Limits...
What makes a protein sequence a prion?

Sabate R, Rousseau F, Schymkowitz J, Ventura S - PLoS Comput. Biol. (2015)

Bottom Line: In many cases, prion structural conversion is driven by the presence of relatively large glutamine/asparagine (Q/N) enriched segments.Several studies suggest that it is the amino acid composition of these regions rather than their specific sequence that accounts for their priogenicity.However, our analysis indicates that it is instead the presence and potency of specific short amyloid-prone sequences that occur within intrinsically disordered Q/N-rich regions that determine their prion behaviour, modulated by the structural and compositional context.

View Article: PubMed Central - PubMed

Affiliation: Departament de Fisicoquímica, Facultat de Farmàcia, Universitat de Barcelona, Barcelona, Spain; Institut de Nanociència i Nanotecnologia (IN2UB), Universitat de Barcelona, Barcelona, Spain.

ABSTRACT
Typical amyloid diseases such as Alzheimer's and Parkinson's were thought to exclusively result from de novo aggregation, but recently it was shown that amyloids formed in one cell can cross-seed aggregation in other cells, following a prion-like mechanism. Despite the large experimental effort devoted to understanding the phenomenon of prion transmissibility, it is still poorly understood how this property is encoded in the primary sequence. In many cases, prion structural conversion is driven by the presence of relatively large glutamine/asparagine (Q/N) enriched segments. Several studies suggest that it is the amino acid composition of these regions rather than their specific sequence that accounts for their priogenicity. However, our analysis indicates that it is instead the presence and potency of specific short amyloid-prone sequences that occur within intrinsically disordered Q/N-rich regions that determine their prion behaviour, modulated by the structural and compositional context. This provides a basis for the accurate identification and evaluation of prion candidate sequences in proteomes in the context of a unified framework for amyloid formation and prion propagation.

Show MeSH

Related in: MedlinePlus

Relationship between amyloid and prion propensities.Average pWALTZ scores of prion (white) and non-prion (red) domains.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4288708&req=5

pcbi-1004013-g002: Relationship between amyloid and prion propensities.Average pWALTZ scores of prion (white) and non-prion (red) domains.

Mentions: The WALTZ algorithm can be run using different levels of stringency or custom defined thresholds. In a typical use, WALTZ high stringency levels (>90%) are employed in order to identify very short and potent segments able to nucleate amyloid formation with high specificity. For example, the analysis of the 758 residue long Tau protein renders a single prediction overlapping with the experimentally validated hexapeptide 591-KVQIIN-596 [26]–[31]. However, the identification and scoring of these strong and short protein stretches, usually flanked by highly soluble residues, does not allow an accurate discrimination between prionic and non-prionic Q/N rich sequences [15]. In the present approach, a sequence is considered for further analysis as a putative PFD candidate only if at least in one of the sliding windows all the 21 residues display values higher than the a given threshold. SUP35 was excluded from the test set, since one tetra- and three hexa-peptides belonging to its PFD sequence were part of the WALTZ training set. We used receiver operating characteristic (ROC) curves and evaluated the area under the curve (AUC) for each particular WALTZ stringency level (between 0 and 100%) and used the derived Youden's index for each plot to identify the threshold and the associated WALTZ score rendering the best predictability. The best values were obtained using a threshold of 35%. Despite this amyloidogenicity value is very low, according to the WALTZ scale, already seven of the non-prion proteins did not exhibit any continuous 21 residues sequence stretch able to pass the threshold. We used the rest of 32 non-prion domains and the 11 prion domains to elaborate the correspondent ROC plot which displays a striking AUC of 0.99 (Fig. 1), employing a WALTZ score cut off of 73.55% to discriminate between prion and non-prion domains (Table 1) according to the associated Youden's index. With these parameters, the approach, which we call now as pWALTZ, has a significance P value <0.0001, a sensitivity of 90.9% and a specificity of 97.4%, with only one false positive (PUF4) and one false negative (PUF2) among the 43 analysed proteins (Table 1) and an overall accuracy of 95.3% (Fig. 1). All the known bona fide yeast prions included in the test set (NEW1, RNQ1, SWI1 and URE2) are correctly classified as positive hits. The approach outperforms composition based algorithms like PAPA (Fig. 1) [15], which displays a 86.0% overall accuracy in the same dataset. As expected, SUP35 is also correctly classified as a prion (Table 1). As shown in Fig. 2, prionic sequences display clearly overall higher pWALTZ values than non-prionic ones. The observed difference is significant, especially if we take into account that we do not include in the comparison those sequences that failed to past the soft 35% initial threshold. An example of the scoring of prion and non-prion sequences is provided in the Supplementary Material (S2 Fig.).


What makes a protein sequence a prion?

Sabate R, Rousseau F, Schymkowitz J, Ventura S - PLoS Comput. Biol. (2015)

Relationship between amyloid and prion propensities.Average pWALTZ scores of prion (white) and non-prion (red) domains.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4288708&req=5

pcbi-1004013-g002: Relationship between amyloid and prion propensities.Average pWALTZ scores of prion (white) and non-prion (red) domains.
Mentions: The WALTZ algorithm can be run using different levels of stringency or custom defined thresholds. In a typical use, WALTZ high stringency levels (>90%) are employed in order to identify very short and potent segments able to nucleate amyloid formation with high specificity. For example, the analysis of the 758 residue long Tau protein renders a single prediction overlapping with the experimentally validated hexapeptide 591-KVQIIN-596 [26]–[31]. However, the identification and scoring of these strong and short protein stretches, usually flanked by highly soluble residues, does not allow an accurate discrimination between prionic and non-prionic Q/N rich sequences [15]. In the present approach, a sequence is considered for further analysis as a putative PFD candidate only if at least in one of the sliding windows all the 21 residues display values higher than the a given threshold. SUP35 was excluded from the test set, since one tetra- and three hexa-peptides belonging to its PFD sequence were part of the WALTZ training set. We used receiver operating characteristic (ROC) curves and evaluated the area under the curve (AUC) for each particular WALTZ stringency level (between 0 and 100%) and used the derived Youden's index for each plot to identify the threshold and the associated WALTZ score rendering the best predictability. The best values were obtained using a threshold of 35%. Despite this amyloidogenicity value is very low, according to the WALTZ scale, already seven of the non-prion proteins did not exhibit any continuous 21 residues sequence stretch able to pass the threshold. We used the rest of 32 non-prion domains and the 11 prion domains to elaborate the correspondent ROC plot which displays a striking AUC of 0.99 (Fig. 1), employing a WALTZ score cut off of 73.55% to discriminate between prion and non-prion domains (Table 1) according to the associated Youden's index. With these parameters, the approach, which we call now as pWALTZ, has a significance P value <0.0001, a sensitivity of 90.9% and a specificity of 97.4%, with only one false positive (PUF4) and one false negative (PUF2) among the 43 analysed proteins (Table 1) and an overall accuracy of 95.3% (Fig. 1). All the known bona fide yeast prions included in the test set (NEW1, RNQ1, SWI1 and URE2) are correctly classified as positive hits. The approach outperforms composition based algorithms like PAPA (Fig. 1) [15], which displays a 86.0% overall accuracy in the same dataset. As expected, SUP35 is also correctly classified as a prion (Table 1). As shown in Fig. 2, prionic sequences display clearly overall higher pWALTZ values than non-prionic ones. The observed difference is significant, especially if we take into account that we do not include in the comparison those sequences that failed to past the soft 35% initial threshold. An example of the scoring of prion and non-prion sequences is provided in the Supplementary Material (S2 Fig.).

Bottom Line: In many cases, prion structural conversion is driven by the presence of relatively large glutamine/asparagine (Q/N) enriched segments.Several studies suggest that it is the amino acid composition of these regions rather than their specific sequence that accounts for their priogenicity.However, our analysis indicates that it is instead the presence and potency of specific short amyloid-prone sequences that occur within intrinsically disordered Q/N-rich regions that determine their prion behaviour, modulated by the structural and compositional context.

View Article: PubMed Central - PubMed

Affiliation: Departament de Fisicoquímica, Facultat de Farmàcia, Universitat de Barcelona, Barcelona, Spain; Institut de Nanociència i Nanotecnologia (IN2UB), Universitat de Barcelona, Barcelona, Spain.

ABSTRACT
Typical amyloid diseases such as Alzheimer's and Parkinson's were thought to exclusively result from de novo aggregation, but recently it was shown that amyloids formed in one cell can cross-seed aggregation in other cells, following a prion-like mechanism. Despite the large experimental effort devoted to understanding the phenomenon of prion transmissibility, it is still poorly understood how this property is encoded in the primary sequence. In many cases, prion structural conversion is driven by the presence of relatively large glutamine/asparagine (Q/N) enriched segments. Several studies suggest that it is the amino acid composition of these regions rather than their specific sequence that accounts for their priogenicity. However, our analysis indicates that it is instead the presence and potency of specific short amyloid-prone sequences that occur within intrinsically disordered Q/N-rich regions that determine their prion behaviour, modulated by the structural and compositional context. This provides a basis for the accurate identification and evaluation of prion candidate sequences in proteomes in the context of a unified framework for amyloid formation and prion propagation.

Show MeSH
Related in: MedlinePlus