Insulin-regulated aminopeptidase (IRAP) is a transmembrane zinc metallopeptidase with many important biological functions and an emerging pharmacological target. Although previous structural studies have given insight on how IRAP recognizes linear peptides, how it recognizes its physiological cyclic ligands remains elusive. Here, we report the first crystal structure of IRAP with the macrocyclic peptide inhibitor HA08 that combines structural elements from angiotensin IV and the physiological substrates oxytocin and vasopressin. The compound is found in the catalytic site in a near canonical substrate-like configuration and inhibits by a competitive mechanism. Comparison with previously solved structures of IRAP along with small-angle X-ray scattering experiments suggests that IRAP is in an open conformation in solution but undergoes a closing conformational change upon inhibitor binding. Stabilization of the closed conformation in combination with catalytic water exclusion by the tightly juxtaposed GAMEN loop is proposed as a mechanism of inhibition.

Presentation of antigenic peptides by MHCI is central to cellular immune responses against viral pathogens. While adaptive immune responses versus SARS-CoV-2 can be of critical importance to both recovery and vaccine efficacy, how protein antigens from this pathogen are processed to generate antigenic peptides is largely unknown. Here, we analyzed the proteolytic processing of overlapping precursor peptides spanning the entire sequence of the S1 spike glycoprotein of SARS-CoV-2, by three key enzymes that generate antigenic peptides, aminopeptidases ERAP1, ERAP2, and IRAP. All enzymes generated shorter peptides with sequences suitable for binding onto HLA alleles, but with distinct specificity fingerprints. ERAP1 was the most efficient in generating peptides 8-11 residues long, the optimal length for HLA binding, while IRAP was the least efficient. The combination of ERAP1 with ERAP2 greatly limited the variability of peptide sequences produced. Less than 7% of computationally predicted epitopes were found to be produced experimentally, suggesting that aminopeptidase processing may constitute a significant filter to epitope presentation. These experimentally generated putative epitopes could be prioritized for SARS-CoV-2 immunogenicity studies and vaccine design. We furthermore propose that this in vitro trimming approach could constitute a general filtering method to enhance the prediction robustness for viral antigenic epitopes.