Defects in 8-oxo-guanine repair pathway cause high frequency of C > A substitutions in neuroblastoma

High Frequency of C > A Substitutions Correlate with Poor Prognosis in Neuroblastoma.

We determined the somatic frequency of each of the six possible base substitutions in our whole-genome sequencing (WGS) data of 86 primary neuroblastoma tumor/normal pairs (17). C > A substitutions were the most abundant base substitution type, including 14 tumors with a C > A substitution frequency higher than 50% (Fig. 2A and SI Appendix, Fig. S1 A and B).

Fig. 2.

High percentages of C > A substitutions in neuroblastoma are correlated with high-stage disease and poor prognosis. (A) Relative substitution frequency of the six possible nucleotide substitutions, indicated with different colors, in a cohort of 86 neuroblastoma tumors. Tumors are ranked from left to right on C > A substitution frequency. (B) Dot-boxplot presenting the C > A substitution frequency for the different stages of neuroblastoma (INSS, st1 = stage1, st2 = stage2, st3 = stage 3, st4 = stage 4, st4s = stage 4s) in the same cohort. C > A substitution frequency is significantly higher (unpaired t test with Welch correction, two-sided) in stage 4 neuroblastoma compared to lower stages. ****P < 0.0001. The central line in the boxplot indicates the median and the box limits indicate the first and third quartile. The whiskers denote the interval within 1.5 times the interquartile range of the median. Outliers are indicated with a cross. The number of patients per stage is indicated below the graph. (C) Kaplan–Meier curve was created using Kaplan scanning for the optimal curves. This curve indicates a significantly worse survival for patients with high compared to low C > A substitution frequency (Bonferroni-corrected P = 2.8 × 10⁻¹⁰).

Since C > A substitutions have been suggested to also occur from sequencing artifacts (18), we reanalyzed available Sanger sequencing results to confirm the C > A substitutions on the original tumor biopsies (17). Sanger sequencing validation of C > A substitutions was similar to validation of other substitutions (SI Appendix, Fig. S1C). In addition, C > A ratios were not enriched in the control DNA (leukocytes) (SI Appendix, Fig. S1D). Finally, similar levels of C > A substitutions were found in tumors and their corresponding organoid cell lines (SI Appendix, Fig. S1E). Therefore, we conclude that the C > A substitutions found in our tumor cohort were authentic.

We next compared the C > A substitution frequencies for the different tumor stages and identified significantly more C > A substitutions in International Neuroblastoma Staging System (INSS) stage 4 tumors compared to the low-stage tumors (unpaired t test with Welch correction, two-sided, P < 0.0001; Fig. 2B). Furthermore, a Kaplan-Meier analysis indicates a significant logrank (P = 2.8 × 10⁻¹⁰) difference in overall survival between patients with a high and a low percentage of C > A substitutions (Fig. 2C), independent of the prognostic factors of age, MYCN status, and INSS stage (SI Appendix, Table S1). These results show that the frequency of C > A substitutions is strongly increased in neuroblastoma tumors and that this correlates with a poor prognosis.

High Levels of C > A Substitutions Correlate with Defects in MUTYH and OGG1.

C > A substitutions can result from defects in the 8-oxoG repair pathway (9). Therefore, we analyzed WGS data of our neuroblastoma cohort for CNL and mutation status of MUTYH, OGG1, and NUDT1. In a subset of tumors, we could identify CNL of MUTYH or OGG1 (SI Appendix, Fig. S2 A and B). OGG1 is located on 3p.25, which is part of the frequently deleted 3p region in neuroblastoma. MUTYH, which is located on 1p.34, was deleted in a subset of tumors with (partial) 1p CNL. NUDT1 CNL was identified in two tumors, which also had a MUTYH CNL (SI Appendix, Table S2). CNL of OGG1 and MUTYH was associated with a high frequency of C > A substitutions (Fig. 3A). Compared to tumors with a normal copy number of OGG1 or MUTYH, a significantly higher C > A substitution frequency is detected in tumors with CNL of MUTYH (unpaired t test with Welch correction, two-sided, P < 0.0001) or OGG1 (unpaired t test with Welch correction, two-sided, P = 0.0418; Fig. 3B). The tumor (N170T) with the highest C > A substitution frequency (82%) carried a germline missense variant in MUTYH (NP_036354.1:p.Ala425Pro) and a CNL of the other MUTYH allele (SI Appendix, Table S2). This missense variant in N170T is not reported in the Single Nucleotide Polymorphism Database (dbSNP) (build 153). In addition, this mutation results in an amino acid change from an alanine to a proline, which may affect the protein conformation, although PolyPhen and SIFT do not predict the variant to be damaging. Another tumor (N701T) with a very high frequency (78%) of C > A substitutions carried a germline missense variant in OGG1 (NP_002533.1:p.Gly308Glu) and a CNL of the other OGG1 allele (SI Appendix, Table S2). For this missense variant, it is predicted that it affects function. Overall, five of the seven tumors with the highest C > A substitution frequencies have multiple alterations in genes of the 8-oxoG repair pathway (SI Appendix, Table S2). These results show that C > A substitution frequencies are increased in neuroblastoma tumors with OGG1 or MUTYH CNL.

Fig. 3.

High frequency of C > A substitutions correlates with loss of MUTYH or OGG1. (A) The circular binary segmentation (CBS) scores for MUTYH and OGG1 are plotted for 86 neuroblastoma tumors. The dots represent individual tumor samples and are colored for relative C > A substitution frequency as indicated in the color scale. The red and brown dashed line indicate the cutoff (CBS < −0.5) for MUTYH CNL and OGG1 CNL, respectively. (B) Dot-boxplot presenting the C > A substitution frequency for neuroblastoma tumors with MUTYH CNL (CBS < −0.5), OGG1 CNL (CBS < −0.5), or wild-type (MUTYH and OGG1 CBS > −0.5). The number of tumors per group is indicated below the graph. C > A substitution frequency is significantly higher (unpaired t test with Welch correction, two-sided) in neuroblastoma tumors with MUTYH CNL or OGG1 CNL compared to wild-type. ****P < 0.0001, *P < 0.05. The central line in the boxplot indicates the median and the box limits indicate the first and third quartile. The whiskers denote the interval within 1.5 times the interquartile range of the median. Outliers are indicated with a cross.

8-oxoG Levels Are Increased in Cell Lines with Defects in MUTYH and OGG1.

To study the relation between 8-oxoG levels and defects in the 8-oxoG repair pathway, we used modified alkaline comet assays (19). In this assay, cells embedded in agarose are lysed and treated with purified recombinant OGG1, which creates DNA nicks where 8-oxoG is present. Subsequent gel electrophoresis will separate the nicked DNA from undamaged DNA. With this assay, the DNA can be visualized as comets (undamaged DNA) and tails (nicked DNA). The average tail moment can be used as a measure for the induced DNA damage and therefore for the amount of 8-oxoG. For some of the tumors, we have established tumor-derived organoid cultures (20), and similar C > A substitution percentages were identified in the organoids compared to the corresponding tumor (SI Appendix, Fig. S1D). Compared to a wild-type organoid, two organoids with a defect in the 8-oxoG repair pathway show significantly longer comet tails (unpaired t test with Welch correction, two-sided, P < 0.0001), indicative of higher levels of 8-oxoG (Fig. 4A and SI Appendix, Fig. S3A). To show that the increased levels of 8-oxoG are indeed caused by the defects in OGG1 and MUTYH, we performed rescue experiments by inducing overexpression of wild-type OGG1 or MUTYH in neuroblastoma cell lines with CNL of OGG1 or MUTYH, respectively. Indeed, we could rescue high levels of 8-oxoG by overexpression (72 h) of wild-type OGG1 or MUTYH, as indicated by a significant decrease in comet tail size (unpaired t test with Welch correction, P < 0.0001; Fig. 4 B and C and SI Appendix, Fig. S3 B and C). In a neuroblastoma tumor (N701T) with one of the highest C > A mutation frequencies, a germline OGG1 missense variant was identified, as mentioned above. To test whether this mutation in OGG1 is indeed inactivating, we also induced overexpression of OGG1 containing the missense variant (p.G308E) in the neuroblastoma cell line with OGG1 CNL. Since overexpression of this variant increases the tail moment, this shows that this variant does not rescue the levels of 8-oxoG (SI Appendix, Fig. S4 A and B), confirming that this is an inactivating event in OGG1. In addition, we also tested the effect of a germline MUTYH missense variant identified in the tumor with the highest C > A substitution frequency (N170T). Induced overexpression of MUTYH containing this missense variant (p.A425P) in the neuroblastoma cell line with MUTYH CNL, resulted in a lower tail moment compared to the uninduced control (SI Appendix, Fig. S4 C and D). Nevertheless, the tail moment remains higher than for induced overexpression without recombinant OGG1. This indicates that this variant does not rescue the levels of 8-oxoG to a similar extent as wild-type MUTYH and suggests that this variant reduces MUTYH activity. Overall, these results indicate that 8-oxoG levels are increased in neuroblastoma organoids and cell lines with defects in MUTYH and OGG1 and that this can be rescued by overexpression of the affected wild-type gene.

Fig. 4.

Defects in the 8-oxoG repair pathway result in higher 8-oxoG levels. Tail moments were calculated from modified comet assays with [red bar (+)] or without [blue bar (−)] recombinant OGG1 (rOGG1) treatment. (A) Modified comet assays were performed for two organoids with defects in the 8-oxoG pathway (AMC717T with MUTYH CNL, AMC700T with OGG1 CNL) and one organoid with a functional 8-oxoG pathway (AMC691T). (B, Upper) OGG1 protein expression in SJNB12 (with OGG1 CNL) transduced with a pIND-OGG1 vector with (+) or without (−) addition of doxycycline (Dox). β-actin is used as a loading control. (Lower) Tail moments calculated from modified comet assays for SJNB12 with (+) or without (−) Dox-induced OGG1 overexpression. (C, Upper) MUTYH protein expression in SJNB10 (with MUTYH CNL) transduced with a pIND-MUTYH + or − Dox. β-actin is used as a loading control. (Lower) Tail moments calculated from modified comet assays for SJNB10 with (+) or without (−) Dox-induced MUTYH overexpression. Error bars indicate the SEM. P values are calculated with unpaired t test with Welch correction, two-sided, ****P < 0.0001, **P < 0.01, *P < 0.05.

Increased Accumulation of C > A Substitutions upon OGG1 or MUTYH Knockout in Neuroblastoma Cell Line.

To validate the correlation between C > A substitutions and defects in the 8-oxoG pathway, we created CRISPR-Cas9 knockout clones of MUTYH, OGG1, or NUDT1 in a neuroblastoma cell line (CHP134). To create these knockout clones, we inserted a puromycin resistance cassette in one of the first exons of these genes to disturb the open reading frame (Fig. 5A) (21). After puromycin selection, we performed single-cell fluorescence-activated cell sorting (FACS) to create knockout clones (Fig. 5B). Knockout clones were genotyped and biallelic editing was confirmed for all selected clones (SI Appendix, Fig. S5A). The selected clones also showed absent expression of the targeted protein (Fig. 5C). Messenger RNA expression of the target gene was also strongly reduced for MUTYH and NUDT1 but only slightly reduced for OGG1, possibly because of inefficient nonsense-mediated decay of OGG1 (SI Appendix, Fig. S5 B–D). Selected knockout and wild-type clones were passaged for a period between 88 and 126 d to accumulate independent mutations and subsequently single-cell FACS sorted to establish subclones. From both clones and subclones (n = 2 per gene), genomic DNA was isolated and subjected to WGS to study the mutational patterns that accumulate between the two clonal steps (SI Appendix, Table S3). MUTYH and OGG1 knockout clones showed a relative higher contribution of C > A substitutions ranging from 43 to 57% compared to 25 to 30% in control clones (Fig. 6A and SI Appendix, Table S3). This difference was statistically significant for the MUTYH knockout clones versus the wild-type clones (unpaired t-test, two tailed, P:0.0076) and the MUTYH knockout clones and OGG1 knockout clones versus the wild-type clones (unpaired t test, two-tailed, P: 0.0086). Also, the total number of base substitutions per genome per day was increased for MUTYH and OGG1 knockout clones compared to the wild-type clones (SI Appendix, Fig. S6A). The substitution frequencies in the NUDT1 knockout clones were similar to the wild-type clones (Fig. 6A). No consistent changes in the number of small insertions and deletions were identified between the different knockout and wild-type clones (SI Appendix, Fig. S6B). These results indicate that knockout of OGG1 and MUTYH, but not NUDT1, increases the accumulation of C > A substitutions in individual neuroblastoma cells.

Fig. 5.

CRISPR-Cas9 editing and clonal selection of MUTYH, OGG1, and NUDT1 knockout (sub)clones in CHP134 cell line. (A) CRISPR-Cas9 editing strategy used for targeting MUTYH, OGG1, or NUDT1 in CHP134. (B) Schematic overview of the strategy used for creating clones and subclones of CRISPR-Cas9–edited and wild-type CHP134. Both clones and subclones were sent for WGS to evaluate the mutations that accumulate during clonal outgrowth. (C) Western blot analysis of MUTYH expression in wild-type and selected MUTYH knockout clones of CHP134. GAPDH is used as a loading control. (D) Western blot analysis of OGG1 expression in wild-type and selected OGG1 knockout clones of CHP134. α-tubulin is used as a loading control. (E) Western blot analysis of NUDT1 expression in wild-type and selected NUDT1 knockout clones of CHP134. α-tubulin is used as a loading control.

Fig. 6.

MUTYH or OGG1 knockout results in accumulation of C > A substitutions in neuroblastoma cell line. (A) Relative substitutions per genome per day that accumulated in MUTYH knockout, OGG1 knockout, NUDT1 knockout, and wild-type CHP134 clones. Substitution types are indicated with colors. (B) Mutational spectra are displayed according to the 96 substitution types defined by the substitution class and its trinucleotide context for the different knockout and wild-type clones. The substitution types are indicated on the x-axis, and the relative contribution of the substitution type is depicted on the y-axis.

Mutational Signatures Analysis in Knockout Lines and Neuroblastoma Tumors.

To further study the mutational patterns of the different knockouts, mutational spectra were analyzed and the relative contributions of mutation signatures were determined. The mutational spectra of the wild-type and knockout clones were visualized by showing the relative contribution of the six possible substitutions in their trinucleotide context (Fig. 6B). Cosine similarity analysis showed that the mutational spectra of the MUTYH and OGG1 knockouts are slightly different but cluster close together and separately from the wild-type and NUDT1 knockouts (SI Appendix, Fig. S7A). The mutational spectra were compared using the COSMIC SigProfiler signatures (6, 22). The OGG1 knockouts have an increased number of C > A substitutions at the trinucleotide contexts characteristic of signature 18, relative to the wild-type clones. The slightly different pattern of the MUTYH knockouts mimics the mutational signature 36 (Fig. 6B and SI Appendix, Fig. S7B).

To relate the mutational spectra of the knockout clones to primary tumor samples, combined relative signature contributions were determined and plotted in a heatmap (Fig. 7). There is a strong contribution of signature 36 in one of the tumors, which clusters with the MUTYH knockout clones (Fig. 7). This tumor (N170T) is the only tumor with both a germline MUTYH missense variant and CNL of the other MUTYH allele. In addition, a large cluster with a high contribution of signature 18 was identified in our neuroblastoma tumor cohort. These tumors cluster together with the OGG1 knockout clones (Fig. 7). More than 60% of the tumors in this cluster harbor OGG1 or MUTYH CNL. Furthermore, signature 18 or signature 36 has the highest relative contribution in 90% of the tumors with OGG1 or MUTYH CNL, while this is only 31% in tumors without defective 8-oxoG repair. Overall, these results indicate that the OGG1 and MUTYH knockout clones have a high contribution of C > A mutational signatures 18 and 36, respectively, and cluster together with neuroblastoma tumors with OGG1 or MUTYH CNL.

Fig. 7.

Mutational signatures in neuroblastoma tumors and CHP134 (knockout) clones. Heatmap of neuroblastoma tumors and CHP134 (knockout) clones showing the relative signature contributions for the COSMIC Single Base Substitutions signatures (indicated below the heatmap). Samples have been clustered according to their relative signature contributions. CHP134 clones are indicated in bold. Signature contributions scores are colored as indicated in the color key on the top left. The right panel shows properties of the patient/tumor (MYCN status, status [alive or death of disease], INSS stage, MUTYH status, and OGG1 status) represented by color as indicated in the legend below. KO means knockout. A germline mutation in MUTYH or OGG1 is indicated with a G.

Clinical Relevance of Defects in 8-oxoG Repair in Neuroblastoma.

To investigate potential therapeutic strategies for defects in the 8-oxoG repair pathway, we performed a compound screen (∼330 compounds; SI Appendix, Table S4 A and B) with the OGG1, MUTYH, and NUDT1 knockout CHP134 clones. No differential sensitivities compared to the CHP134 mother line were identified (SI Appendix, Fig. S8 A and B). These results indicate that defects in OGG1, MUTYH, and NUDT1 do not render CHP134 cells more sensitive to chemotherapeutics and targeted compounds. The high frequency of C > A substitutions in neuroblastoma might contribute to the adaptive capacity of the tumors. We evaluated a previously published dataset of primary and relapse neuroblastoma tumors (4). In this cohort of 23 primary-relapse neuroblastoma tumor pairs, 7 of the 15 tumor evolution events in the RAS-MAPK pathway were C > A substitutions (SI Appendix, Table S5) (4). This indicates that in this small dataset, 47% of the known single-nucleotide variants (SNVs) involved in tumor evolution are C > A substitutions.