Figure 1: APRs and gatekeepers in natural polypeptide sequences.

(a) The scatterplot of the average experimentally determined solubility^5,17 of E. coli proteins grouped by the number of APRs in their sequence displays an anti-correlation, showing that on average protein solubility decreases with increasing number of aggregation prone sequences. (b) A similar plot to A, but now for protein abundance in E. coli in vivo ^5,17. Although the anti-correlation is weaker, the trend suggests that cellular abundance is also limited by the number of APRs in its sequence. (c) Schematic representation depicting the TANGO profile of a typical naturally occurring polypeptide stretch of 100 amino acids, containing two APRs (red) each flanked by an N-terminal as well as a C-terminal gatekeeper (green), as well as one suppressed APR (blue) controlled by a single suppressing gatekeeper. (d) A density plot of the TANGO scores in the set of 634 structurally diverse proteins (red), showing a strong bias towards lower scores. In the absence of gatekeeper residues, this distribution is biased towards higher scores (green). The TANGO scores of the suppressed APRs (blue) show a more even distribution including higher scoring regions. (e) Density plot showing the number of APRs occurring per 100 amino acids in 634 structurally diverse proteins (red line). The blue line shows the number of ‘suppressed APRs’, which are not detected by TANGO in the WT sequence, but show up after mutation of a single Asp, Glu, Lys, Arg or Pro residue to Ala. If mutation of such a residue reveals an APR that is undetectable in the WT sequence, we call that residue a ‘suppressing gatekeeper’. The inset shows the frequency of the amino acid type of suppressing gatekeepers. (f) Density plot of the contribution of gatekeeper residues to the overall thermodynamic stability of the native state of the protein as calculated by FoldX (red). The same plot for the suppressing gatekeepers in the same proteins (blue) is very similar.