azizi, F. (2021). Dietary choline and betaine intake and risk of hypertension development: a 7.4-year follow-up. Food & Function, 12(9), 4072-4078. https://doi.org/10.1039/d0fo03208e
Gupta, S., Petras, L., Tufail, M. U., Rodriguez Salazar, J. D., & Jim, B.(2023). Hypertension in Pregnancy: What We Now Know. Current Opinion in Nephrology and Hypertension, 32(2), 153-164. https://doi.org/10.1097/MNH.0000000000000857
Heikal, L., Starr, A., Hussein, D., Prieto-Lloret, J., Aaronson, P., Dailey, L. A., & Nandi, M. (2018). L-Phenylalanine Restores Vascular Function in Spontaneously Hypertensive Rats Through Activation of the GCH1-GFRP Complex. JACC: Basic to Translational Science, 3(3), 366-377. https://doi.org/10.1016/j.jacbts.2018.01.015
Huang, X., Li, C., Tan, K., Wen, Y., Guo, F., Li, M., Huang, Y., Sun, C. Q., Gozin, M., & Zhang, L. (2021). Applying Machine Learning to Balance Performance and Stability of High Energy Density Materials. iScience, 24(3), 102240. https://doi.org/10.1016/j.isci.2021.102240
Jaiswar, S. P., Gupta, A., Sachan, R., Natu, S. N., & Shaili, M. (2011). Lactic Dehydrogenase: A Biochemical Marker for Preeclampsia-Eclampsia. Journal of Obstetrics and Gynaecology of India, 61(6), 645-648. https://doi.org/10.1007/s13224-011-0093-9
Kasture, V., Sundrani, D., Dalvi, S., Swamy, M., Kale, A., & Joshi, S. (2019). Maternal Omega-3 Fatty Acids and Vitamin E Improve Placental Angiogenesis in Late-Onset but Not Early-Onset Preeclampsia. Molecular and Cellular Biochemistry, 461(1-2), 159-170. https://doi.org/10.1007/s11010-019-03599-4
Liu, G., Deng, W., Cui, W., Xie, Q., Zhao, G., Wu, X., Dai, L., Chen, D., & Yu, B. (2020). Analysis of Amino Acid and Acyl Carnitine Profiles in Maternal and Fetal Serum from Preeclampsia Patients. Journal of Maternal-Fetal & Neonatal Medicine, 33(16), 2743-2750. https://doi.org/10.1080/14767058.2018.1560407
Ma, L. N., Huang, X. B., Muyayalo, K. P., Mor, G., & Liao, A. H. (2020). Lactic Acid: A Novel Signaling Molecule in Early Pregnancy? Front Immunol, 11, 279. https://doi.org/10.3389/fimmu.2020.00279
Mayrink, J., Leite, D. F., Nobrega, G. M., Costa, M. L., & Cecatti, J. G. (2022). Prediction of pregnancy-related hypertensive disorders using metabolomics: a systematic review. BMJ Open, 12(4), e054697. https://doi.org/10.1136/bmjopen-2021-054697
McClements, L., Richards, C., Patel, N., Chen, H., Sesperez, K., Bubb, K. J., Karlstaedt, A., & Aksentijevic, D. (2022). Impact of reduced uterine perfusion pressure model of preeclampsia on metabolism of placenta, maternal and fetal hearts. Scientific Reports, 12(1), 1111. https://doi.org/10.1038/s41598-022-05120-2
Mészáros, B., Kukor, Z., & Valent, S. (2023). Recent Advances in the Prevention and Screening of Preeclampsia. Journal of Clinical Medicine, 12(18), 6020. https://doi.org/10.3390/jcm12186020
Mora-Ortiz, M., Nuñez Ramos, P., Oregioni, A., & Claus, S. P. (2019). NMR metabolomics identifies over 60 biomarkers associated with Type II Diabetes impairment in db/db mice. Metabolomics, 15(6), 89. https://doi.org/10.1007/s11306-019-1548-8
Murphy, M. P., & O'Neill, L. A. J. (2018). Krebs Cycle Reimagined: The Emerging Roles of Succinate and Itaconate as Signal Transducers. Cell, 174(4), 780-784. doi: 10.1016/j.cell.2018.07.030
O'Gorman, N., Wright, D., Syngelaki, A., Akolekar, R., Wright, A., Poon, L. C., & Nicolaides, K. H. (2016). Competing risks model in screening for preeclampsia by maternal factors and biomarkers at 11-13 weeks gestation. American Journal of Obstetrics and Gynecology, 214(1), 103.e1-103.e12. doi: 10.1016/j.ajog.2015.08.034
Pang, Z., Lu, Y., Zhou, G., Hui, F., Xu, L., Viau, C., Spigelman, A. F., MacDonald, P. E., Wishart, D. S., Li, S., & Xia, J. (2024). MetaboAnalyst 6.0: towards a unified platform for metabolomics data processing, analysis and interpretation. Nucleic Acids Research, gkae253. doi: 10.1093/nar/gkae253
Peguero, A., Parra, R. A., Carrillo, S. P., Rojas-Suarez, J., & Figueras, F. (2019). Association of plasma lactate concentration at admission of severe preeclampsia to maternal complications. Pregnancy Hypertension, 17, 89-93. doi: 10.1016/j.preghy.2019.05.003
Prameswari, N., Irwinda, R., Wibowo, N., & Saroyo, Y. B. (2022). Maternal Amino Acid Status in Severe Preeclampsia: A Cross-Sectional Study. Nutrients, 14(5), 1019. doi: 10.3390/nu14051019
Ramesar, S. V., Mackraj, I., Gathiram, P., & Moodley, J. (2011). Sildenafil citrate decreases sFlt-1 and sEng in pregnant l-NAME treated Sprague-Dawley rats. European Journal of Obstetrics & Gynecology and Reproductive Biology, 157(2), 136-140. doi: 10.1016/j.ejogrb.2011.03.005
Rybak-Krzyszkowska, M., Staniczek, J., Kondracka, A., Bogusławska, J., Kwiatkowski, S., Góra, T., Strus, M., & Górczewski, W. (2023). From Biomarkers to the Molecular Mechanism of Preeclampsia-A Comprehensive Literature Review. International Journal of Molecular Sciences, 24(17), 13252. doi: 10.3390/ijms241713252
Salahi, P., Gharabaghi, M., Rocky, A., & Alirezaei, M. (2020). In vivo: maternal betaine supplementation normalized fetal growth in diabetic pregnancy. Archives of Gynecology and Obstetrics, 302(4), 837-844. doi: 10.1007/s00404-020-05665-5
van de Wetering, R., Vorster, J. A., Geyrhofer, S., Harvey, J. E., Keyzers, R. A., & Schenk, S. (2023). Behavioral metabolomics: how behavioral data can guide metabolomics research on neuropsychiatric disorders. Metabolomics, 19(8), 69. doi: 10.1007/s11306-023-02034-6
Wang, Y., Rong, X., Guan, H., Ouyang, F., Zhou, X., Li, F., Tan, X., & Li, D. (2023). The Potential Effects of Isoleucine Restricted Diet on Cognitive Impairment in High-Fat-Induced Obese Mice via Gut Microbiota-Brain Axis. Molecular Nutrition & Food Research, 67(20), e2200767. doi: 10.1002/mnfr.202200767
Wang, Z., Cheng, C., Yang, X., & Zhang, C. (2021). L-phenylalanine attenuates high salt-induced hypertension in Dahl SS rats through activation of GCH1-BH4. PLoS One, 16(4), e0250126. doi: 10.1371/journal.pone.0250126
Witkin, S. S. (2018). Lactic acid alleviates stress: good for female genital tract homeostasis, bad for protection against malignancy. Cell Stress and Chaperones, 23(3), 297-302. doi: 10.1007/s12192-017-0852-3
Xu, H., Feng, P., Sun, Y., Wu, D., Wang, D., Wu, L., Peng, H., & Li, H. (2024). Plasma trimethylamine N-oxide metabolites in the second trimester predict the risk of hypertensive disorders of pregnancy: a nested case-control study. Hypertension Research, 47(3), 778-789. doi: 10.1038/s41440-023-01563-w
Yang, C., Zhao, Y., Ren, D., & Yang, X. (2020). Protective Effect of Saponins-Enriched Fraction of Gynostemma pentaphyllum against High Choline-Induced Vascular Endothelial Dysfunction and Hepatic Damage in Mice. Biological & Pharmaceutical Bulletin, 43(3), 463-473. doi: 10.1248/bpb.b19-00805
Yao, M., Xiao, Y., Yang, Z., Ge, W., Liang, F., Teng, H., Gu, Y., & Yin, J. (2022). Identification of Biomarkers for Preeclampsia Based on Metabolomics. Clinical Epidemiology, 14, 337-360. doi: 10.2147/CLEP.S353019
Yoon, M. S. (2016). The Emerging Role of Branched-Chain Amino Acids in Insulin Resistance and Metabolism. Nutrients, 8(7), 405. doi: 10.3390/nu8070405
Youssef, L., Simões, R. V., Miranda, J., García-Martín, M. L., Paules, C., Crovetto, F., Amigó, N., Cañellas, N., Gratacos, E., & Crispi, F. (2021). Paired maternal and fetal metabolomics reveal a differential fingerprint in preeclampsia versus fetal growth restriction. Scientific Reports, 11(1), 14422. doi: 10.1038/s41598-021-93936-9.
Zhang, Q., Ford, L. A., Goodman, K. D., Freed, T. A., Hauser, D. M., Conner, J. K., Vroom, K. E., & Toal, D. R. (2016). LC-MS/MS method for quantitation of seven biomarkers in human plasma for the assessment of insulin resistance and impaired glucose tolerance. Journal of Chromatography B: Analytical Technologies in Biomed Life Sciences. doi: 10.1016/j.jchromb.2016.10.025
Zhou, X., Han, T. L., Chen, H., Baker, P. N., Qi, H., & Zhang, H. (2017). Impaired mitochondrial fusion, autophagy, biogenesis and dysregulated lipid metabolism is associated with preeclampsia. Experimental Cell Research, 359(1), 195-204. doi: 10.1016/j.yexcr.2017.07.029.

No competing interests reported.

Download PDF

Reviewers invited by journal
11 Aug, 2024
Editor invited by journal
26 Jun, 2024
Editor assigned by journal
25 Jun, 2024
Submission checks completed at journal
25 Jun, 2024
First submitted to journal
22 Jun, 2024

You are reading this latest preprint version

Nuclear magnetic resonance spectroscopy based investigation of placenta and serum metabolic signatures in an L-NAME-induced preeclampsia rat model

Status:

Version 1

Abstract

Introduction

Objective

Methods

Results

Conclusion

Figures

Introduction

Methods

Animal handling and tissue collection

NMR spectra acquisition

Metabolomics data processing

Statistical analysis

Results

Pathophysiological features of L-NAME-induced PE rats

L-NAME-induced metabolic changes

Pathway enrichment analysis

Validation of key biomarkers for PE discrimination

Discussion

Conclusion

Declarations

References

Additional Declarations

Status:

Version 1