(309a) Comprehensive Profiling of the Substrate Specificity of Streptococcus Pyogenes Sortase a.

Sortase-mediated ligation (SML) is a well-established technology to generate protein nanocomplexes and continues to make a significant impact across multiple fields, particularly chemical biology and bioconjugate chemistry. Currently, sortase A from Staphylococcus aureus is the most frequently used enzyme to perform SML. S. aureus sortase recognizes a pentapeptide motif LPxTG (x = any amino acid), cleaves after the threonine residue, and ligate a N-terminal glycine nucleophile, generating a new peptide bond. Unfortunately, the narrow substrate specificity of sortases from the Staphylococcus family (only few variations are accepted) hinders our ability to access complex and orthogonal fusion protein designs. To expand the substrate scope of sortases, significant efforts are dedicated to profile and engineer the substrate specificity of sortase enzymes from diverse families and species. Unfortunately, current profiling tools rely on positional scanning, thus do not fully map the relationship among all the residues of the substrate motif. Moreover, no platform exists to access the interplay between the sortase substrate and nucleophile.

Here, we have developed a yeast-based high-throughput functional screen and combined it with next-gen sequencing and machine learning to comprehensively profile the substrate specificity of sortases in a combinatorial fashion. With this platform, we aim to establish high-throughput predictive rules of sortase substrate recognition, elucidate the interplay between the sort signal and the nucleophile, and engineer sortases with bespoke specificities for host proteins. In the context of a fixed LPET P4 to P1 positions, we show for the first time that the substrate specificity of S. pyogenes sortase A (S. pyo SrtA) extends beyond the P1’ Ala/Gly. S. pyo srt A presents an extended substrate binding pocket that accommodates P2’ and P3’ sort signal residues where P2’ prefers to be aromatic and nonpolar and P3’ prefers charged amino acids. Structure modeling and biochemical characterization support an increased catalytic efficiency of ligation when sort signal sequences are extended beyond P1’.

In the context of an extended sort signal, we mapped the overall substrate specificity of S. pyo srtA on a fully randomized DNA-encoded hexapeptide library, XXX(T/S)XXX, where X represent any amino acid on either side of the P1 threonine or serine. Sorting this library by FACS and reveals new non-canonical substrates and alternative cleavage preferences. Importantly, many of the isolated sequences are faster in SML experiments than the canonical LPETA substrate. Interestingly, in one such substrate, LPRTNLC, we observe a P1’ asparagine, where the enzyme typically favors a small amino acid. Our yeast platform is proving a powerful, generally applicable tool to understand the complex sortase substrate recognition. We are currently coupling our findings with a a support vector machine learning framework to delineate the protease−substrate interaction landscape of sortases. These discoveries will form the basis for engineering sortases with orthogonal specificities.