Changes in tumor metabolism can give an earlier indication of treatment response than changes in tumor size and there has, therefore, been considerable interest in imaging methods that can detect changes in metabolism and other aspects of tumor biology following treatment (1). One of the most widely used functional imaging techniques for monitoring treatment response in the clinic is to measure uptake of the glucose analog, 18Fluoro-2-deoxyglucose, using positron emission tomography (FDG-PET). Studies in lung, breast, oesophageal, lymphoma and ovarian cancer have all demonstrated that reduced FDG uptake can identify early treatment response (1). FDG-PET, however, is not effective in all tumor types and in brain tumors high uptake by surrounding brain tissue can mask high uptake by the tumor itself (2, 3).
Glycolytic rate in human gliomas has been correlated with tumor lactate concentration, suggesting that detection of increased lactate concentration, for example, with 1H MRS, might provide a similar diagnostic readout to that provided by FDG-PET in other tumor types (4, 5). While 1H NMR spectroscopy can detect lactate in glioma (6), this approach is limited by the need to separate the relatively low lactate signal from overlapping lipid resonances using either long echo times (>135 ms), which suppresses signal from lipid, or spectral editing techniques. These limitations have restricted widespread clinical application (7, 8).
Detection of lactate 13C labeling, using 13C magnetic resonance spectroscopic imaging, following injection of hyperpolarized [1-13C]pyruvate offers an alternative and more sensitive way to detect tumor lactate, as well as providing information on lactate dehydrogenase (LDH) activity. In a murine lymphoma model lactate labeling was shown to be due predominantly to exchange of 13C label between injected hyperpolarized [1-13C] pyruvate and the relatively large endogenous lactate pool (9, 10). In drug-treated tumors the exchange showed an early decrease in rate due to loss of the coenzyme (NAD(H)) and a loss of LDH activity (9, 11). Thus the experiment was shown to provide an early indication of treatment response, in much the same way as FDG-PET, and indeed the two techniques have been compared directly (11). While in the clinic FDG-PET is already used successfully in lymphoma for treatment response monitoring (1), this is not the case in glioma, where high uptake in surrounding brain tissue compromises the measurement. There is, therefore, an urgent need to find alternatives to FDG-PET for monitoring treatment response in glioma. We show here that the hyperpolarized [1-13C]pyruvate experiment can be used to detect successful response to radiotherapy in glioblastoma, where tumor lactate labeling was shown to be decreased between 24 and 96 h following treatment.
C6 Glioma Growth
C6 glioma cells from the American Tissue Culture Collection (ATCC, Rockville, MD) were grown in Dulbecco's modified Eagle's medium (DMEM, Gibco) containing 10% fetal bovine serum and 2 mM glutamine (Sigma Aldrich, St. Louis MO). At ∼80% confluence the cells were washed once with PBS, trypsinized with 1 mL of a 0.4% trypsin solution (Gibco) washed twice and then resuspended in ice-cold PBS.
Tumors were implanted into male Wistar rats by cell injection (n = 13).Animals were anesthetized with 2% isoflurane in an oxygen/air (40/60%) mixture and the gas levels were adjusted to maintain breathing rate at 60 breaths/min. The head was held in a stereotaxic jig (BenchMark, Leica Microsystems, St. Louis MO) and a central dorsal incision of ∼2 cm was made. A hole was drilled through the skull at a point 2 mm anterior and 1.5 mm to the right of the bregma.Cells (5 μL containing 106 cells) were injected over the course of 2 min at a depth of 1.5 mm.
60Co Irradiation of Glioma-Bearing Rats
Animals were given an i.p. injection of 80–100 mg/kg Ketamine, and 5–10 mg/kg Xylazine in a solution of 0.9% saline (n = 9). The 60Co irradiator beam was collimated to produce an irradiation field of ∼16 cm by 5 cm. Further lead shielding covered the nose and neck. The whole brain was exposed to a dose of 15 Gy. Exposure was calibrated using thermoluminescent dosimeters embedded in the middle of 3.0 cm3 Lucite blocks (Total Plastics, Baltimore, MD), to approximate absorbed dose in the brain. These gave a dose rate of 153.6 ± 2.4 rad/min in the unshielded areas, and 4.0 and 5.4 rad/min under the nose and body shields, respectively. The remaining four tumor-implanted animals did not receive radiotherapy and were used as controls.
Animals were placed in a purpose built cradle containing a bed, warmed by warm water, and an isoflurane delivery system, with a home-made nosecone-mounted 13C tuned surface coil (diameter 30 mm) placed above the head. The cradle was inserted into a birdcage proton imaging coil and the entire assembly placed in the isocenter of a 4.7 T magnet (Magnex Scientific, Abingdon, UK) connected to a Bruker spectrometer. The transmit-receive 13C surface coil was inductively coupled to a passive surface coil situated below the head (diameter 30 mm) to improve B1 homogeneity over the whole brain (12). Shimming was performed on a volume covering the whole rat head using FASTMAP, giving a 13C linewidth of ∼30 Hz.
Rapid acquisition with relaxation enhancement (RARE) T2-weighted images were acquired from a 40.5 × 40.5 mm field of view, with a data matrix of 256 × 192 zero-filled to 256 × 256, giving a nominal in-plane resolution of 0.158 × 0.158 mm2 and a slice thickness of 2 mm. The repetition time was 5 s, with an effective echo time of 65.16 ms and an inter-echo spacing of 8 ms. Sixteen echoes were acquired per excitation, and four averages per phase encoding line. Fast low-angle shot (FLASH) T1-weighted images were acquired from the same slice with a 256 × 256 data matrix. The repetition time was 300 ms, with an echo time of 4.55 ms. Four averages were acquired per phase encoding line, and the pulse tip angle was calibrated to be 60°. For T1 contrast enhancement, 1 mL of a 50 mM solution of Magnevist (Schering AG, Germany) in PBS was administered i.v. over the course of 30 s. FLASH images were collected before injection and immediately upon completion of the injection.
Hyperpolarization of [1-13C] Pyruvate
Samples of [1-13C]pyruvic acid (30 mg), containing 15 mM of the trityl radical OX-063 (Oxford Instruments, Oxford, UK) and 2 mM of the gadolinium chelate, ProHance (Bracco Diagnostics, Milano, Italy) were polarized at 3.35 T and 1.4 °K in an Oxford Instruments Hypersense polarizer, according to the manufacturer's instructions. The hyperpolarized sample was dissolved in 4.5 mL of a superheated buffer (∼180°C, ∼10 bar) comprising 40 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES), 94 mM NaOH, 30 mM NaCl and 50 mg/L ethylenediaminetetraacetic acid (EDTA). The pH of the dissolved sample was 7.4, with a final [1-13C]pyruvate concentration of 75 mM.
13C MR Spectroscopy
To assess the optimal time for chemical shift imaging (CSI), nonlocalized 13C spectra were used to assess the dynamics of metabolite labeling (n = 2 nontumor bearing animals, n = 2 tumor-bearing animals). Two milliliter of 75 mM hyperpolarized [1-13C]pyruvate were injected into the tail vein via a polyethylene line ∼15 seconds after dissolution. The injection of the hyperpolarized pyruvate was completed in less than 5 seconds. Spectra were acquired with a spectral width of 6009 Hz and collected into 512 complex points, with TR 1 s and a 250 μs gaussian excitation pulse with a flip angle of 5° at the center of the brain. Signal acquisition was begun before injection of the polarized pyruvate, and repeated once per second over the course of 4 min.
13C Chemical Shift Imaging
13C chemical shift images were acquired 20 seconds after the start of pyruvate injection from a 40.5 × 40.5 mm field of view in a 6 mm axial slice through the tumor, with sequential phase encoding giving a data matrix of 16 × 16 and an in plane resolution of 2.53 × 2.53 mm2 (spectral width of 6009 Hz; TR, 43 ms; 250 μs gaussian excitation pulse with a flip angle of 5° in the center of the brain). The total time required to acquire an image was ∼11 seconds.
Imaging Protocol for Detection of Treatment Response
Tumors were imaged beginning at 10 days after glioma cell implantation using T2-weighted RARE images, T1-weighted FLASH images, before and after contrast agent administration, and 13C CSI following administration of hyperpolarized [1-13C] pyruvate (n = 9). At this time the tumors were between 4 and 10 mm in diameter. The contrast agent-enhanced images showed that there was opening of the blood-brain barrier between 10 and 12 days following tumor cell implantation. Twenty-four hours after imaging the animals were irradiated with 15 Gy. Post-treatment imaging was started 24 h after irradiation, and the animals were followed longitudinally and reimaged at combinations of 24 (n = 3), 48 (n = 5), 72 (n = 7) and 96 h (n = 9) after treatment. Most animals were imaged 2–3 times after therapy. The same imaging protocol was used at all time points following irradiation except that the contrast agent-enhanced T1-weighted FLASH images were only acquired at 96 h post-treatment.
13C CSI data were analyzed using Matlab (Mathworks, Natick, MA). The data were Hamming filtered (13) in both spatial directions and, for the production of false color images of labeled lactate and pyruvate, were zero-filled to 256 × 256 to match the resolution of the proton images. Lactate and pyruvate peak integrals were calculated from exponentially line-broadened (30 Hz) and baseline-corrected spectra in absolute value mode. The analysis was also repeated without Hamming filtering. The T2-weighted RARE images were used to define a region of interest comprising the tumor, and the mean intensity of lactate from the region was calculated. This number was normalized to the maximum pyruvate signal in the image, which was most often located in a nearby blood vessel above the brain. A similar analysis was performed using contrast agent-enhanced T1-weighted FLASH images to define the tumor regions that consisted of either the whole tumor or only the enhancing region of the tumor. All values are reported as the mean ± SD. Statistical significance was assessed using unpaired Student's t-test.
The glioma tumors presented as a mass that skewed the symmetry of the brain, leading eventually to breakdown of the blood-brain barrier (BBB) at 10–12 days postcell implantation, as indicated by increased tumor signal intensity in contrast agent-enhanced T1-weighted FLASH images. T2-weighted RARE imaging also showed the oedema resulting from glioma growth (Fig. 1).
Untreated animals died within 14–21 days of tumor implantation (n = 4), whereas animals whose tumors were irradiated with 15 Gy on or around day 10 lived for 2 months after treatment, at which time they were euthanized (n = 9). These animals showed renewed weight gain and increased appetite and activity, although the tumor appeared to continue to increase in size following radiotherapy.The average size of the tumor regions, which were enhanced in T2-weighted RARE images and in contrast agent enhanced T1-weighted FLASH images, increased from 150 ± 40 mm3 to 240 ± 60 mm3 (+70%, n = 9, P < 0.01) at 96 h following radiotherapy.
13C spectra acquired from the heads of non tumor-bearing animals showed a rapid appearance of [1-13C] pyruvate, but only small signals from lactate, alanine and pyruvate hydrate. Higher levels of lactate labeling were observed in spectra of tumor-bearing rats (Fig. 2), where the levels of labeled lactate peaked at between 10 and 20 s and labeled pyruvate peaked at ∼7 s following the start of hyperpolarized [1-13C]pyruvate injection. These peaks in the lactate and pyruvate signal intensities were earlier than those observed in the study by Park et al. (14). Although the pyruvate doses were similar in both studies the pyruvate solution was injected over a period of 12 s in the study by Park et al. (14) as compared to a period of 5 s in this study.
13C chemical shift images acquired between 20 and 30 s following injection of hyperpolarized [1-13C] pyruvate showed large pyruvate signals in various of the three major vessels on the dorsal side of the brain, two vessels on the ventral side and two vessels in the neck (Fig. 3c). Comparison of spectra from tumor voxels with those on the contralateral side of the brain showed that the lactate signal was much higher in the tumor than in brain (Fig. 3a). This was true both for voxels at the top of the brain and at deeper brain locations. The approximate locations of the voxels used in the analysis of the 13C CSI data are outlined in white in the image shown in Fig. 1. There were high pyruvate signals in both tumor and normal brain tissue from voxels at the top of the brain, which are likely due to partial volume effects and contamination of these voxels with the large pyruvate signal in blood vessels. At deeper brain locations the pyruvate signals were much lower and in the tumor were much lower than the lactate signal. The ratios for lactate and pyruvate in tumor versus normal brain tissue were 2.2 ± 0.6 (>1, P = 0.0006, n = 8) for lactate and 1.2 ± 0.5 (>1, P = 0.009, n = 8) for pyruvate. The lactate/pyruvate ratio was 0.7 ± 0.3 for normal brain tissue, when measured from deeper brain voxels, and significantly higher than a ratio of 0.4 ± 0.2 in the voxels from the contralateral hemisphere at the level of tumor (P = 0.008, n = 9), where we expected to see partial volume effects (see Fig. 1).For comparison, the ratio of the lactate signal in normal brain to the maximum pyruvate signal detected in neighboring blood vessels was 0.14 ± 0.13 and this was not affected by radiotherapy.
A decrease in tumor lactate labeling was observed following radiotherapy (Table 1). The animals were followed longitudinally and re-imaged at combinations of 24 (n = 3), 48 (n = 5), 72 (n = 7) and 96 h (n = 9) after treatment. It is difficult to determine from these data whether repeated pyruvate injections in a single animal had an effect on the later measurements. However, the decrease in lactate labeling observed at 24 h in just three animals, although not statistically significant, was comparable to that observed at later time points with more animals, suggesting that if there is an effect of multiple pyruvate injections this is likely to be small. The mean lactate signal intensity in the tumor, normalized to the maximum pyruvate signal intensity in a neighboring blood vessel, decreased by 30% following radiotherapy, from 0.38 ± 0.16 to 0.26 ± 0.09 (n = 9, P < 0.05) when all the data collected post-therapy were averaged for each animal. A similar decrease in lactate labeling was observed in a murine lymphoma model at 24 h following chemotherapy (9, 11). When each individual animal was followed over time, all but one of the animals showed a reduction in lactate labeling after irradiation during the first 72 h following therapy, and the remaining animal also exhibited reduced lactate labeling by 96 h. The decrease in lactate labeling varied considerably between animals (Fig. 4). There was some evidence of heterogeneity in tumor treatment response, with some regions retaining high lactate labeling and others showing little or no lactate signal. Similar results were obtained whether the tumor region was defined by the T2-weighted or contrast agent-enhanced T1-weighted images. The way in which the CSI data were processed did not affect the conclusion that there was less lactate labeling following radiotherapy although it did affect the magnitude of the response. When no spatial filtering was performed the mean lactate signal intensity in the tumor, normalized to the maximum pyruvate signal intensity, decreased from 0.28 ± 0.14 to 0.12 ± 0.07 (P < 0.01, n = 9).
Table 1. Tumor Lactate/Blood Pyruvate Ratios at Different Time Points Following Radiotherapy
Pre (n = 9)
24 h (n = 3)
48 h (n = 5)
72 h (n = 7)
96 h (n = 9)
Post (n = 9)
The post value represents the average lactate/pyruvate ratio over the whole postradiotherapy observation period. All values are reported as the mean ± SD.
The contrast agent-enhanced T1-weighted images showed a more heterogeneous tumor enhancement following radiotherapy, suggesting altered perfusion, which could affect pyruvate uptake. Histogram analysis of the overall enhancement, measured as the ratio of signal intensity in the enhanced region to contralateral brain, showed a broader signal distribution following therapy. The mean enhancement was 1.22 ± 0.04 pretherapy versus 1.33 ± 0.09 96 h post-therapy; (n = 9, P = 0.03). Analysis of the lactate signal from only those well-perfused regions that enhanced in the T1-weighted images showed a similar decrease in lactate signal intensity following therapy (data not shown).
The C6 glioma tumor model replicates many of the morphological features of advanced human glioblastoma (GBM), including a high mitotic index, foci of necrosis, nuclear pleomorphism, oedema and diffuse infiltrative borders and palisading cells surrounding areas of necrosis (15, 16). In addition, as shown here, the model shows evidence of pseudo-progression, where there was an increase in the area showing contrast agent enhancement post-treatment, despite a positive tumor response to treatment. This is a significant problem in the clinic (17), where the question of whether there is active tumor growth or infiltration in addition to necrosis has important consequences for treatment, and a reliable distinction between the two conditions is, therefore, crucial.To date, no single imaging technique has been identified that can recognize and adequately establish the diagnosis of pseudo-progression. We have shown here that the response of the C6 glioma to radiotherapy can be detected through decreased labeling of tumor lactate following i.v. injection of hyperpolarized [1-13C]pyruvate.
13C chemical shift images showed high levels of pyruvate in blood vessels surrounding the brain. There were similar levels of pyruvate in the tumor and normal brain tissue and much higher levels of labeled lactate in the tumor. This pattern of the relative lactate and pyruvate signal intensities are similar to those observed in the studies of Park et al. (14), in tumor-bearing rat brain, and of Hurd et al. (18) in normal rat brain. Inspection of the CSI data set presented by Hurd et al. (18) (Fig. 2 in their article), acquired 36 s after injection of the rat with 280 μmols [1-13C]pyruvate over a period of 17 s, with a resolution of 2.5 × 2.5 × 10 mm, gave a maximal lactate-to-pyruvate ratio of ∼1. Here, where the image resolution was 2.5 × 2.5 × 6 mm, the animals were injected over a period of 5 s with 150 μmols [1-13C]pyruvate and the image was acquired between 20 and 30 s after injection, the lactate-to-pyruvate ratio for comparable normal brain regions was also ∼1 (Fig. 3). As in the study of Hurd et al. (18), voxels near the top of the brain, near to blood vessels, showed large pyruvate signals due to partial volume effects. The images acquired from tumor-bearing rats by Park et al. (14) had a lower resolution (5 × 5 × 10 mm) and were acquired with a slightly different pyruvate dose, injection rate and time of image acquisition. Nevertheless, they reported lactate/pyruvate ratios in tumor and in normal brain tissue which were comparable to those measured here.
Although the 13C chemical shift images provided the required spatial resolution of the pyruvate and lactate signals they lacked the temporal resolution need to provide kinetic measurements of label exchange (9). Furthermore the large number of pulses that would be required would result in significant destruction of the hyperpolarized signal. Collection of a series of slice-selective 1D spectra, an approach that we have used previously in a subcutaneous lymphoma model (9), was not practical here since any tumor slice would also contain a substantial amount of normal brain tissue. This problem could be resolved in the future by using the fast imaging methods that are being developed for hyperpolarized 13C (19, 20). Nevertheless, although we were not able to determine the kinetics of 13C label exchange, by acquiring images at a fixed time point after pyruvate injection we were able to show that there was a decrease in lactate labeling following radiotherapy. This may reflect changes in tumor perfusion, since there was increased heterogeneity in the contrast agent enhanced T1-weighted images post-treatment, although there was no decrease in the overall level of enhancement, in fact there was a small but statistically significant increase. This is consistent with previous reports showing that changes in glioma perfusion following radiotherapy are heterogeneous, and that reduced perfusion generally predicts increased survival (21). Decreased lactate labeling may also be due to a decrease in lactate concentration, a loss of LDH activity and a decrease in coenzyme concentration, as were observed in a drug-treated murine lymphoma (9). Decreased lactate concentrations following successful therapy in gliomas have been reported previously (22, 23).
In conclusion, we have shown that the response of a glioma tumor model to radiotherapy, which is a front line treatment for this disease in the clinic, can be detected between 24 and 96 h of treatment through decreased 13C labeling of tumor lactate following injection of hyperpolarized [1-13C]pyruvate. Response could be detected despite an increase in tumor volume. Since it is difficult to detect treatment response using FDG-PET, translation of this technique to the clinic may enable a distinction to be made between pseudo-progression and true progression, in which case it could be used to guide subsequent treatment.Since clinical trials with hyperpolarized [1-13C]pyruvate in prostate cancer are about to commence this appears to be a realistic prospect.
The authors thank Jan-Henrik Ardenkjaer-Larsen (GE Healthcare) for research support and useful discussion.