Bioinformatics analysis of protein acetylation

Acetylation data were acquired from www.phosphosite.org⁶⁸ between December 2016 to September of 2017. Structure files (www.rcsb.org) for membrane-binding domain families were downloaded in bulk. Scripts were generated to sort these.pdb files into homology clusters based on a structural homology algorithm generated in-house. Family members with less than 50% sequence identity were excluded from the analysis.

Determination of lysine acetylation frequency in membrane-binding and non-binding regions was performed as follows. Based on the availability of defined crystal structures and high resolution experimental data on membrane binding in the literature, proteins were chosen to serve as templates for structural and sequence comparisons within domain families. The domain ranges of the templates were defined based on UNIPROT (uniprot.org) definitions. In the case of BAR domains, N-terminal regions containing membrane-embedded helices were also included within the domains. In the case of EH domain proteins, domain ranges and terminology were defined according to the pioneering work of Daumke et al.⁶⁵. The MIRs and NBRs of templates were defined by mapping the experimental data gathered from the literature onto the crystal structure using PyMOL. Charge density mapping was also used to help visualize and validate definitions determined from empirical data.

For domain family members with available crystal structures, but without empirically well-defined membrane-interaction regions, structural alignments were made using PyMOL (PyMOL Molecular Graphic System, Version 1.7.4 Schrödinger, LLC) against a template from which the domain range and MIR definitions were determined. For BAR domains lacking N-terminal helices in their pdb structures, MPEx¹⁰⁹ was used to define the helix for comparisons. For F-BAR, a number of domains display a kink that rotates the tip region. Alignments in such cases were performed against regions comparable to the template on either side of the kink. The assignments were further verified by electrostatic potential mapping and validation with available literature.

For proteins lacking crystal structures, a sequence alignment method was used. MIRs and NBRs were defined from template domains and ensembles of protein sequences aligned using ClustalOmega (https://www.ebi.ac.uk/Tools/msa/clustalo/)¹¹⁰. Colorations for figures were generated using ESPript (PMID: 24753421 http://espript.ibcp.fr)¹¹¹ to highlight the highly conserved residues among each domain family. To avoid aligning non-functional C2 domains¹¹², we applied a threshold similarity score cutoff of <50% similarity, relative to the C2 templates, to exclude degenerate domains. Thus, for each C2 family member, the domain as defined by UNIPROT was individually aligned with both C2A and C2B of Syt1 using the EMBOSS NEEDLE protein alignment function (https://www.ebi.ac.uk/Tools/psa/emboss_needle/) and clustered with either C2A or C2B based upon similarity scores. Each set was then aligned using ClustalOmega.

Acetylation data were obtained from Phosphosite.org⁶⁸. We generated a script to assign those sites within the defined MIR or NBR. Acetylated lysines in the MIR or NBR were tallied and normalized to the total number of amino acids or lysines in the respective regions. The code for these processes are available in our ^{supplementary material}. As a control to test the effects of our MIR assignments, we expanded and contracted the MIR definitions by two residues and repeated our analysis.