Structure and Protein-Protein Interactions of Ice Nucleation Proteins Drive Their Activity

Abstract

Microbially-produced ice nucleating proteins (INpro) are unique molecular structures with the highest known catalytic efficiency for ice formation. Airborne microorganisms utilize these proteins to enhance their survival by reducing their atmospheric residence times. INpro also have critical environmental effects including impacts on the atmospheric water cycle, through their role in cloud and precipitation formation, as well as frost damage on crops. INpro are ubiquitously present in the atmosphere where they are emitted from diverse terrestrial and marine environments. Even though bacterial genes encoding INpro have been discovered and sequenced decades ago, the details of how the INpro molecular structure and oligomerization foster their unique ice-nucleation activity remain elusive. Using machine-learning based software AlphaFold 2 and trRosetta, we obtained and analysed the first ab initio structural models of full length and truncated versions of bacterial INpro. The modeling revealed a novel beta-helix structure of the INpro central repeat domain responsible for ice nucleation activity. This domain consists of repeated stacks of two beta strands connected by two sharp turns. One beta-strand is decorated with a TxT amino acid sequence motif and the other strand has an SxL[T/I] motif. The core formed between the stacked beta helix-pairs is unusually polar and very distinct from previous INpro models. Using synchrotron radiation circular dichroism, we validated the β-strand content of the central repeat domain in the model. Combining the structural model with functional studies of purified recombinant INpro, electron microscopy and modeling, we further demonstrate that the formation of dimers and higher-order oligomers is key to INpro activity. Using computational docking of the new INpro model based on rigid-body algorithms we could reproduce a previously proposed homodimer structure of the INpro CRD with an interface along a highly conserved tyrosine ladder and show that the dimer model agrees with our functional data. The parallel dimer structure creates a surface where the TxT motif of one monomer aligns with the SxL[T/I] motif of the other monomer widening the surface that interacts with water molecules and therefore enhancing the ice nucleation activity. This work presents a major advance in understanding the molecular foundation for bacterial ice-nucleation activity.

Ice nucleation proteins self-assemble into large fibres to trigger freezing at near 0 °C

“…9 ). The dimerization interface involves an interaction of the stacked tyrosine ladders from the two INP monomers as previously suggested (10, 24). However, in this model the INPs are aligned antiparallel to each other ( Fig.…”

“…Electron microscopy of newly synthesized INPs in a cell-free system shows them as thin molecules of dimensions 4-6 nm in diameter by a few hundred nm in length (25). Negatively stained images of recombinantly produced INP multimers isolated by centrifugation and chromatography show an elongated structure ∼5-7 times longer than a monomer but not much wider (24). The fibres seen in situ in INP-expressing E. coli ( Fig.…”

“…The tendency to form such large structures is one of many factors that makes these proteins difficult to work with (3) and may be part of why, despite many attempts, very little is known about them at a molecular level (22). The size of these structures does, however, make them amenable to size-based separation from other proteins (23, 24). INP multimers are also large enough to be visible using negative stain transmission electron microscopy (TEM) on enriched samples, revealing a fibril-like morphology (24, 25).…”

“…The size of these structures does, however, make them amenable to size-based separation from other proteins (23, 24). INP multimers are also large enough to be visible using negative stain transmission electron microscopy (TEM) on enriched samples, revealing a fibril-like morphology (24, 25).…”

Protein structure prediction in the era of AI: Challenges and limitations when applying to in silico force spectroscopy

“…The broad spread use of AI-based structure prediction leads us to ask the question: How reliable are the structures predicted by such models? Despite the growing number of success stories ( Jumper et al, 2021a ; Jumper et al, 2021b ; Mosalaganti et al, 2021 ; Skolnick et al, 2021 ; Hartmann et al, 2022 ; Varadi et al, 2022 ), researchers are accumulating evidence showing that AI-based structure prediction methods are still not perfect ( Perrakis and Sixma, 2021 ; Outeiral et al, 2022 ), and that there is ample room for improvement. In other words, some results suggest that both AlphaFold and RoseTTAFold are qualitatively great, but in many cases, they lack the level of details that is important to understand a protein function ( Akdel et al, 2021 ; Eisenstein, 2021 ; Callaway, 2022 ).…”