All animal study protocols were approved by the Animal Ethics Committee of the University following the Animal Research: Reporting of In Vivo Experiments (ARRIVE) guidelines as well as in accordance with the U.K. Animals (Scientific Procedures) Act of 1986 and its associated guidelines.

Sprague Dawley rats (n = 63; male, 200–300 g) were purchased from Shantou University Animal Center and divided into four groups. Group A (n = 12) was used for animal modeling verification, Group B (n = 15) was used for magnetic resonance molecular imaging, Group C (n = 15) was used for animal behavior experiment, and Group D (n = 21) was used for histopathological examination.

All of the rats were anesthetized with 4% isoflurane mixed with oxygen to minimize the pain until righting reflex of the rats disappeared before and during operations. The CO-poisoned treatments were given high-purity CO gas (120 ml/kg body weight) by intraperitoneal injection (32–34), while the control treatments were injected with the same dose of air. Since the aim of our study is to investigate the changes in glutamate metabolism in the living brain after carbon monoxide poisoning, the dead rats due to poison were replaced by additional rats and all death data would be excluded.

To ensure the reliability and stability of the animal models, we first confirmed the formation of brain injury in the animal receiving CO treatment before conducting other experiments (Group A). Blood samples were taken from the left ventricle according to the 12 timepoints, as shown in Figure 1. Anesthesia with 2.0–2.5% isoflurane and 1 L/min oxygen were maintained throughout the entire process in all rats after induction anesthesia with 4% isoflurane and 1 L/min oxygen to minimize pain in rats. The carboxyhemoglobin saturation (HbCO%) is a quantitative parameter of verification. It was measured in left ventricle blood using spectrophotometry (35).

All of the magnetic resonance studies were acquired using a 7.0 T animal magnetic resonance scanner (Agilent VnmrJ 3 Imaging, Palo Alto, CA) using a transmitter/receiver body coil (diameter = 160 mm). All of the rats in Group B participated in magnetic resonance scanning according to the schedule in Figure 1 and were anesthetized with 2.0–2.5% isoflurane and 1 L/min oxygen mixture throughout the acquisition process. The respiratory frequency of rats was monitored using a small animal respiratory monitor (MR-compatible small animal monitoring & gating system, Model 1030, SA Instruments, Stony Brook, NY).

The three-dimensional (axial, sagittal, and coronal) localization images were scanned at first to determine the location of the rat brain and to help adjust the position with the following parameters: repetition time/echo time = 24 s/2 ms, the field of view (FOV) = 50 mm × 50 mm, 929 × 2 matrix, 1.0 mm thickness without a gap. Then we obtained the three-dimensional T2-weighted images (T₂WI) using a fast spin echo sequence with the following parameters: repetition time = 3,000 ms, FOV = 35 mm × 35 mm, 2562 × 56 matrix, 2.0 mm thickness, and 0.2 mm gap.

Before GluCEST MRI acquisition, we chose an axial slice with the largest hippocampal area as the target and corrected the static magnetic field (B₀) with three-dimensional gradient filling. Then we modified the high-order gradient filling current based on the derived B₀ maps and further adjusted the radio frequency field (B₁) and center frequency in the pre-scanning. The inhomogeneity of B₀ and B₁ were modified according to the B₀ and B₁ maps from the same target slice. For B₀ correction, the water saturation shift referencing technique (36) was used with the parameters as following: TE/TR = 64/2,000 msec, RARE factor = 8, frequency resolution increments =0.1-ppm and frequency coverage = ±1 ppm. GluCEST sequence was obtained by using frequency-selective continuous wave pulse for pre-saturation and echo plane imaging for image acquisition. The CEST saturation frequencies ranged from−5 to +5 ppm with an interval of 0.2 ppm and the parameters were as follows: B₁5 ₌ .9 μT (250Hz), saturation time = 2s, FOV = 35 mm × 35 mm, TR = 4,100 ms, TE = 42 ms, 1281 × 28 matrix size, 2 average, 2.0-mm thickness, and the total acquisition time is 9 mins.

Following the GluCEST acquisition, ¹H-MRS was acquired using the point-resolved spectroscopy sequence acquisition. The regions of interest (ROIs) were chosen based on three-dimensional T₂WI images. The ¹H-MRS sequence parameters were as follows: TR = 5,000 ms, data matrix = 3203 × 20, ROIs were located at the bilateral hippocampus, parietal cortex, and thalamus; the voxel size is 2.0 mm × 3.0 mm × 10.0 mm.

The quality of MR data is closely related to the homogeneity of the magnetic field, so repeated shimming and B₀ mapping were carried out during the scanning process to ensure the quality of data with a linewidth of <20 Hz.

Throughout the process of modeling, the symptoms and behavioral changes of rats 2 h after modeling were observed and recorded. Moreover, the Morris water maze (MWM) test including the positioning navigation experiment and space exploration experiment were performed before and after modeling (14 and 21 days) of Group C to preliminarily evaluate the effects of CO poisoning on the brain.

All of the rats in Group C were moved into the animal behavioral laboratory in advance to adapt to the environment and reduce tension and discomfort. In the positioning navigation experiment, rats were allowed to swim for up to 120 s to a hidden platform separately 4 times a day for 5 consecutive days and to rest on the platform for 60 s each time regardless of whether they found the platform or not. The space exploration experiment was carried out on the5^th day following the positioning navigation experiment.

The original data collected by the GluCEST sequence were analyzed by a program written using the MATLAB R2013b Software (MathWorks, Natick, MA). The formula for calculating glutamate concentration showed as follows:

(MTR_asym, asymmetric magnetization transfer rate; MTR, magnetization transfer rate; S_sat, signal strength after applying saturated pulse; S₀, signal strength after not applying saturated pulse)

Then the contrast of GluCEST was corrected according to the inhomogeneity of B₀ and B₁. In this study, we used GluCEST% to describe MTRsaym (3.0ppm). The GluCEST% (in direct proportion to glutamate concentration) was mapped to the anatomical image as pseudo-color to get the image, and calculated as the relative glutamate concentration in the ROIs.

The original data of ¹H-MRS were performed in the LCModel and LCMgui software (v. 6.2–4E; S.W. Provender). The appropriate parameters of 1H-MRS were selected for automatic linear fitting and data processing analysis. The relative concentrations of metabolites were obtained by using the concentration of water in rat brain tissue as an internal reference. In the estimated using the LCModel fitting, only metabolite concentration values that have a standard deviation (SD) <20 were considered trustable and used in the analysis.

The video of the MWM experiment was analyzed by EthoVision® XT (Noldus, Wageningen, Netherlands). The latency time, the latency distance, and the cumulative time in the correct quadrant could be obtained in the positioning navigation experiment and the space exploration experiment.

The rats in Group D were anesthetized until their righting reflex disappeared. After cardiac perfusion, with 4% polyformaldehyde solution, their brain tissue was removed, according to the timepoints described in Figure 1, for 48 h and then immobilized in paraffin solution. Fixed wax blocks were cut on a microtome (RM2125, Leica Biosystems, German) until the layers (hippocampus, parietal cortex, and thalamus) appeared. The brain was then continuously sliced at every 5 microns. All of the sections were dyed with hematoxylin and eosin. Finally, all of the sections were dehydrated, rendered transparent, and fixed with resin. Images were collected with the help of a microscope after all pathological sections were dried (IM50, Leica, Olympus Corporation, Japan).

All of the results were analyzed by IBM SPSS Statistics 16.0 software. The results for each quantitative parameter were expressed as the mean ± SD. One-way analysis of variance and Student's t-test was applied to compare among and between groups when the data were normally distributed, and the variance was homogeneous. Otherwise, a non-parametric test was used instead. Pearson correlation analysis was applied when data were normally distributed, and Spearman correlation analysis was applied when data were not normally distributed. The level of p < 0.05 was considered to be statistically significant.

The changes of HbCO% in the left ventricular blood of rats before and after 0.5, 1, 1.5, 2, 2.5, 3, 6, 9, 12, 24, and 72 h of CO exposure are shown in Figure 2, where the HbCO% in the blood increased rapidly, plateaued, and then gradually decreased from the plateau stage to near the pre-exposure level after intraperitoneal injection of high purity CO. The peak value of HbCO% in the blood after exposure was 7.3-fold higher than before exposure. The highly elevated HbCO% level lasted for approximately 2.5 h after exposure.

The changes in brain T₂WI were not obvious after CO poisoning. Punctate hyperintensity (pointed by arrow; Figure 3) could be seen in the cortex and hippocampus. After poisoning, the boundary between the gray and the white matter was blurred (pointed by arrowhead; Figure 3) in the first 24 h. However, these changes were difficult to identify and could not be determined qualitatively and quantitatively.

The magnetic resonance examinations data were normally distributed and their variance was homogeneous, so Student's t-test was applied to compare the poisoned group and the control group at different time points.

Before modeling, there was no significant difference in the GluCEST% value between Group B1 and B2. After modeling, the GluCEST% value of the poisoning group (B2) was significantly higher than the control group (B1); the difference was statistically significant (Figure 4). The differences in the three ROIs (bilateral hippocampus areas, cortical areas, and thalamus areas) continued to the 7^th day after CO-poisoning. In the hippocampus area, the value of GluCEST% increased continuously to the peak value on the 7^th day after CO-poisoning; the value in the cortical areas and thalamus areas reached their first peak in 24 h.

Compared with that before the poisoning, the value in Figure 5 of Glu and Glu+Gln (Glx) from ¹H-MRS increased statistically in bilateral hippocampus and thalamus areas after CO exposure, and the differences in the hippocampus and thalamus persisted 24 and 2 h after modeling respectively. But the Glu and Glu+Gln increased in bilateral cortical areas after CO exposure without statistical differences.

The poisoned rats showed a redder mucosa around their auricle and mouth than the control 5 mins after intraperitoneal injection of high purity CO. They also had weaker hind limbs. Some rats showed shortness of breath, convulsions, coma, and even death. There were no significant abnormal changes in the control rats after injection.

In the animal behavioral data, the escape latency time data were normally distributed and their variance was homogeneous, so we applied Student's t-test to compare the poisoned group and the control group. Meanwhile, the rest of all the other animal behavioral data were not normally distributed, so the nonparametric test was used.

We studied behavioral changes of the CO-exposed effectiveness on Group C rats via MWM (Figure 6). In the orientation navigation experiment, the poisoned rats of 14 days after modeling had significantly longer escape latency time vs. baseline (the control treatments). The poisoned rats had longer swimming distances. Furthermore, the cumulative time in the correct quadrant of the poisoned rats in both poisoning time points was less than baseline. In the space exploration experiment, the cumulative swimming time in the target quadrant of the poisoned rats in both poisoning time points was less than that at baseline; there were no statistical differences in swimming velocity.

Since the magnetic resonance examinations data and the escape latency time data were normally distributed and their variance was homogeneous, Pearson correlation analysis was applied. Meanwhile, the rest of all the other animal behavioral data were not normally distributed, so we used Spearman correlation to analyze the other correlations.

Since 21 days after poisoning was chosen as the end point of our study, we conducted a correlation analysis between the GluCEST data of the three ROIs, respectively, on the 21st day after poisoning and the average MWM data of the 21 days group. Figure 7 showed correlations between GluCEST% values and MWM-related results. There was a positive correlation between the latency time and GluCEST% of parietal cortex (ρ = 0.618, p = 0.014) and thalamus (ρ = 0.518, p = 0.048) when the strong positive correlations between swimming distances and GluCEST% of parietal cortex (ρ = 0.818, p < 0.001) and thalamus (ρ = 0.751, p = 0.001) were observed.

Hematoxylin–eosin staining showed the bilateral hippocampus areas and parietal lobe areas of the poisoned rats showed pyknosis, hyperchromatic cytoplasm, and disappearance of central Nissl bodies within 3 days after exposure. The lesions mentioned above decreased significantly 7, 14, and 21 days after exposure. The thalamic areas of the poisoned rats were only obvious 2 h after exposure, and the lesions were then significantly reduced over time (Figure 8).