Single-nucleotide-resolution genomic maps of O6-methylguanine from the glioblastoma drug temozolomide

Abstract

Temozolomide kills cancer cells by forming O6-methylguanine (O6-MeG), which leads to cell cycle arrest and apoptosis. However, O6-MeG repair by O6-methylguanine-DNA methyltransferase (MGMT) contributes to drug resistance. Characterizing genomic profiles of O6-MeG could elucidate how O6-MeG accumulation is influenced by repair, but there are no methods to map genomic locations of O6-MeG. Here, we developed an immunoprecipitation- and polymerase-stalling-based method, termed O6-MeG-seq, to locate O6-MeG across the whole genome at single-nucleotide resolution. We analyzed O6-MeG formation and repair across sequence contexts and functional genomic regions in relation to MGMT expression in a glioblastoma-derived cell line. O6-MeG signatures were highly similar to mutational signatures from patients previously treated with temozolomide. Furthermore, MGMT did not preferentially repair O6-MeG with respect to sequence context, chromatin state or gene expression level, however, may protect oncogenes from mutations. Finally, we found an MGMT-independent strand bias in O6-MeG accumulation in highly expressed genes. These data provide high resolution insight on how O6-MeG formation and repair are impacted by genome structure and nucleotide sequence. Further, O6-MeG-seq is expected to enable future studies of DNA modification signatures as diagnostic markers for addressing drug resistance and preventing secondary cancers.

Graphical Abstract

Figure 1.

O ⁶ -MeG is induced in distinct trinucleotide patterns by TMZ. (A) Strategy for O⁶-MeG-seq. DNA fragments containing O⁶-MeG are pulled down with an O⁶-MeG specific antibody. The exact site is marked with SuperFi II polymerase stalling at O⁶-MeG. Illumina library preparation and sequencing are used to map O⁶-MeG at single-nucleotide resolution. (B) Base contribution at modification site. Three biological replicates of TMZ-exposed naked DNA, and TMZ-exposed and solvent control (DMSO) LN-229 cells (WT or +MGMT). (C) Position information across 10 bases upstream and downstream of modification sites in three times 1 mM TMZ-exposed LN-229 cells and TMZ-exposed naked DNA (replicate with the highest sequencing depth). The relative heights of the letters corresponding to bases indicate their relative abundance at that site, while the height of the entire stack of letters reflects deviation from randomness at this position with a maximum of two bits. (D) Trinucleotide context frequencies of O⁶-MeG where G is at the modification site. Left: TMZ-exposed LN-229 cells exposed three times to 1 mM TMZ, right: LN-229 WT and +MGMT naked DNA exposed to 1 mM TMZ. Trinucleotide patterns of all exposure conditions are shown in Supplementary Figure S7.

Figure 2.

O ⁶-MeG as precursor of TMZ-related mutational signatures. (A) Extracted O⁶-MeG signatures, termed signature A and B. Signatures were extracted from O⁶-MeG trinucleotide patterns of TMZ-exposed and solvent control LN-229 WT and +MGMT cells using non-negative matrix factorization. (B) Relative contribution of trinucleotide patterns to the extracted signatures A and B. Three biological replicates per exposure condition. (C) Cosine similarities of all COSMIC SBS compared to O⁶-MeG signatures A and B. Conversion of O⁶-MeG signatures to mutational signatures considers reverse complementary trinucleotide contexts of G converted into C to T mutations and assumes no signal for other SBSs. Cosine similarity of 0.9 was used as cut off for high similarity (dashed lines). (D) C to T mutation contexts of O⁶-MeG signature A and COSMIC SBS 11 and 23.

Figure 3.

MGMT does not appear to influence O⁶-MeG distribution in the human genome. Genome-wide distribution of O⁶-MeG in three times 1 mM TMZ-exposed LN-229 WT and +MGMT cells normalized by G-only read depth and genomic G abundance (A and B) and by O⁶-MeG abundance of TMZ-exposed naked DNA (C and D). Spearman correlation of replicates was high (>0.75) for all exposure conditions used in this analysis (Supplementary Figure S12). (A and C) Whole-genome view shows the average O⁶-MeG abundance of three biological replicates per 100 kb bin. Y-axis ranges were capped at the 99th and 1st percentile. (B andD) O⁶-MeG distribution in chromosomes 20 and X of LN-229 WT and +MGMT cells. Faded bands show standard deviation of replicates and centromeric areas were marked by grey background. Y-axis ranges were capped at the 99.9th and 0.1st percentile.

Figure 4.

O ⁶-MeG has an MGMT-independent strand bias towards the non-transcribed strand in expressed genes. (A–H) Gene-specific analysis of O⁶-MeG distribution in three times 1 mM TMZ-exposed LN-229 WT (A, C, E and G) and LN-229 +MGMT (B, D, F and H) cells with respect to gene expression. O⁶-MeG abundance was normalized by gene G abundance (A, B, C and D) and corrected by O⁶-MeG abundance of TMZ-exposed naked DNA (E, F, G and H). (A, B, E and F) O⁶-MeG abundance compared to gene expression. Left panels show O⁶-MeG abundance in gene expression tiers in one replicate. The asterisks indicate the average abundance. Right panels show the means of O⁶-MeG abundance in gene expression tiers of three replicates. (C,D,G andH) Gene-body profiles of O⁶-MeG abundance. Means and 95% confidence interval of three replicates are shown. TSS: transcription start site. TES: transcription end site. For gene bodies, 5% of the gene length was used as bin size (2.2 kb average) while 2.5 kb bin size was used outside the gene body. (I and J) O⁶-MeG abundance per gene was normalized by read depth and G abundance and replicates were averaged. Left panels show LN-229 WT and +MGMT while right panels show the difference of LN-229 +MGMT and WT. Analysis was done for all genes (I) or oncogenes only (J).