Effects of fasting on serial measurements of hyperpolarized [1‐13C]pyruvate metabolism in tumors

Imaging of the metabolism of hyperpolarized [1‐13C]pyruvate has shown considerable promise in preclinical studies in oncology, particularly for the assessment of early treatment response. The repeatability of measurements of 13C label exchange between pyruvate and lactate was determined in a murine lymphoma model in fasted and non‐fasted animals. The fasted state showed lower intra‐individual variability, although the [1‐13C]lactate/[1‐13C]pyruvate signal ratio was significantly greater in fasted than in non‐fasted mice, which may be explained by the higher tumor lactate concentrations in fasted animals. These results indicate that the fasted state may be preferable for the measurement of 13C label exchange between pyruvate and lactate, as it reduces the variability and therefore should make it easier to detect the effects of therapy. © 2016 The Authors. NMR in Biomedicine published by John Wiley & Sons Ltd.


INTRODUCTION
The introduction of hyperpolarized 13 C-labeled cell substrates has created a renewed interest in the field of MRS by allowing the real-time assessment of tissue metabolism in vivo (1). [1-13 C]Pyruvate ([1-13 C]Pyr) is the most well-studied substrate, having shown promise in preclinical studies in oncology, particularly as an imaging marker for the detection of early response to therapy and as a possible marker for tumor grading and early lesion detection (2)(3)(4)(5). Pyr has also translated to the clinic, with a study in prostate cancer (6). As the end product of glycolysis, Pyr can be reduced reversibly by NADH to generate lactate (Lac), in a reaction catalyzed by lactate dehydrogenase (LDH), or transaminated by glutamate to form alanine, in a reaction catalyzed by alanine aminotransferase (ALT). Irreversible decarboxylation of Pyr to carbon dioxide, in a reaction catalyzed by mitochondrial pyruvate dehydrogenase (PDH), may also occur and, in some tissues, is sufficiently fast to give observable labeling of carbon dioxide and bicarbonate with a hyperpolarized 13 C label (7). Data from hyperpolarized 13 C MRSI measurements are often presented as either signal ratios (8) or apparent exchange rates when the metabolite signals are analyzed as a function of time. Both measurements are affected by Pyr delivery to the tumor, tumor cell uptake and subsequent enzyme-catalyzed exchange (2,(9)(10)(11)(12).
Despite extensive preclinical use of hyperpolarized [1][2][3][4][5][6][7][8][9][10][11][12][13] C]Pyr, with promising future clinical applications in oncology (2,13,14), notably in the serial assessment of treatment response, there is a lack of information regarding the repeatability of these measurements. The determination of repeatability is important for an understanding of whether the observed changes in kinetics of hyperpolarized [1-13 C]Pyr conversion to Lac reflect real effects of therapy or, rather, changes in physiology and/or method variability. 18 F-Fluorodeoxyglucose positron emission tomography ( 18 F-FDG-PET), similarly to hyperpolarized [1-13 C]Pyr, also assesses an aspect of glycolysis. Several studies have evaluated the repeatability of 18 F-FDG-PET, as well as the main factors underlying the observed variability. A major factor affecting 18 F-FDG uptake is high blood glucose levels (15), as a result of competition between glucose and 18 F-FDG for transport into the cell and subsequent phosphorylation. Therefore, it is recommended that, for 18 F-FDG-PET examinations, there is some control over blood glucose levels, either by overnight fasting (16) or by the measurement of blood glucose levels at the time of injection (17).
We have investigated here whether the variability and repeatability of measurements with hyperpolarized [1-13 C]Pyr are also affected by fasting.

Tumor implantation and animal preparation
Tumors were established by subcutaneous inoculation of a suspension of 5 × 10 6 EL4 murine lymphoma cells (in a volume of 100 μL) in the left flank of 8-10-week-old C57/Blk6 female mice. Tumors were allowed to grow for 10 days (volume,~2 cm 3 ; maximum diameter, 1.5 cm), after which the mice were divided into two cohorts: fasted and non-fasted. Fasted mice were deprived of food for 18 h prior to the first 13 C spectroscopic examination. Experiments were conducted in accordance with project and personal licenses issued under the United Kingdom Animals (Scientific Procedures) Act, 1986. Protocols were approved by the Cancer Research UK, Cambridge Institute Animal Welfare and Ethical Review Body.

MRI/MRS protocols
Tumor-bearing mice were anesthetized initially by inhalation of a mixture of O 2 (0.3 L/min) and air (0.9 L/min) containing 3% isoflurane (Isoflo, Abbotts Laboratories Ltd, Maidenhead, UK) and were maintained during the MRS experiment with 1-2% isoflurane in O 2 and air, which was delivered via a facemask. Animals were taped into a cradle to minimize breathing-related motion, and placed in a heated MR probe, which maintained the core body temperature at~37°C (monitored by a rectal probe). Respiratory rate and body temperature were monitored during the experiment using a Biotrig physiological monitor (Small Animal Instruments, Stony Brook, NY, USA). A cannula was inserted into a tail vein and its patency was maintained through the use of heparin diluted in sterile physiological saline (100 U/mL). A 20-mm-diameter curved surface coil (Rapid Biomedical GmbH, Rimpar, Germany) was placed over the tumor. The entire assembly was placed in a 13 C/ 1 H volume coil (Rapid Biomedical GmbH) in a 7-T horizontal bore magnet (Agilent, Palo Alto, CA, USA). The tumor was localized using transverse 1 H images acquired using a spin-echo pulse sequence (TR = 1.5 s; TE = 10 ms; field of view, 40 mm × 40 mm; data matrix, 128 × 128; slice thickness, 2 mm; 15 slices).

NMR spectroscopy of tumors
From the total number of animals (n = 30), paired data were only obtained from 16 animals, as the remainder showed a poor signal-to-noise ratio in the 13 C spectra from one of the examinations, reflecting the difficulty in obtaining good sequential tail vein injections over such a relatively short period of time. Paired 13 C spectroscopic data, acquired 4 h apart, were obtained from eight fasted and eight non-fasted animals ( Table 1). The mice were examined on the same day as the tumor model showed rapid growth (18,19), which could alter the metabolic status of the tumor if the animals were examined on subsequent days. After the first examination, mice were recovered from anesthesia and food was provided only to the non-fasted group (Fig. 1).
Following intravenous injection of hyperpolarized [1-13 C]Pyr (10 mL/kg, 82 mM), single transient spectra from a 6-mm-thick slice through the tumor were acquired using a slice-selective excitation pulse with a nominal flip angle of 10°. One hundred and twenty 13 C spectra were acquired every 2 s beginning 10 s after injection. The processing of 13 C spectroscopic data was performed in Matlab (Mathworks, Massachusetts, USA). The integrated peak intensities of hyperpolarized [1-13 C]Pyr and [1-13 C] Lac were fitted to the modified Bloch equation for two-site exchange to calculate the rate constants k L and k P and the apparent spin-lattice relaxation rates, as described previously (2).
where L Z and P Z are the z magnetizations of the 13 C nucleus in the Lac and Pyr carboxyl carbons, respectively, R 1,L and R 1,P are the spin-lattice relaxation rates (1/T 1 ), and L ∞ and P ∞ are the equilibrium magnetizations of Lac and Pyr, respectively, which are effectively equivalent to their concentrations. The relaxation rates for Lac and Pyr were assumed to be the same. We have shown previously that this assumption has little effect on the fitted rate constants [see supplementary fig. 2 in ref.
(2)]. L ∞ /P ∞ was obtained from the ratio of the fitted rate constants. The 13 C-labeled Lac/Pyr ratios were calculated from L Z /P Z at 30 s.

Dynamic contrast-enhanced MRI (DCE-MRI)
Separate cohorts of mice, fasted (n = 7) and non-fasted (n = 7), were used for DCE-MRI measurements of tumor perfusion, in which the mice were injected with Gd 3+ -DTPA (diethylenetriaminepentaacetic acid; 200 mmol/kg; Magnevist, Bayer Schering Pharma, Leverkusen, Germany) ( Table 2). The imaging protocol simulated the protocol used for the 13 C hyperpolarized measurements. Briefly, for each cohort (seven fasted, seven nonfasted), four mice were injected with contrast agent (1st scan) and three mice were injected with unlabeled Pyr (82 mM, 10 mL/kg), followed by injection of contrast agent 4 h later (2nd scan) (Fig. 1). Before each injection, animals were submitted to at least 30 min of anesthesia, which is the same time for which they were anesthetized for the hyperpolarized [1-13 C]Pyr measurements. DCE-MRI data were acquired using a T 1 -weighted spin-echo pulse sequence, as described previously (12). Briefly, first an inversion recovery fast low-angle shot (FLASH) pulse sequence was used to measure the native spin-lattice relaxation rates (R 1 = 1/T 1 ). These inversion recovery data were fitted, pixel-bypixel, to a mono-exponential function to obtain a pre-contrast R 1 map. A dynamic T 1 -weighted spin-echo pulse sequence was then used to collect baseline images (6 or 10) prior to injection of contrast agent. Diluted Gd 3+ -DTPA in sterile saline (0.9% wileyonlinelibrary.com/journal/nbm sodium chloride) was then injected as a bolus through a tail vein catheter over 2-3 s and dynamic T 1 -weighted spin-echo images were acquired for 10 min post-injection. These images were then converted to R 1 relaxation rate maps, as described previously (12). Gd 3+ -DTPA concentration curves were determined for all tumor-containing slices with a manually delineated region of interest (ROI) in the tumor and also in adjacent thigh muscle.
The area under the uptake curve up to 45 s (AUC 45) was calculated for each ROI.

Blood and tissue collection and analysis
In a separate set of experiments (Table 3), tumor-bearing mice (n = 24) were divided into two groups: fasted (n = 12) and nonfasted (n = 12). In each group, the mice were further divided into three subgroups: Group 0, anesthetized for 30 min; Group 1, 30 min of anesthesia plus a bolus injection of 10 mL/kg of 82 mM unlabeled Pyr; Group 2, same as Group 1, plus a 4-h recovery period, a second period of anesthesia for 30 min and another injection of 10 mL/kg of 82 mM unlabeled Pyr (Fig. 1). The timing of anesthesia and Pyr injection was the same as for the 13 C MRS data acquisition protocol (Fig. 1). Blood was collected rapidly by cardiac puncture in all mice from each group (starting at 25 s after Pyr injection) and the tumors were quickly excised and freeze-clamped 30 s after Pyr injection. Blood samples were centrifuged (18 800 g) at 4°C for 5 min in heparin-coated Eppendorf tubes (~5 μL of 1000 U/mL). Pyr was analyzed using a colorimetric assay kit (AbCam, Cambridge, UK) and Lac by an enzymatic assay kit (Siemens Healthcare, Sudbury, UK), using a Siemens Dimension RxL analyzer. Perchloric acid extracts were prepared from the tumor tissues of Groups 1 and 2 for both fasted (n = 6) and non-fasted (n = 6) mice using ice-cold 7% perchloric acid (1: 8 w/v), which were then neutralized with KOH, lyophilized and dissolved in 99.9% deuterium oxide. High-resolution 1 H NMR spectra of plasma and tumor extracts were obtained at 14.1 T (25°C, pH 7.2) in a Bruker 600-MHz NMR spectrometer (Bruker, Ettlingen, Germany) using a 5-mm probe. The acquisition conditions were as follows: 90°pulses; spectral width, 7.3 kHz; acquisition time, 4.5 s; 32 k data points; 64 transients; recycling time, 12.5 s. Chemical shifts were referenced to 3-(trimethylsilyl)-2,2′,3,3′-tetradeuteropropionic acid (TSP) at 0.0 ppm, which was added to the sample at a concentration of 5 mM. Data were analyzed using ACDSpecManager (ACD/Labs, Bracknell, UK). The free induction decays were zero-filled twice and multiplied by an exponential function prior to Fourier transformation. Peak integrals were normalized to the integral of the TSP resonance.

Statistical analysis
Data were analyzed using SPSS (v21, IBM SPSS, Chicago, IL, USA) and GraphPad Prism v6 (GraphPad Software, San Diego, USA). The repeatability of measurements of k P and the [1-13 C]Lac/ [1-13 C]Pyr signal ratio, which was measured at 30 s after injection, were assessed from the two imaging examinations performed 4 h apart. Scatter plots (Measurement 1 versus Measurement 2) and Bland-Altman plots (difference versus mean) were generated. The concordance correlation coefficient (CCC), within-subject (intra-individual) standard deviation (SD) and coefficient of variation (COV) were calculated as described in refs. (20,21). The COVs of k P and the [1-13 C]Lac/[1-13 C]Pyr signal ratio were calculated, for each parameter, as the SD divided by the parameter mean, and expressed as a percentage. This was calculated for each individual and then averaged. COV can be used as a statistical measure of reliability (22) and has been used previously to assess reliability in FDG-PET studies (21). Data were reported as mean ± SD, unless stated otherwise. Statistical significance was tested with Prism using a two-tailed Student's t-test or analysis of variance (ANOVA) (post-hoc test: Tukey) when appropriate. Paired t-tests were used when comparing paired data from the same mouse. The results were considered to be significant when p < 0.05.

Repeatability of 13 C MRS measurements with hyperpolarized [1-13 C]Pyr
Intravenous injection of hyperpolarized [1-13 C]Pyr (10 mL/kg, 82 mM) resulted in tumor signals from [1-13 C]Pyr and [1-13 C]Lac that were similar in intensity to those observed previously in this tumor model (2). The apparent rate constant describing the conversion of Pyr to Lac (k P ) and the [1-13 C]Lac/[1-13 C]Pyr signal ratios at 30 s were calculated. Previous studies in this tumor model have shown that there is substantial labeling of both the Pyr and Lac pools at 30 s after Pyr injection (2). The repeatability of measurements made 4 h apart is shown in Fig. 2 and Table 1. Non-fasted animals showed poorer correlations and higher mean differences between successive measurements than fasted animals. Repeatability was also assessed by calculating the COV for the mean k P and mean [1-13 C]Lac/[1-13 C]Pyr signal ratio using data from both time points, i.e. a total of 16 measurements for each group of animals (Table 1). Animals in the fasted state showed lower intra-individual variability than those in the non-fasted state, with significantly lower COV for the mean k P (p = 0.01). The higher metabolic rate of mice versus Tumor/muscle (thigh) ratio for the area under the contrast agent concentration curve (AUC) at 45 s after injection of contrast agent for fasted and non-fasted EL4 tumor-bearing mice. Acquisitions were performed at the same time points as for the hyperpolarized [1-13 C]pyruvate study, with mice from Group 2 having received a dose of unlabeled non-polarized pyruvate (10 mL/kg, 82 mM) and 30 min of anesthesia 4 h before the contrast agent-enhanced 1 H images were acquired. Mean ± standard error of the mean.
wileyonlinelibrary.com/journal/nbm humans means that 4 h is considered to be a substantial fasting period for these animals and equivalent to an overnight fast in humans. Moreover, in the 4-h period between the MRI examinations, the mice were expected to feed multiple times (24). The mean k P and [1-13 C]Lac/[1-13 C]Pyr signal ratio calculated from the first and second examinations showed no evidence of a difference for both fasted (n = 8) and non-fasted (n = 8) animals. However, when using data from all the mice (n = 30), a higher mean k P and [1-13 C]Lac/[1-13 C]Pyr signal ratio were observed in fasted mice than in non-fasted mice, with the difference in the mean [1-13 C]Lac/[1-13 C]Pyr signal ratio observed between the first measurements in the fasted (n = 11) and non-fasted (n = 17) cohorts achieving statistical significance (p = 0.02) ( Table 1).

DCE-MRI
As both the [1-13 C]Lac/[1-13 C]Pyr signal ratio and k P may be dependent on Pyr delivery (12), tumor perfusion was assessed using DCE-MRI. There was considerable variation in contrast enhancement between the different animals, as assessed from the ratio of the area under the contrast agent uptake curves at 45 s for the tumor and adjacent muscle. There was no evidence of perfusion changes between fasted and non-fasted mice at the two time points (corresponding to the two measurements with hyperpolarized [1-13 C]Pyr) ( Table 2).

Measurement of metabolite concentrations in tissue extracts
Tissue Lac concentration can affect 13 C label exchange following the injection of hyperpolarized [1-13 C]Pyr and, consequently, the [1-13 C]Lac/[1-13 C]Pyr signal ratio and k P (2,25). A significantly lower tumor Lac concentration was observed in non-fasted than in fasted mice after the first Pyr injection (Table 3). There was no evidence of differences in tumor alanine concentrations (Table 3) or in blood Lac levels between fasted and non-fasted mice ( Table 3). The blood Pyr concentration was significantly higher in fasted mice after the first Pyr injection relative to non-fasted animals ( Table 3). There was no evidence of a difference between the groups in terms of tumor weight; however, there was a significant difference in animal weight between fasted and non-fasted animals at baseline (Group 0) ( Table 3).

DISCUSSION
The repeatability and variability of measurements with hyperpolarized [1-13 C]Pyr were determined in fasted and nonfasted EL4 tumor-bearing mice. The initial values for the [1-13 C] Lac/[1-13 C]Pyr signal ratio and k P were more similar to the values measured 4 h later in fasted mice than in non-fasted animals (Fig. 2), giving higher concordance coefficients in fasted animals ( Table 1). Both parameters showed higher COVs in non-fasted mice (Table 1), where the unpredictable dietary intake before and between examinations may have contributed to the higher intra-individual variability in this group. The lower COV in fasted mice (Table 1) is within a range similar to that seen with other clinical imaging techniques, such as FDG-PET (26).
There was a significantly higher mean [1- 13  Tumor concentration Alanine (μmol/g wet tissue) 3.9 ± 1.4 (n = 3) 2.9 ± 1.1 (n = 3) 2.6 ± 0.9 (n = 3) Significantly different from non-fasted 1.  (Table 3). This is consistent with previous in vitro and in vivo studies (2,25,27), which have shown that the exchange rate increases with an increase in Lac concentration. There was no evidence for a difference in tumor perfusion between fasted and non-fasted animals or between the first and second scans in these animals (Table 2), although the error in these measurements was large. Therefore, the higher tumor Lac levels in fasted mice implies that these may have been caused by an increase in tumor glycolysis. This has been observed previously in tumors of fasting animals, where the tumor preferentially consumes glucose in comparison with other tissues and produces increased amounts of Lac (28). The fasted animals also showed higher blood Pyr levels after the first injection of [1-13 C]Pyr. This may be explained by increased gluconeogenesis, which results in increased levels of blood Pyr and alanine (29,30), and occurs independently of the presence of a tumor (31,32). The variability of hyperpolarized [1-13 C]Pyr metabolism may depend on the organ and/or type and grade of tumor being assessed. Hyperpolarized [1-13 C]Pyr was found to give reproducible results in the assessment of normal kidneys of non-fasted rats; however, the fasted state was not examined in this study (33).
In conclusion, if hyperpolarized [1-13 C]Pyr is to be used in the clinic, especially for monitoring response to therapy or disease progression using serial measurements, a knowledge of repeatability is essential. The results presented here suggest that, similar to what is already performed in some clinical centers for 18 F-FDG-PET examinations (16,17,34), fasting prior to the MR examination might be preferred because of the more uniform and constant metabolic status and lower intra-subject variability.