The tubulin database: Linking mutations, modifications, ligands and local interactions

Abstract

Microtubules are polymeric filaments, constructed of α-β tubulin heterodimers that underlie critical subcellular structures in eukaryotic organisms. Four homologous proteins (γ-, δ-, ε- and ζ-tubulin) additionally contribute to specialized microtubule functions. Although there is an immense volume of publicly available data pertaining to tubulins, it is difficult to assimilate all potentially relevant information across diverse organisms, isotypes, and categories of data. We previously assembled an extensive web-based catalogue of published missense mutations to tubulins with >1,500 entries that each document a specific substitution to a discrete tubulin, the species where the mutation was described and the associated phenotype with hyperlinks to the amino acid sequence and citation(s) for research. This report describes a significant update and expansion of our online resource (TubulinDB.bio.uci.edu) to nearly 18,000 entries. It now encompasses a cross-referenced catalog of post-translational modifications (PTMs) to tubulin drawn from public datasets, primary literature, and predictive algorithms. In addition, tubulin protein structures were used to define local interactions with bound ligands (GTP, GDP and diverse microtubule-targeting agents) and amino acids at the intradimer interface, within the microtubule lattice and with associated proteins. To effectively cross-reference these datasets, we established a universal tubulin numbering system to map entries into a common framework that accommodates specific insertions and deletions to tubulins. Indexing and cross-referencing permitted us to discern previously unappreciated patterns. We describe previously unlinked observations of loss of PTM sites in the context of cancer cells and tubulinopathies. Similarly, we expanded the set of clinical substitutions that may compromise MAP or microtubule-motor interactions by collecting tubulin missense mutations that alter amino acids at the interface with dynein and doublecortin. By expanding the database as a curated resource, we hope to relate model organism data to clinical findings of pathogenic tubulin variants. Ultimately, we aim to aid researchers in hypothesis generation and design of studies to dissect tubulin function.

Fig 1

(A) A schematic of an individual sidechain (R) in tubulin (green) illustrates elements that may be altered by a point mutation that changes the amino acid at this position. This may alter: (1) local interactions with other amino acids (purple); (2) the capacity for reversible posttranslational modifications (M) such as phosphorylation or acetylation; and (3) interactions with small molecules including drugs such as paclitaxel (blue). (B) Amino acid interactions encompass: (1) interactions within an individual α- or β- subunit, (2) protein-protein interactions between monomers to form the α-β heterodimer interface; (3) dimer assembly into the microtubule lattice; and (4) interactions with other proteins such as microtubule motors. (C) A schematic illustrates how the universal numbering accommodates insertions or deletions relative to the reference sequence when aligned with homologous sequences harboring insertions or deletions. Amino acid insertions are denoted by “i” following the last conserved residue. This corrects numbering following the insertion to be consistent with the consensus sequence. Similarly, in the case that a tubulin harbors a deletion, the numbering at this site is modified to renumber the following amino acids to be in line with the consensus. (D) The α-tubulin H1’-S2 loop has variation in length as illustrated by an alignment of the consensus sequence with S. cerevisiae tubulin (+1) and C. elegans tubulin (-2) (differences highlighted in yellow). Neither S. cerevisiae (ScTUB1) nor C. elegans (CeTBA1) tubulins harbor K40, a conserved residue in many α-tubulins, including human α1a-tubulin (TBA1A, green highlight), which can be modified by acetylation to increase microtubule stability.

Fig 2. The tubulin database tables.

Records can be expanded beyond the single line entry to view additional information and both the tubulin sequence and original reference are hyperlinked (blue). (A) Validated and predicted sites of tubulin PTMs were collected and organized by universal tubulin numbering to document the position and type of the modification; tubulin isotype and species; whether the entry is validated or predicted, with hyperlinks to the tubulin sequence and original data source. (B) Tubulin structures define amino acids that contribute to contours of small molecule binding pockets, interfaces between tubulin subunits in the microtubule lattice and interactions with associated proteins. Individual entries contain hyperlinks to the relevant PDB file and associated publication. This image illustrates the utility of the UTN: amino acid R402 in vertebrates is equivalent to R406 in S. pombe. This position is located at the interface with microtubule motors. (C) The missense mutations datasets are now organized by universal tubulin numbering. This screenshot shows a previously unappreciated correlation between mutation A174V in human α1-tubulin and A178V in S. pombe. The fission yeast mutation impairs the interaction of α- and γ-tubulin; the human mutation is associated with a neurodevelopmental disorder.

Fig 3. Sequence diversity, sites of mutations, modifications, and interactions in α-tubulin residues 1–440.

Diversity is represented by a sequence logo graphical representation with the corresponding consensus sequence displayed below. An alignment of 90 sequences from organisms represented in the mutation database was used to create the image (see Methods for accession numbers). Amino acids that are represented by lowercase letters meet a lower statistical threshold as a consensus residue and “x” denotes positions that lack a clear consensus. For each position, three rows of asterisks are used to map intersections of indexed data: these indicate positions with documented mutations (red), sites of PTMs (blue); and sites of interaction (green).

Fig 4. Sequence diversity, sites of mutations, modifications, and interactions in β-tubulin residues 1–440.

An alignment of 90 representative β-tubulin sequences was used to create the image (see Methods for accession numbers). The details of this figure are identical to those for Fig 3.

Fig 5. Sequence diversity, sites of mutations, modifications, and interactions in γ-tubulin residues 1–440.

A Clustal Omega alignment of 25 γ-tubulin sequences from organisms represented in the mutation database was used to create the image (see Methods for accession numbers). The details of this figure are identical to those for Fig 3.

Fig 6. Genetic evidence for a proposed model of K40 acetylation and microtubule stability.

(A) Molecular dynamics simulations by the Nogales group suggest that αK40 acetylation increases the likelihood that αE55 (anionic partner, AP) forms a salt bridge with αH283 (cationic partner, CP) in the adjacent dimer to increase microtubule stability. In this model, when αK40 is unacetylated, it forms an internal salt bridge with αE55 sequestering it from interaction with αH283. (B) In budding yeast, alanine substitutions at αE55 and αH283 cause supersensitivity (ss) to the microtubule-disrupting drug benomyl, in-line with loss of the protofilament-spanning salt bridge causing reduced microtubule stability. (C) Replacement of lysine with arginine, an amino acid that can form salt bridges as a cationic partner but cannot be acetylated, makes Tetrahymena hypersensitive to microtubule disruption and less sensitive to Taxol, a stabilizing drug. This phenotype is consistent with reduced stability microtubules when αE55 is sequestered. Substitution with glutamine (an acetylation mimetic) but not arginine is recovered in the single conventional α-tubulin gene in Toxoplasma, illustrating that the cross-filament salt bridge is essential for microtubule stability. (D) CHO cell mutations H283Y and E55K confer resistance to vinblastine and colcemid. These substitutions disrupt the salt bridge between αE55 and αH283 in adjacent protofilaments but would likely act to improve protofilament affinity. The Y283 substitution may still form salt bridges or may participate in novel polar interactions across protofilaments. If correctly positioned, K55 could serve as a cationic partner to form a novel salt bridge with E284. (E) E55K and E55G substitutions documented in cases of lissencephaly would disrupt both proposed salt bridges.