Enhanced polymerase activity permits efficient synthesis by cancer-associated DNA polymerase ϵ variants at low dNTP levels

Abstract

Amino acid substitutions in the exonuclease domain of DNA polymerase ϵ (Polϵ) cause ultramutated tumors. Studies in model organisms suggested pathogenic mechanisms distinct from a simple loss of exonuclease. These mechanisms remain unclear for most recurrent Polϵ mutations. Particularly, the highly prevalent V411L variant remained a long-standing puzzle with no detectable mutator effect in yeast despite the unequivocal association with ultramutation in cancers. Using purified four-subunit yeast Polϵ, we assessed the consequences of substitutions mimicking human V411L, S459F, F367S, L424V and D275V. While the effects on exonuclease activity vary widely, all common cancer-associated variants have increased DNA polymerase activity. Notably, the analog of Polϵ-V411L is among the strongest polymerases, and structural analysis suggests defective polymerase-to-exonuclease site switching. We further show that the V411L analog produces a robust mutator phenotype in strains that lack mismatch repair, indicating a high rate of replication errors. Lastly, unlike wild-type and exonuclease-dead Polϵ, hyperactive variants efficiently synthesize DNA at low dNTP concentrations. We propose that this characteristic could promote cancer cell survival and preferential participation of mutator polymerases in replication during metabolic stress. Our results support the notion that polymerase fitness, rather than low fidelity alone, is an important determinant of variant pathogenicity.

Figure 1.

Characteristics of cancer-associated Polϵ variants used for biochemical studies. (A) Schematic of POLE showing the location of variants studied in this work with the alignment of amino acid sequences of human POLE and yeast Pol2 surrounding the mutation sites shown below. All variants occur at highly conserved residues in the exonuclease domain. (B) The mutator effect of each variant modeled in S. cerevisiae is shown above the x-axis and incidence per 10 000 tumors is shown below. Mutation rate data are from (19) and (20). POLE variant frequency was calculated from published studies (>20 000 tumors; Supplementary Table S2).

Figure 2.

Exonuclease defects of cancer-associated Polϵ variants. Exonuclease activity was assessed using 25 nM P50 single-stranded (A) and P50/T80 partially double-stranded (B) oligonucleotide substrates and 3.125 nM Polϵ. Oligonucleotide sequences are shown above each gel. The fraction of 50-mer remaining is quantified below each lane. The gels shown are representative of three independent experiments. Cancer-associated Polϵ variants are in order of decreasing mutator effect starting with Polϵ-P301R as observed in (20).

Figure 3.

Increased DNA polymerase activity of cancer-associated Polϵ variants. (A) DNA polymerase activity was analyzed using 25 nM P50/T80a4T oligonucleotide substrate and 1 nM Polϵ. The oligonucleotide substrate sequence is shown above the gel. The gel is representative of three independent experiments. (B) The fraction of products greater than 78 nt was quantified (averages of three experiments). Error bars denote standard error. Asterisks indicate P < 0.05 as determined by one-way ANOVA with post-hoc Tukey test. (C) Lack of correlation between exonuclease and polymerase activities of Polϵ variants. Percent primer degraded at 10 min incubation (from Figure 2B) is plotted on x-axis, and percent full-length product at 0.2 min incubation (from panel A) is plotted on y-axis.

Figure 4.

Mismatch processing by Polϵ variants. 3.125 nM Polϵ was incubated with 25 nM P51T/T80 containing a G-T mismatch at the 3′ primer terminus for 30 min at 30°C in the presence of dNTP concentrations corresponding to S-phase (left) and checkpoint-induced (right) levels. The checkpoint dNTPs were as estimated in (33) (114 μM dCTP, 266 μM dTTP, 171 μM dATP, 91 μM dGTP). Products were digested with BsaJI to determine the fraction of mismatch correction vs. extension. The presence of a 51-nt band after restriction digest indicates mismatch correction; products >62-nt remaining after BsaJI digest indicate mismatch extension. The fraction of products > 62-nt in BsaJI-digested samples is shown below the gel. Area corresponding to 52- to 62-nt-long products was excluded from quantification, because these products represent termination of synthesis before the BsaJI restriction site sequence and the adjacent nucleotides are copied. Representative images of three independent experiments are shown. The percentage of mismatch extension events for cancer variants relative to the percentage of mismatch extension events in WT Polϵ reactions is plotted below each gel.

Figure 5.

Bypass of hairpin DNA structures by Polϵ variants. (A) DNA polymerase activity was assayed using 2 nM Polϵ and 25 nM P50/T80H oligonucleotide substrate containing a 6-bp inverted repeat in the template, shown above the gel. A representative gel image from four independent experiments is shown. Hairpin bypass was quantified as the fraction of products greater than 71 nt and averages from four independent experiments are shown. Error bars denote standard error. Asterisks indicate P < 0.05 as determined by one-way ANOVA with post-hoc Tukey test. (B) DNA polymerase activity was assayed with 3.125 nM Polϵ and 25 nM P51T/T80H oligonucleotide substrate containing a G-T mismatch at the primer terminus and a 6-bp inverted repeat in the template, shown above the gel. A representative image from three independent experiments is shown. The fraction of products greater than 71 nt was quantified. Averages from three independent experiments are shown. Error bars denote standard error. Asterisks indicate P < 0.05 as determined by one-way ANOVA with post-hoc Tukey test. (C) The correlation between the percent of primer degraded in the exonuclease assay on P50/T80 substrate at 10 min incubation (x-axis, from Figure 1) and percent hairpin bypassed on P51T/T80H at 10 min incubation (y-axis, from panel B). (D) The relative efficiency of mismatch extension versus proofreading on P51/T80H was determined by incubating the DNA substrate with the polymerases for 30 min and digesting the reaction products with BsaJI as in Figure 4.

Figure 6.

Synergistic interaction of pol2-V426L and MMR deficiency. (A and B) Rates of Can^r mutation and His⁺ reversion in haploid yeast strains carrying chromosomal pol2-V426L and/or mlh1Δ alleles. Data are medians and 95% confidence intervals from Supplementary Table S3. Fold increases over the wild-type strain are shown above the bars. (C) The spectrum of spontaneous can1 mutations in pol2-V426L mlh1Δ strain. Indels are shown in yellow and base substitutions in grey. The location of mutations in the CAN1 sequence is shown in Supplementary Figure S4.

Figure 7.

Insights from structures of Polϵ-V426L and Polϵ-L439V. (A) Surface representation of the finger and palm domain with the polymerase active site (light brown) and the exonuclease domain (pink) (PDB ID: 6fwk). DNA is shown in orange cartoon as it would bind to each active site. In the absence of a structure of Polϵ in the editing mode, the ssDNA was modelled into the exonuclease site of Pol2_CORE by superimposing the exonuclease domain of a euryarchael B family DNA polymerase with single-stranded DNA bound in the exonuclease site (PDB ID: 4flw (86)). Amino acids L439 and V426 are shown in red sticks to illustrate that they are located in the trajectory for the switch of the primer terminus between the polymerase and exonuclease sites. (B) The V426 and L439 are located in a helix-loop-helix motif (blue) that is positioned where the primer terminus is transferred between the polymerase and exonuclease site. Metals bound to the active sites are shown in green. (C) The crystal structure of ^V426LPol2_CORE (PDB ID: 7r3y) (pink) superimposed on the WT exonuclease domain of Pol2_CORE (PDB ID: 6fwk) (blue). Amino acids V426 and L426 (red sticks) are buried in a hydrophobic pocket between the two helices of the helix-loop-helix motif that is located in the trajectory of the primer-end during the switch between the polymerase and exonuclease sites. (D) The crystal structure of ^L439VPol2_CORE (PDB ID: 7r3x) (light blue) superimposed on the WT exonuclease domain of Pol2_CORE (PDB ID: 6fwk) (blue) shows that L439 reaches further into a hydrophobic pocket when compared to V439. (E) Alignment of the WT exonuclease domain of Pol2_CORE (PDB ID: 6fwk) and the exonuclease domain of ^L439VPol2_CORE (PDB ID: 7r3x) shows that valine is located at a less favorable distance from the DNA, suggesting that DNA may have a lower affinity for the exonuclease site of Polϵ-L439V. The DNA (blue) is positioned based on an energy minimized MD simulation (28).

Figure 8.

Efficient synthesis by cancer-associated Polϵ variants at reduced dNTP levels. (A) Reactions were performed with 25 nM P50/T80a4T oligonucleotide substrate and 1 nM Polϵ for 1 min. Fold changes in dNTP concentrations are indicated above the gel (the S-phase ratio of individual dNTPs was maintained in all reactions). A representative image from four independent experiments is shown. PR is Polϵ-P301R; SF is S474F; FS is F382S; LV is L439V; DV is D290V; and VL is V426L. (B) The fraction of products greater than 78-nt was quantified, and averages of four independent experiments were plotted. Error bars denote standard error. Asterisks indicate P < 0.05 compared to WT as determined by one-way ANOVA with post-hoc Tukey test.

Figure 9.

Summary of characteristics relating to the pathogenicity of Polϵ variants. Cancer-associated Polϵ variants have reduced proofreading and increased DNA polymerase activity, leading to decreased fidelity and a strong mutator effect. Hyperactive variants are also tolerant to reduced levels of dNTPs. Future studies will determine whether this characteristic can promote survival (green question mark) and mutagenesis (red question mark) during metabolic stress, with both pro-survival and pro-mutagenic roles contributing to tumorigenesis (orange question mark). Exo, exonuclease domain; Polϵ, polymerase domain.