Nucleotide selectivity defect and mutator phenotype conferred by a colon cancer-associated DNA polymerase δ mutation in human cells

Abstract

Mutations in the POLD1 and POLE genes encoding DNA polymerases δ (Polδ) and ε (Polε) cause hereditary colorectal cancer (CRC) and have been found in many sporadic colorectal and endometrial tumors. Much attention has been focused on POLE exonuclease domain mutations, which occur frequently in hypermutated DNA mismatch repair (MMR)-proficient tumors and appear to be responsible for the bulk of genomic instability in these tumors. In contrast, somatic POLD1 mutations are seen less frequently and typically occur in MMR-deficient tumors. Their functional significance is often unclear. Here we demonstrate that expression of the cancer-associated POLD1-R689W allele is strongly mutagenic in human cells. The mutation rate increased synergistically when the POLD1-R689W expression was combined with a MMR defect, indicating that the mutator effect of POLD1-R689W results from a high rate of replication errors. Purified human Polδ-R689W has normal exonuclease activity, but the nucleotide selectivity of the enzyme is severely impaired, providing a mechanistic explanation for the increased mutation rate in the POLD1-R689W-expressing cells. The vast majority of mutations induced by the POLD1-R689W are GC→TA transversions and GC→AT transitions, with transversions showing a strong strand bias and a remarkable preference for polypurine/polypyrimidine sequences. The mutational specificity of the Polδ variant matches that of the hypermutated CRC cell line, HCT15, in which this variant was first identified. The results provide compelling evidence for the pathogenic role of the POLD1-R689W mutation in the development of the human tumor and emphasize the need to experimentally determine the significance of Polδ variants present in sporadic tumors.

INTRODUCTION

Mutations in genes that control cell division are the primary cause of cancer. Not surprisingly, malfunction of the cellular systems that maintain genome stability increases cancer risk. Among these systems, the pathway controlling accuracy of DNA replication is of utmost importance for preventing mutation and tumorigenesis.¹ It involves accurate selection of nucleotides by replicative DNA polymerases, proofreading of rare errors by the 3′→5′ exonuclease activity of these polymerases, and postreplicative error correction by the DNA mismatch repair (MMR). Inherited mutations in MMR genes cause a predisposition to colorectal cancer (CRC) in Lynch syndrome.^2,3 The MMR gene MLH1 is also somatically inactivated in ~15% of sporadic colorectal, endometrial and gastric tumors.⁴

Germline mutations in the POLD1 and POLE genes encoding the catalytic subunits of replicative DNA polymerases δ (Polδ) and ε (Polε) similarly cause hereditary CRC.^5,6 Furthermore, ultramutated sporadic colon and endometrial tumors (~3% and 8% of all CRC and endometrial cancer cases, respectively) were found to carry somatic changes in Polε.^7–10 The changes affect conserved amino acid residues in the exonuclease domains, suggesting that loss of proofreading must be responsible for the faulty DNA maintenance that leads to cancer. Many of these Polε variants, indeed, have impaired 3′→5′ exonuclease activity and reduced fidelity.¹¹ However, in vivo studies showed that the variant alleles cause very strong mutator effects far exceeding those expected from loss of proofreading.^12,13 The additional biochemical defect(s) that make the cancer-associated Polε variants so mutagenic remain to be identified. Because these POLE mutations primarily occur in tumors that do not have MMR defects, the encoded Polε variants are believed to be responsible for the high number of mutations present in these tumors.^14,15

Somatic POLD1 mutations have also been observed in up to 5% of sporadic colorectal and endometrial tumors,^8–11,16,17 as well as in other cancers (http://www.cbioportal.org/; http://cancer.sanger.ac.uk/cosmic). In contrast to the POLE variants that have garnered much interest, the majority of POLD1 mutations are found in MMR-deficient tumors and do not show a noticeable concentration in the exonuclease domain. Approximately a quarter of these mutations affect conserved amino acid residues in the DNA polymerase motifs. Functional consequences of the vast majority of these mutations is unknown. We have previously shown that the yeast analog of one such variant allele, POLD1-R689W, encodes an error-prone DNA polymerase and causes a catastrophic increase in spontaneous mutagenesis.¹⁸ This prompted us to investigate the effects of the R689W substitution on the function of human Polδ in the present study. The POLD1-R689W mutation was found in MSH6-deficient CRC cell lines DLD-1 and HCT15,^19,20 which were derived from the same tumor²¹ and are among the most hypermutated cell lines known.^19,22 However, due to the presence of the MMR defect in these cells, the potential significance of the POLD1-R689W variant has been overlooked. We demonstrate that expression of POLD1-R689W is strongly mutagenic in human cells, and that the encoded enzyme, Polδ-R689W, has reduced base selectivity leading to frequent nucleotide misincorporation during DNA replication. The error signature of Polδ-R689W matches the mutational pattern of the HCT15 cell line, supporting the idea that the POLD1-R689W mutator played a primary role in the development of the tumor. These findings argue that replicative DNA polymerase mutations that map outside of the exonuclease domain and/or occur in MMR-deficient tumors can be highly significant, and the need for functional analysis of such mutations is crucial.

RESULTS

Expression of POLD1-R689W elevates the mutation rate in MMR-proficient and MMR-deficient human cells

The POLD1-R689W mutation was originally identified in the CRC cell lines DLD-1 and HCT15.^19,20 While these cell lines show a mutator phenotype, the presence of MMR defect and multiple replicative DNA polymerase mutations^19,20 makes it difficult to determine whether the POLD1-R689W variant contributes to the genomic instability. We sought to compare the mutation rate in cells that differ only by the status of the POLD1 allele (wild-type versus POLD1-R689W). We have chosen the CRC cell line HCT116, which has no mutations in the replicative DNA polymerase genes, as a model. These cells have been used extensively for mutagenesis studies^23–25 because of their near-diploid karyotype and the ability to form colonies readily when plated at low densities, which facilitates mutation rate measurements. A derivative of HCT116 is also available (HCT116+ch3), in which the MMR defect originally present in this cell line is corrected by microcell transfer of chromosome 3 containing the MLH1 gene.²⁶ This allowed us to study the effects of POLD1-R689W in closely related MMR-proficient and MMR-deficient cells.

To determine if POLD1-R689W expression elevates the mutation rate, we created stable clonal cell lines by transducing HCT116 and HCT116+ch3 with retroviral vectors expressing the wild-type POLD1 or POLD1-R689W. At least six clones of each transgenic cell line were created. The majority of clonal cell lines had morphology and growth rate similar to the respective parental cell line. In all of these clones, POLD1 was significantly overproduced (Figure 1), indicating that the exogenously introduced alleles served as the primary source of the polymerase. Next, we determined the effects of POLD1-R689W on the rate of mutation at the HPRT1 locus, which confers 6-thioguanine resistance (6-TG^r).²⁷ The 6-TG^r mutation rate in the original HCT116+ch3 and HCT116 cell lines was 1.2 × 10⁻⁶ and 8.1 × 10⁻⁶, respectively (Table 1), consistent with earlier data.²⁴ Overexpression of wild-type POLD1 had no effect on mutagenesis in either cell line (Table 1). In contrast, mutagenesis was increased 2.0-, 3.6-, and 3.9-fold in three independent derivatives of HCT116+ch3 overexpressing POLD1-R689W. Clones with higher levels of overexpression had a slightly higher mutation rate, similar to our previous findings with the yeast analog of POLD1-R689W.¹⁸ MMR deficiency enhanced the mutator effect of POLD1-R689W: the mutation rate was increased 2.9-, 11-, and 12-fold in three HCT116 derivatives expressing POLD1-R689W in comparison to the original HCT116 cells (Table 1). The smaller increase in one of the clones did not appear to result from a lower expression of POLD1-R689W (Figure 1) and is likely a consequence of a suppressor mutation that strong mutators frequently accumulate.²⁸ Overall, the data show that the POLD1-R689W allele imparts a mutator phenotype in the presence and absence of a MMR defect.

Figure 1.

Immunoblots showing p125 (POLD1) levels in clonal cell lines overexpressing POLD1 or POLD1-R689W. The numbers in parenthesis indicate independently derived cell lines. The left and right panels are from separate SDS-PAGE gels.

Table 1.

HPRT1 mutation rate in HCT116-derived cell lines overexpressing POLD1 or POLD1-R689W

Cell line1	MMR status	HPRT1 mutation rate (×10−6)2	Standard error (×10−6)2	Fold Increase3
HCT116+ch3	+	1.2	0.08	1.0
HCT116+ch3 POLD1 (1)	+	0.94	0.17	0.78
HCT116+ch3 POLD1 (2)	+	1.4	0.41	1.2
HCT116+ch3 POLD1-R689W (1)	+	2.4	0.64	2.0
HCT116+ch3 POLD1-R689W (2)	+	4.3*	0.49	3.6
HCT116+ch3 POLD1-R689W (3)	+	4.7*	0.48	3.9
HCT116	−	8.1	1.0	6.8
HCT116 POLD1 (1)	−	8.9	1.6	7.4
HCT116 POLD1 (2)	−	8.8	1.2	7.3
HCT116 POLD1-R689W (1)	−	96*	11	80
HCT116 POLD1-R689W (2)	−	24*	2.6	20
HCT116 POLD1-R689W (3)	−	91*	20	76

¹The numbers in parentheses designate independently derived cell lines.

²The mutation rate was calculated by plotting the mutation frequency as a function of population doublings, and determining the slope of the resulting line and standard error by linear regression using the LINEST function (Microsoft Excel). Asterisks indicate significant difference (p<0.01) from the respective parental cell line, HCT116+ch3 or HCT116. The p-values were calculated with Minitab statistical software using the “Fit Regression Model” function.

³Fold increase in the mutation rate is relative to HCT116+ch3.

Because MMR primarily corrects errors that occur during DNA replication, the effect of MMR status on the mutagenic potential of POLD1-R689W can be used to determine if this mutator promotes replication errors. If the majority of mispairs created by Polδ-R689W are corrected by MMR, a combination of POLD1-R689W with a MMR defect should result in a synergistic increase in the mutation rate. We have used such synergy analysis previously to demonstrate that mutations induced by the yeast analog of POLD1-R689W result from replication errors.²⁹ We analyzed the mutation rates in HCT116- and HCT116+ch3-derived cell lines (Table 1) to determine if this is also true for the human POLD1-R689W. MMR deficiency increases the mutation rate 6.8-fold (compare the values for HCT116 and HCT116+ch3 cell lines). The expression of POLD1-R689W in HCT116+ch3 cells increases the mutation rate 3.2-fold, on average. In HCT116 cells expressing POLD1-R689W, the combination of the MMR defect and POLD1-R689W results, on average, in a 58-fold increase in the mutation rate (compared to HCT116+ch3). This strong synergistic interaction indicates that errors resulting from POLD1-R689W expression are normally corrected by MMR and, therefore, are made during replicative DNA synthesis.

The mutational signature of POLD1-R689W in human cells

We next determined the mutational specificity of human POLD1-R689W. We sequenced the HPRT1 gene in 96 independent 6-TG^r mutant clones that arose in the POLD1-R689W-expressing HCT116 cells (clone 1). The mutation rate in this clone exceeds the mutation rate in the original HCT116 cells 12-fold (Table 1), so at least 92% of the 6-TG^r mutants must have resulted from Polδ-R689W errors. Four of the mutants had no mutation present in the HPRT1 coding sequence and could possibly have a mutation(s) in the 5′ or 3′ untranslated region. An additional 11 mutants likely had large deletions within the HPRT1 locus, as suggested by the failure to amplify some stretches of adjacent exons by PCR or the entire HPRT1 cDNA by rtPCR. Among the remaining 81 mutants, 70 contained base substitutions leading to nonsynonymous amino acid changes, premature stop codons, or ablation of correct exon splicing, and 11 contained −1 or +1 frameshift mutations within the HPRT1 coding sequence (Table 2 and Supplementary Figure S1). A total of 83% of base substitutions were GC→TA transversions or GC→AT transitions (Table 2).

Table 2.

The spectrum of spontaneous HPRT1 mutations in HCT116 POLD1-R689W cell line and the parental HCT116 cells

Mutation	HCT116 POLD1-R689W		HCT116
	Number	Percent of total	Number	Percent of total
Base substitutions	70	86	49	54
GC to AT	26	32	22	24
GC to TA	33	41	18	20
GC to CG	0	<1.2	0	<1.1
AT to GC	8	9.9	9	10
AT to CG	1	1.2	0	<1.1
AT to TA	2	2.5	0	<1.1
Indels	11	14	41	46
Minus 1	6	7.4	18	20
Minus 3	0	<1.2	2	2.2
Plus 1	5	6.2	21	23

The data is based on DNA sequence analysis of 81 and 90 independent HPRT1 mutants of the HCT116 POLD1-R689W (1) cell line and the parental HCT116 cell line, respectively. An additional silent mutation (not included in the table), AT to CG, was observed in one HPRT1 mutant of HCT116 POLD1-R689W (1) that also contained a detectable change. The location of individual mutations in the HPRT1 sequence is shown in Supplementary Figures S1 and S2.

Notably, the expression of POLD1-R689W dramatically changed the specificity of spontaneous mutagenesis in HCT116 cells making it resemble the mutational specificity of HCT15 cells (Figure 2). The spectrum of HPRT1 mutations in the parental HCT116 cells, in which we introduced POLD1-R689W, showed a high frequency of small insertions/deletions (indels) (Table 2), similar to earlier data for this cell line.²⁵ In contrast, spontaneous mutations in HCT15 cells are almost exclusively base substitutions.²⁵ The expression of POLD1-R689W in HCT116 cells decreased the proportion of indels and increased the proportion of base substitutions, particularly GC→TA transversions, making the spectrum very similar to that of HCT15 (Figure 2). The minor differences between HCT15 and HCT116 POLD1-R689W spectra (e.g. still a slightly higher proportion of indels in the latter) are likely due to the fact that up to 8% of mutations in the POLD1-R689W-expressing cells are contributed by the HCT116 background. Indeed, the majority of indels in HCT116 POLD1-R689W cells (nine out of 11) occurred in a run of six consecutive G-C base pairs (Supplementary Figure S1) that is a hot spot for indels in HCT116 (²⁵; Supplementary Figure S2). Taken together, these observations strongly argue that the POLD1-R689W variant plays a major role in shaping the mutational specificity of HCT15 cells. The results also suggest that the additional mutations present in the POLD1 gene in these cells, including the R506H variant in the exonuclease domain, contribute very little, if at all, to the mutational spectrum.

Figure 2.

Expression of POLD1-R689W alters the specificity of spontaneous mutagenesis in HCT116 cells making it resemble the mutational specificity of HCT15. The diagrams show the HPRT1 mutation spectra of HCT116, HCT116 POLD1-R689W and HCT15 cell lines. The number of independent HPRT1 mutants analyzed (n) is indicated for each spectrum. Data for HCT116 POLD1-R689W and HCT116 are from Table 2, and data for HCT15 is from reference 25. Asterisks indicate statistically significant differences in the proportion of indels and GC→TA transversions between HCT116 and HCT116 POLD1-R689W spectra (p=0.000005 for indels and p=0.0043 for GC→TA transversions, Fisher’s exact text). No significant differences were observed between HCT116 POLD1-R689W and HCT15 spectra.

The largest class of HPRT1 mutations induced by POLD1-R689W expression, GC→TA transversions, showed a strong DNA sequence context preference. Nearly all occurred in polypurine/polypyrimidine tracts, with all but one having the polypurine sequence in the non-transcribed strand (Figure 3, left). To further support the premise that the POLD1-R689W variant was primarily responsible for the hypermutability of the original tumor, we analyzed the DNA sequence context of GC→TA transversions present in a set of 26 cancer-related genes in the HCT15 cell line.¹⁹ This set of genes was used as a sample of genomic sequence potentially enriched for mutations that played a role in the development of the tumor. All GC→TA transversions in these genes were analyzed without consideration of their possible functional significance. Twelve GC→TA transversions were found in the 26 genes. Strikingly, all occurred in polypurine/polypyrimidine tracts, with many of the sequences matching precisely the context of POLD1-R689W-induced mutations (Figure 3, right). In contrast, two GC→TA transversions found in the 26 cancer genes in HCT116 cell line¹⁹ did not share this sequence context. Thus, the context analysis suggests that not only does the POLD1-R689W allele determine the specificity of new mutations arising in the HCT15 cell line, but it likely shaped the genome of the original hypermutated tumor as well.

Figure 3.

DNA sequence context of GC→TA transversions in cancer-related genes in the HCT15 cell line matches the mutational specificity of the POLD1-R689W mutator. (A) Expression of POLD1-R689W results in GC→TA transversions at polypurine/polypyrimidine tracts. Genomic sequence context is shown for all GC→TA transversions observed at the HPRT1 gene of the HCT116 POLD1-R689W cells. Each mutated site is shown as a separate entry with the number of mutations at this site in parentheses. The mutated base is underlined. (B) Sequence context of GC→TA transversions found in a set of 26 cancer-related genes in the HCT15 cell line. The set of genes was the same as the one used previously to characterize the mutational specificity of the NCI-60 panel of cell lines¹⁹ and included APC, ARID1A, BRAF, DNMT1, DNMT3A, DNMT3B, EGFR, EPHA3, EPHA5, EPHA7, FBXW7, GRIN2A, KRAS, LRP1B, NF2, NRAS, PBRM1, PIK3CA, POLE, PTEN, SETD2, SPTA1, STAG2, SYNE1, TP53, and TRRAP. Sequence of the G-containing strand is shown. The mutated base is underlined.

Purified human Polδ-R689W has reduced base selectivity

To further understand the mechanisms responsible for the mutator effect of POLD1-R689W, we purified four-subunit human Polδ and Polδ-R689W using an insect cell/baculovirus-based protein production system (Figure 4a) and compared their biochemical properties. On an oligonucleotide substrate, the 3′→5′ exonuclease and DNA polymerase activities of Polδ-R689W and the wild-type Polδ were nearly identical (Figure 4b,c). The intact exonuclease activity indicates that Polδ-R689W must be fully capable of proofreading its errors, consistent with the location of Arg689 outside of the exonuclease domain. Next, we examined whether POLD1-R689W had a base selectivity defect by monitoring the insertion of individual nucleotides, correct or incorrect, across from a template C. The correct dGTP was inserted by the wild-type and mutant Polδ equally well. However, unlike the wild-type enzyme, Polδ-R689W also showed a profound ability to misincorporate dTTP at this position (Figure 4d). The C-dTTP is one of the two mispairs that could lead to GC→TA transversions, the most frequent type of mutation induced by POLD1-R689W expression. Thus, the results of oligonucleotide-based assays provide strong support for the idea that the increased mutation rate in cell lines expressing POLD1-R689W is due to a reduced base selectivity of the encoded polymerase.

Figure 4.

Polδ-R689W is an active and highly error-prone DNA polymerase. (A) Purified Polδ and Polδ-R689W were separated by SDS-PAGE in a 10% Bis-Tris gel and visualized with Coomassie staining. (B) DNA synthesis by Polδ and Polδ-R689W was analyzed by incubating the purified enzymes with all four dNTPs and an oligonucleotide template (see Materials and Methods) for the times indicated. (C) Exonuclease activity was assayed by incubating the purified enzymes with the oligonucleotide substrate and no dNTPs. (D) The efficiency of correct and incorrect nucleotide insertion across from a template C was analyzed by incubating the enzymes and the oligonucleotide substrate for 15 min in the presence of dGTP or dTTP as indicated. Dashed lines indicate that the lanes were not adjacent to each other in the original gel.

DISCUSSION

Hypermutated tumors with replicative DNA polymerase defects carry up to one million clonal mutations in their genomes.^8,10 At such a density of mutations, nearly every gene is affected, and multiple alterations in the DNA maintenance pathways are often present. Identification of mutations responsible for the high level of genome instability requires experimental assessment of their significance. A number of replicative DNA polymerase variants, particularly Polε exonuclease domain variants, have been suggested to be functionally important by bioinformatics analysis and/or by functional studies in yeast or in vitro.^5,11,13 However, the ability to increase the mutation rate in human cells has not been demonstrated for any of the cancer-associated polymerase variants. Prompted by the extraordinary mutator effect of the yeast POLD1-R689W analog,¹⁸ here we analyzed the functional consequences of the POLD1-R689W mutation in human cells. We show that expression of the POLD1-R689W allele is mutagenic in both MMR-deficient and MMR-proficient cells. Synergistic interaction with the MMR defect indicated that POLD1-R689W-induced mutations result from DNA replication errors. Biochemical studies of purified human Polδ-R689W indicated that decreased base selectivity of this polymerase is likely responsible for the mutator effect. Finally, we show that the mutational signature of POLD1-R689W is consistent with the pivotal role of the encoded polymerase in the development of the human tumor, in which this variant was found.

This study indicates that POLD1 mutations seen in sporadic human tumors can be highly significant. Furthermore, DNA polymerase mutations in MMR-deficient tumors can be highly significant, as appears to be the case for POLD1-R689W. While MMR deficiency amplifies the mutator effect, the initial replication error load and mutational specificity are determined by the DNA polymerase variant in such tumors. It is worth noting that the HCT15 and DLD-1 cell lines are deficient only in the MSH6-dependent MMR subpathway. It remains to be determined whether strong DNA polymerase mutators are compatible with full MMR deficiency, such as that caused by inactivation of MSH2 or MLH1. Curiously, the POLD1-R689W variant has been reported in two other sporadic tumors, a hepatocellular carcinoma and another colon tumor (http://www.cbioportal.org/), neither of which was hypermutated. While relative levels of the wild-type and mutant allele expression and/or tissue- and tumor-specific factors could have affected the expression of the mutator phenotype, it is also possible that a combination with a MMR defect is required for POLD1-R689W to cause significant hypermutation. Finally, this study demonstrates that functionally important DNA polymerase mutations can occur outside the exonuclease domain and affect base selectivity rather than the much discussed proofreading. Interestingly, a L606M substitution in the DNA polymerase domain of Polδ, which is analogous to a well-known yeast mutator variant Polδ-L612M,^30–32 has been reported as a somatic mutation in brain tumors of children with hereditary biallelic MMR gene defects,³³ providing another example of a likely significant somatic POLD1 mutation. Future studies of other POLD1 changes reported in sporadic tumors will help identify additional pathogenic variants.

This study also validates the use of the Saccharomyces cerevisiae model system for functional analysis of cancer-associated DNA polymerase mutations. We have previously shown that the yeast pol3-R696W allele mimicking the human POLD1-R689W confers a strong mutator phenotype, interacts synergistically with the MMR defect, and encodes a Polδ variant with normal exonuclease activity but severely reduced base selectivity.^18,29 All of these properties were recapitulated in the present study of the human POLD1-R689W variant. Even the propensity of human Polδ-R689W for misincorporation of dTTP across from a template C on an oligonucleotide template in vitro (Figure 4d) is almost identical to that of yeast Polδ-R696W.¹⁸ Like the human POLD1-R689W, the yeast pol3-R696W mutator produced almost exclusively GC→AT transitions and GC→TA transversions in vivo.^18,29 The only difference was that transitions predominated in the mutational spectrum of msh6 pol3-R696W yeast strains,²⁹ and transversions were slightly more frequent in HCT116 POLD1-R689W cells (Figure 2 and Table 2). This could potentially be due to different reporter genes used or slight differences in the properties of the enzymes. Overall, the remarkable similarity of the effects of human POLD1-R689W and its yeast mimic illustrates the value of the yeast system for the identification of functionally significant polymerase variants. This is particularly important, because the number of POLD1 and POLE mutations that await functional analysis will likely increase substantially as additional cancer genomes are sequenced.

Despite the advantages of the yeast system, mutations that affect poorly conserved amino acid residues might require analysis directly in human cells. The present work describes a simple and rational strategy for such an analysis. An approach has been recently proposed for the characterization of POLD1 mutations, wherein the endogenous protein is inducibly replaced with the mutant variant.³⁴ While elegant and undoubtedly valuable for genes where the effects of mutations need to be analyzed in the absence of the wild-type allele expression, such inducible replacement may not be necessary for POLD1 or POLE variants. As a rule, replicative DNA polymerase variants are present in cancer patients in the heterozygous state. Loss of heterozygosity is not required for the patients with germline POLD1 or POLE mutations to develop tumors.⁵ Studies of sporadic tumors with POLD1 or POLE variants always report the presence of both wild-type and mutant alleles.^7,8,13,35 While we cannot exclude the possibility that non-tumor cells in the sequenced samples were responsible for the wild-type signal in some cases, cell lines established from hypermutated tumors are invariably heterozygous for the DNA polymerase mutations.^19,20,35 This is consistent with the view that the cancer-associated variant alleles encode active error-prone polymerases that function in the heterozygous cells along with the wild-type enzyme. Indeed, when tested in the model systems, the cancer-associated alleles appear to be semi-dominant.^13,29 Thus, a reasonable criterion for the functional significance of POLD1 and POLE variants would be their ability to confer the mutator phenotype in the presence rather than in the absence of the wild-type allele. Development of cell-based assays where the wild-type and mutant alleles are expressed at comparable levels could allow for a more accurate prediction of the in vivo effects of DNA polymerase variants in the future.

The analysis of sequence specificity of POLD1-R689W-induced GC→TA transversions in the HPRT1 gene revealed a strong preference for polypurine/polypyrimidine tracts and a highly asymmetric occurrence of mutations in respect to the two strands (Figure 3). The preference for GC→TA transversions in polypurine/polypyrimidine sequences is also evident in the genomic landscape of the HCT15 cell line (Figure 3), confirming the biological relevance of this signature. However, tracts with purines in the transcribed and non-transcribed strands were nearly equally affected in the HCT15 cells (not shown in Figure 3, but the original data can be found in ¹⁹), suggesting that the bias we observed in the HPRT1 gene is related to the strand-specific function of Polδ in DNA replication rather than the transcriptional asymmetry. A preferential occurrence of mutations in polypurine/polypyrimidine tracts has been observed previously in several other situations where increased mutagenesis results from DNA replication errors. Some examples include mutations induced by the pol3-R696W allele or by treatment with a base analog 6-N-hydroxylaminopurine, a potent inducer of replication errors, in yeast, with a strong bias for G in the non-transcribed strand in both cases.^29,36 Curiously, the same context specificity could be seen for GC→TA transversions present in the genome of a hypermutated CRC cell line HCC2998, which carries a strongly mutagenic POLE-P286R variant (Supplementary Figure S3). The mechanism of preferential mutability of polypurine/polypirimidine tracts remains to be determined. One possibility we have previously discussed³⁶ is that stacking interactions between purine bases contribute to this phenomenon. Regardless of the mechanism, this signature could be useful for tracking the activity of mutator Polδ and Polε variants in human cancers.

MATERIALS AND METHODS

Cell lines

HCT116 and HCT116+ch3 were obtained from Robert Lewis and Michael Brattain (University of Nebraska Medical Center) and were last authenticated by STR profiling and tested for mycoplasma contamination in November 2016. All cell lines were maintained in Dulbecco’s Modified Eagle’s medium (DMEM; Hyclone) containing 10% fetal bovine serum (FBS; Atlanta Biologicals). HCT116+ch3 and its derivatives were grown in media containing 200 μg/ml G418 to prevent loss of chromosome 3.

Clonal cell lines expressing POLD1 or POLD1-R689W were constructed by retroviral transduction. Retroviral particles were created by transfection of HEK293-GP cells (Clontech) with plasmids pVSV-G and MXIVpuro-POLD1 or MXIVpuro-POLD1-R689W using FuGENE6 (Promega) or TurboFect (Thermo Scientific) reagents and recommended manufacturer protocols. Media containing viral particles was collected 24 and 48 h after transfection and passed through a 0.45-μM syringe filter to remove unattached cells. Transductions of HCT116 and HCT116+ch3 were performed when cells were ~60% confluent by replacing the growth media with the retroviral media mixed with an equal volume of fresh DMEM+10%FBS and 10 μg/ml Hexadimethrine bromide (Polybrene). Cells with integrated retroviral vectors were selected for 24 h after transduction with 0.35 μg/ml puromycin (Life Technologies). Individual clones were isolated from separate transduction dishes by serial dilution in 96-well culture plates such that ~10% of the wells were positive for growth. Wells were examined after nine days, and only wells with a single colony were expanded for future use.

Plasmids

Plasmid pVSV-G was from Clontech. Retroviral vectors MXIV-puro-POLD1 and MXIV-puro-POLD1-R689W expressing the POLD1 alleles from the CMV promoter were constructed as follows. The R689W mutation was first introduced into the POLD1 cDNA cloned in the bacterial expression plasmid pET-POLD4/1³⁷ by site-directed mutagenesis. Plasmids pET-POLD4/1 and pET-POLD4/1-R689W were digested with EcoRI, and fragments containing the POLD1 and POLD1-R696W cDNA were cloned into the EcoRI site of pBABE-puro-based retroviral expression vector MXIV-puro.³⁸ For overproduction in insect cells, cDNAs for all full-length human Polδ subunits were obtained from Open Biosystems, cloned into pFastBac-1 transfer vector (Life Technologies), and the R689W mutation was introduced by site-directed mutagenesis.

HPRT1 mutation rate and mutation spectra analysis

The HPRT1 mutation rate was determined by measuring the accumulation of mutants during serial passaging as described.²⁴ To characterize the spectrum of HPRT1 mutations, cells were cleansed of pre-existing mutants by passaging ten times in DMEM+10%FBS containing 100 μM hypoxanthine, 0.4 μM aminopterin, and 16 μM thymidine. Independent cultures were then started in DMEM+10%FBS from <10 cells and expanded to ~4×10⁷ cells. Approximately 1×10⁵ cells from each culture were plated in 10-cm dishes containing DMEM+10%FBS with 30 μM 6-TG and incubated for 14 days. The media was changed every three days. One randomly selected colony was isolated by the use of cloning cylinders and expanded to 1×10⁷ cells. Genomic DNA and total RNA were extracted using the Puregene DNA cell kit (Qiagen) and total RNA isolation kits (Omega Bio-Tek and IBI). The cDNA was produced using the SMARTScribe reverse transcriptase kit and oligo(dT)₁₈ primers. The HPRT1 cDNA was amplified with primers flanking the entire coding region and sequenced. If sequencing of rtPCR products revealed skipped exons, corresponding genomic DNA samples were used as templates to amplify exons and flanking intronic sequence.

Immunoblots

Following dissociation with trypsin, ~5×10⁶ cells were collected by centrifugation at 500 RCF, washed with 1 ml of Phosphate Buffered Saline, and flash-frozen in liquid nitrogen. Cells were re-suspended in 350 μl of NP-40 buffer (50 mM Tris-HCl, pH 7.6, 1% NP-40, 150 mM NaCl, 1 mM MgCl₂) containing 1x complete protease inhibitor cocktail (Roche) and incubated at 4 °C for 30 min with gentle agitation. The lysate was then centrifuged at 10,000 RCF at 4 °C for 10 min. The supernatant was either used immediately for immunoblotting or stored at −80 °C. For immunoblots, 30 μg of protein was separated in a 4–12% bis-tris gel and transferred to a nitrocellulose membrane. This membrane was subsequently cut horizontally at a position corresponding to a protein molecular mass of ~70 kDa. The top and bottom halves were probed with mouse monoclonal anti-POLD1 (Abcam, ab10362) and rabbit polyclonal anti-α-tubulin (Abcam, ab4074) antibodies, respectively. The secondary antibodies were HRP-conjugated goat anti-mouse (Genscript, A00160) and goat anti-rabbit (Genscript, A00160). Proteins were detected by using chemiluminescent SuperSignal West Pico kit (Thermo Scientific) and autoradiography.

Purification of human Polδ and Polδ-R689W

Polδ and Polδ-R689W with 6xHis-tagged p125 subunits were produced in insect cells co-infected with four baculoviruses, each encoding one of the four Polδ subunits. Preparation of high-titer baculoviruses and protein production were done using the Bac-to-Bac baculovirus expression system (Life Technologies) and the manufacturer instructions. Buffer N (20 mM Tris-HCl, pH 7.9, 5% glycerol, 10 mM KH₂PO₄/K₂HPO₄, pH 7.9, 0.005% Nonidet P-40, 3 mM 2-mercaptoethanol, 1 mM phenylmethylsulfonyl fluoride, 2 μg/ml pepstatin, 2 μg/ml leupeptin) was used during all purification steps. The concentration of NaCl (in mM) is denoted by subscripts. Insect cells (~7 g) were lysed in buffer N₁₀₀ containing 1 mM imidazole. After centrifugation at 40,000 RCF for 30 min, the cleared lysate was incubated with 5 ml of Ni-IDA resin (Bio-Rad) for 60 min with gentle agitation. The mix was loaded onto a Biorad econocolumn, washed with N₁₀₀ containing 1 mM imidazole, and Polδ or Polδ-R689W was eluted with a 1 mM to 150 mM imidazole gradient in Buffer N₁₀₀. Fractions containing Polδ or Polδ-R689W were combined with an equal volume of N₀, dialyzed to N₂₅, filtered, and loaded to a Mono S column (GE Life Sciences). Polδ was eluted with an N₂₅ to N₈₀₀ gradient, the NaCl concentration was adjusted to ~250 mM with N₁₀₀, and the fractions were loaded to a 1-ml HiTrap Heparin HP column (GE Life Sciences). Polδ was eluted with an N₂₅₀ to N₈₀₀ gradient. Fractions containing Polδ were dialyzed in N₁₀₀ and loaded to a 1-ml Mono Q column (GE Life Sciences). Polδ was eluted with an N₁₀₀ to N₈₀₀ gradient. Peak fractions were concentrated and stored in 20 mM Tris-HCl, pH 7.9, 10% glycerol, 10 mM KH₂PO₄/K₂HPO₄, pH 7.9, 0.005% NP-40, 2 mM dithiothreitol, 150 mM NaAc, 1 mM phenylmethylsulfonyl fluoride, 2 μg/ml pepstatin and 2 μg/ml leupeptin.

In vitro DNA polymerase and exonuclease assays

The reactions were performed at 37°C with 25 nM Cy5-labeled oligonucleotide substrate (5′-Cy5-CAGCACCACAAACCATACAAAAACA-3′/5′-GCCATTATCGGGTTTCTAATATACTGTTTTTGTATGGTTTGTGGTGCTG-3′),¹⁸ 40 mM Tris-HCl, pH 7.8, 150 mM NaAc, 10 mM MgAc, 0.2 mg/mL bovine serum albumin, 4% polyethylene glycol 8000, 1 mM dithiothreitol, 25 nM Polδ or Polδ-R689W, and 100 μM dNTPs or no dNTPs. Reactions were terminated by placing the tubes on ice and adding 15 μl of formamide loading dye. The products were separated by electrophoresis in a 16% denaturing polyacrylamide gel, and detected and quantified as described.¹⁸