Autosomal-dominant Alzheimer's disease: a review and proposal for the prevention of Alzheimer's disease

Autosomal-dominant Alzheimer's disease has provided significant understanding of the pathophysiology of Alzheimer's disease. The present review summarizes clinical, pathological, imaging, biochemical, and molecular studies of autosomal-dominant Alzheimer's disease, highlighting the similarities and differences between the dominantly inherited form of Alzheimer's disease and the more common sporadic form of Alzheimer's disease. Current developments in autosomal-dominant Alzheimer's disease are presented, including the international Dominantly Inherited Alzheimer Network and this network's initiative for clinical trials. Clinical trials in autosomal-dominant Alzheimer's disease may test the amyloid hypothesis, determine the timing of treatment, and lead the way to Alzheimer's disease prevention.

development of the transgenic animal models used in AD research today. Knowledge of the molecular mechanisms of the identifi ed mutations has catalyzed identifi cation of the causative pathogenic events in AD in humans. Indeed, this avenue of research has provided the most compelling case for a unifying theory of AD.
In addition to contributing to advances in the basic scientifi c understanding of AD, ADAD families represent an ideal population for preventative and treatment trials for several reasons. First, there is near certainty (~100%) regarding development of the disease with a known mutation that enables prevention studies and increases the power of treating minimally or presymptomatic patients. Second, the approximate age at which symptoms are likely to develop can be predicted in individuals who are completely asymptomatic, allowing therapeutic trials years or decades before clinical onset. Finally, ADAD research participants are highly motivated, relatively young, and have minimal co-morbidities. By engaging those at risk for ADAD, uniquely informative scientifi c information about disease progression, biomarkers and changes due to therapeutic treatments are expected to lead to advancements in drug development.
Disease-modifying therapeutics have been largely developed with animal models based on human diseasecausing mutations. ADAD caused by known mutations most closely resembles those models, and therefore is more likely to respond to disease-modifying treatments. Results from treatment trials in ADAD will bridge cellular and mouse therapeutic research with SAD therapeutic research. Because the clinical and pathological phenotypes of ADAD are similar to the more common late-onset AD, drugs that prove successful in the prevention or delay of dementia for ADAD are likely to provide guidance for future prevention and disease modifi cation in late-onset AD. Successful implementation of prevention and symptomatic studies will therefore inform about the causes of AD and will provide guidance for future therapeutic development.
In the present review, we present historical and current information about ADAD, including: discovery of the genetic mutations; clinical, pathological, imaging and biomarker fi ndings; the explosion of understanding about AD based on basic science studies of genetic mutations and development of AD animal models from the mutations; and an international multicenter eff ort to understand the cascade of events leading to AD toward future trials to treat -and even prevent -the onset of dementia in those with mutations.

A brief history of autosomal-dominant Alzheimer's disease research
Provocative supportive evidence indicates that Dr Alois Alzheimer's fi rst case may have been ADAD. Th is case (August D), described in 1906, was early onset, possibly familial, and from a region of Germany associated with the PSEN2 Volga-German mutation [2]. Th e fi rst documented cases of familial AD were identifi ed in earlyonset dementia with pathological confi rmation [3,4]. Other notable early studies identifi ed pedigrees in which more than 10 individuals over fi ve generations were aff ected by early-onset AD [5]. Aff ected individuals developed symptoms before age 60 with progressive amnesia and other signs of cortical cognitive impairment as seen in late-onset SAD [6]. Neuropathological exami na tion of these early cases demonstrated extensive amyloid and neurofi brillary pathology with neuronal loss and gliosis.
In 1963, a case series with early-onset AD in 11 of 26 children with an aff ected parent and no aff ected individuals in the pedigree without an aff ected parent developing the disease suggested that early-onset AD was the result of a fully penetrant autosomal-dominant mutation [7]. Th e search for a dominant mutation focused on chromosome 21, due to the Alzheimer's-like pathology seen in older patients with Down syndrome (trisomy 21). In 1987, a genetic linkage study in four large ADAD families found a gene locus at 21q11.2 to 21q22.2, but not in the 21q22 region associated with the Down syndrome phenotype [8]. Th en, in 1991, a missense point mutation (Val-Ile) at codon position 717 was discovered in the APP gene in a single family with linkage to chromosome 21 [9]. Th is report identifi ed the specifi c mutation in this family and provided a possible mechanistic link between the APP mutations and abnormalities in amyloid processing seen in these families. Most of the variants in APP occur between residues 714 and 717 near the putative site for γ-secretase cleavage [10]. At least 38 additional ADAD APP mutations have since been identifi ed.
One year after the discovery of mutations in APP as a cause of ADAD, four diff erent laboratories identifi ed another locus for ADAD on 14q24 [11][12][13][14]. Th e gene PSEN1 was cloned 3 years later, encoding the protein presenilin 1 [15]. Presenilin 1 is a highly conserved membrane protein required for γ-secretase to produce amyloid-beta (Aβ) from APP [16]. Since the initial fi nding of the PSEN1 mutation, approximately 180 diff er ent mutations that cause ADAD have been identifi ed (http:// www.molgen.ua.ac.be/ADMutations/). Within a year of cloning PSEN1, a gene with substantial nucleotide and amino-acid homology was discovered on the long arm of chromosome 1 in two families [15]. Th is gene, PSEN2, appears to account for only a small percentage of ADAD cases and may be associated with a later age of onset and slower disease progression than mutations in PSEN1 and APP.
Th e discovery of the genetic causes of ADAD catalyzed research on the relationship of ADAD to SAD. Th e clinical, imaging, pathologic and biochemical relation ships have been individually described by groups around the world, each following a relatively small number of aff ected families. While the pathogenic cause of ADAD is an inherited mutation, the molecular pathogenic causes of SAD have not yet been identifi ed. Th erefore, although the two forms of the disease may have fundamentally diff erent initial pathways, they share a remarkably similar pathophysiology. Th ese descriptions have provided key insights into the causes of both SAD and ADAD. Th e characteristics of ADAD compared with the more common sporadic late-onset AD are summarized in Table 1.

Clinical presentation of ADAD
In broad terms, the clinical presentation of ADAD is very similar to that of SAD. Like SAD, most ADAD cases present with an insidious onset of episodic memory diffi culties followed by inexorable progression of cortical cognitive defi cits. Th e most obvious diff erence between familial and sporadic cases of AD is the younger age at onset in individuals with ADAD mutations. Th e youngest ages at onset are with PSEN1 mutations; symptoms typically fi rst appear between the ages of 30 and 50 years, but some families have individuals aff ected in their 20s [17]. APP pedigrees tend to have a later age at onset, typically in the 50s and ranging from 45 to 60 years old. Th e rarer PSEN2 mutations have a wide range of onset with some relatively late-onset cases. Overall survival in ADAD is similar to that of SAD, with the caveat that survival length in very elderly sporadic individuals tends to be lower. If younger onset (<65 years old), and there fore healthier, sporadic cases are compared with ADAD individuals, their survival is not very diff erent. PSEN1 mutation carriers may have slightly shorter survival. Comparisons of disease duration are notoriously diffi cult, particularly as recognition of the onset of problems may be earlier in familial individuals who are aware of their at-risk status -particularly those enrolled in longitudinal studies.
Th e majority of ADAD cases have an amnestic presentation very similar to that seen in sporadic disease, with the fi rst defi cits being in visual and verbal recall and recognition. Longitudinal studies of unaff ected at-risk individuals have suggested that the earliest neuro psycho metric fi ndings involve a fall in verbal memory and performance IQ scores [18], with relatively preserved naming [19]. Atypical language and behavioral presen tations occur in a minority of both sporadic and familial cases.
Neurological signs and symptoms appear to be more common in ADAD. Myoclonus and seizures are both relatively more frequent; myoclonus may be a harbinger of later seizures. A number of PSEN1 mutations are variably associated with a spastic paraparesis (and charac teristic histopathology) and extrapyramidal and cerebellar signs.
APP mutations that cluster within the Aβ coding domain around positions 692 to 694 do tend to have a phenotype that is diff erent to sporadic disease -cerebral hemorrhage is a characteristic feature probably related to extensive amyloid angiopathy. Amyloid angiopathy and seizures are also a feature of the APP duplication pedigrees [20]. Apart from some mutation-specifi c exceptions and the earlier age at onset, ADAD is remarkably similar to SAD, with as yet unexplained heterogeneity being a feature of both forms of the disease.

Neuropathology
Th e principal neuropathological changes in ADADneuronal loss, neurofi brillary tangles, senile plaques, and cerebral amyloid angiopathy (CAA) -mirror those seen in SAD, providing strong support for ADAD as a model for studying AD (Figure 1). In vitro and in vivo studies have shown that dominant mutations frequently increase Aβ42 and Aβ40 deposition and alter the Aβ42/Aβ40 ratio [21]. Postmortem studies confi rmed elevated levels of brain Aβ42 in persons with APP mutations compared with SAD [22]. APP mutations increase Aβ production by diff erent mechanisms. Mutations adjacent to the βsecretase cleavage site increase cleavage by β-secretase, generating increased Aβ40 and Aβ42 from APP [23]. APP mutations around the γ-secretase cleavage sites result in modifi cation of γ-secretase activity, enhancing only the production of Aβ42 [24]. PSEN1 and PSEN2 mutations alter the conformation of the γ-secretase complex, increasing production of Aβ42 from APP [21]. Post mortem studies have shown that PSEN1 and PSEN2 muta tions are related to increased levels of insoluble Aβ42, and to a lesser extent insoluble Aβ40, compared with SAD [25][26][27][28]. A comparable Aβ42/Aβ40 ratio between SAD and PSEN mutations has also been reported [29,30], although other research has reported a signifi cantly increased Aβ42/Aβ40 ratio in PSEN1 and PSEN2 mutations when compared with SAD, primarily due to higher levels of Aβ42 [31].
Distinctive neuropathological features are found in some pathology case reports and may be related to mutation type. Th ese variant pathologies may aff ect the pharmacological response, tolerability, and biomarker measurements of experimental agents in clinical trials into SAD. Th ese include cottonwool plaques, severe CAA, intracerebral hemor rhage, cerebellar plaques, and Lewy bodies. Cottonwool plaques are large, ball-like plaques lacking dense amyloid cores that have been reported with PSEN1 mutations, especially in mutations beyond codon 200 [32]. Cottonwool plaques have been associated with spastic paraparesis and seizures [29]. CAA is common in SAD, but may be more prominent with specifi c ADAD muta tions. Th e Dutch, Flemish, and British APP mutations occurring within the Aβ coding region typically feature severe CAA, with intracerebral hemorrhage occurring in persons with the Dutch mutation. Larger and denser Aβ deposits around vessels or ring-like plaques staining for Aβ42 instead of Aβ40 have been reported with some APP mutations compared with SAD [33,34]. PSEN1 mutations after codon 200 show a higher incidence of severe CAA compared with SAD [29]. Cerebellar plaques with the British APP and some PSEN1 mutations have been reported [22]. Lewy body pathology has been reported in the amygdala and neocortex with some PSEN1 and PSEN2 mutations [35], as has been reported in SAD. Variability in phenotypic and patho logical expression has been reported within families, suggesting that genetic or epigenetic factors might be exerting disease-modifying eff ects [31].

Neuroimaging
A growing number of neuroimaging studies have demonstrated evidence of early alterations in brain structure All brain tissues were routinely fi xed in formalin and were paraffi nembedded. Sections were 12 μm thick. Aβ42 was detected using polyclonal antibody C42 (with formic acid pretreatment), kindly provided by Dr Takaomi Saido (RIKEN Brain Science Institute, Tokyo, Japan). PHF-1 tau was detected using PHF-1 monoclonal antibody (with microwave pretreatment), kindly provided by Dr Peter Davies (Feinstein Institute of Medical Research, New York, USA). and function in carriers of autosomal-dominant mutations prior to the onset of clinical dementia. Early magnetic resonance imaging (MRI) studies demonstrated that hippocampal atrophy was present in presymptomatic and early symptomatic carriers, which paralleled the develop ment of verbal or visual memory defi cits, in a pattern similar to that seen in SAD [36]. More severe medial-temporal lobe atrophy may be present in symptomatic ADAD carriers compared with SAD [37]. Graymatter regional volume loss and decreases in magnetization transfer ratio have also been reported in mildly symptomatic carriers [38]. Longitudinal structural imaging studies have demonstrated an accelerated course of atrophy compared with SAD, in both regional-medial temporal lobe and whole-brain measures [39][40][41] and in cortical thinning, particularly evident in the precuneus and posterior cingulate prior to the diagnosis of dementia [42]. Alterations in white matter structure have also been demonstrated in presymptomatic and early symptomatic carriers, with decreased fractional anisotropy in the fornix and widespread areas of brain visualized with diff usion tensor imaging [43].
Presymptomatic alterations in brain perfusion and metabolism, similar to the patterns reported in SAD, have also been reported among ADAD carriers using nuclear medicine techniques, including single photon emission tomography [44,45] and positron emission tomo graphy (PET) [46,47]. One study demonstrated early glucose fl uorodeoxyglucose-PET hypometabolism in the posterior cingulate cortices, hippocampus and entorhinal cortices of presymptomatic carriers of ADAD mutations, which was present prior to signifi cant atrophy in these regions [48]. Functional MRI techniques have demonstrated alterations in hippocampal activity during episodic memory tasks in presymptomatic ADAD carriers that appear to occur decades prior to dementia [49], similar to the observations in young apolipoprotein E ε4 carriers [50,51], however, this observation was not replicated in a larger population of ADAD mutation carriers in a study employing an implicit novelty encoding paradigm [52].
More recently, PET amyloid imaging studies with Pittsburgh Compound B (PiB) have revealed evidence of fi brillar Aβ deposition in ADAD, including carriers who were up to 10 years younger than the age of onset for their family [53][54][55]. Interestingly, these studies have consistently reported elevated levels of PiB retention in the striatum of presymptomatic ADAD individuals, which occurs more variably in late-onset SAD.

Biomarkers
Th e biochemical changes in the brain, cerebrospinal fl uid (CSF) and blood of persons with AD have been described in detail over the past 30 years. Many biochemical changes in the brain have been documented to occur in the AD process, with those biomarkers associated with amyloid plaques and neurofi brillary tangles being specifi c for pathologically defi ned AD [6,56]. Th e identifi cation of Aβ as the major component of CAA [57] and amyloid deposits in plaques [58] was followed by the fi nding that tau is the major component of neurofi brillary tangles. In addition to AD-specifi c protein deposition, biochemical changes in synaptic, infl ammatory, oxidative, and cell cycle markers occur in the AD brain [59].
Multiple groups have reported that CSF Aβ42 in ADAD participants is reduced to approximately one-half of normal values [60,61], a fi nding remarkably similar to SAD [62,63]. While decreased Aβ42 appears to have remarkable specifi city for pathologic AD and Aβ amyloid osis in the brain [64], CSF Aβ40 is not consis tently diff erent in AD individuals compared with normal individuals. CSF tau and phospho-tau levels are increased almost two-fold in ADAD individuals compared with controls [60,61], again mimicking the CSF profi le in lateronset SAD. Th e relative age at which CSF biomarker changes occur in ADAD has not yet been adequately characterized, although it appears to predate clinical symptoms.
Increases in plasma Aβ42 have been consistently found in ADAD, while there is little, if any, consistently reported diff erence in SAD [65][66][67]. Other blood-based biomarkers have not yet reproducibly diff erentiated ADAD or SAD from controls.

Mutations
Identifi cation of mutations in the substrate APP as well as in the proteases PSEN1 and PSEN2 that cleave APP to produce Aβ peptides provides very strong support for the amyloid hypothesis in AD [68]. Th e mutations in the APP gene are clustered around the three cleavage sites ( Figure 2). Only one mutation (the Swedish mutation) increases Aβ generation by increasing β-secretase process ing of APP. Most of the other mutations aff ect the biophysical properties of the Aβ peptide and have a diverse array of eff ects, but, as indicated in Figure 2, they consistently increase the toxic amyloid potential of the protein, thereby increasing the tendency of Aβ to oligomerize. Th is is particularly clear for the most abundant mutations aff ecting the γ-secretase cleavage sites, which all result in the generation of the longer Aβ42 peptide. Th e rationale for therapeutic strategies that target decreas ing the Aβ generated from the APP protein in these families is obviously strong, and β-secretase or γsecretase inhibitors are predicted to work as they act on the enzymes and not on the APP substrate [69]. For immunization strategies, APP mutations in the Aβ sequence may or may not interfere with the binding of particular antibodies.
In contrast to the localized APP mutations, the presenilin mutations are scattered throughout the presenilin protein, although most are clustered along the diff erent transmembrane domains in the hydrophobic core of the protein (Figure 3). Functionally, most presenilin mutations cause a loss of function of γ-secretase activity; that is, they reduce the cleavage rate of the diff erent substrates of the enzyme [70]. Pathologically, they most probably operate in a similar way as the APP mutations by enhancing the toxic amyloid potential of the residual Aβ peptides that are generated by the mutated presenilin/γsecretase. Indeed, although many mutations lower Aβ40 production, almost all mutations increase or at least do not aff ect the production of the Aβ42 peptide [71]. Th e overall result is a change in the Aβ42:Aβ40 ratio, which increases the tendency to form toxic oligomeric species [72]. γ-Secretase inhibitors may have less eff ect on mutated γ-secretase than on wild-type γ-secretase [73][74][75]. In preparation for treatment trials, individual mutations can be tested in vitro for γ-secretase inhibitor eff ects on γsecretase activity. While it is likely that lowering the total burden of Aβ peptide might be benefi cial, caution is needed because it is possible that some γ-secretase inhibitors could block mainly the wild-type γ-secretase while the mutant presenilin remains operational. β-Secretase inhibitors or vaccination against Aβ avoid this particular issue as they target the wild-type β-secretase or the wild-type Aβ.

Mouse models
Th e creation of AD animal models was crucial to the development of modern anti-amyloid therapeutic The APP mutations are all clustered in or around the amyloid-beta (Aβ) peptide sequence, and this region is therefore displayed enlarged using the single amino acid code. White circles, mutations found; red font, resulting amino acid substitutions. Mutations cluster around the α-secretase, β-secretase and γ-secretase sites as indicated. They have various eff ects on the generation of Aβ as indicated, but their overall eff ect is an increased tendency to generate toxic species. pro grams. Initial eff orts to develop an AD model focused on transgenic mice overexpressing human APP, since no naturally occurring animal models fully recapitulate all of the pathological and functional defi cits in AD. Overexpression of the wild-type APP was insuffi cient to cause a relevant phenotype. With the discovery of the familial APP mutations, however, several animal models using the Swedish, London, Indiana and other mutations have been developed and characterized. Most of these mouse models show consistent amyloid pathology, but often there is poor correlation between the development of morphological brain changes of deposition of amyloid plaques and disturbances in learning and memory function. Mouse models with only presenilin 1 or presenilin 2 mutations have been developed, but they do not develop amyloid pathology in spite of increased production of Aβ42 [76,77]. Th e inability of presenilin mutations to cause amyloid pathology in mice is most probably due to the sequence diff erences of mouse APP compared with human APP, as murine Aβ peptides are less prone to aggregation. Accelerated brain pathology was achieved by combining the genetic liability of human APP mutations with presenilin mutations [78]. In addition, the behavioral disturbances are more pronounced in these bigenic animals [79].

KM
Transgenic models of ADAD are quite diff erent from human models because of species diff erences and the location and increased amount of expression of the mutated protein. Transgenic models can be useful for drug development, however, because they develop amyloidosis and express altered Aβ peptides similar to human carriers of the mutation. Th erapies that show pathological effi cacy should therefore also be able to exhibit similar activity in humans; for example, decreasing overall amyloid peptides and normalizing the Aβ42:Aβ40 ratio. Because most of the treatments currently in clinical trials have been developed in mice carrying an ADAD mutation, they are likely to be more eff ective in ADAD compared with SAD. Finally, although all of the mouse models demonstrate disturbances of amyloid production and metabolism, they are not full models of AD. Conclusions about the therapeutic effi cacy whether the eff ect of the mutation on amyloid-beta (Aβ) production has been investigated: green, mutations that decrease Aβ40 production; yellow, mutations that increase Aβ42 production. In all cases, an increase of the Aβ42/Aβ40 ratio has been found. of drugs tested in mouse models must therefore be made cautiously.

Current treatment trials
Current trials for the common form of AD include approaches to target Aβ by decreasing production [80,81], increasing clearance [82][83][84], and other attempts to ameliorate the toxic eff ects of the amyloid cascade. Alternative targets at various stages of drug development include tau, infl ammation, neurotransmitter modulators, and other approaches. Th e diverse approach to drug discovery in AD is helpful for the fi eld, as there has not yet been a successful disease modifi cation trial. Reasons cited for the lack of clinical trial success over the past decade include inadequate preclinical models, few trials completing phase III studies, few studies with demonstrated pharmacodynamic activity, treating the disease process too late in the disease course, or targeting an insignifi cant mechanism. Treatment trials in ADAD provide an opportunity to address several of these concerns of treating too little, too late -with designs that demonstrate target engagement followed by prevention studies to alter the course of changes that occur in the disease process.
Despite the opportunity for prevention studies in persons destined to develop AD because of ADAD mutations, we are aware of only one such study being performed [85]. Six presymptomatic known PSEN1 muta tion carriers are being treated in an open-label fashion with HMG-CoA reductase inhibitors (either ator vastatin or simvastatin). In addition to cognitive outcome measures, CSF indices (Aβ42, tau, p-tau181, sAPPα, and sAPPβ) are being obtained. In a preliminary report, a lowering of CSF sAPPα and sAPPβ associated with HMG-CoA reductase inhibitors was observed in PSEN1 mutation carriers without an eff ect on Aβ42, tau, or p-tau181. Although small in scale, this biomarker study represents an important initial step towards larger eff orts to explore preventative interventions in ADAD.

The Dominantly Inherited Alzheimer's Network
Owing to the geographically dispersed nature of ADAD families and the relative rarity of the disease, an international network of research centers has been established by the National Institute on Aging to adequately power studies in this uniquely informative population. Th is network, formally known as the Dominantly Inherited Alzheimer's Network (DIAN), will enable international longitudinal studies of persons with dominantly inherited mutations that cause AD. Th is is the fi rst large-scale, multicenter, systematic eff ort to use standardized instruments to identify and uniformly evaluate individuals with dominantly inherited AD. Th e DIAN aims to determine the chronological changes in cognition and biomarkers in relation to clinical onset and progression of dementia in a well-characterized and uniformly studied group of persons at risk for ADAD. Th e DIAN investigators will assess and quantify the ability of clinical, biological and imaging markers to predict and track the progression of AD. Th e DIAN's overriding purpose is to contribute to the search for meaningful therapies for AD by helping elucidate the cascade of events that lead to dementia of the Alzheimer's type.
Th e specifi c aims for the DIAN include the following. First, to establish an international registry of individuals (mutation carriers and noncarriers; presymptomatic and symptomatic) who are biological adult children of a parent with a known causative mutation for AD in the APP, PSEN1, or PSEN2 genes in which the individuals are evaluated in a uniform manner at entry and longitudinally thereafter. Th e second aim is to obtain clinical and cognitive batteries that comprise the Uniform Data Set of the National Institutes of Health-funded Alzheimer's Disease Centers, supplemented by web-based neuropsycho logical tests. A further aim is to implement structural, functional, and amyloid imaging protocols (3T MRI, fl uorodeoxyglucose-PET, PiB-PET). Th e fourth aim is to collect biological fl uids, including blood and CSF, for DNA analysis and assays of putative biomarkers of AD, including Aβ42 and tau -this will also provide a resource for exploratory studies of novel biochemical markers. Finally, the DIAN aims to perform uniform histo pathological examination of cerebral tissue in individuals who come to autopsy.
Th e National Institute on Aging awarded a 6-year grant for the DIAN that funds 10 international performance sites that combine resources and research participants of the individual sites in a uniform and comprehensive manner. Currently, over 400 individuals who are members of families with a known causative mutation for AD (that is, APP, PSEN1, PSEN2) have been identifi ed between the sites and are eligible for participation in the DIAN. Over the fi rst 6 years, sites will recruit, enroll and evaluate these individuals to reach a sample size of 400 participants. Th e DIAN cohort is predicted to comprise 80% asymptomatic individuals (with 50% of these being mutation carriers and 50% noncarriers) and 20% symptomatic individuals. Based on the participant population demographics, the DIAN is expected to enroll 50% of individuals within 3 years of parental age at disease onset, and 30% of individuals within 3 to 10 years before parental age at disease onset. Th e DIAN participants will thus consist of approxi mately 160 asymptomatic mutation carriers, 80 sympto matic AD mutation carriers, and 160 mutation-negative sibling controls.
Data obtained through the DIAN will be used in the design and statistical powering of prevention and treatment studies in ADAD. Additionally, white blood cells are being stored at the National Cell Repository for Alzheimer's Disease to establish immortalized lymphoblastoid cell lines for use in a variety of investigations, including in vitro studies to characterize the pharmacodynamic properties of putative anti-AD agents and their applicability in both ADAD and SAD. Th e DIAN will also provide the infrastructure for the recruitment and retention of subjects, which is critical for the successful performance of clinical trials in this rare, widely dispersed, and informative population.

Design of the DIAN clinical trials
An additional scientifi c aim for the DIAN is to evaluate potential disease-modifying compounds for the treatment of AD. To this end, the DIAN formed a Clinical Trials Committee to direct the design and management of interventional therapeutic trials of DIAN participants. Th e committee will assist in the design and implementation of trials that have the highest likelihood of success while providing advancement of treatments, scientifi c understanding and clinical eff ects of proposed therapies. Specifi cally, the committee's aims are to evaluate trial designs to determine the impact of interventions on biomarker, cognitive, and clinical measures in ADAD, to determine which therapeutic targets are most amenable to treatment at diff erent stages of AD, and to test the hypotheses for the causes of AD (for example, amyloid hypothesis) through therapeutic treatment trials.
Testing interventions for the prevention of AD in presymptomatic persons with inherited ADAD mutations off ers potential for medical and scientifi c advances, but also presents a number of challenges -ethically, scientifi cally, and logistically. ADAD participants tend to be highly motivated for research, perhaps due in large part to altruism. Th at is, they frequently express the hope that even if their participation does not benefi t themselves, perhaps it will benefi t their family members, including their progeny. One key design challenge is the fact that most individuals at risk of carrying an ADAD mutation have not chosen to have genetic testing. In a clinical series of 251 persons at risk for ADAD or frontotemporal lobar degeneration due to mutations in the MAPT gene, only 8.4% requested such testing [86].
Th e DIAN investigators aim to explore disease-modifying treatments in ADAD mutation carriers. Th e ultimate goal is to postpone or prevent the onset of AD symptoms, or to slow the progression of symptoms. Th e limited number of potential participants, however, limits the feasibility of trials with traditional cognitive or clinical outcomes. Th e DIAN will pursue a strategy of assessing the impact of putative disease-modifying treatments on biomarkers of AD in combination with sensitive measures of cognition. Candidate biomarkers include molecular imaging (amyloid PET scanning), functional imaging (fl uorodeoxyglucose-PET) and structural imaging (volumetric MRI measures), as well as biochemical measures in CSF (for example, tau, phospho-tau and Aβ42). Although no biomarker has been validated as a surrogate outcome for regulatory purposes, these biomarkers represent plausible candidate surrogate outcomes being pursued by AD trialists. Th e rationale for accepting surrogate markers with cognitive improvements as viable endpoints is compelling in this genetically determined population.
As the number of preventative studies that might be performed in persons carrying familial AD mutations will be limited, the optimum choice of intervention is critical. Medica tions that prevent neurodegeneration by targeting the causative mechanisms are ideal as they might both prevent the development of pathology and slow progres sion after onset. Active or passive immunotherapy or γ-secretase or β-secretase inhibitors may fulfi ll these criteria. Potential hazards include complications related to established amyloid angiopathy (for example, vaso genic edema), which may be increased in some ADAD mutations, teratogenicity, and other unknown risks of chronic exposure.

Statistical design and analyses
As only a minority of presymptomatic persons at risk for ADAD mutations asks to know their genetic status, enrollment of mutation carriers into prevention studies presents a challenge. Th e availability of treatment trials will undoubtedly infl uence the decision to obtain genetic testing. If genetic testing is required for a treatment trial, participants will need to consider testing for mutation status in order to participate in a study in which they may receive a medication (or placebo) that may help prevent the illness but could also have signifi cant side eff ects. An alternative approach would be to open enrollment to all persons at risk, to not report genetic testing, and to only randomize active drug to mutation carriers with noncarriers receiving blinded placebo. In such a study, the occurrence of side eff ects might unblind participants to their treatment group and therefore to their mutation status. Informed consent for such a trial would need the equivalent of presymptomatic genetic counseling for this possibility.
Th e gold standard for demonstrating effi cacy of an intervention is the prospective randomized, blinded, placebo-controlled study. Additionally, studies might be designed that feature open-label extensions after a prespecifi ed time period and/or a clinical endpoint is reached (such as diagnosis of dementia).
Well-established AD biomarkers, including CSF, PiB, and MRI markers, can be used as endpoints in clinical trials on DIAN presymptomatic mutation carriers. Th e objective of such trials is to determine the effi cacy of novel treatments in altering the rate of change among these biomarkers. In addition, cognitive and global function measures (for example, Clinical Dementia Rating sum of boxes) can be used in clinical trials on DIAN symptomatic participants. Given the potential heterogeneity of the population, baseline co-variants may be critical to maximize effi ciency. In a prevention trial of presymptomatic ADAD participants, sensitive cognitive measures may be used in combination with biomarker changes. Alternatively, the time to the onset of mild cognitive impairment or AD can be reasonably used as an effi cacy endpoint, especially if participants are chosen with appropriate estimates of their age of onset so that enough participants will develop AD during the designed length of follow-up to satisfy the statistical power requirement. Th e high-risk period immediately before clinical and cognitive decline can be determined by the use of biomarkers together with family history and age.
Th e ongoing DIAN longitudinal study provides important baseline and rate of change data for clinical, cognitive, imaging and other biomarkers. Th ese data will increase the ability to power and design clinical trials, and will also provide a pretreatment rate of change for analysis of treatment eff ects. In general, an increase of either the study duration or the frequency and precision of repeated measures will decrease the within-subject variability and will improve the precision of parameter estimates or statistical power over time [87]. In prevention trials in presymptomatic DIAN participants, the duration of the trial as well as the age window of participants relative to their parents' age of disease onset is crucial to allow for adequate biomarker and cognitive change to be detected.
Plans for initial DIAN therapeutic trials include identifying optimal anti-amyloid candidate interventions in development. If indicated, the suitability of specifi c candidate agents may be fi rst assessed with shortduration cerebrospinal fl uid biomarker studies to confi rm target engagement. Th e study population may include all participants at risk, or a subset with more imminent risk as suggested by biomarkers or expected age of onset; both symptomatic and presymptomatic individuals may be included. Study designs that may be implemented include randomized controlled trials with parallel group designs, lasting approximately 2 years. After completion of the placebo-controlled period, all participants can be off ered open-label treatment with continued regular assess ments. Th e primary outcome measure may be a change in amyloid PET signal; this measure provides adequate power to demonstrate a treatment eff ect with group sizes of only 20 to 30 participants [82], and allows a clinically heterogeneous study population. Secondary outcomes would include other imaging and biochemical biomarkers, as well as cognitive and clinical assessments.

Conclusion
A historical precedent highlights what is possible in the approach to prevent end organ damage by early intervention. Although there are challenges in designing and implementing presymptomatic treatment trials for an early-onset genetic disease, we are encouraged by similar successful trials in vascular disease. Th e fi rst clinical improvement in statin treatment for hypercholesterolemia was demonstrated in familial hypercholes terolemia, a genetic, early-onset aggressive form of the more common later-onset hypercholesterol emia that ultimately leads to myocardial infarction and stroke [88]. After 4 to 8 weeks of treatment with mevastatin, patients with familial hypercholesterolemia demonstrated resolving vascular bruits and disappearance of tendonous xanthomas [89]. Further, treatment with mevastatin decreased cholesterol levels in familial hypercholesterolemia patients as well as in nonfamilial hyperlipidemic patients. Taken together, these observations provided the fi rst biological evidence of a direct eff ect of a statin on cholesterol metabolism and clinical fi ndings. Th ese early biomarker studies heralded the future success of a class of anti-cholesterol drugs called statins in reducing heart attacks and strokes for millions of patients worldwide. So too may studies of anti-amyloid treatments in ADAD also lead to breakthroughs that allow for highly eff ective therapies against SAD.
Th erapeutic trials in ADAD are highly likely to produce critical scientifi c information, test fundamental theories, bridge basic science with clinical trials, accelerate therapeutic development for SAD and, perhaps most importantly, off er a chance for ADAD mutation carriers to improve their lives and their children's lives.
Neurosciences, University of California San Diego, Gilman Drive, La Jolla,