Thermodynamic and Evolutionary Coupling between the Native and Amyloid State of Globular Proteins

Abstract

The amyloid-like aggregation propensity present in most globular proteins is generally considered to be a secondary side effect resulting from the requirements of protein stability. Here, we demonstrate, however, that mutations in the globular and amyloid state are thermodynamically correlated rather than simply associated. In addition, we show that the standard genetic code couples this structural correlation into a tight evolutionary relationship. We illustrate the extent of this evolutionary entanglement of amyloid propensity and globular protein stability. Suppressing a 600-Ma-conserved amyloidogenic segment in the p53 core domain fold is structurally feasible but requires 7-bp substitutions to concomitantly introduce two aggregation-suppressing and three stabilizing amino acid mutations. We speculate that, rather than being a corollary of protein evolution, it is equally plausible that positive selection for amyloid structure could have been a driver for the emergence of globular protein structure.

Keywords: amyloid; evolution; protein folding; protein stability.

Graphical abstract

Figure 1

Stability and Aggregation Propensity Are Related (A) Class and kingdom composition of the SCOPe dataset. (B) Boxplot representation of the distribution of APRs in the SCOPe and IDP datasets. (C) Boxplot showing the contribution of APRs to the stability of the native state calculated by FoldX in the SCOPe dataset in function of the predicted aggregation propensity by TANGO. (D–G) Boxplots comparing APRs occurring in domains with one APR to those occurring in domains with more than one APR: the distribution TANGO score of APRs (D), the average main-chain burial (E), the average side-chain burial (F), and the average contribution of an APR to native-state stability (G, ΔG calculated by FoldX, in kilocalories per mole). (H, J, L, and N) Histograms of the melting temperature (Tm) observed in whole-proteome protein stability measurements (Leuenberger et al., 2017) for HeLa cells (H), S. cerevisiae (J), E. coli (L), and T. thermophilus (N). The dotted line indicates the mean Tm of the proteome in question. (I, K, M, and O) Boxplots comparing the normalized TANGO scores of proteins with a high or low Tm value in HeLa cells (I), S. cerevisiae (K), E. coli (M), and T. thermophilus (O). The bottoms and tops of the boxes are the first and third quartiles, and the band inside the box represents the median; the dot indicates the mean. The whiskers encompass the minimum and maximum of the data. Significant differences were computed using a Wilcox rank test. Asterisks denote level of significance: n.s., not significant; ^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗p < 0.001. The source files (Data S1) and R-scripts (Data S2) used to generate this figure are available.

Figure 2

Mutational Energies of APRs and Native Structures Are Coupled (A–G) Schematic representations of the crystal structures of APRs and their cognate native states for lysozyme (A), transthyretin (B), insulin (C), SOD1 (D), Bloom syndrome protein (E), p53 (F), and β2-microglobulin (G). The APR segment is highlighted in color in an otherwise gray native-state structure. The PDB identifiers used are listed in STAR Methods. (H) Density plot of the free energy change in thermodynamic stability (ΔΔG, FoldX, in kilocalories per mole) associated with mutation in the native state versus the APR state. The green box encompasses the APR-destabilizing mutations (ΔΔG at least 1.5 kcal/mol) that may be tolerated at the native structure level at a cutoff of ΔΔG no larger than 1 kcal/mol. The source files (Data S1) and R-script (Data S2) used to generate this panel are available.

Figure 3

Codon Usage and Native-State Stability Reduce the Number of Possible Aggregation-Suppressing Mutations in a Systematic Mutation Scan of the SCOPe Set of Protein Structures (A) Mutational screen for mutations that completely suppress an APR (TANGO score of mutant = 0) for all APRs with a TANGO score of more than 50 in the SCOPe set. The height of the bars indicates the number of times each amino acid is found in the suppressed mutant. (B) Types of suppressing mutations in the deep mutational scanning dataset. Bars depict the number of mutations predicted to be suppressing to the APRs identified in the dataset. Colors indicate amino acid type, and APRs are ordered along the x axis according to their TANGO score. (C) Receiver operating characteristic (ROC) curve for the prediction of the fitness outcome of a mutation (with a fitness score of more than 0.8 corresponding to a tolerated mutation) from the ΔΔG calculated through FoldX. ROC curves were produced for each amino acid type (color-coded as in B), and area under the curve (AUC) is reported for each prediction. (D) Density plot of the change of free energy of the native state (ΔΔG, FoldX, in kilocalories per mole) upon mutation versus the associated change in predicted aggregation propensity (TANGO). The top and right insets show histogram projections of the same data, split by whether (CodonPos = 1) or not (CodonPos = 0) the mutations are accessible through single-base changes in the cognate genes. The segmented lines indicate the means of the respected groups. (E) Venn diagram grouping the mutations from (A) into three groups: APR suppression (TANGO reduced to 0 by the mutation), codon compatible (mutation accessible through a single-base change of the cognate gene), and structure compatible (mutations with a ΔΔG value [FoldX] of less than 0.5 kcal/mol). (F) Heatmap of the percentage of APRs for which suppressing mutations can be identified in function of the contribution to the native-state stability of the APRs (ΔΔG, FoldX, in kilocalories per mole) and its aggregation propensity (TANGO). The source files (Data S1) and R-scripts (Data S2) used to generate this figure are available.

Figure 4

Evolutionary and Protein Engineering Analysis of an APR in the p53DBD (A) Structures of the APR-free adhesin fold (PDB: 1R17) and of the p53-family fold (PDB: 2AC0) with the APR highlighted red and the equivalent strand in the adhesin highlighted yellow. (B) Fraction of charged amino acids and hydrophobicity for the entire fold class B in dataset 1, with fold class B2 members highlighted blue, human p53 red, and bacterial adhesin yellow. (C) Degree of sequence conservation of the p53 DNA-binding domain (DBD) in all craniates per residue (black line) and as a rolling average per five residues (red). Functional residues are marked in blue (DNA-binding) and red (Zn ion coordination), and APR residues are marked in green. (D) APR sequence logos for several phylogenetic groups in dataset 5. (E) Total TANGO score of the ILTIITL sequence and selected mutants. (F) Aggregation kinetics of the indicated peptides monitored by Thioflavin-T fluorescence. (G) Location of selected mutations in the native structure of p53DBD. (H) Backbone alignment of the crystal structures of p53cc (gold) and the WT (blue). (I) Surface rendering of the crystal structure of p53cc with the introduced side chains highlighted. A blue surface marks a positive charge and a red surface a negative charge. (J) Electron density maps (contoured to 1σ) around the zinc binding site (H179) and mutated residues in p53cc.

Figure 5

Biophysical Analysis of p53cc (A) Heat denaturation of WT p53DBD, monitored by intrinsic fluorescence plotted as the barycentric mean (BCM) of the emission spectrum. The black line indicates the fit performed to determine the midpoint of the transition. (B) Same as (A) but for the CC mutant. (C) Average Tm values obtained from five biological replicate measurements. (D) DNA-binding affinity of p53cc and the WT, measured by FA. Error bars indicate standard deviation of five biological replicates, each measured in three technical replicates. (E) Degree of promoter activation of p53cc relative to WT p53 in Saos-2 cells, measured by qPCR. Error bars indicate standard deviation of three biological replicates, each measured in three technical replicates. (F) Aggregation kinetics of p53cc and the WT, measured by p-FTAA fluorescence. (G) Aggregation kinetics of p53cc and the WT, measured by ANS fluorescence. (F and G) Error bars indicate standard deviation of five biological replicates, each measured in three technical replicates. (H) Transmission electron microscopy (TEM) analysis of aggregates of WT p53. (I) TEM analysis of p53cc. (H and I) Representative images from three biological replicates.

Figure 6

Topological Invariance of Amyloid Addiction (A and B) Comparison of APRs that can be fully suppressed by a single codon-compatible mutation (rescuable) with those that cannot, in terms of SCOP class (A) and secondary structure element (B) in which they occur, following the FoldX code for secondary structure annotation (^∗, unclassified; 3, 3-10 helix; a, α helix; b, parallel beta conformation; B, antiparallel beta conformation; E, β strand; c, random coil; m, helix C-cap; n, helix N-cap; T, turn). (C) Example of a conserved APR in an all-α fold (SCOP class a): the VPS9 domain of Rab5 (Homo sapiens; PDB: 1TXU). (D) Example of a conserved APR in the α/β fold (SCOP class c): glutaminyl-peptide cyclotransferase (Homo sapiens; PDB: 2AFW). (E) Example of a conserved APR in an α+β fold (SCOP class d): E3 ubiquitin-protein ligase NRDP1 (Homo sapiens; PDB: 2FZP). (F) Schematic representation illustrating the APR addiction principle. The source files (Data S1) and R-scripts (Data S2) used to generate this figure are available.