Pan-neuroblastoma analysis reveals age- and signature-associated driver alterations

Neuroblastoma is a pediatric malignancy with heterogeneous clinical outcomes. To better understand neuroblastoma pathogenesis, here we analyze whole-genome, whole-exome and/or transcriptome data from 702 neuroblastoma samples. Forty percent of samples harbor at least one recurrent driver gene alteration and most aberrations, including MYCN, ATRX, and TERT alterations, differ in frequency by age. MYCN alterations occur at median 2.3 years of age, TERT at 3.8 years, and ATRX at 5.6 years. COSMIC mutational signature 18, previously associated with reactive oxygen species, is the most common cause of driver point mutations in neuroblastoma, including most ALK and Ras-activating variants. Signature 18 appears early and is continuous throughout disease evolution. Signature 18 is enriched in neuroblastomas with MYCN amplification, 17q gain, and increased expression of mitochondrial ribosome and electron transport-associated genes. Recurrent FGFR1 variants in six patients, and ALK N-terminal structural alterations in five samples, identify additional patients potentially amenable to precision therapy.


Point-by-point response
We thank reviewer #1 and #2 for taking the time to review our revised manuscript "Pan-neuroblastoma analysis reveals age-and signature-associated driver alterations" (NCOMMS-20-08286A). We have addressed each comment as detailed below and revised the manuscript accordingly using the Track Changes feature.
Reviewer #1 (Remarks to the Author): The authors have adequately handled most of the reviewers comments, however a major concern that still holds is the limited amount of novel data presented in the manuscript. Most of the presented data/results (although now tested on a larger samples cohort) have been published before.
Concerning comment 5 of reviewer 1: the added sentence in the manuscript should be rephrased as it is very unclear. [Response]: Our manuscript is a pan-cancer analysis of three major subgroups of neuroblastoma which compares the mutational patterns in the three subgroups defined by age of onset. By integrating data generated from previous studies with an additional >300 exome cases generated from this study, we were able to determine the prevalence of genetic alterations across the three age groups, which will serve as an important resource for the broad research community.
Regarding the specific comment 5, we apologize for the lack of clarity and agree that the statement should be re-worded. The unclear statement was as follows: "Further, FGFR1 expression was above-median for all genes across the cohort among the 169 samples analyzed by RNA-Seq (61st percentile, or a median transcripts per million (TPM) of 16.3), including in the two FGFR1-mutant samples that also had RNA-Seq, indicating that the gene is expressed in neuroblastoma." We have updated this to the following to state the FGFR1 expression in the FGFR1-mutant samples specifically: "Further, the median FGFR1 expression was ranked at the 61 st percentile of expression in this cohort (median transcripts per million (TPM) of 16.3), indicating that FGFR1 is expressed in neuroblastoma. Specifically, in the two FGFR1-mutant samples that also had RNA-Seq, FGFR1 expression was at 79.6 TPM (the 87 th percentile of expression within the sample) or 21.6 TPM (69 th percentile), and the mutant FGFR1 alleles were expressed at similar allele frequencies as in DNA (0.30 VAF in RNA vs. 0.22 in WGS, in one example patient)." Location(s) of changes: Results section "Kinase alterations in FGFR1 and truncated ALK variants" (lines 199-204).
According to Figure 1, several samples present without any genomic aberration (copy number or mutation): what is the percentage of tumor cells in these cases? [Response]: A total of 136 samples (out 685) lacked recurrent SNV, indel, structural or copy alterations reported in Fig. 1, and also had no whole-chromosome copy alterations. We were able to collect pathology review for 130 of these cases and found that tumor purity was 60% or above for all cases. These cases were enriched in low-stage tumors. We added the following text in the revision "Of the 685 samples with WGS or WES, 136 (20%) lacked any of these recurrent alterations and had no whole-chromosome copy alterations. These 136 samples were enriched in low disease stage (only 19% were stage 4 compared to 70% of other samples, P < 2.2 x 10 -16 ) and their paucity of somatic alterations was not caused by low tumor purity-pathology review, available for 130 out of the 136 samples, showed that tumor purity exceeded 60% in all cases." Location(s) of changes: Results section "Landscape of somatic mutations in neuroblastoma" (lines 160-165).
Reviewer #2 (Remarks to the Author): Thank you for asking me to re-review the paper, this time in a different Nature journal. The authors now seem to acknowledge that most of their observations merely confirm known features of the neuroblastoma genome. There remain specific technical and semantic concerns that I have (see discussion further below). Leaving those aside, the issue of novelty, or rather the lack thereof, remains. In the introduction of their rebuttal, the authors highlight three findings that they consider to be novel: A) "Our study provides a comprehensive view of the driver alteration landscape in age-dependent subgroups which had only been reported at a smaller scale in both the sample number and number of events." -What the authors have achieved is to increase the confidence interval around the frequency of drivers. I would not consider this to be a substantial shift in our understanding of the drivers of neuroblastoma. [Response]: Our manuscript is a pan-neuroblastoma analysis of three major subgroups of this disease, which compares the mutational patterns in the three subgroups defined by age of onset. The sample size is an important factor in ensuring the rigor of analysis so that the harmonized data and the results can be an important resource for the research community. B) "FGFR1 is a known cancer gene but its role as a neuroblastoma driver gene is first established through this study as previous studies only reported a singleton event, thus lacking the statistical power to assess the significance of this alteration." -The authors still do not acknowledge in their paper that this mutation has been described before. Furthermore, there is no mentioning of the fact amplification of FGFR3 is a known feature of olfactory neuroblastoma. Implicating FGFR signalling in neuroblastoma would therefore not appear to be a novel revelation. Even if this observation were novel, it would still only be relevant in a tiny proportion of tumours (1%). It does not constitute a significant shift in our understanding of the driver mutations underpinning neuroblastoma. [Response]: The reviewer states that "The authors still do not acknowledge in their paper that this mutation [FGFR1] has been described before." However, both our original and revised manuscripts cite several studies shown that FGFR1 mutations have been reported as recurrent in other cancer types, and in a single neuroblastoma tumor, including the following statement: "FGFR1 N546K was previously reported in a single neuroblastoma patient 1 and therefore has not been considered a driver gene in neuroblastoma. This variant activates MAPK signaling in functional studies from other tumor types 2,3 , and is recurrent in pediatric low-grade glioma 4,5 , indicating it is a driver mutation." Olfactory neuroblastoma (or esthesioneuroblastoma), despite its misleading name, is a clinically and molecularly distinct entity from neuroblastoma. Olfactory neuroblastomas arise from sensory olfactory neuroepithelium [6][7][8] and occur most frequently in adults 9 , whereas neuroblastoma is of neuroendocrine origin 10 and occurs primarily in children 11 . The histology, epidemiology, staging and therapy likewise differ. Further, olfactory neuroblastomas are molecularly distinct from neuroblastoma, lacking MYCN 12 or 3 ALK alterations 13 . Therefore, we feel that FGFR3 amplification in this distinct tumor entity is not relevant to neuroblastoma, and discussing it in the context of our findings would lead to confusion in the narrative. C) "We show for the first time the association of mitochondrial gene expression to mutational signature 18 in neuroblastoma, a significant finding on the potential mutagenic processes of neuroblastoma." -There remain several issues with this association; mainly that it is not adjusted for MYCN amplification. As MYCN amplified tumours have more signature 18 mutations, they are likely to proliferate faster and thus have more mitochondrial usage. Therefore, I suspect that the association will disappear once adjusted for MYCN status. Furthermore, this finding is somewhat undermined by the recent publication of this paper here from the authors -(https://www.nature.com/articles/s41467-020-14682-6) -which circumscribes the association of mitochondrial genes expression and ROS-associated mutations without using the word signature 18.
[Response]: We appreciate this suggestion. In our revised manuscript we have performed the differential gene expression analysis while including MYCN alteration status as a covariate, thus adjusting for its effects (new Supplementary Fig. 18B). The mitochondrial gene expression remained statistically significant after this adjustment, and we have also shown the expression of a few example mitochondrial genes in signature 18-positive vs. -negative samples including only samples lacking MYCN alterations to show that the mitochondrial expression increase remains regardless of MYCN. This is consistent with the fact that several of the mitochondrial genes are found on chromosome 17q, and 17q gains are associated with signature 18 independent of MYCN ( Supplementary Fig. 18A, which was included in both the previous and revised manuscripts).
Further, when including signature 18 status, MYCN status, and 17q gain status all as covariates, signature 18 was not significantly associated with the expression of any gene, and 17q gains but not MYCN were associated with increased mitochondrial gene expression (new Supplementary Fig. 18C, D). This indicates that the link between signature 18 and mitochondrial gene expression is due to 17q gains and not MYCN. We have included this analysis in the revised manuscript. We note here that MYCN's primary effect on ROS is through inducing glutaminolysis, which depletes ROS-protective glutathione, as we and others have shown 14,15 . Thus, 17q gains may increase mitochondrial ROS through gains of mitochondrial genes, while MYCN decreases protection from ROS, which could plausibly explain the additive effect of the two alterations on signature 18 ( Supplementary Fig. 18A).
Regarding the study noted by the reviewer, we note that it was cited in our previous version of the manuscript as follows: "MYCN-altered samples had significantly more signature 18 (Fig. 4E), consistent with recently reported MYCN-induced ROS generation in neuroblastoma 15 ." We would like to point out that the ATRX-MYCN exclusivity study, published by our group, is a functional study which did not perform any mutational signature analysis. Specifically, it did not make any reference to ROS-associated mutagenesis nor signature 18, but focused only on the DNA damage response associated with MYCN-induced ROS. Further, as the reviewer noted during the previous revision, signature 18 and ROS have not been definitively linked in neuroblastoma; thus, our finding that MYCN is associated with signature 18 is distinct from (though potentially related to) that study's focus on MYCN-induced ROS. Finally, the association of 17q gain with signature 18, including the expression of mitochondrial genes found on 17q (ATP5G1, ICT1, and MRPS7, for example), has not been noted in the study in question nor any previous study of which we are aware.
Locations of change(s): Results section "Mutational signature 18 is associated with increased mitochondrial gene expression, 17q gains, and MYCN" (Lines 336-345), Supplementary Fig. 18. 1.3. I think the authors' language ought to be much stronger. These changes have not just been previously identified as "recurrent drivers". These are fundamental features of the neuroblastoma genome that determine risk and thus treatment. [Response]: We note that not all recurrent alterations shown in Fig. 1 are associated with risk and treatment. Of the lesions shown, only segmental copy number changes in chromosomes and MYCN are used to routinely determine risk group designation. Only ALK is used currently to select specific therapeutics, and while TERT alterations are shown to correlate with poor risk, they have not been incorporated into current risk classification schemas. Other alterations identified, including SHANK2 and PTPRD, are not used for clinical risk classification, nor are they associated with survival, as we have shown. Therefore, we do not feel it is warranted to make the statement suggested by the reviewer in this section, which is focused on understanding the recurrent mutational landscape of neuroblastoma.
The information regarding the prognosis of specific driver alterations has already been noted at other more appropriate locations in the manuscript, including the Introduction, and our systematic analysis of survival correlations in the Results section. Please note, for example, the following statements: 1. From the Introduction: "Whole chromosome gains are frequently observed in low-risk neuroblastoma 16 , while gains or losses of chromosome arms (segmental chromosome alterations), including loss of 1p, 3p, 4p and 11q and gain of 1q, 2p and 17q 17,18 , are associated with poor prognosis 19 . MYCN amplification is the most frequent driver in neuroblastoma, occurring in ~20% of cases and conferring poor prognosis [20][21][22] ." 2. From the Results: "MYCN, ALK, ATRX, and Ras pathway alterations; segmental deletion of 1p, 3p, and 11q; and segmental gain of 1q, 2p, 7q, 11q13.3, 12q, and 17q were each significantly associated with poor overall survival ( Supplementary Fig. 9), consistent with previous reports 20, [23][24][25][26][27] .
3. From the Results: "FGFR1 and TERT alterations trended towards an association with worse survival but were not statistically significant (P = 0.09 and P = 0.1, respectively; Supplementary  Fig. 9). In previous studies, TERT has been associated with both worse survival 28 and no difference in survival 29 . These reported findings, together with ours, suggest that TERT alterations alone have a modest effect on patient outcomes." The last-noted statement about TERT alterations shows that nuance is needed in describing the prognostic significance of each alteration. Thus, we feel that making a statement that all Fig. 1 driver alterations are associated with prognosis and treatment would be inaccurate, repetitive of information presented in the Introduction, untimely in a section focused only on the genomic landscape, and deserves its own detailed Results description as we have done.
3.1. I appreciate the authors' efforts here. How did the authors determine whether rearrangements were independent or not? Was this done informatically or manually? The issue that remains is that none of the associations, as the authors highlight themselves, is novel. In particular the mutual exclusivity of MYCN and ATRX has now been reported on by the authors themselves (https://www.nature.com/articles/s41467-020-14682-6).
[Response]: This analysis was done manually by verifying the computational predictions generated from WGS. Using visualization tools that we developed 30,31 , we evaluated the junction reads across the structural variation breakpoints as well as read-depth changes in computationally predicted SVs and CNVs for every patient. This enabled us identify whether CNVs and SVs were joined: (a) directly (where a translocation directly joins the termini of two copy alterations, for example) or (b) through intervening chromosome sequences with sequential structural variants 15 Mb or less apart. Details are provided in the Methods section.
As an example of (b), in one patient a 1p deletion and a 17q gain were apparently linked together, but not directly-an intervening segment of chromosome 1q was a likely bridge between the two. Since all three breakpoints (1p to 1q to 17q) could be joined in a theoretical consensus contig with the intervening 1q sequence less than 15 Mb in length, we considered that the 1p deletion and 17 gain were likely linked and removed the co-mutation of the two in that patient from the statistical calculation.
4.1. This is an odd response. FOXO1 fusions are canonical, disease defining events in alveolar RMS, which are typically caused by balanced translocations (mostly) without a wildly rearranged genome. I am not sure what this has to do with the point I raised.
Based on the numbers the authors present it would seem important that they highlight in the manuscript that their classification of NB by age is, although broadly speaking "ok", imprecise. A stage 4 tumour should not be lumped together with low risk tumours in Group A. This needs to be highlighted to the reader. The numbers also highlight what the problem was. This study is not large enough to demonstrate that SHANK2 rearrangement are significantly less prevalent in low risk tumours which are generally less rearranged than high risk tumours. 1 [Response]: To address the reviewer's concern, we have analyzed driver variant/age group associations among stage 4 samples only, in addition to all samples (revised Supplementary Fig. 15B). When doing this, the PTPRD, t(11;17), 2p gain, and Ras pathway age associations became non-significant, while all other age associations noted in the manuscript (MYCN, ATRX, TERT, 1p deletion, etc.) remained statistically significant. We have noted this in the revised manuscript.
4.2. I disagree with this line of reasoning, mostly because I think some assumptions are invalid. Perhaps let us simplify the discussion here. Can the authors assess the enrichment of PTPRD breakpoints across all samples constructing a statistical model that adjusts for the following confounders: age; MYCN status; rearranged-ness (diploid/tetraploid VS everything else). I suspect that they will not find an enrichment of PTPRD variants in any group. [Response]: As noted in response to the previous comment, we have analyzed driver variant/age group associations among stage 4 samples only, in addition to all samples (revised Supplementary Fig. 15B). As hypothesized by the reviewer, the PTPRD age associations became non-significant as there were only 11 group A (<1.5 years) samples with WGS. We have noted this in the revised manuscript, and the revised statement reads as follows: "By contrast, PTPRD genetic alterations (consisting of gene-disrupting SVs and focal deletions) were significantly higher in groups B and C as they were completely absent from group A ( Supplementary Fig.  15A, B), and Ras pathway mutations were enriched in group B, although there was no significant age group difference for PTPRD and Ras pathway mutations when including only stage 4 samples, suggesting their age specificity was related to higher disease stage (Supplementary Fig. 15B)." Locations of changes(s): Results section "Age-related genomic aberrations" (lines 258-260).

5.2.
It would seem that the authors have not fully appreciated the study that I cited. That particular experiment went through rounds of subcloning individual cells to segregate culture artefacts from signature 18 mutations generated by culture-independent mutational processes. [Response]: We respectfully disagree that subcloning cells in culture is able to "segregate culture artefacts from…culture-independent mutational processes" and we see no claim in the study 32 to this 6 effect. Cell culture artefacts will continue to be generated in cell culture models regardless of the experimental manipulation performed. Indeed, the study in question 32 shows that signature 18 was generated in cell lines from tissues not normally generating signature 18, even when performing subcloning. For example, their Figure 3 shows that BT474 and AU565 breast cancer cell lines generate signature 18, while patient breast tumors do not usually generate signature 18 33 . Further, the authors state that while signature 18 was detected continuously in the two neuroblastoma cell lines studied using their sequential subcloning experiment (NB13 and BE2-M17), signature 18 was also observed as a feature of cell culture in general in the following direct quotation (underline added): "SBS18 [Signature 18] is prominent in neuroblastoma (Alexandrov et al., 2013a) and continued to be generated in all of the neuroblastoma cell lines examined (Figure 3). It was also, however, observed in many daughter clones that were whole-genome sequenced (and thus captured sufficient numbers of mutations) of cell lines in which it was not detected in stocks (Figure 3). It therefore appears to be a common feature of in vitro culture, as previously noted (Rouhani et al., 2016). SBS18 may be generated by DNA damage caused by reactive oxygen species (Viel et al., 2017), and this mechanism could plausibly mediate its manifestation as a consequence of in vitro cell culture 32 ." Further, another study 34 analyzing signatures in iPS cells treated with various perturbations also noted that signature 18 is generated under basal cell culture conditions, as follows: "Last, the ubiquitous background signature present across the control [single-cell iPS clones sequenced after solvent control treatment] samples is similar to COSMIC Signature 18, previously hypothesized to be due to ROS 34 ." Therefore, we feel it worthwhile to show that signature 18 is generated continuously in actual patient tumors, which complements this study of only two neuroblastoma cell lines which explicitly states that signature 18 is a common feature of cell culture, including when performing single-cell subcloning. We maintain that the following statement in our manuscript is valid, and the use of the term "suggested" is warranted, given the caveats stated in the study 32 itself, and its use of only two neuroblastoma cell lines: "We found that signature 18 may be both an early event in neuroblastoma, causing truncal mutations, and an on-going mutational event causing relapse-specific mutations, based on analysis of five patients with matched diagnosis and relapse samples (Fig. 4B), as suggested previously by cultured neuroblastoma cell models 32 ." 5.4. The issue remains that what the authors say that have found, as association of MYCN alterations with signature 18, has been shown before, as the authors themselves recognise (see reference 80 they quoted): "MYCN-altered samples had significantly more signature 18 (Fig. 4E), consistent with recently reported MYCN-induced ROS generation in neuroblastoma." [Response]: As noted by the reviewer during the previous revision, signature 18 has not been explicitly linked to ROS in neuroblastoma. Nor did the study in question 15 analyze MYCN-induced mutagenesis or signature 18, but only ROS generation and the resulting DNA damage response 15 . Further, the main novelty of the signature 18 correlations is to show that 17q gains (and increased expression of 17q mitochondrial genes), is associated with increased signature 18, independent of MYCN status ( Supplementary Fig. 18).
6.1. As I was writing the paragraph initially, I was wondering whether the authors would rebut this using the well known anomaly of BRAF V600E mutations. What I am asking the authors to do is to statistically assess whether signature 18 is causing more driver events that one would expect to find by chance, given the prevalence of signature 18 mutations in each sample. The authors need to consider in their model the actual trinucleotide context and amino acid change required to generate the mutation in question. It would be helpful if the authors were able to rebut my point with data analyses. [Response]: In the revised manuscript, we have noted that there is no significant difference between the percentage of driver mutations caused by signature 18 (52%) compared to the percentage of all mutations genome-wide which were caused by signature 18 (56%). The updated statement is as follows: "The percent of driver SNVs most likely induced by signature 18 (52%) was similar to the percentage of all mutations caused by signature 18 across the 38 samples analyzed (56%, P = 0.64 by Fisher's exact test), indicating that signature 18 was proportionally likely to cause driver SNVs as SNVs in general. This indicates that signature 18 is likely a driver of disease progression in neuroblastoma, in contrast with passenger mutational signatures, such as the kataegis-associated APOBEC signature in osteosarcoma which causes no known driver SNVs in that cancer type 35 ." Location(s) of changes: Results section "Driver mutations associated with signature 18" (lines 392-397).
Thank you for asking me to review this paper one more time. I continue to question the novelty of this paper which merely confirms existing knowledge. I note that Reviewer 1 has the same fundamental concern.
The key issues are (lettering as per rebuttal document): A) The authors continue to insist that their paper merits publication as a substantial resource. At a sample size of n=702 with the MAJORITY being exomes (n=539), this would not be considered an impressive resource in the context of childhood cancer genomics study. The authors are portraying only 205 genomes. Note that in 2012 (!) Molenaar et al had already reported 87 neuroblastoma genomes. A substantial meta-analysis would be a 2017 (!) study of ~500 medulloblastoma genomes (Northcott et al). Perhaps the authors may understand therefore that their study does not appear to be an impressive resource. Another issue with the WGS data of this study is that the WGS is mainly Complete Genomics (i.e. non-Illumina) data, which, as the authors know, is of limited use to the community.
C) The authors had stated that key novelty was: "We show for the first time the association of mitochondrial gene expression to mutational signature 18 in neuroblastoma, a significant finding on the potential mutagenic processes of neuroblastoma." After the adjustments I had asked the authors to perform, this aspect of the analysis has become more questionable: C.1) Association of MYCN/17q gain with signature 18 mutations: To me this association looks like it can be explained by "rearrangedness" (which wound account for MYCN and 17q). Can the authors test whether the degree of "rearrangedness" (eg, number of breakpoints, degree of segmentation) would account for signature 18 mutations?
C.2) In testing the associations of various genomic features with signature 18, how many hypotheses have the authors tested? Is there multiple hypotheses correction?
C.3) As some of the "mitochondrial" genes that are associated with 17q gain reside on 17q, the authors need to show that the expression of 17q "mitochondrial" genes is enriched compared to "non-mitochondrial" 17q genes. Otherwise it would seem very questionable to claim that gains of 17q are associated with increased expression on 17q.
C.4) Looking at Figure 4C, it is not clear to me how the authors have identified "mitochondrial*" genes as significantly associated with signature 18 mutations. Where do the authors show that these are significantly enriched amongst the mRNAs shown in Figure 4C?
All other aspects of the paper reiterate existing knowledge, as previously detailed and graciously acknowledged by the authors.

RESPONSE TO REVIEWER COMMENTS
Reviewer #2 (Remarks to the Author): Thank you for asking me to review this paper one more time. I continue to question the novelty of this paper which merely confirms existing knowledge. I note that Reviewer 1 has the same fundamental concern.
The key issues are (lettering as per rebuttal document): A) The authors continue to insist that their paper merits publication as a substantial resource. At a sample size of n=702 with the MAJORITY being exomes (n=539), this would not be considered an impressive resource in the context of childhood cancer genomics study. The authors are portraying only 205 genomes. Note that in 2012 (!) Molenaar et al had already reported 87 neuroblastoma genomes. A substantial meta-analysis would be a 2017 (!) study of ~500 medulloblastoma genomes (Northcott et al).
Perhaps the authors may understand therefore that their study does not appear to be an impressive resource. Another issue with the WGS data of this study is that the WGS is mainly Complete Genomics (i.e. non-Illumina) data, which, as the authors know, is of limited use to the community.
[Response]: We respectfully disagree that Complete Genomics (CGI) data "is of limited use to the community." Our pan-pediatric landscape paper studying 1,699 pediatric cancers (Ma et al., Nature 2018) included 654 CGI-based WGS samples and has been cited 186 times in just over two years. In that paper, we documented extensive data cleaning techniques that make CGI data highly usable and allow effective detection of SNVs, indels, structural variants, and copy number alterations, which was incorporated into this study. We also note that 58 of the 205 WGS samples were obtained with Illumina sequencing. The reviewer also notes the Molenaar et al. study as an example; we would like to point out that this study also used CGI data for all 87 samples.
While the exome data do not enable identification of structural variants, they do allow sensitive and specific detection of large copy number alterations as documented in the Methods (59 samples had both WGS and exome data, enabling a comparison of the two platforms). The combined copy number, SNV/indel, and survival information (shown in Supplementary Table 1, with raw data also accessible on dbGaP and EGA) for this large cohort offer a valuable resource for the cancer research community.
C) The authors had stated that key novelty was: "We show for the first time the association of mitochondrial gene expression to mutational signature 18 in neuroblastoma, a significant finding on the potential mutagenic processes of neuroblastoma." After the adjustments I had asked the authors to perform, this aspect of the analysis has become more questionable: C.1) Association of MYCN/17q gain with signature 18 mutations: To me this association looks like it can be explained by "rearrangedness" (which wound account for MYCN and 17q). Can the authors test whether the degree of "rearrangedness" (eg, number of breakpoints, degree of segmentation) would account for signature 18 mutations? [Response]: We have performed additional analysis limiting the signature 18 correlation analysis to stage 4 samples only. When doing this, 17q gains and MYCN alterations remain statistically associated with signature 18 (new Supplementary Fig. 18). Further, the structural variant burden was correlated very weakly with the signature 18 burden (Pearson r 2 = 0.05). We therefore conclude that the signature 18 correlations with 17q gain and MYCN are not related to the degree of rearrangements in neuroblastoma.
We have included this information in the revised manuscript.