Applications of PET imaging with the proliferation marker [¹⁸F]-FLT

Abstract

[¹⁸F]-3′-fluoro-3′-deoxythymidine (FLT) is a nucleoside-analog imaging agent for quantifying cellular proliferation that was first reported in 1998. It accumulates during the S-phase of the cell cycle through the action of cytosolic thymidine kinase, TK1. Since TK1 is primarily expressed in dividing cells, FLT uptake is essentially limited to dividing cells. Thus FLT is an effective measure of cell proliferation. FLT uptake has been shown to correlate with the more classic proliferation marker, the monoclonal antibody to Ki-67. Increased cellular proliferation is known to correlate with worse outcome in many cancers. However, the Ki-67 binding assay is performed on a sampled preparation, ex vivo, whereas FLT can be quantitatively measured in vivo using positron emission tomography (PET). FLT is an effective and quantitative marker of cell proliferation, and therefore a useful prognostic predictor in the setting of neoplastic disease. This review summarizes clinical studies from 2011 forward that used FLT-PET to assess tumor response to therapy. The paper focuses on our recommendations for a standardized clinical trial protocol and components of a report so multi center studies can be effectively conducted, and different studies can be compared. For example, since FLT is glucuronidated by the liver, and the metabolite is not transported into the cell, the plasma fraction of FLT can be significantly changed by treatment with particular drugs that deplete this enzyme, including some chemotherapy agents and pain medications. Therefore, the plasma level of metabolites should be measured to assure FLT uptake kinetics can be accurately calculated. This is important because the flux constant (K_FLT) is a more accurate measure of proliferation and, by inference, a better discriminator of tumor recurrence than standardized uptake value (SUV_FLT). This will allow FLT imaging to be a specific and clinically relevant prognostic predictor in the treatment of neoplastic disease.

While molecular imaging using positron emission tomography (PET) remains dominated by use of FDG, there has been a concerted effort to develop new PET imaging agents to quantify specific metabolic processes such as cell proliferation, receptor density and function, and other characteristics that have essential implications for therapeutic decision-making. This review focuses on cellular proliferation in tumors, a process that is often inferred from FDG-PET although there are many reasons for tissues, including tumors, to accumulate and retain this glucose analog. Changes in images of cellular proliferation should accurately reflect response to cytotoxic as well as cytostatic therapy. Changes in cellular energetics measured by FDG may reflect response to therapy, but this is non-specific; for example, apoptosis requires additional energy, as does the action of ABC transporters that are exporting some cancer drugs.

Researchers have focused on uptake of nucleosides that are incorporated into DNA as the most specific way to measure cellular proliferation. Because some nucleosides are also incorporated into RNA, the nucleoside of choice for this assay is thymidine, the only nucleoside that is not a constituent of RNA. Thus tritiated thymidine was one of the first laboratory tests for assaying cellular proliferation in cultured cells and in small animals. Methods were also developed to label thymidine with ¹¹C in either the methyl position or in the urea carbonyl and these radiopharmaceuticals made important early contributions to PET imaging of tumors.^{1, 2} It is important to appreciate that thymidine is taken up by cells, sequentially phosphorylated to the nucleotide, thymidine triphosphate (TTP) by the exogenous (salvage) pathway, then incorporated into DNA. However, unlabeled TTP can also enter the DNA synthetic machinery via methylation of deoxyuridine monophosphate, the endogenous (de novo) pathway, followed again by sequential phosphorylation. Tumors will vary in the extent to which they depend more on one or the other of these pathways. Thus, it is important to remember that thymidine imaging only measures the exogenous pathway and drugs that inhibit the endogenous pathway, such as 5-fluorouracil, may lead to a “flare” increase in dependency on exogenous thymidine.³

Thymidine is rapidly degraded to thymine and deoxyribose phosphate by thymidine phosphorylase, which is present in the cytosol and in blood plasma. The short nuclear half-life of¹¹C and rapid metabolism of thymidine spawned a search for an analog labeled with¹⁸F and that was not rapidly metabolized. This search coincided with development of nucleoside analogs for treatment of HIV based on molecules that were modified in the ribose part of the parent molecule. Azidothymidine (AZT) was initially developed as a cancer therapeutic, then was found much more useful for HIV-AIDS therapy; it served as the prototype molecule for developing¹⁸F labeled thymidine analogs for PET imaging. In fact, two¹⁸F-labeled thymidine nucleosides have been developed and evaluated as potential PET radiopharmaceuticals: FLT and 2′-deoxy-2′-[¹⁸F]fluoro-5-methyl-1-beta-D-arabinofuranosyl uracil (FMAU) (Figure 1). Fluorinated nucleosides are transported into cells via human nucleoside transporters.

Figure 1.

Structure of nucleosides used for imaging cellular proliferation. Thymidine can be labeled with ¹¹C at either the pyridine methyl position or the urea carbonyl.

FLT has an F-for-OH substitution at the 3′ position, making it a chain terminator and poor substrate for DNA polymerase. [¹⁸F]-3′-fluoro-3′-deoxythymidine (FLT) is a good substrate for cytosolic thymidine kinase (TK-1), which is up-regulated by about 10-fold during the S-phase of the cell cycle. FMAU is incorporated into DNA, but unlike FLT, FMAU is a better substrate for mitochondrial thymidine kinase (TK-2), which is not regulated during the cell cycle.^{4, 5} TK enzymes add a charged phosphate group to the nucleoside, much like hexokinase enzymes phosphorylate intracellular FDG and cause it to be retained inside cells for the duration of image acquisition.

This review will focus entirely on FLT, [¹⁸F]-3′-fluoro-3′-deoxythymidine. FLT was developed as a PET imaging agent through a joint effort by Shields and Grierson at the University of Washington PET Program, and Dohmen and Machulla at Eberhard-Karls University, Tübingen, DE.⁶ Since the introduction of FLT in 1998, its preclinical and mechanistic background has been reviewed ^7,8 and numerous comprehensive reviews of FLT as a PET tracer for quantifying cellular proliferation have been published.^{9, 10} The reviews by Soloviev⁹ and Sanghera¹⁰ showed considerable insight regarding the theoretical and practical limitations of FLT-PET and, in the interest of brevity, we have elected to limit this review to an update to these earlier reviews. Table I presents a summary of the most recent clinical trials evaluating FLT-PET as a predictor of tumor response to therapy and follows a similar format to the tables in the previous reviews. Additional considerations of this review will evaluate some of the technical challenges of FLT-PET and how they impact interpretation of scan results, especially with respect to assessment of outcome.

Table I.

Data analysis.

Reference	No. of Patients	Cancer	Correlation Ki-67-FLT	Treatment	Response: Timing/criteria	Study Results
Bholi 25 (Chandigarh, India)	15	Stage IIIB-IV NSCLC	Not Measured	Oral 1st or 2nd/3rd line EGFR-TKI	3 weeks responders or non-responders or at a steady state according to disease	In responders and non responders, treated for 3 weeks, both OS and PFS were better predicted by FDG than by FLT Neither the baseline value nor the change in SUVmax after one cycle of NAC were able to predict response
Woolf 26 (Northwood, UK.)	20 women	Breast	Correlation between pre- chemotherapy Ki-67 and SUVmax 0.604 (P=0.006) Baseline SUVmax measurements of FLT-PET-CT were significantly related to Ki-67	1 cycle neoadjuvant chemotherapy (NAC)
Lee 27 (Goyang, South Korea)	61	Non-Hodgkin lymphoma	Not measured	1 cycle of chemotherapy	18F-FLT pre and post 1 cycle chemotherapy	In aggressive NHL, early interim FLT PET is a significant predictor of (PFS) and (OS). Early (18)F-FLT PET has a potential to identify delayed response and nonfavorable prognosis despite achieving a clinical complete response.
Frings28 (Amsterdam, The Netherlands)	23	NSCLC	Not Measured	None	Non-linear regression showed better fits for the irreversible model Metabolite corrected and plasma blood ratio corrected input function gave in high correlations between SUV and Patlak K i (Pearson coef 0.86–0.96, P<0.001).	Plasma-to-blood ratio correction, metabolite correction and calibration improved the correlation between SUV and Patlak K i significantly, indicating the need for these corrections when K i is used to validate semi-quantitative measures, such as SUV.
Hong 3 (Seoul, Republic of Korea)	18	Metastatic colorectal	Not Measured	Chemo 5-FU	FLT scans obtained before therapy: 24 hours after 5-FU (day 2) and 48 hours after completion of therapy. SUVMAX F-FLT was measured.	The FLT flare observed during 5-FU infusion was associated with poor treatment response The degree of FLT flare might predict outcome infusional 5-FU- based chemo.
Nakajo 29 (Kagoshima, Japan)	28	Colorectal	Not Measured	Surgical Resection	Associations between SUVmax and patho by evaluated using the Mann-Whitney U or Kruskal-Wallis. Differences in diagnostic indexes for detecting nodal metastasis between FLT and FDG estimated using the McNemar exact or χ(2) test.	FLT has the same potential as FDG for diagnosis of primary and nodal foci of CRC despite significantly lower FLT uptake in primary foci.
Tsuyoshi 30 (Fukui, Japan)	3	Recurrent ovarian	Not measured	Gemcitabine-based secondary chemo	FDG vs FLT	SUV FLT decreased earlier than with 1FDG and was better correlated with reduction in size by CT. FLT could become a new standard for monitoring response to this treatment F-FLT PET/CT imaging is not recommended for pretreatment assessment of metastatic gastric cancer as it is not competent enough to evaluate liver and bone metastases; moreover, the high background hepatic uptake may cover the gastric primary tumors located adjacent to the liver.
Zhou 31 (Shanghai, China)	39	Metastatic gastric cancer	Not measured
Hoshikawa 32	88	Head and neck squamous cell cancers (HNSCCs)	Not measured	Chemotherapy/surgery	Final diagnoses of second primary cancer and distant metastasis were established on the basis of histological findings or clinical follow-up.	FLT not recommended to replace FDG for pretreatment metastasis staging in HNSCC because of lower sensitivity and higher background in the liver and marrow. It might provide additional diagnostic specificity and biological information. Study suggests ENT1 expression might not reflect accumulation of FLT in vivo due to BBB permeability.
Shinomiya 33 (Kagawa, Japan) [Abstract Only]	21	Newly diagnosed glioma	Not measured	None
Hoshikawa 34 [Kagawa, Japan)	30	HNSCC	Not Measured	Concurrent chemoradiotherapy	PET images evaluated as SUVmax	FLT PET signal change preceded FDG PET change and the increase of FLT uptake after the therapy can imply recurrence or residual tumor.
Hoeben 35 (Nijmegen, The Netherlands)	48	Head and neck cancer	Not measured	Radiotherapy or chemoradiotherapy	SUVmax for the hottest voxel surrounding 8 voxels in 1 slice calculated	A change in FLT early during radiotherapy or chemoradiotherapy is a strong indicator for long-term outcome. low FLT (SUVmax <3.0) associated with longer survival (10.3 months (0–23.3 months, 95% CI) compared to high FLT (3.4 months (0–8.1 months, 95% CI) (p = 0.027). a cutoff of 20% or 30% for metabolic response measured by changes of TLG and TLP provided suitable response for prediction..
Scheffler 36 (Cologne, Germany)	40	Metastatic NSCLC	Not Measured	Erlotinib
Kahraman 37 (Cologne, Germany)	30	Stage IV NSCLC	Not Measured	Erlotinib
Enslow 38 (Utah, USA)	15	Recurrence of treated grade 2 glioma or worse with a new focus of Gd contrast enhancement on MRI			For FDG PET, SUVmax and the ratio of lesion SUVmax to the SUVmean of contralateral white matter were measured. For FLT PET, SUVmax and Patlak-derived metabolic flux parameter Kimax were measured for the same locus. A 5-point visual confidence scale was applied to FDG PET and FLT PET. Receiver operating curve analysis was applied to visual and quantitative results. Differences between recurrent tumor and radiation necrosis were tested by Kruskal-Wallis analysis. On the basis of follow-up Gd-enhanced MRI, lesion- specific recurrent tumor was defined as a definitive increase in size of the lesion, and radiation necrosis was defined as stability or regression.	Both quantitative and visual determinations allow accurate differentiation between recurrent glioma and radiation necrosis by both FDG and FLT PET. In this small series, FLT PET offers no advantage over FDG PET.
Frings 39 (Amsterdam, The Netherlands)	14	NSCLC	Not Measured	Pemetrexed: a thymidylate synthase (TS) inhibitor		Measuring TS-inhibition 4 h after therapy revealed a non-systematic change in FLT uptake within the tumor. No association with response, (TTP) or (OS).

Tabulation of recent clinical studies of FLT-PET

Results from clinical studies confirm that detecting early changes in FLT uptake is useful for monitoring treatment response. However, a minority of reports caution that FLT uptake in some clinical situations did not accurately reflect cellular proliferation. In some of these human studies, the standard assay of reference for cell proliferation was the Ki-67 assay. Ki-67 is a human protein expressed in the nucleus of dividing cells regardless of the pathway for DNA synthesis. Because Ki-67 is an ex vivo cell biomarker, the assay is limited by the tissue sampling error, rendering it an imperfect standard, particularly in heterogeneous tumors.

Challenges in interpretation of FLT-PET imaging

Mechanisms of FLT distribution, metabolism, and excretion within tissues of the body can impact the FLT-PET image and each of these metabolic processes can be altered by therapy. Thus, to the extent that differences in FLT-PET imaging results before and after therapy are useful as an indicator of early response, the interpretation of scans needs to consider factors other than cellular proliferation supported by the exogenous pathway that might be affected by therapy.

Any thymidine analog measures the exogenous pathway to production of TTP, however, the majority of the thymidine nucleotide in tumor DNA often comes from the endogenous pathway. At best, therefore, imaging of thymidine metabolism only reflects one arm of the DNA synthetic pathway. To the extent that treated tumors change their reliance on the endogenous (de novo) route to the nucleotide, the change in FLT uptake after treatment could be misleading. For example, 5-fluorouracil (5-FU) specifically targets the endogenous DNA pathway, thus dependence on the alternative pathway is more than a theoretical limitation.¹¹

Transport of FLT

After an intravenous injection, FLT crosses cell membranes and enters tissues by nucleoside transporters, both equilibrative (hENT1 and hENT2) and concentrative (hCNT).¹² A passive diffusion mechanism has also been suggested¹³ but it is small compared to nucleoside transport. Some early studies with [³H-methyl]-thymidine showed a remarkable correlation with blood flow at 20 seconds after administration in mice and dogs,¹⁴ suggesting that the authentic nucleoside was a freely diffusible tracer. The correlation of uptake of the tritiated analog with blood flow as measured by [¹⁴C]-iodoantipyrine was demonstrated in a large number of normal organs and in spontaneous tumors with both species, but brain uptake was less that would be expected based on blood flow. However there was measureable uptake in healthy brain above what would be expected based on its intravascular space. A similar study of the role of passive diffusion in uptake of FLT has not been reported and would be an important contribution because it has implications for modeling FLT in the brain. A limited literature supports hENT1 as the most abundant nucleoside transporter influencing FLT uptake in tumors.¹⁵ The order of magnitude increase observed in FLT uptake 24 hours after esophageal carcinoma cells were exposed to 5-FU plus methotrexate has been ascribed to redistribution of the hENT1 to cytosol.¹¹

Metabolism of FLT

Unlike thymidine, FLT is not degraded at the glycosidic linkage by thymidine phosphorylase (Figure 2). This has led some authors to claim that FLT is stable in plasma,⁹ although that statement could be misleading because FLT is a substrate for glucuronidation in the liver.

Figure 2.

Diagram of thymidine (A) and FLT (B) uptake and metabolism, including transport, phosphorylation by TK1, synthesis of thymidine from uridine by TS, and reaction with DNA polymerase. Thymidine, but not FLT, is a substrate for DNA polymerase. FLT is metabolized to its glucuronide in the liver and excreted by the kidney (C). After some cancer drugs treatments the glucuronosyl transferase reaction may be inhibited.

About one-third of the¹⁸F in circulation at 1 hour after administration of FLT is diverted to the glucuronide.¹⁶ Therefore, kinetic data analysis requires correction of the blood input function for levels of this metabolite. In a recent report, the description of the synthesis of a reference standard for the glucuronide permits any laboratory to calibrate an analytical system to measure this primary metabolite of FLT in blood samples.¹⁷ In future studies, it will be particularly important to measure and report FLT-glucuronide levels as a function of time after FLT administration for patients both before and after treatment with chemotherapy drugs. Chemotherapy drugs as well as several pain medications can dramatically influence the level/activity of glucuronosyl transferase enzymes. This drug-induced increase in bioavailable (unconjugated) FLT acts to favor an increase in FLT accumulation in the cells. If one were unaware of this increased bioavailability of FLT, the persistence of elevated SUV after therapy could be misinterpreted as a treatment failure. Thus, measurement of the ratio of conjugated to unconjugated FLT in plasma will allow more reproducible measures of cellular uptake and therefore a more accurate measure of chemotherapy efficacy. Lastly, there is some evidence that FLT nucleotides are subject to enzymatic dephosphorylation¹⁸ that allows labeled FLT to leave the cell, although it is unlikely that this process significantly interferes with interpretation of human imaging studies. The same report suggests that FLT monophosphate can directly efflux from cells.

Data analysis for FLT-PET

SUV and its role

The compilation of studies summarized by Soloviev⁹ and Sanghera,¹⁰ and updated in Table I, are consistent with a general correlation between SUV for FLT and Ki-67.¹⁹ In addition, changes in SUV have been found to correlate with response to treatment in brain, lung, and breast cancers.²⁰ However, it is disappointing to find that many of the more recent clinical studies have not reported measurement of any correlates of response that will be essential to test the clinical significance of FLT. Manuscript reviewers and editors should be vigilant in verifying that appropriate outcomes variables are reported as correlates of FLT-PET imaging (Table I).

Role of kinetic analysis

The image intensity of FLT at any time after injection is related to both the rate of delivery of tracer to that region, a process that is dominated by plasma concentration, blood flow and transporters, and tracer retention due to phosphorylation by TK1. This biochemistry is conceptually analogous to that of FDG, where the delivery is driven by blood flow and GLUT transporters, while phosphorylation is under the control of a family of hexokinase enzymes, and phosphorylation is the rate limiting step in the tracer uptake. A mathematical model similar to that for FDG has been described and validated in human studies, and this model showed that the FLT flux parameter, K_FLT, had a strong correlation with Ki-67²¹ even though the former reflects the exogenous pathway but the latter reflects both exogenous and endogenous synthesis.

Kinetic analysis is essential for accurate interpretation of brain imaging because FLT and thymidine are restricted at the blood-brain-barrier, resulting in a transport-dominated SUV image because transport of FLT is the rate-limiting path to uptake. This is one reason why more detailed studies would be helpful to test whether there is any uptake of FLT by passive diffusion, allowing phosphorylation to be the determining step in the metabolism of FLT. Only a few reports have used kinetic analysis for quantification of FLT-PET for imaging non-brain regions.²² In non-brain regions, FLT uptake is not transport limited; phosphorylation is the rate limiting process so SUV analysis is more justifiable. In general, correlation with Ki-67 is about the same for SUV ^{10, 16} although more as it is for K_FLT in body imaging publications on this topic are urgently needed to test whether that correlation breaks down in tumors dominated by endogenous synthesis or when therapy is directed at that pathway.

The most compelling example of the benefit of kinetic analysis involved a study of previously treated primary brain tumors where the clinical question was to distinguish post radiation treatment effects from true tumor recurrence, a distinction that cannot be made reliably by MR imaging alone.²³ The clinical question is whether or not a patient’s previously treated brain tumor is progressing, or do the MR findings spuriously suggest progression while the patient is actually responding to their therapy.

Accurate discrimination between these two groups remains a particularly important and challenging question because these patients may be candidates for experimental therapeutic intervention trials for primary brain tumor recurrence. Patients who are not truly progressing despite having worrisome MRI findings would effectively reduce the power of clinical trials to detect any clinical benefit from experimental therapies if those subjects were not excluded from enrollment, as they are in effect responding to prior therapy despite the misleading impression from anatomic MRI imaging alone. Moreover, a patient treated in the absence of true progression is at risk for increased side effects without deriving significant clinical benefit.

FDG-PET and FLT-PET were used in this study of potentially recurrent brain tumors and SUV as well as kinetic analysis with metabolite correction were compared for their ability to differentiate between the two groups.²³ For FDG a standard clinical protocol was followed and analysis was SUV, which includes effects of transport as well as flux. The results from this study supported the added benefit of kinetic analysis (Figure 3). The SUV_FLT values overlapped for the patients with radiation necrosis and those with true recurrent tumor. The same degree of overlap was seen with the SUV_FDG analysis. A clear difference was evident in the parametric images from kinetic modeling; the FLT transport parameters showed some overlap between groups but K_FLT showed no overlap between these two groups. These results showed that rigorous data analysis applied to FLT-PET imaging is a promising technique for identifying recurrent tumor, suggesting a significant role for FLT-PET imaging in assessing primary brain tumor treatment response. This result needs to be tested further for confirmation and, if found valid, would create a clinically significant indication for FLT-PET in early phase clinical trials.

Figure 3.

FLT imaging can discriminate between radionecrosis and true tumor recurrence in primary brain tumors if the flux rate constant, K_FLT, is measured. That ability does not exist for FLT SUV or for FDG SUV.

Recommendations for reporting future research

While human studies with FLT clearly support a clinical role for this radiopharmaceutical as an indicator of response, there are important questions that we cannot yet answer with confidence. Larger clinical trials are needed with rigorous clinical evaluation; extended follow up is essential. Correlation with Ki-67 is a useful starting point; this biomarker is a nuclear protein expressed throughout the cell cycle however it suffers from the heterogeneity problem that limits all sampled biomarkers. Weber²⁴ has pointed out the complex relationship between the standard Ki-67 labeling index and expression of mRNA for Ki-67. Thus while correlation with Ki-67 activity has been useful, it is no substitute for a properly powered study of FLT with respect to long term clinical outcome variables including both progression free survival and overall survival.

There are other issues that require further study. For example, will the role of FLT-PET be limited to therapies that target S-phase or will it be equally applicable to treatments that interrupt the cell cycle at other points such as G2/M block? Does the relative rate of glucuronidation vary after a drug therapy that is especially toxic to the liver, such that the fraction of the injected dose of FLT available to produce the SUV image might differ by as much as 50%. If so, the SUV number could be driven by changes in blood metabolite levels more than a change in TK1 activity. This question could be answered if the glucuronide ratio were measured and reported for blood samples at baseline and after treatment in a wider range of drug protocols.

The clinical results for FLT-PET reviewed here also suggest that additional basic research is needed, some of which will be translational from the bedside back to the laboratory bench. In order to use FLT-PET more effectively, we need studies that critically test the validity and limitations of a simple SUV measurement to quantify response. This will require both human and animal studies with dynamic imaging and kinetic analysis as well as SUV. Animal models are clearly the best way to study whether post-therapy changes in FLT images are reflecting TK1 levels or are also responding to changes in nucleoside transporters, either their concentration or their intracellular location. Because FLT uptake is a measure of the exogenous pathway for proliferation, it is particularly important that FLT studies in the setting of exogenous pathway drugs be analyzed separately from those that involve endogenous pathway drugs, as the pathway targeted by therapy can affect FLT uptake.

Conclusions

FLT is clearly a more specific imaging agent than FDG for quantifying cellular proliferation. However, FLT is a more demanding imaging study to evaluate than FDG. Specifically, since FLT is glucuronidated by the liver, and the metabolite is not transported into the cell, the plasma concentration of FLT can be significantly changed by the presence of some chemotherapy agents and pain medications. Therefore, the plasma level of un-metabolized FLT should be measured to assure kinetics can be properly interpreted. This is important because the flux constant K_FLT is a better discriminator of response and recurrence than SUV. Applying this standardization to multicenter studies will allow FLT imaging to become a more clinically relevant measure of treatment response for neoplastic disease.