Purpose

CEST is a new MRI technique that is capable of enhancing the MR sensitivity of dilute metabolites by making use of their exchangeable protons (1). However, a main challenge in CEST is to separate different types of exchanging protons in tissues. For most in vivo studies, conventional line-shape based analysis methods of Z-spectra are not able to separate out the overlapping signals of exchanging protons such as amine, amide, and hydroxyl groups, as well as non-exchangeable aliphatic and aromatic protons (2-4). We used an off-resonance variable delay multiple pulse (VDMP) sequence (5-7) to separate out predominantly fast- (>1 kHz) and slow-exchanging components in Z-spectra.

Methods

The method was verified on solutions of Glutamate (Glu, 50 mM) and Bovine serum albumin (BSA, 10% w) as well as on hair conditioner (MTC phantom), and then applied on mouse brain. The VDMP sequence consisted of a train of Gaussian saturation pulses separated by a mixing time and followed by a RARE acquisition module. Experiments were performed by fixing the pulse number (n=16) and length (10 ms) and increasing the inter-pulse delay for different offsets, thus sensitizing the experiment to the rate of magnetization transfer. For each saturation offset, eight mixing times were collected and the VDMP build-up curves fitted with the coupled Bloch equations for a three-pool model (6), consisting of a water pool, a slow-exchanging pool and a fast-exchanging pool. By fitting the VDMP buildup curves with this model, the concentrations of fast and slow-exchanging components can be obtained. Three parameters were fitted for each offset: the exchange rate k_slow and the concentration x_slow for slow-exchanging protons, plus the concentration x_fast for fast-exchanging protons. An average exchange rate of 2 kHz was assumed for the fast-exchanging protons, considering that the VDMP build-up curve is nearly independent of exchange rate when higher than 1 kHz. Notice that the fast exchange component, x_fast, is also concurrent with the direct water saturation. Offsets between ±1 ppm were excluded in the analysis due to strong interference from direct water saturation. MRI experiments were performed on a horizontal bore 11.7 T Bruker Biospec system.

Results and Discussion

Typical VDMP Z-spectra for the phantoms are plotted in Figs. 1a-d, together with the fast- and slow-exchanging Z-spectra obtained by fitting the VDMP build-up curves (Figs. 1e-h). The different exchange pools can be well separated by this method in BSA solution, such as the amine groups, which show a pattern similar to that of Glu solution in fast-exchanging Z-spectra. The aromatic and amide groups fall in the slow-exchanging pool and can thus be distinguished from the amine group, while only a single broad resonance is seen in conventional Z-spectra (Figs. 1c&d). Fig. 2 shows the full Z-spectra, as well as the Z-spectra for fast- and slow-exchanging protons in the mouse brain recorded using the VDMP sequence with two saturation powers of 5.6 μT and 11.2 μT. The fast-exchanging Z-spectrum shows a constant background (~0.5% and 1% for low and high B1, resp.) in the region studied, i. e., from -8 ppm to 8 ppm, except in the range around water signal. The range of 4 to -2 ppm shows clear asymmetry with respect to the water resonance. Therefore, the region around 0 ppm was assumed to arise mainly from the fast-exchanging protons (FP) from metabolites and mobile proteins, and concurrent with water signal. High-resolution slow- and fast- exchanging proton maps at ±3.6 ppm and 2 ppm are plotted in Fig. 3. The total fast proton (TFP, combined fast-MTC and fast exchanging metabolite protons) map for these off-resonance data shows homogeneous intensity across the brain at the three offsets, which is similar to the total fast-exchanging proton (TFP) map obtained by onVDMP previously (6). The slow-exchanging proton is asymmetric and broad and shows high contrast between the WM and GM in the mouse brain at all frequency offsets measured, which indicates that the slow-exchanging proton pool mainly arises from asymmetric MTC from myelin lipid and macromolecules (7). The complicated asymmetric line-shape of the slow-exchanging pool suggests that caution is required when fitting the Z-spectra using line-shape analysis. In order to apply VDMP method on APT and relayed NOE CEST for mobile proteins, the saturation power has to be low enough to minimize the contribution from the MTC pool (Fig. 1d). Then, the fast-exchanging proton is also not visible, which is the case in the original VDMP study (7).

Conclusion

The current technique can extract the fast- and slow-exchanging protons from Z-spectra, and presents direct evidence that MTC in vivo contains both fast-exchanging and slow-exchanging components.

Acknowledgements

This work was supported by R01EB015032, P41EB015909 and R01EB019934. Lin Chen thanks the China Scholarship Council (201506310130) for financial support.