Review Reports

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Comments and Suggestions for Authors:

The abstract is well organized. The objective, data source, methods, and results are clearly presented; however, the conclusion could be more precise.

Line 89-highlights the research gap and study aim very clearly.

Line 97-should provide a URL if possible.

Line 117: Log-rank tests were used to analyze survival rates, which need reference to the exact version that archives current results.

Line 194: Need a bit more detail about the selection of the 15 most significant co-occurrence features and how they correlated.

The overall analysis and presentation of the results of genomic patterns are very good.

A single database has been used; therefore, an external validation could enhance methodological reliability. 

The six limitations are clearly presented and provide valuable information for future studies.

Author Response

Comments 1: [This study summarized the clinical and molecualr finding of astrocytoma, IDH mutant, cns who grade 3 using the AACR genie database and cbioportal. The study have limitations, which as described at the end of conclusions. Authors could describe in more details the material amd methods of the series, which I understand are explained in the cbioportal website. ]
Response 1: Thank you for this suggestion. We have responded with the following in lines 108-112: cBioPortal uses the mutation calls provided by each contributing publication, with little additional filtering except for standardization of mutation annotations using Genome Nexus. [28-30] cBioPortal does filter out non-synonymous mutations, including Silent, Intron, IGR, 3'UTR, 5'UTR, 3'Flank and 5'Flank mutations, with the exception for mutations in the promoter region of the TERT gene.  

 

Additional comments:

Comments 1: [(1) The classification of gliomas is based on the combination of histopathologic characteristics (H&E and immunophenotype) and molecular variables. Nowadays, it is recommended several key molecular tests, including IDH1/2 mutation, 1p/19q deletion, ATRX and TP53 mutations, CDKN2A/B deletion, etc. However, not all departments of pathology may have all the techniques available. How is this solved? Should this be discussed in the Introduction?

]
Response 1: Thank you for this suggestion. We have added the following to the introduction from lines 58-61: Pathology laboratories that don’t have access to the required technology to perform these molecular characterizations must rely on previous WHO guidelines for classification of IDH-mut A3 (previously Anaplastic Astrocytoma), which was a wholly histological approach). In connection to this point, we emphasized that because GENIE was established before the updated 2021 WHO molecular diagnostic criteria guidelines, tumors are annotated with the outdated anaplastic astrocytoma label, which adds value to our study by taking these tumors and filtering for those that have the correct molecular criteria. Lines 97-102.  Notably, because the GENIE database was established prior to the 2021 WHO reclassification, tumors are still annotated under the outdated term “anaplastic astrocytoma,” and cannot be readily filtered by updated molecular criteria. Accordingly, we also aim to generate a practical and up-to-date reference of the mutational landscape of IDH-mut A3 based on modern diagnostic definitions. We also added to our limitations section by saying that our total number of samples included in the analysis is smaller than the total number of samples labeled as anaplastic astrocytoma, in part because of samples that were missing segmentation data (whether because the centers were unable to profile for that or because they just didn’t send it because it wasn’t a required file). Lines 307 - 311 Furthermore, segmented copy number data, which was used to exclude 1p/19q samples, is not a required aspect for submission to the GENIE database, which means that some samples may have been included that have the IDH-A3 exclusing characteristic of a 1p/19q co-deletion, but was impossible to detect due to the data not being submitted.

Comments 1: [(2) In the Introduction, could you please show the diagnostic algorithm of the diffuse gliomas in adults? As I understand, the target of this study is "astrocytoma, IDH-mutatnt, CNS WHO grade 3". ]
Response 1: Thank you for your response. We have responded by outlining the diagnostic algorithm in lines 55-59:  Astrocytoma, IDH-mutant has several essential criteria, including being a diffusely infiltrating glioma, having an IDH1/2 hotspot mutation, and absence of combined whole arm deletions of 1p and 19q. In addition, loss of ATRX expression or ATRX mutation supports the diagnosis and obviates the need for 1p/19q status assessment. 

 

Comments 1: [(3) Line 102. Please add the R version.]
Response 1: Thank you for your comment. We have added the version in line 117: version 4.4.2

Comments 1: [(4) Please add and cite the R packages that were used.]
Response 1: Thank you for your comment. We added the following in lines 116-119: Statistical analyses were conducted using R/RStudio (R Foundation for Statistical Computing, Vienna, Austria) version 4.4.2, with the cbioportalR, tidyR, knitr, survival, dplyr, GenomeInfoDb, IRanges, S4Vectors, BiocGenerics, and data.table packages [33-47].

 

Comments 1: [(5) Lines 113-121. Did you calculate an overall survival for each gene? How many genes in total?]
Response 1: Thank you for your response. Based on your and other reviewer’s responses, we decided to calculate overall survival only for the top 10 most frequently mutated genes. Before we calculated survival for every single gene, but then needed to apply multiple hypothesis test corrections and lost much of the signal. Here is our updated methods in lines 146-150: 

Survival analyses were performed by generating Kaplan-Meier curves for each gene interrogated within the cohort. For each gene, two curves were constructed: one representing patients harboring a mutation in the specified gene, and the other representing patients with the wild-type allele. Statistical significance between survival distributions was assessed using log-rank tests for the top 10 most frequently mutated genes in the cohort.

 

Comments 1: [(6) Was survival time calculated from time of diagnosis to the last visit time/death?]
Response 1: Thank you for your comment. The GENIE database does not include a specific variable for time of diagnosis. Therefore, we used the patient's age at the time of sample sequencing as a proxy for time of diagnosis. Survival time was calculated as the interval between this proxy for diagnosis and the time of death or last follow-up. Notably, the time of death is provided in GENIE as an age in years (measured from birth), not as an interval from diagnosis. Accordingly, survival time was derived by subtracting the age at sequencing from the age at death or last contact. We have clarified this methodology in the revised manuscript. See lines 151-156: Because there GENIE database does not provide a specific variable for time of diagnosis, we used the patient's age at the time of sample sequencing as a proxy for time of diagnosis. Survival time was calculated as the interval between this proxy for diagnosis and the time of death or last follow-up. Importantly, time of death is provided in GENIE as an age in years (measured from birth), not as an interval from diagnosis. Accordingly, survival time was derived by subtracting the age at sequencing from the age at death or last contact.

Comments 1: [Line 121-123. Did you merge results of different samples for same patient? Should the diagnostic samples only be used? Are all cases diagnostic biopsies?.]

Response 1: Thank you for your response. After applying our filtering criteria, we uploadd the sample IDs into the cbioPortal website to calculator frequency. The tables in the cbioPortal website show mutations on a per-sample basis, meaning the samples are not merged. The exception is with the Kaplan-Meyer survival curves (which we calculated separately from cbioPortal), in which a patient is considered as having a mutation in a gene if any one of their samples have a mutation in that gene. We modified the end of line 110 to make this more clear (see lines 106-110) : This investigation utilized data procured from the publicly accessible, de-identified American Association of Cancer Research (AACR) Project GENIE database. The cBioPortal for Cancer Genomics platform was employed for data retrieval and subsequent analysis, encompassing the determination of frequencies for gene mutations, copy number alterations, and structural variants, all of which are displayed on a per-sample basis. Also see lines 157-161: To account for individuals with multiple samples, mutation counts for each gene were aggregated across all samples from a given patient. The overall frequency of gene alteration on a per-patient basis was determined by classifying a gene as altered if its total count of detected alterations across all samples for that patient was greater than one.

 

Comments 1: [(8) Line 143. Should the IDH mutation be 100% instead of 99.7%]
Response 1: [Thank you for your response. The 99.7% comes from IDH1, whereas IDH2 makes up the remaining 0.3% of the cohort]”

 

Comments 1: As I understand, the cBioPortal data was used. Please explain the material and methods of the molecular techniques. For example, the definition of mutation, type of array, filtering of high-confidence pathological mutations, annotation, etc.]
Response 1: Thank you for this suggestion. We have responded with the following in lines 108-112: cBioPortal uses the mutation calls provided by each contributing publication, with little additional filtering except for standardization of mutation annotations using Genome Nexus. [28-30] cBioPortal filters out non-synonymous mutations, including Silent, Intron, IGR, 3'UTR, 5'UTR, 3'Flank and 5'Flank mutations, with the exception for mutations in the promoter region of the TERT gene.  

 

Comments 1: [(10) Letters of Figure 1 are very small. Please make them bigger in the final version.]
Response 1: Thank you for your comment. We will upload the figure into the journal’s system which should allow a larger displayable size. 

 

Comments 1: [(11) Line 160. What is the definition of amplification?]
Response 1: Thank you for your response. We define amplification as a high-level increase in the copy number of a specific genomic region, typically scored as 2 in copy-number analysis algorithms like GISTIC. We have added clarification in the manuscript to define this in lines 123-126: Amplifications as given in the cBioPortal CNA genes table were defined as genomic regions with a score of 2 with the copy-number analysis algorithm (e.g. GISTIC) employed by cBioPortal, representing a high level increase in the copy number of a specific gene region. Deep deletions (homozygous deletions) were define as genomic regions with a score of -2.

Comments 1: [ATRX was truncating mutation. Does it counts as "loss" for the diagnostic algorithm?]
Response 1: [Thank you for your response. To our knowledge ATRX is generally evaluated immunohistochemically, but that a truncating mutation would count as a loss for the diagnostic algorithm]. We have updated our limitation section to address this point (see lines 312-319:)This investigation is subject to several significant limitations that warrant acknowledgment. Firstly, the analytical scope was confined to genomic data, specifically DNA sequence alterations and copy number variations. The absence of integrated transcriptomic, epigenomic (e.g., methylation arrays), proteomic, or microRNA expression data inherently limits the capacity to draw comprehensive conclusions regarding the functional consequences of the observed genomic alterations and the resultant tumor phenotypes. The lack of immunohistochemistry data also makes it more difficult to infer ATRX loss in the diagnostic algorithm.

Comments 1: [Please confirm that Table 4 is correct]
Response 1: Thank you for your response. We have confirmed that the table is correct. 

 

Comments 1: [Please define "enriched genes" in metastasis in comparison to primary sites.]
Response 1: [Thank you for your response. We have modified the manuscript to define these enriched genes: in lines 245-247:”In this analysis, a gene was considered enriched if the proportion of metastatic samples (as pre-annotated in the GENIE database) harboring a mutation in that gene was significantly greater than the proportion observed in primary samples, based on statistical comparison in the two-sided Fischer exact test.]

 

(15) What genes of Table 2 were used in the survival analysis? Please show the survival curves. If you are just analyzing the genes of Table 2, then you may not need multiple comparision correction.

Comments 1: [What genes of Table 2 were used in the survival analysis? Please show the survival curves. If you are just analyzing the genes of Table 2, then you may not need multiple comparision correction..]
Response 1: Thank you for your response. Originally, all genes were used in the survival analysis, but upon your suggestion we agree that it is better to just include the most frequently mutated genes as found in Table 2. See our following modification and subsequent kaplan meier plots in lines 262-268:Out of the 15 most frequently mutated genes, the only two genes with significantly decreased median patient survival time compared to non-mutated genes were BCOR and KMT2D (See Table 5).

Reviewer 2 Report

Comments and Suggestions for Authors

This study summarized the clinical and molecualr finding of astrocytoma, IDH mutant, cns who grade 3 using the AACR genie database and cbioportal. The study have limitations, which as described at the end of conclusions. Authors could describe in more details the material amd methods of the series, which I understand are explained in the cbioportal website. 

Additional comments:

(1) The classification of gliomas is based on the combination of histopathologic characteristics (H&E and immunophenotype) and molecular variables. Nowadays, it is recommended several key molecular tests, including IDH1/2 mutation, 1p/19q deletion, ATRX and TP53 mutations, CDKN2A/B deletion, etc. However, not all departments of pathology may have all the techniques available. How is this solved? Should this be discussed in the Introduction?

(2) In the Introduction, could you please show the diagnostic algorithm of the diffuse gliomas in adults? As I understand, the target of this study is "astrocytoma, IDH-mutatnt, CNS WHO grade 3".

(3) Line 102. Please add the R version.

(4) Please add and cite the R packages that were used.

(5) Lines 113-121. Did you calculate an overall survival for each gene? How many genes in total?

(6) Was survival time calculated from time of diagnosis to the last visit time/death?

(7) Line 121-123. Did you merge results of different samples for same patient? Should the diagnostic samples only be used? Are all cases diagnostic biopsies?

(8) Line 143. Should the IDH mutation be 100% instead of 99.7%

(9) As I understand, the cBioPortal data was used. Please explain the material and methods of the molecular techniques. For example, the definition of mutation, type of array, filtering of high-confidence pathological mutations, annotation, etc.

(10) Letters of Figure 1 are very small. Please make them bigger in the final version.

(11) Line 160. What is the definition of amplification?

(12) ATRX was truncating mutation. Does it counts as "loss" for the diagnostic algorithm?

(13) Please confirm that Table 4 is correct

(14) Please define "enriched genes" in metastasis in comparison to primary sites.

(15) What genes of Table 2 were used in the survival analysis? Please show the survival curves. If you are just analyzing the genes of Table 2, then you may not need multiple comparision correction.

Author Response

Comments 1: [Thoroughly stratify your study cohort. As you mention in the limitations there are many different institutes involved and by this also many different pathologists which naturally leads to some variation. In addition, the time frame of this collection spans at least two different WHO classifications which include major changes on the diagnosis of IDH-mut glioma. Therefor it makes sense to include as many stratification criteria as possible to obtain the most stringent classification possible. Methylation data would be best, but you already mentioned that this is not available for the cohort. But CNV data is available and by this excluding 1p/19 LOH and the 7p gain / 10q loss signatures would be very beneficial. Also, if histological data is available this would help significantly, as ATRX loss is detectable by immunohistochemistry and could thereby close the gap of ATRX cases.

]
Response 1: Thank you for your response. We better explained the filtering procedure and updated our analysis to exclude 1p/19 LOH and 7 gain/10 lost signatures.  : Due to the diagnostic requirement for both molecular and histopathological criteria in identifying IDH-mutant CNS WHO grade III astrocytomas, and given that histological slides are not available in the AACR GENIE database, we relied on the pre-annotated histological classifications provided. Specifically, we included cases labeled as anaplastic astrocytoma, as this designation most closely corresponds to IDH-mutant WHO grade III astrocytomas in the 2021 WHO CNS classification system. This initial filter identified 678 patients. We then refined the cohort by selecting only tumors with a confirmed mutation in either IDH1 or IDH2. To avoid inclusion of higher-grade tumors, we excluded cases with homozygous deletions in CDKN2A or CDKN2B. Additionally, tumors with 1p/19q co-deletion were excluded, based on the methodology described by Williams et al. [48] and using genomic coordinates obtained from the UCSC Genome Browser [49]. We also excluded samples with both a gain in chromosome 7 and loss in chromosome 10 (+7/-10), as to further exclude possible glioblastoma multiforme tumors that were incorrectly classified as IDH-mut A3. After applying these criteria, the final study cohort consisted of 391 patients and 399 tumor samples.

Unfortunately to our knowledge histological data is not available beyond the initial histological classification in the database, so we included that in our limitations section. Another limitation is the lack of histological and immunohistochemistry data, which would have further aided in the detection of ATRX loss and more robust filtering of our samples. 

Comments 1: [In line 72 you mention that astrocytoma progress into glioblastoma. This is not possible based on the WHO 2021, as all IDH mutant tumors with astrocytic genotype are considered to be astrocytoma. They progress into grade IV astrocytoma. In the same paragraph you also describe secondary glioblastoma to be a different entity, which is not the case. The name secondary glioblastoma should not be used anymore, as these are astrocytoma by WHO 2021 definition. This difference in naming was chosen deliberately to pinpoint the difference in biology and clinical behaviour when compared to glioblastoma. The whole paragraph (line 72-82) is somewhat misleading due to this.]

Response 1: Thank you for your response. You are correct and we thank you for pointing out our error. Please see the update paragraph: Given enough time, IDH-mut A3 will progress into Astrocytoma, IDH-mutant CNS WHO grade 4 (IDH-mut A4), previously known as glioblastoma, IDH-mutant in the 2016 WHO classification. The updated classification reflects the observation that grade 4 IDH-mutant astrocytomas have different responses to treatment and a different mechanism of gliomagenesis than their glioblastoma IDH-wildtype counterparts [11,30]. Unlike IDH-mutant glioblastomas which commonly have promoter mutations in telomerase reverse transcriptase (TERT) and epidermal growth factor receptor (EGFR) amplification [12], IDH-mut A3 evolves into a IDH-mut A4 through homozygous deletions in CDKN2A/B [14]. The prognosis with CDKN2A/B homozygous deletions is so poor that WHO classifies any IDH-mut A3 with deletions in CDKN2A/B as IDH-mut A4 even in cases without necrosis or neovascularization [1]. Nevertheless, IDH-mut A4 tumors generally have longer survival than IDH wild type glioblastomas[14].

Comments 1: [You mention that TP53 and TERT mutations are mutually exclusive (line 195) but this is not apparent from figure 1. This is a bit misleading for the reader, therefore I would suggest to rearrange the cases in figure 1 so that this is more obvious..]
Response 1: [Thank you for your response. With the increase filtering criteria we used, TERT is no longer one of the most frequently mutated genes and does not appear in the figure. However, the Oncoprint diagram shows the most 15 most frequently mutated genes, and these are put in as input into the co-occurence and mutual exclusivity analysis. while most will co-occur rather than be mutually exclusive, some will be found to be mutually exclusive  ]”

•  
•  
•  

Comments 1: [I would also highly recommend to exclude TERT mutations that are not the known promoter mutations from any analysis, as they have to opposite effect of the promoter mutations. Or at least calculate them separately.

.]
Response 1: Thank you for your response. We have updated the manuscript to report that we filtered out non-promoter TERT mutations, indicated in lines 110-114: cBioPortal uses the mutation calls provided by each contributing publication, with little additional filtering except for standardization of mutation annotations using Genome Nexus. [28-30] cBioPortal filters out non-synonymous mutations, including Silent, Intron, IGR, 3'UTR, 5'UTR, 3'Flank and 5'Flank mutations, with the exception for mutations in the promoter region of the TERT gene.  

 

 Comments 1: [Line 63, the enzymes are responsible for demethylation of histones and DNA, not methylation. Line 249 “which aligns with what we IDH-mut A3”; I do think that there is a word missing before IDH-mut A3 

.]
Response 1: [Thank you for your response. WE have corrected these issues in the manuscript]]”

Reviewer 3 Report

Comments and Suggestions for Authors

Dear Authors,

Your manuscript on the mutational characterization of astrocytoma, IDH-mutant, CNS WHO grade III in the AACR GENIE database is an interesting approach to characterize this still difficult diagnosis.  Your approach to separate the grades is valid, but probably also the cause of some inconsistencies in your study, as you also mention in the discussion. To get rid of these I recommend the following:

• Thoroughly stratify your study cohort. As you mention in the limitations there are many different institutes involved and by this also many different pathologists which naturally leads to some variation. In addition, the time frame of this collection spans at least two different WHO classifications which include major changes on the diagnosis of IDH-mut glioma. Therefor it makes sense to include as many stratification criteria as possible to obtain the most stringent classification possible. Methylation data would be best, but you already mentioned that this is not available for the cohort. But CNV data is available and by this excluding 1p/19 LOH and the 7p gain / 10q loss signatures would be very beneficial. Also, if histological data is available this would help significantly, as ATRX loss is detectable by immunohistochemistry and could thereby close the gap of ATRX cases.
• In line 72 you mention that astrocytoma progress into glioblastoma. This is not possible based on the WHO 2021, as all IDH mutant tumors with astrocytic genotype are considered to be astrocytoma. They progress into grade IV astrocytoma. In the same paragraph you also describe secondary glioblastoma to be a different entity, which is not the case. The name secondary glioblastoma should not be used anymore, as these are astrocytoma by WHO 2021 definition. This difference in naming was chosen deliberately to pinpoint the difference in biology and clinical behaviour when compared to glioblastoma. The whole paragraph (line 72-82) is somewhat misleading due to this.
• You mention that TP53 and TERT mutations are mutually exclusive (line 195) but this is not apparent from figure 1. This is a bit misleading for the reader, therefore I would suggest to rearrange the cases in figure 1 so that this is more obvious. I would also highly recommend to exclude TERT mutations that are not the known promoter mutations from any analysis, as they have to opposite effect of the promoter mutations. Or at least calculate them separately.

 

Minor remarks

• Line 63, the enzymes are responsible for demethylation of histones and DNA, not methylation.
• Line 249 “which aligns with what we IDH-mut A3”; I do think that there is a word missing before IDH-mut A3

Best regards

Author Response

Comments 1: [The abstract is well organized. The objective, data source, methods, and results are clearly presented; however, the conclusion could be more precise..]
Response 1: [Thank you for your response. We have removed parts of the conclusion to make it more precise, with the final result being in lines 372-374: This genomic characterization of Astrocytoma, IDH-mutant, CNS WHO grade 3 using the AACR GENIE database confirms key mutational patterns and demonstrates the resource's value for studying this tumor type, provided that correct selection criteria are applied to match the updated WHO classification criteria.

Comments 1: [Line 97-should provide a URL if possible.]
Response 1: [. The cBioPortal for Cancer Genomics platform (https://genie.cbioportal.org/) was employed for data retrieval and subsequent analysis, encompassing the determination of frequencies for gene mutations, copy number alterations, and structural variants, all of which are displayed on a per-sample basis.]

Comments 1: [Line 117: Log-rank tests were used to analyze survival rates, which need reference to the exact version that archives current results.]
Response 1: [Thank you for your response. We have update our manuscript to include this in lines 155-157: Statistical significance between survival distributions was assessed using log-rank tests (from the survival package, version 3.7.0) for the top 10 most frequently mutated genes in the cohort.]”

 

Comments 1: [Line 194: Need a bit more detail about the selection of the 15 most significant co-occurrence features and how they correlated.]

Response 1: Thank you for your response. We have further clarified the selection process as well as described the actual numerical correlations: See lines 245-255:The 15 most frequently mutated genes in the cohort were evaluated for Co-occurrence and exclusivity patterns using the query function in cBioPortal, resulting in 105 pairs tested for co-occurrence and mutual exclusivity. Significant co-occurrence was found between FAT1 and PDGFRA (4 mutations found in both out of 19 mutations found in either, p = 0.016), NOTCH1 and NF1 (3 mutations found in both out of 19 mutations found in either, p = 0.025), KMT2D and PRIKDC (3 mutations found in both out of 13 mutations found in either, p = 0.043), FAT1 and ARID1A (3 mutations found in both out of 18 mutations found in either, p =0.043), and KMT2D and BCOR (3 mutations found in both out of 20 mutations found in either, p=0.044). The only significant mutual exclusivity was found in TP53 and ARID1A (9 found in both out of 179 found in either, p=0.040). In additional, all prior associations lost significance after multiple hypothesis test correction.

Comments 1: [A single database has been used; therefore, an external validation could enhance methodological reliability. 

.]
Response 1: [Thank you for your response. We wanted to confine the scope of this study to GENIE database, but have addressed this in our limitations section. If you think necessary, we would be happy to add validation from another dataset though.   ]”

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

no additional comments

Author Response

No further response is required, as the reviewer had no additional comments

Reviewer 3 Report

Comments and Suggestions for Authors

Dear Authors,

Thank you for addressing all remarks so thoroughly. There are only three small points I want to remark:

Line 78: If an astrocytic tumor progresses into a grade 4 tumor before WHO 2021 it was called a secondary glioblastoma.

Line 81-83. There are no IDH-mutant glioblastoma anymore and if there would be, they would not have TERT promoter mutations or EGFR amplifications. These are hallmarks of glioblastoma IDH-wild type. So, I think this section was somehow swapped around in respect to the addressed tumor entity.

Figure 1 There are 2% of tumors that do not have any IDH1 mutation. They probably have an IDH2 mutation and I would suggest to show this as well, to ensure that it is visible that all tumors are IDH-mut.

Best regards

Author Response

Comments 1: [Line 78: If an astrocytic tumor progresses into a grade 4 tumor before WHO 2021 it was called a secondary glioblastoma.]]
Response 1: Thank you for your response. We have updated the following in line 78 to correct this: Given enough time, IDH-mut A3 will progress into Astrocytoma, IDH-mutant CNS WHO grade 4 (IDH-mut A4), previously known as secondary glioblastoma, IDH-mutant in the 2016 WHO classification. 

Comments 2: [Line 81-83. There are no IDH-mutant glioblastoma anymore and if there would be, they would not have TERT promoter mutations or EGFR amplifications. These are hallmarks of glioblastoma IDH-wild type. So, I think this section was somehow swapped around in respect to the addressed tumor entity.]
Response 2: Thank you for you response. We have updated the following in line 81-82 Unlike IDH-wild type glioblastomas which commonly have promoter mutations in telomerase reverse transcriptase (TERT) and epidermal growth factor receptor (EGFR) amplification [12], IDH-mut A3 evolves into a IDH-mut A4 through homozygous deletions in CDKN2A/B [14]. 

 

Comments 3: [Figure 1 There are 2% of tumors that do not have any IDH1 mutation. They probably have an IDH2 mutation and I would suggest to show this as well, to ensure that it is visible that all tumors are IDH-mut..]
Response 3: Thank you for you response. We have added the IDH2 mutation to figure 1.