This is part 2 of a 4-part series from the 27th International Symposium on ALS/MND. See also part 1 here.
IBM Watson tackles ALS, a new technique reveals the site of TDP-43 binding, and a key protein in myotonic dystrophy pops up in ALS—these were some of the highlights from a session on RNA processing and dysregulation, part of the 27th International Symposium on ALS/MND, held December 7-9, 2016, in Dublin, Ireland.
IBM Watson Discovers New ALS-linked RNA Binding Proteins
RNA binding proteins (RBPs) appear to be central to ALS pathogenesis: TDP-43 and FUS bind RNA, and more than two dozen other RBPs have been linked to the disease. Since there are more than 1,450 RBPs in the human genome, it seems likely there may be still more with as-yet undiscovered links to ALS, according to Robert Bowser of the Barrow Neurologic Institute in Phoenix, Arizona.
To hunt for them, Bowser, first author Nadine Bakkar, and colleagues enlisted the services of IBM Watson, the supercomputer that won fame by beating the all-time human “Jeopardy” champion. Watson’s skill is in making inferences from patterns of words in text, enabling it to identify meaningful connections between entities of interest, such as genes, proteins, and diseases. Its talents have already been brought to bear in oncology, where it can assist doctors in making individualized treatment plans. “Our study was the first to apply it to a neurologic disease,” Bowser told the meeting.
Under the guidance of IBM and Barrow scientists, Watson analyzed published abstracts from the ALS literature related to RBPs. “Watson turns every abstract into a semantic representation, or ‘fingerprint,’” Bowser said. That representation—its branches and nodes, along with its verbal content—can be compared to that of any other abstract. A high structural similarity to abstracts discussing known ALS-related RBPs would signal a possible new candidate.
Watson read all RBP-related literature up through 2012, and was informed about the 8 RBPs known to be involved in ALS up to that point. The researchers then tested its ability to identify the three ALS-related RBPs discovered after 2012: Matrin-3, GLE1, and ARHGEF28. Watson identified Matrin-3 as the top candidate, with the other two appearing in the top 10% of its ranking of all 1445 RBPs.
Next, after training on the entire known ALS-related set, Watson re-analyzed the literature through 2015 to predict other new RBPs in ALS. Within the top 50 were five recently associated with ALS (RBM45, MTHFSD, SMN2, EWSR1 and hnRNPA3). Included in the top 10 were five with no previous association to ALS: hnRNPU, RBMS3, NUPL2, RBM6, and SYNCRIP.
Bowser presented in detail the team’s findings on SYNCRIP (synaptotagmin binding cytoplasmic RNA interacting protein), which is known to play a role in several aspects of messenger RNA maturation. In post-mortem tissue, the team found that the level and distribution of SYNCRIP was altered in ALS patients compared to controls, with increased nuclear immunoreactivity in Purkinje cells in sporadic ALS patients and increased cytoplasmic immunostaining in Purkinje cells in C9ORF72 ALS patients. The group also observed alterations in several others of the six new candidates, he said. No changes were seen in tests of several RBPs from the bottom of the list.
“We are now interested in looking for genetic alterations in these genes in ALS populations,” Bowser said. Watson may also be useful for analyzing other aspects of the ALS literature, he added, to look for convergence in metabolic pathways, leading to new therapeutic targets.
Not every relevant RBP can be discovered through this method, noted Jeffrey Rothstein of Johns Hopkins University in Baltimore, Maryland. His work has shown that the nuclear transport protein RanGAP is also an RBP involved in the disease, but because that RNA-binding function had not been previously identified in the literature, it would not have been analyzed by Watson and thus not linked to ALS.
iCLIP for Study of RNA-Protein Interactions
Over the past decade, Jernej Ule at the Francis Crick Institute in London, United Kingdom, has developed a powerful tool for studying RNA-protein interactions, called iCLIP (individual-nucleotide resolution UV Cross-Linking and ImmunoPrecipitation). iCLIP uses ultraviolet light to cross-link proteins to the RNA they attach to, allowing precise identification of even highly transient binding sites.
Recently Ule’s lab has focused their iCLIP studies on TDP-43. At the meeting, he described their findings that the protein’s low-complexity domain “serves as a docking platform for protein-protein interactions,” which then modulate the ability of TDP-43 to bind target RNAs. Disease-causing mutations in TDP-43 and multiple other ALS-linked RBPs cluster in the low-complexity regions, pointing toward functional changes in docking as potentially instrumental in disease.
In addition, Edgardo Rodriguez-Lebron of the University of Florida, Gainesville, described how he used iCLIP and other techniques to generate the first transcriptome-wide map of binding sites in mouse brain of Matrin-3, an RBP recently linked to ALS. From healthy cerebellum autopsy tissues, they established that Matrin-3 bound to 1,400 transcripts, mainly in introns but also in 5’ and 3’ untranslated regions. One target was the 3’ UTR of TDP-43 mRNA. Binding of the Matrin-3 to that RNA reduced TDP-43 expression, a result the researchers observed not only in vitro but in the CNS of mice injected with adeno-associated virus delivering the human Matrin-3 gene.
Loss of Muscleblind Suppresses FUS Neurodegeneration
Udai Pandey and colleagues from the University of Pittsburgh,Pennsylvania, conducted an unbiased genetic screen for modifiers of FUS mutation toxicity in Drosophila. Among the genes with the strongest effect was muscleblind (mbl). “We were not looking for muscleblind, but it emerged as a strong candidate in the screen,” Ian Casci, a Graduate student who did the screen, said. Muscleblind is the fly homolog of human muscleblind-like (MBNL), whose depletion in multiple tissues is believed to contribute to myotonic dystrophy (DM).
But in the context of FUS mutation, Pandey said, loss of mbl appears to be beneficial. RNAi knockdown of endogenous mbl restored nuclear localization of FUS, rescued FUS-mediated fly neurodegeneration, increased lifespan, and improved neuromuscular junction defects. Expression of several genes altered by mutant FUS returned to normal with knockdown of mbl.
How the depletion of mbl in FUS-mutant flies prevents neurodegeneration, while causing DM in otherwise normal mice, is unclear, Pandey said, but he noted that mbl-deficient flies do not appear to exhibit features of DM.