Quantification of Brain Function Using PET

By Academic Press

Functional imaging of the brain is one of the most rapidly advancing areas of neuroscience and Positron Emission Tomography (PET) plays a major role in this progress. This book provides a comprehensive overview of the current status of PET and state-of-the-art neuroimaging. It is comprised of summaries of the presentations by experts in the field. Topics covered include radiotracer selection, advances in instrumentation, image reconstruction and data analysis, and statistical mapping of brain activity. This book focuses on the accuracy of the functional image and the strategies for addressing clinical, scientific, and diagnostic questions.

Covers the PET imaging process from tracer selection to analysis and interpretation
Contains 79 concise reports with abundant illustrations
The definitive state-of-the-art book for functional neuroscience with PET

• Language: English
• Publisher: Academic Press
• Release date: Jul 17, 1996
• ISBN: 9780080540108

Book preview

Spinks.

PART I

TRACER SELECTION

CHAPTER 1

In Vivo–in Vitro Correlations

An Example from Vesicular Monoamine Transporters

MICHAEL R. KILBOURN,     Division of Nuclear Medicine, Department of Internal Medicine, University of Michigan Medical School, Ann Arbor, Michigan 48109

In the development and evaluation of new radiotracers for in vivo imaging using PET (positron emission tomography), researchers often make use of correlations between in vitro and in vivo measures of binding affinities and regional organ distributions of radioligand binding in animals. Using examples drawn from a series of radioligands for the vesicular monoamine transporter, the value of such correlations in choosing candidate radioligands is discussed. In vitro binding affinities are useful in initial selection of candidate radioligands, but other factors such as nonspecific binding and metabolism may be of equivalent importance. Correlations between in vitro and in vivo distributions of radioligand binding sites, if correctly drawn, can provide useful insights into radiotracer selection. Such comparisons of in vitro and in vivo properties of new compounds are thus important components of an overall program to design, synthesize, and validate a new radiotracer for in vivo imaging applications.

I INTRODUCTION

Positron emission tomography (PET) holds great promise as a noninvasive means to examine the biochemistry of the human body. As the field has matured, more complex physiological systems are being studied, using increasingly complicated imaging instrumentation and increasingly sophisticated pharmacokinetic models. All this has the goal of developing new, valuable imaging protocols for the study of normal and pathophysiological functions of the human body. One crucial component of every imaging study is the radiopharmaceutical. As considerable time, effort, and cost is associated with bringing a new radiopharmaceutical to human use, timely and efficient selection of candidate radiotracers is of paramount importance. The requirements for successful radiotracers vary with the intended use, and the optimum characteristics have been repeatedly discussed, rarely with consensus.

As part of the preclinical evaluation process of radiotracers for specific, saturable high-affinity binding sites (e.g., receptors, transporters, ion channels of enzymes), the in vivo properties of new or potential radiotracers are often compared with independent, in vitro measures of binding affinities or in vitro distributions of high-affinity radioligand binding sites within an organ such as the brain. Such correlations can be extremely helpful in identifying valuable new radiotracers and, perhaps just as important, indicate compounds that in all likelihood would be rather poor choices for further development into PET imaging radiopharmaceuticals. In this chapter, we draw upon our experiences with radioligands for the brain synaptic vesicle monoamine transporter (VMAT2) to provide examples of how in vivo–in vitro correlations can be used to identify good candidate radioligands for human imaging with PET.

II MATERIALS AND METHODS

Syntheses of (±)-α-dihydrotetrabenazine ((±)-α-DTBZ), (+)-α-dihydrotetrabenazine ((+)-α-DTBZ) and a series of 2-alkyl-(±)-α-dihydrotetrabenazine derivatives have been described previously (DaSilva et al., 1993; Lee et al., 1994; Kilbourn et al., 1995b). Radiochemical syntheses of (±)-[¹¹C]tetrabenazine ((±)-[¹¹C]TBZ), (±)-α-[¹¹C]methoxytetrabenazine ((±)-α-[¹¹C]MTBZ), (±)-α-[¹¹C]dihydrotetrabenazine ((±)-α-[¹¹C]DTBZ), and (+)-α-[¹¹C]dihydrotetrabenazine ((+)-α-[¹¹C]DTBZ) were done using [¹¹C]methylation of the appropriate desmethyl precursors, by methods previously described (DaSilva et al., 1993; DaSilva et al., 1993b; Kilbourn et al., 1995b). The (±)-α-[³H]methoxytetrabenazine ([³H]MTBZ) was prepared by custom tritiation (Amersham) of the desmethyl precursor. In vivo regional brain distributions in mice were done using iv injections of radiotracer, sacrifice 15 min after injection, and rapid dissection, counting and weighing of tissue samples (DaSilva and Kilbourn, 1992; DaSilva et al., 1994). Ki· values for unlabeled ligands were determined using an autoradiographic method, with [³H]MTBZ as the radioligand (Vander Borght et al., 1995b). Regional distribution of in vitro binding of [³H]MTBZ in rat brain was determined using quantitative autoradiography, as previously reported (Vander Borght et al., 1995b).

III RESULTS AND DISCUSSION

Using the data we have obtained from our series of in vitro and in vivo experiments, as well as the extensive literature data regarding the in vitro properties of (±)-α-[³H]dihydrotetrabenazine ((±)-α-[³H]DTBZ) (Masuo et al., 1990; Scherman, 1986; Scherman et al., 1986, 1988), we can examine whether in vitro-in vivo correlations can be utilized to properly identify the most appropriate radioligand for in vivo PET imaging of the brain synaptic vesicle monoamine transporter in humans. The vesicular monoamine transporter (VMAT2) is a specific protein located in small synaptic vesicles of the monoaminergic neurons of the brain (Schuldiner, 1994). Its function in dopaminergic, noradrenergic, and serotonergic neurons is to transport monoamines (dopamine, norepinephrine, and serotonin) from the cytoplasm into the lumen of the vesicle, protecting the neurotransmitters from degradative enzymes as well as packaging them for exocytotic release. Monoaminergic neurons are found in many regions of the mammalian brain, providing a wide range of in vitro values with which to correlate in vivo radiotracer binding.

A In Vivo-in Vitro Correlations: Binding Affinities vs. Regional Brain Distribution

By the application of in vitro radioligand binding assays, it is now relatively straightforward to determine the in vitro relative binding affinities of a series of compounds, and this is often a first step in the development of new drugs or radiopharmaceuticals. Do such in vitro assays, which determine the ability of test compounds to compete for binding of a well-characterized, high-affinity radioligand to a specific site, necessarily predict eventual in vivo binding to the same site? Can they be used to predict the best compound, or are they more suited to separate a larger series of compounds into good lead candidates (which can then be followed up) and likely failures?

Our data using the benzoisoquinoline derivatives suggest that the latter option is more appropriate. As shown in Figure 1, there is a reasonable correlation between the in vitro binding affinities of these compounds and the ability of to occupy brain VMAT2 binding sites in vivo when all are administered (via peripheral injection) at the same dose. This type of in vivo assay provides an indirect measure of the efficacy of a new drug to bind to the target site, but of course, provides no other information regarding what would be the eventual biodistribution of these compounds if each were independently radiolabeled. Nevertheless, such correlations can provide important information, useful for the selection of candidate radioligands. It is apparent that the derivatives with higher in vitro binding affinities (lower nM values) are, by and large, more effective at blocking radioligand binding in the rat striatum in vivo, with the exception of the isobutyl derivative 7, which has a reasonably good in vitro affinity (Ki of 33 nM) but is ineffective in vivo. The reasons for the poor in vivo effectiveness of the isobutyl derivative are likely pharmacokinetic (poor brain permeability or high serum protein binding or both), but in any case the isobutyl derivative could easily be ruled out as a potential radioligand, and furthermore the smaller chain alkyl derivatives—such as the methyl compounds—would be most attractive.

FIGURE 1 Correlation between in vitro binding affinities and in vivo ability to block radioligand localization in mouse brain striatum. In vitro binding affinities are for competition with [³H]MTBZ binding to rat brain striatal slices. In vivo values are for inhibition of [¹¹C] MTBZ localization in mouse brain striatum following 10 mg/kg. iv coinjections of competing compounds, expressed as a percentage of control values (saline injection). Compounds: 1, 2-β-methyl-α-dihydrotetrabenazine; 2, tetrabenazine; 3, 2-α-methyl-β-dihydrotetrabenazine; 4, 2-α-ethyl-β-dihydrotetrabenazine; 5, 2-α-rt-propyl-β-dihydrotetrabenazine; 6,2-α-isopropyl-β-dihydrotetrabenazine; 7, 2-α-isobutyl-β-dihydrotetrabenazine. The correlation (r² = 0.994) is shown for compounds 1–6.

Can in vitro affinities be used to identify the best candidate? For numerous benzoisoquinolines, the in vitro binding affinities have now been measured, and several of these ((±)-TBZ, (±)-MTBZ, (±)-DTBZ, (+)-DTBZ, (±)-β-methyl-α-DTBZ) have in vitro binding affinities between 1 and 10 nM. As there has never been a firm description of what in vitro binding affinity is best (or even required) for successful, quantitative in vivo PET imaging, we are left with the question; which one to choose? Fortunately for this discussion, four of these candidate compounds have been carbon-11 labeled and examined in both animals and humans, and we can now take a retrospective look at whether in vitro measures might have predicted a best radioligand.

B In Vivo—in Vitro Correlations: Regional Brain Distributions

The determination of in vitro radioligand binding to various brain regions is easily accomplished, using either tissue homogenate binding or quantitative autoradiography. Both methods are capable of providing good measures of the numbers of specific binding sites for a radioligand: autoradiography can provide regional radioligand binding measures with impressive spatial resolution.

For in vivo radiotracer distributions, both cut-and-count dissection techniques as well as ex vivo autoradiography can be used. In general, these methods provide accurate measures of radiotracer distribution into tissues; the proportion of the total radiotracer concentration that is specifically bound to the site of interest, as compared to radioligand in other pools (free, nonspecifically bound, or bound to other sites), is generally determined in separate studies employing pharmacological blocking of the target binding site. Most often, the relatively simple single time point determinations of radiotracer distributions are utilized as estimates of in vivo specific binding; whether more complicated pharmacokinetic modeling of rodent distribution data (entailing many more animals and probably a metabolite-corrected blood curve) would be more appropriate is an interesting question, which will not be addressed here. Certainly, single time point determinations introduce bias into the estimates of in vivo binding parameters, but the magnitudes of these biases will vary with the characteristics of the radiotracer, and therefore need to be considered for each radiotracer.

Do (or should?) in vitro measures of binding sites correlate with such in vivo estimates of specific binding? If it is assumed that all binding sites measured in vitro are similarly accessible in vivo, then the answer to this question is yes; however, if in vivo some binding sites are effectively masked (perhaps through physio-chemical changes in the tissue or occupation by endogenous ligands), then a direct correlation might not be expected. In the latter case, in vitro does not equal in vivo, and attempting to correlate such data would prove useless. In most instances, researchers interested in developing in vivo radioligands have assumed that the in vivo numbers of binding sites do correlate with in vitro measures and have determined the appropriate correlation coefficients. However, as demonstrated in the following examples, there is a correct and incorrect way to present such correlations; and if done correctly they are quite useful in identifying a better radioligand for in vivo imaging.

Using in vitro autoradiography, we have recently determined the regional rat brain distribution of binding sites for [³H]MTBZ (Vander Borght et al., 1995b). The pattern of binding sites is essentially identical to that of [³H]DTBZ binding determined some years earlier (Masuo et al., 1990), and both are very similar to the pattern of [³H]DTBZ binding measured in mouse brain (Scherman, 1986; Scherman et al., 1986). These would therefore be representative of the in vitro levels of VMAT2 in the rodent brain. For in vivo data, we have determined the regional brain distributions of all four carbon-11 labeled benzoisoquinolines in mouse or rat brain (DaSilva and Kilbourn, 1992; DaSilva et al., 1994; Vander Borght et al., 1995a; Kilbourn et al., 1995b). The in vitro and in vivo data (in this case, for MTBZ binding in the mouse brain) can be plotted against one another, and a very typical correlation graph (similar to that found in many papers dealing with evaluation of in vivo radiotracers) is shown in Figure 2. The general assumption is then made that the in vitro and in vivo data are highly correlated, and thus this would be a validated radioligand.

FIGURE 2 Correlation of in vitro [³H]MTBZ binding with in vivo [¹¹C]MTBZ regional brain localization at a single time point after iv injection, for radioligand binding to VMAT2 of the rodent brain.

But is this an appropriate correlation graph? Note that the dimensions of the axis are significantly different; replotting the same data with identical axes gives a plot (not shown) with a very flat line that would suggest little sensitivity of the in vivo radiotracer measure to rather large changes in the in vitro values. But again, this is not an appropriate method to examine the characteristics of a radioligand, and certainly not suitable for comparing radioligands as the difference between two very flat lines might well be indistinguishable.

Is there a better way to plot the data? We suggest that both the in vitro and in vivo data should be normalized to the respective maximum values, such that the brain region with the highest concentrations of binding sites would be 100, and a value of 0 would represent no measurable binding sites. An ideal ligand, with no nonspecific binding, no unbound ligand distribution, and a linear sensitivity to the numbers of sites present, would thus have a slope of 1 and pass through the origin (as depicted by the dotted line in Figure 3). No ideal in vivo radioligand has yet been made, so the actual correlation line will deviate from the ideal case. As shown in Figure 3 for regional [¹¹C]MTBZ distribution in the mouse brain, the slope may well be less than 1, and the y-intercept greater than 0, which represents a nonspecific distribution (free radioligand and nonspecific binding).

FIGURE 3 Correlation of in vitro and in vivo binding of MTBZ to the VMAT2 of the rodent brain. Values have been expressed as a percentage of the maximal value, with the striatum assigned as 100%. The dotted line indicates an ideal radiotracer that has no nonspecific binding, a linear relationship between in vitro and in vivo values, and a slope of 1 for the correlation line.

Are such normalized plots better suited to discern differences between radioligands? Figure 4 shows such correlation lines for two radioligands, (±)-α-[¹¹C]-DTBZ and (+)-α-[¹¹C]DTBZ. For both radioligands, the in vitro-in vivo correlation coefficients are very high (r² = 0.976 and 0.978, respectively). The differences between the radiotracers can now be clearly appreciated, as the single, biologically resolved isomer (+)-DTBZ shows both a larger slope (0.74, but still not 1) and a lower y-intercept. This is, of course, consistent with using a resolved isomer rather than a racemic mixture, as the inactive radiolabeled isomer contributes to tissue radioactivity levels in a nonspecific fashion. But this type of graphical comparison would be just as useful in distinguishing between any two radioligands that might differ in the percentage of nonspecific binding but that, independently, show in vivo distributions highly correlated with in vitro measures of regional binding distributions.

FIGURE 4 Correlation of in vitro and in vivo binding of (+)- and (±)-DTBZ in the rodent brain. Regional values are expressed as a percentage of the maximal binding in the striatum. Correlation coefficients are (+)-DTBZ, r² = 0.978; (±)-DTBZ, r² = 0.986.

Thus, rather simple correlations between single time point measurements of radiotracer distributions and in vitro binding levels can provide meaningful insights into radiotracer selection. But, do they really translate into significant differences in human PET imaging results? We have had a chance to obtain quantitative PET imaging studies with each of the carbon-11 labeled benzoisoquinolines (Kilbourn et al., 1993; Vander Borght et al., 1995a; Koeppe et al., 1995; Frey et al., 1995; Kilbourn et al., 1995a), as for various reasons they were developed sequentially rather than all at the same time. We can thus, in retrospect, compare the in vivo PET imaging results with the in vitro determined densities of [³H]DTBZ binding sites in the postmortem human brain (Scherman et al., 1989). These correlations are somewhat limited due to the paucity of regions sampled in the postmortem assay of the human brain but, nevertheless, provide interesting results as shown in Figure 5. Whereas little significant difference was found between (±)-α-[¹¹C]MTBZ and (±)-α-[¹¹C]-DTBZ (data not shown; (±)-α-[¹¹C]MTBZ seemed, by visual inspection of images, to demonstrate slightly higher nonspecific binding), it is very clear that movement to the resolved isomer (+)-α-[¹¹C]DTBZ has significantly improved the in vivo PET results. Of course, this is also visually evident on the parametric images of radioligand distribution volumes, due to the lower nonspecific and free distribution for (+)-α-[¹¹C]DTBZ. It is encouraging, however, to see that the animal and human results were so consistent; the simplified animal approach of determining a single time point for radiotracer distribution may have in fact underestimated the actual differences between the radioligands, which was better brought out by the pharmacokinetic modeling.

FIGURE 5 Correlation of in vitro binding of [³H]DTBZ and in vivo binding of (+)- and (±)-[¹¹C]DTBZ in the human brain. In vivo values represent the percentage of maximal binding in the putamen, using distribution volumes (DV) calculated through pharmacokinetic modeling.

The correlations shown here and the differences between the different radiotracers are actually not surprising, and certainly do not exemplify all of the radiotracers that have been or are being developed for in vivo PET imaging. The data we have presented is, however, the only complete set of animal plus human data obtained for four closely related carbon-11 labeled radioligands, with all studies done in the same institution and using the same animals, equipment, analytical methods, and personnel. Our results do indicate that in vitro data has a place in the preclinical evaluation of new radiotracers, but other information—such as radiotracer metabolism, toxicity, and serum protein binding—will also play a role in ultimate tracer selection. In our case, (+)-α-[¹¹C]DTBZ is the radioligand of choice, having the best dynamic range in vivo (i.e., the slope of the correlation line (Fig. 4) closest to 1), lower serum protein binding, and no in vivo metabolites that pass the blood-brain