Neurodegenerative diseases are often associated with genetic mutations that cause repetition of short sequences of nucleotides. In the disorders amyotrophic lateral sclerosis (ALS, also known as motor neuron disease) and frontotemporal dementia, such an expansion in a non-protein-coding region of the C9orf72 gene1,2, leads to aberrant translation products that contain repetitive stretches of glycine and alanine amino-acid residues. These ‘poly(GA)’ products form aggregates in neurons, and have been implicated in the disruption of a key cellular process in which complexes called proteasomes degrade proteins3,4. However, the biochemical basis for this disruption, and how it might promote disease, is poorly understood. Writing in Cell, Guo et al.5 precisely map the organizational and structural features of poly(GA) aggregates and associated macromolecular complexes in neurons using a technique called 3D cryo-electron tomography (cryo-ET), to provide direct visualization of how proteasomes are disrupted by poly(GA) proteins.
Cryo-ET in 3D uses electron microscopy to view very thin, frozen but hydrated sections of a cell from various angles. The resulting images are combined to produce a 3D image called a tomogram. Guo et al. used 3D cryo-ET to visualize neurons that had been genetically engineered to express a poly(GA) tract that contained either 175 or 73 repeats. The tracts were fused with a green fluorescent protein that enabled their precise position to be determined using correlative light microscopy. The engineered protein mimics poly(GA) tracts that are produced from C9orf72 expansion, which take a long time to form in vivo. The authors found that poly(GA) proteins form highly clustered and often bifurcated twisted ribbon structures that are of relatively uniform thickness, but of variable length and width, similar to poly(GA) structures previously observed by conventional electron microscopy in vitro6.
The value of the authors’ work lies not only in their observation of the structure of poly(GA) aggregates in detail in cells, but in their comparison of poly(GA) with aggregates formed through a different genetic expansion — a glutamine repeat (poly(Q)) tract, which causes the neurodegenerative disorder Huntington’s disease, and which the same group analysed by 3D cryo-ET last year7. This comparison revealed structural differences that could explain dissimilarities in pathogenic mechanisms between the conditions.
First, the aggregates formed in each case are themselves structurally distinct. Poly(Q )proteins form fibril structures that show little branching and are less densely packed than poly(GA) ribbons7.
Second, when the authors used powerful computational approaches to search for known macromolecular complexes in each aggregate, they found many proteasomes incorporated in poly(GA) aggregates (Fig. 1). Indeed, biochemical data suggested that as many as 50% of the proteasome complexes in the neuron become highly entangled within poly(GA) ribbons. Removal of proteasomes from their normal location in cells through this sequestration mechanism might explain the reduced proteasomal activity in cells harbouring these aggregates4,8. Complexes called ribosomes, which mediate protein production and are comparable in size to proteasomes, were largely excluded from poly(GA) ribbons, suggesting that poly(GA) aggregates are actively recruited or retained by proteasomes. By contrast, poly(Q) fibrils did not contain proteasomes, but formed close contacts with membranes from multiple types of organelle. This interaction leads to deformation of the membranes around organelles, such as the endoplasmic reticulum. Such deformation might alter pathways involved in protein translation, trafficking and degradation7.
The proteasome consists of a barrel-shaped core particle in which substrate cleavage takes place, and one or two regulatory particles that cap the ends of the barrel, restricting access to the core so that only proteins tagged with the pro-degradation molecule ubiquitin can enter. Regulatory particles have been observed in multiple conformations9,10, indicating that proteasomes progress through a reaction cycle that involves ground, committed and substrate-engaged states. Guo et al. used computational particle averaging to quantify the proteasomal states (technique reviewed in ref. 11)11, and found both ground and substrate-engaged states within poly(GA) aggregates. They also found a large increase in the proportion of doubly capped proteasomes (indicating engagement with substrate) compared with control neurons that did not contain poly(GA) products. Almost one-quarter of the proteasomes within aggregates adopted a conformation recently described9 as substrate-engaged yet stalled, in which the substrate becomes trapped in the barrel. And that proportion rose to 36% for those proteasomes closest to poly(GA) ribbons.
Why might this stalling occur? The authors’ tomographic reconstructions revealed numerous regions of electron density located between a poly(GA) ribbon and the site where the protein RAD23 binds to the proteasome. RAD23 is involved in recruiting ubiquitin-tagged proteins to the proteasome and is known to be enriched in poly(GA) aggregates8. Thus, this electron density could indicate RAD23-associated ubiquitin that is attached to proteins within the aggregate. Which protein or proteins the proteasome is choking on is currently unclear, although possibilities include the poly(GA) peptides themselves, which probably inhibit proteasome activity directly. Regardless of the mechanism, it seems likely that depletion of proteasomal activity in the cell proper would be detrimental to protein-degradation pathways, thereby contributing to cellular toxicity.
These results raise several important questions. First, pathogenesis in cases of ALS driven by C9orf72 expansion has been linked both to changes mediated by poly(GA) formation and to changes caused by reduced production of C9orf72 protein — but what are the relative contributions of each mechanism? C9orf72 is part of a complex involved in autophagy12, a process by which cellular material, including proteins, is degraded and recycled. It is therefore possible that reduced C9orf72 levels conspire with poly(GA)-dependent proteasome inhibition to increase neuronal toxicity. Second, is toxicity promoted by the capture of other proteins within poly(GA) aggregates? One candidate is the autophagy cargo receptor p62, which is known to accumulate in poly(GA) aggregates8. Third, several molecular machines involved in disassembling aggregates do not accumulate in poly(GA) structures, but the reasons for this are unclear.
Finally, although poly(GA) is the most abundant repetitive protein produced by C9orf72 expansion, it is not the only one — mutation can also produce tracts of glycine–arginine (poly(GR)) and proline–arginine (poly(PR)). How do the structures of these other aggregates compare to that of poly(GA) proteins? Most data on poly(GR) and poly(PR) aggregates indicate that they do not accumulate proteasomes, suggesting alternative toxicity mechanisms13,14. Further analysis by 3D cryo-ET, and analysis of natural products of C9orf72 expansion rather than the engineered product used in the current study, might clarify the similarities and differences between the aggregates that occur in patients and the model aggregate structures studied by Guo and colleagues.
In sum, the current work highlights the unprecedented resolution of 3D cryo-ET for visualizing fundamental processes within cells11. Moreover, it sets the stage for a more comprehensive mechanistic understanding of aggregate-associated neurodegenerative diseases.
Nature 555, 449-451 (2018)
DeJesus-Hernandez, M. et al. Neuron 72, 245–256 (2011).
Taylor, J. P., Brown, R. H. Jr & Cleveland, D. W. Nature 539, 197–206 (2016).
Yamakawa, M. et al. Hum. Mol. Genet. 24, 1630–1645 (2015).
Zhang, Y.-J. et al. Acta Neuropathol. 128, 505–524 (2014).
Guo, Q. et al. Cell 172, 696–705 (2018).
Beck, M. & Baumeister, W. Trends Cell Biol. 26, 825–837 (2016).
Chang, Y.-J., Jeng, U.-S., Chiang, Y.-L., Hwang, I.-S. & Chen, Y.-R. J. Biol. Chem. 291, 4903–4911 (2016).
Bauerlein, F. J. B. et al. Cell 171, 179–187 (2017).
Wehmer, M. et al. Proc. Natl Acad. Sci. USA 114, 1305–1310 (2017).
Chen, S. et al. Proc. Natl Acad. Sci. USA 113, 12991–12996 (2016).
Amick, J. & Ferguson, S. M. Traffic 18, 267–276 (2017).
May, S. et al. Acta Neuropathol. 128, 485–503 (2014).
Lin, Y. et al. Cell 167, 789–802 (2016).
Lee, K.-H. et al. Cell 167, 774–788 (2016).