Glutamate diffusion in the rat brain in vivo under light and deep anesthesia conditions

Abstract

Purpose

Glutamate (Glu) is the most abundant neurotransmitter in the human central nervous system and glutamatergic neurotransmission has been implicated in many common and severe neuropsychiatric disorders. In vivo MRS techniques have been developed to measure brain Glu concentration to investigate the pathophysiology of various brain disorders. However, it is difficult to interpret Glu signal changes because Glu plays multiple roles in the brain and is found in multiple microenvironments including cytosolic, vesicular, and extracellular.

Methods

In vivo diffusion-weighted MRS (DW-MRS) with low to very high b-values was performed on the rat prefrontal cortex at 9.4T under both light and deep anesthetic conditions to examine Glu diffusion properties.

Results

Significant alterations in Glu diffusion as well as reduced Glu concentration were observed under deep anesthesia compared with superficial anesthesia in the absence of similar changes in NAA or creatine.

Conclusion

The modifications in Glu diffusion under deep anesthesia might reflect changes in Glu microenvironment. The present work shows that Glu DW-MRS could be an important tool to explore Glu physiology with changing levels of neuronal activity and synaptic function.

1 |. INTRODUCTION

Glutamate (Glu) is the most abundant neurotransmitter in the human central nervous system and glutamate neurotransmission has been implicated in many common and severe brain disorders such as epilepsy,¹ amyotrophic lateral sclerosis,² schizophrenia,³ bipolar disorder,⁴ and Alzheimer’s disease.⁵ Recent genetic work, in particular, is pointing to excitatory synaptic function as a key abnormality in several of these conditions.^6,7 Methods exist to quantify glutamate (Glu) concentrations non-invasively in humans, to map changes in Glu concentration across the whole brain with relatively short timescales,⁸ and even to observe Glu synthesis and subsequent conversion to glutamine (Gln).⁹ However, no non-invasive methods have been shown to reflect Glu changes related to synaptic activity. This challenge is compounded by the fact that Glu plays multiple roles in the brain and is found in multiple microenvironments, including cytosolic, vesicular, and extracellular.

Diffusion-weighted MRS (DW-MRS) provides a non-invasive tool to obtain metabolite-specific micro-environmental information.¹⁰ Diffusion signals from N-acetyl aspartate (NAA) and other metabolites have been previously used to probe axon abnormalities in disease states.^11–13 Glu diffusion can similarly be quantified using sophisticated DW-MRS from low to very high b-values.^14,15

In the current study, we sought to examine the Glu MR signal under differing anesthetic conditions that are known to modulate cerebral metabolism and brain activity in rats.¹⁶ We reasoned that changes in brain activity induced by anesthesia might be systematically reflected in Glu diffusion and concentration measures. Therefore, we performed in vivo DW-MRS using low to very high b-values on the rat brain prefrontal cortex under both light and deep anesthesia. Because of the unique role of Glu in brain function, we hypothesized that we would observe changes in Glu concentration and diffusion during modulation of neuronal activity, reflecting potential changes in metabolism, microstructure, and synaptic function.

2 |. METHODS

2.1 |. Animals

All experimental protocols were performed according to the guidelines of the National Institute of Health and approved by the Institutional Animal Care and Use Committee of our institution. Fourteen male Sprague–Dawley rats (350–400 g) were used for our DW-MRS measurements. The animals were induced by 2.5% inhaled isoflurane in pure oxygen during experiment setup and maintained at 1.5% inhaled isoflurane during anatomy scans and DW-MRS under the light anesthesia condition. The respiratory rate was ~70–90 breaths/min. After DW-MRS under the light anesthesia condition (~50 min), the isoflurane level was switched to 3.5% with a transition time of 30 min. During the transition period, DW-MRS with b = 0 was acquired and saved frame-by-frame. After the animal’s physiology reached stable conditions (~30 min), another round of DW-MRS for both metabolites and water was measured following the same strategy and with the same shimming conditions. The respiratory rate was reduced to 30–50 breaths/min. The total scan time for each rat was ~160 min. Rectal temperatures were maintained at 37 ± 1°C throughout the whole experiment using a cycling waterbed system.

2.2 |. DW-MRS experiments

MRI scans were performed on a 9.4T horizontal animal magnet (Magnex Scientific, Abingdon, UK) interfaced with Varian INOVA console (Palo Alto, CA) at McLean Hospital with a quadrature surface coil (each loop with ~1.5 cm diameter) for both RF excitation and signal reception. The sagittal-axial anatomic image was acquired using a multiple spin-echo sequence (TR/TE = 2000/10 ms, echo number = 4, average = 2) using the anterior commissure as a landmark for axial image positioning. Imaging voxel size was 0.16 × 0.12 × 1 mm³. In vivo DW-MRS was acquired in a large voxel of 5 × 5 × 2 mm³ on the prefrontal cortex (Figure 1) with diffusion gradients parallel to the diagonal of 3 dimensions to achieve maximum strength. The center frequencies of the location modules were set to 2.3 ppm (on the Glu resonance) to minimize the voxel shift displacement for compounds of interest (Glu, tCr, and NAA). 3D gradient shimming was performed before the MRS scan with criteria of water linewidth <14 Hz. The DW-MRS sequence was modified based on a STEAM sequence with TE/TM/TR = 18/50/3000 ms and diffusion time (t_D = 55 ms). Other than the DW-MRS with b₀ control scan, 11 spectra were acquired with increased signal averages from low to high b-values to compensate for reduced SNR. The experimental parameters of b-values (s/mm²) and averages were outlined as follows: 0 (50), 434.8 (50), 1014.5 (50), 1884.1 (50), 3043.5 (50), 4347.8 (50), 5797.1 (50), 8695.7 (100), 11,594.2 (100), 15,942.0 (100), 21,739.1 (150), and 28,985.5 (200). The actual b-value for b₀ was 60 s/mm² calculated based on the MRS sequence and the following b-values were referenced to b₀ and further linearly calibrated using an ethylene glycol phantom with known diffusion coefficient of 1.02e–4 mm²/s as reported in Ligneul et al.¹⁷ The water suppression was performed using variable pulse power and optimized relaxation delays water suppression (VAPOR) scheme¹⁸ using Gauss shape pulse with a suppression bandwidth of 200 Hz. DW-MRS is subject to strong phase variations between averages especially at high b-values because of the diffusion gradient pair. Therefore, scan-by-scan phase and frequency corrections are necessary for DW-MRS. On the other hand, the metabolite signals are almost invisible for a single average at high b-values, so it is inaccurate to use them for the corrections. To improve the phase and frequency corrections at high b-values, we adjusted the water suppression to reserve a moderate residual water signal that was visible in a single shot, but not too strong to distort the baseline. Every average was saved for offline phase, frequency, and eddy current corrections. Averages were acquired interleaved with different b-values and the total scan time of DW-MRS for each anesthesia period was ~50 min. Water DW-MRS was acquired with the same parameters, but without RF pulse for water suppression from the same region. Four averages were acquired for all b-values, and the center frequencies of the location modules were set on the frequency of the water resonance. The water signal was used for eddy current corrections. It is important to include experimental macromolecule (MM) resonances in the LCModel basis sets to address the baseline issue of short-TE MRS.¹⁵ We measured MM by an inversion pulse followed by a delay of 800 ms before the DW-STEAM sequence to null metabolite signals. To better account for MM diffusion effect on metabolite quantification under varied b-values, 3 nulled DW spectra with low, medium, and high b-values (1014.5, 4347.8, and 15,942.0 s/mm², respectively) with 512 averages each were acquired on 2 additional rats. The acquisition strategy was the same as the previously mentioned DW-MRS with moderate water suppression and every average saved for offline corrections.

FIGURE 1.

DW-MRS voxel location and representative DW-MRS spectra with up to high b-values. 1 Hz Lorentz line broadening was applied to the spectra

2.3 |. MRS data processing

All MRS spectra were preprocessed using FID-A.¹⁹ The phase and frequency corrections were processed for each b-value individually before summation. The eddy current correction was performed using the unsuppressed water spectra as reference in LCModel version 6.3.1H²⁰ as well as the quantification. For MM measurements, the residual metabolite peaks of NAA (2.0 ppm), taurine (3.2 and 3.4 ppm), and creatine (3.9 ppm), as identified in Kunz et al.,²¹ were removed using HLSVD in jMRUI 6.0.²² The 3 corresponding macromolecule spectra, as mentioned above, were included in separate LCModel basis sets to fit DW-MRS with low (0, 434.8, 1014.5, 1884.1 s/mm²), moderate (3043.5, 4347.8, 5797.1, 8695.7 s/mm²), and high b-value (11,594.2, 15,942.0, 21.739.1, 28,985.5 s/mm²), respectively. An alternative quantification strategy using only the MM spectrum with the high b-value (15,942.0 s/mm²), which was with the least metabolite contaminations, for all DW spectra LCModel fitting was also performed to quantify metabolite diffusions. The other metabolites included in LCModel basis sets were: alanine, aspartate, creatine, phosphocreatine, GABA, glucose, glutamine, glutamate, glycerylphosphorylcholine, phosphorylcholine, glutathione, myo-inositol, lactate, NAA, NAAG, scyllo-inositol, and taurine. They were simulated using VeSPA²³ with the same parameters as experimental acquisition.

A 2-way repeated measures ANOVA was performed on IBM SPSS Statistics Version 24 to compare the values measured with different anesthesia conditions and b-values, followed by the fitting of diffusion curves with 2 analytical models: (I) empirical bi-exponential decay (fast and slow diffusion components as well as its ratio, i.e., ADC_f, ADC_s, and Ratio_f/s),^14,17 and (II) randomly oriented cylinders with the same diameter of r (D₀, r).^24,25 The calculation was performed on MATLAB Version R2016b (The MathWorks, Natick, MA) for each rat individually under low–high dose isoflurane anesthesia conditions. Repeated t-test was performed in SPSS for comparisons of the above fitting parameters as well as the fitting residuals at low–high dose isoflurane anesthesia conditions.

The concentrations of metabolites of interest were quantified from DW-MRS with b₀. The concentrations were normalized to the first data point of the same rat (i.e., the b₀ spectrum of low anesthesia). One-way repeated measures ANOVA were also performed using SPSS to determine whether metabolites concentrations changed during the transition period from superficial to deep anesthesia.

3 |. RESULTS

A series of representative diffusion-weighted spectra (black curve) as well as the LCModel fitting (blue curve) are shown in Figure 1. One Hz Lorentz line broadening was applied to the spectra. Every spectrum was censored for acquisition and fitting quality in our experiments. Two of 14 rats were excluded for low spectral quality (linewidth >15 Hz at b-value = 0). The SNR, linewidth and CRLBs measured by LCModel under the 2 anesthesia conditions are shown in Table 1. Data quality was generally good, allowing for the signal decrement at higher b-values.

TABLE 1.

Spectral SNR and linewidth, as well as CRLBs of Glu, NAA, and tCr under 2 anesthesia conditions

b_values (s/mm2)	SNR		Linewidth		Glu_CRLB(%)		NAA_CRLB(%)		tCr_CRLB(%)
	Low	High	Low	High	Low	High	Low	High	Low	High
0	14 ± 3	15 ± 4	11 ± 1	10 ± 2	3 ± 1	4 ± 1	3 ± 1	3 ± 1	3 ± 1	3 ± 1
435	14 ± 3	14 ± 4	11 ± 1	10 ± 1	3 ± 1	4 ± 1	3 ± 1	3 ± 1	3 ± 1	3 ± 1
1014	13 ± 2	13 ± 3	11 ± 2	10 ± 2	4 ± 1	4 ± 1	3 ± 1	3 ± 1	3 ± 1	3 ± 1
1884	11 ± 2	12 ± 3	11 ± 1	11 ± 2	4 ± 1	4 ± 1	3 ± 1	3 ± 1	3 ± 1	3 ± 1
3043	10 ± 2	10 ± 3	11 ± 1	11 ± 2	4 ± 1	5 ± 1	4 ± 1	4 ± 1	4 ± 1	4 ± 1
4348	9 ± 2	9 ± 2	12 ± 2	11 ± 2	5 ± 1	5 ± 1	4 ± 1	4 ± 1	4 ± 1	4 ± 1
5797	8 ± 2	8 ± 2	12 ± 1	12 ± 2	5 ± 1	6 ± 1	4 ± 1	4 ± 1	4 ± 1	4 ± 1
8696	9 ± 2	9 ± 2	12 ± 1	12 ± 1	4 ± 1	5 ± 1	4 ± 1	4 ± 1	4 ± 1	4 ± 1
11,594	8 ± 2	8 ± 2	13 ± 2	12 ± 1	5 ± 1	6 ± 1	5 ± 1	4 ± 1	5 ± 1	5 ± 1
15,942	7 ± 2	6 ± 2	13 ± 2	13 ± 2	6 ± 1	6 ± 1	5 ± 1	5 ± 1	5 ± 1	5 ± 1
21,739	7 ± 2	6 ± 2	14 ± 2	13 ± 2	6 ± 1	7 ± 2	6 ± 1	6 ± 2	6 ± 1	6 ± 1
28,986	6 ± 1	6 ± 2	14 ± 2	13 ± 2	6 ± 1	7 ± 2	6 ± 1	6 ± 2	6 ± 1	7 ± 2

The upper row of Figure 2 shows the fittings of the representative spectra of 1 rat with low, moderate, and high b-values, respectively. It can be seen that the baselines were well accounted for by the experimental MM basis sets and that the peaks of Glu, NAA, and tCr were well separated from other resonances and reliably quantified (see CRLBs in Table 1). Note that in Figure 2, the corresponding macromolecule spectrum with low, medium, and high b-values was included in the separate LCModel basis set to fit DW-MRS. The bottom row of Figure 2 shows mono-exponential fittings of MM baseline under superficial and deep anesthesia. The fitted ADCs of MM with light-deep anesthesia were 1.25e–5 and 1.27e–5 mm²/s, respectively, without significant difference (P = 0.9). The ADC values were close to the ones reported in Ligneul et al.,¹⁷ (1.6e–5 and 1.1e–5 mm²/s) with similar diffusion time. The consistency of MM ADC with the literature confirms that there is no artificial signal loss even at high b-values.

FIGURE 2.

(A–C) DW-MRS quantification at low, moderate and high b-values of a representative rat. (D) Mono-exponential fittings of its macromolecule diffusion attenuation under low and high anesthesia conditions

The signal attenuations of Glu, NAA, and tCr were presented in Figure 3 with means and SDs for 12 rats. A 2-way repeated measures ANOVA with a Greenhouse-Geisser correction was performed to compare the values measured with different anesthesia conditions and b-values. Difference of Glu signal attenuation is approaching significance between the 2 isoflurane levels (P = 0.12). Paired t-tests shows significant difference in Glu signal attenuation at several (but not all) b-values between superficial and deep anesthesia. No such differences were observed for NAA or tCr.

FIGURE 3.

Averages and SDs of the signal attenuation of Glu, NAA, and tCr (N = 12) as a function of b value

It has been widely known that the diffusion of metabolites in the brain does not follow mono-exponential decay.^26,27 It has been demonstrated that the deviation from mono-exponentially is likely to be caused by restricted diffusion²⁸ or the ensemble effect of the diffusion along randomly oriented fibers²⁹ instead of the presence of 2 distinct physical compartments. Although it has been shown that brain metabolites primarily diffuse in fibers^30,31 and several models based on simplified assumptions^25,29,32 or Monte Carlo simulation have been proposed,^33,34 currently there is no widely accepted model for intracellular diffusion in biological tissues. Therefore, we used 2 distinct analytical models to explore the change in Glu diffusion signal. We applied these models to diffusion attenuations of each of the 12 rats.

The mean and SDs of individual bi-exponential decay fitting curves of 12 rats under these 2 anesthesia conditions are shown in Figure 4. The fitting parameters and the comparisons between 2 anesthesia conditions are summarized in Tables 2, 3, and 4 for Glu, NAA, and tCr, respectively. With the bi-exponential fitting, we observed a statistically significant Glu ADC reduction for both the fast and slow diffusion components under deep anesthesia as compared with superficial. There was also an increase in the fast-slow component magnitude ratio with deep anesthesia. No significant changes were observed for NAA and tCr diffusion parameters, although we note that the tCr diffusion parameters showed numeric changes in the opposite direction from those of Glu (i.e., nonsignificant ADC elevations in the fast and slow diffusion components, and a small reduction in the fast–slow component magnitude ratio with deep anesthesia). It is well known that the values of ADC in slow component (ADC_s), fast component (ADC_f), and the ratio of the fractions of fast and slow components (f_fast/f_slow), deduced from the bi-exponential fitting model, are highly dependent on one another. Hence, for these 3 diffusion measures of Glu, we performed the false discovery rate (FDR) test originally introduced by Benjamini and Hochberg.³⁵ The Glu results of single MM basis or 3 different MM basis preserve the significance: the FDRs of ratio, ADC_f, and ADC_s are (0.04, 0.04, 0.04) and (0.048, 0.036, 0.036), respectively.

FIGURE 4.

The fittings of 2 analytical models: the means (dash lines) and SDs (shadows) of the individual fitted curves under 2 anesthesia conditions (red and blue for 1.5% and 3.5% isoflurane, respectively)

TABLE 2.

Fitting of Glu diffusion under low-high isoflurane dose with 2 analytical models

		Model parameters
Fitting model	Isoflurane(%)	Ratio_f/s	ADC_f(mm2/s)	ADC_s(mm2/s)	Residuals
Model I: bi-exponential14,17	1.5	0.88	3.74e-4	2.74e-5	6.73e-3
	3.5	1.15	3.01e-4	2.11e-5	6.95e-3
	P	0.040	0.023	0.036	0.86
			D0(mm2/s)	Diameter(μm)
Model II: randomly oriented cylinders with restricted perpendicular diffusion25	1.5		4.89e-4	1.34e-4	1.22e-2
	3.5		4.50e-4	0.31	1.33e-2
	P		0.12	0.084	0.67

TABLE 3.

Fitting of NAA diffusion under low-high isoflurane dose with 2 analytical models

		Model parameters
Fitting model	Isoflurane	Ratio_f/s	ADC_f(mm2/s)	ADC_s(mm2/s)	Residuals
Model I: bi-exponential14,17	1.5%	0.99	3.27e-4	2.81e-5	3.99e-3
	3.5%	1.09	3.04e-4	2.57e-5	5.83e-3
	P	0.46	0.60	0.53	0.47
			D0(mm2/s)	Diameter(μm)
Model II: randomly oriented cylinders	1.5%		5.00e-4	0.12	6.00e-3
with restricted perpendicular diffusion25	3.5%		4.86e-4	0.26	8.56e-3
	P		0.62	0.43	0.48

TABLE 4.

Fitting of tCr diffusion under low-high isoflurane dose with 2 analytical models

		Model parameters
Fitting model	Isoflurane	Ratio_f/s	ADC_f (mm2/s)	ADC_s (mm2/s)	Residuals
Model I: bi-exponential14,17	1.5%	0.94	3.44e-4	2.93e-5	3.47e-3
	3.5%	1.03	3.52e-4	3.21e-5	3.66e-3
	P	0.60	0.90	0.48	0.71
			D0(mm2/s)	Diameter(μm)
Model II: randomly oriented	1.5%		4.61e-4	0.66	4.64e-3
cylinders with restricted perpendicular diffusion25
	3.5%		4.59e-4	1.00	5.43e-3
	P		0.97	0.20	0.42

In the randomly oriented cylinder model with restricted perpendicular diffusion (Model II) the fittings of Glu and NAA converge to a very small diameter. This situation was similar to the fitting of NAA with Model II as reported in the literature.²⁵ The Glu fitting parameters in Models II are with much lower P values between the 2 anesthesia levels when compared to NAA fitting parameters. This is consistent with our results at each b-value and using the bi-exponential model. Interestingly, the fitting of tCr with Model II showed a different trend with more reasonable diameters.

We also continuously acquired spectra without diffusion weighting during the half-hour transition time from low (1.5%) to high dose (3.5%) isoflurane anesthesia. A total of 512 averages with TR = 3 s were acquired and every 64 averages (3 min) were summed together to obtain similar SNR as b₀ spectra of DW-MRS under the 2 conditions. The 8 spectra during transition together with the b₀ spectra of DW-MRS at light and deep anesthesia outline the metabolite concentration alterations between the 2 anesthetic conditions. Changes in Glu concentration are presented in Figure 4 along with those of NAA and tCr for comparison. We observed a statistically significant 10% reduction in Glu concentration between the 2 anesthetic conditions in the absence of similar changes in NAA and tCr.

4 |. DISCUSSION

4.1 |. Measurement quality

Spectral and quantification qualities are critical to DW-MRS studies using high b-values because the strong diffusion gradients not only reduce SNR, but also induce severe eddy current effects. The long scan time used to compensate for the reduced SNR may also introduce instability. Therefore, every spectrum was censored for acquisition and fitting quality in our experiments. Two rats were excluded for low spectral quality (linewidth >15 Hz at b-value = 0). High spectral quality was assured by well-performed shimming as well as abundant SNR using an increased number of averages at higher b-values. From the presented spectra (Figure 1), it can be seen that good line-shape was preserved at high b-values after eddy current correction. The Glu peak at 2.3 ppm is well-resolved at all b-values, and CRLBs are similar compared to those of NAA and tCr at the same conditions. Paired t-tests did not show significant differences between these quality measures under superficial or deep anesthesia. This ensures that the comparisons of diffusion and concentration between the 2 anesthetic levels can be carried out under the same measurement conditions. It has been shown in previously published work¹⁵ that simulated macromolecules in the LCModel basis set has an effect on the quantification of glutamate especially at high b-value. Therefore, we acquired experimental MM spectra with different b-values to account for the baselines under different levels because it is possible that a continuum of MM molecular weights contribute to MM signal³⁶ with larger MM being associated with slower diffusion.¹⁷ On the other hand, we also performed the quantification for all spectra with different b-values using the same LCModel basis in which MM resonances were measured under the highest b-value. Under this condition, MM resonances have almost no metabolite contaminations.²¹ This approach shows a very similar trend (Supporting Information Table S1) and no statistical difference compared to results achieved by 3 different b-value MM baselines (Tables 2, 3, and 4).

It is notable that ADCs in our measurements are slightly higher compared to reported values.^14,17 This could be caused by the condition of animal physiology, different magnetic fields, and acquisition parameters, especially t_D, which has been illustrated to have a large effect on ADC measured values. A longer diffusion time would reduce ADCs and reflect metabolite diffusion from a more restricted pool.^26,37 On the other hand, this also could be caused by artefactual signal loss, especially at high b values. However, our bi-exponential fitting curves showed little deviation from the data points up to high b-values, which indicates that there was no significant artefactual signal loss at high b compared to low b. MM signal attenuation presented in Figure 3 is consistent with the one reported in Ligneul et al.¹⁷ This again confirms that there is no significant artefactual signal attenuation, which could confound results in our measurements.

4.2 |. Metabolite diffusion

In the 2 models presented, bi-exponential fitting provided better fitting results for the diffusion behaviors of intracellular metabolites. The fast and slow components of all 3 metabolites largely follow a 1:1 ratio, which is consistent with previous reports.^14,17 Interestingly, only Glu showed significant changes in slow component ADC (ADC_s), fast component ADC (ADC_f), and the ratio of 2 components when switching from superficial to deep anesthesia. In contrast to ADCs of Glu, the ADCs of NAA showed a trend in the same changing direction whereas those of tCr changed in the opposite direction. Neither of these changes were significant. The unique modulation of Glu diffusion indicates that Glu may experience a microenvironment that differs from those of NAA and tCr. The interpretation of the multi-exponential decay would be complicated considering cell diameters, anisotropic features, and sub-compartments.¹⁴ We cannot determine which changes in the Glu microenvironment could lead to our observations based on the data available to us, beyond noting that changes in brain activity and neurotransmission appear to impact Glu diffusion properties differentially. It is worth bearing in mind that brain metabolite ADC in vivo is not directly dependent on the free diffusion coefficient,^14,38 and the molecular diffusion in the tissue compartments is potentially influenced by a number of factors and has never been clearly assessed in vivo.

As reported previously,^30,31 brain metabolites diffuse mostly in fibers. The ensemble effect of diffusion along large number of randomly oriented fibers as well as the restricted diffusion perpendicular to the fibers accounts for the non-mono-exponential feature of the decay.^28,29 Therefore, we included the randomly oriented cylinder model with perpendicular diffusion in this study (Model II).²⁵ Generally, the cylinder model did not fit as well as the bi-exponential model, probably because of the less fitting parameters and simplified assumptions. However, it may provide interesting new information on structural parameters. As shown in Tables 2 and 3, the fittings of Glu and NAA with the randomly oriented cylinder model converge to diameter ≈0, which is similar to the model proposed in Kroenke et al.²⁹ The observed intracellular diffusion coefficient of NAA was also very close to the values reported in human³⁹ and rodent brain.⁴⁰ The fitting of tCr showed a different trend with more reasonable diameters. Based on Figure 3 reported in Palombo et al.²⁵ and our simulations, the fitted fiber diameter is very vulnerable to the variation of the decay especially when diameter <1 μm. This could explain the difference of fitted parameters comparing our results and other’s.²⁵ The randomly oriented model could be over-simplified, but it showed potential structural differences between NAA/Glu and tCr, which could be explained by the fact that NAA and Glu mostly exist in neurons and tCr in mixed cell types.³³ Both models agreed in showing reductions in Glu diffusion with deep anesthesia. This may indicate microstructural changes, redistribution of the metabolite in different environments, or even slow-down of the transport in cells.³⁸ We cannot interpret this observation without other evidence. Interestingly, the diffusion of NAA showed a similar trend as Glu, but Glu shows greater sensitivity to brain activity modulation.

Glu has more than 1 clearly defined compartment in the brain.⁴¹ Detailed neurochemical and immunohistochemical studies established by the early 1990s that Glu is compartmentalized in neurons with a “metabolic pool” (~50% of total) in the cytosol and a “neurotransmitter pool” (~20–30% of total) in synaptic boutons, almost all in synaptic vesicles.⁴² The remaining brain Glu is found in the cytosol of GABAergic neurons and glial cells (~10% each).⁴² A very small amount of Glu is thought to be in extracellular spaces including the synaptic cleft. The metabolic and synaptic Glu compartments are in very different chemical and physical environments, which could result in completely different MR properties such as diffusion, relaxation and even MR visibility. The neurotransmitter compartment of Glu could contribute to the difference between the diffusion of NAA and Glu in neurons and probably their differential changes during the anesthesia modulation. Much additional work is needed to examine this issue, in particular using ex vivo studies of synaptic vesicles. Nonetheless, our work demonstrates that it is possible to indirectly observe subtle changes in Glu microenvironment and diffusion behavior using 1-H DW-MRS.

Our finding of metabolite ADC changes is not in line with a previous study that observed that the ADCs of all metabolite increased in monkey fronto-parietal lobe when isoflurane level varied from 1% to 2%⁴³ (see also Supporting Information Figure S1). This discrepancy could result from the differences in the studied animal, the magnetic field, the isoflurane dose, and DW-MRS approaches (a diffusion tensor spectroscopy was performed in the previous study).⁴³ However, our study is the first one to perform DW-MRS using multiple b-values to study the anesthesia effect; the unique diffusion property change of Glu is interesting and needs to be studied further by future in vivo or ex vivo studies.

4.3 |. Concentration

Another interesting finding in this work is the reduced Glu concentration during the transition from low to high dose isoflurane. The Glu concentration was observed to remain at a low level during the deep anesthesia scan compared to light anesthesia. Glu is an excitatory neurotransmitter and its synthesis may be affected by neuronal activity (e.g., a reduction in conditions of reduced brain activity). Glu concentration is reported to be reduced following continuous isoflurane exposure.⁴⁴ During the transition time from low dose to high dose isoflurane, our observation of the Glu concentration followed the same trend but with more dramatic reduction. The differences between that study⁴⁴ and ours may be because of the fact that we modulated between 2 levels of anesthesia whereas that study switched isoflurane off and back on. We also observed an increase in lactate concentration during and after the 30 min transition (Supporting Information Figure S2).

4.4 |. Limitations

DW-MRS studies using high b-values are always subject to long scan times because abundant b-value points are necessary for a bi-exponential fitting. Also, at high b-values the diffusion attenuation is strong for metabolites and more averages are needed to compensate for the SNR loss. In our study, 2 DW-MRS experiments had to be completed at superficial and deep anesthesia, as well as a half hour transition time between them. Therefore, the average number has to be balanced with a reasonable total scan time. Higher SNR would definitely improve the quantification of the spectra with high b-values.

Because of the limited SNR of single shot signal especially at high b values, we used moderate water suppression to preserve some water signals to use as the reference for phase and frequency corrections. Therefore, the baseline close to water was subject to larger slopes and the metabolite quantifications for tCho, myo-inositol, and taurine were subject to larger variations compared to the 3 metabolites we reported here. Therefore, we did not include these data in the current manuscript, but it is worthy to note that no significant difference, either in signal intensity at each b value or diffusion parameters deduced from bi-exponential fitting, between the 2 anesthesia levels was observed for all these metabolites. In addition, observation of Gln diffusion is also very important to obtain insights into the Glu-Gln cycle between neurons and astrocytes. However, because of the lower SNR, Gln measurement is subject to greater unreliability. The Gln diffusion measurement did not achieve sufficient quality and was excluded in this study.

Diffusion time and/or length is another important parameter when probing the microenvironment.⁴⁵ In this study, we used a medium TM (50 ms) and t_D (55 ms) to obtain a balance between SNR of Glu and diffusion time. At this diffusion time, most metabolites are limited by relatively short diffusion length. In the future work, we would further observe the effect of diffusion time on Glu diffusion and optimize experimental parameters to explore microstructure changes.

Because of the multiple Glu compartments in neurons, the 2 diffusion models used here may not fully reflect the complexity of Glu behavior. A more complicated model with more diffusion parameters such as diffusion time and well-designed computer simulation strategies^33,34 may provide more details of Glu diffusion as well as those of other metabolites.

5 |. CONCLUSIONS

In the current study, significant Glu diffusion and concentration changes were observed for the first time under different anesthesia levels in the absence of similar changes in NAA or tCr. The unique modulation of Glu diffusion may relate to alteration of Glu microstructure, compartments or molecular dynamics, while the modulation of Glu concentration may reflect changes in the Glu-Gln cycling rate regulated by synaptic activity. The findings in this study could have potential applications in brain disorders that involve abnormal Glu regulation.

Supplementary Material

1

FIGURE S1 Averages and SDs of the signal attenuation of total choline (tCho), myo-inositol (ml), and taurine (Tau) (N = 12) as a function of b-value

FIGURE S2 Concentrations of tCho, ml, Tau, and lactate (Lac) during 2 levels of anesthesia conditions as well as the transition from 1.5% to 3.5% isoflurane

TABLE S1 Comparison of LCModel fitting with 3 different b-value MM baselines and 1 high b-value MM baseline bi-exponential decay fittings of Glu, NAA, and tCr under 2 anesthesia conditions

FIGURE 5.

Concentrations of Glu, NAA, and tCr during 2 levels of anesthesia conditions as well as the transition from 1.5% to 3.5% isoflurane