Inhomogeneous Magnetization Transfer MRI of White Matter Structures in the Hypomyelinated Shiverer Mouse Brain

Abstract

Purpose:

Inhomogeneous magnetization transfer (ihMT) MRI is uniquely sensitive to myelin with lipids as a primary source of its contrast. In this study, we investigated whether ihMT can detect white matter structures in the hypomyelinated shiverer mouse brain, a model of dysmyelination.

Methods:

Conventional MT and ihMT images were acquired from ex vivo Rag2^−/− control and shiverer mouse brains at 7 Tesla using previously reported optimized saturation parameters.

Results:

ihMT ratio (ihMTR) maps revealed hypomyelinated corpus callosum in the shiverer mouse brain, whereas conventional MT ratio (MTR) maps showed no clear contrast. The ihMTR values of the corpus callosum in the shiverer mice were reduced by approximately 40% compared to controls, but remained significantly higher than the ihMTR values of the cortex.

Conclusion:

The finding further confirms ihMT’s high myelin specificity and suggests its use as a marker to detect early myelination or myelin repair.

Introduction:

In the central nervous system, myelin consists of multiple lamellae of thin extended cell membrane of oligodendrocytes that elaborately wrap around axons and functions as an insulator for efficient signal transduction. While approximately 70% of myelin is lipid (by dry weight), proteins (e.g., myelin basic protein (MBP) and proteolipid protein (PLP)) also play important roles in maintaining myelin structures and functions (1,2).

Given the importance of myelin and several neurological diseases that involve myelin pathology, such as multiple sclerosis (MS), non-invasive techniques for imaging myelin are critical. MRI is the primary tool used in the clinic to detect myelin pathology (3,4). Multiple MRI contrasts, each targets myelin through distinct contrast mechanisms, have been reported to be sensitive to myelin (5–12). Among these contrasts, magnetization transfer (MT) MRI (13) has been widely used to image myelin in the central nervous system (CNS). In a recently conducted meta-analysis, Mancini et al. (14) surveyed 43 reports to compare correlations between an extensive list of MRI contrasts and histological measurements of myelin, and their findings suggest that magnetization transfer (MT) MRI is one of the two MRI contrasts that are most sensitive to myelin.

MT MRI is based on the transfer of magnetization from protons bound to semi-solid macromolecules (e.g. myelin lipids or proteins) to free protons (13,15), and additional intrinsic tissue properties can be extracted from MT signals based on biophysical models (16). Recently, inhomogeneous MT (ihMT) MRI has been introduced and showed potentials in detecting myelin with even higher sensitivity than conventional MT MRI (17). In addition to single frequency irradiation as in conventional MT, ihMT employs dual frequency irradiation to achieve additional signal attenuation through decoupling the dipolar order from the Zeeman order (18,19). Due to its enhanced sensitivity to myelin, ihMT MRI is becoming an attractive contrast option for monitoring myelin pathology (20,21) and white matter maturation (22). Recent evidence of strong correlations between ihMT results and fluorescence signals linked to PLP in the mouse brain further supports the use of ihMT as a myelin marker (23).

Although the exact origin of ihMT signals is still under investigation, membrane lipid, with long chains of methylene, is one of the potential sources (17,24). This may explain the sensitivity of ihMT to myelin, as myelin sheath has a much higher concentration of lipids than other cellular membranes (25). The relatively long dipolar order relaxation time constant (T_1D) of myelinated white matter (~ 6 ms, in vivo mouse brain at 11.75 T) (26) compared to gray matter and muscles allows the selection of long T_1D tissue component via T_1D-filtered ihMT imaging (26), which can further enhance specificity to myelin (23,26). In this study, we examined the sensitivity of ihMT to myelin in the shiverer mouse brain, which lacks normal compact myelin due to a mutation in the MBP gene but still present a small amount of non-compact myelin sheath (27,28). Previous studies of the shiverer mouse using conventional MT, T₂, and diffusion MRI showed no apparent myelin contrast in this model (28). Here, we investigated whether ihMT can detect the residual myelin in the shiverer mouse brain, as the result may provide additional insight into the sensitivity of ihMT to myelin.

Methods:

Ex vivo mouse brain experiment

The animal protocol has been approved by the local institutional animal care and use committees. Rag2^−/− (control) and Rag2^−/−-shiverer mice (3 months old, n = 5 in each group) were perfusion fixed with 4% paraformaldehyde. The Rag2^−/− mouse, which lacks mature lymphocytes and present severe immune deficiency, has been used by us to evaluate myelin repair with transplanted glial restricted precursor cells (28). We chose the 3 month time point because myelination has mostly finished in the control mouse brain at 3 months after birth and the Rag2^−/−-shiverer mouse has a life span of 120 days (27,28). Dissected mouse brains were placed in 10 ml syringe filled with Fomblin (perfluoropolyether; Ausimont USA Inc.), which has no MRI visible proton, to match tissue susceptibility and prevent dehydration.

MRI Acquisition

All samples were imaged using a 7 Tesla MRI system with a quadrature transmit volume coil (70 mm diameter) and a 4-channel receive-only phased array cryocoil (Bruker Biospin, Billerica, MA, USA). The ambient temperature of the specimens was kept at 36–36.5 °C using a heater inside the cryocoil designed to maintain a constant surface temperature. The MT and ihMT images were acquired using a RARE sequence with a train of Gaussian saturation pulses (pulse width (PW) = 1.83 ms, bandwidth=1,500 Hz, repetition time (Δt) = 2.62 ms, number of pulses (N) = 200) with single offset frequency for MT and dual offset frequencies for ihMT by alternating the offset frequency of consecutive pulses. The choice of saturation pulse width and repetition time were made to increase T_1D-filtering, which can enhance myelin specificity (23). Based on the optimized ihMT imaging parameters reported by Prevost et al. (26) for the living mouse brain, two sets of data were acquired from ex vivo mouse brains (Table 1) to further examine the effects of saturation pulse amplitude (Exp#1) and offset frequency (Exp#2). Another set of data was acquired with varying time delays between saturation pulses (Exp#3), but the results are only included in the supplementary materials (Table S1) because the B_1,rms values were not kept constant. General imaging parameters were: echo time (TE)/repetition time (TR) = 6 ms/4000 ms, 2 signal averages, echo train length = 6 with center-out phase encoding, a slice thickness of 1.25 mm, and an in-plane resolution of 0.125 × 0.125 mm². Images with no saturation pulse (M₀) were acquired with images with positive/negative or dual frequency saturation pulses (M₊, M₋, and M_±). MTR and ihMTR maps were calculated as: $MTR=1-{\frac{M_{+}}{M}}_0$, ihMTR = ((M₊ + M₋) – 2M_±)/M₀. Co-registered diffusion tensor images were acquired using a diffusion-weighted echo planar imaging (DW-EPI) sequence with the following parameters: TE/TR = 28/4000 ms, 1 signal average, 4 segments, diffusion gradient duration = 5 ms, diffusion time = 15 ms, 30 diffusion encoding directions, b=2000 s/mm², four non-diffusion-weighed images, FOV = 15 × 15 mm, matrix size = 128 × 128, slice thickness = 1.25 mm and an in-plane resolution = 0.125 × 0.125 mm². Co-registered T₂-weighted images were acquired using a RARE sequence with the following parameters TE/TR = 50/3000 ms, 2 signal average, echo train length = 8, the same FOV, matrix size, and slice thickness as the diffusion tensor images.

Table 1:

Parameters of the RF irradiation pulses used for ihMT MRI, including timing, amplitude, and offset frequency (f).

Exp#1	PW = 1.82 ms, Δt =2.62 ms, f = 12 kHz, N=200
	B1,peak (μT)	4	8	12	16
	B1,rms (μT)	1.79	3.58	5.37	7.16
Exp#2	PW = 1.82 ms, Δt =2.62 ms, B1,peak / B1,rms = 12/5.37 μT N=200
	f (kHz)	4	8	12	16	20	24

Abbreviations are: B_1,peak/ B_1,rms: peak/root mean square amplitude of the saturation pulse; PW: pulse width; Δt: pulse repetition time; f: offset frequency; N: number of pulses.

Image Analysis

The images acquired using the DW-EPI sequence were reconstructed from raw data on the scanner console with trajectory correction (Paravision 6.0.1, Bruker Biospin, Billerica, MA, USA) (29,30) and then uploaded to a workstation for rigid motion correction using DTI Studio (http://www.mristudio.org). Diffusion tensor was calculated at each pixel along with the mean diffusivity (MD) and fractional anisotropy (FA) using MRtrix (www.mrtrix.org). Regions of interest (ROIs) for the corpus callosum (CC), cortex (CX), cerebral peduncle (CP), and trigeminal nerve (TN) were manually defined using ROIEditor (http://www.mristudio.org) in the FA image, due to its superior contrast for white matter structures even in hypomyelinated Rag2^−/−-shiverer mouse brains. ROIs for the muscle region was manually defined in the M₀ images. For each subject, average MTR and ihMTR values of each ROI were extracted from the MTR and ihMTR maps.

Contrast to noise ratio (CNR) in the ihMTR/MTR maps were calculated by dividing the difference between CC and CX signals by the standard deviation of signals in the muscle region instead of the background, in which noise was magnified by the calculation of ihMTR and MTR. However, differences in muscle tissue properties, both within the defined ROIs and between subjects, also contributed to signal variations and increased the estimated noise level.

Within each group (control or Rag2^−/−-shiverer), comparisons of ihMTR/MTR values between different experimental conditions were performed using the paired Student’s t-test. Comparisons between the control and Rag2^−/−-shiverer groups were performed using Student’s t-test.

Transmission electron microscopy

Brain sections were immersed in a solution of 0.5% toluidine blue in 1 % sodium borate, washed with distilled water, dehydrated with ethanol, and dried on a hot plate at 60°C. Coronal sections of the corpus callosum were postfixed in OsO4 and embedded in Epon. Thin sections of 70 nm were stained with citrate/uranyl acetate, and images were acquired using a Zeiss Libra transmission electron microscope as described previously (28).

Results:

T₂ and diffusion MRI of the shiverer and control mouse brains

T₂-weighted, MD, and FA images of the control mouse brain (Fig. 1) provided distinct contrasts to separate the corpus callosum and other white matter structures from gray matter structures, e.g. the cortex. In comparison, T₂-weighted and MD images of the Rag2^−/−-shiverer mouse brain lacked sufficient white and gray matter contrasts, but FA images of the Rag2^−/−-shiverer mouse brain still revealed white matter structures as their contrasts reflected the restrictive effects of axonal membranes and their anisotropic arrangement. Based on the FA images, the corpus callosum can be delineated reliably in both control and Rag2^−/−-shiverer mouse brains.

Fig. 1:

Representative T2-weighted, MD [mm²/s], and FA images of ex vivo control and Rag2^−/−-shiverer mouse brains. The blue, purple, and white arrows indicate the locations of the cortex, corpus callosum, and muscle.

ihMT MRI of control mouse brains

Fig. 2 compares images of an ex vivo control mouse brain acquired with different saturation pulse amplitudes and offset frequencies, which modulated the contrasts between myelin-rich white matter structures such as the CC and gray matter such as the CX. Within the range of pulse amplitudes examined here, MTR and ihMTR values of gray and white matter structures increased with the amplitude of the saturation pulse (Fig. 3A–B). With a B_1,rms of 5.37 μT, the ihMTR values of the CC peaked between offset frequencies of 8 to 12 kHz (Fig. 3E), while MTR values decreased as the offset frequency increased from 4 kHz to 24 kHz (Fig. 3F). The TN and CP showed similar trends as observed in the CC. For all the conditions examined in this study, the contrasts between CC and CX in ihMTR were significantly higher than in MTR (Fig. 3C, 3G). In Fig. 3D, the contrast to noise ratio (CNR) between CC and CX in ihMTR maps were mostly higher than in MTR maps except for B_1,rms of 1.79 μT. Fig. 3H shows that the CNR between CC and CX in ihMTR maps was initially lower than in MTR maps at 4 kHz offset frequency but became comparable at 8–12 kHz and remained comparable at 16–24 kHz. Results from Exp#3 on control mouse brains can be found in the supplementary material (Fig. S1).

Fig. 2:

ihMTR maps of an ex vivo control mouse brain at different offset frequencies (the horizontal axis) and pulse amplitude (the vertical axis). The blue, purple, red, and green arrows indicate the corpus callosum, cortex, cerebral peduncle, and trigemeninal nerves, respectively.

Fig. 3:

Changes in ihMTR (A and E) and MTR (B and F) values of the corpus callosum (CC), cerebral peduncle (CP), trigeminal nerve (TN), and cortex (CX) in ex vivo control mouse brain (n=5) with amplitude (Exp#1, left column) and offset frequency (Exp#2, right column) of the saturation pulses. C and G: Contrasts between CC and CX under each condition. D and H: CNR between CC and CX under each condition. **/***/****: p<0.01/0.001/0.0001 by Student’s t-test.

ihMT reveals the CC in Rag2^−/−-shiverer mouse brains

We were able to visually distinguish the CC from the CX in the ihMTR maps of Rag2^−/−-shiverer mouse brains (Fig. 4) acquired with 12kHz offset frequency and B1,rms of 5.37 μT. The locations of the CC were confirmed by comparing the ihMTR results with co-registered FA maps. At the lowest frequency of 4kHz, only the trigeminal nerve (TN) was visible as it is a part of the peripheral nervous system and still has normal myelin not affected by the MBP mutation. Increasing the saturation pulse amplitude enhanced the overall signal.

Fig. 4:

ihMTR maps of an ex vivo Rag2^−/−-shiverer mouse brain at different offset frequencies (the horizontal axis) and pulse amplitude (the vertical axis). The blue, purple, red, and green arrows indicate the corpus callosum, cortex, cerebral peduncle, and trigemeninal nerves, respectively. In the ihMTR map acquired with 12kHz offset frequency and B1,rms of 5.37 μT, the region containing the CC was enlarged and brightened to visualize the subtle contrast between CC and CX.

Fig. 5 shows that the ihMTR values of the CC and CP of the Rag2^−/−-shiverer mouse brains were lower than in the control mouse brains but remained slightly higher than the CX (Table 2), and the contrasts were highest with offset frequencies between 8 and 12 kHz. The ihMTR values of the TN in the Rag2^−/−-shiverer mouse brains were comparable to the control mouse brains. In comparison, MTR values of the CC and CP were similar to or lower than the CX in the Rag2^−/−-shiverer mouse brains for most of the conditions tested here. For most of the conditions examined in the Rag2^−/−-shiverer mouse brains, the contrasts between CC and CX in ihMTR were significantly higher than in MTR (Fig. 5C, 5G). The CNR between CC and CX in ihMTR maps were significantly higher than in MTR maps for B_1,rms greater or equal than 3.58 uT with an offset frequency of 12 kHz (Fig. 5D) and for offset frequency between 12 and 16 kHz with a B_1,rms of 5.37 μT (Fig. 5H). Results from Exp#3 on Rag2^−/−-shiverer mouse brains can be found in supplementary data (Fig. S2).

Fig. 5:

Changes in ihMTR (A and E) and MTR (B and F) values of the corpus callosum (CC), cerebral peduncle (CP), trigeminal nerve (TN), and cortex (CX) in Rag2^−/−-shiverer mouse brain (n=5) with amplitude (Exp#1, left column) and offset frequency (Exp#2, right column) of the saturation pulses. C and G: Contrasts between CC and CX under each condition. D and H: CNRs between CC and CX under each condition. */**/***: p<0.05/0.01/0.001 by Student’s t-test.

Table 2:

MTR and ihMTR values of four structures in the control and Rag2^−/−-shiverer mouse brain.

	ihMTR (12 kHz)		MTR (4 kHz)
	Control	Shiverer	Control	Shiverer
Corpus Callosum	0.071 ± 0.007 (****)	0.044 ± 0.007 (*)	0.64 ± 0.022 (**)	0.55 ± 0.032 (ns)
Cerebral Peduncle	0.104 ± 0.061 (****)	0.059 ± 0.045 (***)	0.67 ± 0.025 (***)	0.54 ± 0.015 (ns)
Trigeminal Nerve	0.130 ± 0.065 (****)	0.120 ± 0.009 (****)	0.67 ± 0.018 (***)	0.63 ± 0.067 (*)
Cortex	0.041 ± 0.046	0.039 ± 0.006	0.57 ± 0.028	0.55 ± 0.019

Values in the parentheses indicate the results of paired t-test between the ihMTR/MTR values of each structure and cortex. Both MTR and ihMTR detected significant differences in CC and CP between control and Rag2^−/−-shiverer mouse brains (t-test, p < 0.001). NS denotes not significant difference detected, and */**/***/**** indicate p<0.05/0.01/0.001/0.0001.

Comparisons of the CC in control and shiverer mouse brains

Comparisons with T₂-weighted, ihMTR, and MTR images of the control and Rag2^−/−-shiverer mouse brains showed that the CC in the Rag2^−/−-shiverer mouse brain can only be detected in ihMTR map (Fig. 6A) (Table 2). The CP in the Rag2^−/−-shiverer mouse brain was also visible in the ihMTR map, but appeared less clearly defined probably due to a lower signal-to-noise ratio at the ventral part of the brain. In comparison, the TN in both Rag2^−/−-shiverer and control mouse brains can be consistently detected. In term of contrasts between CC and CX, while both MTR and ihMTR detected significant differences between control and Rag2^−/−-shiverer mouse brains, ihMTR provided better contrast than MTR (Fig. 6B). ihMTR showed significant differences between CC and CX in the Rag2^−/−-shiverer mouse brains, whereas MTR showed no significant difference (Table 2). Fig. 6C shows normal compact myelin in a control mouse brain and lose noncompact myelin in a Rag2^−/−-shiverer mouse brain.

Fig. 6:

Comparisons of ihMT and MT signals in the control and Rag2^−/−-shiverer mouse brains. A: Representative T₂-weighted (T₂WT), ihMTR, and MTR images showing the corpus callosum (CC, blue arrow), cerebral peduncle (CP, red arrow), trigeminal nerve (TN, green arrow), and cortex (CX). Small inhomogeneities in the hippocampus (yellow arrows) can be seen in MTR maps of both control and shiverer mouse brain. (B) Contrasts between CC and CX in ihMTR and MTR in control and shiverer mouse brains (offset frequency = 12kHz and B_1,rms = 5.37 μT). C: Transmission electron microscopy images of the CC in control and Rag2^−/−-shiverer mouse brains. ***/****: p<0.001/0.0001 by Student’s t-test.

Discussion:

In this study, we mostly followed the optimized ihMT parameters established in previous reports (19,23,26) to detect white matter structures in the hypomyelinated Rag2^−/−-shiverer mouse brain. The peak ihMT contrast between CC and CX was observed at offset frequency between 8 and 12 kHz, which was inline with 10 kHz for in vivo mouse brains (26) but was higher than previously reported optimal frequency of 7 kHz for human brains (19,31), due to lower B_1,rms used for human imaging (19). The timing of saturation pulses used in this study resulted in strong T_1D weighting as previous reports have demonstrated that strong T_1D weighting improves specificity to myelin (23,26). It is important to note that the temperature of specimen was maintained at 36–36.5 °C because T_1D-filtered ihMT is temperature dependent and imaging at room temperature will lower dynamic range and myelin specificity (32,33).

The finding that ihMT MRI detected significant differences between the CC and CX in the Rag2^−/−-shiverer mouse brain, whereas conventional MT MRI showed no clear contrast, is important for sensitive detection of myelin injury and repair in the brain. Although multiple reports have demonstrated that T₂, susceptibility, diffusion, and MT MRI can detect significant differences between Rag2^−/−-shiverer and control mouse brains (34–38), only anisotropy-based contrasts from diffusion MRI have been reported to distinguish white and gray matter in the hypomyelinated Rag2^−/−-shiverer mouse brain as shown in Fig. 1 and (36,38), probably due to the restrictive effects of axonal membranes. As inflammation is not a pathological component in the Rag2^−/−-shiverer model, one possible explanation of the lack of white matter contrast is that the remaining non-functional myelin in the Rag2^−/−- shiverer mouse brain lacks the physical and chemical properties of mature myelin or sufficient quantity for reliable detection using relaxation and conventional MT-based MRI methods. It is, however, necessary to note that the time it took to acquire the ihMTR maps in our experiments was 2.5 times of that for the MTR maps, which gave the MTR results a SNR penalty when compared to ihMTR results.

This points to a potential obstacle in using existing MRI methods to detect early myelination or myelin repair. For example, during early myelination, compaction of new myelin membranes usually starts after there are three layers of myelin membrane wrapping an axon (39), and therefore, a large number of unmyelinated axons or axons with non-compact myelin may be found at this stage and will not be picked up by relaxation and MT MRI. On the other hand, diffusion MRI based markers may also have difficulties in detecting new myelin due to lower sensitivity and specificity than relaxation and MT-based methods (14). Indeed, we previously reported that T₂, conventional MT, and diffusion MRI detected myelin repair by transplanted glial precursor cells in the Rag2^−/−-shiverer mouse brain but only after a substantial delay after robust myelination demonstrated by histology (28).

The enhanced specificity of ihMT to myelin, as shown by Duhamel et al. (23), can potentially assist non-invasive monitoring of early myelination or myelin repair, and this is further supported by our results. The potential lipid origin of ihMT signals is well-suited for detecting early myelination, in which lipids play a critical role (40). In addition, the larger relative differences in ihMTR between normal and dysmyelinated CC than in MTR (Table 2) may permit more nuanced assessment of the progression of myelination than conventional MT MRI.

An important limitation of the study is that our experiments were performed on ex vivo specimens, which have altered tissue physical and chemical environment due to death and chemical fixation compared to in vivo. Previous studies comparing in vivo and ex vivo MT-MRI experiments reported increases in bound pool fraction in addition to changes in tissue T₁, T₂, and diffusivity values (41,42), and these changes are likely to affect ihMT-MRI results as well. Future in vivo experiments are necessary to confirm our findings here. Another limitation is that we did not explore how varying T_1D weighting of the RF saturation scheme, as described in (23), affects our ability to detect residual myelin in the shiverer mouse brains. Optimized RF saturation schemes will benefit future studies in this area.

In summary, we report that ihMT-MRI can reveal hypomyelinated white matter structures with non-compact myelin in the Rag2^−/−-shiverer mouse brains. The finding suggests that ihMT may be used as a marker to detect early myelination or myelin repair.