Abstract: The first sentence in the abstract clearly defines the context (glioma) and the problem associated with it (poor diagnosis). The second sentence directly starts with the method used. Thus, authors did not state their aim clearly in their abstract (actually in the whole manuscript). The authors stated that alleles near TERC (rs1920116) and TERT (rs2736100) loci are associated with higher glioma risk. The new SNPs identified in the study are also associated with longer leukocyte telomere length (LTL). Methods and findings are stated clearly in the abstract.

Title is informative and relevant - summarizes information in the abstract. References: include relevant, recent (2013) papers. References were referenced correctly, and key studies (GWAS studies, TERT promoter (TERTp) mutations and RTEL1 variants in glioma) in the context are included.

Introduction: The authors introduced the topic progressively in the introduction section. They started with glioma and explained that the most aggressive form is glioblastoma (GBM). They mention that the frequency of TERT promoter (TERTp) mutations are quite high in GBM but use high grade glioma and glioblastoma interchangeably within the manuscript. They should mention the differences between GBM and high grade glioma and should clarify why they used “GBM” versus “high grade glioma” patient data in each case. They can also point out why finding a risk allele in GBM is more critical.

The authors also mention GWAS studies and 7 genetic loci including TERT and RTEL1 associated with glioma risk were identified in previous studies. These 2 genes actually constituted the whole body of the manuscript. Thus, it is very reasonable to explain function of these in the introduction. Moreover, the authors mentioned later that all individuals included in the study were of European ancestry. It would be nice to include frequency of TERTp /TERC/RTEL1 mutations in this population in the introduction. Although the authors introduced the concept progressively, the introduction is made out of one paragraph which includes very limited information about the topic. The readers included many new topics after the results section. Acronyms like GWAS and LTL are defined in the abstract, but it would be nice to do it again in the introduction. Furthermore, long version of SNP may be mentioned once in introduction for those who are unfamiliar with the topic. Aim of the paper is only stated partially (To identify new risk loci for glioma…) but research question is not stated in the manuscript at all. I assumed it is “Are there new genetic loci associated with glioma risk” but it is not clear. The authors should have explained why this (identification of new genetic loci) is important in introduction. The authors also mention several new topics such as mechanisms contributing to glioma risk, stress induced and canonical telomere extension which should all be part of introduction. The authors focus on TERT, TERC and RTEL1 throughout the manuscript but they do not explain function of these genes clearly. They should have included an introductory sentence to telomeres, change of their length with age and tumorigenesis in introduction. Later in the manuscript they explain that TERC is the RNA component – which should be done in introduction section. They also mention that mean telomere length (it would be nice to use an acronym for this as well) is a biomarker and epidemiological risk factor for gliomagenesis- which must all be part of introduction. The authors also mention LTL and glioma risk several times in the text including the abstract. It would be nice if they included an example reference focusing on association of rs1920116 (which is mentioned in abstract but not introduction) with LTL with more detail. This would be a nice introduction to the topic before explaining identification of association of this SNP with glioma risk (as well as LTL) in the results section.

Methodology

1. Subject selection: Process of subject selection is generally clear with some confusing points. The authors used only patients with European ancestry. However, they mention Caucasian, white and European ancestry with the text. It would be nicer to start and continue with European ancestry. Furthermore, in the majority of the paper, they used controls based on age, ethnicity and sex, which is appropriate. However, in stage 3 (Mayo clinic data) they selected control group based on a different criteria “matching variables: sex, date of birth (within 2.5 years), self-identified ancestry (Hispanic white, non- hispanic white, American indian, African american, Asian, pacific islander or other) and residence”. The authors should have explained why they included extra criteria at this step, is it possible to decrease the number of controls and select control patients with only European origin? In addition to ancestry, the authors selected patients based on age (>18 years old). Patients and controls had similar histology. They excluded patients with brain tumor and poorly genotypes controls. The authors explained population series starting from series #2, it would be nice to explain series #1 as well. Population series (UCSF AGS) were labeled as 1997-9 (#2), 2001-5(#3),2006-9(#4) whereas Clinic base diagnosis of UCSF Neuro-oncology clinic were labeled as (2002-6 (#3) 2006-10 (#4),2009-12 (#5)) The authors need to clarify these numbers are used in stage 3 (replication) are replication by referring to Supplementary Table 1 (S Table 1) and preferably labeled these differently from UCSF AGS study.

2. Measurement of Variables: Throughout the text, terms glioma and glioblastoma (GBM) were used interchangeably as mentioned above in Introduction section. In S Table 1- most patient data are from GBM cases except 25% of UCSF data (stage 1). The authors need to explain why they progressed to GBM from glioma in patient groups and why they used to glioma or GBM data in each case. The numbers provided in Supplementary Fig1 (S Fig1) and S Table 1 are not consistent at stage 3. In S Fig1, UCSF 388 GBM cases, 511 controls were used in stage 3 whereas the same study is included as UCSF 303 GBM cases, 375 controls in S Table 1. This inconsistency need be corrected. S Fig1 is a nice outline of methodology and it would be nice to refer to this figure at the end of 1st sentence of the method section within the main text (second paragraph). It will be easier to follow methodology. Moreover, stages should be mentioned and explained briefly in the main text. In study design section of online methods, stages are explained in detail and must be referred both in S Fig1 and main text while explaining different stages. In addition, Stage 4 is not explained in methodology section of the main text. S Fig 1 and S Table1 are confusing at stage 4- stage 4 must be rewritten and S Fig1 can refer to the study design section (online methods) within the legend. In Stage 1 of S Fig1, UCSF and TCGA high grade glioma cases were outlined whereas the same numbers were outlined as “UCSF high grade glioma cases (GBM 85%) and TCGA GBM cases” in S Table 1. Thus, to be consistent (S table 1+ S fig1), S fig1 should be corrected in terms of glioma vs GBM cases. Finally, when referring to rs1920116, it would be nice to mention either within the text of legend that it refers to TERC locus.

3. Reliability of Methods: In the section of “Measurement of telomere length” in Online methods, the authors mentioned “to ensure harmonization assays were conducted at one of four centralized labs”. This sentence implies that there is a problem with consistency (reliability, reproducibility). They need to clarify the reason and there are no differences in assay procedures. Furthermore, the authors mention “67% of samples ran on the same lab”. The authors should explain why minority of the samples were ran elsewhere. Was there a problem with consistency? They mention that differences in different labs were due to differences in calibrator or standard DNA used –did not they use the same reference material? In statistical analysis section, the authors used different software including EIGENSTRAT, Impute2 SNPTESTv2, META, STATA at different stages of the study. In “Testing for independence of glioma risk loci and for the presence of interaction” section, they do not mention which software is used for single logistic regression model. Within the section they also refer to Supplementary Figure 2 (S Fig2) and Supplementary Figure 3 (S Fig 3), which do not also mention the name of the software.

4. Replication of the study: The information present in methodology seems appropriate for repeating the experiment elsewhere. However, some confusions mentioned above (section 3) need to be corrected. In addition, abbreviations need to be appropriately used in within the main text and Supplementary data (S data) (see results section).

Results The authors organized results in two parts, i. association of high risk alleles with glioma/ high-grade glioma and ii. association of identified SNPs with LTL. Organization the data is appropriate and eases understanding the link between SNPs- glioma risk and LTL. Additionally, beta estimates for standard error were explained well in footnote of Supplementary Table 2 (S Table 2). However, some revisions are necessary in the results section. For instance, results and discussion are intermixed. Even if no separate titles are present for results and discussion, interpretation of the data should have started after introduction of all data. Examples include “OR estimates- similar to colorectal cancer”, “Conditional analyses suggest that”, “Variants regulating TERC expression- might differ across tissues”. In addition, telomere maintenance, oncogenic progression and telomere length are explained and linked towards the end of the manuscript. This paragraph should be moved to the introduction. Furthermore, while explaining association of OR and age, it would be nice to include percentages of each age group in a table or under the ages (x-axis) in the figure (S Figure 3). Units, titles of the figures and tables are appropriate. Footnotes, miss some critical information/explanation (see below). Rows and columns labelled well, some with missing explanation of abbreviations. Furthermore, categories are grouped well, experimental and control groups are separated well and are self-explanatory. Text in the results section is not repetitive, which is appropriate but text in the footnotes requiring some extra information (see below). Tables and figures are partially clear- problems especially with OR and abbreviations are present in nearly all figures/tables. Within the whole text, abbreviations should be used for glioblastoma (GBM) and odd ratio (OR). First, throughout figures/tables, there is a problem and inconsistency in explaining what OR stands for. For example, for OR, full length phrase was mentioned once in the text then manuscript continued with OR, which is reasonable. However, in Table 1: full name of OR needs to be present in the legend as “odds ratio (OR)”. S Fig2 starts with OR in the legend and continues with full phrase. Authors used full phrase in S Fig3 and Supplementary Figure 5 (SFig 5). “OR” phrase should be revised throughout the whole text including S data. Second, glioblastoma is referred as “glioblastoma” within main text and S data. However, in S Fig1, they referred to glioblastoma cases as GBM without explaining the abbreviation in the legend or footnote. Other abbreviations also need to be added or explained. For example, in S Table 3, “rsID” and “Pos” need to be explained in the footnote. Instead of rsID, SNP ID could be used. In the footnote, “a” is explained as “minor allele freq” which is really bad even for the reputation of the journal. All typos in the text need to be corrected. Moreover, in “c” the authors had written TF (transcription factor) with no explanation. Every reader may not be familiar with TF and it needs to be written in full at least in the first mention in the footnote. In addition, for DNAse HS, DNAse hypersensitivity is explained in the footnote but not connected with HS- should have written “DNAse hypersensitivity (HS)”. When mentioning stage 4 in S Fig1, the authors used the full phrase for LTL and at the last box in the figure they used “LTL”. In this figure, full names of the following should be explained in the footnote: Mayo, GWAS, LTL and GBM and the abbreviations should be used appropriately. In the same figure, the readers should be referred to online methods section for “Illumina”, “Affymetrix”, Taqman and Sequenom. In Supplementary Table 4, pos(37), A1 and A2 should be explained in the footnote. In Table2, details of “number” should be given. The authors did not explain what considered as significant within the text. For example, in S Fig5, they selected “glioma OR for SNPs with p<0,05”, but did not indicate what p value is considered significant.

Discussion The authors included results and their discussion together in most paragraphs. Towards the end of the manuscript, they provided broader discussions of the topic and included a summary (conclusions) at the final paragraph. Results were discussed from multiple angles and placed into context without being over interpreted. The main conclusion in the paper is “a new locus (SNP) in TERC (rs1920116) is associated with high grade glioma (C1)” and results clearly supported this conclusion. The second conclusion is “alleles in/near TERC as well as TERT influence glioma risk and telomere length (C2)”. This conclusion is supported by the finding that SNPs in TERT (rs2736100), TERC and RTEL1 were independently associated with glioma risk, and this association was still present when all SNPs are present together. The authors, however, did not clearly discuss the importance of the study and how these results can be implicated in future studies. A new paragraph discussing importance of the results and possible future studies should be added. In addition, the frequency of these SNPs is not mentioned throughout the paper. Since, TERT promoter (TERTp) mutations are detected very frequently in most solid tumors including glioblastoma (GBM), in would be nice to compare frequency of these SNPs with TERTp mutations in GBM. The authors pointed out a single aim throughout the paper, which is “to identify new risk loci for glioma“. Their conclusion of “SNPs in TERC, TERT and RTEL1 are independently associated with increased glioma risk with modest effect in RTEL1 (C3)” answers this aim perfectly well. The authors also mention “Other loci like EGFR, which is known to be associated with high glioma risk lacks association with LTL, supports hypothesis that multiple mechanisms contribute to glioma risk.” This hypothesis should have been mentioned within the introduction with more detail on the context. Although the authors do not clearly mention other aims, one of their aims should have been “to compare association of risk alleles (TERT, TERC and RTEL1) with high grade glioma”. Their conclusion C1 actually answers this aim. Therefore, by revising the introduction section, the authors could add this aim to the manuscript. Conclusions are supported by both findings and references. As mentioned above, conclusion C2 is supported by the findings. Another conclusion: “Both longer and shorter telomere lengths associate with pathogenicity (C4)” is supported by both results and references provided. For TERC loci, similar OR estimates were recorded in GBM (this manuscript) and colorectal (literature) cancer patients. Furthermore, the authors mentioned another SNP (rs10936599) near TERC loci, which was also determined to be associated with increased glioma risk and LTL in this manuscript, is also associated with colorectal cancer risk. However, this statement is made in a later paragraph than the previous one. Since both statements are about colorectal cancer, the authors could have combined the two in two consecutive statements. Similarly multiple independent TERT SNPs were associated with cancer risk and LTL in glioma (this paper) as well as breast (reference) cancer patients. In addition, the authors mentioned additional SNPs near TERT and TERC, which were found to be associated with idiopathic pulmonary fibrosis(IPF) in the literature, are also associated with shorter telomere length and decreased glioma risk in this manuscript (opposite of findings about rs1920116 and rs2736100). Towards the end of the manuscript, the authors also indicated that mean telomere length is a promising epidemiological risk factor and biomarker for cancer. This statement should have been discussed in the introduction section. Then they should have suggested that the new SNPs identified in this study could be used as a biomarker in the future in the discussion section. Besides IPF, the authors also mention that shorter telomere length is associated with coronary artery disease whereas shorter along with longer telomere length is associated with different cancer types. Therefore, it would be nice to include a comparison of telomere length in hematological and solid tumors in the introduction section before including this result in the discussion section. One major limitation of this study which is also mentioned by the authors towards the end, is that data connecting telomere length in astrocytes and LTL is lacking in the manuscript. They suggest that difference in statistical significance values of associations with glioma risk and LTL may be due to differences in sample size, analytic technique and tissue type used. In addition, they point out that variants regulating TERC expression might differ between tissues due to differential TF expression and tissue specific regulatory mechanisms. Because they included telomere length values measured in leukocytes but not astrocytes, this statement confuses the reader about using LTL data in a glioma study. Due to differences in TF regulated mechanisms in different tissues (leukocytes versus astrocytes), it is possible that telomere length may differ between tissues. To make this clear, the authors should have compared telomere length in astrocytes and leukocytes either in an additional experiment or provided a reference from literature and provided supporting evidence that telomere length is similar in astrocytes and leukocytes.