Introduction

Inhomogeneous magnetization transfer (ihMT), weighted by the dipolar relaxation time T_1D¹, is sensitive to myelination² and to pathological induced alterations³ of myelinated structures. Its high specificity to myelin² is attributed to the mechanisms underlying the proton dipole-dipole interactions along the lipid chain⁴.

Coherent Anti-stoke Raman scattering (CARS) microscopy is a dye-free optical imaging method, that can be tuned to the vibrational modes of CH₂ and CH₃ bonds of myelin lipids, usually used to characterize demyelinating and neurodegenerative models⁵.

In this work we propose a combined ihMT–CARS analysis on a model of oligodendrocyte intoxication with lysophosphatidylcholine (LPC) performed in the spinal cord (SC) of mice.

Methods

LPC mouse model: LPC triggers the selective death of spinal oligodendrocytes, and results in a peak of demyelination 7 days after exposure. Adult Thy1–CFP fluorescent reporter mice were submitted either to focal LPC incubation (LPC, n = 3), or to phosphate-buffered saline (PBS) incubation (Sham, n = 3) for 2 hours following dura matter opening. Incubation was performed in the dorsal tracts of cervical SC. Seven days post incubation spinal cords were extracted and fixed with 4% paraformaldehyde for 12h prior to imaging.

Microscopy Methods: CARS microscopy was done for myelin imaging using OPO femtosecond laser source (80 MHz 150 fs; Coherent, Santa Clara, CA, United States) with excitation wavelengths λ_p = 806 nm and λ_s = 1050 nm. Nonlinear excitation of CFP fluorescence was also optimal at λ_p = 806 nm and allowed axonal density imaging. Figure 1a shows a representative combined image of myelin and axons obtained by the superposition of the 2 contrasts. Average intensity of CARS was determined in rectangle regions of interest (680x333 µm²) along the rostral and caudal axis. Axon density was determined from the average number of axons intersecting the lateral edges of each ROI (100 µm depth). Axon and CARS data were normalized to a reference value, 3000 microns away from the lesion.

MRI Protocol: Experiments were performed on a preclinical 7T scanner (PharmaScan, Bruker). A 3D ihMT-RAGE sequence (ihMT preparation: Hann shaped pulses of pw = 0.5 ms duration; inter-pulse delay Δt = 0. 8 ms; frequency-offset Δf = 10 kHz; number of pulses per burst N_p = 8; burst repetition time BTR = 60 ms; MT pulse duty cycle DC = 6.7%; total saturation time τ = 900 ms and saturation power calculated over BTR, B_1RMS = 6.7 µT; axial readout direction, 73x73x333 µm³ resolution) was used to acquire images centred on the LPC lesion. Quantification was performed in ROIs selected in the SC dorsal tracts. T_1D measurements were performed in a single axial slice (666 µm slice thickness) centred in the lesion using a modified ihMT-RARE sequence (B_1rms = 8 µT; 10 Δt values in the [0.8; 40] ms range) and a bi-T_1D compartment model⁶. In addition, T₁ and T₂ imaging were performed using Variable Flip Angle (VFA; B₁ corrected) and multi spin-echo sequences.

Results

Figures 1b-c show CARS (magenta) and CFP (cyan) coronal images of a sham and an LPC spinal cords. Insets highlight sections in dorsal tracts where normal dense myelinated axons are visible for sham, whereas, for LPC, demyelination, axonal loss, myelin debris and cell infiltration are noticeable. Corresponding ihMTR images (greyscale) presented in Figure 2, show the extent of the demyelinated lesion in both the coronal and axial directions. The rostro-caudal profiles of ihMTR, CARS and CFP signals indicate a signal loss of 34%, 67% and 85 % respectively in the centre of the lesion (Figure 2b), while the PBS-incubated spinal cord presents variations smaller than 5% around the centre of the spinal cord for all metrics (Figure 2a).

Figure 2c, shows coronal ihMTR images at different depths of the dorsal tracts indicating that the lesion propagates deeper than 365 μm. Figure 3 shows correlations between ihMTR, CARS and CFP data normalized. Highest correlation was obtained between CARS and axonal density (r² = 0.96, p < 0.0001). Correlations between ihMTR and axonal density and CARS showed similar strengths (r² = 0.84, p < 0.0001 and r² = 0.86, p < 0.0001 respectively). The quantification of T_1Ds, T₁ and T₂ in the lesion (Figure 4) showed different values than those obtained in the ventral WM taken as reference. Specifically, an increase in both T₁ and T₂, along with a decrease in the fraction of long T_1Ds in the lesion and a shortening of both long and short T_1D components was observed.

Discussion

The rostro-caudal ihMTR profile in the LPC-incubated spinal cord follows the dynamics of demyelination and axonal loss measured by microscopy. Interestingly, the correlation between CARS and ihMTR signals revealed a positive intercept, suggesting that the ihMTR signal does not originate only from myelin, but also from other non-myelin components.

Noteworthy, we observed an increase of T₁ and T₂ and a shortening of long/short T_1D values in line with the damages (cell infiltration, demyelination) observed in the lesion (Figure 1c). CARS data are precious assets to further interpret the signal variations measured with ihMT using advanced biophysical models and quantitative analysis.

Conclusion

IhMT MRI combined with CARS imaging provide a comprehensive characterization of the cellular processes involved in demyelination induced by the LPC model.

Acknowledgements

This work was performed by a laboratory member of France Life Imaging network (grant ANR‐17‐ CE18‐0030, VERISMO project), supported by Carnot Star Institute (CARNOT STAR 2020) and the French Association pour la Recherche sur la Sclérose En Plaques (ARSEP 2020).