Combining 31P-1H Cross Polarization With Magnetization Transfer: A Novel Approach for Myelin Investigation

Abstract

MRI is a crucial tool for studying white matter, which is primarily composed of myelin, a phospholipid-rich sheath surrounding nerve fibers. Myelin damage leads to disrupted neurological function, which is a prominent feature in neurodegenerative diseases like multiple sclerosis. Current MRI techniques for detecting myelin use hydrogen nuclei (¹H) exclusively to generate contrast. Phosphorus (³¹P) is highly concentrated in myelin phospholipids relative to other brain structures. Due to sensitivity of the anisotropic chemical shifts and dipolar couplings to structure and dynamics, ³¹P may provide richer and more specific ways to probe the myelin bilayers. Key experiments aimed at developing MRI compatible probes of myelin ³¹P are demonstrated. First, a solid-state NMR technique, cross polarization (CP) is compared with single pulse excitation in white matter. The ¹H-³¹P CP spectrum retains the morphology sensitive ³¹P powder pattern of the single-pulse spectrum, but lacks the aqueous ³¹P peak. Second, by combining magnetization transfer (MT) with CP, we observe bi-directional polarization exchange between myelin ³¹P and surrounding water, apparently proceeding through a unique ¹H pool that is distinct from the ¹H that typically dominates MT. Pulsed magnetic field gradients are used to isolate magnetization that originates from myelin ³¹P and transferred to aqueous ¹H. This small signal, approximately 1/68,000 that of the ¹H water signal, could offer access to structural and dynamic information from the membrane/water interface not previously available. This two-step transfer process opens new possibilities for understanding myelin and white matter disease and injury. These proof-of-principle findings may have broad implications for both basic neuroscience and clinical imaging. By leveraging ³¹P as a myelin probe, this approach offers a novel tool for studying myelin and could aid in detection and treatment monitoring of white matter disease and injury. Future work will investigate the translation of this technique to MRI.

Keywords: NMR; cross polarization; magnetization transfer; myelin; phosphorus; white matter.

FIGURE 1

(A) Hierarchical view of myelin structure showing (from left to right) an intact neuron, the membrane organization in myelin, lipids within a membrane, a phospholipid molecule, and the phosphorus environment within the phospholipid head group. (B) NMR pulse sequences for single‐pulse excitation and ramped cross‐polarization. (C) Proton‐decoupled ³¹P solid‐state NMR spectra of WM obtained from porcine spinal cord collected with single pulse excitation (green) and cross‐polarization (blue), with the chemical shift tensor labels describing parallel, anti‐parallel and magic angle (iso) orientations relative to B₀.

FIGURE 2

(A) Pictorial representation of the MT‐CP pulse sequence, shown in (B). The sequence begins with a Goldman–Shen T₂ filter during which initially excited magnetization from semi‐solid ¹H quickly decays away due to its short T₂. At the end of the filter, the remaining ¹H_aq magnetization is placed either parallel (red) or anti‐parallel (black) to the static magnetic field. Then a variable magnetization transfer time, τ, allows magnetization to transfer back to the semi‐solid ¹H_s. At the end of the MT time, CP transfers magnetization for detection to the ³¹P head group. (C) Results of this experiment on porcine spinal cord WM. Data points represent integrated ³¹P CP spectra as a function of τ. Points shown in red correspond to experiments having ¹H_aq magnetization aligned parallel to the main magnetic field while black is for anti‐parallel alignment. Solid lines represent fits described in the text. The gray curve represents the difference of the two fit equations.

FIGURE 3

(A) Pictorial explanation of the gCP‐MT pulse sequence, shown in (B). ³¹P magnetization is tipped into the x‐y plane where the encode gradient introduces a magnetization helix. The CP spin lock retains only the projection of the helix along the x axis in the rotating frame and transfers the resulting spatial sinusoidal modulation to ¹H_s. This magnetization is returned to the z axis and is transferred to nearby ¹H_aq via MT, retaining the modulation from the gradient encode. This ¹H_aq magnetization is tipped into the x‐y plane where it is decoded with a second gradient pulse for detection. The area of the decode gradient pulse is smaller by a factor of γ _H/γ _P = ~2.5. A demonstration of the effects of the gradient pulses on the measured ¹H signal is shown in (C). Without either gradient, a signal is observed (orange), but is not reproducible, varying in both amplitude and phase. With only the decode gradient, no signal is observed (blue). With both gradients, a consistent reproducible signal is observed (red).

FIGURE 4

(A) Dependence of the integrated gCP‐MT ¹H_aq signal on the duration of the initial ³¹P excitation pulse. In these experiments the repetition time was 8 s to minimize any T₁ weighting effects. (B) Dependence of the integrated gCP‐MT signal on the magnetization transfer time, τ. The difference of parallel and anti‐parallel MT‐CP data points from Figure 2C are normalized and included in gray to demonstrate the similarity of the cross‐relaxation dynamics in the two experiments.