Different compounds containing a choline moiety can contribute to the main peak at about 3.2 ppm in in vivo 1H-MR spectra. These are free choline, glycerophosphocholine, and phosphocholine, but also contributions from protons in taurine, ethanolamines and myo-inositol may be present at this spectral position . As a convenient shorthand, we refer to this composite resonance as the “choline” signal. Choline-containing metabolites are precursors and breakdown products of the phospholipid phosphatidylcholine, a major cell membrane compound . In prostate cancer cell lines, an increase in choline is observed due to an altered phospholipid metabolism . This alteration is probably due to an increased expression and activity of choline-kinase, a higher rate of choline transport, and an increased phospholipase activity [9, 10].
The choline moiety has nine chemically equivalent protons of three methyl groups resonating as a singlet around 3.19 ppm and two methylene groups, resulting in two multiplets at 4.05 and 3.50 ppm . Because the intensity of these multiplets is very low in in vivo MR spectra of the prostate, it is common to only evaluate the nine-proton singlet at 3.19 ppm. Estimates of the T1 and T2 relaxation times of the choline methyl protons (Table 1) are valuable to determine the effect of the chosen echo time (TE) and repetition time (TR) on the choline signal intensity.
Table 1. In vivo T1 and T2 Relaxation Times of the Prostate Metabolites Choline, Creatine, and Citrate
| || ||T1||T2|
|Heerschap et al. ( ||Choline||0.84 ± 0.09 s|| ||0.23 ± 0.06 s|| |
|Creatine||0.86 ± 0.1 s|| ||0.21 ± 0.1 s|| |
|Citrate||0.34 ± 0.04 s|| || || |
|Heerschap et al. ( ||Citrate|| || ||0.18 ± 0.1 s|| |
|Lowry et al. ( ||Citrate||0.84 ± 0.08 s|| ||0.14 ± 0.02 s (PZ)|| |
|0.12 ± 0.03 s (TZ)|
|Scheenen et al. ( ||Choline|| ||1.1 ± 0.4 s|| ||0.22 ± 0.09 s|
|Citrate|| ||0.47 ± 0.14 s|| ||0.17 ± 0.05 s|
|Chen et al. ( ||Choline|| ||0.96 ± 0.25 s|| || |
|Citrate|| ||0.54 ± 0.14 s|| || |
The production and storage of citrate is one of the main functions of the prostate. Citrate is an intermediate in the tricarboxylic acid cycle. In most organs, citrate is quickly oxidized in the tricarboxylic acid cycle and is therefore only present in low concentrations. In contrast, prostate epithelial cells actively produce citrate and store it in the luminal space, where it is one of the main components of the prostatic fluid . Prostate tissue has high levels of zinc, which inhibits mitochondrial (m-)aconitase activity. This leads to the buildup of a high concentration of citrate . In prostate cancer, a decrease in zinc levels is observed that leads to activation of m-aconitase and the consequential oxidation of citrate . At the same time, the morphology of the prostate gland changes, leading to a loss of luminal space, which might also cause a decrease in the observed (or total) citrate levels.
Citrate contains two methylene groups that are magnetically equivalent (Fig. 2A). The four protons of these groups form a strongly coupled AB spin system. The difference in chemical shifts (Δ), the midpoint of the chemical shifts (δ), and the scalar coupling (J) of this spin system depend on pH [17, 18] and cation concentration  and are approximately 0.15 ppm, 2.61 ppm, and 16.3Hz, respectively (Fig. 2B). Because citrate is a strongly coupled spin system, its shape depends on interpulse timing, pulse shape, TE, and field strength [17-23]. In Table 1, the relaxation times for citrate at 1.5T and 3T are given. The determination of the T2 relaxation time of citrate is less straightforward than for singlets. By increasing the TE in such an experiment, there is not just attenuation of the signal intensity due to T2 relaxation, but also shape and intensity variation due to J-modulation.
Figure 2. A: Schematic chemical structure of citrate. B: Simulated spectral shape of citrate at 600 MHz. Indicated are the scalar coupling constant (J), the chemical shift difference (Δ) and the midpoint of the chemical shifts of the second and third peak (δ).
Download figure to PowerPoint
In the first in vivo prostate 1H-MRS studies at 1.5T, stimulated echo acquisition mode (STEAM) and point resolved spectroscopy (PRESS) techniques were used for volume localization [4, 24]. One advantage of the STEAM for these data is its ability to use a very short TE (at the expense of the general loss of half of the signal in a stimulated echo). In this way, the strongly coupled protons of citrate will have limited phase evolution, which will result in an almost completely in-phase citrate signal. Integration of the area of the citrate peak(s) will then result in maximal signal intensity. For the PRESS sequence, generally longer TEs are used for prostate MRS, and dispersive parts can be present in the spectrum that affect the peak area by cancellation with absorptive parts in simple integration. Several studies have been performed to determine the PRESS pulse timing with maximum absorptive signal at the central lines of the citrate signal. Van der Graaf et al.  used a delay of 7.5 ms between excitation and the first refocusing pulse, and varied the TE. Their optimal TE was 130 ms at 1.5T. One sequence optimization procedure at 3T led to an optimal TE of 75 ms (negative absorptive shape) and 145 ms (positive absorptive shape) with a delay of 25 ms between excitation and refocusing . However, other researchers found an optimal TE of 85 ms  and there are more possibilities that result in a favorable citrate signal . It is therefore not surprising that prostate MRSI acquisition software packages of three MR vendors at 3T are equipped with quite different TEs, varying from 85 to 145 ms . The differences in the spectral shape of citrate in different pulse sequences will lead to variations in the integral of citrate signal at a constant citrate concentration. The influence of interpulse timing is also evident from matrix density simulations [19, 22] illustrated in Figure 3, which shows simulated and in vivo spectra of one patient, using a PRESS sequence at 3T with a TE of 145 ms and a semi-LASER sequence (Localization by Adiabatic Selective Refocusing) at nearly the same TE of 144 ms . In the semi-LASER sequence the excitation pulse is followed by four adiabatic refocusing pulses.
Figure 3. A: Simulated spectral shape of citrate (Cit) using the PRESS sequence with an optimized pulse timing (90°—25 ms—180°—72.5 ms—180°—47.5 ms—echo) at a TE of 145ms at 3T . B: The simulated Cit shape using an optimized semi-LASER sequence at a TE of 144 ms (90°—11 ms—180°—21 ms—180°—29 ms—180°—51 ms—180°—32 ms—echo) . C: T2-weighted image of the prostate of a 71-year-old man with biopsy proven prostate cancer (Gleason score 9). D: In vivo spectrum of the PRESS (TR = 750 ms) of normal tissue. E: Spectrum of corresponding region with the semi-LASER sequence (TR = 2070 ms). The region is indicated in panel C with a blue circle. F: In vivo spectrum of tumor tissue (red circle in panel C) with the PRESS sequence. G, H, and I: The corresponding fits of LCModel using the simulated citrate shape of panels A and B show minimal residuals (J, K, and L). Each spectrum is scaled to maximum intensity.
Download figure to PowerPoint
The optimal TR for citrate detection can be quite short, because of its relatively short T1 (Table 1). It was calculated that the use of a TR of 750 ms, instead of 1500 ms, would lead to an increase in signal-to-noise ratio per unit time of 17% for citrate and a decrease of 6% for choline for 3T data . When a weighted averaged acquisition scheme is used for MRSI, the use of a short TR allows for more averages in the centre of k-space in the same amount of measurement time. In this way, the acquisition of the phase-encoding steps in k-space can be done following the shape of a Hanning filter. Apodization in k-space with a Hanning filter produces a point spread function with widened full-width-at-half-maximum, but strongly decreased signal contamination form more spatially distant signals . The effective voxel size increases, but extraprostatic lipid signal contamination is strongly decreased.
In the physiological range of pH (6.8–7.4), variations in Δ of 3.2 Hz and variations in δ of 0.025 ppm were observed using a 400-MHz magnet . The changes in J are minimal in this pH range [17, 18], but the concentrations of zinc, calcium, and magnesium also influence the value of this coupling , which may have significant effects on the in vivo resonances. Figure 4 shows that small changes in Δ and J can have substantial influences on the spectral shape of citrate. Although a relation was found in vitro between the spectral shape of citrate and the ion concentration or pH, variations in the spectral citrate shape in vivo are difficult to relate to ion concentrations or pH, as these are difficult to measure. For a good fitting, it is necessary to use a model signal that is based on J, δ, and Δ values that closely resemble those present in vivo. At our institution, we use J = −16.2, δ = 2.625, Δ = 0.154, as this led to the smallest residuals for the citrate resonance using LCModel fitting (unpublished data). Using these parameters, the in vivo citrate shape closely resembles the simulated spectrum (Fig. 3). This was the case for both the PRESS spectrum and that obtained with the semi-LASER sequence, indicating that these values are a good approximation of the in vivo coupling parameters.
Figure 4. Simulations with NMRSIM (part of Topspin, Bruker BioSpin Corporation, Billerica, Massachusetts, USA) show the influence of small differences in the scalar coupling and chemical shifts on the citrate shape at a field strength of 3T with an optimized pulse sequence (TE 145 ms) . Line broadening of 1 Hz (black spectra) and 4 Hz (red spectra) were used. Δ was 2.6105 ppm for all spectra. A–D: Δ was kept constant at 0.151 and the scalar coupling was varied between 15.3 Hz and 16.8 Hz. E–H: J was kept constant at 16.3 Hz and Δ was varied between 0.141 ppm and 0.156 ppm. The amplitudes of all spectra are scaled to a reference signal.
Download figure to PowerPoint
The side lobes of the citrate resonance extend to the 3 ppm region and contain a mixture of absorption and dispersion shapes at a TE of 145 ms at 3T , which can result in a negative effect on the total intensity in this region. When a model signal is used for the fitting, this effect can be taken into account; however, when simple peak integration is used, one cannot fully compensate for this effect. Negative components of the citrate signal will be excluded from citrate quantification, and one might need to revert to severe baseline corrections to prevent an influence on other metabolite signals.
The RF pulses that are used to suppress the large lipid and water signals may adversely affect the citrate signals. Often dual frequency-selective pulses are used to suppress both these signals simultaneously, mainly by Mescher-Garwood (MEGA)  or double band-selective inversion with gradient dephasing (BASING) [30, 31] pulses. These pulses selectively invert the lipid and water resonances and are surrounded by crusher gradients. Their bandwidth and position in the frequency domain should be sufficient to invert all lipid signals, but distant enough from the chemical shift of citrate. When the bandwidth of the lipid inversion pulse is too broad, this will cause a decreased signal intensity of citrate (Fig. 5). As a consequence, healthy spectra may get a “cancerous” profile. Therefore, a good adjustment of the dual frequency pulses is essential for obtaining consistent results. Spectrally selective refocusing pulses may be used instead of signal suppression pulses, which prevent refocusing of lipids by simultaneous volume and frequency selection [32, 33]. Care should be taken that these pulses fully excite or refocus the citrate spins and leave the lipid signals untouched.
Figure 5. A–C: In vivo prostate spectra of a healthy volunteer at 3T using PRESS with MEGA pulses for water and lipid suppression. The spectra are from the same location, but the width of the frequency selective inversion bands was 1.40 ppm in panel A, 1.45 ppm in panel B, and 1.55 ppm in panel C, leading to a decrease in the citrate (Cit) intensity, while the intensities of choline (Cho), spermine (Spm), and creatine (Cr) remain unchanged. Consequently, the (Cho+Spm+Cr)/Cit ratio increases influencing the classification of the spectrum. D–F: The same effect in a phantom containing Cit, Cho, Spm, and Cr using MEGA pulses with a width of the inversion bands of 1.35 ppm in D, 1.45 ppm in E, and 1.55 ppm in F. G: Shape of the MEGA pulse in the frequency domain. These (Cho+Spm+Cr)/Cit ratios were determined with LCModel. In line with previously published metabolite ratios, the citrate intensity in LCModel was scaled to the number of protons in citrate, rather than to the magnitude integral at this TE, which is smaller due to cancellation of signal intensity by the strongly coupled pattern.
Download figure to PowerPoint
For proper selective suppression or spectral excitation, the homogeneity of the B0 field is critical. Poor homogeneity will not only negatively affect spectral quality as it causes broadening of the spectral lines; it will also decrease the effectiveness of frequency selective pulses. Broadened or shifted fat and water signals can suffer from diminished suppression and components of these signals may overlap with the resonances of interest. In addition, in the case of shifted or broadened spectral lines of citrate, the metabolite can be influenced by the frequency selective pulses for lipid and water suppression leading to decreased citrate intensities, comparable to the effect shown in Figure 5.
The dependence of the citrate signals on interpulse timing can also be exploited for spectral editing. By varying this timing (at a constant TE) in such a way that citrate is inverted in one measurement and in phase in the next measurement, uncoupled resonances can be removed from the spectrum by subtraction [34, 35]. In this way, rapid citrate imaging without lipid suppression is possible.
The tissue concentration of polyamines in the prostate is relatively high. Because spermine is the dominant polyamine in the prostate, we focus on this compound. Like citrate, polyamines are stored in the luminal space, and a very strong correlation between the citrate and spermine concentration is reported for prostatic fluid specimens . A hypothesis for the strong correlation (r = 0.94) is the formation of complexes between citrate and spermine since citrate is negatively charged, whereas spermine is positively charged. In this way ionic neutrality can be achieved . Polyamines play a role in prostatic growth and differentiation . A decrease in spermine has been suggested as a marker for prostatic malignancy [37, 38]. In prostate cancer, a decrease in spermine or polyamine levels is observed compared with benign tissue using MRS  and high-resolution magic angle spinning experiments . The incorporation of polyamine levels measured with MRS to improve detection of prostate cancer has been proposed and has yielded an increased sensitivity at the same specificity .
Spermine is a coupled spin system and, in addition to its amine groups, contains 10 methylene groups. These methylene protons consist of symmetrical pairs, giving a total of four protons that resonate approximately at 1.81 ppm with further groups of four at 2.11 ppm, 3.13 ppm, 3.12 ppm, and 3.18 ppm . These chemical shifts are pH-dependent  and these quoted chemical shifts were measured at pH 7 . At a higher pH, the amine groups are more protonated, and the chemical shifts are therefore more downfield . Protons near a nitrogen atom show the largest pH dependence. Spermine proton chemical shifts are also sensitive to temperature differences. We performed temperature measurements at 500 MHz with a spermine compound dissolved in water. These measurements showed that protons near a nitrogen atom had the highest temperature dependence (unpublished data). For that reason, when one wants to perform a phantom measurement to determine the shape of spermine (with a certain sequence), the phantom should be measured at body temperature and have a pH in the physiological range. Local chemical shift correction to improve the separation between choline and spermine is hindered by the dependence of the chemical shift of spermine on the environment. Also, usually no water reference measurement is done that could be used for this purpose. The metabolites in the prostate spectrum are unsuitable for this purpose, as the chemical shift of citrate is environment-dependent and choline is not always well separable from spermine.
As with citrate, TE and interpulse timing influence the spectral shape of spermine and leading to dispersive components in the resonances. If dispersive parts are present in the 3.1-ppm region, this can negatively affect the apparent intensity of choline and/or creatine resonances. Furthermore, BASING and MEGA pulses that are used for simultaneous water and lipid suppression invert the 2.1- and 1.8-ppm resonances of spermine . Without these pulses and crushers, the 2.1- and 1.8-ppm resonances could be helpful for decomposition of spermine from the 3.1-ppm region. Figure 6 shows the influence of the MEGA pulses on the spectral shape of spermine. As expected, the resonances at 1.8 and 2.1 ppm are almost completely crushed by the combination of MEGA pulses and crushing gradients. The resonances at 1.8 and 2.1 ppm are scalar coupled to the resonances in the 3.1 ppm region; therefore, the selective refocusing of these upfield groups also refocuses the J-evolution of the downfield protons. This is evident from the difference in spectral shape and intensity of spermine in Figure 6B and 6C, where a different timing for the MEGA pulses is used. These measured spermine shapes can be used as prior knowledge for spectral fitting of the metabolites . Figure 7 demonstrates how the spermine shape can affect the choline and creatine region, showing two spectra from the same location in one volunteer that are measured with a different MEGA pulse timing, resulting in different (Cho+(Spm+)Cr)/Cit ratios.
Figure 6. The spectral shape of spermine measured in a phantom at 3T using PRESS with a TE of 145 ms (90°—25 ms—180°—72.5 ms—180°—47.5 ms—echo). The phantom contained 18 mM spermine, 9 mM ZnCl, 15 mM MgCl2, 18 mM CaCl2, and 60 mM of KCl [41, 71]. The pH was adjusted to 6.8 and the temperature was 310 K. A: The spermine spectral shape without the use of MEGA pulses. B: Spermine spectral shape when two MEGA pulses for combined water and lipid suppression are used surrounding the second refocusing pulse. C: Spectral shape when two MEGA pulses are used surrounding the first refocusing pulse.
Download figure to PowerPoint
Figure 7. A: Spectra of a healthy volunteer of the same voxel in two measurements, containing resonances of choline (Cho), spermine (Spm), creatine (Cr), and citrate (Cit). The difference between the measurements is the timing of the MEGA pulses. In the black spectrum, the MEGA pulses surround the second refocusing pulse (as in Fig. 5B) and in the red spectrum the MEGA pulses surround the first refocusing pulse (as in Fig. 5C). Two different LCModel basis sets were used for the fitting (to take the differences in spermine shapes into account), and the (Cho+Spm+Cr)/Cit amplitude ratios were (B) 0.35 (black and fit) and (C) 0.46 (red and fit). This demonstrates the influence of spectral spermine shape on the (Cho+Spm+Cr)/Cit ratio.
Download figure to PowerPoint
T1 and T2 values of spermine reported in the literature are obtained in vitro and the T2 values were rather short and dependent on the presence of ions and proteins . No in vivo data of relaxation times of spermine spins is available yet.