Abstract
Aim: We aimed to investigate whether 3’-deoxy-3’-[18F]fluorothymidine ([18F]FLT) positron emission tomography (PET) can estimate thymidine kinase 1 (TK1) activity after thymidylate synthase (TS) inhibition by 5-fluorouracil (5-FU) in a mouse tumor model. Materials and Methods: Nude mice with HT29 tumors were injected with phosphate-buffered saline or 5-FU (16.7 or 50 mg/kg). Twenty-four hours later, 2-hour dynamic images were acquired after injection of [18F]FLT. In another group of mice with HT29 cells, static PET images were obtained 110 min after [18F]FLT injection. Results: Kinetic parameters related to [18F]FLT retention increased significantly, whereas the de-phosphorylation of [18F]FLT monophosphate decreased significantly. The standardized uptake value (SUVmean) of HT29 tumors correlated significantly with the net influx constant and the distribution volume for phosphorylated [18F]FLT. There was a significant correlation between the tumor SUVmean and TK1 activity. Conclusion: SUVmean at 110-120 min after [18F]FLT, can quantitatively evaluate kinetic parameters and TK1 activity after TS inhibition.
- 3’-deoxy-3’-[18F]fluorothymidine
- positron emission tomography
- thymidine kinase
- 5-fluorouracil
- thymidylate synthase
Thymidylate synthase (TS), a critical enzyme in the de novo synthesis of 2’-deoxythymidine-5’-monophosphate, is a target for TS inhibitors. Pyrimidines (e.g., 5-fluorouracil [5-FU] and 5-fluorodeoxyuridine) and folate cosubstrates (e.g., raltitrexed, pemetrexed, and nolatrexed) inhibit TS as a cancer chemotherapeutic target (1, 2). However, the overall response rate of 5-FU when used as a single chemotherapeutic agent is only 10-15% in advanced colorectal cancer, and 20-40% when used as part of combination regimens with leucovorin, dihydropyrimidine dehydrogenase, methotrexate, and interferon (3, 4). Identifying a sub-group of patients who might benefit from TS inhibitors is, therefore, important.
Quantification of TS inhibition in tumors currently relies on measurements of biochemical changes that accompany TS inhibition. TS inhibition leads to an increase in intracellular pools of the TS substrate dUMP and its corresponding nucleoside 2’-deoxyuridine. Plasma levels of 2’-deoxyuridine, intra-tumoral TS activity, and measurement of dUMP binding sites have been used as pharmacodynamic markers of TS inhibitor activity (5, 6). However, plasma assays do not provide a direct assessment of TS inhibition within a tumor, and tumor biopsy may not always be possible due to significant morbidity.
Substrate cycles, involving phosphorylation of deoxyribonucleosides by kinases and de-phosphorylation of deoxyribonucleoside 5’-phosphates by a nucleotidase, participate in regulation of the size of pyrimidine deoxyribonucleoside triphosphate pools (7). TS inhibitors can cause a rapid decrease in thymidine phosphate pools. Cancer cells respond to decrease in thymidine phosphate pools by an increase in thymidine kinase 1 (TK1), and nucleoside transporter (8) or possibly by decrease in 5’(3’)-deoxyribonucleotidase activity (7, 9) to increase phosphorylation of nucleosides. Positron emission tomography (PET) studies with 3’-deoxy-3’-[18F]fluorothymidine ([18F]FLT) have demonstrated an increase in [18F]FLT uptake after TS inhibition (10-12). Underlying biochemical features associated with the increase in [18F]FLT uptake in tumors were shown to be re-distribution of nucleoside transporters (13) and an increase in TK1 levels (14). The temporary increase in [18F]FLT retention, referred to as the flare effect, may be used in pharmacodynamic measurements of the effect of TS inhibitors (11, 15).
Recently, we showed that [18F]FLT flare was associated with poor treatment response in patients with metastatic colorectal cancer (16). [18F]FLT flare may indicate that a salvage pathway would circumvent the inhibition of pathways of de novo synthesis (17). Evidence of the contribution of the salvage pathway to clinical resistance to TS inhibition may include reduced activity of TS inhibitors in the presence of extracellular thymidine (18), enhancement of the anticancer activity of 5-FU after thymidine transport inhibition (19), and increased activity of TS inhibition in thymidine kinase–deficient tumors (20).
Thymidine metabolism following TS inhibition, and its effect on [18F]FLT metabolism must be elucidated before [18F]FLT metabolism can be adopted as an imaging biomarker for TS inhibition. TK1 is the key regulator of [18F]FLT retention. Nevertheless, the relationship between TK1 activity and [18F]FLT uptake after TS inhibition is not entirely clear. Furthermore, TS inhibition may result in changes in perfusion, and biodistribution of [18F]FLT in addition to nucleoside transporter and enzymes for thymidine metabolism (21). Therefore, we aimed to study the kinetic modeling of [18F]FLT using compartmental analysis to fully-characterize the kinetics of [18F]FLT in tumor tissue after 5-FU treatment in a mouse tumor model. We also assessed whether a simple semi-quantitative method, the standardized uptake value (SUV), can estimate kinetic parameters from validated compartmental models and TK1 after TS inhibition by 5-FU.
Materials and Methods
Radiopharmaceutical preparation. [18F]FLT was prepared from 5’-O-(DMTr-2’-deoxy-3’-O-nosyl-β-D-threopentafuranosyl)-3-N-BOC-thymine by nucleophilic [18F]fluorination in t-amyl alcohol (22). Radiochemical yields (non decay corrected) and purity were 44.9% ± 3.2% and 99.8% ± 1.1%, respectively. The specific activity of the obtained [18F]FLT was 197 ± 53 TBq/mmol.
Cell cultures and tumor models. HT29 cell line was obtained from the American Type Culture Collection (Rockville, MD, USA). Cells were routinely cultured in RPMI-1640, supplemented with 10% heat-inactivated fetal bovine serum, L-glutamine (2 mM), penicillin (50 IU/mL), and streptomycin (50 μg/mL). Cells were maintained at 37°C in an atmosphere of 5% CO2 in air. Cell lysates were prepared as described previously (14). Protein content was determined by the Bradford assay (Bio-Rad, Hercules, CA, USA).
Athymic male Balb-c/nu mice (age, 6 weeks; weight, 20-30 g) were purchased from Japan Shizuoka Laboratory Center (Hamamatsu, Japan). The research protocol was approved by the Institutional Animal Care and Use Committee at the Asan Institute for Life Science. Mice were maintained in accordance with the guideline issued by this committee. Exponentially growing cells (0.2 ml containing 1×107 cells) were inoculated into the right flank of each mouse. Tumor volumes were measured with calipers and calculated according to the following formula: volume=(π/6)abc, where a, b, and c represent the three orthogonal axes of the tumor, respectively. When tumor volumes reached 80-120 mm3 after injection, mice were subjected to animal PET imaging and TK1 activity analysis.
Animal PET imaging with [18F]FLT after vehicle or 5-FU injection. Mice with HT29 tumors were injected intraperitoneally with a bolus of phosphate-buffered saline (n=7), or 5-FU (Choongwae Pharma Corporation, Seoul, Korea; 16.7 or 50 mg/kg; n=6 for each group). 5-FU doses were chosen based on the results of previous studies on the anti-tumor and adverse effects of 5-FU in mice with human xenograft (23, 24). Twenty-four h later, 2-h dynamic PET images were acquired after tail vein injection of 7.4 MBq (0.2 mCi) of [18F]FLT using a commercially available PET system (microPET Focus 120 system, Siemens Medical Solutions, Inc., Knoxville, TN, USA) (25). A 37-frame dynamic protocol (4×3 s, 6×1 s, 7×6 s, 8×30 s, 1×300 s, and 11×600 s) was used for the emission scans with 128×128×95 matrices and voxel size of 0.432×0.432×0.796 mm. The intraperitoneal injection was expected to result in slow absorption of 5-FU from the peritoneal cavity, allowing mice to be exposed to 5-FU for prolonged periods of time (26). Mice were maintained under isoflurane anesthesia during the uptake and scanning periods. A heating pad and heat lamp were used to maintain body temperature at about 37°C using a rectal probe. PET images were reconstructed by filtered back projection using a Hamming filter at a cut-off frequency of 0.5 cycles per pixel. No attenuation correction was applied.
Another group of 19 mice with HT29 tumors were injected with phosphate-buffered saline (n=7), or 5-FU (16.7 mg/kg, n=6; and 50 mg/kg, n=6). Twenty-four h later, a static PET imaging was performed 110 min after injection of 7.4 MBq (0.2 mCi) of [18F]FLT for 10 min. In this group of mice, TK1 activity in tumors was measured after PET imaging.
Analysis of [18F]FLT PET. To obtain time-activity curves for kinetic analysis, cylindrical volumes-of-interest (VOIs) with a diameter of seven pixels and a length of three slices were drawn in the left ventricle (LV) on the early frame. LV time-activity curves corrected for partial-volume effects with the correction factor described were used as the input function (27). The last frame of the dynamic acquisition was used to define the VOI for analysis of [18F]FLT activity in tumors as described previously (28).
To determine the optimal compartment model for [18F]FLT PET in mice with HT29 tumors, we tested a 3-compartment model and its simplified models (29). The three-compartment model includes the parameters K1, K1/k2, k3, k4. Parameters K1 and k2 are rate constants for influx and efflux of [18F]FLT between plasma and the tissue pool, respectively. The rate constant k3 represents TK1-mediated phosphorylation of [18F]FLT. The parameter k4 represents dephosphorylation of [18F]FLT-phosphate. The blood volume fraction (Vb) was included in the modeling. Three-compartment models with reversible phosphorylation (k4≠0, 3C5P) and irreversible phosphorylation (k4=0, 3C4P) were examined. In addition, a 2-compartment model that combined exchangeable and phosphorylated compartments into a single tissue compartment with three parameters (2C3P: K1, k2, and Vb) was also considered.
Data analysis was performed using PMOD software package (version 2.65; PMOD Technologies, Zurich, Switzerland). Tissue time-activity curves were fitted to the models using the nonlinear least-squares method with the Levenberg-marquardt algorithm, which minimizes the weighted sum of squared error between PET measurements and model solutions. Estimated parameters were restricted to the following value ranges: 0.05-1.0 for K1, 0.1-10.0 for K1/k2, 0.001-1.0 for k3, 0.001-0.2 for k4, and 0.01-0.2 for Vb. Initial values for K1, K1/k2, k3, k4, and Vb were set at 0.1, 1.0, 0.1, 0.01, and 0.05, respectively. The net influx constant (KFLT), and the distribution volume for phosphorylated [18F]FLT nucleotides (DVm) were estimated as follows: KFLT=K1 k3 / (k2 + k3) = K1 k3 / (K1/Vd + k3) where Vd is the early volume of distribution of the reversible compartment, which is given by K1/k2 according to previous reports (29, 30), and DVm=K1/k2×(k3/k4).
We chose an adequate compartment model on the basis of the Akaike's information criterion (AIC) and extra sum-of-squares F test.
For semi-quantitive analysis, the same VOIs for kinetic analysis were used. The mean standardized uptake value (SUVmean) of each tumor was calculated using the following formula: SUVmean=(tumor radioactivity in the tumor VOI, measured as Bq/cc × body weight) divided by injected radioactivity. We examined the correlations between kinetic parameters and SUVmean.
TK1 activity assay. TK1 activity was measured as previously described (14). Tumors were lysed in lysis buffer and incubated for 30 min on ice. Lysates were centrifuged at 10,000 g for 20 min at 4°C, and supernatants were centrifuged again at 100,000 ×g for 1 h at 4°C, to separate mitochondrial fractions. After protein contents were determined, cytosolic fractions were assayed for TK1 activities in a reaction buffer containing 10 μM [Methyl-3H]-thymidine (740 GBq/mmol, Perkin Elmer, MA, USA). Mixtures were then incubated at 37°C with gentle stirring, and samples were removed and added to 5 mM ethylenediaminetetraacetic acid to stop the reaction at 15, 30, and 60 min. For sequestration of labeled nucleotides, each sample was spotted onto DE-81 filters (Whatman, GE healthcare, NJ, USA), dried, and washed in 4 mM ammonium formate and 95% ethanol. Radioactivity was measured by liquid scintillation counting with Ultima Gold F scintillation cocktail solution (Perkinelmer, MA, USA). Activities were calculated with linear time-activity curves and are presented as picomoles of phosphorylated thymidine/minute/mg protein.
Statistical analysis. Data are expressed as mean (SD). The Kruskal-Wallis test was used to assess significant differences among groups, and different treatment groups were compared using the Mann-Whitney test with post-hoc Bonferroni correction. Repeated measurements of analysis of variance with three levels (vehicle and two doses of 5-FU) were used to test the changes in SUVmean between the groups at three different time points (50, 80, and 110 min) after vehicle or 5-FU injection. Correlations were assessed by using Spearman's rank correlation coefficient. p<0.05 was considered to indicate statistical significance. All statistical analyses were conducted by using the IBM SPSS Statistics Version 19 for Windows (SPSS Inc., IBM Company, Somers, NY, USA) statistical package.
Results
Compartment modeling. Figure 1 shows axial and coronal PET images obtained 110 min after [18F]FLT injection, and time-activity curves for HT29 tumors and left ventricular blood pool. Mice injected with vehicle had a low level of [18F]FLT retention (Figure 1A). On the other hand, 5-FU injected mice showed a high level of [18F]FLT retention (Figures 1B and C). Mice injected with 50 mg/kg 5-FU had a continuous accumulation of [18F]FLT in the HT29 tumors over the entire 2 h measurement period (Figure 1C).
Time-activity curves were fitted more accurately when Vb was included in the model (3-compartment model with reversible phosphorylation; k4≠0, 3C5P) in both vehicle and 5-FU treatment group. AIC values from the 3C5P models were lower than those from the 3C4P and 2C3P models in all mice except for two (Table I). The results of F tests were similar to those of AIC analysis.
Correlation matrices for the 3C5P model are shown in Table II. In both vehicle and 5-FU treatment groups, K1/k2 and k3 were highly correlated, and in vehicle treated mice there was also a high correlation between k3 and k4.
Parameters for [18F]FLT uptake in HT29 tumors after vehicle or 5-FU treatment. Figure 2 shows the mean and SD of kinetic parameters estimated with the 3C5P models, and SUVmean of HT29 tumors. Kinetic parameters related to [18F]FLT retention (K1/k2 and k3) were significantly different between the groups (p<0.005; Figures 2B and C). K1/k2 and k3 parameters in 5-FU treated mice were higher, but there was no difference in Vb and K1 (Figure 2A). The k4 parameter, defining the dephosphorylation of [18F]FLT monophosphate, was also significantly different (p<0.01) between the groups with a lower value in the 5-FU treatment groups (Figure 2D). As a result, the kinetic macroparameters (KFLT and DVm) increased dose-dependently with 5-FU treatment (Figures 2E and 2F). The SUVmean of HT29 tumors at 50 and 110 min after [18F]FLT administration also showed a dose-dependent increase following 5-FU treatment (Figures 2G and 2H).
The changes in SUVmean at three different time points (50, 80, and 110 min) after injection of vehicle or 5-FU were significantly different between vehicle or 5-FU treated mice (Figures 2G and 2H, p<0.001). Mice injected with 50 mg/kg 5-FU showed significantly increasing SUVmean up to 110 min after the injection, whereas mice injected with vehicle had significantly decreasing SUVmean over time (p<0.001). The SUVmean of HT29 tumors at 110 min after [18F]FLT administration correlated significantly with KFLT (ρ=0.959, p<0.001) and DVm (ρ=0.905, p<0.001).
Typical [18F]FLT PET images of mice with HT29 tumors and time-activity curves. PET images obtained at 110-120 min after injection of 7.4 MBq of [18F]FLT and time-activity curves for left ventricular (LV) and HT29 tumor-bearing regions in vehicle (A), 16.7 (B), and 50.0 mg/kg (C) 5-FU treated mice under isoflurane anesthesia. High [18F]FLT activity in the tumor was seen following 5-FU treatment. Mice treated with 50.0 mg/kg of 5-FU showed a continuous accumulation of [18F]FLT in the tumors. LV input function and tissue time–activity curves were obtained from VOIs drawn on the left ventricle and tumor-bearing regions (white circle) for kinetic analysis.
Correlation of [18F]FLT uptake and TK1 activity. The tumor SUVmean of vehicle, 16.7 mg/kg and 50 mg/kg 5-FU-treated mice-determined by static imaging of [18F]FLT PET, as was significantly different between groups (p<0.005, Figure 3). Similarly, TK1 activity was also significantly different between the treatment groups (p<0.001, Figure 3). A significant correlation was observed between the tumor [18F]FLT uptake (SUVmean) obtained with the static PET images and TK1 activity (ρ=0.890, p<0.001, Figure 3).
Discussion
In the present study, we examined the kinetics of [18F]FLT uptake in HT29 tumors after 5-FU treatment. Our results showed that the tissue time-activity curves of tumor-bearing regions in vehicle and 5-FU-treated groups were best-described by the 3C5P model. The rate constants related to [18F]FLT retention increased with 5-FU treatment. The k4 parameter, representing de-phosphorylation of phosphorylated [18F]FLT, decreased significantly in the 5-FU treated group. The SUVmean showed a high correlation with kinetic parameters from validated compartmental models (KFLT and DVm) as well as TK1 after TS inhibition by 5-FU. Our results suggest that TS inhibition after 5-FU treatment can be assessed quantitatively with [18F]FLT PET.
The most important finding in the present study is that the kinetic parameters related to [18F]FLT retention, including K1/k2, k3, and macroparameters, increased after 5-FU. Although the high levels of co-variance between K1/k2 and k3 parameters make it difficult to obtain independent estimates, our results may be in agreement with previous studies demonstrating an increased TK1 activity after TS inhibition (8, 14). Our ex vivo study of HT29 tumors consistently confirmed an increase in TK1 activity. However, K1, which is representative of nucleoside transport, did not change after 5-FU treatment. We administered 5-FU by intraperitoneal injection to allow mice exposure to 5-FU for a prolonged period of time (26). In our previous work, we found that [18F]FLT PET may be optimal after 24 h of treatment when continuous infusion of 5-FU is given, because of a higher [18F]FLT flare than that after 1-2 h (14). The [18F]FLT flare observed early after TS inhibition may be largely due to the re-distribution of the nucleoside transporter (10, 13). On the other hand, our study suggests that the effect of increased TK1 activity may predominate later. The absence of change in K1 in our study may be explained by [18F]FLT imaging being carried out 24 h after 5-FU injection.
The present study showed that k4, which represents a significant loss of phosphorylated [18F]FLT nucleotides by de-phosphorylation, was lower than k3, but it should not be ignored in the kinetic modeling of HT29 tumors. Importantly, there was a significant decrease in k4 after 5-FU treatment. The decrease in k4 after 5-FU treatment may not be related to a correlation with k3 because no high level of covariance was observed in the 5-FU treatment group. De-phosphorylation of phosphorylated [18F]FLT nucleotides is catalyzed by 5’(3’)-deoxyribonucleotidase. A decrease in 5’(3’)-deoxyribonucleotidase activity leads to increased phosphorylation of nucleosides. The 5’(3’)-deoxyribo-nucleotidase is a regulator of nucleoside and drug metabolism, and a change in 5’(3’)-deoxyribonucleotidase activity is associated with nucleoside analogue activation and drug resistance (31). Our findings suggest a role for 5’(3’)-deoxyribonucleotidase activity in the effect of 5-FU-mediated suppression of the de novo pathway on the salvage pathway. Among the macrokinetic parameters that can be reproducibly measured from kinetic modeling of [18F]FLT (28), KFLT is calculated on the basis of the assumption that k4 can be ignored. Therefore, DVm may be a better macrokinetic parameter with which to assess [18F]FLT kinetics after TS inhibition.
Kinetic and static parameters of [18F]FLT PET in mice with HT29 tumors obtained by 2 h dynamic imaging 24 h after saline, 16.7 and 50 mg/kg 5-FU treatment. Data are expressed as mean±SD. All parameters except K1 were significantly different between the three groups. Only p-values that were significant with post-hoc Bonferroni correction are shown: *p<0.05, **p<0.01, ***p<0.005). Dose-dependent increase in KFLT, DVm and SUVmean was observed with 5-FU treatment.
The present study demonstrates that [18F]FLT PET parameters with 3C5P model are most suitable for evaluating the effect of 5-FU in HT29 tumors. The most simple method used to quantify [18F]FLT uptake is SUV measurement, which is relevant for routine clinical use. SUV analysis, however, can yield misleading results through the contribution of non-phosphorylated [18F]FLT and washout of phosphorylated [18F]FLT by k4 when imaged at later time points (32). Our dynamic [18F]FLT PET revealed a significant decrease in k4, which was accompanied by significantly increasing SUVmean up to 110 min after 5-FU. Therefore, SUV at later time points could be a simple measure of [18F]FLT metabolism after 5-FU treatment. We validated SUVmean at 110-120 min after 5-FU treatment as a measure of the pharmacodynamic effect, by demonstrating a high correlation with KFLT and DVm. Finally, we demonstrated a high correlation of SUVmean at 110-120 min with TK1 activity in HT29 tumors.
Comparison of kinetic models for [18F]FLT PET in mice implanted with HT29 tumors using the Akaike's information criterion (AIC) and extra sum-of-squares F-test.
A limitation of this study is that we did not evaluate [18F]FLT kinetics after a high effective dose of 165 mg/kg 5-FU (13, 21). In contrast to our results, previous studies have found a decreased [3H]thymidine or [18F]FLT uptake in RIF-1 and HT29 tumors associated with decreased cell viability after a high dose of 5-FU (21, 33). Since compromised cell viability may obscure [18F]FLT flare by 24 h after 5-FU injection, we used lower doses of 5-FU. A more comprehensive study that includes the effect of cell viability on [18F]FLT uptake in addition to TS inhibition is needed. Secondly, our results are applicable only to infusional 5-FU treatment. The timing of [18F]FLT PET after TS inhibition and analysis approaches should be optimized and validated to quantitatively evaluate TS inhibition by other TS inhibitors. Thirdly, our mouse model may not be valid in this situation because of high blood thymidine level enough to cause significant competitive inhibition of [18F]FLT uptake. However, induction of TK1 and cell-cycle transition after 5-FU has been demonstrated as a mechanism for [18F]FLT PET in mice models including HT29 in our previous study (14). In mice, left ventricular activity can be used as an image-derived input function because no [18F]FLT metabolites are found in the plasma (27). Mice models may be more suitable for dynamic imaging analysis. Finally, it was beyond the scope of our study to evaluate quantitative [18F]FLT PET parameters as predictive markers of response to 5-FU.
Estimated correlation of tumoral [18F]FLT uptake 110 min after [18F]FLT injection and TK1 activity in mice with HT29 tumors (ρ=0.890, p<0.001). [18F]FLT uptake measurements were obtained 24 h after intraperitoneal injection of vehicle or 5-FU.
Conclusion
The three-compartment model with reversible phosphorylation best described the tissue time-activity curves of the tumor-bearing regions in vehicle and 5-FU treated groups. [18F]FLT flare after 5-FU is characterized by an increase in K1/k2 and k3 and a decrease in k4. The simplified parameter, SUVmean at 110-120 min after injection, can quantitatively evaluate kinetic parameters and TK1 activity after TS inhibition by 5-FU treatment. Our results suggest that [18F]FLT flare, mediated by a mechanism involving increased TK1 activity, can be used to assess the pharmacodynamics of TS inhibition by 5-FU in vivo. More studies are required to provide insight on the quantitative measurement of [18F]FLT flare as a valuable biomarker in selecting patients who are likely to benefit from TS inhibition.
Average correlation coefficients for [18F]FLT kinetic parameters after vehicle or 5-fluorouracil treatment in mice implanted with HT29 tumors.
Acknowledgements
This study was supported by a grant from the Korea Healthcare Technology R&D Project, Ministry for Health, Welfare & Family Affairs, Republic of Korea (A062254).
Footnotes
-
Disclosure
The Authors declare that they have no conflict of interest.
- Received December 23, 2013.
- Revision received January 17, 2014.
- Accepted January 20, 2014.
- Copyright© 2014 International Institute of Anticancer Research (Dr. John G. Delinassios), All rights reserved
