By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Wiley Online Library will be unavailable on Saturday 7th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 08.00 EDT / 13.00 BST / 17:30 IST / 20.00 SGT and Sunday 8th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 06.00 EDT / 11.00 BST / 15:30 IST / 18.00 SGT for essential maintenance. Apologies for the inconvenience.
Magnetic resonance spectroscopy (MRS) is indispensible for investigators who want to noninvasively observe metabolites in humans. In the past decades valuable information has been obtained in MRS studies. Among many nuclear magnetic resonance active spins, 1H is the most used nucleus because of its higher sensitivity than any other. However, in the early developing stages of MRS, 31P was more popular than 1H (1). The relationship between choline-containing compounds and cancer was first discovered in 31P MRS (2). Even today, 31P MRS has its unique value in measuring phosphorus monoesters (PMEs) and phosphorus diesters (PDEs) in tissues, because 31P spectra are less crowded than 1H spectra and some metabolites like phosphoethanolamine (PE) and glycerophosphoethanolamine (GPE), which are not easily distinguished in 1H MRS, can be directly observed in 31P MRS. Consequently, 31P MRS is still very popular, both in clinical applications and in laboratory investigations (3–8). Whereas direct observation of 31P is usually time consuming because of its low sensitivity, borrowing magnetization from 1H reservoir through polarization transfer is a convenient way of enhancing the sensitivity of 31P MRS. Many polarization techniques, such as nuclear Overhauser effect (9), refocused insensitive nuclei enhanced by polarization transfer (rINEPT; 10), 31P-edited 1H MRS (11), and cross polarization (12) have been tested. Investigations have found that rINEPT is the most sensitive method among these polarization transfer techniques (13). Thus, attention has been paid to this method and improvement has been suggested. As in the PME and PDE molecules, which are mainly phosphocholine (PC), glycerophosphocholine (GPC), PE, and GPE, the homonuclear couplings between the two groups of methylene protons, i.e., CH2O and NCH2, can seriously attenuate the rINEPT signals, strategies have been proposed to refocus the homonuclear couplings during INEPT evolution of the CH2O proton magnetization. Mohebbi and Gonen (14) noticed for the first time the so-called JHH′ losses in rINEPT.
Heteronuclear J couplings between 31P and the methylene protons in the CH2O groups in PME and PDE (Fig. 1), though rather weak (JHP ∼ 6–7 Hz; Ref.15), form the basis for polarization transfer, whereas the homonuclear couplings between the CH2O protons and the NCH2 protons diverge the CH2O magnetization and cause JHH′ losses. For convenience, the molecular structures and the spin systems for PME and PDE are presented in Fig. 1, where the homonulear coupling is designated as JHM. Mohebbi and Gonen proposed to recover JHH′ losses by applying spin lock on the NCH2 magnetization during INEPT evolution. Experiments on a phantom sample showed that selective spin locking can enhance the detection by ∼50%. Years later, Klomp et al. (16) proposed to insert two selective refocusing pulses in the INEPT period and named the modified pulse sequence selectively refocused insensitive nuclei enhanced polarization transfer (srINEPT). They also obtained ∼50% enhancement, similar to the results of Mohebbi and Gonen (14). Technically, the selective refocusing method is more easily implemented than spin-locking method, because calibrating the selective refocusing pulse length and field strength is easier than calibrating the spin locking field strength alone. Hence, the srINEPT method has aroused immediate attention and has been used in MRSI (16–18).
However, the srINEPT can be further improved. Very recently, Klomp and coworkers. (19) proposed a new version of the srINEPT pulse sequence, where the pulse train “soft π-hard π-soft π” in srINEPT pulse sequence is replaced with a single soft π pulse. Whereas the two soft π pulses in the pulse train selectively invert the NCH2 magnetizations to decouple the homonuclear couplings, the single π pulse in the new version selectively inverts the CH2O magnetizations, also to decouple the homonuclear couplings. However, Klomp and coworkers (19) did not provide detailed information about this simplification. In this study, the simplified pulse sequence is compared with the original one. It is found that when the pulse sequence is simplified, protons coupled to the 31P nucleus, but not selectively inverted by the single π pulse, do not contribute to the final spectrum, enabling a better selective filtering of the spectrum. Meanwhile, different shaped (or soft) pulses for selective refocusing have been compared and the ReBurp (20) pulse is recommended for srINEPT experiments.
Two pulse sequences are discussed in this section, which are srINEPT-9 and srINEPT-7, as presented in Fig. 2a,b, respectively. The purpose of introducing soft π pulses into the rINEPT pulse sequence is to remove the homonuclear coupling between CH2O and NCH2 protons (see the structure in Fig. 1) in PME and PDE (16), so that the detection sensitivity can be enhanced. It is convenient to insert two soft π pulses into the echo time (TE1) period to selectively invert the NCH2 resonances, as has been done in the experiments of srINEPT-9 (16) and also in the 31P-edited 1H MRS (11). No doubt at the end of TE1 in srINEPT-9, the JHH′ effect is refocused and the JHP effect is retained. However, there is a simpler way to reach the same goal, which is srINEPT-7, where the pulse train soft π-hard π-soft π in srINEPT-9 is replaced with a single soft π pulse, but the soft π pulse is to selectively invert the CH2O resonances. This is possible since the CH2O protons are the only protons J-coupled to the 31P nucleus and therefore the only protons contributing to the polarization transfer.
A quick analysis can explain how srINEPT-7 works. For simplicity, the two protons in NCH2 are regarded as a pseudo 1/2 spin and are referred to as Spin M, and the two protons in CH2O are referred to a pseudo spin-1/2 H. As H is coupled to both Spin P (i.e., 31P) and Spin M (see Fig. 1) after the excitation by the first nonselective hard (π/2)y pulse, an inphase coherence Hx is created, which contains four components: Hx(++), Hx(+−), Hx(−+), and Hx(−−), where the plus and minus symbols in the parentheses denote the spin status of P (+1/2 or −1/2) and M (+1/2 or −1/2). During the evolution due to homonuclear and heteronuclear couplings, an antiphase coherence of Spin H, 4HyPzMz, develops (21). The evolution of individual components can be described by
where H0 denotes equilibrium magnetization of spin H and δ denotes the chemical shift (in units of radian). Notice that the signs associated with the J couplings correspond to the spin status. At the time point, t = TE1/2 when only Spin H receives a π pulse applied along the x axis, all four components rotate about the x axis by 180°. Then we have
The difference between Eqs. 1 and 2 is that the first sign in the exp function is changed from a minus to a plus.
Since at this time point, spin P also receives a π pulse, the spin status of 31P in the four components must exchange correspondingly, i.e., Hx(++) exchanges with Hx(−+), and Hx(+−) exchanges with Hx(−−), leading to
In the following evolution, all components rotate in the xy-plane in their individual directions according to their spin status.
When t = TE1, we have
Now the four components merge into two components, (H0/2)exp(+iπJHPTE1) and (H0/2)exp(−iπJHPTE1), which do not contain chemical shift information any more, nor information about JHM; but the JHP information is still retained. In other words, at the time point of t = TE1 in the experiment of srINEPT-7, the chemical shift and JHM effects are refocused while the JHP effect is kept, irrespective of how long TE1 is. This is exactly what we want in the srINEPT-9 experiments. When the experiment is set up with TE1 = 1/2JHP, the two components become 180° out of phase, which can be described by the antiphase coherence 2HyPz.
In the above analysis, the selective π pulse was assumed to be applied along the x-axis. The same results will be obtained if a y-pulse is assumed. Although only a simplified H–P–M three spin system is discussed here, the results should be applied to the real NCH2CH2OPO3 molecular system in PME and PDE. Experiments on milk will confirm this prediction.
MATERIALS AND METHODS
The sample used in this study was vitamin D milk (whole milk, Friendly Farms, IL) purchased from local food market and introduced into nuclear magnetic resonance tubes without adding any lock substance. Sample temperature was kept at 4°C and milk was well within the “sell-by” date.
All nuclear magnetic resonance experiments were performed on a Bruker Avance 600 spectrometer with a QXI (1H13C15N31P) probe. The third channel was used for 31P and the 31P π/2 pulse was 22 μs. The B0 field was preshimmed with D2O. Then the lock power and lock gain were turned to the lowest values and the lock sweep was turned off during data acquisition.
In srINEPT experiments, pulse repetition time = 2 s. Take into account the relaxation effects, TE1 = 60 ms and TE2 = 32 ms, being somewhat shorter than 1/2JHP (77 ms) and 1/4JHP (38 ms), respectively. Before the inversion bandwidth was determined, the B1 field strength of the shaped pulse for 180° rotation was calibrated in the same way as the hard pulse. During free induction decay acquisition, proton decoupling was applied using WALTZ-16 composite pulse with rather weak power (1560 Hz) to avoid possible heating effect. Decoupling was applied only during acquisition; hence nuclear Overhauser effect was not attempted. The parameters for the inversion pulses will be given in the corresponding figure captions, together with other specific acquisition parameters.
Relaxation times of 31P were measured using inversion recovery for T1 and spin-echo for T2. Relaxation times of 1H were measured using srINEPT-7 preceded by inversion recovery for T1 and preceded by spin-echo for T2. Detailed description of the relaxation time measurements needs more space and will be published elsewhere.
Bruker software “Shape Tool” was used for simulating the inversion profile of the soft π pulses.
Calibration of the Selective Inversion Pulse
Both srINEPT-9 and srINEPT-7 pulse sequences contain shaped pulses that are used for selective refocusing. To completely remove the homonuclear coupling during the TE1 period, the selective π pulse should be able to invert resonances over a precalculated band of frequencies. This is illustrated in Fig. 3, where the 1H spectrum of milk is shown in the region between 5.5 and 3.0 ppm, which is in agreement with the spectrum reported in Ref.22. According to the chemical shift data available in the literature (15, 16, 23), the NCH2 resonances span a range of 0.44 ppm with PE and GPE around 3.22 ppm, PC at 3.64 ppm, and GPC at 3.66 ppm. The CH2O resonances of the four molecules are at 3.98 (PE), 4.12 (GPE), 4.18 (PC), and 4.32 ppm (GPC), respectively, spanning a slightly narrow range of 0.34 ppm. A soft π pulse for srINEPT-9 should be carefully chosen with an inversion profile not narrower than 0.44 ppm but not wider than 1.10 ppm and centered at the NCH2 resonances around 3.44 ppm. For srINEPT-7, the inversion profile should be: not narrower than 0.34 ppm, not wider than 1.00 ppm and centered at the CH2O resonances at 4.15 ppm. Currently available selective inversion pulses can be optimized to achieve these selectivity ranges. However, not all shapes of selective pulses generate a rectangular frequency band.
Theoretically, the gaussian pulse does not generate a rectangular frequency band, because the Fourier transformation of a gaussian function is also a gaussian function, but gaussian cascade-3 (G3) pulse are theoretically able to generate a flat inversion band (24), similarly to Sinc pulse (25), and ReBurp pulse (20). The inversion profiles of the four shaped π pulses (Reburp, G-3, Sinc, and gaussian) were simulated using the Bruker software Shape Tool as shown in Fig. 4, from which it can be seen that ReBurp and G-3 are better than Sinc and gaussian, because the inversion bands generated by Reburp and G-3 are flat. However, in experiments it is found that ReBurp pulse is more suitable for band inversion than G-3. To show the quality of the ReBurp pulse, in Fig. 5 the inversion profiles generated by the four shaped pulses are compared, which were recorded by using the pulse sequence in Fig. 2c. In experiments the lipid signal in milk at 1.29 ppm was chosen for demonstration of the inversion profile and the offset of the shaped pulse varied from 2.09 to 0.49 ppm. The profiles in Fig. 5 show that the ReBurp pulse can evenly invert the resonances in a range of 0.45 ppm and leave the resonances ±0.3 ppm farther away not affected, better than the other three shaped pulses. It should be pointed out that according to the calculation in Shape Tool, 16 ms ReBurp pulse would result in an inversion band of 0.60 ppm (360 Hz, see Fig. 4). However, practically only 0.45 ppm effective inversion band was obtained.
Comparing srINEPT-7 with srINEPT-9 and Pulse-and-Acquire
By using the ReBurp pulse for selective refocusing, the srINEPT-9 and srINEPT-7 pulse sequences were tested on a milk sample. Figure 6 shows the results in (b) and (c) and compares them with the pulse-and-acquire spectrum in (a). The GPC, GPE, and PC signals can be immediately identified. The PE peak was missing, probably because the PE concentration is very low in milk. In the pulse-and-acquire spectrum (Fig. 6a) there are many peaks other than PME and PDE. The srINEPT-9 spectrum (Fig. 6b) filters out most of them but still leaves some surviving. These surviving peaks, however, do not exist in the srINEPT-7 spectrum (Fig. 6c). The srINEPT-7 pulse sequence shows better filter effect, because it selectively detects those 31P spins that are coupled to protons appearing in the range from 3.98 to 4.32 ppm. 31P signals other than PME and PDE, coupled to protons outside this range, appear in the srINEPT-9 spectrum (Fig. 6b) but are completely filtered out in the srINEPT-7 spectrum (Fig. 6c), because their coupled protons do not experience the proton inversion (soft) pulse necessary for the INEPT polarization transfer. The srINEPT-9 pulse sequence has worse filter effect, because it nonselectively detects all 31P spins that are coupled to protons, and only selectively removes the homonuclear couplings from those protons appearing in the range from 3.22 to 3.66 ppm to enhance the PME and PDE peaks. Those 31P signals other than PME and PDE can still form antiphase coherences with their attached protons, because the 1H π pulse in the middle of the TE1 is nonselective.
One could see slightly enhancement in the srINEPT-7 spectrum. Although theoretically the srINEPT-9 and srINEPT-7 predict the same sensitivity, in practice the srINEPT-7 pulse sequence would prevent some losses which could occur in the srINEPT-9 experiment. In srINEPT-9, the shaped pulse was applied two times. If the shaped pulse has any imperfection, the effect can be accumulated. However, this accumulation effect does not exist in srINEPT-7 experiment, because the shaped pulse is applied only once.
Test srINEPT-7 Under Different Conditions
In experiments, mistuned parameters would also cause signal losses if the pulse sequence is not robust. For a soft π pulse there are three important parameters: the pulse length which determines the inversion bandwidth and affects the rotation angle, the offset which determines the inversion range, and the B1 field strength which determines the rotation angle in inversion and affects the inversion bandwidth.
The srINEPT-7 experiment proved relatively insensitive to change in these parameters in certain ranges. The spectra with varied pulse length, offset, and B1 field attenuation factor are presented in Fig. 7a–c, respectively. In Fig. 7a, where the offset (4.17 ppm) and the B1 field strength (400 Hz) were fixed but the pulse length was changed from 14 to 18 ms, only insignificant changes in signal intensity were observed. The pulse lengths of 14, 15, 16, 17, and 18 ms correspond to inversion bandwidths of 0.69, 0.64, 0.60, 0.58, and 0.54 ppm, respectively, according to the simulation in Bruker Shape Tool software. By taking into account the ∼0.15 ppm difference between the calculated and the effective inversion band width (Fig. 5a), these inversion bandwidths can still cover the CH2O resonances region (0.34 ppm from 3.98 to 4.32 ppm). From Fig. 7a it can be seen that ±2 ms change in the pulse length of the selective 1H π pulse resulted in 11–15% loss in the 31P signal intensity. When there was ±1 ms change in 1H π pulse length, the 31P signal intensity changed only by 4–5%.
The spectra shown in Fig. 7b were recorded with a fixed pulse length (16 ms) and a fixed B1 field strength (400 Hz) while the pulse offset was changed from 4.33 ppm to 4.01 ppm. The theoretical inversion bandwidth of the ReBurp pulse with 16 ms pulse length is 0.60 ppm (calculated using Shape Tool), but in practice the inversion bandwidth is smaller (0.45 ppm, see Fig. 5a). When the offset was shifted to the left or to the right by 0.08 ppm, the spectrum was nearly not affected with the signal intensity changed only by 1–2%; but when the offset was further shifted by ±0.08 ppm, the signals become weaker. The low-field 0.16 ppm shift of the offset seriously affected the GPE peak, which is understandable because the CH2O protons at 4.12 ppm would have not been effectively inverted. The up-field 0.16 ppm shift of the offset seriously affected the GPC peak, which is also understandable because the CH2O protons at 4.32 ppm would have not been effectively inverted.
The spectra shown in Fig. 7c were recorded with a fixed pulse length (16 ms) and a fixed offset (4.17 ppm) while the attenuation factor of the B1 field was changed. In experiments, 400 Hz B1 field strength was achieved with 37.53 dB attenuation. When the attenuation factor was changed with ±0.5 dB, only less than 2% change in 31P signal intensity could be appreciated. When the attenuation factor was changed with ±1 dB, the 31P signal intensity was changed by 1–5%.
Relaxation Time Measurements
In milk at 4°C and at 14.1 T for PC, GPC, and GPE, respectively, T2H has been found to be around 120 ms (104, 143, and 122 ms), T2P has been found to be about 200 ms (195, 212, and 204 ms), and both T1H and T1P have been found to be around 1 s (0.99, 1.13, and 1.20 s for T1H and 0.92, 0.84, and 1.23 s for T1P).
Compared to the in vivo T1 and T2 data of PC, GPC, and GPE in human brain at 3 T (16), the T1 and T2 data of milk at 14 T show that T1H's and T1P's are smaller, T2P for PC is smaller, whereas T2P's for GPC and GPE are larger. These differences may be attributed to the combination of many factors, such as field strength, temperature, molecular size, and microenvironmental viscosity.
It can be seen from Fig. 6 that only less than 30% sensitivity enhancement of srINEPT (either −9 or −7 version) has been observed as compared to pulse-and-acquire, whereas sensitivity enhancement of srINEPT is potentially 2.47-fold and experimentally observed in many low-field other studies (16). This is mostly explained by relaxation effect at high fields. It is known that as the magnetic field is increased, the T2 relaxation times become shorter (26) while T1's increase for protons and might decrease for phosphorus spins. As pointed out by Mancini et al. (13), the sensitivity enhancement factor η is dependent on relaxation times through the T2 factor f(T2) = exp(−TE1/T2H − TE2/T2P) and the T1 factor f(T1) = [1−exp(−pulse repetition time/T1H)]/[1−exp(−pulse repetition time/T1P)]. By use of the average relaxation times (T1H, T1P = 1 s, T2H = 200 ms, and T2P = 200 ms), we have f(T2) = 0.52 and f(T1) = 1. Then, the theoretical enhancement factor of 2.47 is reduced to 1.28. By taking into account other factors (like A factor in Ref.13), it is understandable that 2.47-fold enhancement was not observed in srINEPT experiments for milk at 4°C.
Feasibility on MRI Systems
The feasibility of the srINEPT-7 experiments on a 7 T clinic MRI system has been demonstrated in vitro by Klomp and coworkers (19). At lower fields, the selective pulses have narrower resonance bands (in Hz) and therefore the selective pulse length should be longer. It is generally not difficult to use a longer pulse, and TE1 (60 ms in this study) is long enough to allow for the single selective π pulse in srINEPT-7. However, 60 ms may be too short for the pulse train soft-π-hard-π-soft-π in srINEPT-9, if the selective π pulse needed to be close to 30 ms. Longer pulses might also result in larger power deposition. From these points of view, srINEPT-7 is more easily implemented. There could be a difficulty on some MRI systems, when 1H and 31P pulses cannot be applied simultaneously. Transmitting on both 1H and 31P channels at the same time is a prerequisite for the srINEPT-7 sequence. If this is not possible on clinical MR systems, one would need to revert to the sequential srINEPT-9 variant (16).
Peaks Other than PME and PDE
In the 31P spectra of milk (see Fig. 6) there are many other peaks other than PME and PDE peaks. Full assignments of these peaks, however, have not been available in the literature and are not the goal of this study. Although these peaks are currently not of interest, they could represent some metabolites with biological significance. By comparing the srINEPT-7 spectrum with the srINEPT-9 spectrum, it can be seen how these peaks can be filtered or retained.
In summary, the srINEPT technique for detecting PME and PDE is revisited. Both theory and experiments have indicated that the srINEPT-9 pulse sequence, which is currently applied in clinical investigations for PME and PDE detection, can be simplified to srINEPT-7, provided that simultaneous transmission of radiofrequency pulses on two frequencies is technically possible. Replacing the three π pulses by one π pulse possibly lowers overall power deposition. It is demonstrated with a milk sample that ReBurp pulse performs well as a soft π pulse for 1H selective refocusing. Furthermore, the spectra of srINEPT-7 select just the PME and PDE metabolites, whereas in the srINEPT-9 spectra other metabolites which are not currently used as cancer biomarker are also detected. Tests under different conditions have shown that srINEPT-7 is a robust pulse sequence, because it is not sensitive to small variations in experimental setups. Variations within 6% (±1ms/16ms) in the pulse length, ±0.08 ppm in the offset, and ±1 dB in the B1 field attenuation factor of the 1H ReBurp pulse will cause 31P signal intensity variations in the spectrum of less than 5%.