Diffusion signal attenuation of endogenous brain metabolites is substantially influenced by microstructure. This has been used in some works on water and metabolites diffusion coupled with modeling to extract structural properties such as axonal diameters. However, the attenuation could also be strongly dependent on relaxation times due to the presence of different compartments with different relaxation properties, which may hamper any attempt to model diffusion based on geometry only. A few studies [1][2][3] have indeed reported some dependency of metabolite diffusion properties on echo time (TE). In this work we take advantage of new methodological features to revisit the potential relation between metabolite diffusion and relaxation, by performing diffusion-weighted spectroscopy (DW-MRS) in a large voxel of the mouse brain, and changing TE and mixing time (TM) at two diffusion times, up to very high diffusion-weighting b.

We designed a new sequence consisting of a diffusion-weighting stimulated echo (STE) block, followed by a LASER localization block (Fig. 1A). Because both blocks are isolated from each other, there is no cross-term between diffusion and localization gradients. Diffusion-weighting induced only by the LASER block is small (0.22 ms/µm²) and will be neglected in our study. TE and TM can be varied independently of the delay ∆ between the two diffusion gradients (Fig. 1B). This “STE-LASER” sequence has been implemented on an 11.7 T Bruker scanner (Gmax=752 mT/m), equipped with a cryoprobe.

10 mice were repeatedly scanned (body weight 28-30 g). Animals were anesthetized with 1.2-1.5% isoflurane. Water and metabolite spectra (128 repetitions) were acquired in a 72 µL voxel, for different b-values and different set of parameters TE/TM/∆. Each parameter combination, over the 6 tested (TE/TM/∆=33.4/60/64.2 ms and 33.4/250/254.2 ms up to b=60 ms/µm², 73.4/20/64.2 ms and 73.4/60/64.2 ms up to b=40 ms/µm², 73.4/210/254.2 ms and 73.4/250/254.2 ms up to b=30 ms/µm²), was investigated on 4 mice (Fig. 2).

After scan-to-scan phase correction, spectra were analyzed with LCModel. Signal attenuation was quantified for NAA, total creatine (tCr), choline compounds (tCho), myo-inositol (Ins), taurine (Tau) and macromolecules (MM) (using an experimentally measured MM spectrum included in the basis-set). We also extracted metabolites apparent diffusion coefficient (ADC) from a mono-exponential fit between b=0 and 5 ms/µm², and fast and slow diffusing components (ADC_fast, ADC_slow and slow diffusion fraction f_slow) from a bi-exponential fit between b=0 and 30 ms/µm².

To quantify the influence of TE/TM on diffusion, we performed statistical analysis on raw signal attenuation at each b value, as well as on ADC, ADC_fast, ADC_slow and f_slow for water and metabolites, using analysis of variance (ANOVA) and post-hoc analysis.

The impact of TE and TM is extremely low, whatever ∆. Only a few data points exhibit significant dependency on TE/TM (Fig. 3). Most of these points are isolated, suggesting possible type I error. Anyway, signal difference between different TE/TM is less than 5% in most cases, and can therefore be considered negligible (the most notable exception is the ~7% decrease of NAA attenuation at b=20-30 ms/µm² when increasing TE). Results for fitted parameters are consistent with results on raw data, since no significant dependence on TE/TM is found for ADC, ADC_fast, ADC_slow and f_slow for any metabolites (Table 1 and Table 2). Note that results are very different for water, which is consistent with the existence of several water pools with different T₂ properties (e.g. myelin water, cerebrospinal fluid). For macromolecules, TE visually seems to have some impact, but MM data are noisy at high b-values and long TE/TM. We think this dependency is plausible, e.g. if macromolecules consist in a “continuum” of different weights/T₂.

The absence of correlation between metabolites diffusion and relaxation, while a few previous studies did find a correlation, might be explained by several reasons. For example, Assaf and Cohen worked on excised brain [1] and optic nerve [2], which might change diffusion properties (e.g. if metabolites leak in the extracellular space). Another potential explanation might be due to the different kinds of fibers investigated, with [1][3] being performed in highly myelinated optic nerve and Human white matter, assuming that myelin may have strong surface relaxivity.

In this work we measured little or no dependence of metabolite diffusion properties on TE/TM, in a large voxel of the mouse brain in vivo. This strongly supports the interpretation and modeling of metabolite diffusion based on geometry only, at least under these experimental conditions. Potential correlation between macromolecules relaxation and diffusion, and the possible role of myelin in past studies, remain to be explored in deeper details.