PET imaging has been used as a noninvasive tool to probe many questions with implications for the field of psychopharmacology. Some of these are examination of receptor/transporter availability at many different neuror-eceptor targets in patient populations relative to healthy control subjects, measurement of drug receptor occupancy through competition models, detection of fluctuations in endogenous neurotransmitter levels following phar-macologic or task-based stimulation, and examination of various metabolic processes. This essay is organized in three sections, followed by brief concluding remarks: (1) a brief description of the physical principles underlying the imaging system, (2) a description of the pharmacokinetic methodology common to most PET studies, and (3) a survey of the currently available probes and targets studied with PET.
Positrons are the antiparticles of electrons. They have the same mass as electrons but positive, rather than negative electric charge. Unstable isotopes that emit positrons when they decay can be produced in a cyclotron; many academic medical centers currently have onsite cyclotrons that can produce isotopes for PET imaging. Several of these isotopes can be readily incorporated into small molecules suitable as biologic tracers. Among these are carbon (11C, half-life = 20.4 min), fluorine (18F, halflife = 109.8 min), and oxygen (15O, half-life = 122 s). When a positron is emitted, it briefly interacts with local atoms, losing kinetic energy during the resulting collisions. When its energy is low enough, the positron will be attracted to and captured by an electron; the pair will mutually annihilate and emit a pair of 511 keV photons. Conservation of momentum dictates that the total momentum of the 2 photons is equal to that of the positron-electron pair at the time of their annihilation. As this momentum is small, the photons travel in nearly opposite directions (their momenta cancel each other), approximately along a line from the location of the annihilation event (Fig. 1, left panel). The imaging system contains
Positron Emission Tomography (PET) Imaging. Fig. 1. Left: A schematic detector ring. An annihilation event occurs in an imaged object as seen along the long axis of the subject (from head to toe). If the emitted photons (paths represented by dashed arrows) strike detectors A and B within a coincidence timing window (several nanoseconds for most scanners), a coincidence event is recorded along the line between A and B. Right: A schematic stack of rings, viewed from the side. A tomographic image is formed by combining data from all the rings.
rings of detectors composed of scintillating materials that surround the imaged object. Combined with electronic components that are attached to these detectors, the arrival of photons at the detector arrays can be counted with very high temporal resolution (on the order of nanoseconds). If a pair of photons are detected in two different detectors within a specified time interval (a "coincidence timing window,'' typically 4-12 ns), the scanner treats these as having originated from a single annihilation event occurring somewhere along the line between the two detectors (a "line of response''). Computer-intensive mathematical methods can then be applied to this collection of coincidence event data to infer the original distribution of the annihilation events. Typical modern systems contain from 20 to 40 of these detector rings stacked in parallel to form a cylindrical field of view (Fig. 1, right panel); the data from these can then be recombined to form a three dimensional image of the original distribution of the radiotracer, composed of a series of slices through the object (a "tomogram," literally a "slice picture,'' see Fig. 2).
The preceding description of the data acquisition and reconstruction is greatly simplified - there are many confounds and sources of noise and image degradation that must be accounted for. For example, many of the photons produced by annihilation events interact with the tissues intervening between the locus of the annihilation and the detector; some of these are scattered in different directions and others absorbed completely ("attenuated"). A description of the methods for correcting these confounds is beyond the scope of this chapter (see Cherry et al. 2003; Valk et al. 2003 for detailed expositions) but a key point is that PET utilizes coincidence detection, in contrast to the related technologies such as single photon computed tomography (SPECT). This results in an accurate estimate of the fraction of photons that are attenuated in any line of response through an imaged object, and this, in turn, leads to a level of quantitative accuracy that is not practically attainable with other technologies. Because of this level of quantitative accuracy, the concentration of radiotracer in tissues can be accurately inferred, during a single contiguous interval of time, or dynamically over a sequence of sampling times after injection of the tracer. The implications for psychopharmacology are that detailed mathematical models can be formed of the tracer pharmacokinetics, the interaction between tracer phar-macokinetics and endogenous transmitters or
Positron Emission Tomography (PET) Imaging. Fig. 2. A PET image and coregistered high resolution MRI from the same subject. The data, acquired using the dopamine D2 antagonist radioligand [11C]FLB457, are summed over the entire 90 min of the scan. The saggital MRI slice (bottom right) shows the slice levels for the transverse and coronal views.
exogenously administered drugs, or, of rates of metabolic processes. Physiologic parameters can be estimated by fitting data to these models.
General principles: The ► pharmacokinetic schemes used to estimate parameters both for ► reversible binding receptor radioligands and for many metabolic processes are called compartment models. Compartments can be spatially distinct, but they can also be different states of the radiolabel occupying the same spatial domain, such as bound versus unbound, or parent radiotracer versus metabolic product. Figure 3 is a schematic representation of a standard compartment model used with reversibly binding radioligands.
Corresponding to the compartment model, are systems of linear first-order, ordinary differential equations (ODEs). The ODEs express the temporally dynamic relationship between the input source of radioligand to the tissue (either the concentration in arterial plasma, or the concentration in a tissue devoid of the biologic process of interest, but similar to the tissue of interest in other respects, a ''reference tissue'') and the resulting concentration in the tissue of interest (usually brain or specific brain regions for targets of interest in psychopharmacology). ''First-order'' implies that the models operate according to the convention that movement from a source compartment to a target compartment is proportional to the concentration in the source compartment; the validity of this assumption in the case of reversibly binding receptor ligands is predicated on the use of tracer dose so that the associated mass action law, which is second-order when higher concentrations are used, can be treated as pseudo
Positron Emission Tomography (PET) Imaging. Fig. 3. A 2-
tissue compartment model (2TC). CP radioligand concentration in arterial plasma;CND nondisplaceable compartment, the sum of free and nonspecifically bound tracer in tissue;CS specifically bound ligand, i.e., ligand-receptor complex;^ through k4 are rate constants governing the fractional transfer between compartments per unit time.
first-order. By using some data-fitting method such as least squares minimization to regress observed data onto the model, the rate constants can be estimated. In turn, various mathematical combinations of the rate constants represent physiological parameters of interest. There is an extensive literature on methods for data fitting and parameter estimation in PET (see Slifstein and Laruelle 2001; Slifstein et al. 2004; Valk et al. 2003 for overviews).
Receptor imaging. The form of PET imaging most frequently used in applications relevant to psychophar-macology involves the use of radioligands that bind selectively and reversibly to target neurotransmitter receptors and transporters. These are usually administered as a single ► bolus injection, though there are some tracers amenable to a ► bolus plus constant infusion administration that induces steady-state conditions in ligand concentrations. The compartment configuration shown in Fig. 3 presents a model frequently used with reversible tracers. In the figure, CP represents the concentration of radioligand in the arterial plasma, the input to brain tissue. CND (''nondisplaceable'') is the sum of freely dissolved tracer in the brain and tracer that is nonspecifically bound to membranes. These quantities are combined to a single compartment based on the assumption that equilibration between free and nonspecifically bound ligand occurs on a much more rapid time scale than the specific binding process does, and can therefore be treated as if constantly in equilibrium. Models in PET frequently involve simplifying assumptions of this type in order to insure that they are not over-parameterized for statistical fitting. CS represents specifically bound ligand-receptor complex. The movement of tracer between CP and CND is governed by a transport law, whereas exchange between the states CND and CS is governed by a mass action law. These various states of the radioligand cannot be distinguished in the PET signal; it is comprised of the sum of all sources of radioactive decay in the spatial locus being imaged, and is sometimes referred to as CT (total concentration, Fig. 4). The parameters of greatest interest in receptor imaging are the density of receptors available for binding to the radioligand ( ► Bmax or Bavail, nM) and the ► affinity of the radioligand for receptor (KD_1 where KD (nM) is the ► equilibrium dissociation constant). These quantities cannot be estimated separately from single tracer dose scans - multiple scanning sessions with increased radiotracer concentrations that bind to a significant fraction of receptors would be required for this purpose. The quantity that is readily estimated is the ► binding potential, a parameter that is proportional to the product of Bmax and affinity, or Bmax/KD. There are several possible constants of proportionality, according to which
Positron Emission Tomography (PET) Imaging. Fig. 4. Arterial plasma input and resulting PET data and model fit. These data show the time course of the dopamine Dt receptor tracer [11 C]NNC112 in the striatum of an anesthetized baboon. Plasma input to brain (CP) on left, and the compartments represented in Fig. 3 on the right. The discrete dots represent the measured CT, the three continuous curves are the fit to the model.
of several approaches to experimental design and derivation of the binding potential estimate is used. The version appearing most frequently in the literature, however, is called BPnd (Innis et al. 2007), equal to fND Bmax/KD where fND is the fraction of CND that is freely dissolved, fND = free ligand/ (free + nonspecifically bound ligand). The rate constants can be shown to have the following physiological interpretations: K1 = FE, where F is flow (actually, perfusion, mL cm-3 min-1) and E is the firstpass extraction fraction (unitless), k2 = K1/VND where VND is the nondisplaceable equilibrium distribution volume, CND/CP at equilibrium, (mL cm-3), k3 = konfNDBmax (min-1), where kon is the association rate of the receptor-ligand complex and k4 (min-1), is koff, the dissociation rate of the receptor ligand complex. As Kd is equal to koff/kon, it is apparent that k3/k4 is equivalent to BPND. However, BPND is rarely estimated directly from the fitted values of k3 and k4. Rather, more involved methods utilizing various constraints to make the estimated BPND more statistically robust are employed (see Slifstein and Laruelle 2001; Slifstein et al. 2004; Valk et al. 2003 for further details).
Applications of receptor imaging to psychopharmacology: If another ligand - an endogenous transmitter or an exogenously administered drug - competes with the radi-oligand at the binding site on the receptor, then it can be shown, again invoking tracer dose conditions for the radioligand, that the binding potential is reduced by the factor 1/(1+L/Ki), where L is the concentration of the ligand and Ki is its affinity for the binding site. This in turn implies that the relative difference between the binding potential with and without the competing ligand on board, [BPND(competitor on board)-BPND(base-line)]/BPND(baseline), is equal to L/Ki/(1+L/Ki), the fraction of receptors occupied by the competitor. This technique can be used to infer the occupancy of a receptor by a drug, or to demonstrate that a stimulus induces transmitter release.
In a seminal study, Farde et al. (1988) used the dopamine D2 receptor radioligand [11C] raclopride to measure the occupancy of D2 receptors by the antipsychotic drug ► haloperidol in patients with schizophrenia, and presented the concept of a therapeutic window - the idea that there is a minimal receptor occupancy necessary to achieve antipsychotic efficacy, but a maximum tolerable occupancy above which extrapyramidal symptoms would appear (Fig. 5). The in vivo competitive binding technique has subsequently been used in many published studies examining drug receptor occupancy. While a large number of these have continued to examine D2 occupancy by antipsychotics, the method has been used to look at other receptor systems as well, including various 5-HT receptors, nicotinic and muscarinic ACh receptors, NK1, adenosine 2, histamine, CB1, mu-opioid receptors, 5-HT and DA transporters, and other targets. The approach has been widely used by pharmaceutical companies in both drug development and in postmarketing studies characterizing efficacious occupancies.
A similar imaging technique can be used to infer fluctuations in endogenous neurotransmitters as a result
Positron Emission Tomography (PET) Imaging. Fig. 5. The concept of a therapeutic window, showing a range over which an antipsychotic drug is efficacious, but below the threshold for extrapyramidal symptoms. (After Farde et al. 1988.)
of either pharmacological or task-based stimuli. Again, the system that has been the most amenable to this type of study has been dopamine release at the D2 receptor. Many studies have been performed in which dopamine release and/or inhibition of reuptake has been induced by ► amphetamine, ► methylphenidate, or other compounds, or by tasks hypothesized to induce dopamine release, for example, by incorporating a monetary reward for accurate task execution. There has been considerable evidence that the simple competitive interaction model described earlier is not adequate to explain the decrease in binding potential following amphetamine. In particular, the decrease lasts much longer than the apparent increased dopamine release as measured with ► microdialysis in animal models. On the other hand, the binding decrease is highly correlated with amphetamine dose and with the microdialysis measurements, and does not occur if dopa-mine stores have been pharmacologically depleted prior to the scan, suggesting that the competitive binding plays at least some role in the observed effect. Also, while an effect consistent with the occupancy model has been detected with several radioligands in the benzamide ([123I]IBZM, [11C]raclopride, [18F]fallypride) and catecholamine ([11C]NPA) classes, paradoxical increases in radioligand binding following amphetamine have been observed using the ► butyrophenone radioligand N-[18F]methylspiperone. A number of mechanisms have been proposed to explain both the extended duration of the amphetamine effect and the different results observed with different radioligands, including receptor trafficking and differential responses to internalized receptors, differences in binding sites, and differences in ► pharmacoki-netic properties with concomitant differences in robustness of quantification of the various radioligands. At this time, these issues remain unresolved. Recently, there has been some investigation of the use of agonist, rather than antagonist radioligands, based on the premise that they may be more sensitive to the affinity state of the receptor for endogenous neurotransmitters, and, therefore, endogenous dopamine might compete more successfully with these, leading to greater sensitivity to the differences across conditions or populations in stimulated dopamine release. Studies using anesthetized animals have demonstrated increased sensitivity of [11C]NPA and [11C]PHNO, both D2/3 agonists, to amphetamine stimulation relative to [11C]raclopride, and early studies with [11C]PHNO in healthy human volunteers have also shown increased amphetamine effect in the dorsal stria-tum compared to previously published reports utilizing [11C]raclopride. The competitive binding method has proved to be much more difficult to use with receptors other than the dopamine D2 receptor. Several investigators have been unable to detect the amphetamine effect on binding of dopamine D1 radioligands. Researchers have had mixed results detecting pharmacologically-induced increases in serotonin levels as well, with some investigators reporting decreases in [18F]MPPF or [18F]MEFWAY binding to 5-HT1A receptors in animal models following stimulated increases in serotonin levels either by ► fenfluramine or ► SSRIs, while others have been unable to detect changes with [18F]MPPF or [11C]WAY100,635. See Laruelle (2000) for a comprehensive examination of competitive binding techniques in PET and SPECT, especially as pertains to dopamine D2/3 receptor imaging.
Finally, it is worth noting that receptor imaging has been used to infer some more subtle pharmacological effects, such as the "GABA-shift" observed at the ► benzodiazepine binding site on the GABAa receptor with the PET radioligand [11C] ► flumazenil, in which radioligand binding increased when GABA levels were increased by reuptake inhibition (Frankle et al. 2009), presumably due to increased affinity through allosteric interaction between the GABA and benzodiazepine binding sites.
Metabolism imaging. The models used for metabolism imaging can vary potentially according to the mechanism being studied. In this section, two widely used metabolism tracers are described. [18F]DOPA and [18F]FDG. Both of these ligands are substrates for some, but not all enzymes that act on an endogenous compound, and thus partially follow the same metabolic pathway. Both of these reach a stage in the metabolic process in which they are
Positron Emission Tomography (PET) Imaging. Fig. 6. An irreversible trapping model.
assumed to be irreversibly trapped, and thus the model is based on the assumption of irreversible accumulation. In each case, there has been a considerable body of literature demonstrating that this assumption is oversimplified and that more accurate estimates of the measured process can be obtained using more complex models that account for the further progress of the radiolabeled metabolites after the putative trapping stage. Nonetheless, the trapping models are still widely used due to their simplicity and ease of implementation, and so are described here. A basic compartment model for ► irreversible trapping is shown in Fig. 6. Unlike the receptor models in the previous section, there is no notion of equilibrium. As long as free radioligand is in the tissue, the concentration in the trapped compartment continues to increase as k3 times the free concentration. However, one can envisage a hypothetical steady state in which influx to the tissue from plasma just balances the sum of efflux back to the plasma plus the conversion into the trapped form, so that the free concentration is constant. Under these conditions, the free concentration equals K^(k2 + k3) times the plasma concentration CP, and the steady state rate of conversion into the trapped form, therefore, equals Kxk3/(k2 + k3) times the plasma concentration. This parameter is often referred to as Kin, the steady state uptake rate for an irreversible compartment model.
[18F]DOPA is a substrate for amino acid decarboxylase (AADC), the enzyme that catalyzes l-dihydroxyphenylala-nine (DOPA) into ► dopamine within the dopaminergic terminals. [18F]DOPA readily passes the barrier and cell membranes, and is metabolized by AADC into 6-fluorodopamine (6-FDA), effectively 18F labeled dopamine, in dopaminergic terminals. 6-FDA, like endogenous dopamine, does not cross the blood-brain barrier and so is "trapped" in the proximity of the terminals at the stages of loading into vesicles, release through exocytosis and reuptake into the terminal, the main path followed by endogenous dopamine in the striatum, where reuptake through dopamine transporters is the dominant mode of clearance from the synapse. Thus Kin, estimated with the irreversible trapping model, represents a lumped marker of presynaptic dopaminergic condition. 6-FDA is, however, a substrate for monoamine oxidase (MAO) and for catechol-O-methyltransferase (COMT). Both of these enzymes are present in the extracellular environment, and the radiolabeled metabolic by-products from 6-FDA metabolism can readily diffuse across the blood-brain barrier. Numerous studies have demonstrated that failure to account for this loss of radiolabel from the brain results in underestimation of the true uptake rate, and several approaches have been proposed for its incorporation into the modeling and design of [18F]DOPA experiments. See Cumming and Gjedde (1998) for a detailed discussion of the metabolism of [18F]DOPA.
18F labeled 2-fluoro-2-deoxy-D-glucose (FDG) is arguably the most extensively used PET radioligand. The development of FDG for use with PET followed the groundbreaking work of Sokoloff et al. (1977) with [14C]DG, a 14C labeled tracer which is an analog of glucose and partially follows its metabolic pathway. Thus, FDG is a useful probe for measuring the cerebral glucose metabolism rate (CMRglu). In more recent times, techniques such as fMRI have become more prevalent for studies measuring brain metabolic activity, owing to their better temporal and spatial resolution and the less invasive nature of the procedure. FDG has gone on to be used extensively as a clinical tool in radiological diagnostic procedures in other fields such as oncology. But given the historic nature of the role of FDG in the use of PET to study the brain, a brief description of the model is included here. FDG is a substrate for the same carrier protein that transports glucose into brain tissue. It is also a substrate for hexokinase, the enzyme that metabolizes glucose, and is phosphorylated into FDG-6-PO4. FDG-6-PO4 does not follow further steps of glucose metabolism. It does dephosphorylate, but because dephosphorylation of FDG-6-PO4 is slow compared to the forward process, FDG-6-PO4 is treated as a trapped state. Here, Kin represents the steady state phosphoryla-tion rate of FDG. This is proportional, not identical, to CMRglu, owing to the fact that the transport and phosphorylation rates are different for the two compounds. When multiplied by a conversion factor accounting for this difference (1/LC, the ''lumped constant,'' because it lumps two conversion factors together) then Kin/LC times the plasma glucose concentration is taken as an estimate of brain glucose metabolism. In analogy with observations made about metabolism of [18F]DOPA, the dephosphorylation step, while small, still contributes to
Positron Emission Tomography (PET) Imaging. Table 1. A sample of PET probes currently used in research with human subjects.
D2/3 antagonist; useful in striatum only
Was this article helpful?