DOI 10.1007/s10812-015-0124-x
Journal of Applied Spectroscopy, Vol. 82, No. 3, July, 2015 (Russian Original Vol. 82, No. 3, May–June, 2015)
QUANTIFICATION OF N-ACETYL ASPARTYL GLUTAMATE IN HUMAN BRAIN USING PROTON MAGNETIC RESONANCE SPECTROSCOPY AT 7 T M. Elywa
UDC 543.422.25:611.81
The separation of N-acetyl aspartyl glutamate (NAAG) from N-acetyl aspartate (NAA) and other metabolites, such as glutamate, by in vivo proton magnetic resonance spectroscopy at 7 T is described. This method is based on the stimulated echo acquisition mode (STEAM), with short and long echo time (TE) and allows quantitative measurements of NAAG in the parietal and pregenual anterior cingulate cortex (pgACC) of human brain. Two basesets for the LCModel have been established using nuclear magnetic resonance simulator software (NMR-SIM). Six healthy volunteers (age 25–35 years) have been examined at 7 T. It has been established that NAAG can be separated and quantified in the parietal location and does not get quantified in the pgACC location when using a short echo time, TE = 20 ms. On the other hand, by using a long echo time, TE = 74 ms, NAAG can be quantified in pgACC structures. Keywords: proton magnetic resonance spectroscopy, N-acetylaspartate, N-acetyl aspartyl glutamate, human brain, LCModel. Introduction. N-Acetylaspartylglutamic acid (NAAG) is a neuropeptide that is the third most prevalent neurotransmitter in the human brain [1]. NAAG is produced by the coupling between a glutamic acid (Glu) and the N-acetylaspartic acid (NAA) via a peptide bond. Therefore, it may be difficult to distinguish between them on account of the similarity of the chemical structure. In human studies, it has been suggested that NAAG levels are abnormal in patients with schizophrenia [2] and decreased in amyotrophic lateral sclerosis (ALS) [3]. In a previous study [4], it had been difficult to distinguish NAAG from NAA and Glu using proton magnetic resonance spectroscopy (MRS) at 3 T. As the dominant signal of the NAAG spectrum from the N-acetyl proton resonance peak is located at 2.05 ppm, which is 0.03 ppm away from the corresponding NAA signal at 2.02 ppm [5], a high resolution of metabolite peaks can be obtained using ultrahigh-field strengths such as 7 T [6, 7] or at high fields (2 or 3 T), with unusual shimming [8]. Proton magnetic resonance spectroscopy provides markers for various human brain diseases [9, 10]. Some of the important measurement characteristics of the MRS technique are the signal-to-noise ratio (SNR), spatial and spectral resolution, as well as scan time; thus, several localization methods have been proposed. Single volume spectroscopy (SVS) localization techniques, with pulse sequences based on spin echoes (PRESS) [11] or stimulated echoes (STEAM) [12], provide the highest spectral quality [13], as the lower magnetic field strengths result in a peak overlap of the human brain metabolites, such as, NAA with NAAG. Besides, standard spectroscopic sequences such as STEAM or PRESS will suffer if limited transmitter voltage takes place. This results in a reduced excitation flip angle and consequently insufficient excitation in the volume of interest (VOI). Moreover, previous studies have shown that the increase in magnetic field strength increases the 1H NMR spectral peak separation and SNR and thus provides higher sensitivity when compared with lower field strengths [14, 15]. With standard radio-frequency (RF) amplifiers and volume RF coils at 7 T, short 180° refocusing pulses are difficult to achieve in most parts of the brain. Therefore, earlier, high field MRS studies frequently employed surface coils with locally higher RF field amplitude [16, 17]. However, not all brain areas can be adequately examined using surface coils. As the goal of this study is to quantify the NAAG concentrations in different brain structures, STEAM with short echo time benefits, from reduced signal decay, because of the J-evolution of the coupled spin system, allows a more precise quantification of the metabolites, especially with a short T2 relaxation time [18]. Thus, short echo times have been used in several studies [19]. In the current study, echo time (TE) was selected to be short and long 20 and 74 ms, respectively. Although earlier studies had employed STEAM with ultrashort echo times (6 ms), they were limited by the surface coils that prevented detection in the deep brain regions [20]. Also, a quadrature (transmit/receive) surface RF coil or a half-volume RF coil have been used for the Zagazig University, Zagazig, Egypt; e-mail:
[email protected]. Published in Zhurnal Prikladnoi Spektroskopii, Vol. 82, No. 3, pp. 417–422, May–June, 2015. Original article submitted July 9, 2014. 0021-9037/15/8203-0425 ©2015 Springer Science+Business Media New York
425
study [13]. Therefore, it is important to mention that a big difference between the surface and volume coils is that the volume coils are always much larger than the surface coils and can surround the human head. Although the surface coils have a higher SNR, as they receiver noises only from superficial brain layers, they are usually used as receiver coils only. On the other hand, the volume coils are used both as receiver and transmitter coils and can acquire the signal from deep brain regions. In conclusion, the current method of metabolite quantification proved to be successful to quantify NAAG levels from other metabolites, such as NAA and Glu, by using 1H NMRs with 24 head volume coils at 7 T in some deep human brain GM (gray matter) and WM (white matter) structures. Methods and Materials. All measurements were performed on a 7 T whole body magnetic resonance imaging (MRI) system (Siemens Medical Solutions, Erlangen, Germany). A 24-channel volume head coil, with local, circularly polarized volume excitation (Nova Medical, Inc. USA), was used for acquisition of the human brain spectra. STEAM optimized with variable rate selective excitation (VERSE) [21] pulses (TR/TE/TM = 3000/20/15 ms) has been used since RF-peak power limitations prohibited the use of PRESS, with volume coils at 7 T. Two important parameters– the VERSE factor (Vf) and the pulse duration – have been used to reduce the 90° transmitted voltage and to get the true flip angle for the examined deep brain voxels. Figure 1 shows the STEAM with VERSE sequence consisting of three 90° pulses. The repetition time (TR) is equal to the mixing time (TM) plus the echo time (TE) plus the acquisition time (At) plus the pre-scan time (D1). The SIM-WIN program (teaching version) [22] has been used to make the pulse sequence of two stimulated base-sets for the LC Model. The simulated spectrum is based on two parameters: (1) a RF pulse sequence and (2) a spin system; all parameters are the same for the experimental measurements. Simulated basis-sets using 3000 ms repetition time (TR), 15 ms mixing time (TM), 20 and 74 ms echo time (TE), 3600 Hz acquisition bandwidth, and 2048 sample points have been built. Six healthy subjects (age 25 to 35 years) were studied after giving informed consent according to the procedures approved by the Local Institutional Review Board. All in vivo human brain spectra at 7 T were measured using TR/TE/TM = 3000/20/15 ms and TR/TE/TM = 3000/74/15 ms. Moreover, the volume of interest (VOI) was selected according to the anatomical structures of the pgACC and parietal regions. After eddy current correction, the absolute and relative metabolite concentrations and spectral line width were evaluated by the LC-Model [23, 24] using an unsuppressed water signal from the same VOI, as a reference signal. The evaluated standard deviations (Cramèr-Rao lower bounds, CRLB) were expressed in percent of the estimated concentrations. Quantifications for each metabolite were given as concentrations in millimoles (mM) together with the standard deviation (SD) in percent and full width at half maximum (FWHM) in Hertz (Hz). In addition, the SNRs of all the metabolites were calculated by taking the ratio of the metabolite peak height to twice the root mean square of the standard deviation. The SD was estimated in the range from 10 to 12 ppm (where no signals were expected). Results and Discussion. A comparison between STEAM and STEAM with VERSE shows that the required transmitted voltage for the true 90° flip angle changes as the position of the voxel changes in the sample and the required transmitted voltage decreases by 28% from STEAM to STEAM with VERSE. Thus, all the following in vivo measurements have been done using STEAM with VERSE pulses. Figure 2 shows the fitting spectra, depending on the stimulated and experimental LCModel basis sets. The high matching between the measured and stimulated spectra notes are related to SNR and FWHM values. In experimental spectra they are 30 and 0.023 ppm, respectively, and in the stimulated spectra they are 29 and 0.028 ppm, respectively. This comparison shows that the simulation basis sets are beneficial to use rather than experimental basis set in LC Model. Table 1 indicates the concentrations and SDs of some metabolites. The concentration levels are much closer to each other. Comparison is made between the experimental and stimulated concentration levels of the human brain metabolites of six healthy volunteers. The data were acquired from the pregenual anterior cingulate cortex (pgACC) regions using short TE = 20 ms, where the other scan parameters remained the same. The results of [25] show that the white matter structures contain a considerable amount of NAAG, in accordance with the qualitative observations using the STEAM localization sequence (TR/TE/TM = 6000/20/30 ms). In the current study, it is also observed that the NAAG levels in WM-parietal voxel are 50–60% higher than in the GM-pgACC regions using STEAM (TR/TE/TM = 3000/20/15 ms). In addition, this indicated that in the parietal and occipital regions, the NAAG concentration level is significantly higher than in the frontal white matter, and in the gray matter it is much lower than in the white matter. To differentiate the NAAG concentrations in the GM and WM in the human brain, long and short TE are used. This study shows that the NAAG concentration is higher in the WM-parietal than in GM-pgACC, using TR/TE/TM = 3000/20/15, at 7 T. Moreover, in the GM-pgACC, NAAG can be quantified by using long TE 74 ms only.
426
Fig. 1. STEAM pulses using the NMR simulation program; the sync pulse matched with the same pulse used in the 7 T device, that is, VERSE pulses included; TR/TE/TM = 3000/15/20 ms.
Fig. 2. Spectra and LC Model fits to the measured data in a parietal region using (a) simulated basis set (b) experimental basis set.
TABLE 1. Comparison between the Experimental and Stimulated Concentration Levels of Human Brain Metabolites of Six Healthy Volunteers Acquired from the pgACC Regions Using Short TE = 20 ms. The other scan parameters are the same Metabolite
Cexp ± SD, mM/kg
Cstim ± SD, mM/kg
NAA
12.31 ± 0.90
11.94 ± 0.81
Cr
6.64 ± 0.63
7.21 ± 0.33
Glu
8.93 ± 0.54
9.58 ± 0.65
Gln
2.72 ± 0.21
2.52 ± 0.22
GABA
1.75 ± 0.11
1.94 ± 0.12
427
Fig. 3. LC Model analysis of two experimental data acquired from two different brain structures, parietal and pgACC: a — The LC Model fits with the spectrum acquired from the GM-pgACC voxel at scan parameters (TE = 20 ms, TR = 3000 ms, and TM = 15 ms); b — The LC Model fits with the spectrum acquired from the WM-parietal voxel at the same scan parameters. Figure 3 shows the LC Model analysis of two experimental data acquired from two different brain structures, parietal and pgACC. By using image segmentation it has been found that the selected pgACC voxel contains about 35% WM and about 61% GM. It is found that the parietal voxel contains 65% WM and about 33% GM. Figure 3a shows the LC Model analysis of a spectrum acquired from the GM pgACC voxel using scan parameters (TE = 20 ms, TR = 3000 ms, and TM = 15 ms). The spectrum indicates the overlapping between the metabolite peaks in the range of 2 to 2.45 ppm, of NAAG, with N-acetyl-aspartate (NAA), glutamate (Glu), and Glutamine (Gln) peaks. Therefore, NAAG cannot quantify. Figure 3B indicates the quantification of NAAG and other metabolites and shows the high spectrum resolution in the range of 2–2.45 ppm. As the in vivo nuclear magnetic resonance (NMR) spectroscopy aims to define and quantify the highest possible number of tissue metabolites with a high spatial resolution in the shortest scan time, a high magnetic field of 7 T brings about specific challenges that must be met to take advantage of the increased chemical shift separation, and thus, increased sensitivity. The acquired data shows the high spectral quality, which is the precondition for reliable metabolite quantification. Figure 4 shows the data acquired from GM-pgACC voxels of six health volunteers. One can see the overlap of the NAAG and MM20 peaks using TR/TE/TM = 3000/20/15 ms. However, by using the same TR and TM with long TE = 74 ms, it is found that NAAG can be separated and detected, that is, the NAAG metabolite can be quantified in the GM-pgACC structure using long echo time (TE = 74 ms). The most important signals of MM20 are considered to be around 2 ppm because of CH2 group resonance, where the numbers after macromolecules (MM) indicate the approximate chemical shift of the main peak in ppm. Figure 4a indicates the data acquired from the GM-pgACC tissue structures under scan parameters (TR = 3 s, TE = 20 ms, and TM = 15 ms). The graph shows the inverse relation between the SDs of MM20 and of NAAG. Figure 4b indicates the concentration levels of MM20 and NAAG of a health volunteer and shows the overlap between the metabolite peaks, which leads to NAAG un-quantified at short TE. Figure 5a shows the data acquired from the same health volunteers by using scan parameters TR/TE/TM = = 3000/74/15 ms, that is, at long TE. By increasing the TE, it is found that the overlapping between the MM20 and NAAG signals decreases. Figure 5b represents the high concentration level of NAAG relative to the MM20 of one health volunteer. Subsequently, NAAG can be quantified using long TE = 74 ms, which is highly matched with the result of previous studies [26]. Moreover, the results of the determination of NAAG and MM20 levels show that there is a negative correlation between the SD values of the macromolecule MM20 and that of NAAG with short TE = 20 ms (r = –0.93, p = 0.0076), as
428
Fig. 4. 1H MRS data acquired from the pgACC gray matter (GM) of six healthy subjects at 7 T: a — using STEAM with VERSE of factor = 1.4, voxel size = 3.8 ml, BW = 2800 Hz, vector size = 2048, and averages = 128. Negative relationship between SD of MM20 versus NAAG at TR/TE/TM = 3000/20/15 ms; b — two individual spectra of MM20 and NAAG acquired at the last parameters, showing the overlap.
Fig. 5. Analysis data of the same six volunteers: a — the relation between SD of NAAG versus MM20 at TR/TE/TM = 3000/74/15 ms; b — two individual spectra of MM20 and NAAG acquired at the last parameters of TR/TE/TM = 3000/74/15 ms.
TABLE 2. NAAG, MM20 Concentration Levels of Human Brain Metabolites of Six Healthy Volunteers Acquired from the pgACC Regions Using Short TE (20 ms) and Long TE (74 ms). The other scan parameters are the same TE = 20 ms
TE = 74 ms
Metabolites
C ± SD, mM/kg
CRLB ± SD, %
C ± SD, mM/kg
CRLB ± SD, %
NAAG
0.19 ± 0.23
41.17 ± 6.54
1.31 ± 0.04
10.22 ± 1.30
MM20
6.91 ± 0.16
13.83 ± 1.54
2.75 ± 1.56
49.33 ± 43.65
shown in Fig. 4a. However, Fig. 5a also shows a negative correlation between the SD values of NAAG versus that of MM20 at the long TE = 74 ms and TM = 64 ms (r = –0.87, p = 0.023). The strong negative correlation (r = –0.93) represents the overlapping between MM20 and NAAG and shows that the MM20 and NAAG peaks cannot be resolved by using short TE = 20 ms. On the other hand, the high negative correlation (r = –0.87) shows that the NAAG can be quantified in GM-pgACC, by using long TE = 74 ms. Table 2 shows the absolute concentration and Cramer–Rao lower bound (CRLB) of the NAAG metabolite peaks and the peaks of the macromolecules at a chemical shift of 2.05 ppm. All values are estimated as mean ± SD of the six health volunteers. The results show that NAAG cannot be quantified in GM-pgACC using short TE = 20 ms, with the CRLB = 41.17 ± 6.54. In 429
the same scans, the CRLB of MM20 = 13.83 ± 1.54, which is lower than 20%. This means that the MM20 signal is stronger than the NAAG signal and the overlap occurs between each other. Using long TE = 74 ms, the NAAG can be quantified while the signal of MM20 is weaker than NAAG, as presented in Table 1. The consequence of these results is the relation between NAAG and MM20, allowing one to determine the absolute concentrations of the tissue metabolites in a deep human location (GM-pgACC) when using the LC Model analysis. The developed technique STEAM with VERSE is used not only to quantify the NAAG signal from GM-pgACC, but also to quantify many other metabolites. The NAAG concentration at a GM-pgACC location with short TE shows SD values higher than 20%, but with long TE (TE = 74 ms) it can be detected with SD values less than 20%. As shown in Fig. 3a, the pgACC location contains more gray matter than white matter, that is, the MM20 peak intensity increases and overlap between NAAG and MM20 occurs. Figure 3b shows the LC Model fit of a spectrum acquired from a localized VOI in the parietal location, containing more white matter than gray matter. In this location, the peak intensity of MM20 is not so high and the NAAG can be separated from MM20 and can be quantified. Conclusions. In conclusion, it has been shown that NAA and NAAG can be differentiated through appropriate application of the STEAM with VERSE method. In the normal human brain, the method appears reliable and provides a robust separation of the spin systems. The main advantage of the method is its high SNR relatively to small brain volumes. The method shows promise for use in global or diffuse neurological or psychiatric diseases in which NAAG is deep in the brain. The results also clearly display that it is possible, with very good reproducibility, to quantify NAAG in vivo at 7 T using short TE = 20 ms in the white matter parietal regions, and using long TE = 74 ms to determine it in the gray matter of pgACC. Acknowledgments. I would like to express my profound gratitude and sincere appreciation to Prof. Dr. Oliver Speck for his inspiration and indispensable guidance, and for giving me invaluable knowledge throughout the course of my study. My deep thanks to all the members of the BMMR group at Magdeburg University.
REFERENCES 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22.
430
J. H. Neale, T. Bzdega, and B. Wroblewska, J. Neurochem., 75, 443–452 (2000). G. Tsai, L. Passani, B. Slusher, R. Carter, L. Baer, J. Kleinman, and J. Coyle, Arch. Gen. Psychiatry, 52, 829–836 (1995). G. Tsai, B. Stauch-Slusher, L. Sim, J. Hedreen, J. Rothstein, R. Kuncl, and J. Coyle, Brain Res., 556, 151–156 (1991). R. Edden, M. Pomper, and P. Barker, MRM, 57, 977–982 (2007). V. Govindaraju, K. Young, and A. Maudsley, NMR Biomed., 13, 129–153 (2000). I. Tkáč and R. Gruetter, Appl. Magn. Reson., 29, 139–157 (2005). M. Elywa, S. Mulla-Osman, F. Godenschweger, and O. Speck, J. Appl. Spectrosc., 79, 120–125 (2012). A. El-Sharkawy, M. Schar, P. Bottomley, and R. E. Atala, MAGMA, 19, 223–236 (2006). G. Hasler, J. Vander-Veen, T. Tumonis, N. Meyers, J. Shen, and W. Drevets, Arch. Gen. Psychiatry, 64, 193–200 (2007). P. Mandal, Concept. Magn. Res., 30, 40–64 (2007). P. Bottomley, Ann. N.Y. Acad. Sci., 508, 333–348 (1987). J. Frahm, K. Merfsoldt, and W. Anicke, J. Magn. Reson., 72, 502–508 (1987). G. Adriany and R. Gruetter, J. Magn. Reson., 125, 178–184 (1997). C. Juchem, N. Logothetis, and J. Pfeuffer, Magn. Reson. Imaging, 15, 902–912 (2007). H. Qiao, X. Zhang, X. Zhu, F. Du, and W. Chen, Magn. Reson. Imaging, 24, 1281–1286 (2006). R. Gruetter, S. Weisdorf, V. Rajanayagan, M. Terpstra, H. Merkle, C. Truwit, M. Garwood, S. Nyberg, and K. Ŭgurbil, J. Magn. Reson., 135, 260–264 (1998). I. Tkáč, P. Henry, P. Andersen, C. Keene, W. Low, and R. Gruetter, Magn. Reson. Med., 52, 478–484 (2004). I. Tkáč, R. Rao, M. Georgieff, and R. Gruetter, Magn. Reson. Med., 50, 24–32 (2003). P. Barker, B. Soher, S. Blackband, J. Chatham, V. Mathews, and R. Bryan, NMR Biomed., 6, 89–94 (1993). R. Mekle, V. MlynaRik, G. Gambarota, M. Hergt, G. Krueger, and R. Gruetter, Magn. Reson. Med., 61, 1279–1285 (2009). O. Khorkhordin, K. Zhong, F. Godenschweger, and O. Speck, Proc. Int. Soc. Magn. Reson. Med. (2009). C. Shorn and B. Taylor, NMR Spectroscopy: Data Acquisition, Wiley (2004).
23. D. Rosenberg, F. Macmaster, Y. Mirza, J. Smith, P. Easter, S. Banerjee, R. Bhandari, C. Boyd, M. Lynch, M. Rose, J. Ivey, R. Villafuerte, G. Moore, and P. Renshaw, Biol. Psychiatry, 58, 700–704 (2005). 24. J. Frahm, H. Bruhn, M. Gyngellk, K. Merboldtw, W. Anicke, and R. Sauter, Magn. Reson. Med., 9, 79–93 (1989). 25. P. Pouwels and J. Frahm, NMR Biomed., 10, 73–78 (1997). 26. F. Jessen, N. Fingerhut, A. Sprinkart, K.-U. Kühn, N. Petrovsky, W. Maier, H. Schild, W. Block, Michael, and F. Träber, Schizophrenia Bull., 39, 197–205 (2013).
431