Investigating Plant and Microbe Systems for Understanding the Formation and Modulation of Amyloid and Amyloid-Like Proteins
Liisa Lutter* (email@example.com), Jeff Qu, Samantha Zink, Jose A. Rodriguez, and David S. Eisenberg
University of California–Los Angeles DOE Institute for Genomics and Proteomics
The team aims to investigate plant and microbe systems as a platform for understanding the formation and modulation of amyloid or amyloid-like assemblies. This includes low complexity structures that create phase-separated physical systems that may be functional or pathogenic as well as those that form large macromolecular assemblies, particularly fibrils. Researchers therefore aim to better identify potential amyloid-forming sequences across microbial and plant species and identify compounds or molecules that inherently modulate the amyloid-forming propensity of such segments. The downstream aims of the work involve the identification of functional or pathogenic roles for such assemblies and specifically their roles in plant stress tolerance in response to increased pressures on agriculture resulting from climate change.
The aggregation of proteins into the amyloid state is often associated with disease, but amyloids or amyloid-forming sequences exist throughout the tree of life and have been known to carry out functional roles under normal physiological conditions. The team is therefore interested in the identification of amyloid-forming segments across microbes and plants and their modulation by endogenous molecules.
For amyloid prediction, the team relies on the fact that at their core, amyloid fibrils contain peptide segments assembled into tightly mated and interdigitated structures referred to as ‘steric zippers.’ Algorithms that rely on protein design approaches, such as secondary structure predictors and Rosetta modelers, have accurately predicted amyloid propensity for short segments, but they are computationally expensive to apply on the scale of entire genomes. Machine learning offers a faster and potentially accurate alternative to the prediction of amyloid propensity. Team members analyzed the amyloid-forming propensity of short peptide segments by training a neural network on the calculated Rosetta energy scores of computed six-residue steric zipper structures. Training a network on this data leverages many hours of compute time already invested in zipper structure predictions and promises to replace the expensive process with fast evaluation by the trained network. The network can yield propensity scores for 6-residue segments in seconds rather than hours or days. After training the network on over a million structures, researchers evaluated its performance by scoring its accuracy against a computed set during training and assessed its speed of prediction when scoring entire genomes containing millions of hexapeptide sequences. Experimentally, researchers evaluate predicted scores from the network by probing the propensity of synthetic peptides to form steric zippers in solution. For those segments that form what appear to be amyloid assemblies, researchers characterize their structures by X-ray crystallography or electron diffraction and compare their structures and scores to those predicted by the network. Predictions thus far are competent in either—forming fibrils crystals with cross-beta structures. Separately, the team has applied bioinformatic tools to determine proteins with amyloid sequence features from the Arabidopsis thaliana proteome. Gene ontology data was then used to identify proteins with functional roles in environmental stress responses, including heat and drought tolerance as well as roles in response to pathogens. From these, proteins with previously validated in vivo functional roles were selected for investigating the roles of the propensity of the LCDs to drive protein phase separation and aggregation in the function of the protein with the aim of modulating this behavior and the associated stress-response functions.
Our efforts to identify amyloid modulators have focused on plant natural products identified using open-access databases and computational evaluation of absorption, distribution, metabolism, excretion, and toxicity (ADMET) profiles. Researchers score these compounds based on their drug-like physico-chemical properties and targeting of model amyloids. The effects of highest ranking compounds on known amyloid proteins were evaluated in vitro using a fluorescent reporter Thioflavin T and transmission electron microscopy. Preliminary results show an inhibitory effect on amyloid formation or significant changes to sample morphology by several of the tested plant small molecules. Further characterization will involve determining the effect of the compounds on aggregation and their prion-like spread from cell to cell in biosensor cell assays.
In conclusion, this research will identify plant amyloid-forming proteins and plant molecules that influence amyloid aggregation with implications to their functional roles in environmental stress responses.
Research supported by the BER program of the DOE Office of Science, award DE-FC02-02ER63421.