Clinical and Translational Science (Jul 2021)

Estimating peptide half‐life in serum from tunable, sequence‐related physicochemical properties

  • Marco Cavaco,
  • Javier Valle,
  • Isabel Flores,
  • David Andreu,
  • Miguel A. R. B. Castanho

DOI
https://doi.org/10.1111/cts.12985
Journal volume & issue
Vol. 14, no. 4
pp. 1349 – 1358

Abstract

Read online

Abstract Proteolytic instability is a critical limitation for peptide‐based products. Although significant efforts are devoted to stabilize sequences against proteases/peptidases in plasma/serum, such approaches tend to be rather empirical, unspecific, time‐consuming, and frequently not cost‐effective. A more rational and potentially rewarding alternative is to identify the chemical grounds of susceptibility to enzymatic degradation of peptides so that proteolytic resistance can be tuned by manipulation of key chemical properties. In this regard, we conducted a meta‐analysis of literature published over the last decade reporting experimental data on the lifetimes of peptides exposed to proteolytic conditions. Our initial database contained 579 entries and was curated with regard to amino acid sequence, chemical modification, terminal half‐life (t1/2) or other stability readouts, type of stability assay, and biological application of the study. Although the majority of entries in the database corresponded to (slightly or substantially) modified peptides, we chose to focus on unmodified ones, as we aimed to decipher intrinsic characteristics of peptide proteolytic susceptibility. Specifically, we developed a multivariable regression model to unravel those peptide properties with most impact on proteolytic stability and thus potential t1/2 predicting ability. Model validation was done by two different approaches. First, a library of peptides spanning a large interval of properties that modulate stability was synthesized and their t1/2 in human serum were experimentally determined. Second, the t1/2 of 21 selected peptides approved for clinical use or in clinical trials were recorded and matched with the model‐estimated values. With both approaches, good correlation between experimental and predicted t1/2 data was observed.