The transgenic adenocarcinoma of mouse prostate (TRAMP) mouse is an established and well-studied murine model of prostate cancer (1, 2). The histopathology of TRAMP mouse cancer tissue mimics that of human disease. TRAMP mice develop spontaneous progressive disease that begins with prostatic intraepithelial neoplasia (PIN) and then advances to frank carcinoma lesions in the prostatic lobes. The cancer also frequently metastasizes in this model to the lymph nodes (LN) and lungs, and, to a lesser extent, to the kidneys, adrenal glands, liver, and bone (2). The TRAMP model is being used in numerous laboratories to identify the molecular mechanisms associated with the initiation and progression of metastatic prostate cancer. Currently, almost all TRAMP studies use histopathology as the only parameter to evaluate the progression of the disease as well as the efficacy of the treatment agent being tested. Although this parameter is very useful and informative, the mouse has to be killed. Having a noninvasive, in vivo method would be valuable for following disease progression and treatment response in the same animal. TRAMP mice also serve as a good model system for testing new methods to characterize human prostate cancers.
Combined proton MRI and MR spectroscopic imaging (MRSI) exams have become routine clinical procedures for characterizing prostate cancer in humans (3–6). In particular, proton MRSI allows identification of changes associated with prostate cancer biomarkers, such as an increase in choline and a reduction in citrate. Single-voxel proton MRS data have been demonstrated in the TRAMP model (7), but this required a 7T animal MR system and an extremely long acquisition time (∼2 hr for a single-voxel spectrum). Studying cellular bioenergetics in the TRAMP mouse would also benefit from in vivo MRS studies with non-proton nuclei, such as 13C, but due to the low natural abundance of 13C and its low sensitivity compared to the proton, the signal-to-noise ratio (SNR) is prohibitively low in small animals. With the recent development of a method for retaining dynamic nuclear polarization (DNP) in solution, high-SNR 13C MRI and MRSI data have been demonstrated in rats following injection of a hyperpolarized 13C agent (8–11). Substrates such as hyperpolarized 13C pyruvate (pyr) and 13C acetate allow unique access to observe fluxes in metabolic processes such as glycolysis and the citric acid cycle. In particular, hyperpolarized pyr labeled at the C1 position has been injected into rats and pigs so that 13C1 pyr and its metabolic products 13C1 alanine (ala) and 13C1 lactate (lac) can be observed (10, 11) over a period of approximately 1 min after injection. This technique may enable the acquisition of 13C MRS data with very high temporal resolution (on the order of seconds) and the observation of real-time, tissue-specific metabolic changes.
The MR properties of hyperpolarized agents are inherently different from those of endogenous nonhyperpolarized molecules. One of the most important considerations for imaging of hyperpolarized 13C substrate is that the longitudinal magnetization consumed (by both T1 relaxation and RF pulses) during an MR experiment is not restored by relaxation. Thus RF pulse sequences employed for MRI or MRS experiments of hyperpolarized 13C must be able to utilize the prepolarized magnetization efficiently and quickly. One of the strategies for this situation is to use small-tip-angle excitation pulses to consume only a fraction of the magnetization for each k-space data point acquisition (11, 12). In this study a double spin-echo (DSE) pulse sequence incorporating a small tip-angle excitation RF pulse and hyperbolic-secant refocusing pulses was used for MRSI of the 13C substrate and its metabolic products. A flyback echo-planar readout trajectory was incorporated into this sequence in order to allow the rapid acquisition of multidimensional MRSI datasets (13).
The goal of this study was to use the developed DSE pulse sequence to determine the feasibility of acquiring high-spatial-resolution 3D 13C MRSI data and 2D dynamic MRSI data in TRAMP and normal mice using hyperpolarized 13C1-labeled pyr on a clinical 3T MR scanner.
MATERIALS AND METHODS
Polarizer and Materials
A prototype DNP polarizer that was developed and constructed by GE Healthcare (Malmö, Sweden) was used in this study. The polarizer was installed in the room adjacent to the MR scanner. The compound polarized was a mixture of pyruvic acid-1-13C and the trityl radical (Tris[8-carboxyl-2,2,6,6-tetra[2-(1-hydroxyethyl)]-benzo(1,2-d:4,5-d′)bis(1, 3)dithiole-4-yl]methyl sodium salt) obtained from GE Healthcare (Oslo, Norway). The purity of the pyruvic acid was ∼91% and the degree of labeling was >99%. TRIS/NaOH/EDTA dissolution medium was used to bring the final dissolved sample to the desired concentration and pH level. Two different concentrations of the hyperpolarized pyr solution (79 mM and 250 mM) were used in this study. An aliquot of the pyr solution was used to measure the percent polarization (ranging from 17% to 21% for these experiments), as well as the pH of the final solution (ranging from 6.5 to 8.0). In each of the mice studied, 0.35 ml of the final pyr solution was injected into the mouse over a 12-s period, followed by a 0.2-ml normal saline flush. The injection volume was approximately 10% of the mouse blood volume, which demonstrated no adverse effects. The protocol was approved by the UCSF Institutional Animal Care and Use Committee.
The polarization and subsequent dissolution were described previously (8). Briefly, the sample was loaded into a small cup constructed from PEEK, and then placed inside one end of a long Teflon tube. The sample was lowered into the variable-temperature insert (VTI) and then the VTI was evacuated to ∼0.8 mbar to cool the sample to 1.2 K. The sample was then irradiated by microwave (∼93.9 GHz) in a helium bath at a magnetic field of 3.35T for more than 1 hr. Just prior to the dissolution, the dissolution stick was filled with the dissolution medium, prepressurized with helium gas, and then heated. After the sample was polarized, and at the time of dissolution it was raised 10 cm from the center of the magnetic so that the sample was out of the helium bath. The dissolution stick was then inserted into the Teflon tube and connected to the sample cup to allow the hot solvent to be injected into the sample cup. A custom-made glass flask was used to collect the pyr solution. The glass flask was precooled in an ice bath so that the solution was approximately 40°C at the time of injection. The level of polarization was estimated using a custom-designed low-field NMR spectrometer that provided measurements of free induction decays (FIDs) from an aliquot of the dissolved solution.
RF Coil and Scanner
The RF coil used in the study was a custom-built dual-tuned 1H-13C coil with quadrature 13C and quadrature 1H channels that was designed and constructed for this study. It is based on an earlier design (14) in which a half-Helmholtz and a low-pass birdcage resonator (in this instance with eight rungs) were mounted on a single cylinder so as to share a pair of legs. The present coil differs from that reported in Ref.14 merely in the addition of a second half-Helmholtz unit, to afford quadrature operation for protons. The diameter of the coil was 5 cm and the length of the coil was 8 cm to accommodate the significant size increase of the adult TRAMP mice as tumors developed. All studies were performed on a GE 3T scanner (Waukesha, WI, USA) equipped with multinuclear spectroscopy capability, including a broadband amplifier.
A small-tip excitation followed by a DSE sequence was designed for fast spectroscopic imaging of 13C hyperpolarized pyr and its metabolic products. An SE signal is desirable because it allows for phase-sensitive analysis of the spectroscopic data. Two hyperbolic secant pulses were used instead of the conventional RF refocusing pulses to achieve accurate refocusing and provide insensitivity to B1 variations and/or minor errors in setting transmit gain (15). Since the hyperpolarized magnetization is not replenished by longitudinal relaxation, and any error in the refocusing pulse flip angle (FA) will contribute to the loss of magnetization, the adiabatic characteristic of these pulses is desirable for refocusing the spins in acquisitions that utilize hyperpolarized agents. Using a pair of adiabatic pulses is an important consideration since the pulse pair eliminates the frequency-dependent nonlinear phase that occurs when a single adiabatic pulse is used (16). The hyperbolic-secant refocusing pulse used had a bandwidth of 2.8 kHz, pulse duration of 10 ms, and maximum B1 of 0.17 mT.
To allow acquisition of 3D or time-resolved 2D 13C MRSI data within the very short time before the hyperpolarized magnetization decays by T1 relaxation, a flyback echo-planar trajectory was implemented during signal readout to encode both spatial and frequency information (Fig. 1) (13, 17). The flyback echo-planar trajectory was designed based on the 40 mT/m gradient amplitude and 150 mT/m/ms maximum slew rate available on the 3T MR scanner used in this study. The gradient waveform was designed for a 581-Hz spectral bandwidth to include 13C lac, 13C ala, and 13C pyr without spectral aliasing. The designed bandwidth provided coverage of the 13C metabolites while maintaining adequate SNR efficiency of the readout waveform. A total of 59 readout/rewind lobes were included during each readout for a spectral resolution of 9.83 Hz. With a readout filter of 25000 Hz/2538 points, 16 k-space points were acquired during each repetition time (TR), achieving a 16-fold decrease in acquisition time compared to conventional phase encoding with the same TR. The flyback readout trajectory was designed for a 5.4-mm minimum spatial resolution and an SNR efficiency of 61% (13).
T2-weighted anatomical images were obtained in all three planes using a fast SE (FSE) sequence. Axial and sagittal images were each acquired in 10 min with FOV = 10 cm, 192 × 192 matrix, slice thickness = 2 mm, and NEX = 6. Coronal images were acquired with FOV = 12 cm, 192 × 192 matrix, slice thickness = 1.5 mm, NEX = 8, and scan time = 10 min. The total imaging time required to obtain images in all three planes was thus approximately 30 min.
Time-resolved 2D 13C MRSI data were acquired on one normal mouse from a 2D slice through the left mouse kidney using a 1 × 8 phase-encoding matrix with flyback echo-planar trajectory on the x-axis (16 × 8 effective matrix), 5.4 mm × 5.4 mm × 10 mm spatial resolution (0.292-cc voxel resolution), and a 43.2 mm × 43.2 mm FOV. The echo time (TE) was 35 ms, TR was 375 ms, and each 2D dataset was acquired in 3 s. The FA for the excitation pulses was 5°. A total of 20 time points were acquired.
3D 13C MRSI data were acquired in 10–14 s in one normal mouse and four TRAMP mice using an 8 × 8 × 1 phase-encoding matrix with flyback echo-planar trajectory on the z-axis (8 × 8 × 16 effective matrix), 5 mm × 5 mm × 5.4 mm spatial resolution (0.135 cc voxel resolution), and 40 mm × 40 mm × 86.4 mm FOV to cover the mouse torso and abdomen. The TE for the 3D acquisition was 35 ms (readout began at the center of the second SE), the TR was 160 ms, and the total acquisition time was 10.24 s for three of the TRAMP mouse studies. For one TRAMP mouse, a TE of 140 ms was used with a 3D MRSI acquisition to obtain the second SE symmetrically (18) while using the same flyback echo-planar trajectory (readout was centered on the center of the second SE). In this case the TR was 215 ms and the total acquisition time was 13.76 s. A variable-FA (VFA) scheme was used for the small-tip-angle excitation pulse to effectively use all of the magnetization available (12).
All animal studies were carried out under a protocol approved by the UCSF Institutional Animal Care and Use Committee. TRAMP male mice were raised to 20–35 weeks of age. For each of the TRAMP mouse studied, intravenous access was established by a semipermanent jugular vein catheter placed during a surgical operation that typically occurred one or more days before the MR exam. Just prior to each study, the mouse was anesthetized with isoflurane (2–3%) delivered via O2 flow at 1 liter/min. An i.v. extension tube was then attached to the main catheter and the animal was transferred to the scanner suite, placed on a 37°C heating pad, and positioned in the RF coil. Anesthesia was maintained by a continual delivery of isoflurane (1–1.5%) with oxygen (l liter/min) via a long tube to a nose cone placed over the mouse's snout. Respiratory rate and skin color were monitored by periodic visual inspection (facilitated by slots in the RF coil). One animal was also monitored during scanning by a specially designed, MR-compatible murine respiratory rate monitor (BioVet System, Summit Anesthesia Solutions). After the imaging studies, each animal was placed in an isolated cage and allowed to recover from the anesthesia.
One B6SJL male wild-type mouse was imaged at unknown age (∼24 weeks) and 32.0 g body weight. One C57BL6/FVB male wild-type mouse was imaged at 20 weeks of age and 32.0 g body weight. Four C57BL6/FVB TRAMP mice supplied by the Roswell Park Cancer Institute (Buffalo, NY, USA) were imaged at 35 weeks, 23 weeks, 22 weeks, and 24 weeks of age, and 38 g, 34 g, 25 g, and 32 g of body weight, respectively (Table 1).
Table 1. Summary of the TRAMP and Normal Mice in This Study
Dose of pyruvate (ml/mM)
Kidney lactate SNR
Primary tumor lactate SNR
Kidney lactate/pyruvate ratio
Primary tumor lactate/pyruvate ratio
The SNR and lactate/pyruvate ratio for the 2D dynamic study was derived from spectra acquired at the fifth time points, as shown in Fig. 2. N/A = not applicable.
All 13C MRSI data were processed using custom MRSI processing software developed in our institution with modifications made specifically for this study. The raw data acquired with the DSE sequence were ordered as a 4D array, with the first dimension being k-space values in the flyback direction, the second being time decay, the third being one phase-encoding direction, and the fourth being the other phase-encoding direction. The k-space points in the flyback dimension corresponding to the constant gradient portion of the trajectory were selected out and the data set was reordered so that time decay was the first dimension. The reordered dataset was then processed in the same manner used for the conventional 4D MRSI dataset, with the exception that the k-space points in the flyback dimension were each acquired at a slightly different time point (19). This was corrected during the 4D Fourier reconstruction using the fact that an origin shift in k-space is equivalent to a phase shift in the transformed domain. The time-domain signal was apodized by a 16-Hz Gaussian filter and zero-filled from 59 to 128 points. No apodization in the spatial domains was applied. The total reconstruction time on a SunBlade 1500 UNIX workstation (Sun Microsystems, Inc., Santa Clara, CA, USA) was approximately 5 s for 3D MRSI datasets acquired with the flyback echo-planar trajectory.
SNRs for all studies were calculated from a voxel in the kidney (also a voxel in the tumor in the TRAMP mice studies) using the peak height of the 13C lac resonance and noise estimated from the right and left ends of the spectra that did not contain 13C resonances.
Hyperpolarized 13C MRSI spectra were obtained from two normal mice and four TRAMP mice using the new pulse sequence and coil developed for this study. The 13C lac resonance in the mouse kidney was observed to have high SNR (26.9–83.6) in all studies, and to be high in the tumor for the TRAMP mice studies (Table 1). Alanine levels were often much lower than lac in most tissues (other than liver), and were often not observed in murine tumors. The lac/pyr ratio was also estimated in the kidney voxels and tumor voxels. In all four TRAMP mice studied, lac/pyr ratios were higher in the prostate cancers compared to the kidney.
The time-resolved 2D MRS images acquired from a normal mouse in a slice through the left mouse kidney demonstrated the time course of 13C pyr uptake and conversion to 13C lac in mouse kidney and muscle (Fig. 2). Substantial 13C lac was observed in the kidney spectrum, with relatively lower 13C lac observed in muscle. The representative 2D spectra array shown in Fig. 2b was collected from the fifth temporal time point. The SNR and lac/pyr ratios listed in Table 1 were also calculated from this spectral array.
13C 3D MRSI were acquired at 0.135-cc nominal voxel resolution from a male wild-type mouse with spatial coverage throughout the body of the mouse in 10.24 s. Different 13C metabolite profiles were observed in the mouse liver, kidney, and prostate (Fig. 3). Relatively high amounts of ala were observed in the liver compared to the kidney or normal mouse prostate. Relatively high levels of lac were observed in the kidneys. Lower levels of pyr and lac (13C lac SNR mean = 17.2, 13C pyr SNR = 20.9) and baseline level of ala were observed in the normal mouse prostate compared to the other tissue types. Respiratory motion did not appear to cause any adverse effects in either the T2-weighted FSE images or the 13C MRSI data in the anesthetized mice.
The 3D MRSI data acquired from a TRAMP mouse with a large primary tumor in the abdomen and an LN metastasis demonstrated a different metabolite pattern in the cancer regions as compared to the normal mouse prostate (Fig. 4). The 13C MRSI spectra in kidneys were very similar in the normal mouse and the TRAMP mouse, with high levels of pyr and lac and relative low ala (Figs. 3c and 4b). However, in the TRAMP mouse the primary tumor and LN metastasis showed much higher relative lac levels compared to the kidney in the same animal and as compared to the normal mouse prostate (Figs. 3d and 4c and d).
When the DSE was used at a longer TE (140 ms) to acquire the symmetrically sampled SE in the 3D MRSI study (Fig. 5), the magnitude spectra had similar linewidth (22.6 Hz for lac) as compared to the pure absorption spectra (21 Hz for lac) acquired starting at the center of the echo (Fig. 4). Similarly to the short TE spectra, elevated lac (higher lac/pyr ratio) was observed in the TRAMP mouse prostate tumor compared to that observed in either the normal mouse prostate or the normal mouse kidney (Fig. 5).
This study demonstrates, for the first time, the feasibility of detecting the rapid cellular conversion of pyr to lac in a transgenic mouse model of prostate cancer using a hyperpolarized 13C substrate. The methods developed in this study allowed the acquisition of 3D 13C MRSI in approximately 10 s with a spatial resolution of 0.135 cm3. Prominent 13C lac resonances were detected in the liver, muscle, and kidney, but the highest lac levels were detected in TRAMP prostate tumors. This initial study indicates a strong potential for this relatively new in vivo metabolic imaging technique to become a powerful tool for monitoring the abnormal bioenergetics of prostate cancers.
Although MRSI of hyperpolarized 13C agents requires a drastically different approach, it offers different and complementary information compared to proton MRSI. Prostate 1H MRSI allows detection of small endogenous molecular markers (choline, spermine, and citrate) within the cytosol and extracellular spaces of the prostate at a steady-state level. The increase in choline and decrease in citrate level in prostate cancer tissue observed by 1H MRSI is due to increased cell density, changes in cellular proliferation and phospholipid metabolism, and the loss of citrate production/secretion capacity that is unique to normal prostate epithelial cells (20).
In normal prostate tissue, citrate production (i.e., incompletion of the Krebs cycle) results in much lower energy production. The existence of substantial anaerobic metabolism and lower overall activity is unique to prostate epithelial cells (21). It can be speculated that the loss of citrate producing metabolism is associated with the increased need for energy in the process of malignancy. In addition, increased anaerobic metabolism can also play a role in supplying the energy requirements of cancerous tissue (22). If these and other metabolic transformations can be observed in vivo at different stages of malignancy, they may provide valuable information for the diagnosis, characterization, and treatment planning of prostate cancer.
By taking advantage of the DNP method to increase the MR sensitivity for 13C substrates, this study demonstrates that changes in 13C metabolite levels in the prostate tumor from the TRAMP model may be observed in vivo. Using the DSE MRSI sequence with flyback echo-planar encoding trajectory, time-resolved 2D and 3D spectra were obtained with very high SNR (Table 1). The high SNR obtained indicated that even higher spatial resolution may be feasible for future studies. Also, SNR may be further increased through improved RF coil design and higher-performance gradients. The use of a small tip angle with adiabatic refocusing to acquire hyperpolarized signal appeared to be robust, as demonstrated by very good spectral quality in all studies. Large FA with a refocusing pulse-train approach is another alternative for imaging of hyperpolarized 13C spins (23). In this approach, T2 relaxation would dominate the signal decay during the experiment. It has been shown that a hyperpolarized 13C substrate has much shorter apparent T2 in vivo as compared to in solution (1.3 s vs. 18 s for bis -1,1-(hydroxymethyl)-1-13C-cyclopropane-D8) (23). Thus without accurate knowledge of the T2 relaxation time of 13C1-pyr in tissue, small-tip-angle excitation was chosen as a good approach for the 3D MRSI experiments in this study. The one TRAMP study performed with acquisition of the symmetric echo demonstrated the advantage of having the same linewidth in magnitude mode as compared to pure absorptive mode. It was not clear whether the longer-TE, full-echo acquisition showed improved SNR over the acquisition that acquired only the right-hand side of the echo when the kidney SNR was used for comparison (Table 1) (18, 24). Measurement of in vivo T2 relaxation times and a larger number of datasets acquired with both schemes are required to resolve whether the full-echo acquisition has a robust advantage in SNR for hyperpolarized 13C MRSI.
The use of the flyback echo-planar readout enabled the acquisition of a high-resolution 3D MRSI dataset covering the entire mouse body in less than 14 s. This was highly significant for several different reasons. According to the dynamic studies shown previously (10, 11) and the time-resolved 2D data shown in this study, it is crucial to choose an imaging window for the MRSI experiment before the decay of the 13C signal via T1 relaxation but after the 13C pyr has been utilized by the tissue so its metabolite products and thus the tissue-dependent bioenergetics can be observed. This MRS imaging window is short by conventional MRI standards and is on the order of ∼20 s, where the metabolite levels are stable and signal has not decayed significantly. In addition, to better depict the 13C metabolic signatures of different tissues or organs, adequate spatial resolution (to reduce partial-volume effects) and a large encoding matrix (to cover the necessary anatomy) are required. The flyback readout trajectory designed for this experiment (5.4-mm minimum resolution) provides the capability for a MRSI matrix that is large enough to cover the mouse (8 × 8 × 16) within the desired imaging time.
Two different formulations of the 13C pyruvic-acid/trityl radical mixtures were used in the study that resulted in two different concentrations (79 mM and 250 mM) of the hyperpolarized 13C pyr solution. The injection volumes and injection rates were kept constant for all experiments. The pyr levels observed in the TRAMP study using the 250-mM formulation were much higher compared to the other three TRAMP studies using the 79-mM formulation (note the much lower lac/pyr ratio for the 250-mM study; Table 1), but the lac SNRs were similar. It can be speculated that since the 13C pyr concentrations for both of the injected solutions were significantly higher than that of biological concentration (∼0.2 mM), the lac SNR was limited partly by the 13C pyr uptake rate and partly by its utilization within the tissue. Even at 79 mM, it is possible that the limit had been met. Understanding the possible trade-off in overall SNR and saturation of the biological metabolic rates may be an important consideration for future studies. In addition, the large differences in lac SNR in the tumor between different studies can presumably be attributed to differences in tumor stage in individual mice and the heterogeneity of the TRAMP model.
Although the lac/pyr ratio was observed to be fairly high (0.82), the normal prostate had much lower levels of both pyr and lac compared to the mouse liver, kidney, or tumor (Fig. 3). A complication of the interpretation of normal prostate lac/pyr ratios is the small size of the normal mouse prostate (∼0.029 cm3, measured in a prior study (7) from the same C57BL6/FVB male wild-type mouse used in this study) and the potential for partial voluming of surrounding tissues. In contrast, prostate tumors in TRAMP mice and other mouse organs were significantly larger than the spectroscopic voxels, and often multiple voxels could be placed within the tumor. In the TRAMP prostate tumor, the level of activity appeared to be elevated not only compared to normal mouse prostate, but to the kidney as well, with a much higher level of lac present. This observation may reflect the transformation from normal prostatic citrate-producing metabolism to malignant citrate-utilizing metabolism; however, the exact mechanisms that lead to the observed changes in the lac/pyr ratio still need to be elucidated. It is interesting to note that lac/pyr ratios were also higher in the tumor tissue compared to the kidney for all TRAMP studies (Table 1).
This study clearly demonstrates the feasibility of obtaining 13C MRSI data with high SNR from normal and TRAMP mice after injecting the animals with hyperpolarized 13C1 pyr. The small-tip-angle with adiabatic refocusing pulse sequence utilizing flyback echo-planar readout trajectory allowed rapid acquisition (in approximately 10 s) of large-matrix 13C MRSI datasets. Different 13C metabolic characteristics were observed in the mouse kidney, liver, normal prostate, and prostate tumor. This study shows that key bioenergetic metabolites in a prostate cancer model can be studied in the TRAMP mouse by MR using DNP to enhanced the signal from 13C-labeled substrates.