Fig. 5: Experimental validation of predicted substrates and non-substrates.

a Immunoblot analysis of substrate candidates tested in a cell-based cleavage assay. C-terminal fragment (CTF, ●) accumulation in PS1/2 DKO cells validates their substrate status. For STYK1, the full-length (FL) protein form (○) accumulated as this protein can be cleaved directly by γ-secretase. For ERBB2, a smaller band than the expected molecular weight (MW) of its CTF was found, likely due to subsequent caspase cleavage of the accumulating CTF⁹⁷. b Immunoblot analysis of non-substrate candidates, performed as in (a). Unchanged FL levels validate their non-substrate status. In a, b, β-Actin served as a loading control. c, d Immunoblot analysis of substrate and non-substrate TMD-based peptides (TMD) tested in a cell-free cleavage assay. Cleavage products (CP) were found after incubation at 37 °C for substrate peptides (c), but not non-substrate peptides (d). To control for cleavage specificity, the γ-secretase inhibitor (GSI) DAPT was used⁸¹. For APP, LVMLKKK-Biotin was used as a CP migration standard (M) in (c). Substrates and non-substrates are indicated by their gene name followed by their substrate prediction scores ± standard deviation (see Methods “Aggregation of prediction results”) using the color code for confidence-based substrate classes (Fig. 3b). VCAM1 and RELL2 are outliers in that they proved to be substrate or non-substrate, respectively, contrary to the prediction. Small black, blue, and red lines indicate 105, 34, and 7 kDa MW markers. All tested candidates were from human, except CD68, ICAM1, STYK1, ACLS5, and SLC27A1, which were from mouse (see Methods “Selection of substrate and non-substrate candidates”). Immunoblot analyses in a–d are representative of three independent experiments. e A scatterplot showing the correlation (two-sided Spearman correlation) between substrate prediction scores and recently reported log2 fold change (FC) of CTF accumulation in the presence of a GSI against control, obtained for 85 endogenously expressed proteins in human microglia-like cells⁴⁹. The maximum FC was used for proteins with CTF detection only during inhibition. The regression estimate (solid black line) with 95% confidence interval (gray shaded area) and the chosen substrate identification FC threshold (dashed gray line) are indicated. Source data are provided as a Source Data file.