Predicted Structure of Tau
Precise structural information about tau in solution has been difficult to obtain, since its high solubility precludes a crystalized defined structure. As a result, the structure of all human tau isoforms are yet to be determined, with RCSB Protein Databank turning no useful results upon searching for microtubule-associated tau. For this project, numerous protein-structure prediction servers were utilised with the aim of producing secondary structures for Tau-F (441AA, largest brain-isoform), as reliably as possible.
SWISS-MODEL (Automated mode)- is an integrated service dedicated to protein-structure homology modelling. Unlike other prediction engines, Swissmodel compares the amino acid sequence of a submitted protein with many other known structures in the PDB, to produce possible structures. This pipeline automatically identifies suitable templates based on Blast (Altschul et al.)[1] and HHblits (Remmert et al.)[2]. SWISS-MODEL assesses the quality of predicted models in the form of QMEAN, with higher numbers indicating high reliability of the depicted residues.
A predicted structure was calculated using conserved hypothetical protein TT1887 as a template. This template was calculated to have 16.13% sequence similarity, highly concentrated at the N-terminus of Tau, with a predicted alpha helix spanning Val411 to Leu441. This template's QMEAN score is -1.31 indicating that it is only moderately reliable.
Full results can be found here.
PDB coordinates can be downloaded here.
Val411
Jpred- is a web server that predicts protein secondary structures using a neural network called Jnet. The results predict the presence of either alpha helices (‘H’), beta sheets (‘E’) or random coils (‘-‘). Each residue is placed with a confidence of 1-9, with 9 being the maximum.
Jpred results predict an alpha-helix from AA331 to AA338. This confirms the findings from SWISS-MODEL seen above.
Download full results (Pages 97-108).
PSIPRED- is a simple secondary structure prediction method, using two feed-forward neural networks which perform an analysis on output obtained from PSI-BLAST.
PSIPRED results confirm the prediction of an alpha-helix between AA427-AA438, as seen above. Of great importance is the prediction of beta-sheets at Q277-I279 and Q307-Y310. These amino acid sequences fall within tubulin-binding domain 2 and 3 respectively. In various tauopathies such as Alzheimers (See medical section), these critical regions known as Paired Helical Filaments (PHFs) indeed assume beta-sheet structures required for tau-aggregation and formation of neurofibrillary tangles [3].
Full results can be downloaded here.
Literature VS Prediction
Inconsistencies appear between predicted structures and science literature concerning Tau. In solution, tau behaves as a random Gaussian coil, as judged by many methods such as electron microscopy, X-ray diffraction, NMR spectroscopy etc [4-7]. The ∼120 N-terminal residues are mostly acidic, projecting away from the microtubule surface [8]. The remainder of the protein has a basic character, complementary to the acidic surface of microtubules. The repeats (three or four, depending onthe isoform) in the C-terminus, form the core of the microtubule-binding domain; together with the proline-rich flanking domains [9-11], they regulate growth rates and dynamics of microtubules .
Tau is a natively unfolded protein, characterized by a lack of secondary structure [6-7]. NMR spectroscopy confirms some cases of secondary structures seen in PHF6 (VQIVYK) and PHF6* (VQIINK), known to play a critical role in the aggregation of tauopathies [12-13].
Even though tau is largely unfolded, it retains an intrinsic structure that can be denatured. Specifically, data from Jeganathan S et al. (2006)[14] shows that tau is globally folded in a 'paperclip' manner (Fig. 1). Both the N- and C- termini are folded over each other near the center of the tubulin-binding domains.
In conclusion, although PSIPRED, Jpred and SWISS-MODEL predict the presence of an alpha-helix at the C-terminus, this is not compatible with the scientific literature surrounding tau. Overall the predicted models show a lot of secondary structure present in tau, which is not true. This comes to remind us, that even though structure-prediction models are useful, they should not be treated as fully reliable. Hopefully Ab initio construction of proteins will become readily available in the near future , paralleled by unprecedented technological advancements.
[1] http://www.ncbi.nlm.nih.gov/pubmed/9254694
[2] http://www.ncbi.nlm.nih.gov/pubmed/22198341
[3] http://www.ncbi.nlm.nih.gov/pubmed/15855160
[4] http://www.ncbi.nlm.nih.gov/pubmed/146092
[5] http://www.ncbi.nlm.nih.gov/pubmed/1639844
[6] http://www.ncbi.nlm.nih.gov/pubmed/7929085
[7] http://www.ncbi.nlm.nih.gov/pubmed/10805776
[8] http://www.ncbi.nlm.nih.gov/pubmed/1465130
[9] http://www.ncbi.nlm.nih.gov/pubmed/1918161
[10] http://www.ncbi.nlm.nih.gov/pubmed/8068626
[11] http://www.ncbi.nlm.nih.gov/pubmed/9190213
Figure 1. 'Paperclip' structure of tau.
The polypeptide chain of Tau-F (441AA) is depicted with coloured tubulin binding domains (R1 blue; R2 green ; R3 yellow and R4 in red). The C- and N- termini sequences are shown in gray.
Adapted from Jeganathan S et al. (2006)
© 2015. CELL2008. Group 13: Andreas Millios, Rebecca Johnson, Dilen Ghetia, Fraz Azizi, Dominic Scaglioni, Nayoon Jang, Hannah De Bruijn.