Genomic/proteomic/metabolomics (OMICS) research in AD remains in its infancy. Most studies stall after discovering lists of hundreds of genes whose expression changes in the chosen comparison. Few groups have been able to validate their results, much less translate to clinical practice. The field needs to devise more focused studies that can advance in this way. General research priorities in the OMICS area include definition of useful algorithms for data mining, comparisons of the relative power of mRNA- versus protein-based approaches in capturing a given biological process, and attempts to characterize the role of post-translational modifications and of non-coding regions. Heterogeneity continues to pose challenges to OMICS, both between brain regions and within a given region. This can partly be addressed by laser capture microdissection.
One line of research that has moved past initial gene expression profiling is establishing a new hypothesis about the biology of brain aging as a way of approaching this old question at a new level of analysis. Classic aging studies have established that people slow down in verbal recall and other cognitive domains even with normal aging. A potential underlying mechanism for those changes is emerging from transcriptome comparisons of people across the human age range. In the prefrontal cortex and other brain areas, gene expression changes early in adult life, around age 40. It changes with a characteristic signature. Certain clusters of genes lose expression; these include synaptic plasticity and memory storage genes such as NMDA, AMPA, GABA, serotonin receptor subunits, calmodulin, calbindins, synaptic organizing molecules such as agrin, vesicle transport genes such as RAB GTPases, dynein, clathrin, kinesin, tau, and energy metabolism genes in mitochondria). Clusters of other genes are induced; this includes genes encoding stress response proteins, inflammatory mediators, antioxidant processes, metal binding, DNA repair, neuronal survival, and myelination. Analysis of this phenomenon has shown that the genes that lose expression have accrued disproportionate oxidative damage in their promoters and failed to repair it, generating the hypothesis that learning, memory, and neuronal survival genes are selectively vulnerable to DNA damage with age. Data are beginning to suggest that old people who maintain exceptional mental acuity have expression signatures similar to middle-aged people (Lu et al., 2004).
Since then, further analysis has indicated that on a given chromosome, DNA damage is not distributed randomly but is most intense in the promoter regions of certain genes, particularly so in promoters of aged samples. The promoters' vulnerability is linked to sequence motifs that bind iron. Iron binding to these promoters appears to increase with age and to lead to double-strand breaks. The working hypothesis coming out of this research holds that how the brain ages depends, in part, on its iron homeostasis, in that increasing iron binding tends to damage the promoters of a small set of genes that are critically important to synaptic plasticity and cognition. Why the affected genes fall into these groups remains unclear but may have to do with high expression in particularly active, that is, plastic, brain regions. (Expression of APP itself does not change with age, but that of some of its binding proteins do [e.g., X11, Fe65], possibly affecting its trafficking or endocytosis.)
A second example of a strategy to move beyond lists of differentially expressed genes lies in exploiting differences between brain areas toward understanding regional vulnerability in a given disease. In one study, high expression of a gene in mouse cerebellum, a region spared in frontotemporal dementia, hinted at a protective function. Follow-up work with fly genetics, in-vitro biochemistry, and human autopsy tissue pinpointed the new tau protease PSA. This aminopeptidase degrades tau protein and is expressed at much higher levels in human cerebellar granule neurons than cortical neurons (see ARF related conference story and Karsten et al., 2006) The bottleneck in this candidate-gene approach is apparent in that the original microarray experiment identified 30 genes of interest, and establishing the role for this one alone took years of a multidisciplinary and collaborative effort.
For faster identification of functionally important changes in gene expression, the field needs systems biology approaches, that is, techniques to trace higher-order features of the transcriptome. Such features include mechanisms of co-regulation of groups of genes, and expression networks that establish connectivity maps between genes. Microarray data can be analyzed to obtain a connectivity measure for a given gene that captures the sum of its connections and connection strengths. Comparison of gene connectivity between humans and chimpanzees shows that the connectivity of genes in human cortex diverges widely from that in chimpanzee cortex, whereas in other brain areas, gene connectivity in humans and primates are more similar. Measures of gene expression alone do not show this divergence, suggesting that the cortex in particular has undergone a massive expansion in gene connectivity from chimpanzee to human, and that expression data alone therefore cannot capture essential features of human cortical function.
Connectivity measures can point up relationships that expression data alone don't. For example, the new tau protease PSA shares many connections with the new FTD gene progranulin (Cruts et al., 2006; Baker et al., 2006), even though their expression is not correlated. Topologic overlap of connectivity data can identify previously known functional networks of genes. It can identify connectivity hubs that are specific to human cortex, as well as functionally relevant hubs that are linked to neurologic disease, such as ones centered on the PINK1 and UCHL1 genes (for further reading, see Khaitovich et al., 2004; Coppola and Geschwind, 2006).
A separate area of genomics focuses on identifying additional risk genes for AD. Here, the critical need lies in establishing larger sample collections than were previously used in order to power linkage and association studies or scans sufficiently high so that they can detect alleles that exert small effects or occur at low frequency. Underpowered studies have been holding back progress in AD and other diseases, notably psychiatric ones. In AD it is estimated that 3,000 patient samples will be necessary to detect variants that increase the relative risk by 1.25. Sample sizes in the hundreds have been typically used in the past; collaborative data sharing is necessary to move beyond this limitation. The Psychiatric Disease Initiative at the Broad Institute in Cambridge, Massachusetts, aims to use whole-genome scans to find risk genes for schizophrenia and bipolar disorder. In that initiative, collaborating groups have agreed to pool data to increase the number of available cases/controls to several thousand. To expand the sample base further, the initiative is tapping into the Swedish National Cohort Study of Schizophrenia, which aims to draw 7,500 schizophrenia patients and 7,500 matched controls from the country's national registers.
The study of complex disorders needs a better understanding of genetic variation. One novel resource in this regard is the International HapMap Project. Freely available as a public database, it to date has analyzed more than four million SNPs in samples from Africa, East Asia, and the U.S., offering a denser set of markers than previously available for association studies. Genotyping and analytic capabilities must also improve. New SNP microarrays contain all SNPs typed in HapMap samples. By themselves they still miss substantial fractions of all common human DNA variants, but improved genotype-calling algorithms can increase their coverage and accuracy such that they should detect common variants that increase risk for schizophrenia and bipolar disorder. Presently available arrays cannot capture the effect of rare variants. Applied to other diseases, the arrays already have identified new risk genes, for example, common variants of three complement factor genes that together explain half of the heritability of age-related macular degeneration, the leading cause of blindness in the aged (Maller et al., 2006). Until last year, almost none of thousands of prior papers on macular degeneration had mentioned complement, or vice versa. Moreover, new microarrays are more sensitive at detecting copy number variations, which can lead to mass effects at the RNA and protein levels. Mass effects in neurodegenerative diseases have been shown with rare gene duplications (Singleton et al., 2003; Rovelet-Lecrux et al., 2006), and promoter variants can exert similar, smaller effects that together may account for a significant fraction of the genetic variation in AD risk (Lahiri et al., 2005; Singleton et al., 2004). Finally, environmental factors can tip the balance between whether a person develops or escapes a given disease for which they carry a moderate increase in genetic risk. Epidemiology can identify potential new environmental risk factors, and once a disease's risk genes are known, environmental factors can be studied more precisely.
On the proteomics front, a majority of applied studies in AD research attempt to identify panels of proteins that can detect and distinguish the disease better than the clinical diagnosis, and eventually will be able to identify preclinical AD or predict AD in mid-life. For one such study, see the Wyss-Coray presentation in the Alzforum report on the Translational Biomarkers Workshop. For validation, such research needs access to plasma from larger samples and younger cohorts of people. In the U.S., groups at University of California, San Diego; University of Washington, Seattle; Washington University, St. Louis; Columbia University, New York; and University of Pennsylvania School of Medicine, Philadelphia, are currently collecting samples longitudinally. The ADNI initiative funded by NIA is gearing up to bank fluids at participating centers, and will make samples available. A funding shortage and insufficient industry investment in diagnostic tests are additional bottlenecks in this area.
Both genomics and proteomics are areas of active technology development. Following below is an example of each, and both may become useful to the field at large. One is the RNAi Consortium. Formally established in 2004, this group aims to create an openly available lentiviral RNAi library that can be applied to large-scale, unbiased loss-of-function screens in mammalian cells, much as has been possible in yeast for some years. The consortium has created reagents to knock down most human and some mouse genes. It is currently focusing on validation techniques and on exploring how best to use the reagents for extracting biological information. In this area, the group emphasizes development of high-throughput imaging assays rather than lacZ or luciferase surrogate reporter assays; this will render this library attractive for neuroscience applications. Freeware to analyze thousands of cell images is already available for download.
To date, more than 140,000 reagents have been made, organized by gene groups, such as kinases, proteases, etc. They infect most cell types, including non-dividing cells and neurons, at low multiplicity of infection, and they yield stable expression of the respective shRNA. A robust, automated protocol for how to use the library has also been worked out. Technical hurdles at this stage include how to distinguish true hits from false-positive or off-target effects, obtaining larger numbers of effective hairpins (i.e., individual short RNAi sequences) per gene, knocking down a given gene strongly enough to see a biological effect, and generally validating the library with available funds. The question remains which features of complex diseases can be mimicked in cell-based assays and made amenable to genome-scale RNAi screens. Large-scale screens for a systems approach to pathways involved in a given function of interest are not yet feasible with this library.
Proteomics technologies that could be applied to fundamental questions of neural function and AD include, for example, phosphoproteomics. One uses mass spectrometry-based algorithms to track intracellular signaling circuits by analyzing phosphorylation. This technique allows the scientist to probe a cell's signaling pathways on a global level by quantifying key aspects of phosphorylation, such as the temporal sequence of successively phosphorylated sites, simultaneously for dozens of proteins (see Zhang et al., 2005; Chen and White, 2004). Applied to the research questions in AD, this technology could bring a network perspective to the study of tau phosphorylation. Other questions of interest that could be studied with greater power in this way include APP or ApoE signaling, or a neuron's response to NGF or to a glial signal and vice versa.
Further Topics for Discussion
A topic that was not on this year's workshop agenda generated significant discussion. It is ApoE, a perennially understudied area in AD research. The failure to find a second major risk gene for AD since the discovery of ApoE4 in 1993 only reinforces ApoE's status as the leading genetic risk factor. Yet initial research efforts on ApoE have waned; few groups today investigate it (for review, see ARF ISOA conference report). The role of ApoE in the periphery is better understood than in the brain. It appears to be a stress-response protein; its expression, like that of APP, soars after stroke or injury. The crystal structure of ApoE is available. ApoE shows isoform-specific effects in the brain, but their role in AD pathogenesis is unclear. ApoE4 shows a unique domain interaction, whereas ApoE2 and E3 don't. ApoE2 binds less tightly to the LDL receptor than ApoE3 and 4, which could affect cholesterol recycling and, in turn, synaptic function. Studies examining ApoE in the context of dendritic spines and LTP induction tend to find a favorable effect of ApoE2. Separately from its synaptic effects, ApoE4 enhances amyloid pathology dramatically and is associated with greater damage/poorer recovery in injury models.
ApoE4 carriers have higher levels of corticosteroids and show differences in PET scans even at young ages and without overt cognitive impairments. ApoE2/2 homozygote carriers are rare and can have abnormally low cholesterol levels, but they rarely ever develop AD. This natural form of risk reduction presents an opportunity to understand its mechanism and exploit it for therapies. ApoE is one of the most abundantly produced and released proteins in astrocytes and should be explored for a potential signaling role. Studies modeling ApoE in animals must be aware of differences between lipid handling in mice and humans, which have made guinea pigs a favored model in the cholesterol field. In the brain, cholesterol and ApoE synthesis and turnover occur mostly locally, without much connection to the periphery. A research area is developing around the discovery that ApoE binds to receptors (i.e., LRP) that also bind APP and that use similar adaptor proteins (i.e., Fe65) to form heterodimeric complexes but, again, relevance for ApoE signaling and AD pathogenesis is not clear (see Eibsee report).
Participants debated the importance of understanding the biology of AD more fully versus focusing on a given hypothesis for therapy development. An area of common ground lies in the notion that the biology of statins and their effects on lipid lowering and reducing heart disease risk and mortality were not fully worked out before their long-term secondary prevention trials began. Despite the wide use and success of statins, the field of heart disease needs additional drugs. Likewise, anti-amyloid therapy development is timely even while gaps remain in the amyloid hypothesis. Basic research must lay the groundwork for alternative approaches in the event that secretase inhibition and immunotherapy fail. Even if they succeed, there will be ample need for alternative approaches as most researchers expect an effective AD therapy to have to act on multiple components of the disease. In this regard, emerging research on glia, immune system components, DNA repair, and synaptic maintenance open new horizons for AD.
Participants agreed that ways must be found to push the time frame of when people are treated, and experimental drugs tested, back into the preclinical phase from the mild-to-moderate phase of diagnosed AD that is the typical time of treatment today. Statins, for example, have their strongest effect as preventive agents, and researchers feel that a number of promising AD drugs may be failing trials because the patients were too advanced in their disease. Epidemiology is reaching consensus that metabolic factors exert their strongest effect on dementia risk during middle age. Secondary prevention trials are necessary but require validation of an antecedent marker and a surrogate marker that is based on the drug's action and that changes as a function of its dose (see Alzforum report on translational biomarkers). Cognitive tests are neither precise enough nor practicable for such trials.
Last but not least, participants agreed on a nagging technical problem that impedes AD research. It is the variability of Aβ preparations and detection assays used throughout the field, and a lack of precision in how authors describe Aβ preparations and measurements in publications. To reproduce and compare studied, Aβ preparations in any publication should be defined using consensus language with regard to their origin (synthetic, cell-secreted) and aggregation state and solubility. Likewise, the field should agree on a set of consensus assays for measuring different kinds of Aβ species in tissue sections vs. fluids vs. brain extracts, because using the incorrect assay can mask large fractions of Aβ present in the sample.
- SfN: Return of the Other—Tau Is Back, Part 3
- Translational Biomarkers in Alzheimer Disease Research, Part 5
- ApoE Catalyst Conference Explores Drug Development Opportunities
- An Intricate Dance: α-Secretase and Its Partners
- Translational Biomarkers in Alzheimer Disease Research, Part 2
- Lu T, Pan Y, Kao SY, Li C, Kohane I, Chan J, Yankner BA. Gene regulation and DNA damage in the ageing human brain. Nature. 2004 Jun 24;429(6994):883-91. PubMed.
- Karsten SL, Sang TK, Gehman LT, Chatterjee S, Liu J, Lawless GM, Sengupta S, Berry RW, Pomakian J, Oh HS, Schulz C, Hui KS, Wiedau-Pazos M, Vinters HV, Binder LI, Geschwind DH, Jackson GR. A genomic screen for modifiers of tauopathy identifies puromycin-sensitive aminopeptidase as an inhibitor of tau-induced neurodegeneration. Neuron. 2006 Sep 7;51(5):549-60. PubMed.
- Cruts M, Gijselinck I, van der Zee J, Engelborghs S, Wils H, Pirici D, Rademakers R, Vandenberghe R, Dermaut B, Martin JJ, van Duijn C, Peeters K, Sciot R, Santens P, De Pooter T, Mattheijssens M, Van den Broeck M, Cuijt I, Vennekens K, De Deyn PP, Kumar-Singh S, Van Broeckhoven C. Null mutations in progranulin cause ubiquitin-positive frontotemporal dementia linked to chromosome 17q21. Nature. 2006 Aug 24;442(7105):920-4. PubMed.
- Baker M, Mackenzie IR, Pickering-Brown SM, Gass J, Rademakers R, Lindholm C, Snowden J, Adamson J, Sadovnick AD, Rollinson S, Cannon A, Dwosh E, Neary D, Melquist S, Richardson A, Dickson D, Berger Z, Eriksen J, Robinson T, Zehr C, Dickey CA, Crook R, McGowan E, Mann D, Boeve B, Feldman H, Hutton M. Mutations in progranulin cause tau-negative frontotemporal dementia linked to chromosome 17. Nature. 2006 Aug 24;442(7105):916-9. PubMed.
- Khaitovich P, Muetzel B, She X, Lachmann M, Hellmann I, Dietzsch J, Steigele S, Do HH, Weiss G, Enard W, Heissig F, Arendt T, Nieselt-Struwe K, Eichler EE, Pääbo S. Regional patterns of gene expression in human and chimpanzee brains. Genome Res. 2004 Aug;14(8):1462-73. PubMed.
- Coppola G, Geschwind DH. Technology Insight: querying the genome with microarrays--progress and hope for neurological disease. Nat Clin Pract Neurol. 2006 Mar;2(3):147-58. PubMed.
- Maller J, George S, Purcell S, Fagerness J, Altshuler D, Daly MJ, Seddon JM. Common variation in three genes, including a noncoding variant in CFH, strongly influences risk of age-related macular degeneration. Nat Genet. 2006 Sep;38(9):1055-9. PubMed.
- Singleton AB, Farrer M, Johnson J, Singleton A, Hague S, Kachergus J, Hulihan M, Peuralinna T, Dutra A, Nussbaum R, Lincoln S, Crawley A, Hanson M, Maraganore D, Adler C, Cookson MR, Muenter M, Baptista M, Miller D, Blancato J, Hardy J, Gwinn-Hardy K. alpha-Synuclein locus triplication causes Parkinson's disease. Science. 2003 Oct 31;302(5646):841. PubMed.
- Rovelet-Lecrux A, Hannequin D, Raux G, Le Meur N, Laquerrière A, Vital A, Dumanchin C, Feuillette S, Brice A, Vercelletto M, Dubas F, Frebourg T, Campion D. APP locus duplication causes autosomal dominant early-onset Alzheimer disease with cerebral amyloid angiopathy. Nat Genet. 2006 Jan;38(1):24-6. PubMed.
- Lahiri DK, Ge YW, Maloney B, Wavrant-De Vrièze F, Hardy J. Characterization of two APP gene promoter polymorphisms that appear to influence risk of late-onset Alzheimer's disease. Neurobiol Aging. 2005 Nov-Dec;26(10):1329-41. PubMed.
- Singleton A, Myers A, Hardy J. The law of mass action applied to neurodegenerative disease: a hypothesis concerning the etiology and pathogenesis of complex diseases. Hum Mol Genet. 2004 Apr 1;13 Spec No 1:R123-6. PubMed.
- Zhang Y, Wolf-Yadlin A, Ross PL, Pappin DJ, Rush J, Lauffenburger DA, White FM. Time-resolved mass spectrometry of tyrosine phosphorylation sites in the epidermal growth factor receptor signaling network reveals dynamic modules. Mol Cell Proteomics. 2005 Sep;4(9):1240-50. PubMed.
- Chen WG, White FM. Proteomic analysis of cellular signaling. Expert Rev Proteomics. 2004 Oct;1(3):343-54. PubMed.
No Available Further Reading