Cerebellar ataxias form a large group of movement and balance disorders that result from degeneration of Purkinje cells. Nearly two dozen individual genes have been identified to date that, if mutated, can cause various cerebellar ataxias in humans or mice. But even with such a wealth of genetic clues, the causes of cell death are hard to pin down because of the difficulty identifying the normal functions of disease-related proteins and the pathways they are involved in. The same problem plagues research in all types of neurodegenerative diseases.

That’s where the interactome can help, according to a new study in today’s issue of Cell. Starting with over 50 proteins involved in 23 different inherited ataxias, researchers in the labs of Huda Zoghbi at Baylor College of Medicine in Houston, Texas, and Marc Vidal at the Dana-Farber Cancer Institute in Boston, Massachusetts, used high-throughput yeast two-hybrid screens to construct a disease-related protein-protein interaction network. Their work shows that the apparently unrelated ataxia gene products organize into a highly connected interaction network that can provide clues to the underlying pathology of this group of related diseases.

By zeroing in on this little corner of the global interactome terrain, the researchers showed they could identify novel candidate disease genes as well as genetic modifiers and potential pathological pathways. Their technique is certain to be useful for other neurodegenerative diseases like Parkinson disease, where multiple genes have been implicated in causing or modulating a common clinical phenotype of neurodegeneration. At present, there do not seem to be enough candidate genes or modifiers for Alzheimer disease to generate a network, but that could change.

To link the many known ataxia genes, Baylor postdoctoral fellow and first author Janghoo Lim took advantage of the Vidal lab’s high-throughput yeast two-hybrid assays to identify the interaction partners for the protein products of 23 ataxia-causing genes plus 31 other ataxia-associated proteins. When they tested the proteins against either a whole-genome expression library or an adult brain library, they came up with 770 interactions, of which 96 percent were novel.

Of course, the interactome is only as good as the underlying interactions, and the yeast two-hybrid screen is notorious for generating false positive and negatives. To try to minimize false positives, the researchers used a two-hybrid screen that did not rely on overexpression of proteins. To gauge their success, a representative sample of interactions was tested in biochemical pull-down experiments, with an 83 percent confirmation rate. The interactions were also confirmed by a high rate of concordance in their subcellular distribution annotation in the Gene Ontogeny database.

Analysis of the interactions showed that 13 of the 23 disease-causing proteins organized into a single dominant network, revealing that the proteins were associated directly or indirectly through common intermediates. The 10 proteins with no connections to the main network had fewer interactions than the others. It turns out these 10 were screened using one full-length expression clone each, in contrast to the others which were expressed as multiple overlapping partial clones. Isolated domains are known to pick up more interactions in the yeast two-hybrid screen than full-length proteins. Consistent with this, the 10 outliers did have fewer partners, raising the possibility that a more extensive analysis would reveal more links that would tie them into the larger network. Alternatively, they could represent proteins involved in unrelated pathogenic pathways.

To their own interaction data, the researchers added additional interactions culled from other studies of the same proteins, or their relatives, to produce an extended map network that included all 23 ataxia proteins. Analysis of the extended network showed it was tighter and more interconnected than a model network formed from 30 random disease genes. Many ataxia-causing proteins were linked by shared partners. Some of these shared partners were familiar as disease modifiers previously identified in genetic screens in Drosophila and mouse models of ataxia. In some cases, these genetic modifiers showed up as hubs that linked multiple ataxia genes. Such hubs make interesting jumping-off points for studies aimed at understanding commonalities among different types of ataxia. They also represent potential therapeutic targets.

In their summary, the authors propose that “phenotype-based protein-protein interaction studies can be applied to many human diseases, particularly common disorders that are sporadic in the majority of cases but do result from single gene defects in a small subset of patients,” including Parkinson disease, diabetes, and hypertension.—Pat McCaffrey

Comments

  1. The technical and conceptual tour de force by J. Lim et al. illustrates some of the potential of combining high-throughput genomic and proteomic technologies with rational study design to expand the ever-growing list of disease-related candidate genes for potential pharmacotherapeutic development and intervention. This highly skilled collaborative group of authors designed a state-of-the-art complex biochemical, functional genomics, and bioinformatics-based study to see whether inherited human cerebellar ataxias and ataxic mouse mutants that cause Purkinje cell degeneration share common molecular pathways and importantly, characterize protein-protein interactions that have not been evaluated previously. One of the important outputs of the research is the identification of a coordinated network of proteins involved within human inherited ataxias and ataxic mouse models. This “interactome” is provocative because the groundwork is now laid for protein-protein interactions to be evaluated on a relatively high-throughput level for a wide variety of neurodegenerative disorders, including Alzheimer disease (AD), frontotemporal dementia (FTD), and Parkinson disease (PD), among others.

    Pat McCaffrey has provided an extensive review of the complex series of results of this article, so I have considered some of the implications of this work for the study of mechanisms underlying neurodegenerative disorders. Essentially, a major problem that collectively plagues the field is the lack of understanding of the actual function of proteins encoded by genes known to be mutated in specific neurodegenerative disorders (e.g., ataxins 1-3 in spinocerebellar ataxias 1-3, amyloid-β precursor protein [APP] in AD, and parkin in PD, among many others). Knowledge of the function of these proteins, along with identification of bona fide binding partners, may lead to several exciting discoveries, including, but not limited to, the elucidation of disease mechanisms as well as novel drug design. A caveat of this approach is that the more common sporadic forms of neurodegenerative disorders such as AD and PD may differ from the familial forms that display mutations in specific disease-related genes.

    The experimental design and proof-of-concept that interactomes can be generated from groups of genes implicated in a neurodegenerative disease entity through stringent yeast two-hybrid screens and subsequent bioinformatics and validation strategies described in this report is very exciting, and the authors have now provided a framework whereby other neurodegenerative disorders can be evaluated. Clearly, this work is in its infancy, and many experimental and interpretive questions remain, notably in the yeast two-hybrid screening procedure as well as the choice of the gene products to be included (or excluded) from the interactome investigation. In addition, the majority of progressive onset human neurodegenerative disorders display selective vulnerability of specific cell types, and this may have to be factored into these types of screening studies for greater validity to the human condition. Furthermore, the fact that the present results indicate a similar interactome network for proteins found within inherited human ataxias and ataxic mouse models speaks to the relevance of these mouse models to the disease they are attempting to shed light upon.

  2. This remarkable study moves inherited neurodegenerative disorders into the era of protein networks by demonstrating a surprisingly organized arrangement of protein interactions involving the gene products of 23 inherited ataxias. The compact nature of the interactome thus generated, together with the diversity of new and intriguing interactions discovered, provides a wealth of new molecular targets for mechanistic and therapeutic studies.

  3. It takes a lot of courage and conviction to undertake large-scale yeast two-hybrid (Y2H) studies, because they result in large harvests of data, with a lot of chaff mixed in with the wheat. Lim et al. have performed a large-scale Y2H study to specifically identify the interaction partners of proteins directly and indirectly implicated in neurodegeneration of cerebellar Purkinje cells, using a number of approaches to minimize the high false negative and false positive rates typically associated with Y2H-based investigations. These approaches included parallel Y2H screens using both cDNA and “ORFeome” libraries and complementary bait/prey and prey/bait screens. In addition, a subset of the predicted interactions was supported by an independent cotransfection/coimmunoprecipitation assay.

    The most important result of this study is that a subset of the 23 proteins directly implicated in Purkinje cell degeneration can be grouped into a tight interaction network, implying that mutations in these (seemingly unrelated) proteins induce neurodegeneration by impinging on the same biological process. The validity of the predicted interaction network was supported by two interesting observations: 1) some of the interacting proteins had been previously identified as genetic modifiers of degeneration in related animal models (a completely independent method using biologically relevant assays), and 2) one of the identified interacting proteins, puratrophin-1, was subsequently implicated in a novel autosomal dominant cerebellar ataxia.

    Although important things have been learned by this extensive study, it is unclear if the general approach will be applicable to other neurodegenerative diseases, such as Alzheimer’s. In part, this is due to the relatively few proteins with a strongly supported direct role in AD, thus limiting the extent to which a network could reasonably be developed. It is also possible that the set of disease-associated proteins used in this study might be particularly well suited for this approach. Specifically, I note that in the well-supported interaction subnetwork centered on ATXN1 (Figure 6 in this paper), 4/5 target disease proteins contain glutamine repeats (and hence are susceptible to CAG repeat expansion mutations). It is unknown why these disease-associated proteins contain glutamine repeats, but it is reasonable to assume that these (unexpanded) repeat regions have a biological function, perhaps as an interaction domain allowing modulation of the function of their host protein by other proteins. If this is true, these types of proteins may more readily fall into a sensible interaction network than other sets of disease-associated proteins.

    View all comments by Chris Link

Make a Comment

To make a comment you must login or register.

References

No Available References

Further Reading

Primary Papers

  1. . A protein-protein interaction network for human inherited ataxias and disorders of Purkinje cell degeneration. Cell. 2006 May 19;125(4):801-14. PubMed.