Molecular imaging in Alzheimer's disease: new perspectives on biomarkers for early diagnosis and drug development

Recent progress in molecular imaging has provided new important knowledge for further understanding the time course of early pathological disease processes in Alzheimer's disease (AD). Positron emission tomography (PET) amyloid beta (Aβ) tracers such as Pittsburgh Compound B detect increasing deposition of fibrillar Aβ in the brain at the prodromal stages of AD, while the levels of fibrillar Aβ appear more stable at high levels in clinical AD. There is a need for PET ligands to visualize smaller forms of Aβ, oligomeric forms, in the brain and to understand how they interact with synaptic activity and neurodegeneration. The inflammatory markers presently under development might provide further insight into the disease mechanism as well as imaging tracers for tau. Biomarkers measuring functional changes in the brain such as regional cerebral glucose metabolism and neurotransmitter activity seem to strongly correlate with clinical symptoms of cognitive decline. Molecular imaging biomarkers will have a clinical implication in AD not only for early detection of AD but for selecting patients for certain drug therapies and to test disease-modifying drugs. PET fibrillar Aβ imaging together with cerebrospinal fluid biomarkers are promising as biomarkers for early recognition of subjects at risk for AD, for identifying patients for certain therapy and for quantifying anti-amyloid effects. Functional biomarkers such as regional cerebral glucose metabolism together with measurement of the brain volumes provide valuable information about disease progression and outcome of drug treatment.

(atrophy, volume changes and cortical thickness) by magnetic resonance imaging, but also to visualize and quantify brain pathology (fi brillar Aβ, tau, activated microglia and astrocytosis) as well as functional changes (cerebral glucose metabolism, neurotransmitter and neuroreceptor activity) by positron emission tomography (PET) ( Table 1). Molecular imaging thus provides important insight into the ongoing pathological processes in AD in relation to clinical symptoms and disease progression. An important step forward has been in vivo imaging of Aβ pathology in AD patients. Although the histopatho logical confi rma tion of diagnosis at autopsy is impor tant, it refl ects the end stage of a disease that may have been ongoing for decades.
Th e new molecular imaging techniques provide possibilities to develop early diagnostic biomarkers for early detection of AD at preclinical stages, as well as for monitor ing eff ects of drug therapy. Recent research has thus also changed the view on incorporating biomarkers into the standardized clinical diagnosis of AD as suggested by Dubois and colleagues [7,8] and the recommendations from the National Institute on Aging-Alzheimer Association workgroups on diagnostic guidelines for AD [9,10].

Amyloid imaging in Alzheimer's disease patients
Among the fi rst Aβ PET tracers was Pittsburgh Compound B ( 11 C-PIB) when 16 AD patients were initially scanned in Sweden [11]. Th e high 11 C-PIB retention observed in cortical and subcortical brain regions of mild AD patients compared with age-matched healthy subjects has consistently been confi rmed with 11 C-PIB in several other studies (for a review see [12][13][14]). Several other Aβ PET tracers have also been tested in AD and control patients [12,15] although so far 11 C-PIB is the most explored. 18 F-labeled tracers will probably be more suitable for use in the clinic, with their longer half-life. 18 F-FDDNP was the fi rst 18 F-PET tracer used for visualizing Aβ plaque in AD patients [16], showing lower binding affi nity to Aβ plaques than 11 C-PIB but also suggested to bind to neurofi brillary tangles [16,17]. Th e 18 F-labeled Aβ PET tracers 18 F-fl utemetamol, 18 F-fl orbetapir and 18 F-fl orbetaben have shown promising results in AD patients [18][19][20].
Th e PET Aβ tracers quantify fi brillar Aβ in the brain by binding in the nanomolar range to the Aβ peptide [21]. Th e in vivo 11 C-PIB retention correlates with 3 H-PIB binding as well as levels of Aβ measured in autopsy AD brain tissue [22][23][24][25]. 18 F-fl orbetapir PET imaging has also been shown to correlate with the presence of Aβ amyloid at autopsy [26], as well as 18 F-fl utemetamol PET imaging to amyloid measured in cortical biopsies [27].
A still unknown factor is the relationship between fi brillar Aβ (plaques) and soluble Aβ oligomers. Presently there is no information on how the smaller soluble Aβ oligomers, which are known for triggering synaptic dysfunction [28][29][30], can be visualized in vivo in man with the presently available Aβ tracers. It is therefore a challenge to try to develop PET tracers that can visualize these smaller forms of Aβ in the brain, although the probably lower content of oligomers in AD brains compared with fi brillar Aβ might be a limiting factor. Th e soluble Aβ oligomers are important since they probably can induce and interfere with the neurotransmission in the brain [30,31].

Longitudinal PET amyloid studies in Alzheimer's disease patients
Th ere are still few longitudinal studies of Aβ PET imaging in AD patients. Th ese studies are important to understand the rate of accumulation of amyloid in the brain and are important for evaluation of intervention in antiamyloid drugs. A 2-year follow-up study with 11 C-PIB in AD patients revealed at group level consistent stable fi brillar Aβ levels in the brain [32]. Two additional 1-year and 2-year follow-up studies confi rmed these observations [33,34] as well as a recent 5-year follow-up PET study of the fi rst imaged PIB PET cohort [35]. In the latter study it was evident at the individual level that increased, stable and decreased PIB retention were observed and the disease progression was refl ected in signifi cant decline in cerebral regional cerebral glucose metabolism (rCMRglc) and cognition [35]. In a recent 20-month follow-up study, Villemagne and colleagues reported a 5.7% increase in fi brillar Aβ in AD patients [36]. Th e longitudinal imaging studies mainly support the assumption that the Aβ levels in the AD brain reach a maximal level at the early clinical stage of the disease, although both increase and decline in later stages of the disease cannot be excluded [12,37,38].
Amyloid imaging in mild cognitive impairment patients 11 C-PIB PET studies in mild cognitive impairment (MCI) patients have revealed a bimodal distribution. Both high (PIB + ) and low (PIB -) retention of the PET tracer has been demonstrated [39,40]. PIB + MCI patients seem to have a greater risk to convert to AD after clinical followup compared with PIB -MCI patients [39,41,42]. Figure 1 illustrates high 11 C-PIB retention in a MCI patient (PIB + ) who later converted to AD in comparison with a nonconverting MCI patient (PIB -). PIB + MCI patients show comparably high 11 C-PIB retention to AD patients ( Figure 1). We recently observed a signifi cant increase in brain 11 C-PIB retention in early MCI patients when rescanned after 3 years [35]. Th e MCI patients also showed a decrease in rCMRglc while they remained stable in cognitive function at follow-up [35]. Jack and colleagues [34] and Villemagne and colleagues [36] have also reported annual changes in 11 C-PIB retention. Th ese fi ndings support a continuous increase in Aβ load in the early stage of prodromal AD [35] (Figure 2).

Amyloid imaging in older subjects without cognitive impairment
High Aβ has been measured in older cognitive normal controls (for a review see [43]). Th e reported percentage of positive Aβ PET scans varies from 10 to 50% between diff erent cohorts of studied older people without cog nitive impairment [44,45]. A possible explanation for variation in percentage of Aβ PET-positive cognitive normal subjects could be age but also genetic background (APOE genotype). Aβ alone most probably does not account for the decline in memory in older people. Further longitudinal studies are needed to investigate to what extent these Aβ-positive older people with normal cognition will later convert to AD [46]. In a recent longitudinal study of 159 older subjects with normal cognition and PIB + , PET showed a greater risk for developing symptomatic AD within 2 to 5 years compared with PIBsubjects [47].

Relationship between brain amyloid and cerebrospinal fl uid biomarkers
Th ere is a strong inverse correlation between accumulations of fi brillar Aβ in the brain as measured by 11 C-PIB and levels of Aβ 1-42 in cerebrospinal fl uid (CSF) [39,[48][49][50][51][52][53][54][55]. An inverse correlation between 11 C-PIB retention and CSF Aβ  has been demonstrated in prodromal AD (MCI) earlier than changes in functional parameters (cerebral glucose metabolism, cognition) [54] (Figure 2). Figure 3 illustrates the inverse relationship between Aβ in the brain and the CSF as analyzed with statistical parametric mapping cluster analysis. A positive relation ship has also been observed between 11 C-PIB retention and levels of CSF tau and p-tau [39,50,51,54]. Which of the biomarkers are most sensitive to detect the earliest pathological signs of the disease is still unclear.
Some data suggest that 11 C-PIB PET imaging detects amyloid pathology prior to CSF biomarkers [39,49,54]. Soluble Aβ oligomers might be the most pathogenic in AD. An interesting observation is therefore that AD patients with the APP arctic mutation show no fi brillar Aβ in the brain (PIB-negative) but a reduction of Aβ 42 in CSF as well as a reduction in cerebral glucose metabolisms by PET [56].

Imaging of infl ammatory processes in Alzheimer's disease brain
Infl ammatory processes have been suggested to cause the pathological processes of AD [57,58]. Amyloid has been observed to mobilize and activate microglia [59]. Activated microglia are found in autopsy brain tissue at sites of aggregated Aβ deposition of AD patients. Th e peripheral benzodiazepine receptor PET tracer 11 C-(R)-PK11195 has been used for measuring the transition of microglia from a resting state to an activated state in the brain. An increase in 11 C-(R)-PK11195 binding was described by Cagnin and colleagues in the temporoparietal, cingulated and ento rhinal cortices of AD patients as a sign for strong micro glia activation compared with controls [60]. Edison and colleagues demonstrated high cortical 11 C-(R)-PK11195 binding with reciprocal negative correlation with cogni tive performance in AD patients [61]. In some other studies, a lower level of microglia activation was observed in mild AD and MCI [62,63]. 11 C-DAA-1106 is a new peripheral benzodiazepine PET tracer that has shown increased binding in several brain regions including the frontal, parietal, temporal cortices and striatum of AD patients compared with age-matched controls [64].
Activated astrocytes participate in the infl ammatory processes occurring around the Aβ plaques. An increased number of astrocytes have been measured in autopsy brain tissue from AD patients, especially those with the Swedish APP mutation [65]. A positive correlation has been observed between 3 H-PIB binding and GFAP immunoreactivity in autopsy AD brain tissue [25]. It is assumed that synaptic activity might be coupled to utilization of energy through an interaction between astrocytes and neurons where the astrocytes take up glucose and release lactate to neurons [66].
N-[ 11 C-methyl]-l-deuterodeprenyl ( 11 C-DED) has been shown to irreversibly bind to the enzyme monoaminooxidase B expressed in reactive astrocytes. 11 C-DED has therefore been tested as a PET ligand for measurement of activated astrocytes. Increased 11 C-DED binding was demonstrated in the brain of patients with Creutzfeldt-Jacob disease [67]. We have recently observed by PET an increased 11 C-DED binding in the cortical and subcortical brain regions of MCI patients compared with AD patients and controls [68]. Th ese observations suggest that astrocytosis might be a very early event in the time course of pathological processes in AD (Figure 2). Further studies are needed to explore the relationship between Aβ and infl ammatory processes in the early stages of AD.

Imaging of functional changes in Alzheimer's disease brain
Brain glucose metabolism 2-[ 18 F]-fl uoro-2-deoxy-d-glucose ( 18 F-FDG) has been widely used both in research and clinically for measurement of regional changes in rCMRglc in AD [10]. A reduction of rCMRglc is often observed in the parietal, temporal, frontal and posterior cingulate cortices. Th e decline in rCMRglc is more regional specifi c compared with the increased 11 C-PIB retention in large areas of the AD brain [11,32]. Th e hypometabolism is often more severe in early-onset AD compared with late-onset AD, while no diff erence in regional 11 C-PIB retention has been observed between early-onset and late-onset AD [69]. 11 C-PIB PET seems to detect prodromal AD at an earlier disease stage and better separates between MCI subtypes (amnestic versus nonamnestic) than 18 F-FDG [39,58,70]. Th e decline in rCMRglc follows, in contrast to PIB, the clinical progression of AD and shows a strong correlation with changes in cognition [32,35,58,70]. Figure 3 illustrates the correlation between rCMRglc and episodic memory (Rey Auditory Verbal Learning) and between 11 C-PIB and episodic memory (Rey Auditory Verbal Learning) as analyzed with statistical parametric mapping analysis. Th e 18 F-FDG uptake shows more brain regional specifi c clusters compared with 11 C-PIB [54].

Neurotransmitter and neuroreceptor imaging
Several neurotransmitters are impaired in AD, especially the cholinergic system but also the dopaminergic and serotonergic neurotransmitter. Several PET tracers have been developed and tested for measuring the diff erent neurotransmitters, enzymes and various subtypes of receptors in AD patients [10]. PET tracers are available for studying dopaminergic, serotonergic and cholinergic systems [12] (Table 1). Th e cholinergic neurotransmission has so far been the focus for clinical AD therapy. It is therefore worth mentioning that decreases in nicotinic receptors have been demonstrated by PET in AD patients using 11 C-nicotine [71] and 18 F-fl uoro-A-85380 (α4 nicotinic receptors) [72]. Th e extent of reduction in 11 C-nicotine binding correlated with the reduction in level of attention of the AD patients [71]. Presently there is a great interest to develop selective PET tracers for imaging of the α7 nicotinic receptors in the brain since these receptors interact with Aβ and might therefore be a new target for AD therapy [73].

Imaging biomarkers and drug development
Recent progress in molecular imaging and biomarkers indicates that subtle pathological changes indicative for AD disease might be detected decades prior to clinical diagnosis of AD. Diff erences in the time course are observed between pathological and functional AD imaging biomarkers (Figure 2). PET imaging allows measurement of pathological processes such as depo sition of fi brillar Aβ plaques, levels of activated microglia and astrocytosis. Th ere is a need for further exploration of PET tracers visualizing infl ammatory processes that might occur at very early disease states (Figure 2). Similarly, there is a great need for PET tracers visualizing the accumulation of Aβ oligomers in diff erent stages of AD ( Figure 2). Preclinical data for the new promising PET ligand THK 523 for in vivo tau imaging have recently been presented [74]. Additional PET studies are needed to predict with more accuracy the time course for changes in neurotransmitter function including the nicotinic receptors. Brain atrophy changes (magnetic resonance imaging) correlate closely with cognitive decline and disease progression but less with amyloid load in the brain [14,20,75].
Th e rapid development of molecular imaging will be important not only for early diagnostic biomarkers and early detection of AD [7][8][9]46] but also to select patients for certain drug therapies and to identify diseasemodifying therapies and testing in clinical trials (Table 2). PET imaging bio markers could thereby play an important role in identifying patients with elevated risk of developing AD. In addition, fi brillar Aβ imaging could (together with CSF Aβ 42 ) serve as an inclusion criterion as well as a primary outcome in phase 2 and a secondary outcome in phase 3 drug trials. Measurement of rCMRglc and magnetic resonance imaging atrophy changes are probably most useful for predicting the clinical outcomes of drug therapy.
Th e multi-tracer PET concept off ers unique opportunities in drug trials to study pathological as well as functional processes and to relate these processes in the brain to CSF biomarkers and cognitive outcomes (Figure 4). Th ere is now an increased interest to introduce diff erent biomarkers into clinical trials in AD patients [76], which will be important for all drug candidates in the pipeline for AD trials [77]. Long-term treatment with cholinesterase inhibitors in AD patients has shown signifi cant correlation between the degree of inhibition Signifi cant clusters of covariance between regional cerebral glucose metabolism (rCMRglc) and scores in RAVLtot tests. Data from [54].
of acetylcholinesterase in the brain, the number of nicotinic receptors, rCMRglc and clinical outcome of treatment measured as attentional test performances [78][79][80][81][82]. To evaluate the eff ect of new disease-modifying thera peutics, imag ing of fi brillar amyloid, activated microglia, astrocytosis, tau in addition to rCMRglc and structural brain changes should be applied to determine whether anti-amyloid strategies may clear the amyloid plaques from the brain but also slow down disease progression. A few PET studies in AD patients have shown reduction of brain Aβ measured by 11 C-PIB following anti-amyloid treatment [81,83,84] but the disease-modifying eff ects still have to be proven.

Competing interests
AN is an investigator in clinical trials sponsored by Novartis AB, Jansen-Cilag, Torrey Pines Therapeutics, GSK, Wyeth and Bayer; served on an advisory board for Elan, Pfi zer, GSK, Novartis AB, Lundbeck AB, and GE Health Care; served on an advisory board for Elan, Pfi zer, GSK, Novartis AB, Lundbeck AB, Merck and GE Health Care; received honorarium for lectures from Novartis AB, Pfi zer, Jansen-Cilag, Merck AB, Ely Lilly and Bayer; and received research grants from Novartis AB, Pfi zer, GE Health Care and Johnson & Johnson.

Table 2. Clinical implications of molecular imaging in Alzheimer's disease
To increase the understanding of pathophysiological mechanisms To increase the understanding of time course of disease progression To understand the diff erences in time course between pathology and functional changes To develop diagnostic markers that can predict rate of progression To enable selection of Alzheimer's disease patients to certain therapy To measure brain changes after short-term and long-term therapeutic intervention that correlate with clinical symptoms Figure 4. Multi-tracer positron emission tomography concept to study pathological and functional processes in the brain. Multi-positron emission tomography tracer concept applied in drug trials combined with atrophy studies (magnetic resonance imaging (MRI)), cerebrospinal fl uid (CSF) biomarkers and cognitive testing. 11 C-PMP, acetylcholinesterase; 11 C-PIB, amyloid; 11 C-deprenyl, astrocytosis; 11 C-nicotine, nicotinic receptors;