Passenger hotspot mutations in cancer driven by APOBEC3A and mesoscale genomic features

Abstract

Cancer drivers require statistical modeling to distinguish them from passenger events, which accumulate during tumorigenesis but provide no fitness advantage to cancer cells. The discovery of driver genes and mutations relies on the assumption that exact positional recurrence is unlikely by chance; thus, the precise sharing of mutations across patients identifies drivers. Examining the mutation landscape in cancer genomes, we found that many recurrent cancer mutations previously designated as drivers are likely passengers. Our integrated bioinformatic and biochemical analyses revealed that these passenger hotspot mutations arise from the preference of APOBEC3A, a cytidine deaminase, for DNA stem-loops. Conversely, recurrent APOBEC-signature mutations not in stem-loops are enriched in well-characterized driver genes and may predict new drivers. This demonstrates that mesoscale genomic features need to be integrated into computational models aimed at identifying mutations linked to diseases.

Fig. 1.. The mutational background in cancer shows variation at all scales.

(A) At the large scale, the genome is organized into multi-megabase chromatin domains roughly corresponding to nuclear compartment A (gene-rich, highly transcribed, early-replicating, low-mutation frequency) versus compartment B (gene-poor, repeat-rich, low-transcription, late-replicating, high-mutation frequency). Cancer driver genes tend to occupy compartment A, whereas frequently mutated genes in compartment B are more likely to be passengers. (B) Mutations are generally enriched in late-replicating DNA, whereas APOBEC mutation frequency is unresponsive to large-scale covariates. Error bars denote 95% confidence intervals. (C) At the mesoscale, local DNA secondary structures can also influence mutation frequency; for example, nucleotides exposed in the loop of a DNA hairpin may suffer increased vulnerability to mutagens. (D) For each mutational signature, relative mutation frequency in hairpin loops is shown as a function of how strongly the hairpin is base-paired. APOBEC-associated mutation frequency increases markedly at nucleotides exposed in stable hairpin loops. In contrast, other mutational signatures show little effect of hairpins. Error bars denote 95% confidence intervals. (E) At the small scale, mutational signatures have been defined by the relative abundance of mutations across trinucleotide contexts. (F) Mutational signatures reflect the action of mutagens such as APOBEC activity, UV irradiation, loss of polymerase proofreading (POLE), or failed mismatched repair (MSI). Each signature is displayed using two equivalent visual conventions: The upper rectangles show “fingerprint plots” (4, 5), whereas the lower three-dimensional bar plots show “Lego plots” (1), in which the APOBEC mutational signature can be seen at a glance to occupy the “back-row” TpC motif.

Fig. 2.. Enzymatic activity of APOBEC3A is enhanced in the loop of a DNA hairpin.

(A) In vitro assay for APOBEC cytidine deamination activity. A labeled DNA substrate is incubated with APOBEC enzyme, which can deaminate C to U. Subsequent activity of the enzyme uracil-DNA glycosylase (UNG) generates an abasic site that can be cleaved by heating. (B) Naturally occurring DNA hairpin in the gene NUP93, and a modified version (“NUP93-noHP”) in which one side of the hairpin has been replaced with PolyA, abrogating the secondary structure. (C) Time course of APOBEC3A deamination activity on the NUP93 and NUP93-noHP substrates. APOBEC3A shows a preference for the hairpin-containing substrate. Error bars denote SD. (D) Comparison of deamination activity across three APOBEC3 family members. Whereas APOBEC3A (A3A) prefers the hairpin-containing substrate, APOBEC3B (A3B) shows no such preference, and APOBEC3H haplotype I (A3H-I) shows preference for nonhairpins.

Fig. 3.. Data from in vitro enzymology and tumor bioinformatics converge to illustrate the substrate preferences of APOBEC3A.

(A) A series of NUP93 hairpin substrate analogs in which the base-paired stem is progressively weakened shows that APOBEC3A (A3A) activity depends on the presence of a DNA hairpin. (B) Analysis of APOBEC-dominated tumors sequenced by WGS recapitulates the pattern observed with synthetic substrates: TpC sites exposed in a DNA hairpin loop are mutated at higher frequencies, increasing as the stem base-pairing becomes stronger. Furthermore, the position of the C in the loop affects mutability, with the highest frequencies observed for C’s at the 3′-most position in the loop. No increase in frequency was observed for C’s in the 5′ half of the loop. (C) The naturally occurring NUP93 hairpin site has a 4-nt loop and is an efficient substrate in the in vitro A3A activity assay. Increasing the size of the loop decreases the observed rate of A3A activity. Conversely, a substrate with a 3-nt loop (“NUP93-L3”) shows even stronger activity than the natural hairpin. (D) Statistics from human APOBEC⁺ tumors confirm this trend: Hairpins with 3-nt loops show the highest mutation frequency, and the increase is observed only when the C is at the 3′-most position of the loop (in other words, the Tof the TpC is centered in the loop). Hairpins with longer loops show decreasing APOBEC mutation frequency. Hairpins with TpC centered in the loop show the strongest increase. (E) The precise sequence context around the TpC site can drastically affect substrate fitness for A3A activity. A version of the NUP93 substrate with the TpC moved to the middle of the loop shows lower activity in the in vitro A3A activity assay. However, a single-nucleotide change from T to C (blue arrowheads) restores activity. (F) These findings are mirrored by statistical trends in APOBEC⁺ tumors. For each type of hairpin loop, mutation frequency is influenced by the identity of the other nucleotide(s) in the loop and the stem’s closing base pair. The globally optimal DNA substrate for APOBEC3A is a strongly base-paired hairpin with the 3-nt loop TTC and a C-G closing base pair; in APOBEC⁺ human cancers, these sites are mutated ~200 times as frequently as nonhairpin TpC sites. Error bars in (B), (D), and (F) denote 95% confidence intervals.

Fig. 4.. Mesoscale APOBEC mutational background features allow cancer drivers to be distinguished from passengers.

Shown are the top 100 most frequently mutated coding TpC hotspots in a set of 2572 APOBEC⁺ human cancers sequenced by WXS. Each dot is a hotspot: y-axis position indicates the number of patients mutated at that particular base pair, and x-axis position indicates the substrate optimality, expressed as the relative mutation frequency predicted from the position’s mesoscale features, namely the ability to fold into a hairpin that ideally exposes the mutated position in a short loop, as explored in Fig. 3. Hotspots in known cancer genes (34,35) are colored blue. Hotspots in other genes are colored either red (<4) or gray (>4) by substrate optimality. Hotspots fall into two clear groups: Hotspots at APOBEC3A’s optimal substrates (gray) tend to be in genes unconnected to cancer; many of these are likely to be passenger hotspots. In contrast, hotspots at sites that are not APOBEC3A optimal substrates tend to be in known cancer driver genes. The exceptions (red) represent potential novel drivers. Conversely, MB21D2, currently annotated as a known cancer gene on the strength of a single highly recurrent mutational hotspot, may not play any role in driving cancer, instead merely harboring an APOBEC3A-driven passenger hotspot.