Study of the AlphaFold2 predicted structure of the yeast Gdt1 protein

Details

Supervisors

Morsomme, Pierre ; Soumillion, Patrice ; Deschamps, Antoine

Faculty

Faculté des sciences

Degree label

Master [120] en biochimie et biologie moléculaire et cellulaire, à finalité approfondie

Abstract

Congenital Disorders of Glycosylation (CDGs) represent a group of rare hereditary diseases linked to defects in the glycosylation process. A subtype of CDG is caused by mutations in the TMEM165 human gene. The TMEM165 protein is a membrane transporter belonging to the well-conserved GDT1 family, named after its yeast orthologue (Gdt1p), and characterised by the presence of a highly conserved motif. The GDT1 family gathers proteins with different structural organisations: some, like Dma, Ter1a, and Ter2b, contain only a single motif and three transmembrane helices, whereas others, such as TMEM165 and Gdt1p, contain two motifs and possess six helices. Proteins containing only three transmembrane helices may function as dimers: homodimers for Dma, which does not show a preferential orientation in the membrane, and heterodimers for Ter1a and Ter2b, each presenting a different topology. Despite numerous studies, there is currently no experimentally determined three-dimensional structure for members of the GDT1 protein family, thus limiting the understanding of how they function. Nowadays, obtaining an experimentally determined three-dimensional structure for any protein is a challenge, especially for membrane proteins. Fortunately, over the past five years, there have been an emergence of algorithms using artificial intelligence to improve protein structure prediction. The validation of AlphaFold and its subsequent public availability enabled the prediction of the structure of all proteins in databases, including those of the GDT1 family. Observation of the structures predicted by AlphaFold for this protein family led to the two questions of my Master’s thesis. On the one hand, AlphaFold predicts for Gdt1p, TMEM165, Ter1a, Ter2b, Dma and other proteins belonging to this family, structures consistent with those anticipated years earlier. The first objective of my Master’s thesis was to trace the evolutionary history of the GDT1 family by forcing the evolution of Dma. To that end, mutations were introduced into its sequence to give it a specific orientation, in order to prevent its homodimerization. Then, by expressing simultaneously two Dma mutants with opposite orientations, it could allow its heterodimerization, in a similar way to what is observed for the Ter1a-Ter2b dimer. However, the various mutations tested did not enable me to obtain a functional Dma-Dma heterodimer. On the other hand, some previous observations related to the genetic manipulation of Gdt1p could be understood under the light of the structure predicted by AlphaFold. The predicted structure of Gdt1p contains a pseudoknot. Its presence in the latter could explain why, in proteins possessing six transmembrane helices belonging to the GDT1 family, the C-terminal part remains short, and why it was not possible to obtain a recombinant Gdt1p protein tagged with GFP at its C-terminal end. Indeed, the stability of a protein with a knot significantly relies on the degree of freedom of the end closest to it. The two other objectives of my Master’s thesis therefore aim to verify the presence of this knot by taking advantage of the observation that the C-terminus of Gdt1p is kept short: adding an extension to this end could destabilize the protein. To validate this hypothesis, Gdt1p was first cleaved at different positions, creating two halves capable of heterodimerization (split-Gdt1p) and lacking the pseudoknot. Then, the effect of adding different C-terminal extensions to the C-terminal half of the split-Gdt1p compared to their addition to the complete protein was evaluated. As an overall conclusion, it was easier to genetically add extensions on the splits than on the complete protein, and, when additions could be performed on both full-size and split-Gdt1p, the heterodimers showed a more functional phenotype. Further experiments would be necessary to definitively validate the presence of the pseudoknot in Gdt1p structure.