The future of amyloid-beta imaging: a tale of radionuclides and tracer proliferation

Abstract

Purpose of review

This review will focus on the coming proliferation of amyloid-beta imaging tracers and give an opinion on how the Alzheimer’s disease field can develop a systematic means of evaluating which tracers are useful and how the useful tracers compare to each other.

Recent findings

Several new tracers have been reported to be useful for human amyloid-beta imaging. The most recent of these are labeled with fluorine-18. Compared with the 20 min half-life of carbon-11 used in the most widely used tracer, Pittsburgh Compound-B, the 110 min half-life of fluorine-18 allows for wider utilization in research and clinical settings.

Summary

It is likely that more than one fluorine-18-labeled tracer will come into common use. The use of preclinical and clinical ‘bridging studies’ to [C-11]Pittsburgh Compound-B could be a means to determine whether the sizable body of knowledge already gained in [C-11]Pittsburgh Compound-B studies can be applied to the understanding of these new tracers and to form a basis for the comparison among them. This approach could save resources and help sort out a potentially bewildering onslaught of new amyloid-beta imaging tracers.

Introduction

It feels a bit odd to be writing about the future of amyloid-beta (Aβ) imaging when the past history of this technology is brief, only about a decade old. However, this is a quickly evolving field fueled by both research and commercial interests and much has happened over the past 10 years. The field is at a crossroads and careful choices will be important to lead to a systematic and rigorous development strategy of new imaging tracers. These choices will have important implications not only for Aβ imaging, but also for antiamyloid drug development.

The first attempt to image brain Aβ deposition in Alzheimer’s disease was reported by Friedland et al. [1] in 1997 using a monoclonal antibody fragment labeled with Tc-99m for single photon emission computed tomography (SPECT). Although this first attempt failed, it was instructive for subsequent efforts. The antibody fragment was too large to meet the stringent pharmacokinetic demands placed on an in-vivo neuroimaging agent. The limited brain uptake of this Fab fragment doomed the initial effort of imaging Aβ in living humans. The second attempt to image Aβ in the brain of Alzheimer’s disease patients used PET and was presented in preliminary form by Barrio et al. [2] in 1999, and the first full report of this work appeared in 2002 [3]. This tracer is a fluorinated derivative of a nonspecific cellular membrane dye, 1,1-dicyano-2-[6-(dimethylamino)naphthalen-2-yl]propene (DDNP) called [F-18]FDDNP [4]. [F-18]FDDNP appears to have achieved some success as an Aβ imaging agent, but it has been sparingly utilized because of a relatively high background signal [5] and a relatively low specific binding signal [6] providing a limited dynamic range. In addition, there is limited experience with this tracer worldwide at present, as all published reports of human studies come from a single research group [3,6,7]. However, one important characteristic of [F-18]FDDNP is that it is labeled with the 110 min half-life positron emitter, fluorine-18. We will return to this topic in more depth below. The third report of human in-vivo Aβ imaging using a tracer now known as Pittsburgh Compound-B ([C-11]PiB) was presented in preliminary form in 2002 [8], with the full report in 2004 [9]. This tracer has been used in over 40 centers and 3000 participants around the world, including 22 sites in the United States and Canada, seven sites in Europe, ten sites in Japan and Korea and two sites in Australia, and it has been the subject of many independent reports (for a review, see [10]). A key feature of PiB relevant to this discussion is that it is labeled with the 20 min half-life positron emitter, carbon-11. The significance of this also will be discussed below.

Although [F-18]FDDNP was the first fluorine-18-labeled tracer and [C-11]PiB was the first carbon-11-labeled tracer, the results from human studies with at least nine other agents have been reported [11–17]. A task that is already upon the researchers who develop these tracers and will soon be upon the Alzheimer’s disease field as a whole is that of separating the good, the bad and the indifferent on this growing list of Aβ imaging agents. In this review, we will suggest a time-honored approach for the evaluation of amyloid imaging tracers. From the outset, the overarching philosophy is that the future will see common use of more than one Aβ imaging agent. This eventuality will be fueled by developers of tracers using those tracers for their own research purposes as well as by commercial incentives that will bring more than one imaging company into the Aβ imaging arena in the same way that most drug classes have representation by several drug companies. Although a multiplicity of agents can lead to confusion, it is not necessarily a bad thing. A positive aspect of this is the fact that no one tracer is likely to hold all of the advantages and that accessibility to tracers will vary depending on whether they were developed in the academic or commercial realm. However, we are now at a point in the development of this field that we should establish some guiding principles that will promote good research practices and facilitate meaningful interpretation of results across tracers.

The first principle for acceptance of any Aβ imaging tracer must be that it effectively images brain Aβ deposits. To be effective, the quantity of tracer measured in a given brain area must be proportional to the amount of Aβ in that area and there must be sufficient signal-to-noise ratio to detect the tracer when the levels of Aβ have become clinically significant (i.e. the level would be detectable using current postmortem methodologies). These are both the most important and the most difficult criteria for a tracer to meet. There are published data for PiB demonstrating these properties in both postmortem and in-vivo studies [18–20]. It seems reasonable and necessary that when a claim is made for specific detection of any structure in human brain (normal or pathological), postmortem studies should be conducted with brain tissue that has been carefully characterized by independent and quantitative postmortem measures. For example, Aβ can be quantified biochemically by using an enzyme-linked immunosorbent assay [21], immunohistochemically by using antibodies specific for various Aβ epitopes [22,23] or histologically using histochemical stains such as thioflavin-S [24], X-34 [25] or silver stains such as the Bielschowsky method [26]. Any tracer that is claimed to bind Aβ deposits in vivo should also bind to these structures in appropriately prepared postmortem tissue. This can be done in homogenates of frozen tissue using classic receptor binding approaches or in frozen tissue sections using quantitative autoradiographic techniques. The use of fixed tissue may not be appropriate for this purpose because of potential fixation artifacts. In addition to showing that the amount of tracer bound to brain tissue is proportional to the amount of Aβ in the tissue, this exercise can also yield indirect information on the specificity and sensitivity of the tracer. For example, a tracer with high white matter binding may not show proportionality with the amount of amyloid in the tissue due to a large influence of nonspecific white matter binding. Furthermore, one can begin to assess the amount of Aβ content required to give a specific binding signal at least twice that of the nonspecific background signal. This can be considered the ‘best case scenario’ for sensitivity, as many factors will conspire to reduce the sensitivity in vivo. The importance of quantitative homogenate binding or autoradiographic assays cannot be overemphasized. Demonstration of binding by qualitative autoradiographic techniques alone is not sufficient and proper quantitation of autoradiographs is a demanding process. There are several examples in the literature in which qualitative autoradiographic evidence of tracer binding to Aβ deposits did not predict success in vivo [12,27,28]. This may be because nonphysiological conditions are often used in these autoradiographic assays. In addition, the spatial resolution and signal-to-noise ratio that can be achieved with high-resolution microscopic autoradiography cannot be achieved during in-vivo PET or SPECT studies.

Of course, the sine qua non for the validation of an in-vivo Aβ tracer is the in-vivo–postmortem correlation study in humans. Autopsy cases can be difficult to acquire and the postmortem studies are demanding to perform in detail. Fortunately, the amount of Aβ (as determined by [C-11]PiB) does not change appreciably over 2 years during clinical Alzheimer’s disease [29]. This means that Alzheimer’s disease brain tissue obtained a year or more after the in-vivo study should be very informative with respect to the correlation of tracer retention and Aβ content of the tissue. Two in-vivo–postmortem correlation studies have been reported for [C-11]PiB and both showed excellent correspondence between the in-vivo and postmortem measures of Aβ [20,30]. No other Aβ tracer has yet been validated by correlation between in-vivo and postmortem data, but it is expected that these data will be forthcoming for some of the newer tracers.

A second principle for acceptance of any Aβ imaging tracer should be that it can be used reproducibly across many research and clinical settings. Most of the Aβ tracers mentioned above have thus far been used at only one or two sites. As mentioned above, [C-11]PiB is currently in use at more than 30 sites. Several tracers including [F-18]BAY94-9172 [31], [F-18]AV-45 [17], [F-18]AH110690 (or [F-18]3′F-PiB) [16], along with two compounds listed on clinicaltrials.gov called [C-11]AZD2995 and [C-11]AZD2184 (http://clinicaltrials.gov/ct2/show/NCT00692705) have recently been or soon will be employed in multisite phase II trials. Reproducibility across many sites is an indication of the robust nature of the tracer. Such independent validations are a time-honored part of the scientific process and will be a regulatory requirement for any commercial tracer. A byproduct of the use of a tracer at many sites is that the tracer will be studied in many participants. Despite thousands of participants having already been scanned with the most widely studied tracer, [C-11]PiB, we continue to learn more about the capabilities and limitations of this tracer, and a similar number of studies may be required of the other tracers before we have as good an understanding of their performance across a wide range of human studies. In addition, to fully understand the behavior of a tracer in neurodegenerative disease, the research population must be widened beyond Alzheimer’s disease patients and controls. Studies on populations with mild cognitive impairment [32–35], early onset familial autosomal dominant Alzheimer’s disease [36–39], fronto-temporal lobar dementia [40–43], Parkinson’s disease and dementia with Lewy bodies [42,44], cerebral amyloid angiopathy [45], prion disease and other atypical dementias [7,46] and, in particular, cognitively normal aging [33,47–52] have all greatly extended our understanding of Aβ imaging in general and, in particular, of the performance of [C-11]PiB in these special populations. It will be critical to gain similar broad experience with other useful tracers.

A third principle for acceptance of any Aβ imaging tracer must be that it is widely accessible and appropriate for the particular task at hand. It is this principle that brings us back to the discussion of carbon-11 and fluorine-18. The key difference between these radionuclides is their rate of decay or their decay ‘half-life’. This physical parameter determines both how quickly the radiolabeled form disappears from the body and how far they can be distributed from the point of radiochemical production. The decay half-live of carbon-11 is approximately 20 min and that of fluorine-18 is approximately 110 min. Each has advantages. The main advantage of carbon-11 is that it decays so rapidly that sequential imaging studies in the same participant can be performed on the same day. For example, a 90 min [C-11]PiB study can be immediately followed by a study of cerebral metabolism using [F-18] fluorodeoxyglucose (FDG). This would require two separate days with a fluorine-18-labeled Aβ tracer paired with FDG. Another advantage of carbon-11 is that one can benefit from the wealth of experience already gained with [C-11]PiB by using that tracer. However, of the approximately 2000 PET scanner sites in the United States, only a small minority (<10%) have the on-site capability of producing high specific activity carbon-11-labeled products, and the situation is comparable worldwide. The majority of PET scanner sites rely on external production of fluorine-18 radiotracers (such as FDG) by regional cyclotron facilities that distribute the radiotracers to local scanners. The approximately 110 min half-life of fluorine-18 allows distribution within a 2–4 h travel radius. Thus, the need for a good fluorine-18 Aβ imaging tracer is mainly a matter of widening the availability (and thus making the tracer a commercially viable product as well). Most of the new Aβ tracers in the human study phases of development are labeled with fluorine-18. Because the properties of [C-11]PiB appear quite sufficient for imaging Aβ with a carbon-11 tracer, the goal seems to be finding a fluorine-18 tracer with performance characteristics similar to those of [C-11]PiB. This is very important to advancing the field of Aβ imaging, because the availability of a good fluorine-18 Aβ tracer will quickly increase the availability of this new technology by more than 10-fold.

So, how do we enter the future of amyloid imaging? It seems that the carbon-11-labeled tracer, [C-11]PiB, will continue to hold a particular niche in research centers with on-site cyclotron capabilities that are doing detailed imaging studies. In our opinion, the excellent signal-to-noise characteristics of [C-11]PiB and the wealth of validation studies and broad range of geographic and clinical experience make this the tracer of choice when one has the technical capabilities of producing it on-site. But how will we decide which of the many fluorine-18 Aβ tracers in development will serve a useful purpose both in research centers without carbon-11 capabilities and in clinical settings? How will we compare results from fluorine-18 tracers back to the wealth of data available from [C-11]PiB? If, as we predicted above, several fluorine-18-labeled Aβ tracers come into use, how will we compare results obtained from one with another? Will it be practical to expect each agent to be studied in as many patients, in as many different centers and in patients with as broad a clinical base as has already been done with [C-11]PiB before they can be approved for widespread use? Should we require several in-vivo–postmortem correlative studies prior to approval? These would not be completely unreasonable suggestions, but there is an alternative that has been suggested in the reports of several new tracers. The performance of these new tracers is often discussed in relation to the performance of separately published studies using [C-11]PiB. This is a logical way to proceed from the known to the unknown. However, it is often conducted in a nonrigorous manner, with no direct comparison ever performed.

That is, comparisons are being made in the absence of data. These comparative studies could readily be done for any new Aβ tracer. Detailed analysis of the binding properties of any new tracer could be quickly accomplished in homogenates of postmortem Alzheimer’s disease brain. A classic cross-competition study using nonradioactive PiB to compete with radiolabeled test tracer and the corresponding experiment using nonradioactive test tracer to compete with radiolabeled PiB will give important data on whether the two tracers share a common binding site on Aβ and give a relative comparison of affinity. It is possible that a new Aβ tracer binds to a site on Aβ distinct from the binding site for PiB [53]. Unfortunately, these tracers will not be able to take advantage of the wealth of data already available for [C-11]PiB and it falls upon the developers of these tracers to produce an equivalent body of data to validate their tracer.

More important than the in-vitro comparisons, the short half-life of [C-11]PiB facilitates direct in-vivo comparison of this tracer to putative fluorine-18-labeled tracers. It is relatively straightforward to administer [C-11]PiB and follow this immediately with a fluorine-18-labeled Aβ tracer as is often done in same day [C-11]PiB/[F-18]FDG experiments. Having these direct comparison data could remove much of the necessity to acquire a large body of data similar to that already available for [C-11]PiB. To be sure, a sufficient body of new studies will be necessary to establish the safety and efficacy of each individual new tracer, but the number of studies, the number of sites involved and the types of patients studied could be lessened if one can be confident how the new tracer performs relative to the most widely studied tracer currently available. In addition, having this one piece of common ground (i.e. performance relative to [C-11]PiB) obviates the need to compare each new fluorine-18-labeled tracer directly to every other fluorine-18-labeled tracer in the same patients – an obviously impractical comparison to make.

Conclusion

In summary, as we look to the immediate future, we are on the verge of a proliferation of new Aβ imaging tracers. The majority of these will be fluorine-18-labeled tracers. This is fueled by the commercial and scientific need to expand the availability of this technology to the more than 90% of PET scanners that do not have the on-site capabilities to produce high specific activity carbon-11-labeled tracers. It is likely that more than one fluorine-18-labeled tracer will come into common use. The Alzheimer’s disease field will need to develop a systematic means of evaluating which tracers are useful and how the useful tracers compare to each other. One approach would be to demand, for each new tracer, the assembly of a degree of preclinical and clinical experience similar to the best understood, currently used tracer, [C-11]PiB. Another method would be to use preclinical and clinical ‘bridging studies’ to [C-11]PiB as a means to determine whether the knowledge already gained in [C-11]PiB studies can be applied to the understanding of the new tracer and to develop a common ground among new tracers. This concept is not new. It is an established method of drug development after the establishment of an effective example in a given class of compounds. This approach could save resources and help sort out a potentially bewildering onslaught of new Aβ imaging tracers. Careful evaluation at this juncture will go a long way to ensuring that the future of Aβ imaging is long and productive.