Evolution-based screening enables genome-wide prioritization and discovery of DNA repair genes

Abstract

DNA repair is critical for genome stability and is maintained through conserved pathways. Traditional genome-wide mammalian screens are both expensive and laborious. However, computational approaches circumvent these limitations and are a powerful tool to identify new DNA repair factors. By analyzing the evolutionary relationships between genes in the major DNA repair pathways, we uncovered functional relationships between individual genes and identified partners. Here we ranked 17,487 mammalian genes for coevolution with 6 distinct DNA repair pathways. Direct comparison to genetic screens for homologous recombination or Fanconi anemia factors indicates that our evolution-based screen is comparable, if not superior, to traditional screening approaches. Demonstrating the utility of our strategy, we identify a role for the DNA damage-induced apoptosis suppressor (DDIAS) gene in double-strand break repair based on its coevolution with homologous recombination. DDIAS knockdown results in DNA double-strand breaks, indicated by ATM kinase activation and 53BP1 foci induction. Additionally, DDIAS-depleted cells are deficient for homologous recombination. Our results reveal that evolutionary analysis is a powerful tool to uncover novel factors and functional relationships in DNA repair.

Keywords: DDIAS; DNA repair; evolution; genome integrity; homologous recombination.

Fig. 1.

Genetic constituents of most DNA repair pathways exhibit evolutionary rate covariation. (A) Seven mammalian DNA repair pathways are responsible for repairing DNA lesions. In pathway 1, SSBs (indicated by a missing nucleotide), abasic sites, and alkylation damage are repaired by the BER pathway. In pathway 2, bulky adducts (indicated by the red star) are repaired by NER. In pathways 3 to 5, DSBs (indicated by the 2 nicks) are repaired using HR, NHEJ, and meiosis. In pathway 6, replication errors (indicated by the red nucleotide mismatches) are repaired using MMR. In pathway 7, interstrand crosslinks (indicated by the red line) are repaired by the FA pathway. (B) The genes mediating the 7 mammalian DNA repair pathways were tested for elevated ERC. Statistical significance was determined by permutation test, where the P value is the computed probability of the observed mean ERC or greater from 1,000 equally sized groups of randomly sampled genes. n is the number of genes in the indicated group (the gene list is found in SI Appendix, Table S1).

Fig. 2.

Evolutionary rate covariation between HR genes. The ERC values between HR gene pairs in each DNA processing step were calculated and plotted using a heat map (ranging from 0 [no covariation] to 1 [positive covariation]). Genes are hierarchically clustered.

Fig. 3.

Genome-wide ranking of 17,487 genes for coevolution with HR and FA enriches for known HR and FA factors when compared to functional screens. (A) Genes were ranked by mean ERC with HR. Each bin size contains 874 genes (5% of genes analyzed). The number of HR genes in each bin is plotted relative to the total number of HR genes (n = 39) (blue bar). A scan statistic was used to find an enriched window of known HR genes, and significance was determined by permutation test (1,000 nulls; ***P < 0.001, 0 to 20%). Note that only HR genes present in both screens (the ERC dataset and the screen described in B) were included. (B) Genes were ranked by relative drGFP in a published siRNA screen (n = 21,121) for HR factors (19). Each bin contains 1,056 genes (5% of genes analyzed). The number of HR genes in each bin is plotted relative to the total number of HR genes (n = 39) (blue bar). A scan statistic was used to find an enriched window of known HR genes, and significance was determined by permutation test (1,000 nulls; ***P < 0.001, 0 to 5%). (C) Genes were ranked by mean ERC with FA. Each bin size contains 874 genes (5% of genes analyzed). The number of FA genes in each bin is plotted relative to the total number of FA genes (n = 20) (pink bar). A scan statistic was used to find an enriched window of known FA genes, and significance was determined by permutation test (1,000 nulls; ***P < 0.001, 0 to 10%). Note that only FA genes present in both screens (the ERC dataset and the screen described in D) were included. (D) Genes were ranked by MMC sensitivity in a published shRNA screen (n = 32,293) for FA factors (20). Each bin contains 2,464 targets (5% of targets analyzed). The number of FA genes in each bin is plotted relative to the total number of FA genes (n = 20) (pink bar). A scan statistic was used to find an enriched window of known FA genes, and significance was determined by permutation test (1,000 nulls; ***P < 0.001, 0 to 25%).

Fig. 4.

DDIAS depletion results in defective DSB repair. (A) Violin plots show the distribution of ERC values between DDIAS and major DNA repair pathways. Overlaid box plots indicate the quartiles of each distribution, and a white dot represents the median. Permutation P values are listed reflecting the significance of DDIAS’s ERC elevation with each pathway. (B) siDDIAS-U2OS cells were Western blotted for DDIAS, pATM-S1981, ATM, pCHK1-S345, CHK1, and alpha tubulin (Left). pATM signal was quantified and normalized to alpha tubulin signal. The mean of 3 experiments is plotted with SEM (Right; *P < 0.05). (C) The 53BP1 foci were quantified in siDDIAS-U2OS cells. The average number of 53BP1 foci (Bottom Left) and percentage of cells ≥5 53BP1 foci (Bottom Right) were quantified from 2 independent experiments (n = 200 cells per condition). Means are plotted with SEM graphed (*P < 0.05; ***P < 0.001). (D) DDIAS depletion using 2 independent siRNAs results in reduced HR in U2OS-SCR cells. Schematic of the SCR-GFP reporter system is shown (Top). U2OS-SCR-GFP cells were treated with siRNA targeting DDIAS (siDDIAS 1 and 2; see SI Appendix, Fig. S9, for corresponding Western blot showing DDIAS depletion). Untransfected cells and I-SceI–transfected cells treated with mirin were also measured. Mean %GFP+ is plotted with SE. (***P < 0.001.) n.s., not significant.