A novel approach for testing the teratogenic potential of chemicals on the platform of metabolomics: studies employing HR-MAS nuclear magnetic resonance spectroscopy

Abstract

NMR based metabolomics offers a complementary approach that gives information on whole-organism functional integrity over time after drug exposure. Hence the objective was to develop a quick, reliable method for testing the teratogenic potential of a new chemical entity (NCE) on the platform of metabonomics, as an alternative to conventional procedures. Time mated Charles Foster rats (n = 9) were injected with cyclophosphamide (0, 5, 15 and 30 mg kg⁻¹) at different dose levels on day 11 of the pregnancy, through an i.p. route. On day 12 of the pregnancy, embryos were procured from six rats out of the 9 pregnant rats from each group, using a per abdominal approach. These embryos were then undertaken for morphological studies and NMR experiments using a high resolution-magic angle spin (HR-MAS) probe. The remaining three rats were followed until term to procure full term pups, which were used for conventional observations and studies. Multivariate unsupervised principal component analysis of the ¹H NMR spectra from the rat embryos revealed a dose dependent cluster separation between the controls and treated specimens, which was further confirmed by supervised partial least-squares discriminant analysis with an R² of 0.77 and Q² of 0.72. Those embryos which were found to have malformations/anomalies, significantly presented a few upregulated (aspartate) and a few downregulated (creatine, choline and glycine) metabolite levels. This work revealed that observed teratogenic ailments and resulting metabolic profiles have a definite correlation, demonstrating the suitability of this procedure for testing teratogenicity, which may give a new dimension to the prioritization of lists having a large number of chemicals to be tested for their teratogenicity. Malformations observed among the full term pups also validated the findings from the embryos.

---

Introduction

Testing new chemical entities (NCEs) for their ‘teratogenic’ potential is a crucial step before human exposure occurs. The conventional method available for teratogenicity testing involves administration of a test compound, during the period of major organogenesis, to rats and collection of the offspring at term. The uterine contents are noted and the foetuses are observed for various anomalies and malformations. However, the conventional method is very time and cost intensive, and the number of drugs and chemicals are increasing substantially. With this hike, and with the ethical involvement in animal experiments and public demand for safer commodities, quick testing with accuracy, precision and expedience is not feasible, rather it would be clearly unrealistic to attempt or to perform a complete conventional test on each and every chemical.

Hence, to overcome this situation, a variety of techniques for testing teratogenicity that do not utilize laboratory animals have been developed during the last 30 years. The culture of invertebrate embryos^1,2 and vertebrate embryos^3,4 as sub mammalian systems, and isolated whole embryo culture of both pre- and post-implantation^5–9 and mammalian embryonic organ culture^10–12 as mammalian systems, have been developed. Out of these methods, emphasis has been given to post-implantation rat and mouse embryo culture. Two more tests systems, micromass culture and embryonic stem cell culture were introduced in the nineties.^13–15 From the overview outlined above, it is intelligible that in vitro teratology has shown an impressive development during the last few decades. However, the in vitro system has still to play a great role in a wider area, for which it needs further development, as from the practical perspective these tests present many limitations and boundaries in themselves, due to which they could not be justified as suitable for the purpose. The limitations of these systems are related to the fact that they not only require high technical skills and high cost, but are also relatively complex and cover only a part of the total gestation or organogenesis. Maternal xenobiotic metabolism is absent from the study, the placental barrier is absent, and complex in vivo interactions and whole animal pharmacokinetics are ignored; only a limited period of embryogenesis can be studied, development in vitro may be abnormal, and most of the methods do not justify an endpoint, which is the most important clause for a test system and can define the teratogenicity.

Metabolomics offers a complementary approach that gives overall information on an organism’s functional integrity over time, after exposure to drugs, on various spectra.¹⁶ Toxicants usually disrupt the normal composition and flux of endogenous biochemicals in or through key intermediate cellular metabolic pathways. By measuring and mathematically modelling changes in the levels of metabolic products found in biological fluids and tissues, metabolomics offers fresh insight into the effects of diet, drugs and disease.¹⁷ If a significant number of trace molecules are monitored, then the overall pattern produced by them in no way limits its usefulness. The fingerprint produced by groups of such molecules may be proven more consistent than any other marker or end point.

Recent trends in instrumentation have emphasized ¹H nuclear magnetic resonance (¹H NMR) based ‘metabonomics’ as a robust method to categorize organ specific toxicity, and as a tool for monitoring the onset and progression of toxicological effects and identifying biomarkers of toxicity. A future challenge, however, is to describe the cellular metabolome for the purpose of understanding cellular function (i.e. metabolomics). NMR, meant for recognizing chemical structure, was being used for the analysis of biofluids a few years back,^18–20 but now it is being applied to tissue systems for gathering various information.^21,22 For generating complete information from tissues, the HR-MAS NMR technique has great advantages over the gold standard method, as the sample can be analyzed in its native form and HR-MAS NMR also provides better resolution of the spectrum, comparable to that of solution-state or conventional NMR spectroscopy,²³ without the loss of metabolite concentration that occurs during solvent extraction for solution state NMR.²⁴ A high field ¹H NMR compiled with pattern recognition technology (PRT) and principal component analysis (PCA) can give such comprehensive information regarding the changes in metabolic pathways due to toxicants and their concentrations. Thus, evaluation of the changes in metabolic consequences due to toxicity can help in the toxicity assessment of a compound.

With the above background information, and considering the capability and advantages of ¹H NMR for providing information on metabolic changes in cells and tissues, it was anticipated that metabolomics can help in describing the function and, subsequently, predicting the teratogenicity and embryotoxicity of a new chemical entity (NCE). Hence the present work is aimed at developing a procedure for testing the teratogenicity of newly formulated drugs and chemicals/NCEs on the platform of metabolomics using ¹H NMR, which can bypass the inherent shortcomings of in vitro systems and will involve less time, cost and laboratory animals. To validate this procedure, cyclophosphamide was used as a test substance because cyclophosphamide (CP), an alkylating agent, is a widely studied and well known pro-teratogen. CP causes embryonic resorptions, foetal malformations, and anomalies involving morphological, visceral and skeletal changes in rats and mice, and in humans too.²⁵ It provides an opportunity, as a model teratogen, for understanding teratogenicity and the mechanism through which the teratogenicity is caused.

Materials and methods

Animals

Young adult male and female Charles Foster rats were obtained from the National Laboratory Animal Centre (NLAC) of the CSIR-Central Drug Research Institute, Lucknow (India), after getting approval from the Institutional Animal Ethics Committee. The rats were maintained in the experimental room of the animal house at CDRI, at standard husbandry conditions i.e. standard temperature (23 ± 2 °C), relative humidity (55 ± 10%) and 12 h light and dark cycle. After 10 days acclimatisation, the male and female rats (1 : 2) were kept for mating in mating cages at late evening. On the subsequent mornings, the females were checked for the presence of sperm, and the day on which the sperm was positively found was designated as day ‘0’ of gestation. Thus, a total of 36 pregnant rats were obtained and divided into 4 groups of nine rats each (n = 9). The first group of rats, which served as a control, were given 0.5 ml of distilled water (DW), whereas the rats of the remaining 3 groups, which served as treated groups, were administered cyclophosphamide (CP) at dose levels of 5, 15 and 30 mg per kg of body weight, respectively. The DW and CP were given to their respective groups on day 11 of gestation, through an i.p. route. Six pregnant rats from each control and CP treated group were followed for 24 h after the administration of the vehicle control or CP, and then their embryos were explanted using a per abdominal approach for further investigation. Later these embryos were deep-frozen using liquid nitrogen and stored for NMR studies at −80 °C. The remaining dams (n = 3) from each group were followed until term (i.e. up to day 20 of gestation) for the in life observations and their foetuses were then taken out using a per abdominal approach, for further investigations and to look for any anomalies or malformations. The observations of the dams and their pups were carried out mainly to ascertain the doses, to confirm whether a single exposure can induce teratogenic effects or not, when given on a particular day of gestation (i.e. day 11), and also to examine whether the effect provoked can last until term or not.

Observations

Observations of dams. Cage-side observations were made daily for each dam, to observe any signs of behavioural changes, reactions to the treatment or ill health, including changes in the skin, fur coat, any discharge from orifices or the general activity of the animal, if any, until the total period of their following was complete.

Body weight gain, and food and water intake were measured on gestational days 0, 6 and 12 for the dams followed until day 12 of gestation, whereas, for the dams followed until term these parameters were measured on days 0, 6, 11 and 20 of gestation, to get their periodical body weight gain and also to confirm their appetite. Before collecting the embryos or foetuses, the ovary and uterine contents of the dams were observed thoroughly for the number of corpora lutea, implants and resorptions.

Observations of foetuses. Foetuses obtained from the dams were weighed and their crown to rump lengths were measured, followed by examination for possible gross morphological anomalies. Later these foetuses were divided into two groups for study of the visceral anomalies and malformations using a serial slicing method,⁴ and the skeletal anomalies and malformations using the Dawson Alizarin Red Preparation.²⁶

Observations of embryos. Embryos explanted from dams of the control and CP treated groups were immediately placed in dry sterile Falcon Petri plates, and then their body weight and crown to rump length were measured. Later, they were placed under a stereo zoom microscope, where each and every embryo was observed for their heartbeat and yolk sac circulation to see whether they were alive or not, using the method of Warner et al.,²⁷ as well as examining them for the presence of any other gross morphological anomalies or malformations. Soon after, all embryos were flash frozen in liquid nitrogen after autopsy and stored at −80 °C until the NMR experiment was conducted.

Sample preparation and NMR experimental conditions

At the time of the NMR experiment, the frozen rat embryo was thawed at room temperature and placed into a 4 mm zirconia rotor with a 50 μl capacity. 20 μl of D₂O was added to the rotor with the embryo for locking of the spectrometer frequency. All NMR experiments were carried out on a Bruker Avance-II 400 MHz spectrometer, operating at a ¹H frequency of 400.13 MHz, equipped with a 5 mm HR-MAS ¹³C–¹H Z gradient probe with a magic-angle gradient, at a temperature of 300 K and a sample spinning rate of 4000 ± 1 Hz. Data acquisition and processing were done with Bruker Topspin 2.1.

One dimensional NMR analysis

¹H NMR spectra were acquired using the standard 1D (noesypr 1D) pulse program, a 2.0 s relaxation delay (d1), a 3.9 s acquisition time (AQ), 64K FID data points, a spectral width of 8012.82 Hz and a number of scans of 64, as the standard setting. For experiments using presaturation, the transmitter offset was manually set at 4.689 ppm in order to achieve optimal suppression of the residual water signal. The acquired FIDs were Fourier transformed to yield spectra with 64K data points. Manual phase correction and automatic polynomial baseline correction were always used. The signal of lactate was calibrated at 1.33 ppm for all standard 1D (noesy 1D) spectra. The pulse width P1 (15.60 μs) and power level PL9 (56.22 dB) were calculated using command pulsecal for each sample. The calculated P1 and PL9 were almost the same for each sample. 1D T2 relaxation edited ¹H NMR spectra were acquired using the CPMG pulse sequence [RD-90°-{τ-180°-τ}_n acquisition]²⁸ and simple presaturation of the water peak was used as a T2 filter to suppress broad signals arising from the macromolecules. All 1D spectra were acquired with 128 transients, 32k complex data points, a spectral width (SW) of 8012.82 Hz and a 3 s relaxation delay, with 14 minutes of acquisition time in each CPMG experiment.

Characterization of the metabolites in the HR-MAS ¹H NMR spectra

Characterization of the metabolites in the NMR spectra of the embryos was carried out on the basis of chemical shift, coupling constant and multiplicities of the signal, as reported in the literature,^29–31 by comparison with the standard spectra of each metabolite as reported in the Biological Magnetic Resonance Bank (BMRB, http://www.bmrb.wisc.edu) or Human Metabolome Data Base (HMDB, http://www.hmdb.ca).^32,33

Quantification of the metabolites in the Carr–Purcell–Meiboom–Gill (CPMG) spectra

The quantification of embryonic metabolites using an external reference will introduce substantial errors, as the spectra are T₁ and T₂ dependent and correction factors for both have to be incorporated for precise calculation. For this purpose, HR-MAS NMR spectra, using identical conditions, were obtained for each metabolite. The gravimetrically weighed metabolite (1.0 mg), namely, lactate, alanine, aspartate, aspargine, taurine, choline, glycine or creatine (Sigma Aldrich), was prepared as a mixture in a stock solution (D₂O, 1 mg mL⁻¹) and 50 μl of the stock solution was subjected to HR-MAS NMR measurements. A standard artificial signal was generated through using the software QUANTAS and all of the prior recorded spectra were added in. This approach compensates for the scaling according to the experimental conditions of all the pure metabolites, and the other recorded spectra obtained from all the embryos were then quantified accordingly.³⁴

Statistical analysis

Univariate analysis. The metabolite concentration and the outcome of pregnancy, such as number of corpora lutea, number of implantations, number of resorptions, number of live/still births, and foetal weight and length, were expressed as the mean values ± SD (standard deviation). The statistical significance was determined using a one-way ANOVA post hoc Bonferroni multiple comparison test (Graph pad prism 5.0, San Diego, USA) between all the four groups. A probability (P-value) of ≤0.05 was taken to indicate statistical significance.

Multivariate analysis. The noesypr 1D ¹H NMR spectral range of 0.69 ppm to 4.56 ppm was taken for binning the data to produce a series of sequentially integrated regions of 0.01 ppm in width, which gave 387 equal segments of a spectrum in the sum of intensities integration mode, and spectral intensities were scaled to the biggest bucket using Bruker AMIX software (version 3.5.5. Bruker Biospin, Germany). The resulting data binning matrices or bucket files having normalized integral values were exported into Microsoft Office Excel 2007. These were further imported to ‘The Unscrambler X’ software package (version 10.0.1, Camo USA, Norway) for multivariate principal component analysis (PCA) and partial least-squares discriminant analysis (PLS-DA).

Results

Results obtained from the dams

General observations and pregnancy outcomes. The results obtained from the dams of all the groups terminated on day 12 of gestation revealed a steady gain in body weight during the entire period of follow up i.e. from day 0 to 12 of the pregnancy. However, no significant dose dependent variations in body weight gain were observed among the CP treated dams (Fig. 1a). Food and water intake was found to be satisfactory among these dams. A good number of embryos were encountered from each treated group, as well as the control group, except for the high dose CP (30 mg kg⁻¹ i.p.) group. This group’s dams showed a high incidence (75%) of resorptions (post implantation loss), subsequently bringing down the number of embryos (n = 9) to that lesser than the other treated groups, as well as the control group (Table 3).

Fig. 1 (a) Graph showing body weight gain among dams of the control and cyclophosphamide treated groups that were followed until day 12 of gestation. (b) Body weight changes in the dams treated with a control or various doses of CP and followed until term (i.e. day 20 of gestation). All of the results are expressed as mean ± SEM.

The remaining 12 dams of the control and all the CP treated groups, followed until term, also showed an overall gain in their body weight, however, a dose dependent decrease in body weight relative to the control group was observed among all of the CP (5, 15 and 30 mg kg⁻¹ i.p.) treated dams (Fig. 1b). Though changes in food and water intake were observed among the control and CP treated dams, they were well within the normal range. The pregnancy outcomes of all treated group dams showed a comparable number of corpora lutea and implants with the control (Table 1). However, a drastic situation regarding the number of resorptions was noticed, which revealed 8.33, 19.23 and 52.17 percent resorptions (post implantation loss) among the low (CP, 5 mg kg⁻¹ i.p.), intermediate (CP, 15 mg kg⁻¹ i.p.) and high (CP, 30 mg kg⁻¹ i.p.) dose groups, respectively, and a significant (p < 0.05) increase in the number of resorptions from low to high dose CP groups. This dire condition has led to the reduction of the average live births among the dams of the high dose (CP; 30 mg kg⁻¹ i.p.) group (3.33), whereas among low dose (CP; 5 mg kg⁻¹ i.p.) and intermediate dose (CP; 15 mg kg⁻¹ i.p.) group dams the average numbers of live births were 8.7 and 7, respectively (Table 1).

Table 1 Pregnancy outcomes from control and various cyclophosphamide (CP) treated dams followed until term and observations of their foetuses^a

Parameters	Control	Treated groups
	Distilled water (0.5 ml i.p.)	CP-LD (5 mg kg^−1 i.p.)	CP-ID (15 mg kg^−1 i.p.)	CP-HD (30 mg kg^−1 i.p.)
a Values calculated based on litter. The comparison was made using one way ANOVA followed by a Bonferroni multiple comparison test. *p < 0.05 versus DW-control. **p < 0.01 versus DW-control. ^#p < 0.05 versus CP-HD. ^##p < 0.01 versus CP-HD. DW = distilled water; CP-LD = cyclophosphamide low dose; CP-ID = cyclophosphamide intermediate dose; CP-HD = cyclophosphamide high dose.
Total no. of animals studied	3	3	3	3
No. of corpora lutea	25	24	28	24
Mean ± SD (range)	8.33 ± 3.05 (5–11)	8 ± 2.0 (6–10)	9.33 ± 1.52 (8–11)	8 ± 1.0 (7–9)
No. of implantations	24	24	26	23
Mean ± SD (range)	8 ± 2.64 (5–10)	8 ± 2.0 (8–10)	8.66 ± 0.577 (8–9)	7.66 ± 0.577 (7–8)
No of resorptions	0	2	5	12
Mean ± SD (range)	0 ± 0 (0–0)	0.66 ± 0.577^# (0–1)	1.66 ± 0.577 (1–2)	4 ± 2.0 (2–6)
No. of live births	24	22	21	10
Mean ± SD (range)	8 ± 2.65 (5–10)	7.33 ± 1.53 (7–9)	7 ± 1.0 (6–8)	3.33 ± 2.51 (1–6)
Mean fetal weight (g)	3.30 ± 0.009	3.27 ± 0.039^##	3.17 ± 0.004*	3.13 ± 0.060**
Mean fetal length (cm) (crown to rump length)	3.12 ± 0.040^##	3.04 ± 0.062^##	2.90 ± 0.051	2.70 ± 0.135
Post implantation loss (%)	0	8.33	19.23	52.17
Anomalies
(a) External (%)	0	2 (9.09)	4 (19.05)	4 (40.00)
(b) Visceral (%)	0	0	2 (20.00)	2 (40.00)
(c) Skeletal (%)	0	1 (9.09)	2 (18.18)	2 (40.00)
(d) Total (%)	0	3 (13.64)	8 (38.10)	8 (80.00)

---

Observations among foetuses. Significant (p < 0.01) decreases in mean foetal weight and crown to rump length were observed in the high dose CP (30 mg kg⁻¹ i.p.) group vis-a-vis the control (DW, 0.5 ml i.p.) group (Table 1). Whereas, on analyzing the fetal anomalies and malformations, it also appeared that external, visceral and skeletal anomalies were observed in almost all of the treated groups, yet the incidences and variety of anomalies were dose dependent and at a maximum (80%) in the high dose CP (30 mg kg⁻¹ i.p.) group (Fig. 2a). The anomalies and malformations included cases of curling of the tail, scoliosis with a totally deformed body appearance, cleft lip, vaulted cranium, wrist drop, exencephaly, cleft palate, hydropic cyst in right kidney, increased intercostal space and poorly ossified cranium (Table 2). One observation noticeable with the increase of the dose, was that the variety and intensity of the anomalies were also increased and that their effects can last until term.

Fig. 2 Histograms showing the total and percentage anomalies and malformations found among: (a) foetuses from dams treated with distilled water (control) or various doses of CP, and followed until term (i.e. day 20 of gestation); (b) embryos from dams treated with distilled water (control) or various doses of CP, and followed until day 12 of gestation.

Table 2 A breakdown of the anomalies and malformations found among the offspring of dams from various groups treated with a vehicle control or various doses of cyclophosphamide^a

Groups (doses)	Number of young
	With external anomalies				With anatomical anomalies
	Examined	Affected		Description	Visceral anomalies				Skeletal anomalies
					Examined	Affected		Description	Examined	Affected		Description
		No.	%			No.	%			No.	%
a Number of dams per group = 3; route = intraperitoneal (i.p.); DW = distilled water; CP-LD = cyclophosphamide low dose; CP-ID = cyclophosphamide intermediate dose; CP-HD = cyclophosphamide high dose.
Control (0.5 ml DW)	24	0	0	—	12	0	0	—	12	0	0	—
CP-LD (5 mg kg^−1)	22	2	9.09	Wrist drop (1)	11	0	0	—	11	1	9.09	Increased intercostal space (1)
				Curling of tail (1)
CP-ID (15 mg kg^−1)	21	4	19.05	Cleft lip (1)	10	2	20	Cleft palate (1)	11	2	18.18	Poorly ossified cranium (1)
				Vaulted Cranium (1)
				Curling of tail (1)				Hydropic cyst in right kidney (1)				Increased intercostal space (1)
				Wrist drop (1)
CP-HD (30 mg kg^−1)	10	4	40.00	Curling of tail (1)	5	2	40	Cleft palate (2)	5	2	40	Scoliosis with curling of tail (1)
				Scoliosis with totally deformed body appearance and curling of tail (1)
				exencephaly (1)								Increased intercostal space (1)
				Cranial heamatoma with wrist drop (1)

---

Results from embryos. The embryos from all of the groups (Fig. 3a and b) showed normal heartbeats and yolk sac circulation. Their body weight, crown to rump length and somite number were well within the normal ranges (Fig. 3c–e). The CP treated groups revealed in total eight cases of anomalies or malformations (Table 3), out of which one case each of stunted growth and curling of tail were found among the low dose (CP, 5 mg kg⁻¹ i.p.) group, revealing 7.4% total anomalies and malformations, and in the intermediate dose (CP, 15 mg kg⁻¹ i.p.) group, a total of 10.34% anomalies was observed with two cases of stunted growth and one case of neural tube defect (NTD). In the high dose (CP, 30 mg kg⁻¹ i.p.) group, one embryo possessed craniofacial defects (CFD) and the second case was of NTD, whereas the 3^rd case presented neural tube defects along with CFD, thus presenting in total three embryos (33.34%) with malformations (Fig. 2b). Here two things were noticed – one was that the number and variety of malformations and anomalies were increased with the augmentation of dose, and the other was the incidence of multiple malformations due to a high dose (CP; 30 mg kg⁻¹ i.p.). Among the control group, one embryo, out of 37, showed poor growth (ESI Fig. 2†).

Fig. 3 Histograms showing the results of various observations of the embryos from dams treated with distilled water (control) or various doses of CP, and followed until day 12 of gestation – (a) heartbeat score plot, (b) yolk sac circulation score plot, (c) average embryonic weight, (d) average embryonic length (crown to rump) in cm, and (e) number of somites.

Table 3 Anomalies and malformations observed among the embryos of dams from various groups treated with a vehicle control or various doses of cyclophosphamide (CP)^a

Groups (doses)	No. of		Anomalies & malformation types	Total number of anomalies & malformations	Total percentage of anomalies & malformations (%)
	Embryos studied	Mortalities/resorptions (%)
a Number of dams per group = 6; route = intraperitoneal (i.p.); DW = distilled water; CP-LD = cyclophosphamide low dose; CP-ID = cyclophosphamide intermediate dose; CP-HD = cyclophosphamide high dose; B.wt. = bodyweight.
DW-control (0.5 ml/100 g B.wt.)	37	0/0 (0.0)	Stunted growth (1)	1	2.70
CP-LD (5 mg kg^−1 B.wt.)	27	1/6 (18.18)	Stunted growth (1)	2	7.40
			Curling of tail (1)
CP-ID (15 mg kg^−1 B.wt.)	29	2/9 (23.68)	Stunted growth (2)	3	10.34
			Neural tube defect (1)
CP-HD (30 mg kg^−1 B.wt)	9	2/27 (75.0)	Neural tube + Craniofacial defect (1)	3	33.34
			Neural tube defect (1)
			Craniofacial defect (1)

---

Metabolic profiles of the embryos using HR-MAS ¹H NMR analysis

Standard 1D ¹H NMR (noesypr 1D) and CPMG ¹H NMR spectra of the embryos allowed identification and assignment of 19 endogenous metabolites (Fig. 4a) viz. acetate, alanine, aspargine, aspartate, choline, creatine, fatty acids (triacylglycerols (TAGs)), glycerophosphocholine (GPC), glutamine, glutamate, glycine, 3-hydroxyglutaric acid, isoleucine, lactate, leucine, lysine, phosphocholine, taurine and valine. Representative ¹H NMR spectra of DW-control, CP-LD (low dose CP; 5 mg kg⁻¹ i.p.), CP-ID (intermediate dose CP; 15 mg kg⁻¹ i.p.) and CP-HD (high dose CP; 30 mg kg⁻¹ i.p.) treated embryos show the variability in the metabolites among the different groups of samples (Fig. 4b).

Fig. 4 HR-MAS ¹H NMR spectra of the embryos, showing various assignments of the metabolites. (a) Complete assignment of the metabolites present within the 0.8–4.2 ppm region of the CP-HD treated rat embryo noesypr 1D HR-MAS ¹H NMR spectrum; (b) representative ¹H NMR spectra for the DW-control, CP-LD, CP-ID and CP-HD embryos, showing assignment of the metabolites present in the 0.6–4.4 ppm region. The metabolites, which showed major differences, were marked on the spectra. Abbreviations: Val, valine; Leu, leucine; Ile, isoleucine; Cho, choline; PCho, phosphocholine; GPC, glycerophosphocholine; and TCC, total choline content.

Principal component analysis (PCA) of the NMR spectra

The results were presented as a 2D principal component scores plot (Fig. 5a; each point represents an individual sample) and loadings plots, in which the metabolite signals are shown as positive loadings or negative loadings to indicate differential changes of the metabolites among groups of samples (Fig. 5b and c).

Fig. 5 (a) Two dimensional scattered principal component analysis (PCA) score plot; (b) PC-1 loading and (c) PC-2 loading plots for typical HR-MAS ¹H NMR spectra of rat embryos, showing the differences in the metabolic profiles of embryos from dams exposed to the DW-control or to various doses of CP (5, 15 and 30 mg kg⁻¹ i.p.).

Multivariate statistical principal component analysis (PCA) indicated that the explained variance, represented by PC1 and PC2, was 61% among the dataset. The PC1 loading plot contributed 45% to the explained variance. The positive loadings in PC1 were because of major contributions from fatty acids, alanine, glutamine, glycerophosphoryl choline (GPC), phosphorylcholine (PC), aspartate, taurine and glycine, whereas, the minor contributors were asparagine and creatine. Negative loadings were contributed by lactate, choline, lycine and 3-hydroxyglutaric acid (Fig. 5b). The PC2 loading plot contributed only 16% of the explained variance. The positive loadings in PC2 were because of major contributions from the total choline content (TCC) and creatine, and minor contributions from taurine, lysine and lactate. While negative loadings were because of the fatty acids, valine, leucine, isoleucine, aspartate, asparagine and glycine (Fig. 5c).

Partial least-squares discriminant analysis (PLS-DA) of the NMR spectra

PLS-DA facilitated characterization and better cluster separation of the metabolites, distinguishing the DW-control, the intermixed group of CP-LD and CP-ID, and the CP-HD group from each other. The cross validated 2D score plot showed statistically significant differences between the samples of the four groups (Fig. 6). A full cross validation was applied to describe the quality of the mathematical model using the observed and cross-validation parameters R² and Q², respectively. The values of R² = 0.772 and Q² = 0.718 from the PLS-DA model indicate that the model is quite robust, with little overfitting and a Q² of 72% validation ability.

Fig. 6 Comparison of the DW-control, CP-LD, CP-ID and CP-HD embryos using 2D partial least-squares discriminant analysis (PLS-DA) validated score plots showing R² and Q² values.

Quantitative analysis of the CPMG spectra

On quantitative analysis of the CPMG spectra of various groups’ embryos, a significant (p < 0.05) dose dependent decrease in the level of total choline content and increase in the aspartate level were observed for the CP treated embryos in comparison to the control. Embryonic levels of glycine and creatine were also found to be reduced significantly (p < 0.05) in the CP-HD group dams vis-a-vis the CP-ID group (Table 4).

Table 4 Qualitative and quantitative variability in the metabolites of embryos from the control and various CP treated rats, using CPMG ¹H NMR spectra along with their chemical shifts (ppm) for quantification^a

Metabolites	Chemical shift (ppm)	DW-control (DW, 0.5 ml i.p.)	CP-LD (5 mg kg^−1 i.p.)	CP-ID (15 mg kg^−1 i.p.)	CP-HD (30 mg kg^−1 i.p.)
a Values expressed as mean values ± SD in mmol L^−1 from a 50 μl sample volume of rat embryo. The comparison was made using one way ANOVA followed by a Bonferroni multiple comparison test. a, b, c, d, e and f denote a statistical significance of P ≤ 0.05, i.e. a = DW-control vs. CP-LD; b = DW-control vs. CP-ID; c = DW-control vs. CP-HD; d = CP-LD vs. CP-ID; e = CP-LD vs. CP-HD; f = CP-ID vs. CP-HD. DW = distilled water; CP-LD = cyclophosphamide low dose; CP-ID = cyclophosphamide intermediate dose; CP-HD = cyclophosphamide high dose; TCC = total choline content.
Alanine	1.48 (β-CH3)	10.1 ± 6.4	13.4 ± 9.3	14.8 ± 10.3	5.7 ± 3.1
Asparagine	2.80 (β-CH)	4.7 ± 2.8	7.7 ± 5.2	6.7 ± 3.8	6.1 ± 3.1
Aspartate	2.65 (β′-CH)	1.2 ± 1.9^b,c	3.6 ± 1.9^d,e	16.2 ± 12.2^f	31.0 ± 13.3
Creatine	3.94 (N-CH2)	6.1 ± 4.2	8.4 ± 6.0	9.3 ± 5.6^f	3.2 ± 1.5
Glycine	3.56 (α-CH2)	6.8 ± 4.5	8.8 ± 5.9	9.7 ± 6.0^f	3.2 ± 1.2
3-Hydroxy glutaric acid	2.34 (CH)	42.3 ± 24.0	53.4 ± 35.2	59.4 ± 36.0	22.7 ± 10.1
Lactate	1.33 (β-CH3)	55.3 ± 29.8	64.6 ± 42.3	69.1 ± 47.9	24.72 ± 14.5
Taurine	3.41 (N-CH2)	16.5 ± 11.7	22.0 ± 17.1	26.3 ± 16.7	14.2 ± 4.9
TCC	3.21–3.24 (N(CH3)3)	8.17 ± 2.6^a,b,c	3.88 ± 2.9	3.8 ± 2.3	1.3 ± 0.4

---

Discussion

This study has demonstrated a method for testing the teratogenicity of a new chemical entity (NCE) on the platform of metabolomics which can bypass the inherent shortcoming and limitations of in vitro systems, and will involve less time, cost and laboratory animals. Here, rats have been used as the testing species, cyclophosphamide (CP) as a ‘teratogen’, and embryonal metabolites as the end point.

The reason for selecting rats as the testing species is due to the fact that all of the regulatory agencies recommend ‘rodents’ and ‘lagomorphs’ as animal species for testing teratogenicity, and among rodents, the rat is the highly preferred species. Other advantages associated with rats are their easy availability, large litter sizes, ease of handling, high fertility rate, short gestation period, low spontaneous malformation rate, genetic stability and lack of seasonal breeding.³⁵

Cyclophosphamide is one of the most commonly used anti-neoplastic drugs of the alkylating compound group.³⁶ It is also teratogenic in various laboratory animal species,^25,37,38 as well as in humans,³⁹ and the teratogenicity of which can be characterized by stunted growth, and central nervous system, facial, and skeletal anomalies. Although it is a proteratogen, it provides an opportunity for use as a model teratogen for understanding teratogenicity and the associated mechanisms.

Organisms often respond in complex and unpredictable ways to stimuli that cause disease or injury. Various bio-components, such as creatine, lactate, glutamate, glutamine, alanine, aspartate, aspargine, taurine, fatty acid and choline, have been reported to have pivotal roles during embryonal growth and development,^40–50 and variation in any of them may cause an ailment in the developing embryos and foetuses. Keeping this fact in mind, and the view of Nicholson and Lindon,¹⁷ by measuring and mathematically modelling changes in the levels of metabolic products found in biological fluids and tissues, metabolomics offers a fresh insight into the effects of diet, drugs and disease. The design of the present work was based on studying changes in the levels of metabolites among embryos due to chemical insult in utero in the form of a teratogen, through obtaining metabolic peaks using ¹H NMR spectroscopy. For the embryos explanted from the dams of all the four test groups after 24 hours of vehicle control or CP treatment, we could perform further investigations regarding the qualitative and quantitative changes in their metabolites due to CP exposures with the help of HR-MAS ¹H NMR.

In the present study, multivariate unsupervised PCA was used to separate the HR-MAS ¹H NMR spectra of the embryos from various treated groups based on the changes in their metabolic profiles, using score and loading plots. A 2D principal component scores plot showed that most of the major variations were among the spectra of the control and various CP treated group embryos, and observation of these variations could be achieved from unsupervised data treatment. This was explored to find relevant features which showed the differences between different sets of embryos. A dose dependent cluster separation observed (Fig. 5a) for the DW-control, CP-LD, CP-ID and CP-HD group embryos, revealed the variations among their spectra. In addition, better cluster separation between the DW-control and CP-HD embryo spectra along PC1 showed significant differences in the metabolic profiles of the two. However, only a little variation was seen in the pattern of metabolites between samples of the CP-LD and CP-ID groups, as intermixing was remarked among their spectra. The cross validation of the unsupervised 2D principal component scores plot was achieved using supervised PLS-DA, typically explaining better separation as the value of cross validation parameter Q² was lower than R².

The score plots along with their loading plots (Fig. 5b and c), revealed raised levels of glutamine, fatty acids and aspartate in the CP treated groups, whilst on the other hand the levels of lysine, glycine, choline and lactate were found to be decreased in embryos from dams treated with CP, when compared with the control. The 2D loading plots also showed that changes in the levels of metabolites, such as aspargine, creatine and 3-hydroxybutyric acid, were observed among the various treatment groups, yet these metabolites were not the major confounders.

From results of the quantitative analysis of various metabolites (Table 2), it appears that these NMR findings are in concordance with the reports of previous researchers, and that the incidences of anomalies and malformations found among the embryos and foetuses of the present study are due to in utero exposure to different doses of CP (Tables 3 and 4). The present study revealed that anomalies and malformations in the embryos (ESI Fig. 2†) and foetuses (ESI Fig. 1†) due to CP treatment were mainly in the form of neural tube defects, craniofacial defects, exencephaly, cleft palate, cleft lip and stunted growth. Fisher et al.,⁵¹ reported that choline is essential for normal neural tube closure in early pregnancy. Reduced choline content intercepts the methyl group metabolism resulting in hypomethylation, which is reported to influence neural tube closure⁴⁴ resulting in neural tube defects (NTD) or craniofacial defects (CFD). A significant dose dependent down regulated choline content in the various CP treated embryos has been noticed, which may be the reason for the occurrence of NTD or CFD in the present study. A similar relationship was observed between choline content and cleft palate formation, also. A reduced choline content results in hypomethylation of DNA.⁵² DNA methylation occurs at cytosine bases followed by guanine (CpG sites),⁵³ and influences many cellular events⁵⁴ involved in the normal formation of lip, palate and alveolus. In this context, reductions in choline content may also be a reason for the occurrence of cleft palate among the CP treated group foetuses. Niculescu et al.⁵⁵ have reported that choline deficiency inhibits cell proliferation, which may result in neural tube defects, craniofacial defects and stunted growth among embryos. In our study, the intensity of choline has been found to be significantly reduced on the one hand, and on the other hand cases of NTD, craniofacial defects and stunted growth were observed. This can be linked to the reports of previous authors who studied the critical role of choline on normal embryonic development.

In the context of the NTD and exencephaly, the second possible cause is an increased signal intensity of aspartate. Chen et al.⁵⁶ have reported that aspartate stimulates N-methyl-D-aspartate (NMDA) receptors responsible for the excitation of neurons, which may bring about excitotoxicity. Moreover, since the brain enzymes of neonates are not developed enough to combat this excitotoxicity, this might have resulted in neuronal cell death and apoptosis,⁴⁵ which may have caused NTD in embryos and exencephaly among foetuses exposed to various doses of CP. As creatine deficiency leads to neurological symptoms in early infancy and severe neurodevelopmental delay,^42,48 the reduced signal intensity of creatine in the embryos from high dose CP treated dams may also be a contributing factor in bringing about NTD, exencephaly and other neurological problems. Another metabolite reduced in the embryos of the CP treated dams was glycine. As supplementation of glycine enhances embryonic development,⁴¹ its reduced level of signal intensity among the embryos of the CP-high dose treated pregnant rats may also be the reason for the developmental issues in the embryos and foetuses from dams of the same group.

In the present study, embryos which were found with malformations/anomalies of mild to moderate condition have presented these above metabolites in altered conditions, when their spectra obtained from NMR were analysed. Thus, it appears that the alterations in metabolites, as shown by NMR spectral analysis, and the teratogenic incidences found among the embryos and foetuses are in concordance to each other, also validating the findings. Finally, it can be surmised that alteration of various metabolites from their normal profile in an embryo after intrauterine exposure to cyclophosphamide is the key finding in this work. Hence, alteration of the metabolic fingerprint of an organism may be the end point for defining teratogenicity.

This will also give a new dimension to the study of birth defects, helping in the prioritization lists having a large number of chemicals to be tested for their teratogenicity potential, subsequently providing quality and safe commodities for human consumption and hence ensuring a better life style. Due to providing quick and correct information, this method may also have a good translational value as unnecessary abortions due to misinformation about the teratogenicity of drugs may be avoided.

Conflict of interest

There are no conflicts of interest in connection with the submitted article.

Abbreviations

CPMG	Carr–Purcell–Meiboom–Gill
HR-MAS	High resolution-magic angle spinning
PLS-DA	Partial least-squares discriminant analysis
PCA	Principal component analysis
DW-control	Distilled water control
CP	Cyclosphosphamide
CP-HD	Cyclophosphamide high dose
CP-ID	Cyclophosphamide intermediate dose
CP-LD	Cyclophosphamide low dose
TCC	Total choline content
NCE	New chemical entity
NTD	Neural tube defects
CFD	Craniofacial defects

Acknowledgements

We are thankful to Dr Raja Roy, Professor, Centre of Biomedical Research, SGPGIMS Campus, Raibarely Road, Lucknow, India for extending his valuable guidance and advice. We acknowledge the Indian Council of Medical Research (ICMR), New Delhi, India, for providing financial support.

References

1. U. Deppe, E. Schierenberg, T. Cole, C. Krieg, D. Schmitt, B. Yoder and G. Von Ehrenstein, Cell lineages of the embryo of the nematode Caenorhabditis elegans, Proc. Natl. Acad. Sci. U. S. A., 1978, 75, 376–380.
2. E. Johnson, A subvertebrate system for rapid determination of potential teratogenic hazards, J. Environ. Pathol. Toxicol., 1980, 4, 153–156.
3. W. J. Birge, J. A. Black, A. G. Westerman and B. A. Ramey, Fish and amphibian embryos—a model system for evaluating teratogenicity, Toxicol. Sci., 1983, 3, 237–242.
4. J. Wilson and F. Fraser, Handbook of Teratology, Plenum, NY, 1977, vol. 4, pp. 1–4.
5. J. Daniel Jr, Methods in mammalian embryology, 1971.
6. D. New and K. Stein, Cultivation of post-implantation mouse and rat embryos on plasma clots, J. Embryol. Exp. Morphol., 1964, 12, 101–111.
7. G. Morriss and C. Steele, Comparison of the effects of retinol and retinoic acid on postimplantation rat embryos in vitro, Teratology, 1977, 15, 109–119.
8. D. New, Development of explanted rat embryos in circulating medium, J. Embryol. Exp. Morphol., 1967, 17, 513–525.
9. D. New, Studies on mammalian fetuses in vitro during the period of organogenesis, in The mammalian fetus in vitro, Springer, 1973, pp. 15–65.
10. E. L. Wee, L. G. Wolfson and E. F. Zimmerman, Palate shelf movement in mouse embryo culture: evidence for skeletal and smooth muscle contractility, Dev. Biol., 1976, 48, 91–103.
11. O. Trowell, A modified technique for organ culture in vitro, Exp. Cell Res., 1954, 6, 246–248.
12. D. Neubert and H.-J. Barrach, Organotropic effects and dose-response relationships in teratology, Methods in prenatal toxicology, Thieme, Stuttgart, 1977, pp. 405–412.
13. G. Daston, D. Baines and J. E. Yonker, Chick embryo neural retinal cell culture as a screen for developmental toxicity, Toxicol. Appl. Pharmacol., 1991, 109, 352–366.
14. O. P. Flint, In vitro tests for teratogens: desirable endpoints, test batteries and current status of the micromass teratogen test, Reprod. Toxicol., 1993, 7, 103–111.
15. E. Genschow, H. Spielmann, G. Scholz, A. Seiler, N. Brown, A. Piersma, M. Brady, N. Clemann, H. Huuskonen and F. Paillard, The ECVAM international validation study on in vitro embryotoxicity tests: results of the definitive phase and evaluation of prediction models, Altern. Lab. Anim., 2002, 30, 151–176.
16. J. K. Nicholson, J. Connelly, J. C. Lindon and E. Holmes, Metabonomics: a platform for studying drug toxicity and gene function, Nat. Rev. Drug Discovery, 2002, 1, 153–161.
17. J. K. Nicholson and J. C. Lindon, Systems biology: metabonomics, Nature, 2008, 455, 1054–1056.
18. M. Bollard, E. Holmes, J. Lindon, S. Mitchell, D. Branstetter, W. Zhang and J. Nicholson, Investigations into Biochemical Changes Due to Diurnal Variation and Estrus Cycle in Female Rats Using High-Resolution ¹H NMR Spectroscopy of Urine and Pattern Recognition, Anal. Biochem., 2001, 295, 194–202.
19. E. Holmes, A. W. Nicholls, J. C. Lindon, S. C. Connor, J. C. Connelly, J. N. Haselden, S. J. Damment, M. Spraul, P. Neidig and J. K. Nicholson, Chemometric models for toxicity classification based on NMR spectra of biofluids, Chem. Res. Toxicol., 2000, 13, 471–478.
20. J. C. Lindon, J. K. Nicholson, E. Holmes and J. R. Everett, Metabonomics: metabolic processes studied by NMR spectroscopy of biofluids, Concepts Magn. Reson., 2000, 12, 289–320.
21. J. K. Nicholson, J. C. Lindon and E. Holmes, ‘Metabonomics’: understanding the metabolic responses of living systems to pathophysiological stimuli via multivariate statistical analysis of biological NMR spectroscopic data, Xenobiotica, 1999, 29, 1181–1189.
22. S. Howells, R. Maxwell, A. Peet and J. Griffiths, An investigation of tumor 1H nuclear magnetic resonance spectra by the application of chemometric techniques, Magn. Reson. Med., 1992, 28, 214–236.
23. O. Beckonert, M. Coen, H. C. Keun, Y. Wang, T. M. Ebbels, E. Holmes, J. C. Lindon and J. K. Nicholson, High-resolution magic-angle-spinning NMR spectroscopy for metabolic profiling of intact tissues, Nat. Protoc., 2010, 5, 1019–1032.
24. I. Lowe, Free induction decays of rotating solids, Phys. Rev. Lett., 1959, 2, 285.
25. P. E. Mirkes, Cyclophosphamide teratogenesis: a review, Teratog., Carcinog., Mutagen., 1985, 5, 75–88.
26. A. B. Dawson, A note on the staining of the skeleton of cleared specimens with Alizarin Red S, Biotech. Histochem., 1926, 1, 123–124.
27. C. Warner, T. Salder, J. Shockey and M. Kate Smith, A comparison of the in vivo and in vitro response of mammalian embryos to a teratogenic insult, Toxicology, 1983, 28, 271–282.
28. S. Meiboom and D. Gill, Modified spin-echo method for measuring nuclear relaxation times, Rev. Sci. Instrum., 1958, 29, 688–691.
29. B. Sitter, S. Lundgren, T. F. Bathen, J. Halgunset, H. E. Fjosne and I. S. Gribbestad, Comparison of HR MAS MR spectroscopic profiles of breast cancer tissue with clinical parameters, NMR Biomed., 2006, 19, 30–40.
30. D. Olmo, Metabolic profile of chronic liver disease by NMR spectroscopy of human biopsies, Int. J. Mol. Med., 2011, 27, 111–117.
31. S. K. Bharti, A. Behari, V. K. Kapoor, N. Kumari, N. Krishnani and R. Roy, Magic angle spinning NMR spectroscopic metabolic profiling of gall bladder tissues for differentiating malignant from benign disease, Metabolomics, 2013, 9, 101–118.
32. J. L. Markley, M. E. Anderson, Q. Cui, H. Eghbalnia, I. A. Lewis, A. D. Hegeman, J. Li, C. F. Schulte, M. R. Sussman and W. M. Westler, New bioinformatics resources for metabolomics, in Pacific Symposium on Biocomputing, 2007, pp. 157–168.
33. D. S. Wishart, C. Knox, A. C. Guo, R. Eisner, N. Young, B. Gautam, D. D. Hau, N. Psychogios, E. Dong and S. Bouatra, HMDB: a knowledgebase for the human metabolome, Nucleic Acids Res., 2009, 37, D603–D610.
34. S. K. Bharti and R. Roy, Quantitative ¹H NMR spectroscopy, TrAC, Trends Anal. Chem., 2012, 35, 5–26.
35. J. M. Manson and Y. Kang, Test methods for assessing female reproductive and developmental toxicology, Princ. Methods Toxicol., 1994, 27, 989–1034.
36. A. D. Huitema, K. D. Smits, R. A. Mathôt, J. H. Schellens, S. Rodenhuis and J. H. Beijnen, The clinical pharmacology of alkylating agents in high-dose chemotherapy, Anti-Cancer Drugs, 2000, 11, 515–533.
37. S. H. Gilani and M. Chatzinoff, Embryopathic effects of cyclophosphamide, Environ. Res., 1983, 31, 296–301.
38. R. Padmanabhan and S. Singh, Axial Skeletal Malformations Associated with Cranioschisis Aperta and Exencephaly: The Result of Experimental Intervention after the Neural Tube Closure in Rats, Acta Orthop., 1983, 54, 104–112.
39. K. K. Vaux, N. C. Kahole and K. L. Jones, Cyclophosphamide, methotrexate, and cytarabine embropathy: Is apoptosis the common pathway?, Birth Defects Res., Part A, 2003, 67, 403–408.
40. J. Brown and D. Whittingham, The roles of pyruvate, lactate and glucose during preimplantation development of embryos from F1 hybrid mice in vitro, Development, 1991, 112, 99–105.
41. K. Moore and K. R. Bondioli, Glycine and alanine supplementation of culture medium enhances development of in vitro matured and fertilized cattle embryos, Biol. Reprod., 1993, 48, 833–840.
42. L. L. Sandell, X.-J. Guan, R. Ingram and S. M. Tilghman, Gatm, a creatine synthesis enzyme, is imprinted in mouse placenta, Proc. Natl. Acad. Sci. U. S. A., 2003, 100, 4622–4627.
43. M. Lane and D. K. Gardner, Removal of embryo-toxic ammonium from the culture medium by in situ enzymatic conversion to glutamate, J. Exp. Zool., 1995, 271, 356–363.
44. S. H. Zeisel, Choline: critical role during fetal development and dietary requirements in adults, Annu. Rev. Nutr., 2006, 26, 229.
45. T. S. Lee, Aspartame-a bitter sweet sweetener, NaturoDoc, http://www.naturodoc.com/library/nutrition/aspartame.htm, 2012.
46. A. Van den Pol, K. Obrietan, V. Cao and P. Trombley, Embryonic hypothalamic expression of functional glutamate receptors, Neuroscience, 1995, 67, 419–439.
47. J. A. Sturman, Taurine in development, J. Nutr., 1988, 118, 1169–1176.
48. O. Braissant, H. Henry, A.-M. Villard, O. Speer, T. Wallimann and C. Bachmann, Creatine synthesis and transport during rat embryogenesis: spatiotemporal expression of AGAT, GAMT and CT1, BMC Dev. Biol., 2005, 5, 9.
49. Y. Rezk, C. Huff and B. Rizk, Effect of glutamine on preimplantation mouse embryo development in vitro, Am. J. Obstet. Gynecol., 2004, 190, 1450–1454.
50. S. D. Liberles and S. L. Schreiber, Apoptosis-inducing natural products found in utero during murine pregnancy, Chem. Biol., 2000, 7, 365–372.
51. M. C. Fisher, S. H. Zeisel, M.-H. Mar and T. W. Sadler, Perturbations in choline metabolism cause neural tube defects in mouse embryos in vitro, FASEB J., 2002, 16, 619–621.
52. S. H. Zeisel, Importance of methyl donors during reproduction, Am. J. Clin. Nutr., 2009, 89, 673S–677S.
53. R. Holliday and G. Grigg, DNA methylation and mutation, Mutat. Res., Fundam. Mol. Mech. Mutagen., 1993, 285, 61–67.
54. A. Jeltsch, Beyond Watson and Crick: DNA methylation and molecular enzymology of DNA methyltransferases, ChemBioChem, 2002, 3, 274–293.
55. M. D. Niculescu, C. N. Craciunescu and S. H. Zeisel, Dietary choline deficiency alters global and gene-specific DNA methylation in the developing hippocampus of mouse fetal brains, FASEB J., 2006, 20, 43–49.
56. P. E. Chen, M. T. Geballe, P. J. Stansfeld, A. R. Johnston, H. Yuan, A. L. Jacob, J. P. Snyder, S. F. Traynelis and D. J. Wyllie, Structural features of the glutamate binding site in recombinant NR1/NR2A N-methyl-D-aspartate receptors determined by site-directed mutagenesis and molecular modeling, Mol. Pharmacol., 2005, 67, 1470–1484.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c5ra00671f

---