Low complexity (LCD) protein domains stack to form dynamic liquid droplets in the cell; alas, some pathogenic mutations in these regions promote aggregation beyond this temporary physiological state. What flips the switch? Something as seemingly fleeting as a small hydrogen bond, according to researchers led by Steven McKnight and Glen Liszczak, University of Texas Southwestern Medical Center, Dallas. In July 1 Science, they reported that mutations in proline, which release the backbone nitrogen-hydrogen bond, improperly stabilize LCD peptide aggregates. Experimental capping of this nitrogen-normalized peptide solubility. This was true for point mutations in tau, neurofilament lumen (NfL), and RNA-binding proteins TDP-43 and hnRNPA2. All are linked to neurodegenerative diseases.

  • Proline mutations allow nitrogen-hydrogen bonding in the protein backbone.
  • This reinforces the crossed β-sheets between the peptides, forming fibrils.
  • Main chain nitrogen methylation normalizes protein structure.

“This work [provides] a structural link between gene mutation and pathology, wrote David Eisenberg, University of California, Los Angeles (full comment below).

“These in vitro studies suggest that…hydrogen bonds of the peptide backbone can influence aberrant phase transitions of LCDs and could potentially be tuned to control self-association and aggregation of disease-related proteins,” Dorothee wrote. Dormann, Johannes Gutenberg-Universität Mainz, Germany (full comments below). Ben Wolozin of Boston University agreed, noting that “this is one of the cumulative factors that leads to TDP-43 aggregation.”

The LCDs of TDP-43, tau, and hnRNA2 self-aggregate forming labile cross-β-sheets, allowing proteins to partially slip out of the cytosol as liquid droplets, much like oil suspended in the air. (March 2019 news; August 2017 news; February 2018 news). What drives this liquid-liquid phase separation?

First author Xiaoming Zhou suspected skeletal hydrogen bonding because it is crucial for β-sheet formation. He created 23 variants of the LCD region of TDP-43, a 150 amino acid peptide, each with a modified amino acid in the conserved part of the protein. Synthetic amino acids contained a methylated backbone nitrogen to prevent intermolecular hydrogen bonding. Zhou then used microscopy to analyze each peptide’s ability to undergo phase separation in neutral isotonic buffer.

Nine variants failed to merge into liquid droplets. All of the mutated residues were in the region that forms the core of the β-sheet (December 2021 news). “This removal of a single hydrogen bond from the spine could disrupt the entire structure shows how weak—and therefore dynamic and reversible—self-association is,” Gregory Petsko, Harvard Medical School, Boston, and Scott Small , Columbia University, New York, wrote in an editorial in the same issue of Science. The researchers believe that the hydrogen bonds of the peptide backbone are crucial for β-sheet self-aggregation and phase separation.

Amino acid side chains can also influence peptide structure. Zhou systematically replaced every amino acid in the conserved region of TDP-43 LCD with glycine. Curiously, the most profound change occurred in the proline 320 mutant, which rushed into tangles rather than coalescing into liquid droplets. This mutation released the nitrogen-hydrogen bond from the main chain of the residue. When the scientists capped this nitrogen by incorporating an N-methyl glycine into the LCD peptide, droplets again formed. The authors concluded that the release of this nitrogen from the skeleton stabilized the labile aggregates and precipitated (see image below).

Hydrogen Velcro? Backbone hydrogen bonding (dashed lines) slightly holds low complexity domains together (blue squiggles). These domains remain labile due to occasional proline (left), which cannot bind. Substitution of any other residue allows the H bond of the main (middle) chain to stabilize the crossed β-sheet. This is reversed by methylating the backbone nitrogen (right) to restore the H bond. [Courtesy of Zhou et al., Science, 2022.]

Could mutations in the DNA sequence encoding proline explain why variants of other aggregate-prone proteins increase the risk of neurodegenerative disease? Indeed, conversion of proline to another amino acid in tau LCDs is known to cause tauopathies, proline exchanges in NfL cause the hereditary peripheral neuropathy of Charcot-Marie-Tooth disease, and in the hnRNPA2, amyotrophic lateral sclerosis and the bone disorder Paget’s disease (Jordanova et al., 2003; Shin et al., 2008; Qi et al., 2017).

Specifically, tau middle domain peptides that harbor P301S, P301T, or P301L had a strong propensity to aggregate and sequester the dye thioflavin-T, which alters its fluorescence after attaching to aggregated proteins. Similarly, six NfL proline mutants readily polymerized to form amorphous tangles rather than wild-type filaments. The LCD fragment of hnRNPA2 containing P298L aggregated and coalesced into misshapen liquid droplets, consistent with previous research (February 2018 news). For each of these mutant proteins, the incorporation of N-methylated amino acids to block further hydrogen bonding reversed its stability and solubility compared to wild type (see image below).

Out of phase. Pathogenic proline mutations cause NfL to randomly tangle together (top left), tau to aggregate when it shouldn’t (middle left), and hnRNPA2 to coalesce into liquid droplets that appear distorted (at the bottom left). Main-chain nitrogen methylation of the mutated amino acid restored all three proteins to their wild-type conformations (right panels). [Courtesy of Zhou et al., Science, 2022.]

Any caveats from these experiments? They only involved the LCD fragment of a given protein, not the full sequence. “What is very clear with a small peptide like this is more ambiguous when studying the whole protein,” Wolozin told Alzforum. The pathological relevance of these findings remains uncertain.

Dan Li, Shanghai Jiao Tong University, China, inquired about the detection of labile crossed β-sheets in cells or tissues. Although it does not provide a direct intracellular view, deep mutation scanning allows for the systematic study of thousands of variants of protein structure and aggregation within cells (Starita et al., 2017). “It is possible to reveal structural patterns and the in vivo aggregation state of full-length proteins containing LCDs and determine their effect on cell shape,” agreed Benedetta Bolognesi, Institute of Bioengineering of Catalonia, Barcelona , Spain (full reviews below).

Using deep mutation scanning, Bolognesi found that out of 52,000 mutations in the LCD of TDP-43, the conserved region was a hotspot for the creation of toxic mutants that formed solid aggregates rather than clusters. liquid-like droplets (Bolognesi et al., 2019). Interestingly, the pattern of interactions between LCD mutations suggested two structures in this region: a β-sheet at residues 311-315 and an α-helix at 324-331.

Researchers debate the secondary structure in which the conserved region of TDP-43 contorts. Some have reported that the α-helices of TDP-43, not the β-sheets, drive liquid-liquid phase separation (news September 2016). This raises the question of whether crossed β-sheets actually control LCD phase separation, as Zhou’s study did not include structural experiments. “Another possibility could be that the introduced backbone methyl groups interfere with the helical conformation and thus interfere with the phase separation of TDP-43,” Dormann wrote. Laboratories directly involved in this debate declined to comment.—Chelsea Weidman Burke

News Quotes

  1. More evidence for distinct TDP-43 droplets

  2. More Tau droplets

  3. Do reversible amyloids cause liquid-liquid phase separation?

  4. The double spiral distinguishes TDP-43 from other amyloids

  5. How does a neuron avoid the aggregation of liquid protein droplets?

  6. The helical tail dominates the TDP-43 packaging

Mutation Quotes

  1. MAPT P301S

  2. MAPT P301T

  3. MAPT P301L

paper quotes

  1. .
    Mutations in the neurofilament light chain (NEFL) gene cause severe early-onset Charcot-Marie-Tooth disease.
    Brain. 2003 Mar;126(Pt 3):590-7.
  2. .
    NEFL Pro22Arg mutation in Charcot-Marie-Tooth disease type 1.
    J Hum Genet. 2008;53(10):936-940. Published online August 29, 2008
  3. .
    Early-onset Paget’s disease of bone associated with a novel hnRNPA2B1 mutation.
    Fabric Calcif Int. 2017 Aug;101(2):159-169. Published online April 7, 2017
  4. .
    Variant Interpretation: Functional Testing to the Rescue.
    Am J Hum Genet. 2017 Sep 7;101(3):315-325.
  5. .
    The mutational landscape of a prion-like domain.
    Nat Common. 2019 Sep 13;10(1):4162.