Required Reading—InteracTomes for AD, Aging, APP

Fundamental knowledge gaps in Alzheimer disease research include a clear grasp of the normal role of the amyloid precursor protein (APP), and the molecular basis for the increased risk for AD as people age. Two recent papers took a systems approach to deciphering both. In the February 6 Journal of Neuroscience, Daniel Geschwind and colleagues at the University of California, Los Angeles, describe a weighted analysis to uncover significant differences in gene expression between Alzheimer disease and control samples, and also between normal samples at different ages. Their methodology uncovers genes and gene networks that may be central to normal aging and to AD pathophysiology. Researchers led by Gerold Schmitt-Ulms at the University of Toronto, Canada, took a slightly different approach to identify gene products that may be important in AD biology. They used in vivo cross-linking to identify APP binding partners. Their paper appears in this month’s Molecular and Cellular Proteomics and identifies more than 30 new potential partners for APP.

Geschwind and colleagues used a technique called weighted gene coexpression network analysis (WGCNA) to re-examine microarray analysis of transcript levels in the CA1 region of the hippocampus in AD patients (see Blalock et al., 2004 and ARF related news story) and also in the frontal lobes of normal people between ages 26 and 106 (see Lu et al., 2004 and ARF related news story) . The WGCNA approach arranges genes into transcriptional modules, where genes in each module have expression patterns that are more similar to each other than to the patterns of genes in other modules (see Zhang and Horvath, 2005). The approach, therefore, links together genes that may be functionally related.

First author Jeremy Miller and colleagues first applied the approach to the AD sample set, which consisted of eight controls, six incipient, eight moderate, and six severe AD cases—all age-matched. The analysis identified 12 modules, which were characterized for function according to gene database annotations. Ten of the 12 modules represented functional categories that are associated with AD pathology, including synaptic transmission, mitochondrial function, extracellular transport, and neuronal transmission. Within each module the researchers also identified several hub genes, defined as those with 15 or more interactions with other genes in the module. These included mitochondrial membrane proteins involved in ion transport (VDAC1, VDAC3, ATP5F1), a voltage-gated calcium channel subunit (CACNB2), a glycine receptor subunit (GLRB), and some unknown genes (FLJ14346 and LOC152719).

From the frontal lobe dataset, the researchers identified three modules, two with genes that mostly increased in expression with age, and one in which gene expression predominantly decreased with age. To see if any gene changes are common between AD and aging, Miller and colleagues compared the AD and aging modules. Interestingly, they found that there was statistically significant overlap between genes in the three aging modules and genes in the three modules from the AD sample set. Two of these module pairs also had overlapping functional gene categories, and eight genes were identified as hub genes in both analyses. The most interesting hub gene was, perhaps, Cdk5, which has been implicated in both tau and Aβ pathology (see ARF related news story).

In these analyses, the researchers built the networks starting with the 5,000 most variable transcripts irrespective of their relationship to disease status. Focusing on genes known to be involved in AD yielded a slightly different picture. This local network analysis showed that presenilin 1 is a hub gene in the AD sample set, but not in the aging set. “Although this effect may reflect differences in brain regions, microarrays, or AD progression, a changed role for PSEN1 in the AD network is consistent with its established role in the disease,” write the authors. Also of interest is the fact that in both AD and aging samples, presenilin expression highly correlates with expression of genes that regulate myelination or are related to oligodendrocytes. “These results provide a new set of evidence supporting the hypothesis that demyelination and oligodendrocyte dysfunction may play a role in AD progression,” write the authors.

It remains to be seen whether the gene networks and hubs identified in this study represent cause or effect. “In either case, hubs of modules correlated with AD progression play key roles in processes disrupted in AD, and understanding the function of such genes may lead to a better understanding of AD progression,” write the authors. A prime example is the YWHAZ gene coding for 14.3.3 γ, a member of a protein family that is involved in cell signaling, cell cycle regulation, and cytoskeletal structure. This protein family has been linked to a variety of neurodegenerative disorders, including Huntington’s and prion diseases.

The APP interactome identified by Schmitt-Ulms and colleagues in Toronto was obtained in an entirely different way. First authors Yu Bai, Kelly Markham, and colleagues used in vivo time-controlled transcardiac perfusion cross-linking in mice, a method that pumps controlled amounts of formaldehyde through the circulatory system and results in limited protein cross-linking. After the perfusion, the researchers isolated APP complexes by immunoprecipitation with APP antibodies. They then digested the proteins and identified the resulting peptides by mass spectroscopy.

To distinguish proteins that bind specifically to APP and not to APLP1 and APLP2 homologs, Bai and colleagues used a variety of different antibodies. For APP they used one antibody that recognizes the intracellular C-terminal end and one that binds the extracellular domain flanked by α- and β-secretase sites. The latter does not recognize APLP1 and 2. The researchers also used two antibodies specific for APLP1 and 2. They identified three types of interactors: those that are non-specific, those involving a single APP family member, and those that bind to more than one family member. The last group was small, containing, in addition to the APP proteins themselves, only one other protein, the RasGAP-activating-like protein 1.

The researchers identified a total of 33 proteins that bound exclusively to APP. Twelve were pulled down by both antibodies. The C-terminal antibody pulled down an additional 10 proteins, most of them cytoplasmic. The antibody directed against the extracellular epitope also pulled down 10 additional proteins, mostly extracellular. The 12 proteins pulled down by both antibodies are either membrane proteins or reside in the ER. Overall, the pattern of hits suggests that the strategy identifies physiologically relevant interactions rather than non-specific cross-links. But to confirm the specificity, the researchers turned to an internal control system for quantifying interactomes. Called iTRAQ, for isobaric Tags for Relative and Absolute Quantitation, the system relies on chemical tags of the same mass to serve as tracers in four separate experiments. This allowed the researchers correct data for the relative abundance of peptides in individual samples. The data obtained using the iTRAQ analysis was in “excellent agreement with data obtained following analysis of individual IPs and as such strongly argue that mere sampling bias was not the underlying cause for differences seen above in interactome data of APP and family members,” write the authors.

Some proteins in the interactome were conspicuous by their absence. α-, β- and γ-secretases did not show up, for example. But the authors note that “short-lived interactions of catalytic nature (e.g., proteolytic enzymes) commonly escape detection following chemical cross-linking.” The interactome did include F-spondin, a potential ligand for APP (see ARF related news story); the remaining proteins were potential novel APP partners. The authors chose to focus on LINGO-1, a type 1 transmembrane protein that has been linked to myelination and axon regeneration, to test the validity of the interactome. Antisense probes showed that LINGO-1 and APP co-localize in the brain, predominantly in the CA1 and CA3 region of the hippocampus, and cell culture experiments suggest a functional interaction between the two proteins. Knocking down LINGO-1 in HEK293 cells expressing APP with the Swedish mutation reduced β-secretase cleavage and Aβ production by about 30 percent and increased α-secretase cleavage by about 70 percent. In contrast, overexpression of LINGO-1 had the opposite effects, increasing Aβ production.

While this example serves to show that the interactome has some validity, the authors recognize that intensive validation will be required to appreciate the roles of the proteins in this network. “It is hoped that further investigations of this network will reveal an ‘Achilles’ heel’ in the molecular biology of APP that can be exploited for diagnostic or therapeutic purposes,” write the authors. The same may be said for the networks uncovered by Geschwind and colleagues.—Tom Fagan

Comments

  1. One of the driving questions in Alzheimer’s research has been its nature—is
    AD simply an extension of normal aging, or is it a disease unto itself?
    If it is a disease unto itself, what changes over time in the brain
    to make the aged tissue so much more vulnerable to attack? Using high-level
    bioinformatic approaches, Miller, Oldham, and Geschwind take a closer look
    at the interplay between aging and AD using transcriptional profiles from
    previously published data sets.

    On a technical scale, this work is extremely thorough and careful. It is a
    terrific example of a well-reasoned meta-analysis in its truest form, paying
    attention to statistical probabilities at each stage of the analyses (e.g.,
    probability of overlap between the two studies based on the discovery power
    in either). Rather than simply comparing lists of genes from two studies,
    the authors stripped the information from each study down to its most
    raw form (at least as raw as they could get it, i.e., CEL files from Affymetrix
    array scans), and used a consistent probe level algorithm across both
    studies to create new data sets that are more comparable.

    Using the innovative WGCNA approach to clustering, the authors established
    modules of genes based on similar behavior across disease states or aging.
    Within these modules, the authors established connectivity among genes and
    identified “hub” genes that appeared to be most often linked to other genes
    within the module. Among these, two genes stood out: VDAC1 in the
    mitochondrial module of AD, and YWHAZ, a fairly uncharacterized, but
    extremely abundant gene product in brain in a module that stubbornly refused
    to reveal a functional association. Further, well-known PSEN1 was found to
    be related to myelinating processes in a “guilt-by-association” module of
    myelin-related gene products. The authors found that two of the modules from the AD data set (synaptic and mitochondrial) were related to a single module
    from the aging study that contained both functional categories.

    The implication here is that aging and AD do share some common processes.
    However, the question remains as to whether AD represents a frankly
    different process than aging. That’s because there are both intriguing agreements (discussed in the paper), and disagreements (for instance, that the
    mitochondrial and synaptic genes split into two distinct modules in AD, but
    resided in one aging module) between the two studies compared here.

    It is remarkable that these relationships were seen even despite the confounds
    on a technical level of different labs, and times, as well as the biological
    confounds of different tissue type (hippocampal CA1 versus forebrain cortex). It
    would be interesting to know, in future studies using this approach, what the
    level of agreement would be between two studies attempting to examine the
    same disease state—such as AD—in different brain regions. Our own
    observations are that changes are more consistent at the functional group
    level as opposed to the per gene level.

    Other interesting further work might include determining the consequences of
    various cutoff decisions in the course of implementing the algorithm, e.g.,
    number of presence calls, variability for inclusion, lack of connectivity
    for exclusion, should number of connections for determining hub genes be
    adjusted for number of genes in module, etc. Also interesting would be to determine “anti-modules,” functional groups that remain static across groups.

    View all comments by Eric Blalock

  2. This interesting and technically innovative paper addresses the still enigmatic physiological function of APP. The study sheds new light on APP by comparing its “interactome” (i.e., the sum of its interactions with other proteins) with those of the APP paralogs, the APP-like proteins 1 and 2.

    To achieve the goal of catching as many as possible interacting proteins in the act, a new way of cross-linking is employed. This method, termed "time-controlled transcardiac perfusion cross-linking," basically glues all protein complexes together while the animal still lives. This is advantageous as it reduces the risk of mapping artifactual interactions, which often occur after tissue is extracted, and it should better enable the identification of transient and low-affinity interactions. After immuno-affinity purification of the cross-linked protein complexes, the authors employ isobaric tagging and quantitative mass spec to map differences in the complexes formed by the three paralogous proteins.

    This strategy poses a twofold advantage. First, non-specific interactions with other proteins—a notorious problem for membrane proteins—are identified as they will likely be the same for a group of proteins with similar structural make-up. Second, any interactors identified that are specific for APP may yield clues to why APP but not its cousins can go awry in Alzheimer disease.

    Indeed, the results obtained here are interesting. Firstly, it is always great to find interactions previously observed with other approaches, and a number of known APP binding proteins are identified. This includes cystatin C which has been proposed as a genetic marker for AD (see Alzgene). Secondly, a number of novel interacting proteins are identified. An intriguing case is LINGO-1, a neuronal transmembrane protein. The authors go on to demonstrate that siRNA-mediated reduction of LINGO-1 in APPsw-cells caused a reduction in Aβ production, while LINGO-1 overexpression caused increased β-secretase cleavage of APP. In summary, the combination of techniques used here carries a lot of promise for the elucidation of membrane-protein complexes, which should contribute to a better understanding of their role in disease processes and as drug targets.

    View all comments by Gerard Drewes

  3. Bai et al. use a novel approach for identifying APP family member-candidate interacting proteins. For identification of interacting proteins, the investigators used transcardiac perfusion cross-linking, followed by immunoprecipitation of the cross-linked complexes from mouse brain with APP family member-specific antibodies and LC/MS/MS analysis of recovered tryptic peptides. Validation of the putative interaction of APP with one of the identified proteins, LINGO-1, a transmembrane protein that binds p75, could only be demonstrated in immune complexes recovered from cross-linked mouse brain proteins, suggesting that this interaction is weak, transient, or mediated by an intermediate protein. However, manipulations of LINGO-1 protein levels in HEK293 APPSwe cells using siRNA or overexpression produced opposite effects on levels of APP proteolytic fragments, suggesting that this interaction is physiologically relevant for APP processing.

    The data in this study confirm the previously identified interaction of APP family members with each other, as well as the interaction of APP with the membrane-associated proteins: calsyntenins (alcadeins) and PrP; the ER proteins BIP and calnexin, and the secreted proteins F-spondin and Cystatin C. One advantage of this approach is the identification of proteins that are located in the membrane or in the vicinity of membranes, including the plasma membrane. However, a striking absence of previously identified APP C-terminal binding proteins was observed using this experimental approach. In the case of the FE65 proteins, this cannot be attributed to lack of importance of this protein family for APP biology, since the FE65/FE65L1 double knockout mice display a neuronal positioning defect in the developing cortex that is remarkably similar to the APP triple knockout mice (Guénette et al., 2006; Herms et al., 2004).

    Given that APLP2 and APLP1 were found in APP immunoprecipitates, it is unfortunate that no proteins common to all three APP family member immune complexes were identified. Such proteins would be good candidates for studies aimed at elucidating the molecular mechanisms underlying the phenotypes observed in APP/APLP1/APLP2 triple knockout mice. Nevertheless, it remains possible that disruptions of APP family member interactions with proteins identified by Bai et al. contribute to the type II lissencephaly-like phenotype observed in the APP family member knockout mice. In this respect, SPARC-L1, an extracellular protein known to localize to the surface of radial glial fibers in the outer layers of the developing cortex, is particularly interesting. That is because it was recently shown to block adhesion of migrating neurons to radial glia, thereby terminating radial glia-mediated neuronal migration (Gongidi et al., 2004). Continued adhesion of neurons to radial glia in mice deficient for the APP family members may lead to neuronal overmigration and heterotopia formation, similar to what is observed in the developing cortex of APP/APLP1/APLP2 triple knockout mice (Herms et al., 2004).

    Finally, the identification of calsyntenins (alcadeins) as APP cross-linked proteins in the absence of X11L (X11β or Mint-2) is interesting, given that alcadein α1 (Alcα1) was identified as an interactor for X11L in a two-hybrid screen (Araki et al., 2003). Alcadein was initially proposed to form an APP-X11L-Alcα1 tripartite complex because all three proteins could be recovered from brain membrane fractions in immune complexes isolated with an APP C-terminal antibody.

    However, the binding domain within X11L for alcadein was identified as the same domain that mediates binding to APP, the PTB domain of X11L; thus X11L cannot act as an adapter protein in the formation of this complex. The data provided by Bai et al. suggest that the tripartite complex is formed by the interaction of the N-terminal domains of APP and alcadein with binding of either alcadein or the APP C-terminus to X11L. Furthermore, this may allow for complex formation of secreted APP or alcadein with X11L-bound, membrane-associated alcadein or APP, respectively (Araki et al., 2004).

    Unique APP and APLP-interacting proteins have been identified using different experimental strategies. Therefore, it seems reasonable to consider all candidate interactors when thinking of the APP/APLP interactome, and how it may shed light on our understanding of the physiological functions of APP and/or the APLPs through their protein-protein interactions.

    References:

    Guénette S, Chang Y, Hiesberger T, Richardson JA, Eckman CB, Eckman EA, Hammer RE, Herz J. Essential roles for the FE65 amyloid precursor protein-interacting proteins in brain development. EMBO J. 2006 Jan 25;25(2):420-31. PubMed.

    Herms J, Anliker B, Heber S, Ring S, Fuhrmann M, Kretzschmar H, Sisodia S, Müller U. Cortical dysplasia resembling human type 2 lissencephaly in mice lacking all three APP family members. EMBO J. 2004 Oct 13;23(20):4106-15. PubMed.

    Gongidi V, Ring C, Moody M, Brekken R, Sage EH, Rakic P, Anton ES. SPARC-like 1 regulates the terminal phase of radial glia-guided migration in the cerebral cortex. Neuron. 2004 Jan 8;41(1):57-69. PubMed.

    Araki Y, Tomita S, Yamaguchi H, Miyagi N, Sumioka A, Kirino Y, Suzuki T. Novel cadherin-related membrane proteins, Alcadeins, enhance the X11-like protein-mediated stabilization of amyloid beta-protein precursor metabolism. J Biol Chem. 2003 Dec 5;278(49):49448-58. PubMed.

    Araki Y, Miyagi N, Kato N, Yoshida T, Wada S, Nishimura M, Komano H, Yamamoto T, De Strooper B, Yamamoto K, Suzuki T. Coordinated metabolism of Alcadein and amyloid beta-protein precursor regulates FE65-dependent gene transactivation. J Biol Chem. 2004 Jun 4;279(23):24343-54. PubMed.

    View all comments by Suzanne Guenette

Make a Comment

To make a comment you must login or register.

References

News Citations

  1. New Microarray Data Offer Grist for AD Hypothesizing Mills
  2. After 40, DNA Damage Accrues in Genes, Hampering Expression
  3. Tangles, Neurodegeneration, Plaques—p25 Does it All
  4. AβPP Ligand Found? Neurons Respond to F-Spondin

Paper Citations

  1. Blalock EM, Geddes JW, Chen KC, Porter NM, Markesbery WR, Landfield PW. Incipient Alzheimer's disease: microarray correlation analyses reveal major transcriptional and tumor suppressor responses. Proc Natl Acad Sci U S A. 2004 Feb 17;101(7):2173-8. PubMed.
  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.
  3. Zhang B, Horvath S. A general framework for weighted gene co-expression network analysis. Stat Appl Genet Mol Biol. 2005;4:Article17. PubMed.

Further Reading

Primary Papers

  1. Miller JA, Oldham MC, Geschwind DH. A systems level analysis of transcriptional changes in Alzheimer's disease and normal aging. J Neurosci. 2008 Feb 6;28(6):1410-20. PubMed.
  2. Bai Y, Markham K, Chen F, Weerasekera R, Watts J, Horne P, Wakutani Y, Bagshaw R, Mathews PM, Fraser PE, Westaway D, St George-Hyslop P, Schmitt-Ulms G. The in vivo brain interactome of the amyloid precursor protein. Mol Cell Proteomics. 2008 Jan;7(1):15-34. PubMed.

Annotate

To make an annotation you must Login or Register.