Protein O-fucosyltransferase (POFUT) enzymes mediate direct O-fucosylation on secreted and membrane-associated proteins such as thrombospondin-1 and Notch-11,2. Dysregulation of O-fucosylation has been linked to numerous cancers and developmental disorders3. POFUTs play a key role in modulating the secretion of their substrates, by O-fucosylating specific substrate protein domains only when correctly folded, and as such are thought to be involved in non-canonical protein quality control4. However, as O-glycosylation is critically understudied, the true diversity of proteins involved in O-fucosylation is largely uncharacterised5.
We recently identified human POFUT3/POFUT4 as the enzymes responsible for O-fucosylation of Elastin Microfibril Interface (EMI) domain-containing proteins6. Here, we combined Alphafold3 protein structural prediction and the structural alignment tool Foldseek to identify new substrates for POFUT3/POFUT4. Each EMI-domain was searched iteratively in Foldseek to detect structurally homologous protein domains. Numerous proteins with EMI-like domains were identified, including MFAP2/MFAP5, which have key roles in stabilising the extracellular matrix and maintaining blood vessel integrity. The structure of the EMI domain is highly conserved in these proteins despite low primary sequence homology. As such, we hypothesised they may be recognised by the structurally-specific POFUT3/POFUT4 enzymes. HEK293 cells were transiently transfected with plasmids expressing the putative substrates in the presence or absence of POFUT3/POFUT4, to identify interactions and quantify O-fucosylation. We have confirmed O-fucosylation within several EMI-like domains and have observed these modifications at high stoichiometry within human visceral adipose tissue. Interestingly, O-fucosylation of MFAP2/MFAP5 is only observed in cells co-transfected with POFUT3, but not with POFUT4, a selectivity that has not been seen for EMI domains. We also observed that secretion of MFAP2/MFAP5 is significantly decreased in POFUT3 KO HEK293 cells, indicating that POFUT3 may retain its known role in modulating secretion for these putative substrates.
Hence, we have shown the potential for structural homology tool Foldseek to identify new substrates for domain-specific substrate-enzyme interactions. Future work will investigate the function of this O-fucosylation through POFUT3/POFUT4 knockout mice, and streamline this workflow for application to other domain-specific protein-protein interactions. A methodology for rapid identification of putative protein-protein interactions would provide a significant broadening of our knowledge of the human interactome.