INTRODUCTION

An imbalance in excitatory/inhibitory neurotransmission has been implicated in the pathogenesis of autism. We tested this hypothesis by measuring changes in concentration of glutamate (Glu), glutamine (Gln) and GABA in the brain. Concentrations of tested compounds were evaluated in two rat models of pharmacologically-induced autism using Magnetic Resonance Spectroscopy (MRS), Nuclear Magnetic Resonance (NMR) and High Performance Liquid Chromatography (HPLC).

MATERIALS AND METHODS

On the 11th day of gestation, female rats were orally administrated valproic acid (VPA, n=11) or thalidomide (Thal, n=13) to induce autism in their offspring. There were 10 control subjects used in the test. 30-day old pups underwent MRS studies using the 7 T system. Spectra were acquired using short echo time PRESS sequence (TR/TE=2500/20 ms, 512 averages, 2048 points, scan time=17 min.) with VAPOR water suppression, the outer volume suppression, frequency drift correction (flip angle 5°) and eddy current correction. Metabolite concentrations were estimated using the LC-Model software. The amino acids from rat hippocampi were extracted for NMR studies using the HCl-Bligh and Dyer procedure. Three-trimethylsilyl propionic acid was used as an internal reference. NMR spectra were acquired at 25ºC using 500 MHz spectrometer. For chromatographic measurements. HPLC with fluorescence detection after derivatization in a timed reaction with o-phtalaldehyde plus mercaptoethanol was used. Derivatized samples were injected into a 250x4.6mm 5μm column with a mobile phase of 50mM phosphate buffer (KH₂PO₄/K₂HPO₄) containing 10% v/v methanol at pH 6.2 and methanol.Concentrations of Glu, Gln, GABA and their ratios with respect to Glu were used for statistical analysis. Univariate statistical analysis of the ratios was performed using General Linear Model (GLM) test. P value below 0.05 was considered as statistically significant. Orthogonal Partial Least Squares Discriminant Analysis (OPLS-DA) was used to build models for differentiation between the two animal models of autism and control groups for all methods.

RESULTS

Ratios of concentrations of Glu, Gln and GABA were analyzed only for the control group. Hotelling test showed significant differences for Gln/Glu and GABA/Glu ratios between all methods: MRS, NMR and HPLC (Fig.1). OPLS-DA analysis of amino acid concentrations per 1mg of protein for the three groups, the control and two animal models of autism, allowing to build models only for NMR and HPLC data. For the NMR data, models consisted of one predictive and one orthogonal component for both VPA and Thal vs. the control groups (Fig.2). For model of VPA vs. control group R2cum was at 95% and Q2cum was at 57%, while for Thal vs. control group, 80% and 77% respectively. For VPA vs. control group, 87.5% samples were classified correctly (p=0.004), while for Thal vs. control 100% (p<0.0001). Both models were validated using CV-ANOVA, p=0.04 and p=0.0001, respectively for VPA and Thal groups vs. control. For HPLC studies, the model consisted of one predictive and one orthogonal component for both VPA and Thal vs. control groups (Fig.3). For the model of VPA vs. control group, R2cum was at 96% and Q2cum was at 58%, while for Thal vs. control, 92% and 75% respectively. For VPA vs. control group 88.9% samples were classified correctly (p=0.001), while for Thal vs. control 100% (p<0.0001). Both models were validated using CV-ANOVA, p=0.02 and p=0.0001, respectively for VPA and Thal groups vs. control.

DISCUSSION

The differences between two in vitro methods could have resulted from measuring errors, which were assumed to be at 10% and 5% for NMR and HPLC, respectively. NMR measured concentration of free forms of amino acids, while HPLC measured their derivatives. The size of voxel used for MRS was large enough to measure the spectra. The measuring error could be 20-25% because spectra also included signals from CSF and blood. We expected similar results for the ratios of amino acids in all methods but due to measuring errors and limitation of each method they were not.

CONCLUSION

In vitro and in vivo methods could not be compared directly because of different measurement units and because we used post mortem extraction procedure of the tissue for the in vitro studies and live ones for the in vivo. OPLS-DA analysis of amino acid concentrations allowed to build a model to differentiate control and diseased groups in the in vitro studies.

Acknowledgements

This study was supported by the Polish National Science Centre, grant no. 2014/15/B/NZ4/04490

References

No reference found.