Proteome-wide identification of the endogenous ADP-ribosylome of mammalian cells and tissue

Abstract

Although protein ADP-ribosylation is involved in diverse biological processes, it has remained a challenge to identify ADP-ribose acceptor sites. Here, we present an experimental workflow for sensitive and unbiased analysis of endogenous ADP-ribosylation sites, capable of detecting more than 900 modification sites in mammalian cells and mouse liver. In cells, we demonstrate that Lys residues, besides Glu, Asp and Arg residues, are the dominant in vivo targets of ADP-ribosylation during oxidative stress. In normal liver tissue, we find Arg residues to be the predominant modification site. The cellular distribution and biological processes that involve ADP-ribosylated proteins are different in cultured cells and liver tissue, in the latter of which the majority of sites were found to be in cytosolic and mitochondrial protein networks primarily associated with metabolism. Collectively, we describe a robust methodology for the assessment of the role of ADP-ribosylation and ADP-ribosyltransferases in physiological and pathological states.

Figure 1. Proteome-wide identification of endogenous ADP-ribosylation sites in human cell culture.

(a) Schematic representation of the peptide-based enrichment strategy. HeLa cells were treated with genotoxic stress and digested into peptides. Tryptic digested peptides were treated with PARG enzyme to convert multimeric ADP-ribosylation into monomeric counterparts, and subsequently ADP-ribosylated peptides were enriched using GST-Af1521 macrodomain. Enriched peptides were analysed by high-resolution LC-MS/MS on an Orbitrap Q-Exactive HF instrument and the data was further processed by bioinformatic software tools. (b) Comparison of HeLa cells exposed with 500 μM H₂O₂. During cellular knock-down of PARG enzyme (siPARG) an abundant PAR signal is observed, while under physiological conditions the PAR signal is significantly weaker (compare second lane with fourth lane on gel). Previous methods for characterizing ADP-ribosylation solely worked under siPARG conditions while the methodology described here is applicable to physiological conditions. (c) Optimization of the incubation time and amount of PARG enzyme required for converting multimeric ADP-ribosylated peptides into monomeric counterparts. (d) Validation experiment that confirms PARG treatment does not remove MAR from investigated peptides. (e) Venn diagram of identified ADP-ribosylation sites identified in two biological replicate analyses. A strong overlap in identified sites signifies high reproducibility in the developed method. (f) GO functional annotation of significantly regulated proteins in the combined data set reveal strong enrichment of proteins involved in DNA repair processes compared with annotated GO genes across the entire human genome (indicated P-values<0.005). (g) GO term annotation enrichment for cellular distribution of proteins harbouring ADP-ribosylation sites.

Figure 2. Lysine residues are in vivo targets of ADP-ribosylation in human cells.

(a) Distribution of ADP-ribosylation acceptor sites compared with their distribution in the genome. (b) Assessment of peptide glycation by free ADP-ribose. Distribution of log2 transformed SILAC ratios, and ADP-ribosylation acceptor sites, from forward and reverse SILAC experiments as outlined in Supplementary Fig. 2a. No increased SILAC ratios were observed for the different acceptor sites when cells were treated with PAR, supporting the notion, that the observed modifications are not derived from glycation. (c) In vitro PARylation of identified protein targets. Purified full-length human ARTD1 was incubated with recombinantly expressed proteins in the presence of ³²P-NAD⁺ and double-stranded DNA oligomer. Samples were resolved by SDS–PAGE, stained with Coomassie (CB; lower panel) and ³²P-incorporation was detected by autoradiography (P32; upper panel). (d) HeLa cells stably expressing SRSF1, SRSF2 CBX4 or TAF15 as GFP-fusion proteins were treated with H₂O₂ and PARP (that is, ADP-ribosylation) inhibitor olaparib as control experiment. Lysates were subjected to Af1521 WT pull-down or GFP-immunoprecipitation and subsequently analysed by immunoblotting with GFP antibody. Error bars are 95% confidence intervals with n=4.

Figure 3. SILAC ratios of ADP-ribosylation sites.

(a) Boxplot analysis of logarithmized H/L SILAC ratios from six SILAC experiments representing HeLa cells treated with H₂O₂ in a temporal manner (see Supplementary Fig. 2a). Strongest regulation of ADP-ribosylation sites is observed when cells are treated for 5–10 min of genotoxic stress. (b) Densiometric evaluation of IF analysis of HeLa cells treated with H₂O₂ for different time points. The strongest abundance in PAR signal is observed after 5–10 min treatment of H₂O₂, in good correlation with observed increase in SILAC ratios on MS analysis (Fig. 2a). Experiments were performed in triplicates. (c) Venn diagram depicts overlap between identified ADP-ribosylated proteins compared with previously reported Asp and Glu ADP-ribosylated proteins. (d) Venn diagram depicts overlap between identified ADP-ribosylated proteins compared with previously reported Asp and Glu ADP-ribosylated proteins using an in vitro strategy. (e) Comparison of identified ARTD1 ADP-ribosylation sites across different experiments as indicated. (f) Multi-scatter plot of measured peptide signal intensities from triplicate ADP-ribosylation experiments. A strong Pearson correlation signifies high reproducibility in the measured abundance of ADP-ribosylated peptide species. (g) Abundance measurement for seven ADP-ribosylation sites, demonstrating that lysine residue K498 is abundantly modified in ARTD1. Error bars are 95% confidence intervals with n=3.

Figure 4. Proteome-wide identification of endogenous ADP-ribosylation sites in mammalian tissue.

(a) Experimental setup for mammalian liver tissue analysis. In total, liver samples derived from three mice were investigated and prepared as indicated. Mouse images were adapted from the Servier Image Bank under the Creative Commons licences CC-BY. (b) Venn diagram of identified ADP-ribosylation sites from three liver samples. A strong overlap between analysed sampled signifies good reproducibility in the identified ADP-ribosylation sites across investigated tissue samples. (c) Distribution of ADP-ribosylation sites across proteins. (d) Distribution of ADP-ribosylated amino acids. The majority (86%) of identified sites reside on arginine residues indicative of differential ART activity compared with cell culture analysis (Fig. 1g). (e) GO term annotation enrichment for cellular distribution of proteins identified in liver samples harbouring ADP-ribosylation sites.