Discovery of lipid-mediated protein-protein interactions in living cells using metabolic labeling with photoactivatable clickable probes

Abstract

Protein-protein interactions (PPIs) are essential and pervasive regulatory elements in biology. Despite the development of a range of techniques to probe PPIs in living systems, there is a dearth of approaches to capture interactions driven by specific post-translational modifications (PTMs). Myristoylation is a lipid PTM added to more than 200 human proteins, where it may regulate membrane localization, stability or activity. Here we report the design and synthesis of a panel of novel photocrosslinkable and clickable myristic acid analog probes, and their characterization as efficient substrates for human N-myristoyltransferases NMT1 and NMT2, both biochemically and through X-ray crystallography. We demonstrate metabolic incorporation of probes to label NMT substrates in cell culture and in situ intracellular photoactivation to form a covalent crosslink between modified proteins and their interactors, capturing a snapshot of interactions in the presence of the lipid PTM. Proteomic analyses revealed both known and multiple novel interactors of a series of myristoylated proteins, including ferroptosis suppressor protein 1 (FSP1) and spliceosome-associated RNA helicase DDX46. The concept exemplified by these probes offers an efficient approach for exploring the PTM-specific interactome without the requirement for genetic modification, which may prove broadly applicable to other PTMs.

Fig. 1. X3, X8 and X10 probes label myristoylated proteins in cells. (A) Diagram showing cellular transformations leading to protein myristoylation. Initiator methionine aminopeptidase (MetAP) (co-translational myristoylation, shown) or an endoprotease (post-translational myristoylation) reveal an N-terminal glycine via proteolysis, which is then myristoylated by NMT. (B) Myristoylated proteome labeling strategy: diazirine probe is incubated with cells in culture and incorporated into the proteome, photocrosslinked by irradiation at 365 nm, and following lysis alkyne groups are functionalized with dye and/or biotin labels, optionally including a trypsin-digestible arginine linker. Proteins can be visualized by in-gel fluorescence or enriched on streptavidin/neutravidin beads for western blotting and proteomics. (C) Cells were treated with a panel of myristate analogs and visualized by in-gel fluorescence. A myristoylated protein labeling pattern is observed for all probes without UV-irradiation (UV), YnMyr (5 μM) and X3/X8/X10 (100 μM each), whilst X3/X8/X10 probes are crosslinked to cellular proteins upon UV irradiation. (D) IMP-366 (NMTi) treatment inhibits myristoylated proteome labeling by X10; upon UV-irradiation, X10 labeling is NMT-independent and is likely dominated by lipid–protein crosslinks. For the unadjusted contrast range see Fig. S1. (E) Cells were treated with a panel of myristate analogs, and labeled proteins enriched on streptavidin-conjugated beads. Western blots of NMT substrate PSMC1 before (input) and after enrichment (pull-down) indicate labeling with X3/X8/X10 (100 μM each) probes comparable to YnMyr (5 μM). Equivalent blotting for NMT1 labeling (a buried site) shows photocrosslinking-dependent enrichment only with denaturation at 95 °C pre-CuAAC ligation.

Scheme 1. Synthesis of photocrosslinkable clickable myristate probes. (A) Left, myristic acid (Myr). Right, YnMyr probe. (B–E) Synthesis schemes for X3 (B), X8 (C), X10 (D) and X10-CoA (E). Full synthesis and characterization details are provided in the ESI.

Fig. 2. Incorporation of photocrosslinkable clickable probes into the myristoylated proteome. (A) Experimental design for quantitative SILAC proteomics validating myristoylated proteome labeling by X3/X8/X10 (100 μM each) compared to positive control YnMyr (5 μM) and negative controls obtained by 0.5 μM NMTi treatment. Treated cells were lysed, ligated to an azido-arginine-biotin capture reagent and enriched on neutravidin-agarose resin. (B) Pearson correlation between myristoylated protein enrichment (green) using either photocrosslinkable clickable probes X3/X8/X10 or YnMyr. Grey – background (non-myristoylated) proteins. Axes are differences in protein enrichment between vehicle (DMSO) and NMTi-treated samples for X3/X8/X10 (X-axes) and YnMyr (Y-axis). (C) Structures and masses of PTMs searched to identify X3/X8/X10/YnMyr-related modifications. (D) Left – Total number of Myr′, Myr′-N2, and Myr-Xlink (see (C)) modified peptide-spectrum matches (PSMs, top) or proteins (bottom) identified using PEAKS search engine for X3/X8/X10 or YnMyr. Right – All PSMs or proteins detected in the experiment for a given probe.

Fig. 3. Myristate probes X3/X8/X10 form crosslinks with myristate-binding proteins. (A–C) Representative MS/MS fragmentation spectra and structural positions of crosslinked amino acids (red) for ACAT2 ((A), PDB:1WL4, orange – acetyl-CoA), HMOX2 ((B), PDB:5UC9, yellow – myristic acid) and HMOX1 ((C), PDB:1N45, brown – heme) structures.

Fig. 4. X10-CoA is an NMT1/2 substrate. (A) Rates of enzymatic transfer of Myr or X10 by HsNMT1 or HsNMT2 to a synthetic peptide measured in vitro by CPM assay; both natural and artificial substrates show similar catalytic efficiencies (V_max/K_M ± SD, RFU min⁻¹ μM⁻¹ × 10⁶) and Michaelis constants (K_M ± SD, μM); (B) crystal structure of HsNMT1 (orange) in complex with X10-CoA (green) showing the overall structure and an enlarged view of the X10 moiety (chain A) with a potential hydrogen bond between the diazirine and T268. Color code: blue (nitrogen), red (oxygen), yellow (sulfur), and orange (protein structure).

Fig. 5. Identification of PPIs by photocrosslinking. (A) Scheme of X10 photocrosslinking experiment to identify PPIs of TST-tagged myristoylated proteins. (B–D) Fold protein enrichment from cells transfected with mCitrine-TST (B), DDX46-TST (C) or FSP1-TST (D) in triplicates, relative to all other samples combined (19 proteins × 3 biological replicates) on the X-axis, plotted against statistical significance (Y-axis). (E and F) Validation of DDX46 (E) and FSP1 (F) interactors. Pulldowns for DDX46-TST and FSP1-TST were probed with TRA2B and TOMM40 antibodies, respectively. α-Tubulin or GAPDH were used as loading controls. Normalization was performed by dividing the TRA2B or TOMM40 antibody signal by the DDX46 or FSP1 signal, respectively. Data are shown as mean ± s.e.m., n = 3–4. ****p = 0.0001, ns – no significant difference, calculated by the Student's t-test (two-tailed, unpaired).