Imaging of Cell Proliferation: Status and Prospects

Radiotracers Indirectly Related to Cell Proliferation.

Much of radiotracer imaging in oncology is based on the accelerated intermediary metabolism (such as energy production, protein synthesis, and phospholipid synthesis) needed to support tumor growth (10). Several radiotracers for the substrates used in these processes have been developed; among them is ¹⁸F-FDG, the mainstay of clinical PET. Because of the general tendency of intermediary metabolism to track cell growth, the accumulation in tumors, or at least the one-way flux from blood into tumors, of these radiotracers may be positively correlated with tumor cell proliferation. However, the relationships are indirect, and changes in tumor intermediary metabolism are likely to lag behind the events (e.g., cell cycle arrest) directly involved in modulating cell proliferation. Some studies comparing tumor ¹⁸F-FDG uptake or use with in vitro cell proliferation assays have revealed strong positive correlations (11,12); however, in most cases, the relationships have been negative or relatively weak (11,13–16). Rather than a marker for tumor cell proliferation, ¹⁸F-FDG is generally considered to be indicative of viable cell density (15–18).

Some studies have shown correlations of the tumor uptake of radiotracers of amino acid and phospholipid metabolism with cell proliferation. Among the various amino acids that have been used for imaging in humans, 2 short-lived PET tracers, l-[¹¹C-methyl]methionine and l-[1-¹¹C]tyrosine, have been studied the most (10). [¹¹C-methyl]MET uptake was found to correlate positively with tumor cell proliferation in carcinomas of the brain and lungs (19–21) but did not correlate with tumor histologic grade in head and neck carcinomas or non-Hodgkin's lymphoma (22,23). Studies with l-[1-¹¹C]tyrosine in human brain tumors failed to show a correlation with cell proliferation (24). Various amino acids have also been labeled with longer-lived radionuclides. For example, the uptake of ¹²³I-iodo-α-methyltyrosine was found in a study of 32 patients with soft-tissue sarcomas to correlate moderately well (r = 0.6–0.7) with the mitotic index, the Ki-67 proliferative index, and tumor cellularity (25).

Several studies have revealed positive correlations between the uptake of radiolabeled tracers of phospholipid metabolism (choline and acetate) and the proliferation of cultured tumor cells (16,26–28), but verification in vivo is currently lacking. One PET study of 18 patients with prostate cancer revealed no significant correlation between the tumor uptake of ¹¹C-choline and Ki-67 staining (29).

Radiotracers Directly Related to Cell Proliferation.

The replication of nuclear DNA is the central process required for cell division. Of the 4 nucleosides required for DNA synthesis, thymidine is the only one that is not also incorporated into RNA. Thymidine nucleotides are derived either through the methylation of dUMP in a reaction catalyzed by thymidylate synthase (TS) or through the phosphorylation of exogenous TdR. These 2 routes, known, respectively, as the de novo and salvage pathways of DNA synthesis, are diagrammed in Figure 2. The de novo pathway is not a viable alternative for monitoring DNA synthesis because the relevant precursors (deoxyuridine, uridine, and uracil) are routed into both DNA and RNA. (Like other nucleotides, dUMP, the direct precursor for TS, cannot enter cells from the outside.) Thus, efforts to measure DNA synthesis have focused on tracers for the thymidine salvage pathway.

The synthesis of nuclear DNA occurs in the cytosol. It is important to note that DNA synthesis also occurs in the mitochondria. Different isoenzymes of TK, known as TK1 and TK2, catalyze the phosphorylation of TdR to TMP in the cytosol and the mitochondria, respectively (Fig. 2). TK1 expression is closely regulated in coordination with that of other key enzymes of DNA synthesis (e.g., TS and DNA polymerase) during the cell cycle. The concentrations of TK1 increase as much as 10- to 20-fold beginning at the G₁-phase/S-phase transition, continue at high levels throughout the S, G₂, and M phases, and then decline rapidly as the enzyme is destroyed at the onset of the G₀ or G₁ phase (30,31). TK2, on the other hand, is constitutively expressed, and its activity is not regulated in conjunction with the cell cycle (32).

The development of PET and the earlier success of the ex vivo TdR assay for DNA synthesis led to the development of [¹¹C-methyl]TdR in the 1970s. It became known that the accurate interpretation of TdR PET images requires arterial blood sampling and metabolite analysis as well as kinetic modeling, making this approach impractical for routine clinical use. Investigators have also synthesized several TdR analogs, some of which are designed to resist catabolism. The chemical structures and radiolabeling positions of TdR, the thymine base, and the analogs that have received the most study are shown in Figure 3.

The most thoroughly developed methodology for the image-based measurement of cell proliferation uses ¹¹C-TdR in conjunction with quantitative, dynamic PET. The history of that development as well as the scientific and technical considerations underlying the methodology were thoroughly reviewed by Mankoff et al. (4) and Wells et al. (5). We present only a brief summary here. The synthesis of [¹¹C-methyl]TdR was first reported in 1972 (3). Animal studies in the late 1970s and early 1980s revealed good uptake and retention of the radiolabel in tumors as well as organs with rapid cell turnover (33–35). Concern that the rapid disappearance of [¹¹C-methyl]TdR from the circulation after intravenous injection might indicate that the distribution of the radiotracer simply reflected perfusion was alleviated in a study by Shields et al., who compared the tumor uptake of [³H-methyl]TdR with blood flow in mice and dogs (35). By 1 min, the correlation between uptake and flow disappeared, indicating that some mechanism (e.g., metabolism) other than blood flow was responsible for [¹¹C-methyl]TdR retention in tissues. Subsequent pilot studies in humans, however, showed that ¹¹C-methyl-labeled catabolites dominated the circulating activity within minutes after injection and that these acidic compounds accumulated in tumors and various nontumor tissues (36–38). On the basis of the hypothesis that radiolabel lost to catabolites would be rapidly removed from the body as the endpoint catabolite ¹¹C-CO₂, rapid procedures for synthesizing ring-labeled 2-¹¹C-TdR were developed (39,40). However, it was found that because CO₂ is diluted in the body's large bicarbonate pool, ¹¹C-CO₂ and H-¹¹C-CO₃⁻ comprise a relatively large percentage of the total activity in blood and other tissues for more than 1 h after the injection of 2-¹¹C-TdR (41,42). Although 2-¹¹C-TdR did provide some improvement over [¹¹C-methyl]TdR with regard to the percentage of activity in the DNA synthesis pathway and the uniformity of the catabolite background (42,43), it was still necessary to account for labeled breakdown products to reliably interpret PET images.

In the 1990s and early 2000s, validation studies were presented for a multicompartment model of 2-¹¹C-TdR in somatic and brain tumors that can be used to estimate TdR flux (i.e., one-way net clearance) from blood into DNA (44–46). Key assumptions and approximations underlying the use of the model (Fig. 4) to describe PET-derived time–activity curves for 2-¹¹C-TdR include the following. First, parenchymal cell membranes do not present a significant barrier to the movement of TdR, implying that radiotracer in the interstitial tissue space mixes freely with that in the intracellular space. This assumption is based on the fact that nucleoside transport is dominated by rapid, facilitated diffusion (47). Second, TdR and TMP, thymidine diphosphate, and thymidine triphosphate (TTP) nucleotides within tumors can be treated as a single compartment; that is, their interconversion is rapid compared with the rates at which TdR reenters the circulation and TTP is incorporated into DNA (48). Third, the incorporation of ¹¹C-TTP into DNA is rate limiting; that is, once the radiolabel enters DNA, it is effectively trapped there during the time scale of the PET experiment (≤1 h). This assumption is certainly well founded (48). Fourth, non-CO₂ catabolites within tumors can be treated as a single compartment. This simplification is justified on the basis of the fact that such catabolites contribute only a small percentage of the total activity in tumors after the injection of 2-¹¹C-TdR and thus have relatively little effect on the estimation of the other parameters in the model. Fifth, non-CO₂ catabolites are not trapped within tumors, an approximation based on the authors' (44–46) own observations. Sixth, the transport of labeled CO₂/HCO₃⁻ is reversible, and some ¹¹C-CO₂/HCO₃⁻ may be incorporated into molecular species that are effectively trapped in tumors during the PET experiment. Both assumptions are supported by empiric observations (49,50). Seventh, the TdR, catabolite, and CO₂ compartment sets are independent (i.e., not coupled) and are driven by 3 separate blood time–activity input functions, each of which is measured by blood sampling and metabolite analysis. This treatment assumes that labeled TdR catabolites within tumors originate predominantly from the circulation instead of being formed internally and that catabolites are not reutilized for DNA synthesis. The authors (44–46) based these assumptions on published information (48), their own observations, and simulations showing that parameter estimation for the TdR component of the model is relatively insensitive to moderate amounts of intratumoral catabolism of 2-¹¹C-TdR.

Calculation of the DNA synthesis rate requires combining the model with a measurement of the TdR concentration in blood (TdR_blood) as well as some additional assumptions. Parameters of the TdR component of the model permit the calculation of one-way clearance of TdR from blood into DNA. The equation for this “flux constant” (K_TdR) is shown in Figure 4. The product [(s + dn)/s] × TdR_blood (μmol/mL) × K_TdR (mL/min/g) yields total TdR flux into DNA in units of μmol/min/g of tissue. The factor (s + dn)/s is the ratio of total TdR flux into DNA (salvage + de novo pathways) to salvage pathway flux. Assumptions underlying the conversion of K_TdR to TdR flux (which is directly proportional to the DNA synthesis rate (48)) include the following. First, the concentration of native TdR in the tumor interstitial space is approximately the same as that in the blood. This assumption is probably justified for somatic tumors, in which small molecules are exchanged freely across capillary walls, but may not be valid in brain tumors that conserve a portion of the blood–brain barrier (BBB). Second, the amount of TMP synthesis via the salvage pathway relative to that via the de novo pathway is predictable. An extensive study by Shields et al. showed that, independently of cell type and species, the 2 pathways tend to contribute similarly, depending on the external concentration of TdR (51). It should be noted, however, that the effects of antitumor treatment on the ratio of the salvage pathway to the de novo pathway were not examined; therefore, extrapolation of salvage pathway flux to total TdR flux may not be valid during therapy, especially when that therapy is specifically intended to suppress the de novo pathway.

An alternative approach to the analysis of ¹¹C-TdR PET studies has been developed by Gunn et al. (52). Rather than a compartmental model, their method uses spectral analysis, with or without a second scan after the intravenous injection of ¹¹C-HCO₃⁻, to obtain separate estimates of the ¹¹C-CO₂/HCO₃⁻, intravascular ¹¹C-TdR, and parenchymal ¹¹C-TdR components of the tumor time–activity curves. Like the compartmental model approach, this method requires determination of the arterial time–activity curves for ¹¹C-TdR and catabolites. However, unlike the compartmental model approach, it does not support the calculation of TdR flux into DNA.

Finally, we note that TdR is a substrate for the synthesis of both nuclear DNA and mitochondrial DNA. Although it is true that the cytosolic synthesis of nuclear DNA via TK1 dominates TdR use in S-phase cells, the S-phase fraction may be small, even in rapidly growing tumors. Therefore, it should be kept in mind that rates of use of TdR estimated by ¹¹C-TdR PET reflect a combination of nuclear DNA synthesis and mitochondrial DNA synthesis, with the latter providing a background that may attenuate percentage changes in ¹¹C-TdR incorporation, for example, during anticancer treatments.

The routine use of ¹¹C-TdR for clinical imaging has several problems, including dealing with catabolites, the short half-life of ¹¹C, and relatively difficult synthesis. Labeled TdR must be made for each patient, and distribution networks such as those that now routinely provide ¹⁸F-labeled agents are not possible. These problems have limited ¹¹C-TdR use to a few research centers and have propelled a search for TdR analogs that can be labeled with ¹⁸F and that are resistant to degradation. A wide variety of such analogs has been studied over the years (53–55); of these, ¹⁸F-FLT is one of the more promising tracers.

FLT was originally synthesized as an antineoplastic and antiretroviral agent and is similar in concept to azidothymidine, the first approved anti-HIV agent. It acts as a chain terminator in the synthesis of DNA. Initial clinical trials of FLT therapy were performed in patients with AIDS. Marrow suppression and occasionally fatal hepatotoxicity at routine pharmacologic doses (about 100 mg/d) led to the abandonment of FLT as a therapeutic agent. Nevertheless, the information gained provided a strong background to support the development of ¹⁸F-FLT as a radiotracer, because the doses used are orders of magnitude lower. The synthesis of FLT with ¹⁸F was initially accomplished with a mesylate precursor and was subsequently done with the nosylate and anhydrides (56–58). These precursors are now commercially available, and automated devices for the synthesis of ¹⁸F-FLT can be purchased.

Replacement of the hydroxy group at the 3′ position of deoxyribose converts TdR to FLT (Fig. 3). This modification prevents cleavage of the sugar from the thymine base by thymidine phosphorylase, making FLT highly resistant to breakdown, a major advantage for in vivo imaging. On the anabolic side of the equation, FLT is a good substrate for TK1 and a very poor substrate for TK2 (59). Additional kinases convert FLT monophosphate to FLT diphosphate and FLT triphosphate, and the FLT nucleotides can also be dephosphorylated by 5-nucleotidase (5). Importantly, fluorination at the 3′ position of the sugar inhibits the incorporation of FLT triphosphate into DNA. Thus, trapping of FLT in the form of its nucleotides is reversible, a factor that must be considered when ¹⁸F-FLT is used as an imaging agent for cell proliferation (60).

Initial studies in dogs and humans demonstrated that ¹⁸F-FLT was taken up and retained in organs (such as the bone marrow) and tumors with high proliferative rates (53). Imaging studies in human patients demonstrated accumulation of activity in the liver (Fig. 5), unlike the pattern noted in other species. As was known from studies of azidothymidine, ¹⁸F-FLT undergoes extensive glucuronidation in the human liver (60–63). On average, about 25% of the plasma activity is present as ¹⁸F-FLT-glucuronide at 60 min after injection. This factor must be considered in measurement of the blood level of the tracer for use in kinetic modeling (60,61,64). Although clearance can vary from patient to patient, ¹⁸F-FLT-glucuronide can be reasonably estimated by simple separation of a blood sample obtained at 60 min.

The kinetic modeling of ¹⁸F-FLT retention is analogous to that used for ¹⁸F-FDG; that is, it requires only a simple 2-compartment model for tissue. FLT-glucuronide does not accumulate in tissues outside the bloodstream; therefore, its contribution to PET images can be combined with that of intravascular ¹⁸F-FLT as an additive term in the equation for total activity seen by PET (61). Compared with the model for 2-¹¹C-TdR (Fig. 4), the ¹⁸F-FLT model lacks the 2 compartment sets representing catabolites. As is the case for TdR, K₁ and k₂ represent transport into and out of the tumor, respectively. Unlike the situation with 2-¹¹C-TdR, however, the exchanging tissue compartment represents only the free ¹⁸F-FLT nucleoside, not nucleoside plus nucleotides, whereas the second tissue compartment corresponds to ¹⁸F-FLT nucleotides, not DNA. Thus, for ¹⁸F-FLT, k₃ is the rate constant for TK1-modulated phosphorylation, not incorporation into DNA. The dephosphorylation of ¹⁸F-FLT phosphate can be detected by 120 min, and the rate constant for this process is designated k₄ (61). This term cannot be accurately measured in shorter imaging intervals and, in some instances, has been set to zero (62).

¹⁸F-FLT uptake has been compared with the measurement of Ki-67 levels in tumors by several investigators. The most extensive studies have been done in patients imaged with ¹⁸F-FLT before resection of lung cancer. The correlations between the ¹⁸F-FLT standardized uptake value (SUV) and Ki-67 were statistically significant (r = 0.87 and r = 0.84, as determined by Buck et al. (65) and Vesselle et al. (66), respectively). Yamamoto et al. found a similar significant correlation (r = 0.77, P < 0.0002) but also found that Ki-67 was slightly better correlated with ¹⁸F-FDG (r = 0.81, P < 0.0001) (67). In contrast,Yap et al. found only ¹⁸F-FLT, not ¹⁸F-FDG, to correlate with Ki-67 levels (68). In brain tumors, on the other hand, Ki-67 levels correlated better with ¹⁸F-FLT uptake (r = 0.84, P < 0.0001) than with ¹⁸F-FDG uptake (r = 0.51, P = 0.07) (69). Although the tumor uptake of ¹⁸F-FLT was found to be predictive of Ki-67 levels in breast cancer (70), it was not found to correlate with such measurements in esophageal cancer (71). In summary, whereas ¹⁸F-FLT retention correlates well with measures of proliferation obtained from biopsy specimens, this correlation is neither perfect nor seen in every tumor type. Furthermore, it is not known to what extent the correlation between ¹⁸F-FLT incorporation and cell proliferation is maintained during antitumor therapy.

A theoretic limitation of ¹⁸F-FLT as a radiotracer for the salvage pathway of DNA synthesis is that ¹⁸F-FLT is not incorporated into DNA; therefore, its uptake may be misleading if TK1 activity is not rate limiting for the incorporation of exogenous TdR into DNA. Cleavage of the sugar from the thymine base is inhibited by fluorination of either the 3′ position (as with ¹⁸F-FLT) or the 2′ position in deoxyribose (Fig. 3); however 2′ fluorination does not inhibit the incorporation of the triphosphate nucleotide into DNA.

1-(2′-Deoxy-2′-fluoro-1-β-d-arabinofuranosyl)-thymine (FMAU) is a 2′-fluoropyrimidine that has been developed for cell proliferation imaging with labeling either in the 5-methyl group of the pyrimidine base with ¹¹C (72) or at the 2′-fluoro position of the sugar with ¹⁸F (73). Preclinical studies have shown that FMAU retention in tumors and nontumor tissues with rapid cell turnover (e.g., marrow and small intestine), reflects incorporation into DNA (74–77). FMAU is highly resistant to catabolism in both animals and humans, with the injected compound dominating time–activity curves in blood during the first hour after injection (75,76,78). Preliminary clinical studies have shown tumor uptake of ¹¹C- or ¹⁸F-FMAU in a variety of cancers (78,79) comparable to that seen in human studies with ¹⁸F-FLT (65,80–83). In humans, ¹¹C- or ¹⁸F-FMAU has higher liver, kidney, and myocardial uptake but much lower marrow uptake and rate of urinary excretion than ¹⁸F-FLT (compare Figs. 5 and 6), suggesting a potential role for FMAU in cases requiring assessment of bone metastasis or the pelvic region. Another potential advantage is that, in part because of its rapid blood clearance, ¹¹C- or ¹⁸F-FMAU uptake in tumors reaches a plateau by about 10 min after bolus injection of the radiotracer. Tehrani et al. defined a tumor retention ratio equal to the area under the tumor time–activity curve (AUC) from injection to time t divided by the product t × the area under the blood time–activity curve from injection to t (84). Blood curves were determined by peripheral venous sampling. For patients with prostate (n = 6) or brain (n = 4) tumors, the investigators found that the tumor retention ratio at t = 11 min correlated strongly with the one-way clearance of ¹⁸F-FMAU from blood into the trapping compartment, as determined by compartmental modeling of 60-min dynamic PET and arterial sampling data. Thus, a single PET image of a tumor together with serial venous blood sampling during the first 11 min after injection may yield information comparable to that obtained from a full kinetic analysis, a major advantage for clinical applications. The rapid kinetics of FMAU also imply a potential clinical role for ¹¹C-FMAU both in reducing patient radiation doses and in facilitating same-day imaging with ¹⁸F-FDG or other PET tracers.

The primary limitation of ¹¹C- or ¹⁸F-FMAU appears to be that it is a relatively poor substrate for TK1 and a relatively good substrate for TK2 (85,86), probably accounting for its localization in mitochondrion-rich human myocardium (Fig. 6). Because TK1 is so highly expressed in dividing cells (87), cytosolic DNA synthesis may dominate FMAU uptake in aggressively growing tumors. However, because of background incorporation into mitochondrial DNA, ¹¹C- or ¹⁸F-FMAU retention may be less sensitive than ¹⁸F-FLT retention to changes in cell proliferation. It is clear that a key question affecting the utility of ¹¹C- or ¹⁸F-FMAU is the extent to which tumor uptake reflects TK1 versus TK2 activity.

Preliminary studies suggested that, as with ¹⁸F-FLT, monophosphorylation is the rate-limiting or trapping step for ¹¹C- or ¹⁸F-FMAU; that is, ≥80% of the radiolabel recovered from tissues was either in FMAU or in DNA (75). However, the DNA incorporation step may become rate limiting for FMAU incorporation during therapies that specifically impair triphosphate nucleotide incorporation into DNA, implying a potential advantage for ¹¹C- or ¹⁸F-FMAU over ¹⁸F-FLT in situations in which TK1 activity is not rate limiting for the incorporation of exogenous TdR into DNA.

Halogenation at the 5 position of the uracil base imitates CH₃ and thus converts deoxyuridine into a TdR analog (Fig. 3). This process, plus the availability of suitable iodine and bromine isotopes (¹²³I, ¹³¹I, and the β⁺-emitters ¹²⁴I and ⁷⁶Br), has led to the development of iodo- and bromodeoxyuridine radiotracers for in vivo imaging of cell proliferation (88–90). The principal limitation of this approach is rapid in vivo dehalogenation of the radiotracers. Various manipulations designed to reduce the radiohalogen background, including saline flush and diuresis (91), late-phase imaging (24 h) to allow washout of the radiohalogen (92), biomodulation to block dehalogenation (93), and molecular modification of the radiotracers to reduce dehalogenation (55), have met with limited success (94,95).

Additional TdR analogs that are resistant to catabolism and that have shown promise as potential imaging agents for cell proliferation include 1-(2′-deoxy-2′-fluoro-β-d-arabinofuranosyl-uracil)bromouracil (FBAU) labeled with either ¹⁸F or ⁷⁶Br (half-life, 16 h) and 4′-[¹¹C-methyl]thiothymidine. On the basis of the similar sizes of the methyl and bromo moieties, FBAU behaves similarly to FMAU. Studies in rats have shown that it is rapidly incorporated into DNA in proliferating tissues, such as the small intestine, spleen, and marrow (96,97). A recent study with ¹⁴C-labeled thiothymidine in mice bearing syngeneic mammary tumors showed high uptake in the spleen and duodenum and moderate uptake in tumors (98). The percentages of activity in DNA at 1 h were greater than 70% for the spleen and duodenum and greater than 50% for tumors.

There are large differences in the systemic pharmacokinetics and biodistribution of ¹¹C-TdR and its analogs among humans and the various species of animals available for preclinical testing. These variations, which are illustrated in Figure 6 for the TdR analog ¹¹C- or ¹⁸F-FMAU, cause difficulties in predicting radiolabel biodistribution and radiation doses in humans from studies in animal models.

Mice and rats have circulating levels of TdR (∼1 μM) that are about an order of magnitude greater than those in humans (99) and are high enough to cause significant competitive inhibition of radiolabeled TdR and TdR analog uptake in murine tissues (75). This factor contributes to the generally low organ and tissue uptake seen with ¹¹C-TdR and its analogs in mice and rats (35,75,100). Some investigators have injected thymidine phosphorylase to deplete circulating TdR before radiotracer administration, and this procedure results in a dramatic increase in marrow uptake (75). The initial biodistribution studies of ¹⁸F-FLT were done with normal Swiss Webster mice, and relatively high uptake of the tracer was noted in proliferative tissues, such as the marrow and spleen. When studies were done with BALB/c mice bearing implanted tumors, no concentration of ¹⁸F-FLT above the background was demonstrated. Although the variation between strains of mice has not been formally tested, it may reflect differences in native TdR levels.

In dogs, the biodistribution of radiolabeled TdR and its analogs at 1 h after injection is similar to that in rats, with the most notable differences being modestly higher kidney uptake of ¹¹C-TdR, higher marrow uptake of ¹⁸F-FLT (53), and modestly higher myocardial uptake of ¹⁸F-FMAU (76).

We are aware of only one imaging study of ¹¹C-TdR or its analogs in nonhuman primates, that being with ¹¹C-FMAU (101). Scans of 2 cynomolgus monkeys showed much higher accumulation of radiolabel in the liver and kidneys in monkeys than in rodents or dogs, a difference also seen with ¹⁸F-FMAU in humans (78).

A striking difference among the various species examined is the much higher myocardial uptake of ¹¹C-TdR and ¹⁸F-FMAU in humans. TK2, the mitochondrial form of the enzyme, is present at relatively high levels in the human heart, contributing to the uptake of ¹¹C-TdR in this organ. ¹⁸F-FLT and ¹⁸F-FMAU show variable cardiac uptake, depending on the relative affinity of these compounds for TK2 and the relative myocardial expression of TK2 among species. In humans, little ¹⁸F-FLT is retained because of its very low affinity for TK2 (59), but ¹⁸F-FMAU shows high retention, reflecting its relatively high affinity for TK2 (86).

Humans also have much higher hepatic uptake of ¹⁸F-FLT and ¹¹C- or ¹⁸F-FMAU than do mice or dogs (53,76,78,100,102). For ¹⁸F-FLT, this finding reflects differences in glucuronidation. As discussed earlier, in humans the kinetic modeling of ¹⁸F-FLT retention must take into account the fraction of ¹⁸F-FLT-glucuronide seen in the blood, which reaches a level of about 25% at 1 h. ¹¹C- or ¹⁸F-FMAU has extremely fast clearance from the blood in humans, with 95% of the tracer being cleared by 10 min, reflecting very high liver retention. ¹⁸F-FMAU clearance is much slower in dogs, with little uptake being seen in the liver above the background. The mechanism causing the high liver uptake of ¹¹C- or ¹⁸F-FMAU in humans has not been determined.

It is clear that one must take into account the major differences in native TdR levels in the blood and the resulting effects on tracer metabolism, clearance, and retention in tissues when deciding to study radiolabeled TdR or its analogs in an animal model. Rodents and dogs have been used the most for studying these tracers, but their limitations must be kept in mind when one is planning studies and interpreting results.

MRI.

Most of the effort to develop MRI for the examination of tumor cell proliferation has focused on magnetic resonance spectroscopy in brain tumors. Empiric correlations have been reported between tumor growth rates and magnetic resonance spectroscopy measurements of relative choline, N-acetylaspartate, or mobile lipid concentrations; the mapping of local cell growth rates within tumors appears to be feasible (103–105). There also have been some attempts to establish empiric correlations between tumor proliferative activity and proton relaxation times, T1 and T2, on the basis of differences in tumor interstitial or intravascular water content (106).

Optical Imaging.

Fluorescence-labeled molecular probes targeted at molecules expressed specifically in dividing cells can, in principle, be developed (107). Alternatively, reporter genes driving fluorophore or bioluminescence enzyme expression and triggered by conditions related to cell proliferation can be built into tumor cells (108). At this point, however, in vivo optical imaging of tumor cell proliferation is mostly confined to monitoring of tumor growth and spread in transplant-bearing or transgenic mice with tumor cells engineered to express either a fluorescence or a bioluminescence marker (109). With proper calibration, the detected light is proportional to cell number, so tumor cell growth rates can be determined by serial imaging in the same mouse (110,111).

TdR.

Only a few studies have examined the use of ¹¹C-TdR PET to assess the tumor response to treatment. Shields et al. reported on the use of 2-¹¹C-TdR in 3 patients with lung cancer and 2 patients with soft-tissue sarcomas (116). Scans were obtained before and at 1, 6, and 10 wk after a single course of chemotherapy (various multidrug regimens) and compared with same-day ¹⁸F-FDG scans. Responses were judged from histopathologic examination or overall clinical assessment. The tumor uptake of both 2-¹¹C-TdR (SUV at 20–60 min after injection) and ¹⁸F-FDG (SUV at 30–60 min after injection) declined by the first follow-up scan in the 4 patients who had complete or partial responses. 2-¹¹C-TdR uptake was unchanged in the one nonresponding patient, whereas ¹⁸F-FDG uptake increased. Kinetic analysis enhanced the distinctions in TdR responses; the 3 patients with complete responses all showed 100% reductions in the blood-to-DNA flux constant measured at 1 wk. 2-¹¹C-TdR and ¹⁸F-FDG tumor responses over 10 wk are illustrated for one of the responding patients in Figure 7.

A study that illustrated a very early flair response with 2-¹¹C-TdR was reported by Wells et al. (126). Five patients with intrahepatic metastases from colon cancer and 7 control patients with advanced gastrointestinal malignancies were studied twice each with 2-¹¹C-TdR. The 5 test patients were rescanned 1 h after the oral administration of a TS inhibitor. Images were corrected for labeled catabolites and analyzed for tumor uptake of 2-¹¹C-TdR at 50–60 min after injection. Treated patients all showed marked increases in the tumor SUV (43% ± 10%, mean ± SEM), whereas control patients showed small overall decreases (−8% ± 3%; P < 0.005). That study dramatically illustrated the potential of cell proliferation imaging to provide immediate information on the pharmacodynamics of appropriate chemotherapeutic drugs.

FLT.

In recent studies, small-animal PET or direct tissue assays were used to correlate ¹⁸F-FLT with various types of antineoplastic therapy in tumor-bearing mice. Sugiyama et al. studied the effects on ¹⁸F-FLT and ¹⁸F-FDG uptake of x-ray irradiation of a murine sarcoma (experiment 1) and photodynamic therapy of a human cervical carcinoma xenograft (experiment 2) (127). In experiment 1, ¹⁸F-FLT uptake had declined 30% at 6 h and remained at that level, whereas ¹⁸F-FDG uptake was unchanged until 3 d after treatment. Measurements were obtained in treated and control animals at 24 h in the photodynamic therapy experiment; ¹⁸F-FLT uptake was reduced 75% at that time, whereas ¹⁸F-FDG uptake was unchanged. In both experiments, changes in ¹⁸F-FLT uptake reflected declines in cell proliferation measured ex vivo. Leyton et al. reported declines in ¹⁸F-FLT uptake of 24% ± 8% and 49% ± 8% at 1 and 2 d, respectively, after the administration of cisplatin to mice bearing radiation-induced fibrosarcomas (128). The same researchers also found a dose-dependent reduction in ¹⁸F-FLT uptake in an implanted human cervical carcinoma 4 d after the administration of a TK1 inhibitor (129). Barthel et al. observed 52% and 28% reductions in ¹⁸F-FLT uptake and ¹⁸F-FDG uptake, respectively, at 24 h after the administration of 5-fluorouracil to mice bearing radiation-induced sarcomas (100). Finally, Waldherr et al. administered an epidermal growth factor receptor tyrosine kinase inhibitor to mice bearing human cervical cancer xenografts and observed tumor growth arrest as well as an 80% decline in ¹⁸F-FLT uptake at 1 wk compared with a 30% decline for ¹⁸F-FDG (130). An additional experiment showed that the ¹⁸F-FLT response was present by 48 h. In summary, 5 separate studies with 6 different types of treatment in 5 different types of tumors all revealed early (i.e., within a few days) and marked reductions in ¹⁸F-FLT uptake. In studies in which ¹⁸F-FLT was compared with ¹⁸F-FDG, ¹⁸F-FLT always provided earlier or quantitatively more distinct evidence of treatment efficacy.

Clinically, investigators have begun to report small pilot studies of treatment response assessment with ¹⁸F-FLT. Pio et al. scanned 14 patients with breast cancer before treatment, after the first cycle and again after the completion of chemotherapy (131). Responses were assessed with CT. Rapid equilibration of ¹⁸F-FLT uptake was observed. A change (decline) in the tumor SUV after the first course of treatment correlated with a subsequent change in tumor size (r = 0.8), leading the investigators to conclude that ¹⁸F-FLT scans beginning 10 min after injection accurately predict the tumor response. More recently, Kenny et al. reported on changes in tumor uptake and trapping of ¹⁸F-FLT and ¹⁸F-FDG in 13 breast cancer patients treated with 5-fluorouracil, epirubicin, and cyclosporine (1). Follow-up scans were obtained 6–12 d after the initiation of treatment, and responses were defined by CT (Response Evaluation Criteria in Solid Tumors [RESIST]). Images from 2 patients are shown in Figure 1. The tumor SUV at 90 min and the Patlak slope K_i (Patlak slope, equivalent to net clearance from blood into the trapping compartment) were significantly reduced on follow-up in all 6 patients who had complete or partial responses. Average decreases in responding lesions were 41% for SUV and 53% for K_i, whereas nonresponding tumors showed average increases of 3.1% and 1.9% for SUV and K_i, respectively. There was a significant difference in ¹⁸F-FLT parameter changes between patients who responded and those who did not (P = 0.02 for both SUV and K_i).

Two studies of patients with non-Hodgkin's lymphoma examined the utility of ¹⁸F-FLT in predicting responses to multiagent immunotherapy and chemotherapy (rituximab plus cyclosporine, adriamycin, vincristine, and prednisone [R-CHOP]). Herrmann et al. scanned 6 patients before and again at 1 and 6 wk after R-CHOP treatment (132). The tumor uptake of ¹⁸F-FLT (SUV measured at ≥45 min after injection) declined by averages of 77% and 85% at those time points (132). Another group of 16 patients had follow-up scans at 2 d after rituximab treatment and again at 2 d after R-CHOP treatment; there was no significant change after the immunotherapy component, but there was an average decline of 32% after the administration of chemotherapy. Kim et al. studied 33 patients with aggressive non-Hodgkin's lymphoma (133). ¹⁸F-FLT follow-up scans were obtained before the second cycle of R-CHOP therapy and again after the completion of therapy. Responses were assessed with conventional imaging, biochemistry, or ¹⁸F-FDG. For ¹⁸F-FLT, the average changes in the tumor SUV were −71% after the first cycle of therapy and −88% after the final cycle. Patients with complete responses had normal ¹⁸F-FLT scans after both the first and the final cycles.

Finally, Chen et al. performed ¹⁸F-FLT follow-up scans at 1–2 wk and again at 6 wk after the administration of combined antiangiogenesis and chemotherapy (bevicuzimab and irinotecan) in 19 patients with gliomas (134). A metabolic response was defined as a decrease in tumor uptake (SUV) of ¹⁸F-FLT of greater than 25%. Mean survival times in 9 metabolic responders and 10 nonresponders were 10 and 5 mo, respectively. Early and late ¹⁸F-FLT evaluations were the 2 most powerful independent predictors of survival among all of the variables tested, including MRI.

Although it is still far too early to draw conclusions, the initial clinical evaluations of ¹⁸F-FLT have consistently been very encouraging. In each of the 5 pilot studies cited here, the tumor uptake of ¹⁸F-FLT declined dramatically at the earliest time of follow-up, suggesting that cell proliferation imaging with ¹⁸F-FLT can be a powerful tool for the individualization of anticancer therapy.