The "adductome": A limited repertoire of adducted proteins in human cells

Abstract

Proteins form adducts with nucleic acids in a variety of contexts, and these adducts may be cytotoxic if not repaired. Here we apply a proteomic approach to identification of proteins adducted to DNA or RNA in normally proliferating cells. This approach combines RADAR fractionation of proteins covalently bound to nucleic acids with quantitative mass spectrometry (MS). We demonstrate that "RADAR-MS" can quantify induction of TOP1- or TOP2-DNA adducts in cells treated with topotecan or etoposide, respectively, and also identify intermediates in physiological adduct repair. We validate RADAR-MS for discovery of previously unknown adducts by determining the repertoires of adducted proteins in two different normally proliferating human cell lines, CCRF-CEM T cells and GM639 fibroblasts. These repertoires are significantly similar with one another and exhibit robust correlations in their quantitative profiles (Spearman r = 0.52). A very similar repertoire is identified by the classical approach of CsCl buoyant density gradient centrifugation. We find that in normally proliferating human cells, the repertoire of adducted proteins - the "adductome" - is comprised of a limited number of proteins belonging to specific functional groups, and that it is greatly enriched for histones, HMG proteins and proteins involved in RNA splicing. Treatment with low concentrations of formaldehyde caused little change in the composition of the repertoire of adducted proteins, suggesting that reactive aldehydes generated by ongoing metabolic processes may contribute to protein adduction in normally proliferating cells. The identification of an endogenous adductome highlights the importance of adduct repair in maintaining genomic structure and the potential for deficiencies in adduct repair to contribute to cancer.

Keywords: Aldehyde dehydrogenase (ADH); Crosslink; DNA binding protein; Etoposide; Formaldehyde; HMG protein; Histone; Mass spectrometry; Proteomics; RADAR; RNA binding protein; Topoisomerase; Topotecan.

Fig. 1.. RADAR-SILAC analysis identifies adducts formed by TOP1 and TOP2.

(A) Schematic of fractionation: Cells are lysed in chaotropic salts and detergent, then nucleic acids and adducted proteins precipitated with ethanol or isopropanol. (B) RADAR-SILAC analysis of CCRF-CEM cells treated with the TOP1 poison, topotecan. Proteins enriched in inhibitor-treated cells are shown in red. (C) RADAR-SILAC analysis of CCRF-CEM CCRF-CEM cells treated the TOP2 poison, etoposide. Proteins enriched in inhibitor-treated cells are shown in red. (D) Above, diagram of TOP1. Tryptic peptides used to quantify recovery from the N-terminal (residues 205–216, 224–239, 252–271 and 300–310,) and C-terminal (residues 643–750, 693–700, 701–712 and 736–742) are highlighted in yellow and green, respectively. Catalytic tyrosine Y723 indicated in red. Below, ratios of recovery of the indicated peptides from the C- and N-terminal regions of TOP1, based on intensity. Each dot represents relative recovery in one of 28 different experiments reported in the MaxQB database (whole cell), and squares represent relative recovery in one of 11 different MS analyses of RADAR-fractionated cells. The Mann-Whitney test was used to determine p value.

Fig. 2.. RADAR-MS identifies a limited set of adducted proteins in normally proliferating human cell lines.

(A) Venn diagram of shared and unshared proteins in two untreated samples of CCRF-CEM T cells, NTa and NTb, containing 1449 and 1511 proteins, respectively; and a combined repertoire of 1664 non-overlapping proteins. Repertoire size and percent repertoire indicated in parentheses. (B) Pairwise comparison of the IBAQ intensities of the 1296 proteins shared between untreated samples NTa and NTb (Spearman r=0.93; p<0.0001). (C) Venn diagram comparing repertoires of adducted proteins in untreated CCRF-CEM T cells (unshaded) and GM639 fibroblasts (shaded) as determined by RADAR fractionation. Repertoire size and percent of the GM639 repertoire indicated in parentheses. Hypergeometric p=3.6e-227 was calculated based on the following parameters: population size=9000 ([21]; see Methods); sample size A=1296; sample size B=518; set=384; expected successes=75; observed/expected: 175/38; enrichment=5-fold. (D) Pairwise comparison of the intensities of the RADAR shared repertoire of CCRF-CEM T cells and GM639 fibroblasts (Spearman r=0.52; p<0.0001).

Fig. 3.. RADAR fractionation enriches RNA binding proteins.

(A) Pairwise comparison of the IBAQ intensities of the 342 protein pairs common to the CCRF-CEM whole cell proteome and the RADAR shared repertoire (Spearman r=0.029; p=0.595). (B) Venn diagram illustrating overlap of repertoires of adducted proteins as determined by RADAR fractionation (unshaded) and CsCl buoyant density gradient fractionation (shaded). Repertoire size indicated in parentheses. Hypergeometric p=2.3e-41 was calculated based on the following parameters: population size=6282 sample size A=384; sample size B=96; set=54; expected successes=5.9; observed/expected: 54/5.9; enrichment=9.2-fold. (C) Pie chart indicating the fraction of proteins classified as DBP, RBP or dual binders in the repertoires of the CCRF-CEM whole cell proteome and the RADAR shared repertoire. (D) Pie chart indicating the fraction of proteins classified as DBP, RBP or dual binders in the repertoire determined by buoyant density gradient centrifugation.

Fig. 4.. Formaldehyde treatment has little effect on the composition of the repertoire of adducted proteins.

(A) Venn diagram of shared and private (unshared) proteins among the 100 proteins with greatest response slopes in Groups A and B. Hypergeometric p=1.8e-15 was calculated based on the following parameters: population size=625 (all proteins with positive intensities); sample size A=100; sample size B=100; set=46; expected successes=16; observed/expected: 46/16; enrichment=2.9-fold. (B) Fractions of MS signal contributed by the 46 shared proteins ranked in the top 100 based upon response slopes. (C) Quantitation of contribution to signal of four proteins/classes of proteins that contributed to 90% of the increased signal at the highest dose of formaldehyde.

Fig. 5.. Proteins enriched in the repertoire of adducted proteins.

(A) Frequency distribution of IBAQ scores in the RADAR shared repertoire of CCRF-CEM T cells and GM639 fibroblasts. (B) Pie chart indicating the fraction of MS signal contributed by proteins classified as DBP, RBP or dual binders in the CCRF-CEM whole cell proteome and the RADAR shared repertoire. (A pie chart based on repertoire rather than MS signal is presented in Fig. 3C.) (C) The fifty most abundant proteins in the RADAR repertoire of unperturbed CCRF-CEM T cells which account for over 70% of the total MS signal. Orange diamonds mark species shared with 30 topmost proteins of RNA-crosslinked proteome of MCF7 cells recovered after XRNAX fractionation (see discussion).

Fig. 6.. Pathway analysis of adducted proteins.

(A) Highly significant functional enrichments in the RADAR shared repertoire based on analysis of protein-protein interaction networks (STRING database). (B) Pie chart indicating the fraction of MS signal contributed by proteins associated with the GO terms RNA splicing and translation in the CCRF-CEM whole cell proteome and the RADAR shared repertoire.