Elsevier

Molecular Imaging & Biology

Volume 5, Issue 6, November–December 2003, Pages 363-375
Molecular Imaging & Biology

Article
Relationships between radiotracer properties and image quality in molecular imaging of the brain with positron emission tomography

https://doi.org/10.1016/j.mibio.2003.09.009Get rights and content

Abstract

In molecular imaging of the brain, many factors affect the reliability of the quantitative information that can be derived from the imaging process. This article discusses factors impacting on the imaging quality that are related to the radiotracer per se. Following a brief summary of key concepts in receptor quantification, a number of these factors are discussed, including selectivity, affinity, delivery, and lipophilicity. Concepts discussed in the theoretical section are then illustrated, by reviewing a recent comparative evaluation of four agents developed to label the serotonin transporter ([11C]ADAM, [11C]DASB, [11C]DAPA, and [11C]AFM). Specifically, the relationship between affinity and lipophilicity, measured in vitro, and several scanning parameters are investigated. These include peripherical metabolism, brain uptake, required scanning time, nonspecific binding, and binding potential. It is shown that, within a given structural family, affinity and lipophilicity are associated with scan outcome in a relatively predictable manner.

Introduction

Molecular imaging of the brain with positron emission tomography (PET) is a rapidly expanding field of clinical investigations. In addition to improvements in instrumentation and image analysis, the availability of suitable radiotracers is a key factor driving this expansion. Radiotracers are being developed to image a growing numbers of receptors, transporters, enzymes, and other molecular targets. The development of adequate PET radiotracers represents, however, a formidable challenge, given the large number of requirements that must be fulfilled. The candidate radiotracer should be suitable for high specific activity labeling with 11C or 18F. It should not be toxic. It should cross the blood–brain barrier (BBB). This implies an appropriate lipophilicity (logP 1.5–4), low molecular weight (<450), and an absence of active efflux mechanism for the tracer. Its metabolism should not produce radiolabeled metabolites that enter the brain. Its free fraction in the plasma should be measurable. Its binding should be selective. Its affinity should be high enough to provide adequate signal-to-noise ratio, but not too high, so that binding equilibrium can be reached within the time frame of the scan. It should not be too lipophilic, so that its nonspecific binding remains low.

As noted, these requirements might be conflicting, inasmuch as many chemical and pharmacologic parameters are associated with in vivo behaviors that might affect the final image qualities in opposite directions. For example, lipophilicity promotes brain penetration, which is good, but increases nonspecific binding, which is detrimental. Perhaps the greatest difficulty in ligand development stems from the fact that the in vitro chemical and pharmacologic properties of candidate compounds do not necessarily predict their success as imaging agents. Specifically, no set of physicochemical properties guarantees brain penetration or low nonspecific binding. Failure can be reasonably predicted, but not success. As a result, the field is still essentially progressing by trial and error, and experiments involving testing of radiolabeled compounds in rodents and nonhuman primates remain the cornerstones of ligand development.

In this article, we review the qualities of PET radiotracers that are desirable to provide reliable quantification of receptors and transporters in the brain. For a ligand to be useful in clinical investigations, the information collected by the camera (i.e., the images of the distribution of this ligand in the brain) should be suitable to quantitative analysis. Pictures must be translated into numbers that relate to well-defined biologic entities. Short of a validated quantitative analysis, the information is of little value for clinical investigations. Given the importance of quantification in molecular imaging of the brain, the first section of this article will briefly review the nature of the outcome measure derived from quantitative analysis of these studies. The second section will discuss the key properties of a radiotracer that are required for quantitative analysis. In the third section, these concepts will be illustrated by presenting a comparative evaluation of four new radiotracers developed to image the serotonin transporter (SERT).

For neuroreceptor quantification, the outcome measure of experiments performed at tracer doses is the binding potential (BP, mL/g),1 which is equal to the ratio of the receptor density (Bmax, nM, or fmol per gram of tissue) and affinity (1/KD, nM, or fmol per mL of brain water).BP=BmaxKD

Reversible interaction between ligand and receptors involves bimolecular association of ligand with receptor and unimolecular dissociation of receptor–ligand complex. According to the law of mass action, the rate of change of products is proportional to the concentration of reactants:d([LR])dt=kon[L][R]−koff[LR]where [L], [R], and [LR] indicate free concentrations of radioligand, available receptors, and ligand receptor complex, respectively, and kon (association rate constant), koff (dissociation rate constant) are the constants of proportionality.

Binding equilibrium is a key concept in neuroreceptor imaging. At the molecular level, equilibrium is a state in which association and dissociation of the ligand receptor complex per unit time are equal, which implieskon[L][R]=koff[LR].

Using a more pharmacologic notation where the free, [L], is denoted F, the bound, [LR], is denoted B, the total concentration of receptor, [R] + [LR], is denoted Bmax, and the koff/kon ratio is denoted KD, and rearrangement of Equation 3 leads to the Michaelis-Menten equilibrium equation, in which Bmax and KD determine the equilibrium bound quantity at a given free level.B=BmaxF(KD+F)

At tracer dose, F is much smaller than KD, and Equation 4 simplifies toB=BmaxFKD.

By rearrangement, Equation 5 leads to the most fundamental equation for neuroreceptor imagingBP=BmaxKD=BFin which it is established that, at tracer dose, BP, the Bmax/KD ratio, is equal to the ratio of the specifically bound to the free tracer at equilibrium. Analyses of PET neuroreceptor imaging experiments aim at extracting this ratio from the data. As will be discussed below, the determination the denominator in this ratio is often much more problematic than that of its numerator.

A three-compartment model (also known as the two tissue-compartment model) is the most widely used configuration to derive BP. This model includes the arterial plasma compartment (Ca), the intracerebral free and nonspecifically bound compartment (nondisplaceable compartment, C2), and the specifically bound compartment (C3).

The equilibrium distribution volume of compartment i (Vi, mL g−1) is defined as the ratio of the tracer concentration in this compartment to the free arterial concentration at equilibrium,Vi=Cif1Cawhere f1 is the plasma free fraction of the parent compound. V2 and V3 are defined as the distribution volumes of the second (nondisplaceable) and third compartments (specific), respectively. VT is defined as the total regional equilibrium distribution volume, equal to the sum of V2 and V3. A region of reference is a region with negligible density of the receptors of interest (V3 = 0). Therefore, VT in the region of reference provides an estimation of V2. Several methods are available to derive the distribution volumes from the data set, the most frequently used being the kinetic, equilibrium, and graphic approaches.2 In the kinetic method, distribution volumes and receptor parameters are calculated from the derivation of the rate constants that govern the transfer of tracer between compartments. In the three-compartment model described above, K1 and k2 are the rate constants for transit of the radiotracer between plasma and nondisplaceable compartment, while k3 and k4 are rate constants for the transit of the radiotracer between nondisplaceable and specific binding compartment. The derivation of the relationship between k3, k4 and receptor parameters can be found in Slifstein and Laruelle2 and yieldsk3=f2konBmaxk4=koff

Equation 5 indicates that BP is equal to the ratio of the bound to the free in the vicinity of the receptors at equilibrium. While the bound is equal to C3, the free is not equal to C2 (a compartment that includes both free and nonspecific binding), but rather to f2C2, with f2 being the free fraction in the second compartment.BP=BmaxKD=BF=C3f2C2

The parameter f2 cannot be directly measured. However, at equilibrium, and under the assumption that the tracer crosses the BBB by passive diffusion, the free equilibrates on both sides of the BBB. It follows that at equilibriumf1Ca=f2C2.

This equivalence leads to a new expression for BP, in which it is recognized that the term C3/f1Ca is the distribution volume of the third compartment, V3 (Equation 7).BP=BmaxKD=BF=C3f2C2=C3f1Ca=V3

Because a reliable f1 measurement is difficult to obtain for many tracers, the term f1 is often neglected and assumed to be a constant across subjects. This leads to a more practical definition of BP and V3, denoted here BP′ and V3BP′=f1BmaxKD=f1BF=C3Ca=V3′.

Similarly, while V2 is the distribution volume of the nondisplaceable compartment relative to the free plasma concentration, V2′ is the distribution volume of the nondisplaceable compartment relative to the total plasma concentration. Note that Equations 7 and 11 establish that f2 is the inverse of V2f2=1V2.

BP and BP′ essentially relate the bound in the brain to the tracer concentration in the plasma, and thus require the measurement of both. A third expression of BP, which does not require the measurement of plasma concentration, consists in expressing BP relative to the free and nonspecific binding in the brain (the C3/C2 ratio at equilibrium). Here, the denominator is the tracer concentration in the nondisplaceable compartment. We use the term specific to nonspecific partition coefficient and the notation BP″ or V3″ for this quantity.BP″=f2BmaxKD=BmaxV2KD=f2BF=C3C2=V3V2=V3

In summary, the ubiquitous term “binding potential” is used to designate an equilibrium ratio between the specific binding and a given concentration parameter. The nature of the concentration parameter varies. Table 1 summarizes the three types of BP that are encountered in the PET literature, as well as their kinetic, equilibrium, and graphic derivation. Although BP as defined in Equation 1 (or V3) is the only outcome measure that is exclusively related to receptor parameters, it is almost never used as an outcome measure. The two other estimators of BP (V3′ and V3″) include a term unrelated to receptor parameters. In the PET neuroreceptor literature, the term BP is most often used to designate BP″ (V3″).

To provide data suitable for the derivation of BP and V3″, a radiotracer must meet a large number of specifications. Factors that affect the information content of the images are discussed in this section. In other terms, we consider here the requirements that come after it is established that the radiotracer is suitable for high specific activity labeling, is not toxic, and crosses the BBB.

Ideally, the radiotracer should bind with appropriate affinity to only one class (or subclass) of receptors. However, one should note that some useful radiotracers bind with high affinity to more than one type of receptor. Examples includes [11C]N-methyl-spiperone ([11C]NMSP) or [18F]setoperone, both of which bind to the dopamine D2 receptors and serotonin 5-HT2A receptors.3., 4. In this case, the signal recorded in the striatum corresponds to binding to D2 receptors, while the signal recorded in the neocortex corresponds to binding to 5-HT2A receptors. Another example is [123I]β-CIT, which binds to both the dopamine transporter (DAT) and the SERT.5 In the striatum and the midbrain, the signal of [123I]β-CIT, corresponds mainly to DAT and SERT binding, respectively.6 Thus, differences in receptor localization in the brain might, in some instances, enable the use of nonselective radiotracers. Ultimately, the selectivity margin required depends on the respective density of receptor sites in a given region. Assuming that, in a given region, a radiotracer binds to two types of receptors (A and B), the BP in that region will be given byBP=Bmax AKDA+Bmax BKDB.

Thus, the contribution of each type of receptor to the total BP can be readily calculated by the knowledge of the affinity of the tracer for each receptor and of the relative densities of the receptors in this region. For example, [11C]raclopride binds with similar affinity to D2 and D3 receptors.7 In the human ventral striatum, postmortem data suggest that 30% of the total D2-like receptors are of the D3 subtype.8 Therefore, it should be anticipated that 30% of [11C]raclopride BP corresponds to binding to D3 receptors in the VST (note that differential occupancy of D2 receptors and D3 receptors by endogenous dopamine might affect this prediction). For hypothetical ligands with 10-fold higher affinity for D3 compared to D2, the contribution would be 81%. With a 100-fold difference, it would be 97%. The point is that the level of selectivity required for a particular radiotracer depends on the expected distribution of available receptors in the regions of interest.

The affinity required for a useful radiotracer also depends on the concentration of sites in the target region. Obviously, the lower the target density, the higher the required affinity. Typically, useful ligands display affinities between 1 and 0.01 nM. Some targets, such as the DAT, can be visualized with ligands with much lower affinity (such as [11C]cocaine), because of their high concentration.9 Although the density of the target provides some indication of the desired range of affinity, the affinity required for adequate signal-to-background ratio is also a function of the nonspecific binding. Ultimately, what will matter in imaging is the ratio of specific to nonspecific binding, a ratio which, at equilibrium, is equal to V3″. Generally, V3″ equal to or higher than 0.5 is desirable for reliable quantification (although the reproducibility of the measurement is also affected by the size of the region of interest). Equation 15 shows that V3″ is equal to Bmax/V2KD. Thus, V3″ increases as Bmax and the affinity (1/KD) increases, and decreases as the nonspecific binding (V2) increases. An increase in affinity from ligand A to B will translate in a similar increase in image quality (i.e., V3″) only if the nonspecific binding remains constant. As increased affinity is often associated with increased lipophilicity, which in turn, increases the nonspecific binding, high-affinity ligands do not necessarily perform better than their lower affinity congeners as far as V3″ is concerned. The comparative evaluation of the four SERT radiotracers presented below illustrates this point: although [11C]ADAM and [11C]DAPA have higher affinity for SERT relative to [11C]DASB, their V3″ is, in fact, lower, due to higher nonspecific binding.

The other important point regarding affinity is related to the duration of the scan. Everything else being equal, increase in affinity results in increase in scan duration required to measure receptor parameters. This phenomenon is due to the fact that, at tracer dose, increase in affinity is associated with a longer time needed to reach equilibrium. Equilibrium is reached when the rates of association and dissociation to and from the receptors are equal, that is, when the Michaelis-Menten equation (Equation 4) is satisfied. To clarify this point, it is important to understand what equilibrium means in the context of a single bolus experiment, which is the typical mode of administration in PET (Figure 1).

Following a single bolus injection, the tracer concentration in the plasma and in the brain is constantly changing. Equation 4 prescribes that a given level of binding equilibrium (Be) corresponds to a given level of free radiotracer (F). During the initial phase following bolus injection of the tracer (uptake phase), the actual bound (Ba) is lower than the equilibrium level (Be): under these conditions, the rate of association of the tracer to the receptors is higher than the rate of dissociation from the receptors, and the net result is an increase in Ba. This phase will last as long as Ba is lower than its equilibrium value (Be). After a variable period of time, the specifically bound will equilibrate (Ba = Be and association = dissociation). Specific binding thus reaches its peak. This equilibrium state will not last, as the plasma concentration and the free level in the brain will continue to decline, so that Be will become lower than Ba. As a result, the rate of dissociation will become higher than the rate of association, the bound concentration will decrease, and the tracer will leave the brain (washout period). The time at which the equilibrium point occurs will depend on multiple factors. It will increase as the peripheral clearance and the regional blood flow decrease, and as the affinity and receptor density increase. As proper derivation of distribution volumes, and therefore BP and V3″, usually requires scanning at least up to a point where the washout phase is captured, increased affinity is associated with longer scanning time. Thus, the affinity of a radiotracer must represent a compromise between the need to measure high signal-to-noise ratio and the need to do so in a reasonable amount of time.

To illustrate this point, simulations were performed in which variation in ligand affinity was related to minimal scanning time. Time–activity curves were formed using the modeled version of a typical arterial plasma input function obtained during a human [11C]DASB study. This input function was then convolved with an impulse response functions having the following k values: K1 = 0.37, k2 = 0.037, k4 = 0.34, and k3 ranging from 0 to 1 in increments of 0.1. Eleven curves were thus generated, with V3″ values ranging from 0 to 2.9 (V3 = k3/k4 ratio, Table 1). These idealized curves were then sampled at the midframe times of our PET [11C]DASB protocol. Total simulated scanning time was 120 minutes. Gaussian distributed noise with zero mean and SD at time t equal to 5% of the noise-free curve value at time t was added. At each level of k3, 100 noisy simulations were performed. Each curve was then fitted by nonlinear least squares regression, using a one-tissue compartment model to generate estimates of the total distribution volume (VT). The final scan frame was then removed from the data and fitting was performed again. This procedure was repeated down to scan duration of 40 minutes (a total of nine scan durations were fitted for each curve). Thus, a total of 9,900 curve-fitting procedures were performed (11 V3″ levels×9 scan durations×100 noise distributions). For each fit, the resulting estimate of VT was normalized to the VT derived with the full dataset. Time independence was considered achieved at time t, if, for frame ending at time t and all subsequent frames, the following two criteria were fulfilled: (1) the average normalized VT was between 95% and 105% of the reference VT (small bias), and (2) the SD of the normalized VT was less than 10% (small error).

Results are presented in Figure 2, which displays the ideal (i.e., noise-free) time–activity curves for each nonzero level of V3″, the time of peak specific binding, and the minimal scanning time required to reach time independence criteria under this noise model. It can be appreciated that the time to reach peak and the minimal scanning time increase as V3″ increases. As all other parameters, including V2, were held constant, the increase in V3″ could be due to increase in either Bmax or affinity. In other terms, for experiments performed at tracer doses, minimal scan duration increases as the regional density of sites increases and as the tracer affinity increases. This simulation also shows that, at this level of noise, the minimal scanning time required to derive VT is longer than the time it takes to reach peak equilibrium. Interestingly, the relationship between these two times was not constant. At V3″ of 0.3, time to peak was 44 minutes, and minimal scanning time was 50 minutes (i.e., 13% longer). At V3″ of 2.9, time to peak was 72 minutes, and minimal scanning time was 110 minutes (i.e., 52% longer).

It is important to stress that this result is affected by the noise level introduced in the simulation. Under the hypothetical situation of a ligand behaving exactly according to the model and of a noise-free experiment, VT could be derived after just a few minutes. Noise is affected by multiple factors, including the size of the ROI under study and subject head motion, and, to a lesser extent, the injected dose and the brain uptake. In addition, the simulation presented here did not take into account the fact that noise increases with the duration of the experiment, because of isotope decay and increasing head motion due to subject fatigue. Because of the combination of factors involved in the determination of the minimal scanning time, this quantity should be estimated on real data. Until this determination has been performed, results derived with a ligand should be viewed with caution.

Another point related to the affinity is the need to maintain a negligible level of receptor occupancy during the scan. Typically, it is considered that an occupancy lower than 5% is required to fulfill the linearity requirement of most analytical methods. The knowledge of the in vivo affinity might permit correction for nonnegligible occupancy, but this requires the use of complex nonlinear models, the reliability of which is not as well established as that associated with linear tracer dose models. Furthermore, nonnegligible occupancy might be associated with pharmacologic effects, which might be undesirable. Thus, higher affinity ligands require lower mass doses and higher specific activity, which might be difficult to obtain routinely. This issue is particularly important in small animal imaging. In conclusion, high affinity is generally a plus for a candidate radiotracer, providing that scan duration remains reasonable and that negligible occupancy can be maintained.

The quality of the input function is also an important parameter defining the usefulness of a tracer. The plasma clearance describes the rate at which the tracer disappears from the blood. Everything else being equal, high clearance is associated with lower brain uptake and shorter minimal scan duration. Lower brain uptake is generally not a problem for tracers with adequate BBB permeability. Shorter minimal scan duration is a positive consequence of fast clearance, as the sequence of events described above (uptake phase, equilibrium point, washout phase) will play out in a shorter time frame. Thus, a fast clearance enables the use of higher affinity tracers. The relationship between rate of plasma clearance and total scan duration is also illustrated in the example of SERT ligands presented below.

Another important property of the peripheral metabolism is that it does not generate radiolabeled metabolites that enter the brain. Sophisticated models have been developed to deal with this issue,10 but it is naturally best to avoid this problem. When the metabolites do possess pharmacologic activity, the problem becomes extremely difficult to solve.

The importance of a high plasma free fraction warrants some discussion. A high plasma free fraction will be associated with higher brain uptake, but again, this factor is rarely a problem for ligands with adequate BBB permeability. The importance of high plasma free fraction is simply due to the fact that it makes this measurement more reliable, and that the knowledge of f1 is important in relating the experimental data to underlying biologic parameters.

As described in the previous section, the only way to estimate free fraction in the brain (f2) is by measuring free fraction in the plasma (f1) and applying Equation 11. The derivation of f2 is essential to estimate the in vivo affinity of a tracer in a manner that will enable meaningful comparison with in vitro data. For example, when calculating the in vivo affinity of the DA D2 receptor radiotracer [123I]IBF,11 we used three concentration parameters (the free plasma, the total plasma, and a region of reference) and obtained three different KD values (0.081, 0.84, and 5.5 nM). The KD value measured relative to the plasma free fraction (0.081 nM) closely approximated the in vitro values (0.06 nM at 22°C, 0.10 nM at 37°C). The KD values measured relative to the total plasma and to the region of reference were off by factors of 10 and 50, respectively. The absence of correction for the free fraction is the reason why the initial estimates of the in vivo affinity of [11C]raclopride were off by a factor of 10.12., 13. In the example below, the f1 of [11C]ADAM and [11C]DAPA was too low to be measured reliably. Therefore, nothing could be said with confidence about the in vivo affinity of these tracers. In contrast, f1 of [11C]DASB and [11C]ADAM were measurable, thus permitting the comparison between their in vivo and in vitro affinities.

The derivation of the in vivo affinity of radiotracers is not just an academic exercise. For radiotracers that cannot be administered in humans at pharmacologic doses (and this is the case for the majority of them), appropriate estimation of KD and free fractions is required to calculate the target occupancy achieved during the scan, and to develop limits to the injected mass.

Beside the issue of deriving the in vivo KD, the measurement of free fraction is also important in studies comparing groups of subjects in which the pathology might be associated with alterations in plasma free binding (such as alcoholism). Short of controlling for this factor, differences in BP could be attributed to differences in receptor density, when they are, in fact, due to difference in plasma protein binding.

Lipophilicity is a critical parameter, as it impacts on plasma free fraction, the ability to cross the BBB and nonspecific binding. Highly lipophilic compounds will generally be associated with lower free fraction in the plasma and the brain. The inconvenience and limitations imposed by a low free fraction in the plasma were discussed above. The importance of low nonspecific binding cannot be overemphasized. Coming back again to Equation 15, increases in nonspecifc binding (V2) will linearly affect the signal-to-noise ratio in the image, as measured with V3″ or any other type specific to the nonspecific ratio. Most of the failures in ligand development result from an unfavorable combination of target density, ligand affinity, and nonspecific binding, all parameters embodied in V3″.

Although lipophilicity is a parameter affecting nonspecific binding, the level of nonspecific binding cannot be predicted by this parameter alone. Other factors, presently unknown, affect this quantity. However, within a given structural family, lipophilicity appears to be associated with nonspecific binding in a relatively predictable manner, as illustrated with the four SERT compounds discussed below.

Also note that the activity uptake in a region of reference is not a rigorous measure of nonspecific binding, as it is influenced by the plasma clearance. For proper comparison between tracers, nonspecific binding should be expressed in terms of “nonspecifc binding potential,” that is, the constant relating free to nonspecifc binding, that is, f2.

Table 2 presents logP, f1, V2′, V2 and f2 for some radiotracers routinely used in neuroreceptor imaging studies. All values (including LogP) were measured in our laboratory, under standard conditions. Table 2 illustrates that logP is neither predictive of f1 (r2 = 0.05), nor of f2 (r2 = 0.15). For example, [11C]raclopride is less lipophilic than [18F]fallypride, but the nonspecific binding of [11C]raclopride is higher than that of [18F]fallypride (this is true both in the plasma and in the brain). Furthermore, f1 and f2 are not correlated (r2<0.01). For example, [11C]MDL 100,907 has a very high free fraction in plasma, but a very low free fraction in the brain. Thus, logP is not a good predictor of nonspecific binding, when ligands of different chemical structures are compared. It is, however, a useful predictor when ligands with similar structures are compared (see the example of SERT compounds below).

[11C]Raclopride, [18F]fallypride, and [11C]WAY 100,635 display V2′ values lower than unity: at equilibrium, the activity in the region of reference is lower than that in the plasma (and, assuming a blood-to-plasma partition coefficient of 1, lower than that in the blood). For these ligands, the free fraction in the brain is actually higher than that in the blood. Conversely, compounds such as [11C]MDL 100,907 and [11C]DASB display a higher free fraction in the blood than the brain. The first situation (V2′ lower than unity) is obviously better, because a lower V2 promotes a high signal-to-noise contrast in the images. However, for these ligands, correction for the activity present in the vasculature is important. Furthermore, if the low nonspecific binding is associated with fast plasma clearance (such as in the case of [11C]WAY 100,635), both factors conspire to decrease the activity in the region of reference to a point where noise in this measure become a significant concern.

In this third section, the principles discussed above are illustrated by presenting a comparison of four new candidate radiotracers developed to image the SERT. [11C]McN 5652 was the first PET radiotracer successfully developed and used to image SERT density in humans.14., 15., 16., 17., 18., 19., 20., 21. For example, [11C]McN 5652 has been used to study 5-HT innervation in MDMA (“ecstasy”) abusers22., 23. and in patients with mood disorder.24 However, [11C]McN 5652 as a PET radiotracer is associated with some limitations:20 (1) the brain uptake is protracted, requiring prolonged acquisition time; (2) the nonspecific binding is relatively high, precluding the reliable quantification of SERT in regions of lower density such as the limbic system; and (3) the plasma free fraction is too low to be measured with accuracy, making it impossible to control for this variable in clinical studies.

More recently, a new group of PET SERT radiotracers from the diarylsulfide class of compounds has been introduced. These include, among others, [11C]ADAM,25., 26., 27. [11C]DASB,28., 29., 30. [11C]DAPA,31 and [11C]AFM.32 The chemical structures of these compounds are presented Figure 3. All of these compounds showed in vivo distribution in rat brain consistent with the known regional density of SERT. Their specific binding was selectively blocked by selective 5-HT reuptake inhibitors. Therefore, all appeared to be suitable candidate radiotracers to label SERT in vivo with PET. The aim of the investigation reviewed here was to compare the imaging performance of these tracers. Details of the experimental procedures and results can be found in Huang et al.26 All experiments (in vitro and in vivo) were carried out in the same laboratory and in the same animals.

Table 3 lists the Ki values of the radiotracers at 22°C and 37°C, as well as the logP values. DASB Ki was significantly higher than the other three compounds (AFM, DAPA, and ADAM). Temperature had no significant effect on the Ki of ADAM, DAPA, and AFM, but significantly decreased the affinity of DASB. Significant between tracer differences were observed for logP (P<0.0001, Table 3). DASB and AFM were significantly less lipophilic than ADAM and DAPA.

At this point, the prediction could be made that AFM should provide higher V3″ compared to DAPA and ADAM, because the affinity of the three compounds is similar, and the lipophilicity of AFM is lower. It is difficult to predict how DASB will behave relative to the other three compounds: it is less lipophilic, but its affinity is lower. It can, however, be predicted that the scanning time should be shorter for DASB than for the three others, due to its lower affinity. Integrating affinity and lipophilicity information for these four compounds, and keeping in mind that low lipophilicity and high affinity are desirable, it appears that AFM provides the best combination of both parameters (Figure 4). Compared to ADAM and DAPA, the lower lipophilicity of DASB is achieved at the price of a lower affinity. In contrast, the lower lipophilicity of AFM is achieved without a decreased affinity.

Clearance and f1 values are presented in Table 4. The clearance of [11C]AFM and [11C]DASB were faster than the clearance of [11C]DAPA and [11C]ADAM. This observation reinforces the prediction that [11C]DASB should require a shorter scanning time, and further suggests that [11C]AFM might be a superior ligand compared to [11C]DAPA and [11C]ADAM. [11C]DASB f1 was significantly higher than that of [11C]AFM, and both were higher than the f1 of [11C]ADAM and [11C]DAPA. Values of f1 provided for [11C]ADAM and [11C]DAPA are only an approximation, due to high binding of the tracers to the ultracentrifugation filter. The higher free fraction of [11C]DASB and [11C]AFM is consistent with their lower lipophilicity (a significant relationship was found between logP and f1, Figure 5). At this point, [11C]DASB and [11C]AFM continue to emerge as the tracers of choice.

[11C]DASB showed significantly lower total brain uptake compared to the other three tracers. This observation could be attributed partly to its fast peripheral clearance, but was mostly due to a lower nonspecific binding (see below). For this group of compounds, the nonspecific binding is the dominant term of the total brain uptake.

Over time, activity concentrated in brain regions with high SERT densities, that is, midbrain, thalamus, and striatum (Figure 6). Intermediate levels were found in the hippocampus, temporal, and cingulate cortices, and lower values were found in other cortical regions. Lowest values were observed in the cerebellum. With all tracers, activity accumulation was readily noticeable in the central structures associated with high SERT densities (midbrain, thalamus, and striatum). However, the higher density of SERT in medial temporal lobe structures such as hippocampus compared to the neocortex translated into a visually detectable activity contrast only in the [11C]DASB and [11C]AFM images (Figure 6). Thus, from a simple examination of the image, it appeared that [11C]DASB and [11C]AFM provide more information about SERT compared to the other two tracers.

[11C]DASB peaked significantly earlier than [11C] ADAM, [11C]AFM, and [11C]DAPA, and washed out significantly faster, an observation consistent with the expectations developed so far. Data were analyzed using a one-tissue compartment model. The minimal scanning time required to reach time independent derivation of VT (average±SD of all ROIS) was 45±13 minutes for [11C]ADAM, 37±11 minutes for [11C]DASB, 69±9 minutes for [11C]DAPA and 64±16 for [11C]AFM. Although this observation was expected for [11C]DASB, the shorter scanning time required for [11C]ADAM was not predicted based on data collected so far.

Values of cerebellar VT′ (i.e., V2′) are presented in Table 5. [11C]DASB showed significantly lower cerebellar VT′ compared to the other three ligands, which did not significantly differ from one another. This observation was consistent with the respective lipophilicity of these compounds. Because reliable measures of f1 were available only for [11C]DASB and [11C]AFM, the free fraction in the nondisplaceable distribution volume, f2, could reliably be derived only for these two tracers. [11C]DASB f2 was 0.77±0.08%, significantly higher than that of [11C]AFM (0.32±0.04%), which was consistent with their respective lipophilicity.

Rank order of BP′ values was [11C]AFM > [11C]DAPA > [11C]ADAM > [11C]DASB, an observation in complete agreement with the affinity values. The Bmax/KD ratios could be calculated for [11C]AFM and [11C]DASB, but not for the other three ligands, given that their free fraction could not be measured with accuracy. For example, in the thalamus, [11C]AFM and [11C]DASB Bmax/KD values were 742±168 mL g−1 and 225±30 mL g−1, respectively. Assuming that these tracers binds to the same number of binding sites, these data indicate that the in vivo affinity of [11C]AFM is 3.3 times higher than that of [11C]DASB. The ratio of affinity deduced from the imaging experiment is in good agreement with the ratio of affinities measured in vivo (Table 6).

Radiotracer rank order of V3″ values was [11C]AFM > [11C]DASB≈[11C]DAPA > [11C]ADAM. As previously stated, V3″ (=f2Bmax/KD) is a decisive outcome measure of the present study for comparing these SERT ligands. In a given region (Bmax being constant), higher V3″ means higher affinity (lower KD) and/or lower nonspecific binding (higher f2), improved signal-to-noise ratio and better measurement reliability. To increase V3″ above values observed with [11C]McN 5652 might not be critical (or even desirable) to measure SERT in regions of high density, such as the midbrain. However, it is essential for the measurement of SERT availability in regions of lower density, such as the neocortex and the limbic system. According to this criteria, [11C]AFM emerged as the leading compound.

Ultimately, this better image quality derives from the combination of its high affinity and low lipophilicity. Coming back to the components of V3″ (=f2Bmax/KD), and recalling that Bmax is constant for these two ligands, we see that, compared to [11C]DASB, the higher lipophilicity of [11C]AFM results in an increase in nonspecific binding by a factor of 2.5 (ratios of respective f2), while the higher affinity results in an increase in specific binding by a factor of 3.3 (ratios of respective in vivo KD, Table 6). Thus, [11C]AFM V3″ is higher than that of [11C]DASB by a margin of 3.3/2.5 = 1.32, that is, a 32% increase. To what extent this increase in V3″ will provide more reliable imaging of SERT in regions of low to moderate SERT density in humans remains to be established. In addition, this increase in signal to noise will be obtained at the expense of longer scanning time. It is thus likely that the choice of [11C]DASB vs. [11C]AFM in clinical investigations will depend on the question being asked, as is the case for the choice between [11C]raclopride and [18F]fallypride when imaging D2 receptors.

Another attractive feature of [11C]AFM is that this ligand can be radiolabeled with 18F. The use of [18F]AFM might make the issue of scan duration less critical, and thus make it possible to take full advantage of the higher V3″ offered by this ligand. In addition, [18F]AFM would make SERT imaging available in PET centers without an on-site cyclotron, given that the distribution network for 18F labeled PET compounds is currently established.

Section snippets

Conclusion

In this article, basic properties of a radiotracer that impact on the qualities of neuroreceptor quantification have been reviewed. These requirements could be summarized in two propositions: (1) V3″should be as high as possible; and (2) V3″ should be measurable within an experimental time frame as short as possible. Clearly, these two propositions involve conflicting requirements as far as affinity is concerned, as increasing the affinity will increase V3″ and the time required to measure it.

Acknowledgements

This work was supported by the National Alliance for Research in Schizophrenia and Depression (NARSAD), the Lieber Center for Schizophrenia Research at Columbia University and the Public Health Service (NIMH K02 MH01603-0, NIMH MH59342-01, NIAAA IP50 AA-1287001).

References (41)

  • M.A. Mintun et al.

    A quantitative model for the in vivo assessment of drug binding sites with positron emission tomography

    Ann. Neurol.

    (1984)
  • J.L. Neumeyer et al.

    [123i]2β-carbomethoxy-3β-(4-iodophenyl)tropane (β-cit): High affinity spect radiotracer of monoamine reuptake sites in brain

    J. Med. Chem.

    (1991)
  • M. Laruelle et al.

    Spect imaging of dopamine and serotonin transporters with [123-i]β-cit: pharmacological characterization of brain uptake in nonhuman primates

    Synapse.

    (1993)
  • B. Levant

    The d-3 dopamine receptor: neurobiology and potential clinical relevance

    Pharmacol. Rev.

    (1997)
  • E.V. Gurevich et al.

    Mesolimbic dopamine d-3 receptors and use of antipsychotics in patients with schizophrenia—a postmortem study

    Arch. Gen. Psychiatry.

    (1997)
  • J.S. Fowler et al.

    Mapping cocaine binding sites in human and baboon in vivo

    Synapse.

    (1989)
  • J.C. Price et al.

    Analyses of [(18)f]altanserin bolus injection pet data. Ii: consideration of radiolabeled metabolites in humans

    Synapse.

    (2001)
  • L. Farde et al.

    Quantitative analysis of d2 dopamine receptor binding in the living human brain by pet

    Science.

    (1986)
  • L. Farde et al.

    Kinetic analysis of central [11c]raclopride binding to d2 dopamine receptors studied by pet—a comparison to the equilibrium analysis

    J. Cereb. Blood Flow Metab.

    (1989)
  • M. Suehiro et al.

    A pet radiotracer for studying serotonin uptake sites: carbon-11-mcn-5652z

    J. Nucl. Med.

    (1993)
  • Cited by (210)

    View all citing articles on Scopus
    View full text