The serial 3D images of [1-13C]lactate acquired after injection of prepolarized [1-13C]pyruvate clearly demonstrated the feasibility of dynamically measuring real-time changes in the formation of 13C-lactate in a TRAMP mouse model in vivo. Through quantitative assessment of parameters obtained from the dynamic 13C-lactate curves that do not rely on any prior model assumptions, we were able to characterize regional, intratumoral and disease-related variations in lactate formation in the TRAMP mouse model. While the presence of such heterogeneity is not surprising based on the histological characteristics of these tumors, this strategy for noninvasive metabolic imaging, which covers the majority of the murine body at a high spatial resolution, could be useful in determining the more aggressive part of the tumor for planning and monitoring novel experimental therapies in TRAMP mice, as well as in differentiating stages of disease that are more or less likely to respond to a given treatment. Although still preliminary, these encouraging results in the TRAMP mouse model demonstrate the potential application of this technique to improve the characterization of human prostate cancers noninvasively.
The acquisition scheme utilized in this study demonstrated the feasibility of detecting changes in pyruvate to lactate conversion in real time at a spatial resolution of 3.125×3.125×4 mm in only 3.5 s. The advantage of this technique lies in the fact that frequency selecting for only the lactate resonance preserves the amount of hyperpolarized pyruvate signal available that would have otherwise been depleted from multiple excitations. This allows for the acquisition of multiple time points at a high resolution. A flyback echo-planar readout trajectory was implemented to acquire eight lines of k-space per excitation, therefore requiring only 64 excitations for the entire volume of 16,384 voxels. For the 10° flip angle utilized in these experiments, this resulted in consumption of at least 63% of the initial available 13C-lactate magnetization and provided enough signal for the duration of the imaging volume. Smaller flip angles (<5°) would preserve >80% of the signal, resulting in a dynamic time course that is proportional to the total amount of lactate accumulation. The images from these low flip-angle acquisitions, however, did not possess sufficient SNR as shown in . Employing larger flip angles (20–25°) should, theoretically, both fully saturate the signal from the previous time point and increase SNR; however, we observed similar SNR values between acquisitions acquired with 10° and 25° flip angles, probably because the slightly lower expected SNR due to the 10° flip angle was balanced by some residual 13C-lactate signal from the previous time point.
The data from this study showed that the maximum amount of lactate produced was increased for more advanced disease in these TRAMP mouse tumors. The interpretation of lactate levels in the prostate region in the wild-type mouse, however, was complicated by the small size of the normal mouse prostate (~0.029 cm3) and the potential for partial voluming of surrounding tissues, as well as the low SNR in this region. In contrast, prostate tumors in TRAMP mice were significantly larger than the lactate imaging voxels, and an array of voxels could be obtained from even the smaller-size tumors. The levels of lactate observed in most tumors appeared to be elevated compared to both normal mouse prostate and the surrounding anatomy, indicating that hyperpolarized lactate has the potential to be a very sensitive biomarker for prostate cancer detection and staging, with much higher levels of 13C-lactate present in mice with more advanced disease. There was no overlap between tumors classified as early-stage and advanced-stage disease using the MLS parameter derived from the curves. Not only were the MLS levels stage dependent, but also the values varied substantially within the tumor, with increases in the standard deviation of MLS values for more advanced-stage disease compared to that of earlier stage disease. This supported the heightened heterogeneity that was visually observed and quantified (in terms of standard deviation) with disease progression. The lower 13C-lactate SNR in mice with earlier disease development may have limited the accuracy of the FWHM parameter measurement. Nonetheless, FWHM values were similar between tumor stages when removing voxels that did not reach a certain noise criteria, as were time-to-peak values, indicating comparable kinetics and time course of conversion between stages. Consequently, the increase in areas observed with tumor stage was driven mostly by the heightened MLS.
There are several main factors affecting the observed 13
C-lactate signal, including the residual magnetization due to the flip angle, the amount of 13
C-pyruvate present, the T1 relaxation times of the labeled pyruvate and lactate, the conversion rate of 13
C-pyruvate to 13
C-lactate, as well as hemodynamics. The MLS in the kidney region was both higher and occurred earlier than the lactate signal from the tumor region in all TRAMP mice, although the different time courses observed in these regions make it difficult to directly compare the amount of lactate formed. Early arrival may be due to a number of factors including higher levels of 13
C-pyruvate present for conversion to 13
C-lactate at earlier time points, an increased rate of conversion from 13
C-pyruvate to 13
C-lactate and/or a faster arrival time of the bolus due to elevated blood flow in the kidneys. The location of maximum lactate in the kidney region corresponded to the central bright region on the T2-FSE image where the long T2 signal is indicative of high fluid content. This suggests that the heightened lactate levels we see early on may be due to washout of lactate from the blood and/or other tissues such as heart and muscle. The conversion of labeled pyruvate to lactate in normal mice followed a time course that was very similar to previous data from similar experiments conducted in TRAMP mice and rat kidneys [16
]. The timing of the MLS in the kidney from the dynamic lactate imaging data coincided within a few seconds of the maximum 13
C-pyruvate concentration from the 1D dynamic spectroscopic data, while the slower time course of 13
C-lactate production in the tumor region peaked when the 13
C-pyruvate concentration has nearly halved (). Therefore, the lactate signal observed in the kidney region was most likely dominated by the arrival of pyruvate, whereas the local conversion of new lactate most likely influenced the observed signal in the prostate/tumor region. This further suggests that the labeled lactate observed in the kidneys was closely related to the blood flow into the organ, while labeled lactate observed within the tumor region was converted from the uptake of 13
C-pyruvate by tumor cells where LDH activity is elevated [24
]. Although it was not the aim of this work to study the metabolism of murine kidneys, the comparison of the dynamic lactate curves between different organ tissues highlights the wealth of information provided by this technique for investigating real-time metabolism.
In conclusion, this study demonstrated the feasibility of using 3D dynamic 13C-lactate imaging to characterize lactate formation in vivo at a high spatial and temporal resolution in a transgenic mouse model of prostate cancer. Parameters quantified from the dynamic 13C-lactate curves elucidated differences within individual tumors as well as between tumors with varying levels of disease progression. This new acquisition offers novel spatially resolved dynamic data of 13C-lactate and may improve our understanding of in vivo lactate formation in these disease models. Further studies will attempt to model the effect of hyperpolarized pyruvate concentration on this data and investigate whether the differences that were seen in the dynamic 13C-lactate imaging data were related to blood supply, substrate uptake and/or the rate of conversion of pyruvate to lactate.